Compiler Explorer

Source code

; ModuleID = 'ld-temp.o'
source_filename = "ld-temp.o"
target datalayout = "e-p:64:64-p1:64:64-p2:32:32-p3:32:32-p4:64:64-p5:32:32-p6:32:32-p7:160:256:256:32-p8:128:128:128:48-p9:192:256:256:32-i64:64-v16:16-v24:32-v32:32-v48:64-v96:128-v192:256-v256:256-v512:512-v1024:1024-v2048:2048-n32:64-S32-A5-G1-ni:7:8:9"
target triple = "amdgcn-amd-amdhsa"

%"struct.__llvm_libc_22_0_0_git::DataEnvironment" = type { ptr }
%class.LlvmLibcCharacterConverterUTF32To8Test_OneByte = type { %"class.__llvm_libc_22_0_0_git::testing::Test" }
%"class.__llvm_libc_22_0_0_git::testing::Test" = type { ptr, ptr, ptr }
%"class.__llvm_libc_22_0_0_git::FixedVector" = type { %"struct.__llvm_libc_22_0_0_git::cpp::array.6", i64 }
%"struct.__llvm_libc_22_0_0_git::cpp::array.6" = type { [64 x %"struct.__llvm_libc_22_0_0_git::AtExitUnit"] }
%"struct.__llvm_libc_22_0_0_git::AtExitUnit" = type { ptr, ptr }
%"struct.rpc::Client" = type { %"struct.rpc::Process" }
%"struct.rpc::Process" = type { i32, ptr, ptr, ptr, ptr, [128 x i32] }
%"struct.__llvm_libc_22_0_0_git::testing::TestOptions" = type <{ ptr, i8, i8, [6 x i8] }>
%"class.__llvm_libc_22_0_0_git::IntegerToString" = type { %"struct.__llvm_libc_22_0_0_git::cpp::array.59", i64 }
%"struct.__llvm_libc_22_0_0_git::cpp::array.59" = type { [11 x i8] }
%"struct.rpc::Header" = type { i64, i32 }
%"struct.rpc::Buffer" = type { [8 x i64] }
%"class.__llvm_libc_22_0_0_git::IntegerToString.21" = type { %"struct.__llvm_libc_22_0_0_git::cpp::array.17", i64 }
%"struct.__llvm_libc_22_0_0_git::cpp::array.17" = type { [21 x i8] }
%"struct.__llvm_libc_22_0_0_git::testing::TestLogger" = type { i8 }
%"struct.rpc::Port" = type <{ ptr, i64, i32, i32, i32, i8, i8, [2 x i8] }>

$_ZN22__llvm_libc_22_0_0_git7testing4TestD2Ev = comdat any

$_ZN51LlvmLibcCharacterConverterUTF32To8Test_InvalidStateD0Ev = comdat any

$_ZN22__llvm_libc_22_0_0_git7testing4Test5SetUpEv = comdat any

$_ZN22__llvm_libc_22_0_0_git7testing4Test8TearDownEv = comdat any

$_ZNK51LlvmLibcCharacterConverterUTF32To8Test_InvalidState7getNameEv = comdat any

$_ZN60LlvmLibcCharacterConverterUTF32To8Test_CantPushMidConversionD0Ev = comdat any

$_ZNK60LlvmLibcCharacterConverterUTF32To8Test_CantPushMidConversion7getNameEv = comdat any

$_ZN47LlvmLibcCharacterConverterUTF32To8Test_FourByteD0Ev = comdat any

$_ZNK47LlvmLibcCharacterConverterUTF32To8Test_FourByte7getNameEv = comdat any

$_ZN48LlvmLibcCharacterConverterUTF32To8Test_ThreeByteD0Ev = comdat any

$_ZNK48LlvmLibcCharacterConverterUTF32To8Test_ThreeByte7getNameEv = comdat any

$_ZN46LlvmLibcCharacterConverterUTF32To8Test_TwoByteD0Ev = comdat any

$_ZNK46LlvmLibcCharacterConverterUTF32To8Test_TwoByte7getNameEv = comdat any

$_ZN46LlvmLibcCharacterConverterUTF32To8Test_OneByteD0Ev = comdat any

$_ZNK46LlvmLibcCharacterConverterUTF32To8Test_OneByte7getNameEv = comdat any

$_ZN22__llvm_libc_22_0_0_git7testing8internal4testIcEEbPNS1_10RunContextENS0_8TestCondET_S6_PKcS8_NS1_8LocationE = comdat any

$_ZN22__llvm_libc_22_0_0_git7testing8internal4testIiEEbPNS1_10RunContextENS0_8TestCondET_S6_PKcS8_NS1_8LocationE = comdat any

$_ZN22__llvm_libc_22_0_0_git7testing8internal4testIbEEbPNS1_10RunContextENS0_8TestCondET_S6_PKcS8_NS1_8LocationE = comdat any

$_ZN22__llvm_libc_22_0_0_git17stdc_at_exit_funcEPv = comdat any

$_ZN3rpc6Client4openILj1660944385EEENS_4PortILb0EEEv = comdat any

$_ZN3rpc4PortILb0EE5closeEv = comdat any

$_ZN22__llvm_libc_22_0_0_git3cpp6string14NULL_CHARACTERE = comdat any

@_ZN22__llvm_libc_22_0_0_git3appE = internal unnamed_addr addrspace(1) global %"struct.__llvm_libc_22_0_0_git::DataEnvironment" zeroinitializer, align 8
@__fini_array_end = external hidden addrspace(1) global [0 x i64], align 8
@__fini_array_start = external hidden addrspace(1) global [0 x i64], align 8
@__init_array_end = external hidden addrspace(1) global [0 x i64], align 8
@__init_array_start = external hidden addrspace(1) global [0 x i64], align 8
@llvm.global_ctors = appending global [1 x { i32, ptr, ptr }] [{ i32, ptr, ptr } { i32 65535, ptr @_GLOBAL__sub_I_utf32_to_8_test.cpp, ptr null }]
@LlvmLibcCharacterConverterUTF32To8Test_OneByte_Instance = internal addrspace(1) global %class.LlvmLibcCharacterConverterUTF32To8Test_OneByte zeroinitializer, align 8
@.str = private unnamed_addr addrspace(4) constant [12 x i8] c"cr.isFull()\00", align 1
@.str.1 = private unnamed_addr addrspace(4) constant [5 x i8] c"true\00", align 1
@.str.2 = private unnamed_addr addrspace(4) constant [82 x i8] c"/home/janplehr/git/llvm-project/libc/test/src/__support/wchar/utf32_to_8_test.cpp\00", align 1
@.str.3 = private unnamed_addr addrspace(4) constant [19 x i8] c"popped.has_value()\00", align 1
@.str.4 = private unnamed_addr addrspace(4) constant [34 x i8] c"static_cast<char>(popped.value())\00", align 1
@.str.5 = private unnamed_addr addrspace(4) constant [4 x i8] c"'A'\00", align 1
@.str.6 = private unnamed_addr addrspace(4) constant [13 x i8] c"cr.isEmpty()\00", align 1
@.str.7 = private unnamed_addr addrspace(4) constant [4 x i8] c"'B'\00", align 1
@.str.8 = private unnamed_addr addrspace(4) constant [6 x i8] c"false\00", align 1
@LlvmLibcCharacterConverterUTF32To8Test_TwoByte_Instance = internal addrspace(1) global %class.LlvmLibcCharacterConverterUTF32To8Test_OneByte zeroinitializer, align 8
@.str.10 = private unnamed_addr addrspace(4) constant [33 x i8] c"static_cast<int>(popped.value())\00", align 1
@.str.11 = private unnamed_addr addrspace(4) constant [5 x i8] c"0xc3\00", align 1
@.str.12 = private unnamed_addr addrspace(4) constant [14 x i8] c"!cr.isEmpty()\00", align 1
@.str.13 = private unnamed_addr addrspace(4) constant [5 x i8] c"0xbf\00", align 1
@.str.14 = private unnamed_addr addrspace(4) constant [5 x i8] c"0xd6\00", align 1
@.str.15 = private unnamed_addr addrspace(4) constant [5 x i8] c"0x8e\00", align 1
@LlvmLibcCharacterConverterUTF32To8Test_ThreeByte_Instance = internal addrspace(1) global %class.LlvmLibcCharacterConverterUTF32To8Test_OneByte zeroinitializer, align 8
@.str.17 = private unnamed_addr addrspace(4) constant [5 x i8] c"0xea\00", align 1
@.str.18 = private unnamed_addr addrspace(4) constant [5 x i8] c"0xb0\00", align 1
@.str.19 = private unnamed_addr addrspace(4) constant [5 x i8] c"0x95\00", align 1
@.str.20 = private unnamed_addr addrspace(4) constant [5 x i8] c"0xe2\00", align 1
@.str.21 = private unnamed_addr addrspace(4) constant [5 x i8] c"0x99\00", align 1
@.str.22 = private unnamed_addr addrspace(4) constant [5 x i8] c"0xbb\00", align 1
@LlvmLibcCharacterConverterUTF32To8Test_FourByte_Instance = internal addrspace(1) global %class.LlvmLibcCharacterConverterUTF32To8Test_OneByte zeroinitializer, align 8
@.str.24 = private unnamed_addr addrspace(4) constant [5 x i8] c"0xf0\00", align 1
@.str.25 = private unnamed_addr addrspace(4) constant [5 x i8] c"0x9f\00", align 1
@.str.26 = private unnamed_addr addrspace(4) constant [5 x i8] c"0xa4\00", align 1
@.str.27 = private unnamed_addr addrspace(4) constant [5 x i8] c"0xa1\00", align 1
@.str.28 = private unnamed_addr addrspace(4) constant [5 x i8] c"0x92\00", align 1
@.str.29 = private unnamed_addr addrspace(4) constant [5 x i8] c"0x84\00", align 1
@LlvmLibcCharacterConverterUTF32To8Test_CantPushMidConversion_Instance = internal addrspace(1) global %class.LlvmLibcCharacterConverterUTF32To8Test_OneByte zeroinitializer, align 8
@.str.31 = private unnamed_addr addrspace(4) constant [15 x i8] c"cr.push(utf32)\00", align 1
@.str.32 = private unnamed_addr addrspace(4) constant [2 x i8] c"0\00", align 1
@.str.33 = private unnamed_addr addrspace(4) constant [4 x i8] c"err\00", align 1
@.str.34 = private unnamed_addr addrspace(4) constant [3 x i8] c"-1\00", align 1
@LlvmLibcCharacterConverterUTF32To8Test_InvalidState_Instance = internal addrspace(1) global %class.LlvmLibcCharacterConverterUTF32To8Test_OneByte zeroinitializer, align 8
@.str.36 = private unnamed_addr addrspace(4) constant [18 x i8] c"c1.isValidState()\00", align 1
@.str.37 = private unnamed_addr addrspace(4) constant [18 x i8] c"c2.isValidState()\00", align 1
@.str.38 = private unnamed_addr addrspace(4) constant [18 x i8] c"c3.isValidState()\00", align 1
@.str.39 = private unnamed_addr addrspace(4) constant [18 x i8] c"c4.isValidState()\00", align 1
@.str.40 = private unnamed_addr addrspace(4) constant [18 x i8] c"c5.isValidState()\00", align 1
@.str.41 = private unnamed_addr addrspace(4) constant [18 x i8] c"c6.isValidState()\00", align 1
@.str.42 = private unnamed_addr addrspace(4) constant [18 x i8] c"c7.isValidState()\00", align 1
@.str.43 = private unnamed_addr addrspace(4) constant [18 x i8] c"c8.isValidState()\00", align 1
@.str.44 = private unnamed_addr addrspace(4) constant [18 x i8] c"c9.isValidState()\00", align 1
@.str.45 = private unnamed_addr addrspace(4) constant [19 x i8] c"c10.isValidState()\00", align 1
@.str.46 = private unnamed_addr addrspace(4) constant [19 x i8] c"c11.isValidState()\00", align 1
@_ZTV46LlvmLibcCharacterConverterUTF32To8Test_OneByte = internal unnamed_addr addrspace(1) constant { [8 x ptr addrspace(1)] } { [8 x ptr addrspace(1)] [ptr addrspace(1) null, ptr addrspace(1) null, ptr addrspace(1) addrspacecast (ptr @_ZN22__llvm_libc_22_0_0_git7testing4TestD2Ev to ptr addrspace(1)), ptr addrspace(1) addrspacecast (ptr @_ZN46LlvmLibcCharacterConverterUTF32To8Test_OneByteD0Ev to ptr addrspace(1)), ptr addrspace(1) addrspacecast (ptr @_ZN22__llvm_libc_22_0_0_git7testing4Test5SetUpEv to ptr addrspace(1)), ptr addrspace(1) addrspacecast (ptr @_ZN22__llvm_libc_22_0_0_git7testing4Test8TearDownEv to ptr addrspace(1)), ptr addrspace(1) addrspacecast (ptr @_ZN46LlvmLibcCharacterConverterUTF32To8Test_OneByte3RunEv to ptr addrspace(1)), ptr addrspace(1) addrspacecast (ptr @_ZNK46LlvmLibcCharacterConverterUTF32To8Test_OneByte7getNameEv to ptr addrspace(1))] }, align 8, !type !0, !type !1, !type !2, !type !3, !type !4, !type !5, !type !6, !type !7, !type !8, !type !9
@_ZTV46LlvmLibcCharacterConverterUTF32To8Test_TwoByte = internal unnamed_addr addrspace(1) constant { [8 x ptr addrspace(1)] } { [8 x ptr addrspace(1)] [ptr addrspace(1) null, ptr addrspace(1) null, ptr addrspace(1) addrspacecast (ptr @_ZN22__llvm_libc_22_0_0_git7testing4TestD2Ev to ptr addrspace(1)), ptr addrspace(1) addrspacecast (ptr @_ZN46LlvmLibcCharacterConverterUTF32To8Test_TwoByteD0Ev to ptr addrspace(1)), ptr addrspace(1) addrspacecast (ptr @_ZN22__llvm_libc_22_0_0_git7testing4Test5SetUpEv to ptr addrspace(1)), ptr addrspace(1) addrspacecast (ptr @_ZN22__llvm_libc_22_0_0_git7testing4Test8TearDownEv to ptr addrspace(1)), ptr addrspace(1) addrspacecast (ptr @_ZN46LlvmLibcCharacterConverterUTF32To8Test_TwoByte3RunEv to ptr addrspace(1)), ptr addrspace(1) addrspacecast (ptr @_ZNK46LlvmLibcCharacterConverterUTF32To8Test_TwoByte7getNameEv to ptr addrspace(1))] }, align 8, !type !10, !type !11, !type !12, !type !13, !type !14, !type !5, !type !6, !type !7, !type !8, !type !9
@_ZTV48LlvmLibcCharacterConverterUTF32To8Test_ThreeByte = internal unnamed_addr addrspace(1) constant { [8 x ptr addrspace(1)] } { [8 x ptr addrspace(1)] [ptr addrspace(1) null, ptr addrspace(1) null, ptr addrspace(1) addrspacecast (ptr @_ZN22__llvm_libc_22_0_0_git7testing4TestD2Ev to ptr addrspace(1)), ptr addrspace(1) addrspacecast (ptr @_ZN48LlvmLibcCharacterConverterUTF32To8Test_ThreeByteD0Ev to ptr addrspace(1)), ptr addrspace(1) addrspacecast (ptr @_ZN22__llvm_libc_22_0_0_git7testing4Test5SetUpEv to ptr addrspace(1)), ptr addrspace(1) addrspacecast (ptr @_ZN22__llvm_libc_22_0_0_git7testing4Test8TearDownEv to ptr addrspace(1)), ptr addrspace(1) addrspacecast (ptr @_ZN48LlvmLibcCharacterConverterUTF32To8Test_ThreeByte3RunEv to ptr addrspace(1)), ptr addrspace(1) addrspacecast (ptr @_ZNK48LlvmLibcCharacterConverterUTF32To8Test_ThreeByte7getNameEv to ptr addrspace(1))] }, align 8, !type !15, !type !16, !type !17, !type !18, !type !19, !type !5, !type !6, !type !7, !type !8, !type !9
@_ZTV47LlvmLibcCharacterConverterUTF32To8Test_FourByte = internal unnamed_addr addrspace(1) constant { [8 x ptr addrspace(1)] } { [8 x ptr addrspace(1)] [ptr addrspace(1) null, ptr addrspace(1) null, ptr addrspace(1) addrspacecast (ptr @_ZN22__llvm_libc_22_0_0_git7testing4TestD2Ev to ptr addrspace(1)), ptr addrspace(1) addrspacecast (ptr @_ZN47LlvmLibcCharacterConverterUTF32To8Test_FourByteD0Ev to ptr addrspace(1)), ptr addrspace(1) addrspacecast (ptr @_ZN22__llvm_libc_22_0_0_git7testing4Test5SetUpEv to ptr addrspace(1)), ptr addrspace(1) addrspacecast (ptr @_ZN22__llvm_libc_22_0_0_git7testing4Test8TearDownEv to ptr addrspace(1)), ptr addrspace(1) addrspacecast (ptr @_ZN47LlvmLibcCharacterConverterUTF32To8Test_FourByte3RunEv to ptr addrspace(1)), ptr addrspace(1) addrspacecast (ptr @_ZNK47LlvmLibcCharacterConverterUTF32To8Test_FourByte7getNameEv to ptr addrspace(1))] }, align 8, !type !20, !type !21, !type !22, !type !23, !type !24, !type !5, !type !6, !type !7, !type !8, !type !9
@_ZTV60LlvmLibcCharacterConverterUTF32To8Test_CantPushMidConversion = internal unnamed_addr addrspace(1) constant { [8 x ptr addrspace(1)] } { [8 x ptr addrspace(1)] [ptr addrspace(1) null, ptr addrspace(1) null, ptr addrspace(1) addrspacecast (ptr @_ZN22__llvm_libc_22_0_0_git7testing4TestD2Ev to ptr addrspace(1)), ptr addrspace(1) addrspacecast (ptr @_ZN60LlvmLibcCharacterConverterUTF32To8Test_CantPushMidConversionD0Ev to ptr addrspace(1)), ptr addrspace(1) addrspacecast (ptr @_ZN22__llvm_libc_22_0_0_git7testing4Test5SetUpEv to ptr addrspace(1)), ptr addrspace(1) addrspacecast (ptr @_ZN22__llvm_libc_22_0_0_git7testing4Test8TearDownEv to ptr addrspace(1)), ptr addrspace(1) addrspacecast (ptr @_ZN60LlvmLibcCharacterConverterUTF32To8Test_CantPushMidConversion3RunEv to ptr addrspace(1)), ptr addrspace(1) addrspacecast (ptr @_ZNK60LlvmLibcCharacterConverterUTF32To8Test_CantPushMidConversion7getNameEv to ptr addrspace(1))] }, align 8, !type !25, !type !26, !type !27, !type !28, !type !29, !type !5, !type !6, !type !7, !type !8, !type !9
@_ZTV51LlvmLibcCharacterConverterUTF32To8Test_InvalidState = internal unnamed_addr addrspace(1) constant { [8 x ptr addrspace(1)] } { [8 x ptr addrspace(1)] [ptr addrspace(1) null, ptr addrspace(1) null, ptr addrspace(1) addrspacecast (ptr @_ZN22__llvm_libc_22_0_0_git7testing4TestD2Ev to ptr addrspace(1)), ptr addrspace(1) addrspacecast (ptr @_ZN51LlvmLibcCharacterConverterUTF32To8Test_InvalidStateD0Ev to ptr addrspace(1)), ptr addrspace(1) addrspacecast (ptr @_ZN22__llvm_libc_22_0_0_git7testing4Test5SetUpEv to ptr addrspace(1)), ptr addrspace(1) addrspacecast (ptr @_ZN22__llvm_libc_22_0_0_git7testing4Test8TearDownEv to ptr addrspace(1)), ptr addrspace(1) addrspacecast (ptr @_ZN51LlvmLibcCharacterConverterUTF32To8Test_InvalidState3RunEv to ptr addrspace(1)), ptr addrspace(1) addrspacecast (ptr @_ZNK51LlvmLibcCharacterConverterUTF32To8Test_InvalidState7getNameEv to ptr addrspace(1))] }, align 8, !type !30, !type !31, !type !32, !type !33, !type !34, !type !5, !type !6, !type !7, !type !8, !type !9
@.str.47 = private unnamed_addr addrspace(4) constant [47 x i8] c"LlvmLibcCharacterConverterUTF32To8Test.OneByte\00", align 1
@.str.48 = private unnamed_addr addrspace(4) constant [47 x i8] c"LlvmLibcCharacterConverterUTF32To8Test.TwoByte\00", align 1
@.str.49 = private unnamed_addr addrspace(4) constant [49 x i8] c"LlvmLibcCharacterConverterUTF32To8Test.ThreeByte\00", align 1
@.str.50 = private unnamed_addr addrspace(4) constant [48 x i8] c"LlvmLibcCharacterConverterUTF32To8Test.FourByte\00", align 1
@.str.51 = private unnamed_addr addrspace(4) constant [61 x i8] c"LlvmLibcCharacterConverterUTF32To8Test.CantPushMidConversion\00", align 1
@.str.52 = private unnamed_addr addrspace(4) constant [52 x i8] c"LlvmLibcCharacterConverterUTF32To8Test.InvalidState\00", align 1
@.str.9 = private unnamed_addr addrspace(4) constant [9 x i8] c"equal to\00", align 1
@.str.6.10 = private unnamed_addr addrspace(4) constant [2 x i8] c":\00", align 1
@.str.7.11 = private unnamed_addr addrspace(4) constant [11 x i8] c": FAILURE\0A\00", align 1
@_ZN22__llvm_libc_22_0_0_git7testing4Test5StartE = internal unnamed_addr addrspace(1) global ptr null, align 8
@_ZN22__llvm_libc_22_0_0_git7testing4Test3EndE = internal unnamed_addr addrspace(1) global ptr null, align 8
@.str.8.26 = private unnamed_addr addrspace(4) constant [6 x i8] c"\1B[32m\00", align 1
@.str.9.27 = private unnamed_addr addrspace(4) constant [1 x i8] zeroinitializer, align 1
@.str.10.28 = private unnamed_addr addrspace(4) constant [6 x i8] c"\1B[31m\00", align 1
@.str.11.29 = private unnamed_addr addrspace(4) constant [5 x i8] c"\1B[0m\00", align 1
@.str.12.30 = private unnamed_addr addrspace(4) constant [14 x i8] c"[==========] \00", align 1
@.str.13.31 = private unnamed_addr addrspace(4) constant [9 x i8] c"Running \00", align 1
@.str.14.32 = private unnamed_addr addrspace(4) constant [6 x i8] c" test\00", align 1
@.str.15.33 = private unnamed_addr addrspace(4) constant [2 x i8] c"s\00", align 1
@.str.16 = private unnamed_addr addrspace(4) constant [21 x i8] c" from 1 test suite.\0A\00", align 1
@.str.17.34 = private unnamed_addr addrspace(4) constant [14 x i8] c"[ RUN      ] \00", align 1
@.str.18.35 = private unnamed_addr addrspace(4) constant [14 x i8] c"[  FAILED  ] \00", align 1
@.str.19.36 = private unnamed_addr addrspace(4) constant [14 x i8] c"[       OK ] \00", align 1
@.str.20.37 = private unnamed_addr addrspace(4) constant [3 x i8] c" (\00", align 1
@.str.21.38 = private unnamed_addr addrspace(4) constant [26 x i8] c"unknown - try rerunning)\0A\00", align 1
@.str.22.39 = private unnamed_addr addrspace(4) constant [6 x i8] c" ms)\0A\00", align 1
@.str.23 = private unnamed_addr addrspace(4) constant [6 x i8] c" us)\0A\00", align 1
@.str.24.40 = private unnamed_addr addrspace(4) constant [6 x i8] c" ns)\0A\00", align 1
@.str.25.41 = private unnamed_addr addrspace(4) constant [5 x i8] c"Ran \00", align 1
@.str.26.42 = private unnamed_addr addrspace(4) constant [9 x i8] c" tests. \00", align 1
@.str.27.43 = private unnamed_addr addrspace(4) constant [8 x i8] c" PASS: \00", align 1
@.str.28.44 = private unnamed_addr addrspace(4) constant [8 x i8] c" FAIL: \00", align 1
@.str.29.45 = private unnamed_addr addrspace(4) constant [15 x i8] c"No tests run.\0A\00", align 1
@.str.30 = private unnamed_addr addrspace(4) constant [22 x i8] c"No matching test for \00", align 1
@.str.34.12 = private unnamed_addr addrspace(4) constant [11 x i8] c"Expected: \00", align 1
@.str.35 = private unnamed_addr addrspace(4) constant [11 x i8] c"Which is: \00", align 1
@.str.36.13 = private unnamed_addr addrspace(4) constant [7 x i8] c"To be \00", align 1
@.str.37.14 = private unnamed_addr addrspace(4) constant [3 x i8] c": \00", align 1
@_ZN22__llvm_libc_22_0_0_git7testing4argvE = internal unnamed_addr addrspace(1) global ptr null, align 8
@_ZN22__llvm_libc_22_0_0_git7testing4envpE = internal unnamed_addr addrspace(1) global ptr null, align 8
@.str.53 = private unnamed_addr addrspace(4) constant [17 x i8] c"--gtest_color=no\00", align 1
@.str.1.54 = private unnamed_addr addrspace(4) constant [19 x i8] c"--gtest_print_time\00", align 1
@.str.2.55 = private unnamed_addr addrspace(4) constant [9 x i8] c"--gtest_\00", align 1
@_ZN22__llvm_libc_22_0_0_git3cpp6string14NULL_CHARACTERE = internal addrspace(4) constant i8 0, comdat, align 1
@_ZN12_GLOBAL__N_13ptrE = internal unnamed_addr addrspace(1) global ptr addrspacecast (ptr addrspace(1) @_ZN12_GLOBAL__N_16memoryE to ptr), align 8
@_ZN12_GLOBAL__N_16memoryE = internal addrspace(1) global [65336 x i8] zeroinitializer, align 8
@_ZN22__llvm_libc_22_0_0_git16atexit_callbacksE = internal unnamed_addr addrspace(1) global %"class.__llvm_libc_22_0_0_git::FixedVector" zeroinitializer, align 8
@__llvm_rpc_client = protected addrspace(1) global %"struct.rpc::Client" zeroinitializer, align 8
@__llvm_libc_clock_freq = protected local_unnamed_addr addrspace(4) global i64 100000000, align 8

; Function Attrs: convergent mustprogress nounwind
define protected amdgpu_kernel void @_begin(i32 noundef %argc, ptr noundef %argv, ptr noundef %env) local_unnamed_addr #0 {
entry:
  %_begin.kernarg.segment = call nonnull align 16 dereferenceable(280) ptr addrspace(4) @llvm.amdgcn.kernarg.segment.ptr()
  %argc.kernarg.offset2 = bitcast ptr addrspace(4) %_begin.kernarg.segment to ptr addrspace(4), !amdgpu.uniform !42
  %argc.load = load i32, ptr addrspace(4) %argc.kernarg.offset2, align 16, !invariant.load !42, !noundef !42
  %argv.kernarg.offset = getelementptr inbounds i8, ptr addrspace(4) %_begin.kernarg.segment, i64 8, !amdgpu.uniform !42
  %0 = load <2 x i64>, ptr addrspace(4) %argv.kernarg.offset, align 8, !invariant.load !42
  %env.load4 = extractelement <2 x i64> %0, i32 1
  %1 = inttoptr i64 %env.load4 to ptr
  %2 = ptrtoint ptr %1 to i64
  store atomic i64 %2, ptr addrspace(1) @_ZN22__llvm_libc_22_0_0_git3appE monotonic, align 8
  %3 = load i64, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @_ZN22__llvm_libc_22_0_0_git16atexit_callbacksE, i64 1024), align 8, !tbaa !43
  %cmp.i.not.i.i = icmp ne i64 %3, 64
  br i1 %cmp.i.not.i.i, label %if.end.i.i.i, label %_ZN22__llvm_libc_22_0_0_git6atexitEPFvvE.exit, !amdgpu.uniform !42

if.end.i.i.i:                                     ; preds = %entry
  %inc.i.i.i = add i64 %3, 1
  %arrayidx.i.i.i.i = getelementptr inbounds %"struct.__llvm_libc_22_0_0_git::AtExitUnit", ptr addrspace(1) @_ZN22__llvm_libc_22_0_0_git16atexit_callbacksE, i64 %3
  store <2 x i64> <i64 ptrtoint (ptr @_ZN22__llvm_libc_22_0_0_git17stdc_at_exit_funcEPv to i64), i64 ptrtoint (ptr @_ZN22__llvm_libc_22_0_0_gitL25call_fini_array_callbacksEv to i64)>, ptr addrspace(1) %arrayidx.i.i.i.i, align 8, !tbaa !49
  store i64 %inc.i.i.i, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @_ZN22__llvm_libc_22_0_0_git16atexit_callbacksE, i64 1024), align 8, !tbaa !43
  br label %_ZN22__llvm_libc_22_0_0_git6atexitEPFvvE.exit, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git6atexitEPFvvE.exit:    ; preds = %if.end.i.i.i, %entry
  %cmp3.not.i = icmp ne i64 sub (i64 ptrtoint (ptr addrspacecast (ptr addrspace(1) @__init_array_end to ptr) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(1) @__init_array_start to ptr) to i64)), 0
  br i1 %cmp3.not.i, label %for.body.preheader.i, label %Flow5, !amdgpu.uniform !42

for.body.preheader.i:                             ; preds = %_ZN22__llvm_libc_22_0_0_git6atexitEPFvvE.exit
  %argv.load3 = extractelement <2 x i64> %0, i32 0
  %4 = inttoptr i64 %argv.load3 to ptr
  %sub.ptr.div.i = ashr exact i64 sub (i64 ptrtoint (ptr addrspacecast (ptr addrspace(1) @__init_array_end to ptr) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(1) @__init_array_start to ptr) to i64)), 3
  %umax.i = tail call i64 @llvm.umax.i64(i64 %sub.ptr.div.i, i64 1)
  br label %for.body.i, !amdgpu.uniform !42

Flow5:                                            ; preds = %Flow, %_ZN22__llvm_libc_22_0_0_git6atexitEPFvvE.exit
  br label %_ZN22__llvm_libc_22_0_0_gitL25call_init_array_callbacksEiPPcS1_.exit, !amdgpu.uniform !42

for.body.i:                                       ; preds = %for.body.preheader.i, %for.body.i
  %lsr.iv1 = phi ptr addrspace(1) [ %scevgep, %for.body.i ], [ @__init_array_start, %for.body.preheader.i ], !amdgpu.uniform !42
  %lsr.iv = phi i64 [ %lsr.iv.next, %for.body.i ], [ %umax.i, %for.body.preheader.i ]
  %5 = load i64, ptr addrspace(1) %lsr.iv1, align 8, !tbaa !51
  %6 = inttoptr i64 %5 to ptr
  tail call void %6(i32 noundef %argc.load, ptr noundef %4, ptr noundef %1) #33
  %lsr.iv.next = add i64 %lsr.iv, -1
  %scevgep = getelementptr i8, ptr addrspace(1) %lsr.iv1, i64 8
  %exitcond.not.i = icmp eq i64 %lsr.iv.next, 0
  br i1 %exitcond.not.i, label %Flow, label %for.body.i, !amdgpu.uniform !42

Flow:                                             ; preds = %for.body.i
  br label %Flow5, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_gitL25call_init_array_callbacksEiPPcS1_.exit: ; preds = %Flow5
  ret void
}

; Function Attrs: convergent mustprogress nounwind
define internal void @_ZN22__llvm_libc_22_0_0_gitL25call_fini_array_callbacksEv() #1 {
entry:
  %cmp.not3 = icmp ne i64 sub (i64 ptrtoint (ptr addrspacecast (ptr addrspace(1) @__fini_array_end to ptr) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(1) @__fini_array_start to ptr) to i64)), 0
  br i1 %cmp.not3, label %for.body.preheader, label %Flow2, !amdgpu.uniform !42

for.body.preheader:                               ; preds = %entry
  %sub.ptr.div = ashr exact i64 sub (i64 ptrtoint (ptr addrspacecast (ptr addrspace(1) @__fini_array_end to ptr) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(1) @__fini_array_start to ptr) to i64)), 3
  br label %for.body, !amdgpu.uniform !42

Flow:                                             ; preds = %for.body
  br label %Flow2, !amdgpu.uniform !42

for.cond.cleanup:                                 ; preds = %Flow2
  ret void

Flow2:                                            ; preds = %Flow, %entry
  br label %for.cond.cleanup, !amdgpu.uniform !42

for.body:                                         ; preds = %for.body.preheader, %for.body
  %lsr.iv = phi ptr addrspace(1) [ %scevgep1, %for.body ], [ getelementptr (i8, ptr addrspace(1) @__fini_array_start, i64 add (i64 sub (i64 ptrtoint (ptr addrspacecast (ptr addrspace(1) @__fini_array_end to ptr) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(1) @__fini_array_start to ptr) to i64)), i64 -8)), %for.body.preheader ], !amdgpu.uniform !42
  %i.04 = phi i64 [ %dec, %for.body ], [ %sub.ptr.div, %for.body.preheader ]
  %0 = load i64, ptr addrspace(1) %lsr.iv, align 8, !tbaa !51
  %1 = inttoptr i64 %0 to ptr
  tail call void %1() #33
  %dec = add i64 %i.04, -1
  %scevgep1 = getelementptr i8, ptr addrspace(1) %lsr.iv, i64 -8
  %cmp.not = icmp eq i64 %dec, 0
  br i1 %cmp.not, label %Flow, label %for.body, !amdgpu.uniform !42
}

; Function Attrs: mustprogress nocallback nofree nosync nounwind speculatable willreturn memory(none)
declare i64 @llvm.umax.i64(i64, i64) #2

; Function Attrs: convergent mustprogress nounwind
define protected amdgpu_kernel void @_start(i32 noundef %argc, ptr noundef %argv, ptr noundef %envp, ptr noundef captures(none) %ret) local_unnamed_addr #3 {
entry:
  %_start.kernarg.segment = call nonnull align 16 dereferenceable(288) ptr addrspace(4) @llvm.amdgcn.kernarg.segment.ptr()
  %argc.kernarg.offset35 = bitcast ptr addrspace(4) %_start.kernarg.segment to ptr addrspace(4), !amdgpu.uniform !42
  %argc.load = load i32, ptr addrspace(4) %argc.kernarg.offset35, align 16, !invariant.load !42, !noundef !42
  %argv.kernarg.offset = getelementptr inbounds i8, ptr addrspace(4) %_start.kernarg.segment, i64 8, !amdgpu.uniform !42
  %0 = load <3 x i64>, ptr addrspace(4) %argv.kernarg.offset, align 8, !invariant.load !42
  %argv.load36 = extractelement <3 x i64> %0, i32 0
  %1 = inttoptr i64 %argv.load36 to ptr
  %envp.load37 = extractelement <3 x i64> %0, i32 1
  %2 = inttoptr i64 %envp.load37 to ptr
  %ref.tmp.i = alloca %"struct.__llvm_libc_22_0_0_git::testing::TestOptions", align 8, addrspace(5)
  store ptr %1, ptr addrspace(1) @_ZN22__llvm_libc_22_0_0_git7testing4argvE, align 8, !tbaa !52
  store ptr %2, ptr addrspace(1) @_ZN22__llvm_libc_22_0_0_git7testing4envpE, align 8, !tbaa !52
  call void @llvm.lifetime.start.p5(ptr addrspace(5) %ref.tmp.i) #34
  %cmp67.i.i = icmp sgt i32 %argc.load, 1
  br i1 %cmp67.i.i, label %for.body.i.i.preheader, label %Flow43, !amdgpu.uniform !42

for.body.i.i.preheader:                           ; preds = %entry
  br label %for.body.i.i, !amdgpu.uniform !42

Flow43:                                           ; preds = %Flow42, %entry
  %3 = phi i8 [ %retval.sroa.6.1.i.i.lcssa, %Flow42 ], [ 0, %entry ]
  %4 = phi i8 [ %retval.sroa.3.1.i.i.lcssa, %Flow42 ], [ 1, %entry ]
  %5 = phi ptr [ %retval.sroa.0.1.i.i.lcssa, %Flow42 ], [ null, %entry ]
  br label %main.exit, !amdgpu.uniform !42

for.body.i.i:                                     ; preds = %for.body.i.i.preheader, %cleanup.i.i
  %i.071.i.i = phi i32 [ %inc.i.i, %cleanup.i.i ], [ 1, %for.body.i.i.preheader ]
  %retval.sroa.6.070.i.i = phi i8 [ %retval.sroa.6.1.i.i, %cleanup.i.i ], [ 0, %for.body.i.i.preheader ]
  %retval.sroa.3.069.i.i = phi i8 [ %retval.sroa.3.1.i.i, %cleanup.i.i ], [ 1, %for.body.i.i.preheader ]
  %retval.sroa.0.068.i.i = phi ptr [ %retval.sroa.0.1.i.i, %cleanup.i.i ], [ null, %for.body.i.i.preheader ]
  %6 = addrspacecast ptr %1 to ptr addrspace(1)
  %idxprom.i.i = zext nneg i32 %i.071.i.i to i64
  %arrayidx.i.i = getelementptr inbounds ptr, ptr addrspace(1) %6, i64 %idxprom.i.i, !amdgpu.uniform !42
  %7 = load ptr, ptr addrspace(1) %arrayidx.i.i, align 8, !tbaa !55, !amdgpu.noclobber !42
  %scevgep = getelementptr i8, ptr %7, i64 -1
  br label %for.cond.i.i.i.i, !amdgpu.uniform !42

for.cond.i.i.i.i:                                 ; preds = %for.body.i.i, %for.cond.i.i.i.i
  %phi.broken = phi i64 [ 0, %for.body.i.i ], [ %9, %for.cond.i.i.i.i ]
  %lsr.iv = phi ptr [ %scevgep23, %for.cond.i.i.i.i ], [ %scevgep, %for.body.i.i ]
  %scevgep23 = getelementptr i8, ptr %lsr.iv, i64 1, !amdgpu.uniform !42
  %8 = load i8, ptr %scevgep23, align 1, !tbaa !57
  %cmp.i.i.i.i = icmp eq i8 %8, 0
  %9 = call i64 @llvm.amdgcn.if.break.i64(i1 %cmp.i.i.i.i, i64 %phi.broken)
  %10 = call i1 @llvm.amdgcn.loop.i64(i64 %9)
  br i1 %10, label %_ZN22__llvm_libc_22_0_0_git3cpp11string_viewC2EPKc.exit.i.i, label %for.cond.i.i.i.i

_ZN22__llvm_libc_22_0_0_git3cpp11string_viewC2EPKc.exit.i.i: ; preds = %for.cond.i.i.i.i
  %scevgep23.lcssa = phi ptr [ %scevgep23, %for.cond.i.i.i.i ]
  %.lcssa47 = phi i64 [ %9, %for.cond.i.i.i.i ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa47)
  %sub.ptr.lhs.cast.i.i.i.i = ptrtoint ptr %scevgep23.lcssa to i64
  %sub.ptr.rhs.cast.i.i.i.i = ptrtoint ptr %7 to i64
  %sub.ptr.sub.i.i.i.i = sub i64 %sub.ptr.lhs.cast.i.i.i.i, %sub.ptr.rhs.cast.i.i.i.i
  %cmp.i.i26.i.i = icmp eq i64 %sub.ptr.sub.i.i.i.i, sub (i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) getelementptr inbounds nuw (i8, ptr addrspace(4) @.str.53, i64 16) to ptr) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.53 to ptr) to i64))
  %11 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.i.i26.i.i)
  %12 = extractvalue { i1, i64 } %11, 0
  %13 = extractvalue { i1, i64 } %11, 1
  br i1 %12, label %for.body.i.i.i.i.i.preheader, label %Flow41

for.body.i.i.i.i.i.preheader:                     ; preds = %_ZN22__llvm_libc_22_0_0_git3cpp11string_viewC2EPKc.exit.i.i
  br label %for.body.i.i.i.i.i, !amdgpu.uniform !42

Flow41:                                           ; preds = %_ZN22__llvm_libc_22_0_0_git3cpp11string_view13compareMemoryEPKcS3_m.exit.loopexit.i.i.i.i, %_ZN22__llvm_libc_22_0_0_git3cpp11string_viewC2EPKc.exit.i.i
  %14 = phi i1 [ %22, %_ZN22__llvm_libc_22_0_0_git3cpp11string_view13compareMemoryEPKcS3_m.exit.loopexit.i.i.i.i ], [ true, %_ZN22__llvm_libc_22_0_0_git3cpp11string_viewC2EPKc.exit.i.i ]
  call void @llvm.amdgcn.end.cf.i64(i64 %13)
  %15 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %14)
  %16 = extractvalue { i1, i64 } %15, 0
  %17 = extractvalue { i1, i64 } %15, 1
  br i1 %16, label %if.else.i.i, label %cleanup.i.i

for.body.i.i.i.i.i:                               ; preds = %for.body.i.i.i.i.i.preheader, %for.body.i.i.i.i.i
  %phi.broken44 = phi i64 [ 0, %for.body.i.i.i.i.i.preheader ], [ %20, %for.body.i.i.i.i.i ]
  %lsr.iv27 = phi ptr [ %scevgep28, %for.body.i.i.i.i.i ], [ %7, %for.body.i.i.i.i.i.preheader ], !amdgpu.uniform !42
  %lsr.iv25 = phi ptr addrspace(4) [ %scevgep26, %for.body.i.i.i.i.i ], [ @.str.53, %for.body.i.i.i.i.i.preheader ], !amdgpu.uniform !42
  %lsr.iv24 = phi i64 [ %lsr.iv.next, %for.body.i.i.i.i.i ], [ sub (i64 add (i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) getelementptr inbounds nuw (i8, ptr addrspace(4) @.str.53, i64 16) to ptr) to i64), i64 -1), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.53 to ptr) to i64)), %for.body.i.i.i.i.i.preheader ]
  %18 = load i8, ptr %lsr.iv27, align 1, !tbaa !57
  %19 = load i8, ptr addrspace(4) %lsr.iv25, align 1, !tbaa !57
  %tobool.not.i.i.i.i.i = icmp ne i8 %18, %19
  %exitcond.not.i.i.i.i.i = icmp eq i64 %lsr.iv24, 0
  %or.cond.i.i.i.i.i = select i1 %tobool.not.i.i.i.i.i, i1 true, i1 %exitcond.not.i.i.i.i.i
  %lsr.iv.next = add i64 %lsr.iv24, -1
  %scevgep26 = getelementptr i8, ptr addrspace(4) %lsr.iv25, i64 1
  %scevgep28 = getelementptr i8, ptr %lsr.iv27, i64 1
  %20 = call i64 @llvm.amdgcn.if.break.i64(i1 %or.cond.i.i.i.i.i, i64 %phi.broken44)
  %21 = call i1 @llvm.amdgcn.loop.i64(i64 %20)
  br i1 %21, label %_ZN22__llvm_libc_22_0_0_git3cpp11string_view13compareMemoryEPKcS3_m.exit.loopexit.i.i.i.i, label %for.body.i.i.i.i.i

_ZN22__llvm_libc_22_0_0_git3cpp11string_view13compareMemoryEPKcS3_m.exit.loopexit.i.i.i.i: ; preds = %for.body.i.i.i.i.i
  %.lcssa50 = phi i8 [ %18, %for.body.i.i.i.i.i ]
  %.lcssa49 = phi i8 [ %19, %for.body.i.i.i.i.i ]
  %.lcssa48 = phi i64 [ %20, %for.body.i.i.i.i.i ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa48)
  %22 = icmp ne i8 %.lcssa50, %.lcssa49
  br label %Flow41, !amdgpu.uniform !42

if.else.i.i:                                      ; preds = %Flow41
  %cmp.i.i36.i.i = icmp eq i64 %sub.ptr.sub.i.i.i.i, sub (i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) getelementptr inbounds nuw (i8, ptr addrspace(4) @.str.1.54, i64 18) to ptr) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.1.54 to ptr) to i64))
  %23 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.i.i36.i.i)
  %24 = extractvalue { i1, i64 } %23, 0
  %25 = extractvalue { i1, i64 } %23, 1
  br i1 %24, label %for.body.i.i.i39.i.i.preheader, label %Flow39

for.body.i.i.i39.i.i.preheader:                   ; preds = %if.else.i.i
  br label %for.body.i.i.i39.i.i, !amdgpu.uniform !42

Flow39:                                           ; preds = %_ZN22__llvm_libc_22_0_0_git3cpp11string_view13compareMemoryEPKcS3_m.exit.loopexit.i.i47.i.i, %if.else.i.i
  %26 = phi i1 [ %34, %_ZN22__llvm_libc_22_0_0_git3cpp11string_view13compareMemoryEPKcS3_m.exit.loopexit.i.i47.i.i ], [ true, %if.else.i.i ]
  call void @llvm.amdgcn.end.cf.i64(i64 %25)
  %27 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %26)
  %28 = extractvalue { i1, i64 } %27, 0
  %29 = extractvalue { i1, i64 } %27, 1
  br i1 %28, label %if.else4.i.i, label %Flow40

for.body.i.i.i39.i.i:                             ; preds = %for.body.i.i.i39.i.i.preheader, %for.body.i.i.i39.i.i
  %phi.broken45 = phi i64 [ 0, %for.body.i.i.i39.i.i.preheader ], [ %32, %for.body.i.i.i39.i.i ]
  %lsr.iv33 = phi ptr [ %scevgep34, %for.body.i.i.i39.i.i ], [ %7, %for.body.i.i.i39.i.i.preheader ], !amdgpu.uniform !42
  %lsr.iv31 = phi ptr addrspace(4) [ %scevgep32, %for.body.i.i.i39.i.i ], [ @.str.1.54, %for.body.i.i.i39.i.i.preheader ], !amdgpu.uniform !42
  %lsr.iv29 = phi i64 [ %lsr.iv.next30, %for.body.i.i.i39.i.i ], [ sub (i64 add (i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) getelementptr inbounds nuw (i8, ptr addrspace(4) @.str.1.54, i64 18) to ptr) to i64), i64 -1), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.1.54 to ptr) to i64)), %for.body.i.i.i39.i.i.preheader ]
  %30 = load i8, ptr %lsr.iv33, align 1, !tbaa !57
  %31 = load i8, ptr addrspace(4) %lsr.iv31, align 1, !tbaa !57
  %tobool.not.i.i.i43.i.i = icmp ne i8 %30, %31
  %exitcond.not.i.i.i45.i.i = icmp eq i64 %lsr.iv29, 0
  %or.cond.i.i.i46.i.i = select i1 %tobool.not.i.i.i43.i.i, i1 true, i1 %exitcond.not.i.i.i45.i.i
  %lsr.iv.next30 = add i64 %lsr.iv29, -1
  %scevgep32 = getelementptr i8, ptr addrspace(4) %lsr.iv31, i64 1
  %scevgep34 = getelementptr i8, ptr %lsr.iv33, i64 1
  %32 = call i64 @llvm.amdgcn.if.break.i64(i1 %or.cond.i.i.i46.i.i, i64 %phi.broken45)
  %33 = call i1 @llvm.amdgcn.loop.i64(i64 %32)
  br i1 %33, label %_ZN22__llvm_libc_22_0_0_git3cpp11string_view13compareMemoryEPKcS3_m.exit.loopexit.i.i47.i.i, label %for.body.i.i.i39.i.i

_ZN22__llvm_libc_22_0_0_git3cpp11string_view13compareMemoryEPKcS3_m.exit.loopexit.i.i47.i.i: ; preds = %for.body.i.i.i39.i.i
  %.lcssa53 = phi i8 [ %30, %for.body.i.i.i39.i.i ]
  %.lcssa52 = phi i8 [ %31, %for.body.i.i.i39.i.i ]
  %.lcssa51 = phi i64 [ %32, %for.body.i.i.i39.i.i ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa51)
  %34 = icmp ne i8 %.lcssa53, %.lcssa52
  br label %Flow39, !amdgpu.uniform !42

if.else4.i.i:                                     ; preds = %Flow39
  %cmp.not.i.i.i = icmp uge i64 %sub.ptr.sub.i.i.i.i, sub (i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) getelementptr inbounds nuw (i8, ptr addrspace(4) @.str.2.55, i64 8) to ptr) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.2.55 to ptr) to i64))
  %35 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.not.i.i.i)
  %36 = extractvalue { i1, i64 } %35, 0
  %37 = extractvalue { i1, i64 } %35, 1
  br i1 %36, label %for.body.i.i.i.i.preheader, label %Flow

for.body.i.i.i.i.preheader:                       ; preds = %if.else4.i.i
  br label %for.body.i.i.i.i, !amdgpu.uniform !42

Flow:                                             ; preds = %_ZN22__llvm_libc_22_0_0_git3cpp11string_view13compareMemoryEPKcS3_m.exit.loopexit.i.i.i, %if.else4.i.i
  %38 = phi ptr [ %44, %_ZN22__llvm_libc_22_0_0_git3cpp11string_view13compareMemoryEPKcS3_m.exit.loopexit.i.i.i ], [ %7, %if.else4.i.i ]
  call void @llvm.amdgcn.end.cf.i64(i64 %37)
  br label %Flow40, !amdgpu.uniform !42

for.body.i.i.i.i:                                 ; preds = %for.body.i.i.i.i.preheader, %for.body.i.i.i.i
  %phi.broken46 = phi i64 [ 0, %for.body.i.i.i.i.preheader ], [ %41, %for.body.i.i.i.i ]
  %i.011.i.i.i.i = phi i64 [ %inc.i.i.i.i, %for.body.i.i.i.i ], [ 0, %for.body.i.i.i.i.preheader ]
  %arrayidx.i.i.i.i = getelementptr i8, ptr %7, i64 %i.011.i.i.i.i, !amdgpu.uniform !42
  %39 = load i8, ptr %arrayidx.i.i.i.i, align 1, !tbaa !57
  %arrayidx1.i.i.i.i = getelementptr i8, ptr addrspace(4) @.str.2.55, i64 %i.011.i.i.i.i, !amdgpu.uniform !42
  %40 = load i8, ptr addrspace(4) %arrayidx1.i.i.i.i, align 1, !tbaa !57
  %tobool.not.i.i.i.i = icmp ne i8 %39, %40
  %inc.i.i.i.i = add nuw i64 %i.011.i.i.i.i, 1
  %exitcond.not.i.i.i.i = icmp eq i64 sub (i64 add (i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) getelementptr inbounds nuw (i8, ptr addrspace(4) @.str.2.55, i64 8) to ptr) to i64), i64 -1), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.2.55 to ptr) to i64)), %i.011.i.i.i.i
  %or.cond.i.i.i.i = select i1 %tobool.not.i.i.i.i, i1 true, i1 %exitcond.not.i.i.i.i
  %41 = call i64 @llvm.amdgcn.if.break.i64(i1 %or.cond.i.i.i.i, i64 %phi.broken46)
  %42 = call i1 @llvm.amdgcn.loop.i64(i64 %41)
  br i1 %42, label %_ZN22__llvm_libc_22_0_0_git3cpp11string_view13compareMemoryEPKcS3_m.exit.loopexit.i.i.i, label %for.body.i.i.i.i

_ZN22__llvm_libc_22_0_0_git3cpp11string_view13compareMemoryEPKcS3_m.exit.loopexit.i.i.i: ; preds = %for.body.i.i.i.i
  %.lcssa56 = phi i8 [ %39, %for.body.i.i.i.i ]
  %.lcssa55 = phi i8 [ %40, %for.body.i.i.i.i ]
  %.lcssa54 = phi i64 [ %41, %for.body.i.i.i.i ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa54)
  %43 = icmp eq i8 %.lcssa56, %.lcssa55
  %44 = select i1 %43, ptr %retval.sroa.0.068.i.i, ptr %7
  br label %Flow, !amdgpu.uniform !42

Flow40:                                           ; preds = %Flow, %Flow39
  %45 = phi i8 [ %retval.sroa.6.070.i.i, %Flow ], [ 1, %Flow39 ]
  %46 = phi ptr [ %38, %Flow ], [ %retval.sroa.0.068.i.i, %Flow39 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %29)
  br label %cleanup.i.i, !amdgpu.uniform !42

cleanup.i.i:                                      ; preds = %Flow41, %Flow40
  %retval.sroa.0.1.i.i = phi ptr [ %retval.sroa.0.068.i.i, %Flow41 ], [ %46, %Flow40 ]
  %retval.sroa.3.1.i.i = phi i8 [ 0, %Flow41 ], [ %retval.sroa.3.069.i.i, %Flow40 ]
  %retval.sroa.6.1.i.i = phi i8 [ %retval.sroa.6.070.i.i, %Flow41 ], [ %45, %Flow40 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %17)
  %inc.i.i = add nuw nsw i32 %i.071.i.i, 1
  %exitcond.not.i.i = icmp eq i32 %inc.i.i, %argc.load
  br i1 %exitcond.not.i.i, label %Flow42, label %for.body.i.i, !amdgpu.uniform !42

Flow42:                                           ; preds = %cleanup.i.i
  %retval.sroa.0.1.i.i.lcssa = phi ptr [ %retval.sroa.0.1.i.i, %cleanup.i.i ]
  %retval.sroa.3.1.i.i.lcssa = phi i8 [ %retval.sroa.3.1.i.i, %cleanup.i.i ]
  %retval.sroa.6.1.i.i.lcssa = phi i8 [ %retval.sroa.6.1.i.i, %cleanup.i.i ]
  br label %Flow43, !amdgpu.uniform !42

main.exit:                                        ; preds = %Flow43
  %47 = call ptr @llvm.amdgcn.addrspacecast.nonnull.p0.p5(ptr addrspace(5) %ref.tmp.i)
  store ptr %5, ptr addrspace(5) %ref.tmp.i, align 8
  %tmp.coerce.sroa.2.0.ref.tmp.ascast.sroa_idx.i = getelementptr inbounds nuw i8, ptr addrspace(5) %ref.tmp.i, i32 8
  %48 = insertelement <2 x i8> poison, i8 %4, i32 0
  %49 = insertelement <2 x i8> %48, i8 %3, i32 1
  store <2 x i8> %49, ptr addrspace(5) %tmp.coerce.sroa.2.0.ref.tmp.ascast.sroa_idx.i, align 8
  %call1.i = call fastcc noundef range(i32 0, 2) i32 @_ZN22__llvm_libc_22_0_0_git7testing4Test8runTestsERKNS0_11TestOptionsE(ptr noundef nonnull align 8 dereferenceable(10) %47) #33
  call void @llvm.lifetime.end.p5(ptr addrspace(5) %ref.tmp.i) #34
  %50 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop, !amdgpu.uniform !42

51:                                               ; preds = %ComputeEnd
  %ret.load38 = extractelement <3 x i64> %0, i32 2
  %52 = inttoptr i64 %ret.load38 to ptr
  %53 = addrspacecast ptr %52 to ptr addrspace(1)
  %54 = atomicrmw or ptr addrspace(1) %53, i32 %.lcssa monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %55, !amdgpu.uniform !42

55:                                               ; preds = %51, %ComputeEnd
  call void @llvm.amdgcn.end.cf.i64(i64 %72)
  ret void

ComputeLoop:                                      ; preds = %ComputeLoop, %main.exit
  %Accumulator = phi i32 [ 0, %main.exit ], [ %59, %ComputeLoop ]
  %ActiveBits = phi i64 [ %50, %main.exit ], [ %62, %ComputeLoop ]
  %56 = call i64 @llvm.cttz.i64(i64 %ActiveBits, i1 true)
  %57 = trunc i64 %56 to i32
  %58 = call i32 @llvm.amdgcn.readlane.i32(i32 %call1.i, i32 %57)
  %59 = or i32 %Accumulator, %58
  %60 = shl i64 1, %56
  %61 = xor i64 %60, -1
  %62 = and i64 %ActiveBits, %61
  %63 = icmp eq i64 %62, 0
  br i1 %63, label %ComputeEnd, label %ComputeLoop, !amdgpu.uniform !42

ComputeEnd:                                       ; preds = %ComputeLoop
  %.lcssa = phi i32 [ %59, %ComputeLoop ]
  %64 = trunc i64 %50 to i32
  %65 = lshr i64 %50, 32
  %66 = trunc i64 %65 to i32
  %67 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %64, i32 0)
  %68 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %66, i32 %67)
  %69 = icmp eq i32 %68, 0
  %70 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %69)
  %71 = extractvalue { i1, i64 } %70, 0
  %72 = extractvalue { i1, i64 } %70, 1
  br i1 %71, label %51, label %55
}

; Function Attrs: cold convergent mustprogress noreturn nounwind
define protected amdgpu_kernel void @_end(i32 noundef %retval) local_unnamed_addr #4 !section_prefix !58 {
entry:
  %_end.kernarg.segment = call nonnull align 16 dereferenceable(264) ptr addrspace(4) @llvm.amdgcn.kernarg.segment.ptr()
  %retval.kernarg.offset1 = bitcast ptr addrspace(4) %_end.kernarg.segment to ptr addrspace(4), !amdgpu.uniform !42
  %retval.load = load i32, ptr addrspace(4) %retval.kernarg.offset1, align 16, !invariant.load !42, !noundef !42
  tail call fastcc void @_ZN22__llvm_libc_22_0_0_git4exitEi(i32 noundef %retval.load) #35
  unreachable
}

; Function Attrs: mustprogress nofree norecurse nosync nounwind willreturn memory(readwrite, argmem: write, inaccessiblemem: none)
define internal void @_GLOBAL__sub_I_utf32_to_8_test.cpp() #5 {
entry:
  store ptr null, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @LlvmLibcCharacterConverterUTF32To8Test_OneByte_Instance, i64 16), align 8, !tbaa !59
  store <2 x i64> <i64 ptrtoint (ptr addrspace(1) getelementptr inbounds nuw inrange(-16, 48) (i8, ptr addrspace(1) @_ZTV46LlvmLibcCharacterConverterUTF32To8Test_OneByte, i64 16) to i64), i64 0>, ptr addrspace(1) @LlvmLibcCharacterConverterUTF32To8Test_OneByte_Instance, align 8
  %0 = load ptr, ptr addrspace(1) @_ZN22__llvm_libc_22_0_0_git7testing4Test3EndE, align 8, !tbaa !63
  %cmp.i = icmp ne ptr %0, null
  br i1 %cmp.i, label %if.end.i, label %Flow66, !amdgpu.uniform !42

Flow66:                                           ; preds = %if.end.i, %entry
  %1 = phi i1 [ false, %if.end.i ], [ true, %entry ]
  br i1 %1, label %if.then.i, label %_ZN22__llvm_libc_22_0_0_git7testing4Test7addTestEPS1_.exit, !amdgpu.uniform !42

if.then.i:                                        ; preds = %Flow66
  store ptr addrspacecast (ptr addrspace(1) @LlvmLibcCharacterConverterUTF32To8Test_OneByte_Instance to ptr), ptr addrspace(1) @_ZN22__llvm_libc_22_0_0_git7testing4Test5StartE, align 8, !tbaa !63
  br label %_ZN22__llvm_libc_22_0_0_git7testing4Test7addTestEPS1_.exit, !amdgpu.uniform !42

if.end.i:                                         ; preds = %entry
  %Next.i = getelementptr inbounds nuw i8, ptr %0, i64 8
  store ptr addrspacecast (ptr addrspace(1) @LlvmLibcCharacterConverterUTF32To8Test_OneByte_Instance to ptr), ptr %Next.i, align 8, !tbaa !64
  br label %Flow66, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git7testing4Test7addTestEPS1_.exit: ; preds = %if.then.i, %Flow66
  %2 = load i64, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @_ZN22__llvm_libc_22_0_0_git16atexit_callbacksE, i64 1024), align 8, !tbaa !43
  %cmp.i.not.i.i = icmp ne i64 %2, 64
  br i1 %cmp.i.not.i.i, label %if.end.i.i.i, label %__cxa_atexit.exit, !amdgpu.uniform !42

if.end.i.i.i:                                     ; preds = %_ZN22__llvm_libc_22_0_0_git7testing4Test7addTestEPS1_.exit
  %inc.i.i.i = add i64 %2, 1
  %arrayidx.i.i.i.i = getelementptr inbounds %"struct.__llvm_libc_22_0_0_git::AtExitUnit", ptr addrspace(1) @_ZN22__llvm_libc_22_0_0_git16atexit_callbacksE, i64 %2
  store <2 x i64> <i64 ptrtoint (ptr @_ZN22__llvm_libc_22_0_0_git7testing4TestD2Ev to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(1) @LlvmLibcCharacterConverterUTF32To8Test_OneByte_Instance to ptr) to i64)>, ptr addrspace(1) %arrayidx.i.i.i.i, align 8, !tbaa !49
  store i64 %inc.i.i.i, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @_ZN22__llvm_libc_22_0_0_git16atexit_callbacksE, i64 1024), align 8, !tbaa !43
  br label %__cxa_atexit.exit, !amdgpu.uniform !42

__cxa_atexit.exit:                                ; preds = %if.end.i.i.i, %_ZN22__llvm_libc_22_0_0_git7testing4Test7addTestEPS1_.exit
  %3 = phi i64 [ 64, %_ZN22__llvm_libc_22_0_0_git7testing4Test7addTestEPS1_.exit ], [ %inc.i.i.i, %if.end.i.i.i ]
  store ptr null, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @LlvmLibcCharacterConverterUTF32To8Test_TwoByte_Instance, i64 16), align 8, !tbaa !59
  store <2 x i64> <i64 ptrtoint (ptr addrspace(1) getelementptr inbounds nuw inrange(-16, 48) (i8, ptr addrspace(1) @_ZTV46LlvmLibcCharacterConverterUTF32To8Test_TwoByte, i64 16) to i64), i64 0>, ptr addrspace(1) @LlvmLibcCharacterConverterUTF32To8Test_TwoByte_Instance, align 8
  %cmp.i1 = icmp ne ptr addrspace(1) @LlvmLibcCharacterConverterUTF32To8Test_OneByte_Instance, addrspacecast (ptr null to ptr addrspace(1))
  br i1 %cmp.i1, label %if.end.i2, label %Flow65, !amdgpu.uniform !42

Flow65:                                           ; preds = %if.end.i2, %__cxa_atexit.exit
  %4 = phi i1 [ false, %if.end.i2 ], [ true, %__cxa_atexit.exit ]
  br i1 %4, label %if.then.i4, label %_ZN22__llvm_libc_22_0_0_git7testing4Test7addTestEPS1_.exit5, !amdgpu.uniform !42

if.then.i4:                                       ; preds = %Flow65
  store ptr addrspacecast (ptr addrspace(1) @LlvmLibcCharacterConverterUTF32To8Test_TwoByte_Instance to ptr), ptr addrspace(1) @_ZN22__llvm_libc_22_0_0_git7testing4Test5StartE, align 8, !tbaa !63
  br label %_ZN22__llvm_libc_22_0_0_git7testing4Test7addTestEPS1_.exit5, !amdgpu.uniform !42

if.end.i2:                                        ; preds = %__cxa_atexit.exit
  store ptr addrspacecast (ptr addrspace(1) @LlvmLibcCharacterConverterUTF32To8Test_TwoByte_Instance to ptr), ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @LlvmLibcCharacterConverterUTF32To8Test_OneByte_Instance, i64 8), align 8, !tbaa !64
  br label %Flow65, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git7testing4Test7addTestEPS1_.exit5: ; preds = %if.then.i4, %Flow65
  store ptr addrspacecast (ptr addrspace(1) @LlvmLibcCharacterConverterUTF32To8Test_TwoByte_Instance to ptr), ptr addrspace(1) @_ZN22__llvm_libc_22_0_0_git7testing4Test3EndE, align 8, !tbaa !63
  %cmp.i.not.i.i6 = icmp ne i64 %3, 64
  br i1 %cmp.i.not.i.i6, label %if.end.i.i.i7, label %__cxa_atexit.exit11, !amdgpu.uniform !42

if.end.i.i.i7:                                    ; preds = %_ZN22__llvm_libc_22_0_0_git7testing4Test7addTestEPS1_.exit5
  %inc.i.i.i8 = add i64 %3, 1
  %arrayidx.i.i.i.i9 = getelementptr inbounds %"struct.__llvm_libc_22_0_0_git::AtExitUnit", ptr addrspace(1) @_ZN22__llvm_libc_22_0_0_git16atexit_callbacksE, i64 %3
  store <2 x i64> <i64 ptrtoint (ptr @_ZN22__llvm_libc_22_0_0_git7testing4TestD2Ev to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(1) @LlvmLibcCharacterConverterUTF32To8Test_TwoByte_Instance to ptr) to i64)>, ptr addrspace(1) %arrayidx.i.i.i.i9, align 8, !tbaa !49
  store i64 %inc.i.i.i8, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @_ZN22__llvm_libc_22_0_0_git16atexit_callbacksE, i64 1024), align 8, !tbaa !43
  br label %__cxa_atexit.exit11, !amdgpu.uniform !42

__cxa_atexit.exit11:                              ; preds = %if.end.i.i.i7, %_ZN22__llvm_libc_22_0_0_git7testing4Test7addTestEPS1_.exit5
  %5 = phi i64 [ 64, %_ZN22__llvm_libc_22_0_0_git7testing4Test7addTestEPS1_.exit5 ], [ %inc.i.i.i8, %if.end.i.i.i7 ]
  store ptr null, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @LlvmLibcCharacterConverterUTF32To8Test_ThreeByte_Instance, i64 16), align 8, !tbaa !59
  store <2 x i64> <i64 ptrtoint (ptr addrspace(1) getelementptr inbounds nuw inrange(-16, 48) (i8, ptr addrspace(1) @_ZTV48LlvmLibcCharacterConverterUTF32To8Test_ThreeByte, i64 16) to i64), i64 0>, ptr addrspace(1) @LlvmLibcCharacterConverterUTF32To8Test_ThreeByte_Instance, align 8
  %cmp.i12 = icmp ne ptr addrspace(1) @LlvmLibcCharacterConverterUTF32To8Test_TwoByte_Instance, addrspacecast (ptr null to ptr addrspace(1))
  br i1 %cmp.i12, label %if.end.i13, label %Flow64, !amdgpu.uniform !42

Flow64:                                           ; preds = %if.end.i13, %__cxa_atexit.exit11
  %6 = phi i1 [ false, %if.end.i13 ], [ true, %__cxa_atexit.exit11 ]
  br i1 %6, label %if.then.i15, label %_ZN22__llvm_libc_22_0_0_git7testing4Test7addTestEPS1_.exit16, !amdgpu.uniform !42

if.then.i15:                                      ; preds = %Flow64
  store ptr addrspacecast (ptr addrspace(1) @LlvmLibcCharacterConverterUTF32To8Test_ThreeByte_Instance to ptr), ptr addrspace(1) @_ZN22__llvm_libc_22_0_0_git7testing4Test5StartE, align 8, !tbaa !63
  br label %_ZN22__llvm_libc_22_0_0_git7testing4Test7addTestEPS1_.exit16, !amdgpu.uniform !42

if.end.i13:                                       ; preds = %__cxa_atexit.exit11
  store ptr addrspacecast (ptr addrspace(1) @LlvmLibcCharacterConverterUTF32To8Test_ThreeByte_Instance to ptr), ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @LlvmLibcCharacterConverterUTF32To8Test_TwoByte_Instance, i64 8), align 8, !tbaa !64
  br label %Flow64, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git7testing4Test7addTestEPS1_.exit16: ; preds = %if.then.i15, %Flow64
  %cmp.i.not.i.i17 = icmp ne i64 %5, 64
  br i1 %cmp.i.not.i.i17, label %if.end.i.i.i18, label %__cxa_atexit.exit22, !amdgpu.uniform !42

if.end.i.i.i18:                                   ; preds = %_ZN22__llvm_libc_22_0_0_git7testing4Test7addTestEPS1_.exit16
  %inc.i.i.i19 = add i64 %5, 1
  %arrayidx.i.i.i.i20 = getelementptr inbounds %"struct.__llvm_libc_22_0_0_git::AtExitUnit", ptr addrspace(1) @_ZN22__llvm_libc_22_0_0_git16atexit_callbacksE, i64 %5
  store <2 x i64> <i64 ptrtoint (ptr @_ZN22__llvm_libc_22_0_0_git7testing4TestD2Ev to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(1) @LlvmLibcCharacterConverterUTF32To8Test_ThreeByte_Instance to ptr) to i64)>, ptr addrspace(1) %arrayidx.i.i.i.i20, align 8, !tbaa !49
  store i64 %inc.i.i.i19, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @_ZN22__llvm_libc_22_0_0_git16atexit_callbacksE, i64 1024), align 8, !tbaa !43
  br label %__cxa_atexit.exit22, !amdgpu.uniform !42

__cxa_atexit.exit22:                              ; preds = %if.end.i.i.i18, %_ZN22__llvm_libc_22_0_0_git7testing4Test7addTestEPS1_.exit16
  %7 = phi i64 [ 64, %_ZN22__llvm_libc_22_0_0_git7testing4Test7addTestEPS1_.exit16 ], [ %inc.i.i.i19, %if.end.i.i.i18 ]
  store ptr null, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @LlvmLibcCharacterConverterUTF32To8Test_FourByte_Instance, i64 16), align 8, !tbaa !59
  store <2 x i64> <i64 ptrtoint (ptr addrspace(1) getelementptr inbounds nuw inrange(-16, 48) (i8, ptr addrspace(1) @_ZTV47LlvmLibcCharacterConverterUTF32To8Test_FourByte, i64 16) to i64), i64 0>, ptr addrspace(1) @LlvmLibcCharacterConverterUTF32To8Test_FourByte_Instance, align 8
  %cmp.i23 = icmp ne ptr addrspace(1) @LlvmLibcCharacterConverterUTF32To8Test_ThreeByte_Instance, addrspacecast (ptr null to ptr addrspace(1))
  br i1 %cmp.i23, label %if.end.i24, label %Flow63, !amdgpu.uniform !42

Flow63:                                           ; preds = %if.end.i24, %__cxa_atexit.exit22
  %8 = phi i1 [ false, %if.end.i24 ], [ true, %__cxa_atexit.exit22 ]
  br i1 %8, label %if.then.i26, label %_ZN22__llvm_libc_22_0_0_git7testing4Test7addTestEPS1_.exit27, !amdgpu.uniform !42

if.then.i26:                                      ; preds = %Flow63
  store ptr addrspacecast (ptr addrspace(1) @LlvmLibcCharacterConverterUTF32To8Test_FourByte_Instance to ptr), ptr addrspace(1) @_ZN22__llvm_libc_22_0_0_git7testing4Test5StartE, align 8, !tbaa !63
  br label %_ZN22__llvm_libc_22_0_0_git7testing4Test7addTestEPS1_.exit27, !amdgpu.uniform !42

if.end.i24:                                       ; preds = %__cxa_atexit.exit22
  store ptr addrspacecast (ptr addrspace(1) @LlvmLibcCharacterConverterUTF32To8Test_FourByte_Instance to ptr), ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @LlvmLibcCharacterConverterUTF32To8Test_ThreeByte_Instance, i64 8), align 8, !tbaa !64
  br label %Flow63, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git7testing4Test7addTestEPS1_.exit27: ; preds = %if.then.i26, %Flow63
  store ptr addrspacecast (ptr addrspace(1) @LlvmLibcCharacterConverterUTF32To8Test_FourByte_Instance to ptr), ptr addrspace(1) @_ZN22__llvm_libc_22_0_0_git7testing4Test3EndE, align 8, !tbaa !63
  %cmp.i.not.i.i28 = icmp ne i64 %7, 64
  br i1 %cmp.i.not.i.i28, label %if.end.i.i.i29, label %__cxa_atexit.exit33, !amdgpu.uniform !42

if.end.i.i.i29:                                   ; preds = %_ZN22__llvm_libc_22_0_0_git7testing4Test7addTestEPS1_.exit27
  %inc.i.i.i30 = add i64 %7, 1
  %arrayidx.i.i.i.i31 = getelementptr inbounds %"struct.__llvm_libc_22_0_0_git::AtExitUnit", ptr addrspace(1) @_ZN22__llvm_libc_22_0_0_git16atexit_callbacksE, i64 %7
  store <2 x i64> <i64 ptrtoint (ptr @_ZN22__llvm_libc_22_0_0_git7testing4TestD2Ev to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(1) @LlvmLibcCharacterConverterUTF32To8Test_FourByte_Instance to ptr) to i64)>, ptr addrspace(1) %arrayidx.i.i.i.i31, align 8, !tbaa !49
  store i64 %inc.i.i.i30, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @_ZN22__llvm_libc_22_0_0_git16atexit_callbacksE, i64 1024), align 8, !tbaa !43
  br label %__cxa_atexit.exit33, !amdgpu.uniform !42

__cxa_atexit.exit33:                              ; preds = %if.end.i.i.i29, %_ZN22__llvm_libc_22_0_0_git7testing4Test7addTestEPS1_.exit27
  %9 = phi i64 [ 64, %_ZN22__llvm_libc_22_0_0_git7testing4Test7addTestEPS1_.exit27 ], [ %inc.i.i.i30, %if.end.i.i.i29 ]
  store ptr null, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @LlvmLibcCharacterConverterUTF32To8Test_CantPushMidConversion_Instance, i64 16), align 8, !tbaa !59
  store <2 x i64> <i64 ptrtoint (ptr addrspace(1) getelementptr inbounds nuw inrange(-16, 48) (i8, ptr addrspace(1) @_ZTV60LlvmLibcCharacterConverterUTF32To8Test_CantPushMidConversion, i64 16) to i64), i64 0>, ptr addrspace(1) @LlvmLibcCharacterConverterUTF32To8Test_CantPushMidConversion_Instance, align 8
  %cmp.i34 = icmp ne ptr addrspace(1) @LlvmLibcCharacterConverterUTF32To8Test_FourByte_Instance, addrspacecast (ptr null to ptr addrspace(1))
  br i1 %cmp.i34, label %if.end.i35, label %Flow62, !amdgpu.uniform !42

Flow62:                                           ; preds = %if.end.i35, %__cxa_atexit.exit33
  %10 = phi i1 [ false, %if.end.i35 ], [ true, %__cxa_atexit.exit33 ]
  br i1 %10, label %if.then.i37, label %_ZN22__llvm_libc_22_0_0_git7testing4Test7addTestEPS1_.exit38, !amdgpu.uniform !42

if.then.i37:                                      ; preds = %Flow62
  store ptr addrspacecast (ptr addrspace(1) @LlvmLibcCharacterConverterUTF32To8Test_CantPushMidConversion_Instance to ptr), ptr addrspace(1) @_ZN22__llvm_libc_22_0_0_git7testing4Test5StartE, align 8, !tbaa !63
  br label %_ZN22__llvm_libc_22_0_0_git7testing4Test7addTestEPS1_.exit38, !amdgpu.uniform !42

if.end.i35:                                       ; preds = %__cxa_atexit.exit33
  store ptr addrspacecast (ptr addrspace(1) @LlvmLibcCharacterConverterUTF32To8Test_CantPushMidConversion_Instance to ptr), ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @LlvmLibcCharacterConverterUTF32To8Test_FourByte_Instance, i64 8), align 8, !tbaa !64
  br label %Flow62, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git7testing4Test7addTestEPS1_.exit38: ; preds = %if.then.i37, %Flow62
  %cmp.i.not.i.i39 = icmp ne i64 %9, 64
  br i1 %cmp.i.not.i.i39, label %if.end.i.i.i40, label %__cxa_atexit.exit44, !amdgpu.uniform !42

if.end.i.i.i40:                                   ; preds = %_ZN22__llvm_libc_22_0_0_git7testing4Test7addTestEPS1_.exit38
  %inc.i.i.i41 = add i64 %9, 1
  %arrayidx.i.i.i.i42 = getelementptr inbounds %"struct.__llvm_libc_22_0_0_git::AtExitUnit", ptr addrspace(1) @_ZN22__llvm_libc_22_0_0_git16atexit_callbacksE, i64 %9
  store <2 x i64> <i64 ptrtoint (ptr @_ZN22__llvm_libc_22_0_0_git7testing4TestD2Ev to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(1) @LlvmLibcCharacterConverterUTF32To8Test_CantPushMidConversion_Instance to ptr) to i64)>, ptr addrspace(1) %arrayidx.i.i.i.i42, align 8, !tbaa !49
  store i64 %inc.i.i.i41, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @_ZN22__llvm_libc_22_0_0_git16atexit_callbacksE, i64 1024), align 8, !tbaa !43
  br label %__cxa_atexit.exit44, !amdgpu.uniform !42

__cxa_atexit.exit44:                              ; preds = %if.end.i.i.i40, %_ZN22__llvm_libc_22_0_0_git7testing4Test7addTestEPS1_.exit38
  %11 = phi i64 [ 64, %_ZN22__llvm_libc_22_0_0_git7testing4Test7addTestEPS1_.exit38 ], [ %inc.i.i.i41, %if.end.i.i.i40 ]
  store ptr null, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @LlvmLibcCharacterConverterUTF32To8Test_InvalidState_Instance, i64 16), align 8, !tbaa !59
  store <2 x i64> <i64 ptrtoint (ptr addrspace(1) getelementptr inbounds nuw inrange(-16, 48) (i8, ptr addrspace(1) @_ZTV51LlvmLibcCharacterConverterUTF32To8Test_InvalidState, i64 16) to i64), i64 0>, ptr addrspace(1) @LlvmLibcCharacterConverterUTF32To8Test_InvalidState_Instance, align 8
  %cmp.i45 = icmp ne ptr addrspace(1) @LlvmLibcCharacterConverterUTF32To8Test_CantPushMidConversion_Instance, addrspacecast (ptr null to ptr addrspace(1))
  br i1 %cmp.i45, label %if.end.i46, label %Flow, !amdgpu.uniform !42

Flow:                                             ; preds = %if.end.i46, %__cxa_atexit.exit44
  %12 = phi i1 [ false, %if.end.i46 ], [ true, %__cxa_atexit.exit44 ]
  br i1 %12, label %if.then.i48, label %_ZN22__llvm_libc_22_0_0_git7testing4Test7addTestEPS1_.exit49, !amdgpu.uniform !42

if.then.i48:                                      ; preds = %Flow
  store ptr addrspacecast (ptr addrspace(1) @LlvmLibcCharacterConverterUTF32To8Test_InvalidState_Instance to ptr), ptr addrspace(1) @_ZN22__llvm_libc_22_0_0_git7testing4Test5StartE, align 8, !tbaa !63
  br label %_ZN22__llvm_libc_22_0_0_git7testing4Test7addTestEPS1_.exit49, !amdgpu.uniform !42

if.end.i46:                                       ; preds = %__cxa_atexit.exit44
  store ptr addrspacecast (ptr addrspace(1) @LlvmLibcCharacterConverterUTF32To8Test_InvalidState_Instance to ptr), ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @LlvmLibcCharacterConverterUTF32To8Test_CantPushMidConversion_Instance, i64 8), align 8, !tbaa !64
  br label %Flow, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git7testing4Test7addTestEPS1_.exit49: ; preds = %if.then.i48, %Flow
  store ptr addrspacecast (ptr addrspace(1) @LlvmLibcCharacterConverterUTF32To8Test_InvalidState_Instance to ptr), ptr addrspace(1) @_ZN22__llvm_libc_22_0_0_git7testing4Test3EndE, align 8, !tbaa !63
  %cmp.i.not.i.i50 = icmp ne i64 %11, 64
  br i1 %cmp.i.not.i.i50, label %if.end.i.i.i51, label %__cxa_atexit.exit55, !amdgpu.uniform !42

if.end.i.i.i51:                                   ; preds = %_ZN22__llvm_libc_22_0_0_git7testing4Test7addTestEPS1_.exit49
  %inc.i.i.i52 = add i64 %11, 1
  %arrayidx.i.i.i.i53 = getelementptr inbounds %"struct.__llvm_libc_22_0_0_git::AtExitUnit", ptr addrspace(1) @_ZN22__llvm_libc_22_0_0_git16atexit_callbacksE, i64 %11
  store <2 x i64> <i64 ptrtoint (ptr @_ZN22__llvm_libc_22_0_0_git7testing4TestD2Ev to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(1) @LlvmLibcCharacterConverterUTF32To8Test_InvalidState_Instance to ptr) to i64)>, ptr addrspace(1) %arrayidx.i.i.i.i53, align 8, !tbaa !49
  store i64 %inc.i.i.i52, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @_ZN22__llvm_libc_22_0_0_git16atexit_callbacksE, i64 1024), align 8, !tbaa !43
  br label %__cxa_atexit.exit55, !amdgpu.uniform !42

__cxa_atexit.exit55:                              ; preds = %if.end.i.i.i51, %_ZN22__llvm_libc_22_0_0_git7testing4Test7addTestEPS1_.exit49
  ret void
}

; Function Attrs: mustprogress nofree norecurse nosync nounwind willreturn memory(none)
define internal void @_ZN22__llvm_libc_22_0_0_git7testing4TestD2Ev(ptr nonnull readnone align 8 captures(none) %this) unnamed_addr #6 comdat align 2 {
entry:
  ret void
}

; Function Attrs: cold inlinehint mustprogress noreturn nounwind memory(inaccessiblemem: write)
define internal void @_ZN51LlvmLibcCharacterConverterUTF32To8Test_InvalidStateD0Ev(ptr noundef nonnull readnone align 8 captures(none) dereferenceable(24) %this) unnamed_addr #7 comdat align 2 !section_prefix !58 {
entry:
  tail call void @llvm.trap()
  unreachable
}

; Function Attrs: mustprogress nofree norecurse nosync nounwind willreturn memory(none)
define internal void @_ZN22__llvm_libc_22_0_0_git7testing4Test5SetUpEv(ptr nonnull readnone align 8 captures(none) %this) unnamed_addr #6 comdat align 2 {
entry:
  ret void
}

; Function Attrs: mustprogress nofree norecurse nosync nounwind willreturn memory(none)
define internal void @_ZN22__llvm_libc_22_0_0_git7testing4Test8TearDownEv(ptr nonnull readnone align 8 captures(none) %this) unnamed_addr #6 comdat align 2 {
entry:
  ret void
}

; Function Attrs: convergent mustprogress nofree norecurse nounwind
define internal void @_ZN51LlvmLibcCharacterConverterUTF32To8Test_InvalidState3RunEv(ptr noundef nonnull readonly align 8 captures(none) dereferenceable(24) %this) unnamed_addr #8 align 2 {
entry:
  %Ctx.i = getelementptr inbounds nuw i8, ptr %this, i64 16
  %0 = load ptr, ptr %Ctx.i, align 8, !tbaa !59
  %call.i = tail call fastcc noundef zeroext i1 @_ZN22__llvm_libc_22_0_0_git7testing8internal4testIbEEbPNS1_10RunContextENS0_8TestCondET_S6_PKcS8_NS1_8LocationE(ptr noundef %0, i1 noundef zeroext true, i1 noundef zeroext true, ptr noundef addrspacecast (ptr addrspace(4) @.str.36 to ptr), ptr noundef addrspacecast (ptr addrspace(4) @.str.1 to ptr), i32 193) #33
  %1 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %call.i)
  %2 = extractvalue { i1, i64 } %1, 0
  %3 = extractvalue { i1, i64 } %1, 1
  br i1 %2, label %_ZN22__llvm_libc_22_0_0_git8internal18CharacterConverter12isValidStateEv.exit19, label %UnifiedReturnBlock

_ZN22__llvm_libc_22_0_0_git8internal18CharacterConverter12isValidStateEv.exit19: ; preds = %entry
  %sunkaddr = getelementptr inbounds i8, ptr %this, i64 16
  %4 = load ptr, ptr %sunkaddr, align 8, !tbaa !59
  %call.i254 = tail call fastcc noundef zeroext i1 @_ZN22__llvm_libc_22_0_0_git7testing8internal4testIbEEbPNS1_10RunContextENS0_8TestCondET_S6_PKcS8_NS1_8LocationE(ptr noundef %4, i1 noundef zeroext false, i1 noundef zeroext false, ptr noundef addrspacecast (ptr addrspace(4) @.str.37 to ptr), ptr noundef addrspacecast (ptr addrspace(4) @.str.8 to ptr), i32 197) #33
  %5 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %call.i254)
  %6 = extractvalue { i1, i64 } %5, 0
  %7 = extractvalue { i1, i64 } %5, 1
  br i1 %6, label %_ZN22__llvm_libc_22_0_0_git8internal18CharacterConverter12isValidStateEv.exit36, label %Flow189

_ZN22__llvm_libc_22_0_0_git8internal18CharacterConverter12isValidStateEv.exit36: ; preds = %_ZN22__llvm_libc_22_0_0_git8internal18CharacterConverter12isValidStateEv.exit19
  %sunkaddr173 = getelementptr inbounds i8, ptr %this, i64 16
  %8 = load ptr, ptr %sunkaddr173, align 8, !tbaa !59
  %call.i257 = tail call fastcc noundef zeroext i1 @_ZN22__llvm_libc_22_0_0_git7testing8internal4testIbEEbPNS1_10RunContextENS0_8TestCondET_S6_PKcS8_NS1_8LocationE(ptr noundef %8, i1 noundef zeroext true, i1 noundef zeroext true, ptr noundef addrspacecast (ptr addrspace(4) @.str.38 to ptr), ptr noundef addrspacecast (ptr addrspace(4) @.str.1 to ptr), i32 201) #33
  %9 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %call.i257)
  %10 = extractvalue { i1, i64 } %9, 0
  %11 = extractvalue { i1, i64 } %9, 1
  br i1 %10, label %_ZN22__llvm_libc_22_0_0_git8internal18CharacterConverter12isValidStateEv.exit53, label %Flow188

_ZN22__llvm_libc_22_0_0_git8internal18CharacterConverter12isValidStateEv.exit53: ; preds = %_ZN22__llvm_libc_22_0_0_git8internal18CharacterConverter12isValidStateEv.exit36
  %sunkaddr174 = getelementptr inbounds i8, ptr %this, i64 16
  %12 = load ptr, ptr %sunkaddr174, align 8, !tbaa !59
  %call.i260 = tail call fastcc noundef zeroext i1 @_ZN22__llvm_libc_22_0_0_git7testing8internal4testIbEEbPNS1_10RunContextENS0_8TestCondET_S6_PKcS8_NS1_8LocationE(ptr noundef %12, i1 noundef zeroext false, i1 noundef zeroext false, ptr noundef addrspacecast (ptr addrspace(4) @.str.39 to ptr), ptr noundef addrspacecast (ptr addrspace(4) @.str.8 to ptr), i32 204) #33
  %13 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %call.i260)
  %14 = extractvalue { i1, i64 } %13, 0
  %15 = extractvalue { i1, i64 } %13, 1
  br i1 %14, label %_ZN22__llvm_libc_22_0_0_git8internal18CharacterConverter12isValidStateEv.exit70, label %Flow187

_ZN22__llvm_libc_22_0_0_git8internal18CharacterConverter12isValidStateEv.exit70: ; preds = %_ZN22__llvm_libc_22_0_0_git8internal18CharacterConverter12isValidStateEv.exit53
  %sunkaddr175 = getelementptr inbounds i8, ptr %this, i64 16
  %16 = load ptr, ptr %sunkaddr175, align 8, !tbaa !59
  %call.i263 = tail call fastcc noundef zeroext i1 @_ZN22__llvm_libc_22_0_0_git7testing8internal4testIbEEbPNS1_10RunContextENS0_8TestCondET_S6_PKcS8_NS1_8LocationE(ptr noundef %16, i1 noundef zeroext true, i1 noundef zeroext true, ptr noundef addrspacecast (ptr addrspace(4) @.str.40 to ptr), ptr noundef addrspacecast (ptr addrspace(4) @.str.1 to ptr), i32 208) #33
  %17 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %call.i263)
  %18 = extractvalue { i1, i64 } %17, 0
  %19 = extractvalue { i1, i64 } %17, 1
  br i1 %18, label %_ZN22__llvm_libc_22_0_0_git8internal18CharacterConverter12isValidStateEv.exit87, label %Flow186

_ZN22__llvm_libc_22_0_0_git8internal18CharacterConverter12isValidStateEv.exit87: ; preds = %_ZN22__llvm_libc_22_0_0_git8internal18CharacterConverter12isValidStateEv.exit70
  %sunkaddr176 = getelementptr inbounds i8, ptr %this, i64 16
  %20 = load ptr, ptr %sunkaddr176, align 8, !tbaa !59
  %call.i266 = tail call fastcc noundef zeroext i1 @_ZN22__llvm_libc_22_0_0_git7testing8internal4testIbEEbPNS1_10RunContextENS0_8TestCondET_S6_PKcS8_NS1_8LocationE(ptr noundef %20, i1 noundef zeroext false, i1 noundef zeroext false, ptr noundef addrspacecast (ptr addrspace(4) @.str.41 to ptr), ptr noundef addrspacecast (ptr addrspace(4) @.str.8 to ptr), i32 211) #33
  %21 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %call.i266)
  %22 = extractvalue { i1, i64 } %21, 0
  %23 = extractvalue { i1, i64 } %21, 1
  br i1 %22, label %_ZN22__llvm_libc_22_0_0_git8internal18CharacterConverter12isValidStateEv.exit104, label %Flow185

_ZN22__llvm_libc_22_0_0_git8internal18CharacterConverter12isValidStateEv.exit104: ; preds = %_ZN22__llvm_libc_22_0_0_git8internal18CharacterConverter12isValidStateEv.exit87
  %sunkaddr177 = getelementptr inbounds i8, ptr %this, i64 16
  %24 = load ptr, ptr %sunkaddr177, align 8, !tbaa !59
  %call.i269 = tail call fastcc noundef zeroext i1 @_ZN22__llvm_libc_22_0_0_git7testing8internal4testIbEEbPNS1_10RunContextENS0_8TestCondET_S6_PKcS8_NS1_8LocationE(ptr noundef %24, i1 noundef zeroext true, i1 noundef zeroext true, ptr noundef addrspacecast (ptr addrspace(4) @.str.42 to ptr), ptr noundef addrspacecast (ptr addrspace(4) @.str.1 to ptr), i32 215) #33
  %25 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %call.i269)
  %26 = extractvalue { i1, i64 } %25, 0
  %27 = extractvalue { i1, i64 } %25, 1
  br i1 %26, label %_ZN22__llvm_libc_22_0_0_git8internal18CharacterConverter12isValidStateEv.exit121, label %Flow184

_ZN22__llvm_libc_22_0_0_git8internal18CharacterConverter12isValidStateEv.exit121: ; preds = %_ZN22__llvm_libc_22_0_0_git8internal18CharacterConverter12isValidStateEv.exit104
  %sunkaddr178 = getelementptr inbounds i8, ptr %this, i64 16
  %28 = load ptr, ptr %sunkaddr178, align 8, !tbaa !59
  %call.i272 = tail call fastcc noundef zeroext i1 @_ZN22__llvm_libc_22_0_0_git7testing8internal4testIbEEbPNS1_10RunContextENS0_8TestCondET_S6_PKcS8_NS1_8LocationE(ptr noundef %28, i1 noundef zeroext false, i1 noundef zeroext false, ptr noundef addrspacecast (ptr addrspace(4) @.str.43 to ptr), ptr noundef addrspacecast (ptr addrspace(4) @.str.8 to ptr), i32 218) #33
  %29 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %call.i272)
  %30 = extractvalue { i1, i64 } %29, 0
  %31 = extractvalue { i1, i64 } %29, 1
  br i1 %30, label %_ZN22__llvm_libc_22_0_0_git8internal18CharacterConverter12isValidStateEv.exit138, label %Flow183

_ZN22__llvm_libc_22_0_0_git8internal18CharacterConverter12isValidStateEv.exit138: ; preds = %_ZN22__llvm_libc_22_0_0_git8internal18CharacterConverter12isValidStateEv.exit121
  %sunkaddr179 = getelementptr inbounds i8, ptr %this, i64 16
  %32 = load ptr, ptr %sunkaddr179, align 8, !tbaa !59
  %call.i275 = tail call fastcc noundef zeroext i1 @_ZN22__llvm_libc_22_0_0_git7testing8internal4testIbEEbPNS1_10RunContextENS0_8TestCondET_S6_PKcS8_NS1_8LocationE(ptr noundef %32, i1 noundef zeroext true, i1 noundef zeroext true, ptr noundef addrspacecast (ptr addrspace(4) @.str.44 to ptr), ptr noundef addrspacecast (ptr addrspace(4) @.str.1 to ptr), i32 222) #33
  %33 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %call.i275)
  %34 = extractvalue { i1, i64 } %33, 0
  %35 = extractvalue { i1, i64 } %33, 1
  br i1 %34, label %_ZN22__llvm_libc_22_0_0_git8internal18CharacterConverter12isValidStateEv.exit155, label %Flow182

_ZN22__llvm_libc_22_0_0_git8internal18CharacterConverter12isValidStateEv.exit155: ; preds = %_ZN22__llvm_libc_22_0_0_git8internal18CharacterConverter12isValidStateEv.exit138
  %sunkaddr180 = getelementptr inbounds i8, ptr %this, i64 16
  %36 = load ptr, ptr %sunkaddr180, align 8, !tbaa !59
  %call.i278 = tail call fastcc noundef zeroext i1 @_ZN22__llvm_libc_22_0_0_git7testing8internal4testIbEEbPNS1_10RunContextENS0_8TestCondET_S6_PKcS8_NS1_8LocationE(ptr noundef %36, i1 noundef zeroext false, i1 noundef zeroext false, ptr noundef addrspacecast (ptr addrspace(4) @.str.45 to ptr), ptr noundef addrspacecast (ptr addrspace(4) @.str.8 to ptr), i32 225) #33
  %37 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %call.i278)
  %38 = extractvalue { i1, i64 } %37, 0
  %39 = extractvalue { i1, i64 } %37, 1
  br i1 %38, label %_ZN22__llvm_libc_22_0_0_git8internal18CharacterConverter12isValidStateEv.exit172, label %Flow

_ZN22__llvm_libc_22_0_0_git8internal18CharacterConverter12isValidStateEv.exit172: ; preds = %_ZN22__llvm_libc_22_0_0_git8internal18CharacterConverter12isValidStateEv.exit155
  %sunkaddr181 = getelementptr inbounds i8, ptr %this, i64 16
  %40 = load ptr, ptr %sunkaddr181, align 8, !tbaa !59
  %41 = tail call fastcc noundef zeroext i1 @_ZN22__llvm_libc_22_0_0_git7testing8internal4testIbEEbPNS1_10RunContextENS0_8TestCondET_S6_PKcS8_NS1_8LocationE(ptr noundef %40, i1 noundef zeroext false, i1 noundef zeroext false, ptr noundef addrspacecast (ptr addrspace(4) @.str.46 to ptr), ptr noundef addrspacecast (ptr addrspace(4) @.str.8 to ptr), i32 229) #33
  br label %Flow, !amdgpu.uniform !42

Flow:                                             ; preds = %_ZN22__llvm_libc_22_0_0_git8internal18CharacterConverter12isValidStateEv.exit172, %_ZN22__llvm_libc_22_0_0_git8internal18CharacterConverter12isValidStateEv.exit155
  call void @llvm.amdgcn.end.cf.i64(i64 %39)
  br label %Flow182, !amdgpu.uniform !42

Flow182:                                          ; preds = %Flow, %_ZN22__llvm_libc_22_0_0_git8internal18CharacterConverter12isValidStateEv.exit138
  call void @llvm.amdgcn.end.cf.i64(i64 %35)
  br label %Flow183, !amdgpu.uniform !42

Flow183:                                          ; preds = %Flow182, %_ZN22__llvm_libc_22_0_0_git8internal18CharacterConverter12isValidStateEv.exit121
  call void @llvm.amdgcn.end.cf.i64(i64 %31)
  br label %Flow184, !amdgpu.uniform !42

Flow184:                                          ; preds = %Flow183, %_ZN22__llvm_libc_22_0_0_git8internal18CharacterConverter12isValidStateEv.exit104
  call void @llvm.amdgcn.end.cf.i64(i64 %27)
  br label %Flow185, !amdgpu.uniform !42

Flow185:                                          ; preds = %Flow184, %_ZN22__llvm_libc_22_0_0_git8internal18CharacterConverter12isValidStateEv.exit87
  call void @llvm.amdgcn.end.cf.i64(i64 %23)
  br label %Flow186, !amdgpu.uniform !42

Flow186:                                          ; preds = %Flow185, %_ZN22__llvm_libc_22_0_0_git8internal18CharacterConverter12isValidStateEv.exit70
  call void @llvm.amdgcn.end.cf.i64(i64 %19)
  br label %Flow187, !amdgpu.uniform !42

Flow187:                                          ; preds = %Flow186, %_ZN22__llvm_libc_22_0_0_git8internal18CharacterConverter12isValidStateEv.exit53
  call void @llvm.amdgcn.end.cf.i64(i64 %15)
  br label %Flow188, !amdgpu.uniform !42

Flow188:                                          ; preds = %Flow187, %_ZN22__llvm_libc_22_0_0_git8internal18CharacterConverter12isValidStateEv.exit36
  call void @llvm.amdgcn.end.cf.i64(i64 %11)
  br label %Flow189, !amdgpu.uniform !42

Flow189:                                          ; preds = %Flow188, %_ZN22__llvm_libc_22_0_0_git8internal18CharacterConverter12isValidStateEv.exit19
  call void @llvm.amdgcn.end.cf.i64(i64 %7)
  br label %UnifiedReturnBlock, !amdgpu.uniform !42

UnifiedReturnBlock:                               ; preds = %entry, %Flow189
  call void @llvm.amdgcn.end.cf.i64(i64 %3)
  ret void
}

; Function Attrs: mustprogress nofree norecurse nosync nounwind willreturn memory(none)
define internal noundef ptr @_ZNK51LlvmLibcCharacterConverterUTF32To8Test_InvalidState7getNameEv(ptr nonnull readnone align 8 captures(none) %this) unnamed_addr #6 comdat align 2 {
entry:
  ret ptr addrspacecast (ptr addrspace(4) @.str.52 to ptr)
}

; Function Attrs: mustprogress nocallback nofree nosync nounwind willreturn memory(argmem: readwrite)
declare void @llvm.lifetime.start.p5(ptr addrspace(5) captures(none)) #9

; Function Attrs: mustprogress nocallback nofree nosync nounwind willreturn memory(argmem: readwrite)
declare void @llvm.lifetime.end.p5(ptr addrspace(5) captures(none)) #9

; Function Attrs: cold inlinehint mustprogress noreturn nounwind memory(inaccessiblemem: write)
define internal void @_ZN60LlvmLibcCharacterConverterUTF32To8Test_CantPushMidConversionD0Ev(ptr noundef nonnull readnone align 8 captures(none) dereferenceable(24) %this) unnamed_addr #7 comdat align 2 !section_prefix !58 {
entry:
  tail call void @llvm.trap()
  unreachable
}

; Function Attrs: convergent mustprogress nofree norecurse nounwind
define internal void @_ZN60LlvmLibcCharacterConverterUTF32To8Test_CantPushMidConversion3RunEv(ptr noundef nonnull readonly align 8 captures(none) dereferenceable(24) %this) unnamed_addr #8 align 2 {
entry:
  %Ctx.i = getelementptr inbounds nuw i8, ptr %this, i64 16
  %0 = load ptr, ptr %Ctx.i, align 8, !tbaa !59
  %call.i = tail call fastcc noundef zeroext i1 @_ZN22__llvm_libc_22_0_0_git7testing8internal4testIiEEbPNS1_10RunContextENS0_8TestCondET_S6_PKcS8_NS1_8LocationE(ptr noundef %0, i32 noundef 0, i32 noundef 0, ptr noundef addrspacecast (ptr addrspace(4) @.str.31 to ptr), ptr noundef addrspacecast (ptr addrspace(4) @.str.32 to ptr), i32 181) #33
  %1 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %call.i)
  %2 = extractvalue { i1, i64 } %1, 0
  %3 = extractvalue { i1, i64 } %1, 1
  br i1 %2, label %_ZN22__llvm_libc_22_0_0_git8internal18CharacterConverter8pop_utf8Ev.exit, label %UnifiedReturnBlock

_ZN22__llvm_libc_22_0_0_git8internal18CharacterConverter8pop_utf8Ev.exit: ; preds = %entry
  %sunkaddr = getelementptr inbounds i8, ptr %this, i64 16
  %4 = load ptr, ptr %sunkaddr, align 8, !tbaa !59
  %call.i52 = tail call fastcc noundef zeroext i1 @_ZN22__llvm_libc_22_0_0_git7testing8internal4testIbEEbPNS1_10RunContextENS0_8TestCondET_S6_PKcS8_NS1_8LocationE(ptr noundef %4, i1 noundef zeroext true, i1 noundef zeroext true, ptr noundef addrspacecast (ptr addrspace(4) @.str.3 to ptr), ptr noundef addrspacecast (ptr addrspace(4) @.str.1 to ptr), i32 183) #33
  %5 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %call.i52)
  %6 = extractvalue { i1, i64 } %5, 0
  %7 = extractvalue { i1, i64 } %5, 1
  br i1 %6, label %_ZN22__llvm_libc_22_0_0_git8internal18CharacterConverter4pushEDi.exit12, label %Flow

_ZN22__llvm_libc_22_0_0_git8internal18CharacterConverter4pushEDi.exit12: ; preds = %_ZN22__llvm_libc_22_0_0_git8internal18CharacterConverter8pop_utf8Ev.exit
  %sunkaddr15 = getelementptr inbounds i8, ptr %this, i64 16
  %8 = load ptr, ptr %sunkaddr15, align 8, !tbaa !59
  %9 = tail call fastcc noundef zeroext i1 @_ZN22__llvm_libc_22_0_0_git7testing8internal4testIiEEbPNS1_10RunContextENS0_8TestCondET_S6_PKcS8_NS1_8LocationE(ptr noundef %8, i32 noundef -1, i32 noundef -1, ptr noundef addrspacecast (ptr addrspace(4) @.str.33 to ptr), ptr noundef addrspacecast (ptr addrspace(4) @.str.34 to ptr), i32 187) #33
  br label %Flow, !amdgpu.uniform !42

Flow:                                             ; preds = %_ZN22__llvm_libc_22_0_0_git8internal18CharacterConverter4pushEDi.exit12, %_ZN22__llvm_libc_22_0_0_git8internal18CharacterConverter8pop_utf8Ev.exit
  call void @llvm.amdgcn.end.cf.i64(i64 %7)
  br label %UnifiedReturnBlock, !amdgpu.uniform !42

UnifiedReturnBlock:                               ; preds = %entry, %Flow
  call void @llvm.amdgcn.end.cf.i64(i64 %3)
  ret void
}

; Function Attrs: mustprogress nofree norecurse nosync nounwind willreturn memory(none)
define internal noundef ptr @_ZNK60LlvmLibcCharacterConverterUTF32To8Test_CantPushMidConversion7getNameEv(ptr nonnull readnone align 8 captures(none) %this) unnamed_addr #6 comdat align 2 {
entry:
  ret ptr addrspacecast (ptr addrspace(4) @.str.51 to ptr)
}

; Function Attrs: cold inlinehint mustprogress noreturn nounwind memory(inaccessiblemem: write)
define internal void @_ZN47LlvmLibcCharacterConverterUTF32To8Test_FourByteD0Ev(ptr noundef nonnull readnone align 8 captures(none) dereferenceable(24) %this) unnamed_addr #7 comdat align 2 !section_prefix !58 {
entry:
  tail call void @llvm.trap()
  unreachable
}

; Function Attrs: convergent mustprogress nofree norecurse nounwind
define internal void @_ZN47LlvmLibcCharacterConverterUTF32To8Test_FourByte3RunEv(ptr noundef nonnull readonly align 8 captures(none) dereferenceable(24) %this) unnamed_addr #8 align 2 {
entry:
  %Ctx.i = getelementptr inbounds nuw i8, ptr %this, i64 16
  %0 = load ptr, ptr %Ctx.i, align 8, !tbaa !59
  %call.i = tail call fastcc noundef zeroext i1 @_ZN22__llvm_libc_22_0_0_git7testing8internal4testIbEEbPNS1_10RunContextENS0_8TestCondET_S6_PKcS8_NS1_8LocationE(ptr noundef %0, i1 noundef zeroext true, i1 noundef zeroext true, ptr noundef addrspacecast (ptr addrspace(4) @.str to ptr), ptr noundef addrspacecast (ptr addrspace(4) @.str.1 to ptr), i32 130) #33
  %1 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %call.i)
  %2 = extractvalue { i1, i64 } %1, 0
  %3 = extractvalue { i1, i64 } %1, 1
  br i1 %2, label %_ZN22__llvm_libc_22_0_0_git8internal18CharacterConverter8pop_utf8Ev.exit, label %UnifiedReturnBlock

_ZN22__llvm_libc_22_0_0_git8internal18CharacterConverter8pop_utf8Ev.exit: ; preds = %entry
  %sunkaddr = getelementptr inbounds i8, ptr %this, i64 16
  %4 = load ptr, ptr %sunkaddr, align 8, !tbaa !59
  %call.i569 = tail call fastcc noundef zeroext i1 @_ZN22__llvm_libc_22_0_0_git7testing8internal4testIbEEbPNS1_10RunContextENS0_8TestCondET_S6_PKcS8_NS1_8LocationE(ptr noundef %4, i1 noundef zeroext true, i1 noundef zeroext true, ptr noundef addrspacecast (ptr addrspace(4) @.str.3 to ptr), ptr noundef addrspacecast (ptr addrspace(4) @.str.1 to ptr), i32 132) #33
  %5 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %call.i569)
  %6 = extractvalue { i1, i64 } %5, 0
  %7 = extractvalue { i1, i64 } %5, 1
  br i1 %6, label %if.end13, label %Flow372

if.end13:                                         ; preds = %_ZN22__llvm_libc_22_0_0_git8internal18CharacterConverter8pop_utf8Ev.exit
  %sunkaddr324 = getelementptr inbounds i8, ptr %this, i64 16
  %8 = load ptr, ptr %sunkaddr324, align 8, !tbaa !59
  %call.i572 = tail call fastcc noundef zeroext i1 @_ZN22__llvm_libc_22_0_0_git7testing8internal4testIiEEbPNS1_10RunContextENS0_8TestCondET_S6_PKcS8_NS1_8LocationE(ptr noundef %8, i32 noundef poison, i32 noundef 240, ptr noundef addrspacecast (ptr addrspace(4) @.str.10 to ptr), ptr noundef addrspacecast (ptr addrspace(4) @.str.24 to ptr), i32 133) #33
  %9 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %call.i572)
  %10 = extractvalue { i1, i64 } %9, 0
  %11 = extractvalue { i1, i64 } %9, 1
  br i1 %10, label %if.end21, label %Flow371

if.end21:                                         ; preds = %if.end13
  %sunkaddr325 = getelementptr inbounds i8, ptr %this, i64 16
  %12 = load ptr, ptr %sunkaddr325, align 8, !tbaa !59
  %call.i575 = tail call fastcc noundef zeroext i1 @_ZN22__llvm_libc_22_0_0_git7testing8internal4testIbEEbPNS1_10RunContextENS0_8TestCondET_S6_PKcS8_NS1_8LocationE(ptr noundef %12, i1 noundef zeroext true, i1 noundef zeroext true, ptr noundef addrspacecast (ptr addrspace(4) @.str.12 to ptr), ptr noundef addrspacecast (ptr addrspace(4) @.str.1 to ptr), i32 134) #33
  %13 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %call.i575)
  %14 = extractvalue { i1, i64 } %13, 0
  %15 = extractvalue { i1, i64 } %13, 1
  br i1 %14, label %_ZN22__llvm_libc_22_0_0_git8internal18CharacterConverter8pop_utf8Ev.exit39, label %Flow370

_ZN22__llvm_libc_22_0_0_git8internal18CharacterConverter8pop_utf8Ev.exit39: ; preds = %if.end21
  %sunkaddr326 = getelementptr inbounds i8, ptr %this, i64 16
  %16 = load ptr, ptr %sunkaddr326, align 8, !tbaa !59
  %call.i580 = tail call fastcc noundef zeroext i1 @_ZN22__llvm_libc_22_0_0_git7testing8internal4testIbEEbPNS1_10RunContextENS0_8TestCondET_S6_PKcS8_NS1_8LocationE(ptr noundef %16, i1 noundef zeroext true, i1 noundef zeroext true, ptr noundef addrspacecast (ptr addrspace(4) @.str.3 to ptr), ptr noundef addrspacecast (ptr addrspace(4) @.str.1 to ptr), i32 136) #33
  %17 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %call.i580)
  %18 = extractvalue { i1, i64 } %17, 0
  %19 = extractvalue { i1, i64 } %17, 1
  br i1 %18, label %if.end40, label %Flow369

if.end40:                                         ; preds = %_ZN22__llvm_libc_22_0_0_git8internal18CharacterConverter8pop_utf8Ev.exit39
  %sunkaddr327 = getelementptr inbounds i8, ptr %this, i64 16
  %20 = load ptr, ptr %sunkaddr327, align 8, !tbaa !59
  %call.i583 = tail call fastcc noundef zeroext i1 @_ZN22__llvm_libc_22_0_0_git7testing8internal4testIiEEbPNS1_10RunContextENS0_8TestCondET_S6_PKcS8_NS1_8LocationE(ptr noundef %20, i32 noundef 159, i32 noundef 159, ptr noundef addrspacecast (ptr addrspace(4) @.str.10 to ptr), ptr noundef addrspacecast (ptr addrspace(4) @.str.25 to ptr), i32 137) #33
  %21 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %call.i583)
  %22 = extractvalue { i1, i64 } %21, 0
  %23 = extractvalue { i1, i64 } %21, 1
  br i1 %22, label %if.end49, label %Flow368

if.end49:                                         ; preds = %if.end40
  %sunkaddr328 = getelementptr inbounds i8, ptr %this, i64 16
  %24 = load ptr, ptr %sunkaddr328, align 8, !tbaa !59
  %call.i586 = tail call fastcc noundef zeroext i1 @_ZN22__llvm_libc_22_0_0_git7testing8internal4testIbEEbPNS1_10RunContextENS0_8TestCondET_S6_PKcS8_NS1_8LocationE(ptr noundef %24, i1 noundef zeroext true, i1 noundef zeroext true, ptr noundef addrspacecast (ptr addrspace(4) @.str.12 to ptr), ptr noundef addrspacecast (ptr addrspace(4) @.str.1 to ptr), i32 138) #33
  %25 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %call.i586)
  %26 = extractvalue { i1, i64 } %25, 0
  %27 = extractvalue { i1, i64 } %25, 1
  br i1 %26, label %_ZN22__llvm_libc_22_0_0_git8internal18CharacterConverter8pop_utf8Ev.exit69, label %Flow367

_ZN22__llvm_libc_22_0_0_git8internal18CharacterConverter8pop_utf8Ev.exit69: ; preds = %if.end49
  %sunkaddr329 = getelementptr inbounds i8, ptr %this, i64 16
  %28 = load ptr, ptr %sunkaddr329, align 8, !tbaa !59
  %call.i591 = tail call fastcc noundef zeroext i1 @_ZN22__llvm_libc_22_0_0_git7testing8internal4testIbEEbPNS1_10RunContextENS0_8TestCondET_S6_PKcS8_NS1_8LocationE(ptr noundef %28, i1 noundef zeroext true, i1 noundef zeroext true, ptr noundef addrspacecast (ptr addrspace(4) @.str.3 to ptr), ptr noundef addrspacecast (ptr addrspace(4) @.str.1 to ptr), i32 140) #33
  %29 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %call.i591)
  %30 = extractvalue { i1, i64 } %29, 0
  %31 = extractvalue { i1, i64 } %29, 1
  br i1 %30, label %if.end69, label %Flow366

if.end69:                                         ; preds = %_ZN22__llvm_libc_22_0_0_git8internal18CharacterConverter8pop_utf8Ev.exit69
  %sunkaddr330 = getelementptr inbounds i8, ptr %this, i64 16
  %32 = load ptr, ptr %sunkaddr330, align 8, !tbaa !59
  %call.i594 = tail call fastcc noundef zeroext i1 @_ZN22__llvm_libc_22_0_0_git7testing8internal4testIiEEbPNS1_10RunContextENS0_8TestCondET_S6_PKcS8_NS1_8LocationE(ptr noundef %32, i32 noundef 164, i32 noundef 164, ptr noundef addrspacecast (ptr addrspace(4) @.str.10 to ptr), ptr noundef addrspacecast (ptr addrspace(4) @.str.26 to ptr), i32 141) #33
  %33 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %call.i594)
  %34 = extractvalue { i1, i64 } %33, 0
  %35 = extractvalue { i1, i64 } %33, 1
  br i1 %34, label %if.end78, label %Flow365

if.end78:                                         ; preds = %if.end69
  %sunkaddr331 = getelementptr inbounds i8, ptr %this, i64 16
  %36 = load ptr, ptr %sunkaddr331, align 8, !tbaa !59
  %call.i597 = tail call fastcc noundef zeroext i1 @_ZN22__llvm_libc_22_0_0_git7testing8internal4testIbEEbPNS1_10RunContextENS0_8TestCondET_S6_PKcS8_NS1_8LocationE(ptr noundef %36, i1 noundef zeroext true, i1 noundef zeroext true, ptr noundef addrspacecast (ptr addrspace(4) @.str.12 to ptr), ptr noundef addrspacecast (ptr addrspace(4) @.str.1 to ptr), i32 142) #33
  %37 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %call.i597)
  %38 = extractvalue { i1, i64 } %37, 0
  %39 = extractvalue { i1, i64 } %37, 1
  br i1 %38, label %_ZN22__llvm_libc_22_0_0_git8internal18CharacterConverter8pop_utf8Ev.exit99, label %Flow364

_ZN22__llvm_libc_22_0_0_git8internal18CharacterConverter8pop_utf8Ev.exit99: ; preds = %if.end78
  %sunkaddr332 = getelementptr inbounds i8, ptr %this, i64 16
  %40 = load ptr, ptr %sunkaddr332, align 8, !tbaa !59
  %call.i602 = tail call fastcc noundef zeroext i1 @_ZN22__llvm_libc_22_0_0_git7testing8internal4testIbEEbPNS1_10RunContextENS0_8TestCondET_S6_PKcS8_NS1_8LocationE(ptr noundef %40, i1 noundef zeroext true, i1 noundef zeroext true, ptr noundef addrspacecast (ptr addrspace(4) @.str.3 to ptr), ptr noundef addrspacecast (ptr addrspace(4) @.str.1 to ptr), i32 144) #33
  %41 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %call.i602)
  %42 = extractvalue { i1, i64 } %41, 0
  %43 = extractvalue { i1, i64 } %41, 1
  br i1 %42, label %if.end98, label %Flow363

if.end98:                                         ; preds = %_ZN22__llvm_libc_22_0_0_git8internal18CharacterConverter8pop_utf8Ev.exit99
  %sunkaddr333 = getelementptr inbounds i8, ptr %this, i64 16
  %44 = load ptr, ptr %sunkaddr333, align 8, !tbaa !59
  %call.i605 = tail call fastcc noundef zeroext i1 @_ZN22__llvm_libc_22_0_0_git7testing8internal4testIiEEbPNS1_10RunContextENS0_8TestCondET_S6_PKcS8_NS1_8LocationE(ptr noundef %44, i32 noundef 161, i32 noundef 161, ptr noundef addrspacecast (ptr addrspace(4) @.str.10 to ptr), ptr noundef addrspacecast (ptr addrspace(4) @.str.27 to ptr), i32 145) #33
  %45 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %call.i605)
  %46 = extractvalue { i1, i64 } %45, 0
  %47 = extractvalue { i1, i64 } %45, 1
  br i1 %46, label %if.end107, label %Flow362

if.end107:                                        ; preds = %if.end98
  %sunkaddr334 = getelementptr inbounds i8, ptr %this, i64 16
  %48 = load ptr, ptr %sunkaddr334, align 8, !tbaa !59
  %call.i608 = tail call fastcc noundef zeroext i1 @_ZN22__llvm_libc_22_0_0_git7testing8internal4testIbEEbPNS1_10RunContextENS0_8TestCondET_S6_PKcS8_NS1_8LocationE(ptr noundef %48, i1 noundef zeroext true, i1 noundef zeroext true, ptr noundef addrspacecast (ptr addrspace(4) @.str.6 to ptr), ptr noundef addrspacecast (ptr addrspace(4) @.str.1 to ptr), i32 146) #33
  %49 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %call.i608)
  %50 = extractvalue { i1, i64 } %49, 0
  %51 = extractvalue { i1, i64 } %49, 1
  br i1 %50, label %_ZN22__llvm_libc_22_0_0_git8internal18CharacterConverter4pushEDi.exit107, label %Flow361

_ZN22__llvm_libc_22_0_0_git8internal18CharacterConverter4pushEDi.exit107: ; preds = %if.end107
  %sunkaddr335 = getelementptr inbounds i8, ptr %this, i64 16
  %52 = load ptr, ptr %sunkaddr335, align 8, !tbaa !59
  %call.i611 = tail call fastcc noundef zeroext i1 @_ZN22__llvm_libc_22_0_0_git7testing8internal4testIbEEbPNS1_10RunContextENS0_8TestCondET_S6_PKcS8_NS1_8LocationE(ptr noundef %52, i1 noundef zeroext true, i1 noundef zeroext true, ptr noundef addrspacecast (ptr addrspace(4) @.str to ptr), ptr noundef addrspacecast (ptr addrspace(4) @.str.1 to ptr), i32 151) #33
  %53 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %call.i611)
  %54 = extractvalue { i1, i64 } %53, 0
  %55 = extractvalue { i1, i64 } %53, 1
  br i1 %54, label %_ZN22__llvm_libc_22_0_0_git8internal18CharacterConverter8pop_utf8Ev.exit140, label %Flow360

_ZN22__llvm_libc_22_0_0_git8internal18CharacterConverter8pop_utf8Ev.exit140: ; preds = %_ZN22__llvm_libc_22_0_0_git8internal18CharacterConverter4pushEDi.exit107
  %sunkaddr336 = getelementptr inbounds i8, ptr %this, i64 16
  %56 = load ptr, ptr %sunkaddr336, align 8, !tbaa !59
  %call.i616 = tail call fastcc noundef zeroext i1 @_ZN22__llvm_libc_22_0_0_git7testing8internal4testIbEEbPNS1_10RunContextENS0_8TestCondET_S6_PKcS8_NS1_8LocationE(ptr noundef %56, i1 noundef zeroext true, i1 noundef zeroext true, ptr noundef addrspacecast (ptr addrspace(4) @.str.3 to ptr), ptr noundef addrspacecast (ptr addrspace(4) @.str.1 to ptr), i32 153) #33
  %57 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %call.i616)
  %58 = extractvalue { i1, i64 } %57, 0
  %59 = extractvalue { i1, i64 } %57, 1
  br i1 %58, label %if.end135, label %Flow359

if.end135:                                        ; preds = %_ZN22__llvm_libc_22_0_0_git8internal18CharacterConverter8pop_utf8Ev.exit140
  %sunkaddr337 = getelementptr inbounds i8, ptr %this, i64 16
  %60 = load ptr, ptr %sunkaddr337, align 8, !tbaa !59
  %call.i619 = tail call fastcc noundef zeroext i1 @_ZN22__llvm_libc_22_0_0_git7testing8internal4testIiEEbPNS1_10RunContextENS0_8TestCondET_S6_PKcS8_NS1_8LocationE(ptr noundef %60, i32 noundef poison, i32 noundef 240, ptr noundef addrspacecast (ptr addrspace(4) @.str.10 to ptr), ptr noundef addrspacecast (ptr addrspace(4) @.str.24 to ptr), i32 154) #33
  %61 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %call.i619)
  %62 = extractvalue { i1, i64 } %61, 0
  %63 = extractvalue { i1, i64 } %61, 1
  br i1 %62, label %if.end144, label %Flow358

if.end144:                                        ; preds = %if.end135
  %sunkaddr338 = getelementptr inbounds i8, ptr %this, i64 16
  %64 = load ptr, ptr %sunkaddr338, align 8, !tbaa !59
  %call.i622 = tail call fastcc noundef zeroext i1 @_ZN22__llvm_libc_22_0_0_git7testing8internal4testIbEEbPNS1_10RunContextENS0_8TestCondET_S6_PKcS8_NS1_8LocationE(ptr noundef %64, i1 noundef zeroext true, i1 noundef zeroext true, ptr noundef addrspacecast (ptr addrspace(4) @.str.12 to ptr), ptr noundef addrspacecast (ptr addrspace(4) @.str.1 to ptr), i32 155) #33
  %65 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %call.i622)
  %66 = extractvalue { i1, i64 } %65, 0
  %67 = extractvalue { i1, i64 } %65, 1
  br i1 %66, label %_ZN22__llvm_libc_22_0_0_git8internal18CharacterConverter8pop_utf8Ev.exit170, label %Flow357

_ZN22__llvm_libc_22_0_0_git8internal18CharacterConverter8pop_utf8Ev.exit170: ; preds = %if.end144
  %sunkaddr339 = getelementptr inbounds i8, ptr %this, i64 16
  %68 = load ptr, ptr %sunkaddr339, align 8, !tbaa !59
  %call.i627 = tail call fastcc noundef zeroext i1 @_ZN22__llvm_libc_22_0_0_git7testing8internal4testIbEEbPNS1_10RunContextENS0_8TestCondET_S6_PKcS8_NS1_8LocationE(ptr noundef %68, i1 noundef zeroext true, i1 noundef zeroext true, ptr noundef addrspacecast (ptr addrspace(4) @.str.3 to ptr), ptr noundef addrspacecast (ptr addrspace(4) @.str.1 to ptr), i32 157) #33
  %69 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %call.i627)
  %70 = extractvalue { i1, i64 } %69, 0
  %71 = extractvalue { i1, i64 } %69, 1
  br i1 %70, label %if.end164, label %Flow356

if.end164:                                        ; preds = %_ZN22__llvm_libc_22_0_0_git8internal18CharacterConverter8pop_utf8Ev.exit170
  %sunkaddr340 = getelementptr inbounds i8, ptr %this, i64 16
  %72 = load ptr, ptr %sunkaddr340, align 8, !tbaa !59
  %call.i630 = tail call fastcc noundef zeroext i1 @_ZN22__llvm_libc_22_0_0_git7testing8internal4testIiEEbPNS1_10RunContextENS0_8TestCondET_S6_PKcS8_NS1_8LocationE(ptr noundef %72, i32 noundef 146, i32 noundef 146, ptr noundef addrspacecast (ptr addrspace(4) @.str.10 to ptr), ptr noundef addrspacecast (ptr addrspace(4) @.str.28 to ptr), i32 158) #33
  %73 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %call.i630)
  %74 = extractvalue { i1, i64 } %73, 0
  %75 = extractvalue { i1, i64 } %73, 1
  br i1 %74, label %if.end173, label %Flow355

if.end173:                                        ; preds = %if.end164
  %sunkaddr341 = getelementptr inbounds i8, ptr %this, i64 16
  %76 = load ptr, ptr %sunkaddr341, align 8, !tbaa !59
  %call.i633 = tail call fastcc noundef zeroext i1 @_ZN22__llvm_libc_22_0_0_git7testing8internal4testIbEEbPNS1_10RunContextENS0_8TestCondET_S6_PKcS8_NS1_8LocationE(ptr noundef %76, i1 noundef zeroext true, i1 noundef zeroext true, ptr noundef addrspacecast (ptr addrspace(4) @.str.12 to ptr), ptr noundef addrspacecast (ptr addrspace(4) @.str.1 to ptr), i32 159) #33
  %77 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %call.i633)
  %78 = extractvalue { i1, i64 } %77, 0
  %79 = extractvalue { i1, i64 } %77, 1
  br i1 %78, label %_ZN22__llvm_libc_22_0_0_git8internal18CharacterConverter8pop_utf8Ev.exit200, label %Flow354

_ZN22__llvm_libc_22_0_0_git8internal18CharacterConverter8pop_utf8Ev.exit200: ; preds = %if.end173
  %sunkaddr342 = getelementptr inbounds i8, ptr %this, i64 16
  %80 = load ptr, ptr %sunkaddr342, align 8, !tbaa !59
  %call.i638 = tail call fastcc noundef zeroext i1 @_ZN22__llvm_libc_22_0_0_git7testing8internal4testIbEEbPNS1_10RunContextENS0_8TestCondET_S6_PKcS8_NS1_8LocationE(ptr noundef %80, i1 noundef zeroext true, i1 noundef zeroext true, ptr noundef addrspacecast (ptr addrspace(4) @.str.3 to ptr), ptr noundef addrspacecast (ptr addrspace(4) @.str.1 to ptr), i32 161) #33
  %81 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %call.i638)
  %82 = extractvalue { i1, i64 } %81, 0
  %83 = extractvalue { i1, i64 } %81, 1
  br i1 %82, label %if.end193, label %Flow353

if.end193:                                        ; preds = %_ZN22__llvm_libc_22_0_0_git8internal18CharacterConverter8pop_utf8Ev.exit200
  %sunkaddr343 = getelementptr inbounds i8, ptr %this, i64 16
  %84 = load ptr, ptr %sunkaddr343, align 8, !tbaa !59
  %call.i641 = tail call fastcc noundef zeroext i1 @_ZN22__llvm_libc_22_0_0_git7testing8internal4testIiEEbPNS1_10RunContextENS0_8TestCondET_S6_PKcS8_NS1_8LocationE(ptr noundef %84, i32 noundef 132, i32 noundef 132, ptr noundef addrspacecast (ptr addrspace(4) @.str.10 to ptr), ptr noundef addrspacecast (ptr addrspace(4) @.str.29 to ptr), i32 162) #33
  %85 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %call.i641)
  %86 = extractvalue { i1, i64 } %85, 0
  %87 = extractvalue { i1, i64 } %85, 1
  br i1 %86, label %if.end202, label %Flow352

if.end202:                                        ; preds = %if.end193
  %sunkaddr344 = getelementptr inbounds i8, ptr %this, i64 16
  %88 = load ptr, ptr %sunkaddr344, align 8, !tbaa !59
  %call.i644 = tail call fastcc noundef zeroext i1 @_ZN22__llvm_libc_22_0_0_git7testing8internal4testIbEEbPNS1_10RunContextENS0_8TestCondET_S6_PKcS8_NS1_8LocationE(ptr noundef %88, i1 noundef zeroext true, i1 noundef zeroext true, ptr noundef addrspacecast (ptr addrspace(4) @.str.12 to ptr), ptr noundef addrspacecast (ptr addrspace(4) @.str.1 to ptr), i32 163) #33
  %89 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %call.i644)
  %90 = extractvalue { i1, i64 } %89, 0
  %91 = extractvalue { i1, i64 } %89, 1
  br i1 %90, label %_ZN22__llvm_libc_22_0_0_git8internal18CharacterConverter8pop_utf8Ev.exit230, label %Flow351

_ZN22__llvm_libc_22_0_0_git8internal18CharacterConverter8pop_utf8Ev.exit230: ; preds = %if.end202
  %sunkaddr345 = getelementptr inbounds i8, ptr %this, i64 16
  %92 = load ptr, ptr %sunkaddr345, align 8, !tbaa !59
  %call.i649 = tail call fastcc noundef zeroext i1 @_ZN22__llvm_libc_22_0_0_git7testing8internal4testIbEEbPNS1_10RunContextENS0_8TestCondET_S6_PKcS8_NS1_8LocationE(ptr noundef %92, i1 noundef zeroext true, i1 noundef zeroext true, ptr noundef addrspacecast (ptr addrspace(4) @.str.3 to ptr), ptr noundef addrspacecast (ptr addrspace(4) @.str.1 to ptr), i32 165) #33
  %93 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %call.i649)
  %94 = extractvalue { i1, i64 } %93, 0
  %95 = extractvalue { i1, i64 } %93, 1
  br i1 %94, label %if.end222, label %Flow350

if.end222:                                        ; preds = %_ZN22__llvm_libc_22_0_0_git8internal18CharacterConverter8pop_utf8Ev.exit230
  %sunkaddr346 = getelementptr inbounds i8, ptr %this, i64 16
  %96 = load ptr, ptr %sunkaddr346, align 8, !tbaa !59
  %call.i652 = tail call fastcc noundef zeroext i1 @_ZN22__llvm_libc_22_0_0_git7testing8internal4testIiEEbPNS1_10RunContextENS0_8TestCondET_S6_PKcS8_NS1_8LocationE(ptr noundef %96, i32 noundef 161, i32 noundef 161, ptr noundef addrspacecast (ptr addrspace(4) @.str.10 to ptr), ptr noundef addrspacecast (ptr addrspace(4) @.str.27 to ptr), i32 166) #33
  %97 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %call.i652)
  %98 = extractvalue { i1, i64 } %97, 0
  %99 = extractvalue { i1, i64 } %97, 1
  br i1 %98, label %if.end231, label %Flow349

if.end231:                                        ; preds = %if.end222
  %sunkaddr347 = getelementptr inbounds i8, ptr %this, i64 16
  %100 = load ptr, ptr %sunkaddr347, align 8, !tbaa !59
  %call.i655 = tail call fastcc noundef zeroext i1 @_ZN22__llvm_libc_22_0_0_git7testing8internal4testIbEEbPNS1_10RunContextENS0_8TestCondET_S6_PKcS8_NS1_8LocationE(ptr noundef %100, i1 noundef zeroext true, i1 noundef zeroext true, ptr noundef addrspacecast (ptr addrspace(4) @.str.6 to ptr), ptr noundef addrspacecast (ptr addrspace(4) @.str.1 to ptr), i32 167) #33
  %101 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %call.i655)
  %102 = extractvalue { i1, i64 } %101, 0
  %103 = extractvalue { i1, i64 } %101, 1
  br i1 %102, label %_ZN22__llvm_libc_22_0_0_git8internal18CharacterConverter8pop_utf8Ev.exit260, label %Flow

_ZN22__llvm_libc_22_0_0_git8internal18CharacterConverter8pop_utf8Ev.exit260: ; preds = %if.end231
  %sunkaddr348 = getelementptr inbounds i8, ptr %this, i64 16
  %104 = load ptr, ptr %sunkaddr348, align 8, !tbaa !59
  %105 = tail call fastcc noundef zeroext i1 @_ZN22__llvm_libc_22_0_0_git7testing8internal4testIbEEbPNS1_10RunContextENS0_8TestCondET_S6_PKcS8_NS1_8LocationE(ptr noundef %104, i1 noundef zeroext false, i1 noundef zeroext false, ptr noundef addrspacecast (ptr addrspace(4) @.str.3 to ptr), ptr noundef addrspacecast (ptr addrspace(4) @.str.8 to ptr), i32 171) #33
  br label %Flow, !amdgpu.uniform !42

Flow:                                             ; preds = %_ZN22__llvm_libc_22_0_0_git8internal18CharacterConverter8pop_utf8Ev.exit260, %if.end231
  call void @llvm.amdgcn.end.cf.i64(i64 %103)
  br label %Flow349, !amdgpu.uniform !42

Flow349:                                          ; preds = %Flow, %if.end222
  call void @llvm.amdgcn.end.cf.i64(i64 %99)
  br label %Flow350, !amdgpu.uniform !42

Flow350:                                          ; preds = %Flow349, %_ZN22__llvm_libc_22_0_0_git8internal18CharacterConverter8pop_utf8Ev.exit230
  call void @llvm.amdgcn.end.cf.i64(i64 %95)
  br label %Flow351, !amdgpu.uniform !42

Flow351:                                          ; preds = %Flow350, %if.end202
  call void @llvm.amdgcn.end.cf.i64(i64 %91)
  br label %Flow352, !amdgpu.uniform !42

Flow352:                                          ; preds = %Flow351, %if.end193
  call void @llvm.amdgcn.end.cf.i64(i64 %87)
  br label %Flow353, !amdgpu.uniform !42

Flow353:                                          ; preds = %Flow352, %_ZN22__llvm_libc_22_0_0_git8internal18CharacterConverter8pop_utf8Ev.exit200
  call void @llvm.amdgcn.end.cf.i64(i64 %83)
  br label %Flow354, !amdgpu.uniform !42

Flow354:                                          ; preds = %Flow353, %if.end173
  call void @llvm.amdgcn.end.cf.i64(i64 %79)
  br label %Flow355, !amdgpu.uniform !42

Flow355:                                          ; preds = %Flow354, %if.end164
  call void @llvm.amdgcn.end.cf.i64(i64 %75)
  br label %Flow356, !amdgpu.uniform !42

Flow356:                                          ; preds = %Flow355, %_ZN22__llvm_libc_22_0_0_git8internal18CharacterConverter8pop_utf8Ev.exit170
  call void @llvm.amdgcn.end.cf.i64(i64 %71)
  br label %Flow357, !amdgpu.uniform !42

Flow357:                                          ; preds = %Flow356, %if.end144
  call void @llvm.amdgcn.end.cf.i64(i64 %67)
  br label %Flow358, !amdgpu.uniform !42

Flow358:                                          ; preds = %Flow357, %if.end135
  call void @llvm.amdgcn.end.cf.i64(i64 %63)
  br label %Flow359, !amdgpu.uniform !42

Flow359:                                          ; preds = %Flow358, %_ZN22__llvm_libc_22_0_0_git8internal18CharacterConverter8pop_utf8Ev.exit140
  call void @llvm.amdgcn.end.cf.i64(i64 %59)
  br label %Flow360, !amdgpu.uniform !42

Flow360:                                          ; preds = %Flow359, %_ZN22__llvm_libc_22_0_0_git8internal18CharacterConverter4pushEDi.exit107
  call void @llvm.amdgcn.end.cf.i64(i64 %55)
  br label %Flow361, !amdgpu.uniform !42

Flow361:                                          ; preds = %Flow360, %if.end107
  call void @llvm.amdgcn.end.cf.i64(i64 %51)
  br label %Flow362, !amdgpu.uniform !42

Flow362:                                          ; preds = %Flow361, %if.end98
  call void @llvm.amdgcn.end.cf.i64(i64 %47)
  br label %Flow363, !amdgpu.uniform !42

Flow363:                                          ; preds = %Flow362, %_ZN22__llvm_libc_22_0_0_git8internal18CharacterConverter8pop_utf8Ev.exit99
  call void @llvm.amdgcn.end.cf.i64(i64 %43)
  br label %Flow364, !amdgpu.uniform !42

Flow364:                                          ; preds = %Flow363, %if.end78
  call void @llvm.amdgcn.end.cf.i64(i64 %39)
  br label %Flow365, !amdgpu.uniform !42

Flow365:                                          ; preds = %Flow364, %if.end69
  call void @llvm.amdgcn.end.cf.i64(i64 %35)
  br label %Flow366, !amdgpu.uniform !42

Flow366:                                          ; preds = %Flow365, %_ZN22__llvm_libc_22_0_0_git8internal18CharacterConverter8pop_utf8Ev.exit69
  call void @llvm.amdgcn.end.cf.i64(i64 %31)
  br label %Flow367, !amdgpu.uniform !42

Flow367:                                          ; preds = %Flow366, %if.end49
  call void @llvm.amdgcn.end.cf.i64(i64 %27)
  br label %Flow368, !amdgpu.uniform !42

Flow368:                                          ; preds = %Flow367, %if.end40
  call void @llvm.amdgcn.end.cf.i64(i64 %23)
  br label %Flow369, !amdgpu.uniform !42

Flow369:                                          ; preds = %Flow368, %_ZN22__llvm_libc_22_0_0_git8internal18CharacterConverter8pop_utf8Ev.exit39
  call void @llvm.amdgcn.end.cf.i64(i64 %19)
  br label %Flow370, !amdgpu.uniform !42

Flow370:                                          ; preds = %Flow369, %if.end21
  call void @llvm.amdgcn.end.cf.i64(i64 %15)
  br label %Flow371, !amdgpu.uniform !42

Flow371:                                          ; preds = %Flow370, %if.end13
  call void @llvm.amdgcn.end.cf.i64(i64 %11)
  br label %Flow372, !amdgpu.uniform !42

Flow372:                                          ; preds = %Flow371, %_ZN22__llvm_libc_22_0_0_git8internal18CharacterConverter8pop_utf8Ev.exit
  call void @llvm.amdgcn.end.cf.i64(i64 %7)
  br label %UnifiedReturnBlock, !amdgpu.uniform !42

UnifiedReturnBlock:                               ; preds = %entry, %Flow372
  call void @llvm.amdgcn.end.cf.i64(i64 %3)
  ret void
}

; Function Attrs: mustprogress nofree norecurse nosync nounwind willreturn memory(none)
define internal noundef ptr @_ZNK47LlvmLibcCharacterConverterUTF32To8Test_FourByte7getNameEv(ptr nonnull readnone align 8 captures(none) %this) unnamed_addr #6 comdat align 2 {
entry:
  ret ptr addrspacecast (ptr addrspace(4) @.str.50 to ptr)
}

; Function Attrs: cold inlinehint mustprogress noreturn nounwind memory(inaccessiblemem: write)
define internal void @_ZN48LlvmLibcCharacterConverterUTF32To8Test_ThreeByteD0Ev(ptr noundef nonnull readnone align 8 captures(none) dereferenceable(24) %this) unnamed_addr #7 comdat align 2 !section_prefix !58 {
entry:
  tail call void @llvm.trap()
  unreachable
}

; Function Attrs: convergent mustprogress nofree norecurse nounwind
define internal void @_ZN48LlvmLibcCharacterConverterUTF32To8Test_ThreeByte3RunEv(ptr noundef nonnull readonly align 8 captures(none) dereferenceable(24) %this) unnamed_addr #8 align 2 {
entry:
  %Ctx.i = getelementptr inbounds nuw i8, ptr %this, i64 16
  %0 = load ptr, ptr %Ctx.i, align 8, !tbaa !59
  %call.i = tail call fastcc noundef zeroext i1 @_ZN22__llvm_libc_22_0_0_git7testing8internal4testIbEEbPNS1_10RunContextENS0_8TestCondET_S6_PKcS8_NS1_8LocationE(ptr noundef %0, i1 noundef zeroext true, i1 noundef zeroext true, ptr noundef addrspacecast (ptr addrspace(4) @.str to ptr), ptr noundef addrspacecast (ptr addrspace(4) @.str.1 to ptr), i32 86) #33
  %1 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %call.i)
  %2 = extractvalue { i1, i64 } %1, 0
  %3 = extractvalue { i1, i64 } %1, 1
  br i1 %2, label %_ZN22__llvm_libc_22_0_0_git8internal18CharacterConverter8pop_utf8Ev.exit, label %UnifiedReturnBlock

_ZN22__llvm_libc_22_0_0_git8internal18CharacterConverter8pop_utf8Ev.exit: ; preds = %entry
  %sunkaddr = getelementptr inbounds i8, ptr %this, i64 16
  %4 = load ptr, ptr %sunkaddr, align 8, !tbaa !59
  %call.i439 = tail call fastcc noundef zeroext i1 @_ZN22__llvm_libc_22_0_0_git7testing8internal4testIbEEbPNS1_10RunContextENS0_8TestCondET_S6_PKcS8_NS1_8LocationE(ptr noundef %4, i1 noundef zeroext true, i1 noundef zeroext true, ptr noundef addrspacecast (ptr addrspace(4) @.str.3 to ptr), ptr noundef addrspacecast (ptr addrspace(4) @.str.1 to ptr), i32 88) #33
  %5 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %call.i439)
  %6 = extractvalue { i1, i64 } %5, 0
  %7 = extractvalue { i1, i64 } %5, 1
  br i1 %6, label %if.end13, label %Flow286

if.end13:                                         ; preds = %_ZN22__llvm_libc_22_0_0_git8internal18CharacterConverter8pop_utf8Ev.exit
  %sunkaddr250 = getelementptr inbounds i8, ptr %this, i64 16
  %8 = load ptr, ptr %sunkaddr250, align 8, !tbaa !59
  %call.i442 = tail call fastcc noundef zeroext i1 @_ZN22__llvm_libc_22_0_0_git7testing8internal4testIiEEbPNS1_10RunContextENS0_8TestCondET_S6_PKcS8_NS1_8LocationE(ptr noundef %8, i32 noundef poison, i32 noundef 234, ptr noundef addrspacecast (ptr addrspace(4) @.str.10 to ptr), ptr noundef addrspacecast (ptr addrspace(4) @.str.17 to ptr), i32 89) #33
  %9 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %call.i442)
  %10 = extractvalue { i1, i64 } %9, 0
  %11 = extractvalue { i1, i64 } %9, 1
  br i1 %10, label %if.end21, label %Flow285

if.end21:                                         ; preds = %if.end13
  %sunkaddr251 = getelementptr inbounds i8, ptr %this, i64 16
  %12 = load ptr, ptr %sunkaddr251, align 8, !tbaa !59
  %call.i445 = tail call fastcc noundef zeroext i1 @_ZN22__llvm_libc_22_0_0_git7testing8internal4testIbEEbPNS1_10RunContextENS0_8TestCondET_S6_PKcS8_NS1_8LocationE(ptr noundef %12, i1 noundef zeroext true, i1 noundef zeroext true, ptr noundef addrspacecast (ptr addrspace(4) @.str.12 to ptr), ptr noundef addrspacecast (ptr addrspace(4) @.str.1 to ptr), i32 90) #33
  %13 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %call.i445)
  %14 = extractvalue { i1, i64 } %13, 0
  %15 = extractvalue { i1, i64 } %13, 1
  br i1 %14, label %_ZN22__llvm_libc_22_0_0_git8internal18CharacterConverter8pop_utf8Ev.exit39, label %Flow284

_ZN22__llvm_libc_22_0_0_git8internal18CharacterConverter8pop_utf8Ev.exit39: ; preds = %if.end21
  %sunkaddr252 = getelementptr inbounds i8, ptr %this, i64 16
  %16 = load ptr, ptr %sunkaddr252, align 8, !tbaa !59
  %call.i450 = tail call fastcc noundef zeroext i1 @_ZN22__llvm_libc_22_0_0_git7testing8internal4testIbEEbPNS1_10RunContextENS0_8TestCondET_S6_PKcS8_NS1_8LocationE(ptr noundef %16, i1 noundef zeroext true, i1 noundef zeroext true, ptr noundef addrspacecast (ptr addrspace(4) @.str.3 to ptr), ptr noundef addrspacecast (ptr addrspace(4) @.str.1 to ptr), i32 92) #33
  %17 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %call.i450)
  %18 = extractvalue { i1, i64 } %17, 0
  %19 = extractvalue { i1, i64 } %17, 1
  br i1 %18, label %if.end40, label %Flow283

if.end40:                                         ; preds = %_ZN22__llvm_libc_22_0_0_git8internal18CharacterConverter8pop_utf8Ev.exit39
  %sunkaddr253 = getelementptr inbounds i8, ptr %this, i64 16
  %20 = load ptr, ptr %sunkaddr253, align 8, !tbaa !59
  %call.i453 = tail call fastcc noundef zeroext i1 @_ZN22__llvm_libc_22_0_0_git7testing8internal4testIiEEbPNS1_10RunContextENS0_8TestCondET_S6_PKcS8_NS1_8LocationE(ptr noundef %20, i32 noundef 176, i32 noundef 176, ptr noundef addrspacecast (ptr addrspace(4) @.str.10 to ptr), ptr noundef addrspacecast (ptr addrspace(4) @.str.18 to ptr), i32 93) #33
  %21 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %call.i453)
  %22 = extractvalue { i1, i64 } %21, 0
  %23 = extractvalue { i1, i64 } %21, 1
  br i1 %22, label %if.end49, label %Flow282

if.end49:                                         ; preds = %if.end40
  %sunkaddr254 = getelementptr inbounds i8, ptr %this, i64 16
  %24 = load ptr, ptr %sunkaddr254, align 8, !tbaa !59
  %call.i456 = tail call fastcc noundef zeroext i1 @_ZN22__llvm_libc_22_0_0_git7testing8internal4testIbEEbPNS1_10RunContextENS0_8TestCondET_S6_PKcS8_NS1_8LocationE(ptr noundef %24, i1 noundef zeroext true, i1 noundef zeroext true, ptr noundef addrspacecast (ptr addrspace(4) @.str.12 to ptr), ptr noundef addrspacecast (ptr addrspace(4) @.str.1 to ptr), i32 94) #33
  %25 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %call.i456)
  %26 = extractvalue { i1, i64 } %25, 0
  %27 = extractvalue { i1, i64 } %25, 1
  br i1 %26, label %_ZN22__llvm_libc_22_0_0_git8internal18CharacterConverter8pop_utf8Ev.exit69, label %Flow281

_ZN22__llvm_libc_22_0_0_git8internal18CharacterConverter8pop_utf8Ev.exit69: ; preds = %if.end49
  %sunkaddr255 = getelementptr inbounds i8, ptr %this, i64 16
  %28 = load ptr, ptr %sunkaddr255, align 8, !tbaa !59
  %call.i461 = tail call fastcc noundef zeroext i1 @_ZN22__llvm_libc_22_0_0_git7testing8internal4testIbEEbPNS1_10RunContextENS0_8TestCondET_S6_PKcS8_NS1_8LocationE(ptr noundef %28, i1 noundef zeroext true, i1 noundef zeroext true, ptr noundef addrspacecast (ptr addrspace(4) @.str.3 to ptr), ptr noundef addrspacecast (ptr addrspace(4) @.str.1 to ptr), i32 96) #33
  %29 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %call.i461)
  %30 = extractvalue { i1, i64 } %29, 0
  %31 = extractvalue { i1, i64 } %29, 1
  br i1 %30, label %if.end69, label %Flow280

if.end69:                                         ; preds = %_ZN22__llvm_libc_22_0_0_git8internal18CharacterConverter8pop_utf8Ev.exit69
  %sunkaddr256 = getelementptr inbounds i8, ptr %this, i64 16
  %32 = load ptr, ptr %sunkaddr256, align 8, !tbaa !59
  %call.i464 = tail call fastcc noundef zeroext i1 @_ZN22__llvm_libc_22_0_0_git7testing8internal4testIiEEbPNS1_10RunContextENS0_8TestCondET_S6_PKcS8_NS1_8LocationE(ptr noundef %32, i32 noundef 149, i32 noundef 149, ptr noundef addrspacecast (ptr addrspace(4) @.str.10 to ptr), ptr noundef addrspacecast (ptr addrspace(4) @.str.19 to ptr), i32 97) #33
  %33 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %call.i464)
  %34 = extractvalue { i1, i64 } %33, 0
  %35 = extractvalue { i1, i64 } %33, 1
  br i1 %34, label %if.end78, label %Flow279

if.end78:                                         ; preds = %if.end69
  %sunkaddr257 = getelementptr inbounds i8, ptr %this, i64 16
  %36 = load ptr, ptr %sunkaddr257, align 8, !tbaa !59
  %call.i467 = tail call fastcc noundef zeroext i1 @_ZN22__llvm_libc_22_0_0_git7testing8internal4testIbEEbPNS1_10RunContextENS0_8TestCondET_S6_PKcS8_NS1_8LocationE(ptr noundef %36, i1 noundef zeroext true, i1 noundef zeroext true, ptr noundef addrspacecast (ptr addrspace(4) @.str.6 to ptr), ptr noundef addrspacecast (ptr addrspace(4) @.str.1 to ptr), i32 98) #33
  %37 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %call.i467)
  %38 = extractvalue { i1, i64 } %37, 0
  %39 = extractvalue { i1, i64 } %37, 1
  br i1 %38, label %_ZN22__llvm_libc_22_0_0_git8internal18CharacterConverter4pushEDi.exit77, label %Flow278

_ZN22__llvm_libc_22_0_0_git8internal18CharacterConverter4pushEDi.exit77: ; preds = %if.end78
  %sunkaddr258 = getelementptr inbounds i8, ptr %this, i64 16
  %40 = load ptr, ptr %sunkaddr258, align 8, !tbaa !59
  %call.i470 = tail call fastcc noundef zeroext i1 @_ZN22__llvm_libc_22_0_0_git7testing8internal4testIbEEbPNS1_10RunContextENS0_8TestCondET_S6_PKcS8_NS1_8LocationE(ptr noundef %40, i1 noundef zeroext true, i1 noundef zeroext true, ptr noundef addrspacecast (ptr addrspace(4) @.str to ptr), ptr noundef addrspacecast (ptr addrspace(4) @.str.1 to ptr), i32 103) #33
  %41 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %call.i470)
  %42 = extractvalue { i1, i64 } %41, 0
  %43 = extractvalue { i1, i64 } %41, 1
  br i1 %42, label %_ZN22__llvm_libc_22_0_0_git8internal18CharacterConverter8pop_utf8Ev.exit110, label %Flow277

_ZN22__llvm_libc_22_0_0_git8internal18CharacterConverter8pop_utf8Ev.exit110: ; preds = %_ZN22__llvm_libc_22_0_0_git8internal18CharacterConverter4pushEDi.exit77
  %sunkaddr259 = getelementptr inbounds i8, ptr %this, i64 16
  %44 = load ptr, ptr %sunkaddr259, align 8, !tbaa !59
  %call.i475 = tail call fastcc noundef zeroext i1 @_ZN22__llvm_libc_22_0_0_git7testing8internal4testIbEEbPNS1_10RunContextENS0_8TestCondET_S6_PKcS8_NS1_8LocationE(ptr noundef %44, i1 noundef zeroext true, i1 noundef zeroext true, ptr noundef addrspacecast (ptr addrspace(4) @.str.3 to ptr), ptr noundef addrspacecast (ptr addrspace(4) @.str.1 to ptr), i32 105) #33
  %45 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %call.i475)
  %46 = extractvalue { i1, i64 } %45, 0
  %47 = extractvalue { i1, i64 } %45, 1
  br i1 %46, label %if.end106, label %Flow276

if.end106:                                        ; preds = %_ZN22__llvm_libc_22_0_0_git8internal18CharacterConverter8pop_utf8Ev.exit110
  %sunkaddr260 = getelementptr inbounds i8, ptr %this, i64 16
  %48 = load ptr, ptr %sunkaddr260, align 8, !tbaa !59
  %call.i478 = tail call fastcc noundef zeroext i1 @_ZN22__llvm_libc_22_0_0_git7testing8internal4testIiEEbPNS1_10RunContextENS0_8TestCondET_S6_PKcS8_NS1_8LocationE(ptr noundef %48, i32 noundef poison, i32 noundef 226, ptr noundef addrspacecast (ptr addrspace(4) @.str.10 to ptr), ptr noundef addrspacecast (ptr addrspace(4) @.str.20 to ptr), i32 106) #33
  %49 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %call.i478)
  %50 = extractvalue { i1, i64 } %49, 0
  %51 = extractvalue { i1, i64 } %49, 1
  br i1 %50, label %if.end115, label %Flow275

if.end115:                                        ; preds = %if.end106
  %sunkaddr261 = getelementptr inbounds i8, ptr %this, i64 16
  %52 = load ptr, ptr %sunkaddr261, align 8, !tbaa !59
  %call.i481 = tail call fastcc noundef zeroext i1 @_ZN22__llvm_libc_22_0_0_git7testing8internal4testIbEEbPNS1_10RunContextENS0_8TestCondET_S6_PKcS8_NS1_8LocationE(ptr noundef %52, i1 noundef zeroext true, i1 noundef zeroext true, ptr noundef addrspacecast (ptr addrspace(4) @.str.12 to ptr), ptr noundef addrspacecast (ptr addrspace(4) @.str.1 to ptr), i32 107) #33
  %53 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %call.i481)
  %54 = extractvalue { i1, i64 } %53, 0
  %55 = extractvalue { i1, i64 } %53, 1
  br i1 %54, label %_ZN22__llvm_libc_22_0_0_git8internal18CharacterConverter8pop_utf8Ev.exit140, label %Flow274

_ZN22__llvm_libc_22_0_0_git8internal18CharacterConverter8pop_utf8Ev.exit140: ; preds = %if.end115
  %sunkaddr262 = getelementptr inbounds i8, ptr %this, i64 16
  %56 = load ptr, ptr %sunkaddr262, align 8, !tbaa !59
  %call.i486 = tail call fastcc noundef zeroext i1 @_ZN22__llvm_libc_22_0_0_git7testing8internal4testIbEEbPNS1_10RunContextENS0_8TestCondET_S6_PKcS8_NS1_8LocationE(ptr noundef %56, i1 noundef zeroext true, i1 noundef zeroext true, ptr noundef addrspacecast (ptr addrspace(4) @.str.3 to ptr), ptr noundef addrspacecast (ptr addrspace(4) @.str.1 to ptr), i32 109) #33
  %57 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %call.i486)
  %58 = extractvalue { i1, i64 } %57, 0
  %59 = extractvalue { i1, i64 } %57, 1
  br i1 %58, label %if.end135, label %Flow273

if.end135:                                        ; preds = %_ZN22__llvm_libc_22_0_0_git8internal18CharacterConverter8pop_utf8Ev.exit140
  %sunkaddr263 = getelementptr inbounds i8, ptr %this, i64 16
  %60 = load ptr, ptr %sunkaddr263, align 8, !tbaa !59
  %call.i489 = tail call fastcc noundef zeroext i1 @_ZN22__llvm_libc_22_0_0_git7testing8internal4testIiEEbPNS1_10RunContextENS0_8TestCondET_S6_PKcS8_NS1_8LocationE(ptr noundef %60, i32 noundef 153, i32 noundef 153, ptr noundef addrspacecast (ptr addrspace(4) @.str.10 to ptr), ptr noundef addrspacecast (ptr addrspace(4) @.str.21 to ptr), i32 110) #33
  %61 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %call.i489)
  %62 = extractvalue { i1, i64 } %61, 0
  %63 = extractvalue { i1, i64 } %61, 1
  br i1 %62, label %if.end144, label %Flow272

if.end144:                                        ; preds = %if.end135
  %sunkaddr264 = getelementptr inbounds i8, ptr %this, i64 16
  %64 = load ptr, ptr %sunkaddr264, align 8, !tbaa !59
  %call.i492 = tail call fastcc noundef zeroext i1 @_ZN22__llvm_libc_22_0_0_git7testing8internal4testIbEEbPNS1_10RunContextENS0_8TestCondET_S6_PKcS8_NS1_8LocationE(ptr noundef %64, i1 noundef zeroext true, i1 noundef zeroext true, ptr noundef addrspacecast (ptr addrspace(4) @.str.12 to ptr), ptr noundef addrspacecast (ptr addrspace(4) @.str.1 to ptr), i32 111) #33
  %65 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %call.i492)
  %66 = extractvalue { i1, i64 } %65, 0
  %67 = extractvalue { i1, i64 } %65, 1
  br i1 %66, label %_ZN22__llvm_libc_22_0_0_git8internal18CharacterConverter8pop_utf8Ev.exit170, label %Flow271

_ZN22__llvm_libc_22_0_0_git8internal18CharacterConverter8pop_utf8Ev.exit170: ; preds = %if.end144
  %sunkaddr265 = getelementptr inbounds i8, ptr %this, i64 16
  %68 = load ptr, ptr %sunkaddr265, align 8, !tbaa !59
  %call.i497 = tail call fastcc noundef zeroext i1 @_ZN22__llvm_libc_22_0_0_git7testing8internal4testIbEEbPNS1_10RunContextENS0_8TestCondET_S6_PKcS8_NS1_8LocationE(ptr noundef %68, i1 noundef zeroext true, i1 noundef zeroext true, ptr noundef addrspacecast (ptr addrspace(4) @.str.3 to ptr), ptr noundef addrspacecast (ptr addrspace(4) @.str.1 to ptr), i32 113) #33
  %69 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %call.i497)
  %70 = extractvalue { i1, i64 } %69, 0
  %71 = extractvalue { i1, i64 } %69, 1
  br i1 %70, label %if.end164, label %Flow270

if.end164:                                        ; preds = %_ZN22__llvm_libc_22_0_0_git8internal18CharacterConverter8pop_utf8Ev.exit170
  %sunkaddr266 = getelementptr inbounds i8, ptr %this, i64 16
  %72 = load ptr, ptr %sunkaddr266, align 8, !tbaa !59
  %call.i500 = tail call fastcc noundef zeroext i1 @_ZN22__llvm_libc_22_0_0_git7testing8internal4testIiEEbPNS1_10RunContextENS0_8TestCondET_S6_PKcS8_NS1_8LocationE(ptr noundef %72, i32 noundef 187, i32 noundef 187, ptr noundef addrspacecast (ptr addrspace(4) @.str.10 to ptr), ptr noundef addrspacecast (ptr addrspace(4) @.str.22 to ptr), i32 114) #33
  %73 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %call.i500)
  %74 = extractvalue { i1, i64 } %73, 0
  %75 = extractvalue { i1, i64 } %73, 1
  br i1 %74, label %if.end173, label %Flow269

if.end173:                                        ; preds = %if.end164
  %sunkaddr267 = getelementptr inbounds i8, ptr %this, i64 16
  %76 = load ptr, ptr %sunkaddr267, align 8, !tbaa !59
  %call.i503 = tail call fastcc noundef zeroext i1 @_ZN22__llvm_libc_22_0_0_git7testing8internal4testIbEEbPNS1_10RunContextENS0_8TestCondET_S6_PKcS8_NS1_8LocationE(ptr noundef %76, i1 noundef zeroext true, i1 noundef zeroext true, ptr noundef addrspacecast (ptr addrspace(4) @.str.6 to ptr), ptr noundef addrspacecast (ptr addrspace(4) @.str.1 to ptr), i32 115) #33
  %77 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %call.i503)
  %78 = extractvalue { i1, i64 } %77, 0
  %79 = extractvalue { i1, i64 } %77, 1
  br i1 %78, label %_ZN22__llvm_libc_22_0_0_git8internal18CharacterConverter8pop_utf8Ev.exit200, label %Flow

_ZN22__llvm_libc_22_0_0_git8internal18CharacterConverter8pop_utf8Ev.exit200: ; preds = %if.end173
  %sunkaddr268 = getelementptr inbounds i8, ptr %this, i64 16
  %80 = load ptr, ptr %sunkaddr268, align 8, !tbaa !59
  %81 = tail call fastcc noundef zeroext i1 @_ZN22__llvm_libc_22_0_0_git7testing8internal4testIbEEbPNS1_10RunContextENS0_8TestCondET_S6_PKcS8_NS1_8LocationE(ptr noundef %80, i1 noundef zeroext false, i1 noundef zeroext false, ptr noundef addrspacecast (ptr addrspace(4) @.str.3 to ptr), ptr noundef addrspacecast (ptr addrspace(4) @.str.8 to ptr), i32 119) #33
  br label %Flow, !amdgpu.uniform !42

Flow:                                             ; preds = %_ZN22__llvm_libc_22_0_0_git8internal18CharacterConverter8pop_utf8Ev.exit200, %if.end173
  call void @llvm.amdgcn.end.cf.i64(i64 %79)
  br label %Flow269, !amdgpu.uniform !42

Flow269:                                          ; preds = %Flow, %if.end164
  call void @llvm.amdgcn.end.cf.i64(i64 %75)
  br label %Flow270, !amdgpu.uniform !42

Flow270:                                          ; preds = %Flow269, %_ZN22__llvm_libc_22_0_0_git8internal18CharacterConverter8pop_utf8Ev.exit170
  call void @llvm.amdgcn.end.cf.i64(i64 %71)
  br label %Flow271, !amdgpu.uniform !42

Flow271:                                          ; preds = %Flow270, %if.end144
  call void @llvm.amdgcn.end.cf.i64(i64 %67)
  br label %Flow272, !amdgpu.uniform !42

Flow272:                                          ; preds = %Flow271, %if.end135
  call void @llvm.amdgcn.end.cf.i64(i64 %63)
  br label %Flow273, !amdgpu.uniform !42

Flow273:                                          ; preds = %Flow272, %_ZN22__llvm_libc_22_0_0_git8internal18CharacterConverter8pop_utf8Ev.exit140
  call void @llvm.amdgcn.end.cf.i64(i64 %59)
  br label %Flow274, !amdgpu.uniform !42

Flow274:                                          ; preds = %Flow273, %if.end115
  call void @llvm.amdgcn.end.cf.i64(i64 %55)
  br label %Flow275, !amdgpu.uniform !42

Flow275:                                          ; preds = %Flow274, %if.end106
  call void @llvm.amdgcn.end.cf.i64(i64 %51)
  br label %Flow276, !amdgpu.uniform !42

Flow276:                                          ; preds = %Flow275, %_ZN22__llvm_libc_22_0_0_git8internal18CharacterConverter8pop_utf8Ev.exit110
  call void @llvm.amdgcn.end.cf.i64(i64 %47)
  br label %Flow277, !amdgpu.uniform !42

Flow277:                                          ; preds = %Flow276, %_ZN22__llvm_libc_22_0_0_git8internal18CharacterConverter4pushEDi.exit77
  call void @llvm.amdgcn.end.cf.i64(i64 %43)
  br label %Flow278, !amdgpu.uniform !42

Flow278:                                          ; preds = %Flow277, %if.end78
  call void @llvm.amdgcn.end.cf.i64(i64 %39)
  br label %Flow279, !amdgpu.uniform !42

Flow279:                                          ; preds = %Flow278, %if.end69
  call void @llvm.amdgcn.end.cf.i64(i64 %35)
  br label %Flow280, !amdgpu.uniform !42

Flow280:                                          ; preds = %Flow279, %_ZN22__llvm_libc_22_0_0_git8internal18CharacterConverter8pop_utf8Ev.exit69
  call void @llvm.amdgcn.end.cf.i64(i64 %31)
  br label %Flow281, !amdgpu.uniform !42

Flow281:                                          ; preds = %Flow280, %if.end49
  call void @llvm.amdgcn.end.cf.i64(i64 %27)
  br label %Flow282, !amdgpu.uniform !42

Flow282:                                          ; preds = %Flow281, %if.end40
  call void @llvm.amdgcn.end.cf.i64(i64 %23)
  br label %Flow283, !amdgpu.uniform !42

Flow283:                                          ; preds = %Flow282, %_ZN22__llvm_libc_22_0_0_git8internal18CharacterConverter8pop_utf8Ev.exit39
  call void @llvm.amdgcn.end.cf.i64(i64 %19)
  br label %Flow284, !amdgpu.uniform !42

Flow284:                                          ; preds = %Flow283, %if.end21
  call void @llvm.amdgcn.end.cf.i64(i64 %15)
  br label %Flow285, !amdgpu.uniform !42

Flow285:                                          ; preds = %Flow284, %if.end13
  call void @llvm.amdgcn.end.cf.i64(i64 %11)
  br label %Flow286, !amdgpu.uniform !42

Flow286:                                          ; preds = %Flow285, %_ZN22__llvm_libc_22_0_0_git8internal18CharacterConverter8pop_utf8Ev.exit
  call void @llvm.amdgcn.end.cf.i64(i64 %7)
  br label %UnifiedReturnBlock, !amdgpu.uniform !42

UnifiedReturnBlock:                               ; preds = %entry, %Flow286
  call void @llvm.amdgcn.end.cf.i64(i64 %3)
  ret void
}

; Function Attrs: mustprogress nofree norecurse nosync nounwind willreturn memory(none)
define internal noundef ptr @_ZNK48LlvmLibcCharacterConverterUTF32To8Test_ThreeByte7getNameEv(ptr nonnull readnone align 8 captures(none) %this) unnamed_addr #6 comdat align 2 {
entry:
  ret ptr addrspacecast (ptr addrspace(4) @.str.49 to ptr)
}

; Function Attrs: cold inlinehint mustprogress noreturn nounwind memory(inaccessiblemem: write)
define internal void @_ZN46LlvmLibcCharacterConverterUTF32To8Test_TwoByteD0Ev(ptr noundef nonnull readnone align 8 captures(none) dereferenceable(24) %this) unnamed_addr #7 comdat align 2 !section_prefix !58 {
entry:
  tail call void @llvm.trap()
  unreachable
}

; Function Attrs: convergent mustprogress nofree norecurse nounwind
define internal void @_ZN46LlvmLibcCharacterConverterUTF32To8Test_TwoByte3RunEv(ptr noundef nonnull readonly align 8 captures(none) dereferenceable(24) %this) unnamed_addr #8 align 2 {
entry:
  %Ctx.i = getelementptr inbounds nuw i8, ptr %this, i64 16
  %0 = load ptr, ptr %Ctx.i, align 8, !tbaa !59
  %call.i = tail call fastcc noundef zeroext i1 @_ZN22__llvm_libc_22_0_0_git7testing8internal4testIbEEbPNS1_10RunContextENS0_8TestCondET_S6_PKcS8_NS1_8LocationE(ptr noundef %0, i1 noundef zeroext true, i1 noundef zeroext true, ptr noundef addrspacecast (ptr addrspace(4) @.str to ptr), ptr noundef addrspacecast (ptr addrspace(4) @.str.1 to ptr), i32 50) #33
  %1 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %call.i)
  %2 = extractvalue { i1, i64 } %1, 0
  %3 = extractvalue { i1, i64 } %1, 1
  br i1 %2, label %_ZN22__llvm_libc_22_0_0_git8internal18CharacterConverter8pop_utf8Ev.exit, label %UnifiedReturnBlock

_ZN22__llvm_libc_22_0_0_git8internal18CharacterConverter8pop_utf8Ev.exit: ; preds = %entry
  %sunkaddr = getelementptr inbounds i8, ptr %this, i64 16
  %4 = load ptr, ptr %sunkaddr, align 8, !tbaa !59
  %call.i309 = tail call fastcc noundef zeroext i1 @_ZN22__llvm_libc_22_0_0_git7testing8internal4testIbEEbPNS1_10RunContextENS0_8TestCondET_S6_PKcS8_NS1_8LocationE(ptr noundef %4, i1 noundef zeroext true, i1 noundef zeroext true, ptr noundef addrspacecast (ptr addrspace(4) @.str.3 to ptr), ptr noundef addrspacecast (ptr addrspace(4) @.str.1 to ptr), i32 52) #33
  %5 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %call.i309)
  %6 = extractvalue { i1, i64 } %5, 0
  %7 = extractvalue { i1, i64 } %5, 1
  br i1 %6, label %if.end13, label %Flow200

if.end13:                                         ; preds = %_ZN22__llvm_libc_22_0_0_git8internal18CharacterConverter8pop_utf8Ev.exit
  %sunkaddr176 = getelementptr inbounds i8, ptr %this, i64 16
  %8 = load ptr, ptr %sunkaddr176, align 8, !tbaa !59
  %call.i312 = tail call fastcc noundef zeroext i1 @_ZN22__llvm_libc_22_0_0_git7testing8internal4testIiEEbPNS1_10RunContextENS0_8TestCondET_S6_PKcS8_NS1_8LocationE(ptr noundef %8, i32 noundef poison, i32 noundef 195, ptr noundef addrspacecast (ptr addrspace(4) @.str.10 to ptr), ptr noundef addrspacecast (ptr addrspace(4) @.str.11 to ptr), i32 53) #33
  %9 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %call.i312)
  %10 = extractvalue { i1, i64 } %9, 0
  %11 = extractvalue { i1, i64 } %9, 1
  br i1 %10, label %if.end21, label %Flow199

if.end21:                                         ; preds = %if.end13
  %sunkaddr177 = getelementptr inbounds i8, ptr %this, i64 16
  %12 = load ptr, ptr %sunkaddr177, align 8, !tbaa !59
  %call.i315 = tail call fastcc noundef zeroext i1 @_ZN22__llvm_libc_22_0_0_git7testing8internal4testIbEEbPNS1_10RunContextENS0_8TestCondET_S6_PKcS8_NS1_8LocationE(ptr noundef %12, i1 noundef zeroext true, i1 noundef zeroext true, ptr noundef addrspacecast (ptr addrspace(4) @.str.12 to ptr), ptr noundef addrspacecast (ptr addrspace(4) @.str.1 to ptr), i32 54) #33
  %13 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %call.i315)
  %14 = extractvalue { i1, i64 } %13, 0
  %15 = extractvalue { i1, i64 } %13, 1
  br i1 %14, label %_ZN22__llvm_libc_22_0_0_git8internal18CharacterConverter8pop_utf8Ev.exit39, label %Flow198

_ZN22__llvm_libc_22_0_0_git8internal18CharacterConverter8pop_utf8Ev.exit39: ; preds = %if.end21
  %sunkaddr178 = getelementptr inbounds i8, ptr %this, i64 16
  %16 = load ptr, ptr %sunkaddr178, align 8, !tbaa !59
  %call.i320 = tail call fastcc noundef zeroext i1 @_ZN22__llvm_libc_22_0_0_git7testing8internal4testIbEEbPNS1_10RunContextENS0_8TestCondET_S6_PKcS8_NS1_8LocationE(ptr noundef %16, i1 noundef zeroext true, i1 noundef zeroext true, ptr noundef addrspacecast (ptr addrspace(4) @.str.3 to ptr), ptr noundef addrspacecast (ptr addrspace(4) @.str.1 to ptr), i32 56) #33
  %17 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %call.i320)
  %18 = extractvalue { i1, i64 } %17, 0
  %19 = extractvalue { i1, i64 } %17, 1
  br i1 %18, label %if.end40, label %Flow197

if.end40:                                         ; preds = %_ZN22__llvm_libc_22_0_0_git8internal18CharacterConverter8pop_utf8Ev.exit39
  %sunkaddr179 = getelementptr inbounds i8, ptr %this, i64 16
  %20 = load ptr, ptr %sunkaddr179, align 8, !tbaa !59
  %call.i323 = tail call fastcc noundef zeroext i1 @_ZN22__llvm_libc_22_0_0_git7testing8internal4testIiEEbPNS1_10RunContextENS0_8TestCondET_S6_PKcS8_NS1_8LocationE(ptr noundef %20, i32 noundef 191, i32 noundef 191, ptr noundef addrspacecast (ptr addrspace(4) @.str.10 to ptr), ptr noundef addrspacecast (ptr addrspace(4) @.str.13 to ptr), i32 57) #33
  %21 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %call.i323)
  %22 = extractvalue { i1, i64 } %21, 0
  %23 = extractvalue { i1, i64 } %21, 1
  br i1 %22, label %if.end49, label %Flow196

if.end49:                                         ; preds = %if.end40
  %sunkaddr180 = getelementptr inbounds i8, ptr %this, i64 16
  %24 = load ptr, ptr %sunkaddr180, align 8, !tbaa !59
  %call.i326 = tail call fastcc noundef zeroext i1 @_ZN22__llvm_libc_22_0_0_git7testing8internal4testIbEEbPNS1_10RunContextENS0_8TestCondET_S6_PKcS8_NS1_8LocationE(ptr noundef %24, i1 noundef zeroext true, i1 noundef zeroext true, ptr noundef addrspacecast (ptr addrspace(4) @.str.6 to ptr), ptr noundef addrspacecast (ptr addrspace(4) @.str.1 to ptr), i32 58) #33
  %25 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %call.i326)
  %26 = extractvalue { i1, i64 } %25, 0
  %27 = extractvalue { i1, i64 } %25, 1
  br i1 %26, label %_ZN22__llvm_libc_22_0_0_git8internal18CharacterConverter4pushEDi.exit47, label %Flow195

_ZN22__llvm_libc_22_0_0_git8internal18CharacterConverter4pushEDi.exit47: ; preds = %if.end49
  %sunkaddr181 = getelementptr inbounds i8, ptr %this, i64 16
  %28 = load ptr, ptr %sunkaddr181, align 8, !tbaa !59
  %call.i329 = tail call fastcc noundef zeroext i1 @_ZN22__llvm_libc_22_0_0_git7testing8internal4testIbEEbPNS1_10RunContextENS0_8TestCondET_S6_PKcS8_NS1_8LocationE(ptr noundef %28, i1 noundef zeroext true, i1 noundef zeroext true, ptr noundef addrspacecast (ptr addrspace(4) @.str to ptr), ptr noundef addrspacecast (ptr addrspace(4) @.str.1 to ptr), i32 63) #33
  %29 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %call.i329)
  %30 = extractvalue { i1, i64 } %29, 0
  %31 = extractvalue { i1, i64 } %29, 1
  br i1 %30, label %_ZN22__llvm_libc_22_0_0_git8internal18CharacterConverter8pop_utf8Ev.exit80, label %Flow194

_ZN22__llvm_libc_22_0_0_git8internal18CharacterConverter8pop_utf8Ev.exit80: ; preds = %_ZN22__llvm_libc_22_0_0_git8internal18CharacterConverter4pushEDi.exit47
  %sunkaddr182 = getelementptr inbounds i8, ptr %this, i64 16
  %32 = load ptr, ptr %sunkaddr182, align 8, !tbaa !59
  %call.i334 = tail call fastcc noundef zeroext i1 @_ZN22__llvm_libc_22_0_0_git7testing8internal4testIbEEbPNS1_10RunContextENS0_8TestCondET_S6_PKcS8_NS1_8LocationE(ptr noundef %32, i1 noundef zeroext true, i1 noundef zeroext true, ptr noundef addrspacecast (ptr addrspace(4) @.str.3 to ptr), ptr noundef addrspacecast (ptr addrspace(4) @.str.1 to ptr), i32 65) #33
  %33 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %call.i334)
  %34 = extractvalue { i1, i64 } %33, 0
  %35 = extractvalue { i1, i64 } %33, 1
  br i1 %34, label %if.end77, label %Flow193

if.end77:                                         ; preds = %_ZN22__llvm_libc_22_0_0_git8internal18CharacterConverter8pop_utf8Ev.exit80
  %sunkaddr183 = getelementptr inbounds i8, ptr %this, i64 16
  %36 = load ptr, ptr %sunkaddr183, align 8, !tbaa !59
  %call.i337 = tail call fastcc noundef zeroext i1 @_ZN22__llvm_libc_22_0_0_git7testing8internal4testIiEEbPNS1_10RunContextENS0_8TestCondET_S6_PKcS8_NS1_8LocationE(ptr noundef %36, i32 noundef poison, i32 noundef 214, ptr noundef addrspacecast (ptr addrspace(4) @.str.10 to ptr), ptr noundef addrspacecast (ptr addrspace(4) @.str.14 to ptr), i32 66) #33
  %37 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %call.i337)
  %38 = extractvalue { i1, i64 } %37, 0
  %39 = extractvalue { i1, i64 } %37, 1
  br i1 %38, label %if.end86, label %Flow192

if.end86:                                         ; preds = %if.end77
  %sunkaddr184 = getelementptr inbounds i8, ptr %this, i64 16
  %40 = load ptr, ptr %sunkaddr184, align 8, !tbaa !59
  %call.i340 = tail call fastcc noundef zeroext i1 @_ZN22__llvm_libc_22_0_0_git7testing8internal4testIbEEbPNS1_10RunContextENS0_8TestCondET_S6_PKcS8_NS1_8LocationE(ptr noundef %40, i1 noundef zeroext true, i1 noundef zeroext true, ptr noundef addrspacecast (ptr addrspace(4) @.str.12 to ptr), ptr noundef addrspacecast (ptr addrspace(4) @.str.1 to ptr), i32 67) #33
  %41 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %call.i340)
  %42 = extractvalue { i1, i64 } %41, 0
  %43 = extractvalue { i1, i64 } %41, 1
  br i1 %42, label %_ZN22__llvm_libc_22_0_0_git8internal18CharacterConverter8pop_utf8Ev.exit110, label %Flow191

_ZN22__llvm_libc_22_0_0_git8internal18CharacterConverter8pop_utf8Ev.exit110: ; preds = %if.end86
  %sunkaddr185 = getelementptr inbounds i8, ptr %this, i64 16
  %44 = load ptr, ptr %sunkaddr185, align 8, !tbaa !59
  %call.i345 = tail call fastcc noundef zeroext i1 @_ZN22__llvm_libc_22_0_0_git7testing8internal4testIbEEbPNS1_10RunContextENS0_8TestCondET_S6_PKcS8_NS1_8LocationE(ptr noundef %44, i1 noundef zeroext true, i1 noundef zeroext true, ptr noundef addrspacecast (ptr addrspace(4) @.str.3 to ptr), ptr noundef addrspacecast (ptr addrspace(4) @.str.1 to ptr), i32 69) #33
  %45 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %call.i345)
  %46 = extractvalue { i1, i64 } %45, 0
  %47 = extractvalue { i1, i64 } %45, 1
  br i1 %46, label %if.end106, label %Flow190

if.end106:                                        ; preds = %_ZN22__llvm_libc_22_0_0_git8internal18CharacterConverter8pop_utf8Ev.exit110
  %sunkaddr186 = getelementptr inbounds i8, ptr %this, i64 16
  %48 = load ptr, ptr %sunkaddr186, align 8, !tbaa !59
  %call.i348 = tail call fastcc noundef zeroext i1 @_ZN22__llvm_libc_22_0_0_git7testing8internal4testIiEEbPNS1_10RunContextENS0_8TestCondET_S6_PKcS8_NS1_8LocationE(ptr noundef %48, i32 noundef 142, i32 noundef 142, ptr noundef addrspacecast (ptr addrspace(4) @.str.10 to ptr), ptr noundef addrspacecast (ptr addrspace(4) @.str.15 to ptr), i32 70) #33
  %49 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %call.i348)
  %50 = extractvalue { i1, i64 } %49, 0
  %51 = extractvalue { i1, i64 } %49, 1
  br i1 %50, label %if.end115, label %Flow189

if.end115:                                        ; preds = %if.end106
  %sunkaddr187 = getelementptr inbounds i8, ptr %this, i64 16
  %52 = load ptr, ptr %sunkaddr187, align 8, !tbaa !59
  %call.i351 = tail call fastcc noundef zeroext i1 @_ZN22__llvm_libc_22_0_0_git7testing8internal4testIbEEbPNS1_10RunContextENS0_8TestCondET_S6_PKcS8_NS1_8LocationE(ptr noundef %52, i1 noundef zeroext true, i1 noundef zeroext true, ptr noundef addrspacecast (ptr addrspace(4) @.str.6 to ptr), ptr noundef addrspacecast (ptr addrspace(4) @.str.1 to ptr), i32 71) #33
  %53 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %call.i351)
  %54 = extractvalue { i1, i64 } %53, 0
  %55 = extractvalue { i1, i64 } %53, 1
  br i1 %54, label %_ZN22__llvm_libc_22_0_0_git8internal18CharacterConverter8pop_utf8Ev.exit140, label %Flow

_ZN22__llvm_libc_22_0_0_git8internal18CharacterConverter8pop_utf8Ev.exit140: ; preds = %if.end115
  %sunkaddr188 = getelementptr inbounds i8, ptr %this, i64 16
  %56 = load ptr, ptr %sunkaddr188, align 8, !tbaa !59
  %57 = tail call fastcc noundef zeroext i1 @_ZN22__llvm_libc_22_0_0_git7testing8internal4testIbEEbPNS1_10RunContextENS0_8TestCondET_S6_PKcS8_NS1_8LocationE(ptr noundef %56, i1 noundef zeroext false, i1 noundef zeroext false, ptr noundef addrspacecast (ptr addrspace(4) @.str.3 to ptr), ptr noundef addrspacecast (ptr addrspace(4) @.str.8 to ptr), i32 75) #33
  br label %Flow, !amdgpu.uniform !42

Flow:                                             ; preds = %_ZN22__llvm_libc_22_0_0_git8internal18CharacterConverter8pop_utf8Ev.exit140, %if.end115
  call void @llvm.amdgcn.end.cf.i64(i64 %55)
  br label %Flow189, !amdgpu.uniform !42

Flow189:                                          ; preds = %Flow, %if.end106
  call void @llvm.amdgcn.end.cf.i64(i64 %51)
  br label %Flow190, !amdgpu.uniform !42

Flow190:                                          ; preds = %Flow189, %_ZN22__llvm_libc_22_0_0_git8internal18CharacterConverter8pop_utf8Ev.exit110
  call void @llvm.amdgcn.end.cf.i64(i64 %47)
  br label %Flow191, !amdgpu.uniform !42

Flow191:                                          ; preds = %Flow190, %if.end86
  call void @llvm.amdgcn.end.cf.i64(i64 %43)
  br label %Flow192, !amdgpu.uniform !42

Flow192:                                          ; preds = %Flow191, %if.end77
  call void @llvm.amdgcn.end.cf.i64(i64 %39)
  br label %Flow193, !amdgpu.uniform !42

Flow193:                                          ; preds = %Flow192, %_ZN22__llvm_libc_22_0_0_git8internal18CharacterConverter8pop_utf8Ev.exit80
  call void @llvm.amdgcn.end.cf.i64(i64 %35)
  br label %Flow194, !amdgpu.uniform !42

Flow194:                                          ; preds = %Flow193, %_ZN22__llvm_libc_22_0_0_git8internal18CharacterConverter4pushEDi.exit47
  call void @llvm.amdgcn.end.cf.i64(i64 %31)
  br label %Flow195, !amdgpu.uniform !42

Flow195:                                          ; preds = %Flow194, %if.end49
  call void @llvm.amdgcn.end.cf.i64(i64 %27)
  br label %Flow196, !amdgpu.uniform !42

Flow196:                                          ; preds = %Flow195, %if.end40
  call void @llvm.amdgcn.end.cf.i64(i64 %23)
  br label %Flow197, !amdgpu.uniform !42

Flow197:                                          ; preds = %Flow196, %_ZN22__llvm_libc_22_0_0_git8internal18CharacterConverter8pop_utf8Ev.exit39
  call void @llvm.amdgcn.end.cf.i64(i64 %19)
  br label %Flow198, !amdgpu.uniform !42

Flow198:                                          ; preds = %Flow197, %if.end21
  call void @llvm.amdgcn.end.cf.i64(i64 %15)
  br label %Flow199, !amdgpu.uniform !42

Flow199:                                          ; preds = %Flow198, %if.end13
  call void @llvm.amdgcn.end.cf.i64(i64 %11)
  br label %Flow200, !amdgpu.uniform !42

Flow200:                                          ; preds = %Flow199, %_ZN22__llvm_libc_22_0_0_git8internal18CharacterConverter8pop_utf8Ev.exit
  call void @llvm.amdgcn.end.cf.i64(i64 %7)
  br label %UnifiedReturnBlock, !amdgpu.uniform !42

UnifiedReturnBlock:                               ; preds = %entry, %Flow200
  call void @llvm.amdgcn.end.cf.i64(i64 %3)
  ret void
}

; Function Attrs: mustprogress nofree norecurse nosync nounwind willreturn memory(none)
define internal noundef ptr @_ZNK46LlvmLibcCharacterConverterUTF32To8Test_TwoByte7getNameEv(ptr nonnull readnone align 8 captures(none) %this) unnamed_addr #6 comdat align 2 {
entry:
  ret ptr addrspacecast (ptr addrspace(4) @.str.48 to ptr)
}

; Function Attrs: cold inlinehint mustprogress noreturn nounwind memory(inaccessiblemem: write)
define internal void @_ZN46LlvmLibcCharacterConverterUTF32To8Test_OneByteD0Ev(ptr noundef nonnull readnone align 8 captures(none) dereferenceable(24) %this) unnamed_addr #7 comdat align 2 !section_prefix !58 {
entry:
  tail call void @llvm.trap()
  unreachable
}

; Function Attrs: convergent mustprogress nofree norecurse nounwind
define internal void @_ZN46LlvmLibcCharacterConverterUTF32To8Test_OneByte3RunEv(ptr noundef nonnull readonly align 8 captures(none) dereferenceable(24) %this) unnamed_addr #8 align 2 {
entry:
  %Ctx.i = getelementptr inbounds nuw i8, ptr %this, i64 16
  %0 = load ptr, ptr %Ctx.i, align 8, !tbaa !59
  %call.i = tail call fastcc noundef zeroext i1 @_ZN22__llvm_libc_22_0_0_git7testing8internal4testIbEEbPNS1_10RunContextENS0_8TestCondET_S6_PKcS8_NS1_8LocationE(ptr noundef %0, i1 noundef zeroext true, i1 noundef zeroext true, ptr noundef addrspacecast (ptr addrspace(4) @.str to ptr), ptr noundef addrspacecast (ptr addrspace(4) @.str.1 to ptr), i32 23) #33
  %1 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %call.i)
  %2 = extractvalue { i1, i64 } %1, 0
  %3 = extractvalue { i1, i64 } %1, 1
  br i1 %2, label %_ZN22__llvm_libc_22_0_0_git8internal18CharacterConverter8pop_utf8Ev.exit, label %UnifiedReturnBlock

_ZN22__llvm_libc_22_0_0_git8internal18CharacterConverter8pop_utf8Ev.exit: ; preds = %entry
  %sunkaddr = getelementptr inbounds i8, ptr %this, i64 16
  %4 = load ptr, ptr %sunkaddr, align 8, !tbaa !59
  %call.i179 = tail call fastcc noundef zeroext i1 @_ZN22__llvm_libc_22_0_0_git7testing8internal4testIbEEbPNS1_10RunContextENS0_8TestCondET_S6_PKcS8_NS1_8LocationE(ptr noundef %4, i1 noundef zeroext true, i1 noundef zeroext true, ptr noundef addrspacecast (ptr addrspace(4) @.str.3 to ptr), ptr noundef addrspacecast (ptr addrspace(4) @.str.1 to ptr), i32 25) #33
  %5 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %call.i179)
  %6 = extractvalue { i1, i64 } %5, 0
  %7 = extractvalue { i1, i64 } %5, 1
  br i1 %6, label %if.end13, label %Flow114

if.end13:                                         ; preds = %_ZN22__llvm_libc_22_0_0_git8internal18CharacterConverter8pop_utf8Ev.exit
  %sunkaddr102 = getelementptr inbounds i8, ptr %this, i64 16
  %8 = load ptr, ptr %sunkaddr102, align 8, !tbaa !59
  %call.i182 = tail call fastcc noundef zeroext i1 @_ZN22__llvm_libc_22_0_0_git7testing8internal4testIcEEbPNS1_10RunContextENS0_8TestCondET_S6_PKcS8_NS1_8LocationE(ptr noundef %8, i8 noundef signext poison, i8 noundef signext 65, ptr noundef addrspacecast (ptr addrspace(4) @.str.5 to ptr), i32 26) #33
  %9 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %call.i182)
  %10 = extractvalue { i1, i64 } %9, 0
  %11 = extractvalue { i1, i64 } %9, 1
  br i1 %10, label %if.end21, label %Flow113

if.end21:                                         ; preds = %if.end13
  %sunkaddr103 = getelementptr inbounds i8, ptr %this, i64 16
  %12 = load ptr, ptr %sunkaddr103, align 8, !tbaa !59
  %call.i185 = tail call fastcc noundef zeroext i1 @_ZN22__llvm_libc_22_0_0_git7testing8internal4testIbEEbPNS1_10RunContextENS0_8TestCondET_S6_PKcS8_NS1_8LocationE(ptr noundef %12, i1 noundef zeroext true, i1 noundef zeroext true, ptr noundef addrspacecast (ptr addrspace(4) @.str.6 to ptr), ptr noundef addrspacecast (ptr addrspace(4) @.str.1 to ptr), i32 27) #33
  %13 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %call.i185)
  %14 = extractvalue { i1, i64 } %13, 0
  %15 = extractvalue { i1, i64 } %13, 1
  br i1 %14, label %_ZN22__llvm_libc_22_0_0_git8internal18CharacterConverter4pushEDi.exit17, label %Flow112

_ZN22__llvm_libc_22_0_0_git8internal18CharacterConverter4pushEDi.exit17: ; preds = %if.end21
  %sunkaddr104 = getelementptr inbounds i8, ptr %this, i64 16
  %16 = load ptr, ptr %sunkaddr104, align 8, !tbaa !59
  %call.i188 = tail call fastcc noundef zeroext i1 @_ZN22__llvm_libc_22_0_0_git7testing8internal4testIbEEbPNS1_10RunContextENS0_8TestCondET_S6_PKcS8_NS1_8LocationE(ptr noundef %16, i1 noundef zeroext true, i1 noundef zeroext true, ptr noundef addrspacecast (ptr addrspace(4) @.str to ptr), ptr noundef addrspacecast (ptr addrspace(4) @.str.1 to ptr), i32 31) #33
  %17 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %call.i188)
  %18 = extractvalue { i1, i64 } %17, 0
  %19 = extractvalue { i1, i64 } %17, 1
  br i1 %18, label %_ZN22__llvm_libc_22_0_0_git8internal18CharacterConverter8pop_utf8Ev.exit50, label %Flow111

_ZN22__llvm_libc_22_0_0_git8internal18CharacterConverter8pop_utf8Ev.exit50: ; preds = %_ZN22__llvm_libc_22_0_0_git8internal18CharacterConverter4pushEDi.exit17
  %sunkaddr105 = getelementptr inbounds i8, ptr %this, i64 16
  %20 = load ptr, ptr %sunkaddr105, align 8, !tbaa !59
  %call.i193 = tail call fastcc noundef zeroext i1 @_ZN22__llvm_libc_22_0_0_git7testing8internal4testIbEEbPNS1_10RunContextENS0_8TestCondET_S6_PKcS8_NS1_8LocationE(ptr noundef %20, i1 noundef zeroext true, i1 noundef zeroext true, ptr noundef addrspacecast (ptr addrspace(4) @.str.3 to ptr), ptr noundef addrspacecast (ptr addrspace(4) @.str.1 to ptr), i32 33) #33
  %21 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %call.i193)
  %22 = extractvalue { i1, i64 } %21, 0
  %23 = extractvalue { i1, i64 } %21, 1
  br i1 %22, label %if.end49, label %Flow110

if.end49:                                         ; preds = %_ZN22__llvm_libc_22_0_0_git8internal18CharacterConverter8pop_utf8Ev.exit50
  %sunkaddr106 = getelementptr inbounds i8, ptr %this, i64 16
  %24 = load ptr, ptr %sunkaddr106, align 8, !tbaa !59
  %call.i196 = tail call fastcc noundef zeroext i1 @_ZN22__llvm_libc_22_0_0_git7testing8internal4testIcEEbPNS1_10RunContextENS0_8TestCondET_S6_PKcS8_NS1_8LocationE(ptr noundef %24, i8 noundef signext poison, i8 noundef signext 66, ptr noundef addrspacecast (ptr addrspace(4) @.str.7 to ptr), i32 34) #33
  %25 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %call.i196)
  %26 = extractvalue { i1, i64 } %25, 0
  %27 = extractvalue { i1, i64 } %25, 1
  br i1 %26, label %if.end57, label %Flow109

if.end57:                                         ; preds = %if.end49
  %sunkaddr107 = getelementptr inbounds i8, ptr %this, i64 16
  %28 = load ptr, ptr %sunkaddr107, align 8, !tbaa !59
  %call.i199 = tail call fastcc noundef zeroext i1 @_ZN22__llvm_libc_22_0_0_git7testing8internal4testIbEEbPNS1_10RunContextENS0_8TestCondET_S6_PKcS8_NS1_8LocationE(ptr noundef %28, i1 noundef zeroext true, i1 noundef zeroext true, ptr noundef addrspacecast (ptr addrspace(4) @.str.6 to ptr), ptr noundef addrspacecast (ptr addrspace(4) @.str.1 to ptr), i32 35) #33
  %29 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %call.i199)
  %30 = extractvalue { i1, i64 } %29, 0
  %31 = extractvalue { i1, i64 } %29, 1
  br i1 %30, label %_ZN22__llvm_libc_22_0_0_git8internal18CharacterConverter8pop_utf8Ev.exit80, label %Flow

_ZN22__llvm_libc_22_0_0_git8internal18CharacterConverter8pop_utf8Ev.exit80: ; preds = %if.end57
  %sunkaddr108 = getelementptr inbounds i8, ptr %this, i64 16
  %32 = load ptr, ptr %sunkaddr108, align 8, !tbaa !59
  %33 = tail call fastcc noundef zeroext i1 @_ZN22__llvm_libc_22_0_0_git7testing8internal4testIbEEbPNS1_10RunContextENS0_8TestCondET_S6_PKcS8_NS1_8LocationE(ptr noundef %32, i1 noundef zeroext false, i1 noundef zeroext false, ptr noundef addrspacecast (ptr addrspace(4) @.str.3 to ptr), ptr noundef addrspacecast (ptr addrspace(4) @.str.8 to ptr), i32 39) #33
  br label %Flow, !amdgpu.uniform !42

Flow:                                             ; preds = %_ZN22__llvm_libc_22_0_0_git8internal18CharacterConverter8pop_utf8Ev.exit80, %if.end57
  call void @llvm.amdgcn.end.cf.i64(i64 %31)
  br label %Flow109, !amdgpu.uniform !42

Flow109:                                          ; preds = %Flow, %if.end49
  call void @llvm.amdgcn.end.cf.i64(i64 %27)
  br label %Flow110, !amdgpu.uniform !42

Flow110:                                          ; preds = %Flow109, %_ZN22__llvm_libc_22_0_0_git8internal18CharacterConverter8pop_utf8Ev.exit50
  call void @llvm.amdgcn.end.cf.i64(i64 %23)
  br label %Flow111, !amdgpu.uniform !42

Flow111:                                          ; preds = %Flow110, %_ZN22__llvm_libc_22_0_0_git8internal18CharacterConverter4pushEDi.exit17
  call void @llvm.amdgcn.end.cf.i64(i64 %19)
  br label %Flow112, !amdgpu.uniform !42

Flow112:                                          ; preds = %Flow111, %if.end21
  call void @llvm.amdgcn.end.cf.i64(i64 %15)
  br label %Flow113, !amdgpu.uniform !42

Flow113:                                          ; preds = %Flow112, %if.end13
  call void @llvm.amdgcn.end.cf.i64(i64 %11)
  br label %Flow114, !amdgpu.uniform !42

Flow114:                                          ; preds = %Flow113, %_ZN22__llvm_libc_22_0_0_git8internal18CharacterConverter8pop_utf8Ev.exit
  call void @llvm.amdgcn.end.cf.i64(i64 %7)
  br label %UnifiedReturnBlock, !amdgpu.uniform !42

UnifiedReturnBlock:                               ; preds = %entry, %Flow114
  call void @llvm.amdgcn.end.cf.i64(i64 %3)
  ret void
}

; Function Attrs: mustprogress nofree norecurse nosync nounwind willreturn memory(none)
define internal noundef ptr @_ZNK46LlvmLibcCharacterConverterUTF32To8Test_OneByte7getNameEv(ptr nonnull readnone align 8 captures(none) %this) unnamed_addr #6 comdat align 2 {
entry:
  ret ptr addrspacecast (ptr addrspace(4) @.str.47 to ptr)
}

; Function Attrs: convergent mustprogress nofree norecurse nounwind
define internal fastcc noundef zeroext i1 @_ZN22__llvm_libc_22_0_0_git7testing8internal4testIcEEbPNS1_10RunContextENS0_8TestCondET_S6_PKcS8_NS1_8LocationE(ptr noundef writeonly captures(none) %Ctx, i8 noundef signext %LHS, i8 noundef signext range(i8 65, 67) %RHS, ptr noundef %RHSStr, i32 range(i32 26, 35) %Loc.coerce1) unnamed_addr #8 comdat {
entry:
  %buffer.i.i.i = alloca %"class.__llvm_libc_22_0_0_git::IntegerToString", align 8, addrspace(5)
  %buffer.i.i.i91.i = alloca %"class.__llvm_libc_22_0_0_git::IntegerToString", align 8, addrspace(5)
  %buffer.i.i.i.i = alloca %"class.__llvm_libc_22_0_0_git::IntegerToString", align 8, addrspace(5)
  %cmp = icmp eq i8 %LHS, %RHS
  %cmp.inv = xor i1 %cmp, true
  %0 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.inv)
  %1 = extractvalue { i1, i64 } %0, 0
  %2 = extractvalue { i1, i64 } %0, 1
  br i1 %1, label %if.end.i, label %Flow3892

if.end.i:                                         ; preds = %entry
  store i8 1, ptr %Ctx, align 1, !tbaa !65
  %spec.select.i = tail call i64 @llvm.usub.sat.i64(i64 sub (i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) getelementptr inbounds nuw (i8, ptr addrspace(4) @.str.9, i64 8) to ptr) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.9 to ptr) to i64)), i64 2)
  %cmp.i.not.i.i = icmp uge i64 sub (i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) getelementptr inbounds nuw (i8, ptr addrspace(4) @.str.9, i64 8) to ptr) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.9 to ptr) to i64)), 3
  br i1 %cmp.i.not.i.i, label %if.then.i.i.i, label %Flow3891, !amdgpu.uniform !42

if.then.i.i.i:                                    ; preds = %if.end.i
  %inc.i.i.i.i = add nuw i64 %spec.select.i, 1
  %cmp2.i.i.i.i = icmp ult i64 %spec.select.i, 1676976733973595600
  br i1 %cmp2.i.i.i.i, label %if.then3.i.i.i.i, label %if.end4.i.i.i.i, !amdgpu.uniform !42

if.then3.i.i.i.i:                                 ; preds = %if.then.i.i.i
  %mul.i.i.i.i = mul nuw i64 %inc.i.i.i.i, 11
  %div18.i.i.i.i = lshr i64 %mul.i.i.i.i, 3
  br label %if.end4.i.i.i.i, !amdgpu.uniform !42

Flow3891:                                         ; preds = %Flow3890, %if.end.i
  %3 = phi ptr [ %cond.i.i, %Flow3890 ], [ addrspacecast (ptr addrspace(4) @_ZN22__llvm_libc_22_0_0_git3cpp6string14NULL_CHARACTERE to ptr), %if.end.i ]
  br label %if.end.i.i.i, !amdgpu.uniform !42

if.end4.i.i.i.i:                                  ; preds = %if.then3.i.i.i.i, %if.then.i.i.i
  %new_capacity.addr.0.i.i.i.i = phi i64 [ %div18.i.i.i.i, %if.then3.i.i.i.i ], [ %inc.i.i.i.i, %if.then.i.i.i ]
  %sub.i.i = add i64 %new_capacity.addr.0.i.i.i.i, 7
  %div2.i.i = and i64 %sub.i.i, -8
  %4 = load ptr, ptr addrspace(1) @_ZN12_GLOBAL__N_13ptrE, align 8, !tbaa !55
  %add.ptr.i.i = getelementptr inbounds nuw i8, ptr %4, i64 %div2.i.i
  store ptr %add.ptr.i.i, ptr addrspace(1) @_ZN12_GLOBAL__N_13ptrE, align 8, !tbaa !55
  %sub.ptr.lhs.cast.i.i = ptrtoint ptr %add.ptr.i.i to i64
  %sub.ptr.sub.i.i = sub i64 %sub.ptr.lhs.cast.i.i, ptrtoint (ptr addrspacecast (ptr addrspace(1) @_ZN12_GLOBAL__N_16memoryE to ptr) to i64)
  %cmp.i.i = icmp ugt i64 %sub.ptr.sub.i.i, 65335
  %cond.i.i = select i1 %cmp.i.i, ptr null, ptr %4
  %5 = icmp ne i64 0, %spec.select.i
  br i1 %5, label %loadstoreloop2192.preheader, label %Flow3890, !amdgpu.uniform !42

loadstoreloop2192.preheader:                      ; preds = %if.end4.i.i.i.i
  br label %loadstoreloop2192, !amdgpu.uniform !42

Flow3890:                                         ; preds = %Flow3889, %if.end4.i.i.i.i
  br label %Flow3891, !amdgpu.uniform !42

loadstoreloop2192:                                ; preds = %loadstoreloop2192.preheader, %loadstoreloop2192
  %6 = phi i64 [ %8, %loadstoreloop2192 ], [ 0, %loadstoreloop2192.preheader ]
  %7 = getelementptr i8, ptr %4, i64 %6
  store i8 0, ptr %7, align 1
  %8 = add i64 %6, 1
  %9 = icmp uge i64 %8, %spec.select.i
  br i1 %9, label %Flow3889, label %loadstoreloop2192, !amdgpu.uniform !42

Flow3889:                                         ; preds = %loadstoreloop2192
  br label %Flow3890, !amdgpu.uniform !42

Flow3892:                                         ; preds = %2702, %entry
  call void @llvm.amdgcn.end.cf.i64(i64 %2)
  br label %_ZZN22__llvm_libc_22_0_0_git7testing8internal4testIcEEbPNS1_10RunContextENS0_8TestCondET_S6_PKcS8_NS1_8LocationEENKUlbNS_3cpp11string_viewEE_clEbSB_.exit, !amdgpu.uniform !42

if.end.i.i.i:                                     ; preds = %Flow3891
  %cmp.not.i6.i.i.i = icmp ne ptr %3, addrspacecast (ptr addrspace(4) @_ZN22__llvm_libc_22_0_0_git3cpp6string14NULL_CHARACTERE to ptr)
  br i1 %cmp.not.i6.i.i.i, label %if.then.i.i.i.i, label %_ZN22__llvm_libc_22_0_0_git3cpp6stringC2Emc.exit.i, !amdgpu.uniform !42

if.then.i.i.i.i:                                  ; preds = %if.end.i.i.i
  %arrayidx.i.i.i.i = getelementptr inbounds nuw i8, ptr %3, i64 %spec.select.i
  store i8 0, ptr %arrayidx.i.i.i.i, align 1, !tbaa !57
  br label %_ZN22__llvm_libc_22_0_0_git3cpp6stringC2Emc.exit.i, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git3cpp6stringC2Emc.exit.i: ; preds = %if.then.i.i.i.i, %if.end.i.i.i
  %10 = icmp ne i64 0, %spec.select.i
  br i1 %10, label %loadstoreloop2190.preheader, label %Flow3888, !amdgpu.uniform !42

loadstoreloop2190.preheader:                      ; preds = %_ZN22__llvm_libc_22_0_0_git3cpp6stringC2Emc.exit.i
  br label %loadstoreloop2190, !amdgpu.uniform !42

Flow3888:                                         ; preds = %Flow3887, %_ZN22__llvm_libc_22_0_0_git3cpp6stringC2Emc.exit.i
  br label %split2189, !amdgpu.uniform !42

loadstoreloop2190:                                ; preds = %loadstoreloop2190.preheader, %loadstoreloop2190
  %11 = phi i64 [ %13, %loadstoreloop2190 ], [ 0, %loadstoreloop2190.preheader ]
  %12 = getelementptr i8, ptr %3, i64 %11
  store i8 32, ptr %12, align 1
  %13 = add i64 %11, 1
  %14 = icmp uge i64 %13, %spec.select.i
  br i1 %14, label %Flow3887, label %loadstoreloop2190, !amdgpu.uniform !42

Flow3887:                                         ; preds = %loadstoreloop2190
  br label %Flow3888, !amdgpu.uniform !42

split2189:                                        ; preds = %Flow3888
  %15 = tail call i32 @llvm.amdgcn.mbcnt.lo(i32 -1, i32 0)
  %16 = tail call noundef i32 @llvm.amdgcn.mbcnt.hi(i32 -1, i32 %15)
  %sh_prom.i.i.i.i = zext i32 %16 to i64
  br label %for.cond.i.i1.i, !amdgpu.uniform !42

for.cond.i.i1.i:                                  ; preds = %split2189, %Flow3886
  %17 = phi i32 [ %55, %Flow3886 ], [ poison, %split2189 ]
  %18 = phi i64 [ %56, %Flow3886 ], [ poison, %split2189 ]
  %19 = phi i32 [ %57, %Flow3886 ], [ poison, %split2189 ]
  %retval.sroa.6.0.i.i.i = phi i32 [ %60, %Flow3886 ], [ undef, %split2189 ]
  %retval.sroa.2.0.i.i.i = phi i64 [ %59, %Flow3886 ], [ undef, %split2189 ]
  %retval.sroa.8.0.i.i.i = phi i32 [ %58, %Flow3886 ], [ undef, %split2189 ]
  %index.0.i.i.i = phi i32 [ %62, %Flow3886 ], [ 0, %split2189 ]
  %20 = load i32, ptr addrspace(1) @__llvm_rpc_client, align 8, !tbaa !68
  %cmp.not.i.i.i = icmp ult i32 %index.0.i.i.i, %20
  %spec.store.select.i.i.i = select i1 %cmp.not.i.i.i, i32 %index.0.i.i.i, i32 0
  %21 = tail call noundef i64 @llvm.amdgcn.ballot.i64(i1 true)
  %22 = and i32 %spec.store.select.i.i.i, 31
  %23 = lshr i64 %21, %sh_prom.i.i.i.i
  %24 = trunc i64 %23 to i32
  %conv4.i.i.i.i.i = and i32 %24, 1
  %shl.i.i.i.i.i = shl nuw i32 %conv4.i.i.i.i.i, %22
  %25 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop, !amdgpu.uniform !42

26:                                               ; preds = %ComputeEnd
  %sunkaddr = getelementptr inbounds i8, ptr addrspace(1) %2931, i64 40
  %27 = atomicrmw or ptr addrspace(1) %sunkaddr, i32 %.lcssa4195 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %28, !amdgpu.uniform !42

28:                                               ; preds = %26, %ComputeEnd
  %29 = phi i32 [ %27, %26 ], [ poison, %ComputeEnd ]
  call void @llvm.amdgcn.end.cf.i64(i64 %2940)
  %30 = call i32 @llvm.amdgcn.readfirstlane.i32(i32 %29)
  %31 = or i32 %30, %.lcssa4196
  %shl5.i.i.i.i.i = shl nuw i32 1, %22
  %and.i.i.i.i.i = and i32 %shl5.i.i.i.i.i, %31
  %tobool3.i.i.i.i = icmp ne i32 %and.i.i.i.i.i, 0
  %32 = tail call i64 @llvm.amdgcn.ballot.i64(i1 %tobool3.i.i.i.i)
  %and.i.i.i.i.i.i = and i64 %32, %21
  %cmp.i.not.i.i.i = icmp ne i64 %21, %and.i.i.i.i.i.i
  br i1 %cmp.i.not.i.i.i, label %if.then.i.i.i.i2, label %Flow3885, !amdgpu.uniform !42

if.then.i.i.i.i2:                                 ; preds = %28
  fence syncscope("agent") acquire
  %33 = load <2 x i64>, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !75
  %34 = extractelement <2 x i64> %33, i32 0
  %35 = inttoptr i64 %34 to ptr
  %36 = extractelement <2 x i64> %33, i32 1
  %37 = inttoptr i64 %36 to ptr
  %idxprom.i.i.i.i = zext i32 %spec.store.select.i.i.i to i64
  %arrayidx.i.i.i.i3 = getelementptr inbounds nuw i32, ptr %35, i64 %idxprom.i.i.i.i, !amdgpu.uniform !42
  %38 = load atomic i32, ptr %arrayidx.i.i.i.i3 monotonic, align 4
  %39 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %38)
  %arrayidx.i45.i.i.i = getelementptr inbounds nuw i32, ptr %37, i64 %idxprom.i.i.i.i, !amdgpu.uniform !42
  %40 = load atomic i32, ptr %arrayidx.i45.i.i.i monotonic, align 4
  %41 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %40)
  %cmp.i46.not.i.i.i = icmp ne i32 %39, %41
  br i1 %cmp.i46.not.i.i.i, label %if.then12.i.i.i, label %Flow3883, !amdgpu.uniform !42

if.then12.i.i.i:                                  ; preds = %if.then.i.i.i.i2
  fence syncscope("agent") release
  %42 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %21, i1 true)
  %iszero.i.i.i.i.i.i.i = icmp ne i64 %21, 0
  %cmp2.i.i.i.i.i.i = icmp eq i64 %42, %sh_prom.i.i.i.i
  %cmp.i.i.i.i.i.i = select i1 %iszero.i.i.i.i.i.i.i, i1 %cmp2.i.i.i.i.i.i, i1 false
  %conv4.i.i51.i.i.i = zext i1 %cmp.i.i.i.i.i.i to i32
  %shl.i.i52.i.i.i = shl nuw i32 %conv4.i.i51.i.i.i, %22
  %xor.i.i.i.i.i = xor i32 %shl.i.i52.i.i.i, -1
  %43 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2427, !amdgpu.uniform !42

44:                                               ; preds = %ComputeEnd2428
  %sunkaddr3272 = getelementptr inbounds i8, ptr addrspace(1) %2931, i64 40
  %45 = atomicrmw and ptr addrspace(1) %sunkaddr3272, i32 %.lcssa4197 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %46, !amdgpu.uniform !42

46:                                               ; preds = %44, %ComputeEnd2428
  call void @llvm.amdgcn.end.cf.i64(i64 %2958)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %Flow3883, !amdgpu.uniform !42

if.end14.i.i.i:                                   ; preds = %Flow3883
  %47 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %21, i1 true)
  %iszero.i.i.i.i.i.i = icmp ne i64 %21, 0
  %cmp2.i.i.i.i.i = icmp eq i64 %47, %sh_prom.i.i.i.i
  %cmp.i.i.i.i.i = select i1 %iszero.i.i.i.i.i.i, i1 %cmp2.i.i.i.i.i, i1 false
  %48 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.i.i.i.i.i)
  %49 = extractvalue { i1, i64 } %48, 0
  %50 = extractvalue { i1, i64 } %48, 1
  br i1 %49, label %if.then16.i.i.i, label %if.end22.i.i.i

if.then16.i.i.i:                                  ; preds = %if.end14.i.i.i
  %51 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !76
  %opcode.i.i.i = getelementptr inbounds nuw %"struct.rpc::Header", ptr %51, i64 %idxprom.i.i.i.i, i32 1
  store i32 1660944387, ptr %opcode.i.i.i, align 8, !tbaa !77
  %arrayidx21.i.i.i = getelementptr inbounds nuw %"struct.rpc::Header", ptr %51, i64 %idxprom.i.i.i.i
  store i64 %21, ptr %arrayidx21.i.i.i, align 8, !tbaa !80
  br label %if.end22.i.i.i, !amdgpu.uniform !42

Flow3884:                                         ; preds = %if.end22.i.i.i, %Flow3883
  %52 = phi i32 [ %39, %if.end22.i.i.i ], [ %retval.sroa.8.0.i.i.i, %Flow3883 ]
  %53 = phi i64 [ %21, %if.end22.i.i.i ], [ %retval.sroa.2.0.i.i.i, %Flow3883 ]
  %54 = phi i32 [ %spec.store.select.i.i.i, %if.end22.i.i.i ], [ %retval.sroa.6.0.i.i.i, %Flow3883 ]
  br label %cleanup26.i.i.i, !amdgpu.uniform !42

if.end22.i.i.i:                                   ; preds = %if.then16.i.i.i, %if.end14.i.i.i
  call void @llvm.amdgcn.end.cf.i64(i64 %50)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %Flow3884, !amdgpu.uniform !42

Flow3885:                                         ; preds = %cleanup26.i.i.i, %28
  %55 = phi i32 [ %54, %cleanup26.i.i.i ], [ %17, %28 ]
  %56 = phi i64 [ %53, %cleanup26.i.i.i ], [ %18, %28 ]
  %57 = phi i32 [ %52, %cleanup26.i.i.i ], [ %19, %28 ]
  %58 = phi i32 [ %52, %cleanup26.i.i.i ], [ %retval.sroa.8.0.i.i.i, %28 ]
  %59 = phi i64 [ %53, %cleanup26.i.i.i ], [ %retval.sroa.2.0.i.i.i, %28 ]
  %60 = phi i32 [ %54, %cleanup26.i.i.i ], [ %retval.sroa.6.0.i.i.i, %28 ]
  %61 = phi i1 [ %cmp.i46.not.i.i.i, %cleanup26.i.i.i ], [ true, %28 ]
  br i1 %61, label %for.inc.i.i.i, label %Flow3886, !amdgpu.uniform !42

cleanup26.i.i.i:                                  ; preds = %Flow3884
  br label %Flow3885, !amdgpu.uniform !42

for.inc.i.i.i:                                    ; preds = %Flow3885
  %inc.i.i.i = add i32 %spec.store.select.i.i.i, 1
  br label %Flow3886, !amdgpu.uniform !42

Flow3886:                                         ; preds = %for.inc.i.i.i, %Flow3885
  %62 = phi i32 [ %inc.i.i.i, %for.inc.i.i.i ], [ poison, %Flow3885 ]
  %63 = phi i1 [ false, %for.inc.i.i.i ], [ true, %Flow3885 ]
  br i1 %63, label %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i, label %for.cond.i.i1.i, !amdgpu.uniform !42

_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i: ; preds = %Flow3886
  %.lcssa4200 = phi i32 [ %55, %Flow3886 ]
  %.lcssa4199 = phi i64 [ %56, %Flow3886 ]
  %.lcssa4198 = phi i32 [ %57, %Flow3886 ]
  fence acquire
  %64 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %mul.i.i.i.i.i.i = shl i32 %.lcssa4200, 6
  %idxprom.i19.i.i.i.i.i = zext i32 %mul.i.i.i.i.i.i to i64
  %arrayidx.i20.i.i.i.i.i = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %64, i64 %idxprom.i19.i.i.i.i.i
  %arrayidx.i22.i.i.i.i.i = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %arrayidx.i20.i.i.i.i.i, i64 %sh_prom.i.i.i.i
  store i64 sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.2 to ptr), i64 81) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.2 to ptr) to i64)), ptr %arrayidx.i22.i.i.i.i.i, align 8, !tbaa !82
  %spec.select.i.i.i.i.i.i.i = tail call i64 @llvm.umin.i64(i64 sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.2 to ptr), i64 81) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.2 to ptr) to i64)), i64 56)
  %65 = and i64 %spec.select.i.i.i.i.i.i.i, 15
  %66 = sub i64 %spec.select.i.i.i.i.i.i.i, %65
  %67 = icmp ne i64 %66, 0
  br i1 %67, label %loop-memcpy-expansion2308.preheader, label %Flow3882, !amdgpu.uniform !42

loop-memcpy-expansion2308.preheader:              ; preds = %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i
  %68 = shl nuw nsw i64 %sh_prom.i.i.i.i, 6
  %69 = add nuw nsw i64 %68, 8
  %scevgep3235 = getelementptr i8, ptr %64, i64 %69
  %70 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i, 6
  %scevgep3236 = getelementptr i8, ptr %scevgep3235, i64 %70
  br label %loop-memcpy-expansion2308, !amdgpu.uniform !42

Flow3882:                                         ; preds = %Flow3881, %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i
  br label %loop-memcpy-residual-header2311, !amdgpu.uniform !42

loop-memcpy-expansion2308:                        ; preds = %loop-memcpy-expansion2308.preheader, %loop-memcpy-expansion2308
  %loop-index2309 = phi i64 [ %73, %loop-memcpy-expansion2308 ], [ 0, %loop-memcpy-expansion2308.preheader ]
  %71 = getelementptr i8, ptr addrspace(4) @.str.2, i64 %loop-index2309, !amdgpu.uniform !42
  %72 = load <4 x i32>, ptr addrspace(4) %71, align 1
  %scevgep3237 = getelementptr i8, ptr %scevgep3236, i64 %loop-index2309
  store <4 x i32> %72, ptr %scevgep3237, align 1
  %73 = add i64 %loop-index2309, 16
  %74 = icmp uge i64 %73, %66
  br i1 %74, label %Flow3881, label %loop-memcpy-expansion2308, !amdgpu.uniform !42

Flow3880:                                         ; preds = %Flow3878, %Flow3879
  %75 = phi i64 [ %2842, %Flow3878 ], [ %2839, %Flow3879 ]
  %76 = phi i64 [ %2844, %Flow3878 ], [ %2840, %Flow3879 ]
  br label %post-loop-memcpy-expansion2307, !amdgpu.uniform !42

loop-memcpy-residual2310:                         ; preds = %loop-memcpy-residual2310.preheader, %loop-memcpy-residual2310
  %residual-loop-index2312 = phi i64 [ %78, %loop-memcpy-residual2310 ], [ 0, %loop-memcpy-residual2310.preheader ]
  %scevgep3234 = getelementptr i8, ptr addrspace(4) getelementptr inbounds nuw (i8, ptr addrspace(4) @.str.2, i64 48), i64 %residual-loop-index2312, !amdgpu.uniform !42
  %77 = load i8, ptr addrspace(4) %scevgep3234, align 1
  %scevgep3233 = getelementptr i8, ptr %scevgep3232, i64 %residual-loop-index2312
  store i8 %77, ptr %scevgep3233, align 1
  %78 = add i64 %residual-loop-index2312, 1
  %79 = icmp uge i64 %78, %65
  br i1 %79, label %Flow3878, label %loop-memcpy-residual2310, !amdgpu.uniform !42

Flow3878:                                         ; preds = %loop-memcpy-residual2310
  br label %Flow3880, !amdgpu.uniform !42

post-loop-memcpy-expansion2307:                   ; preds = %Flow3880
  %tobool.not.i.i.i.i.i.i = icmp eq i32 %.lcssa4198, 0
  %conv.i.i.i.i.i.i = zext i1 %tobool.not.i.i.i.i.i.i to i32
  fence release
  %80 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %idxprom.i23.i.i.i.i.i = zext i32 %.lcssa4200 to i64
  %arrayidx.i24.i.i.i.i.i = getelementptr inbounds nuw i32, ptr %80, i64 %idxprom.i23.i.i.i.i.i
  store atomic i32 %conv.i.i.i.i.i.i, ptr %arrayidx.i24.i.i.i.i.i monotonic, align 4
  %81 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !84
  %arrayidx.i.i5.i.i = getelementptr inbounds nuw %"struct.rpc::Header", ptr %81, i64 %idxprom.i23.i.i.i.i.i, !amdgpu.uniform !42
  %82 = load i64, ptr %arrayidx.i.i5.i.i, align 8, !tbaa !80
  br label %while.cond.i.i.i.i, !amdgpu.uniform !42

while.cond.i.i.i.i:                               ; preds = %post-loop-memcpy-expansion2307, %Flow3877
  %phi.broken = phi i64 [ 0, %post-loop-memcpy-expansion2307 ], [ %99, %Flow3877 ]
  %lsr.iv3224 = phi i64 [ %97, %Flow3877 ], [ 25, %post-loop-memcpy-expansion2307 ]
  %lsr.iv3219 = phi ptr addrspace(4) [ %96, %Flow3877 ], [ getelementptr inbounds nuw (i8, ptr addrspace(4) @.str.2, i64 56), %post-loop-memcpy-expansion2307 ]
  %port.sroa.37.0.i.i = phi i32 [ %95, %Flow3877 ], [ %conv.i.i.i.i.i.i, %post-loop-memcpy-expansion2307 ]
  %idx.0.i.i.i.i = phi i64 [ %94, %Flow3877 ], [ 56, %post-loop-memcpy-expansion2307 ]
  %cmp.i.i.i.i = icmp ult i64 %idx.0.i.i.i.i, sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.2 to ptr), i64 81) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.2 to ptr) to i64))
  %83 = tail call i64 @llvm.amdgcn.ballot.i64(i1 %cmp.i.i.i.i)
  %and.i.i.i.i6.i.i = and i64 %83, %82
  %tobool.not.i.i.i.i = icmp ne i64 %and.i.i.i.i6.i.i, 0
  %84 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i22.i.i.i = getelementptr inbounds nuw i32, ptr %84, i64 %idxprom.i23.i.i.i.i.i, !amdgpu.uniform !42
  %85 = load atomic i32, ptr %arrayidx.i22.i.i.i monotonic, align 4
  %86 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %tobool.not.i.i.i.i)
  %87 = extractvalue { i1, i64 } %86, 0
  %88 = extractvalue { i1, i64 } %86, 1
  br i1 %87, label %cond.false.i8.i.i.i.i, label %Flow3877

cond.false.i8.i.i.i.i:                            ; preds = %while.cond.i.i.i.i
  %89 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %85)
  %90 = icmp ne i32 %89, %port.sroa.37.0.i.i
  br i1 %90, label %while.body.i.i21.i.i.i.i.preheader, label %Flow3876, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i.preheader:               ; preds = %cond.false.i8.i.i.i.i
  br label %while.body.i.i21.i.i.i.i, !amdgpu.uniform !42

Flow3876:                                         ; preds = %Flow3875, %cond.false.i8.i.i.i.i
  br label %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i:                         ; preds = %while.body.i.i21.i.i.i.i.preheader, %while.body.i.i21.i.i.i.i
  tail call void @llvm.amdgcn.s.sleep(i32 2)
  %91 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i22.i.i.i.i = getelementptr inbounds nuw i32, ptr %91, i64 %idxprom.i23.i.i.i.i.i, !amdgpu.uniform !42
  %92 = load atomic i32, ptr %arrayidx.i.i.i22.i.i.i.i monotonic, align 4
  %93 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %92)
  %cmp.i.not.i.i23.i.i.i.i = icmp eq i32 %93, %port.sroa.37.0.i.i
  br i1 %cmp.i.not.i.i23.i.i.i.i, label %Flow3875, label %while.body.i.i21.i.i.i.i, !amdgpu.uniform !42

Flow3875:                                         ; preds = %while.body.i.i21.i.i.i.i
  br label %Flow3876, !amdgpu.uniform !42

Flow3877:                                         ; preds = %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i, %while.cond.i.i.i.i
  %94 = phi i64 [ %add.i.i.i.i, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i ], [ poison, %while.cond.i.i.i.i ]
  %95 = phi i32 [ %conv.i.i26.i.i.i.i, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i ], [ poison, %while.cond.i.i.i.i ]
  %96 = phi ptr addrspace(4) [ %scevgep3220, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i ], [ poison, %while.cond.i.i.i.i ]
  %97 = phi i64 [ %lsr.iv.next3225, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i ], [ poison, %while.cond.i.i.i.i ]
  %98 = phi i1 [ false, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i ], [ true, %while.cond.i.i.i.i ]
  call void @llvm.amdgcn.end.cf.i64(i64 %88)
  %99 = call i64 @llvm.amdgcn.if.break.i64(i1 %98, i64 %phi.broken)
  %100 = call i1 @llvm.amdgcn.loop.i64(i64 %99)
  br i1 %100, label %cond.false.i.i.i, label %while.cond.i.i.i.i

_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i: ; preds = %Flow3876
  fence acquire
  br i1 %cmp.i.i.i.i, label %if.then.i.i.i.i.i.i.i4, label %Flow3874, !amdgpu.uniform !42

if.then.i.i.i.i.i.i.i4:                           ; preds = %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i
  %sub.i.i.i.i.i.i.i = sub nuw i64 sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.2 to ptr), i64 81) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.2 to ptr) to i64)), %idx.0.i.i.i.i
  %spec.select.i.i.i36.i.i.i.i = tail call i64 @llvm.umin.i64(i64 %sub.i.i.i.i.i.i.i, i64 64)
  %101 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %102 = and i64 %spec.select.i.i.i36.i.i.i.i, 15
  %103 = sub i64 %spec.select.i.i.i36.i.i.i.i, %102
  %104 = icmp ne i64 %103, 0
  br i1 %104, label %loop-memcpy-expansion2314.preheader, label %Flow3873, !amdgpu.uniform !42

loop-memcpy-expansion2314.preheader:              ; preds = %if.then.i.i.i.i.i.i.i4
  %arrayidx.i20.i35.i.i.i.i = getelementptr %"struct.rpc::Buffer", ptr %101, i64 %idxprom.i19.i.i.i.i.i
  %arrayidx.i22.i38.i.i.i.i = getelementptr %"struct.rpc::Buffer", ptr %arrayidx.i20.i35.i.i.i.i, i64 %sh_prom.i.i.i.i
  br label %loop-memcpy-expansion2314, !amdgpu.uniform !42

Flow3873:                                         ; preds = %Flow3872, %if.then.i.i.i.i.i.i.i4
  br label %loop-memcpy-residual-header2317, !amdgpu.uniform !42

loop-memcpy-expansion2314:                        ; preds = %loop-memcpy-expansion2314.preheader, %loop-memcpy-expansion2314
  %loop-index2315 = phi i64 [ %107, %loop-memcpy-expansion2314 ], [ 0, %loop-memcpy-expansion2314.preheader ]
  %scevgep3221 = getelementptr i8, ptr addrspace(4) %lsr.iv3219, i64 %loop-index2315, !amdgpu.uniform !42
  %105 = load <4 x i32>, ptr addrspace(4) %scevgep3221, align 1
  %106 = getelementptr i8, ptr %arrayidx.i22.i38.i.i.i.i, i64 %loop-index2315
  store <4 x i32> %105, ptr %106, align 1
  %107 = add i64 %loop-index2315, 16
  %108 = icmp uge i64 %107, %103
  br i1 %108, label %Flow3872, label %loop-memcpy-expansion2314, !amdgpu.uniform !42

Flow3871:                                         ; preds = %Flow3870, %loop-memcpy-residual-header2317
  br label %Flow3874, !amdgpu.uniform !42

loop-memcpy-residual2316:                         ; preds = %loop-memcpy-residual2316.preheader, %loop-memcpy-residual2316
  %residual-loop-index2318 = phi i64 [ %110, %loop-memcpy-residual2316 ], [ 0, %loop-memcpy-residual2316.preheader ]
  %scevgep3230 = getelementptr i8, ptr addrspace(4) %scevgep3229, i64 %residual-loop-index2318, !amdgpu.uniform !42
  %109 = load i8, ptr addrspace(4) %scevgep3230, align 1
  %scevgep3228 = getelementptr i8, ptr %scevgep3227, i64 %residual-loop-index2318
  store i8 %109, ptr %scevgep3228, align 1
  %110 = add i64 %residual-loop-index2318, 1
  %111 = icmp uge i64 %110, %102
  br i1 %111, label %Flow3870, label %loop-memcpy-residual2316, !amdgpu.uniform !42

Flow3870:                                         ; preds = %loop-memcpy-residual2316
  br label %Flow3871, !amdgpu.uniform !42

_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i: ; preds = %Flow3874
  %conv.i.i26.i.i.i.i = xor i32 %port.sroa.37.0.i.i, 1
  fence release
  %112 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %arrayidx.i24.i29.i.i.i.i = getelementptr inbounds nuw i32, ptr %112, i64 %idxprom.i23.i.i.i.i.i
  store atomic i32 %conv.i.i26.i.i.i.i, ptr %arrayidx.i24.i29.i.i.i.i monotonic, align 4
  %add.i.i.i.i = add i64 %idx.0.i.i.i.i, 64
  %scevgep3220 = getelementptr i8, ptr addrspace(4) %lsr.iv3219, i64 64
  %lsr.iv.next3225 = add i64 %lsr.iv3224, -64
  br label %Flow3877, !amdgpu.uniform !42

cond.false.i.i.i:                                 ; preds = %Flow3877
  %.lcssa4194 = phi i64 [ %99, %Flow3877 ]
  %port.sroa.37.0.i.i.lcssa = phi i32 [ %port.sroa.37.0.i.i, %Flow3877 ]
  %.lcssa4193 = phi i32 [ %85, %Flow3877 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa4194)
  %113 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %.lcssa4193)
  %114 = icmp ne i32 %113, %port.sroa.37.0.i.i.lcssa
  %115 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %114)
  %116 = extractvalue { i1, i64 } %115, 0
  %117 = extractvalue { i1, i64 } %115, 1
  br i1 %116, label %while.body.i.i.i.i.preheader, label %Flow3869

while.body.i.i.i.i.preheader:                     ; preds = %cond.false.i.i.i
  br label %while.body.i.i.i.i, !amdgpu.uniform !42

Flow3869:                                         ; preds = %Flow3868, %cond.false.i.i.i
  call void @llvm.amdgcn.end.cf.i64(i64 %117)
  br label %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit, !amdgpu.uniform !42

while.body.i.i.i.i:                               ; preds = %while.body.i.i.i.i.preheader, %while.body.i.i.i.i
  %phi.broken3893 = phi i64 [ 0, %while.body.i.i.i.i.preheader ], [ %121, %while.body.i.i.i.i ]
  tail call void @llvm.amdgcn.s.sleep(i32 2)
  %118 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i8.i.i = getelementptr inbounds nuw i32, ptr %118, i64 %idxprom.i23.i.i.i.i.i, !amdgpu.uniform !42
  %119 = load atomic i32, ptr %arrayidx.i.i.i8.i.i monotonic, align 4
  %120 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %119)
  %cmp.i.not.i.i.i.i = icmp eq i32 %120, %port.sroa.37.0.i.i.lcssa
  %121 = call i64 @llvm.amdgcn.if.break.i64(i1 %cmp.i.not.i.i.i.i, i64 %phi.broken3893)
  %122 = call i1 @llvm.amdgcn.loop.i64(i64 %121)
  br i1 %122, label %Flow3868, label %while.body.i.i.i.i

Flow3868:                                         ; preds = %while.body.i.i.i.i
  %.lcssa4192 = phi i64 [ %121, %while.body.i.i.i.i ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa4192)
  br label %Flow3869, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit: ; preds = %Flow3869
  fence acquire
  tail call void @llvm.amdgcn.wave.barrier()
  fence syncscope("agent") release
  %123 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %.lcssa4199, i1 true)
  %iszero.i.i.i.i.i13.i.i = icmp ne i64 %.lcssa4199, 0
  %cmp2.i.i.i.i14.i.i = icmp eq i64 %123, %sh_prom.i.i.i.i
  %cmp.i.i.i.i15.i.i = select i1 %iszero.i.i.i.i.i13.i.i, i1 %cmp2.i.i.i.i14.i.i, i1 false
  %124 = and i32 %.lcssa4200, 31
  %conv4.i.i.i19.i.i = zext i1 %cmp.i.i.i.i15.i.i to i32
  %shl.i.i.i20.i.i = shl nuw i32 %conv4.i.i.i19.i.i, %124
  %xor.i.i.i21.i.i = xor i32 %shl.i.i.i20.i.i, -1
  %125 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2431, !amdgpu.uniform !42

126:                                              ; preds = %ComputeEnd2432
  %div8.i.i.i16.i.i = lshr i32 %.lcssa4200, 5
  %idxprom.i.i.i17.i.i = zext nneg i32 %div8.i.i.i16.i.i to i64
  %127 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i17.i.i
  %sunkaddr3273 = getelementptr inbounds i8, ptr addrspace(1) %127, i64 40
  %128 = atomicrmw and ptr addrspace(1) %sunkaddr3273, i32 %.lcssa4191 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %129, !amdgpu.uniform !42

129:                                              ; preds = %126, %ComputeEnd2432
  call void @llvm.amdgcn.end.cf.i64(i64 %2975)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %for.cond.i.i1.i13, !amdgpu.uniform !42

for.cond.i.i1.i13:                                ; preds = %129, %Flow3867
  %130 = phi i32 [ %168, %Flow3867 ], [ poison, %129 ]
  %131 = phi i64 [ %169, %Flow3867 ], [ poison, %129 ]
  %132 = phi i32 [ %170, %Flow3867 ], [ poison, %129 ]
  %retval.sroa.6.0.i.i.i14 = phi i32 [ %173, %Flow3867 ], [ undef, %129 ]
  %retval.sroa.2.0.i.i.i16 = phi i64 [ %172, %Flow3867 ], [ undef, %129 ]
  %retval.sroa.8.0.i.i.i17 = phi i32 [ %171, %Flow3867 ], [ undef, %129 ]
  %index.0.i.i.i18 = phi i32 [ %175, %Flow3867 ], [ 0, %129 ]
  %133 = load i32, ptr addrspace(1) @__llvm_rpc_client, align 8, !tbaa !68
  %cmp.not.i.i.i19 = icmp ult i32 %index.0.i.i.i18, %133
  %spec.store.select.i.i.i20 = select i1 %cmp.not.i.i.i19, i32 %index.0.i.i.i18, i32 0
  %134 = tail call noundef i64 @llvm.amdgcn.ballot.i64(i1 true)
  %135 = and i32 %spec.store.select.i.i.i20, 31
  %136 = lshr i64 %134, %sh_prom.i.i.i.i
  %137 = trunc i64 %136 to i32
  %conv4.i.i.i.i.i24 = and i32 %137, 1
  %shl.i.i.i.i.i25 = shl nuw i32 %conv4.i.i.i.i.i24, %135
  %138 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2435, !amdgpu.uniform !42

139:                                              ; preds = %ComputeEnd2436
  %sunkaddr3274 = getelementptr inbounds i8, ptr addrspace(1) %2985, i64 40
  %140 = atomicrmw or ptr addrspace(1) %sunkaddr3274, i32 %.lcssa4185 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %141, !amdgpu.uniform !42

141:                                              ; preds = %139, %ComputeEnd2436
  %142 = phi i32 [ %140, %139 ], [ poison, %ComputeEnd2436 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %2994)
  %143 = call i32 @llvm.amdgcn.readfirstlane.i32(i32 %142)
  %144 = or i32 %143, %.lcssa4186
  %shl5.i.i.i.i.i26 = shl nuw i32 1, %135
  %and.i.i.i.i.i27 = and i32 %shl5.i.i.i.i.i26, %144
  %tobool3.i.i.i.i28 = icmp ne i32 %and.i.i.i.i.i27, 0
  %145 = tail call i64 @llvm.amdgcn.ballot.i64(i1 %tobool3.i.i.i.i28)
  %and.i.i.i.i.i.i29 = and i64 %145, %134
  %cmp.i.not.i.i.i30 = icmp ne i64 %134, %and.i.i.i.i.i.i29
  br i1 %cmp.i.not.i.i.i30, label %if.then.i.i.i.i31, label %Flow3866, !amdgpu.uniform !42

if.then.i.i.i.i31:                                ; preds = %141
  fence syncscope("agent") acquire
  %146 = load <2 x i64>, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !75
  %147 = extractelement <2 x i64> %146, i32 0
  %148 = inttoptr i64 %147 to ptr
  %149 = extractelement <2 x i64> %146, i32 1
  %150 = inttoptr i64 %149 to ptr
  %idxprom.i.i.i.i32 = zext i32 %spec.store.select.i.i.i20 to i64
  %arrayidx.i.i.i.i33 = getelementptr inbounds nuw i32, ptr %148, i64 %idxprom.i.i.i.i32, !amdgpu.uniform !42
  %151 = load atomic i32, ptr %arrayidx.i.i.i.i33 monotonic, align 4
  %152 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %151)
  %arrayidx.i45.i.i.i34 = getelementptr inbounds nuw i32, ptr %150, i64 %idxprom.i.i.i.i32, !amdgpu.uniform !42
  %153 = load atomic i32, ptr %arrayidx.i45.i.i.i34 monotonic, align 4
  %154 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %153)
  %cmp.i46.not.i.i.i35 = icmp ne i32 %152, %154
  br i1 %cmp.i46.not.i.i.i35, label %if.then12.i.i.i36, label %Flow3864, !amdgpu.uniform !42

if.then12.i.i.i36:                                ; preds = %if.then.i.i.i.i31
  fence syncscope("agent") release
  %155 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %134, i1 true)
  %iszero.i.i.i.i.i.i.i37 = icmp ne i64 %134, 0
  %cmp2.i.i.i.i.i.i38 = icmp eq i64 %155, %sh_prom.i.i.i.i
  %cmp.i.i.i.i.i.i39 = select i1 %iszero.i.i.i.i.i.i.i37, i1 %cmp2.i.i.i.i.i.i38, i1 false
  %conv4.i.i51.i.i.i40 = zext i1 %cmp.i.i.i.i.i.i39 to i32
  %shl.i.i52.i.i.i41 = shl nuw i32 %conv4.i.i51.i.i.i40, %135
  %xor.i.i.i.i.i42 = xor i32 %shl.i.i52.i.i.i41, -1
  %156 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2440, !amdgpu.uniform !42

157:                                              ; preds = %ComputeEnd2441
  %sunkaddr3275 = getelementptr inbounds i8, ptr addrspace(1) %2985, i64 40
  %158 = atomicrmw and ptr addrspace(1) %sunkaddr3275, i32 %.lcssa4187 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %159, !amdgpu.uniform !42

159:                                              ; preds = %157, %ComputeEnd2441
  call void @llvm.amdgcn.end.cf.i64(i64 %3012)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %Flow3864, !amdgpu.uniform !42

if.end14.i.i.i101:                                ; preds = %Flow3864
  %160 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %134, i1 true)
  %iszero.i.i.i.i.i.i102 = icmp ne i64 %134, 0
  %cmp2.i.i.i.i.i103 = icmp eq i64 %160, %sh_prom.i.i.i.i
  %cmp.i.i.i.i.i104 = select i1 %iszero.i.i.i.i.i.i102, i1 %cmp2.i.i.i.i.i103, i1 false
  %161 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.i.i.i.i.i104)
  %162 = extractvalue { i1, i64 } %161, 0
  %163 = extractvalue { i1, i64 } %161, 1
  br i1 %162, label %if.then16.i.i.i106, label %if.end22.i.i.i105

if.then16.i.i.i106:                               ; preds = %if.end14.i.i.i101
  %164 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !76
  %opcode.i.i.i107 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %164, i64 %idxprom.i.i.i.i32, i32 1
  store i32 1660944387, ptr %opcode.i.i.i107, align 8, !tbaa !77
  %arrayidx21.i.i.i108 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %164, i64 %idxprom.i.i.i.i32
  store i64 %134, ptr %arrayidx21.i.i.i108, align 8, !tbaa !80
  br label %if.end22.i.i.i105, !amdgpu.uniform !42

Flow3865:                                         ; preds = %if.end22.i.i.i105, %Flow3864
  %165 = phi i32 [ %152, %if.end22.i.i.i105 ], [ %retval.sroa.8.0.i.i.i17, %Flow3864 ]
  %166 = phi i64 [ %134, %if.end22.i.i.i105 ], [ %retval.sroa.2.0.i.i.i16, %Flow3864 ]
  %167 = phi i32 [ %spec.store.select.i.i.i20, %if.end22.i.i.i105 ], [ %retval.sroa.6.0.i.i.i14, %Flow3864 ]
  br label %cleanup26.i.i.i43, !amdgpu.uniform !42

if.end22.i.i.i105:                                ; preds = %if.then16.i.i.i106, %if.end14.i.i.i101
  call void @llvm.amdgcn.end.cf.i64(i64 %163)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %Flow3865, !amdgpu.uniform !42

Flow3866:                                         ; preds = %cleanup26.i.i.i43, %141
  %168 = phi i32 [ %167, %cleanup26.i.i.i43 ], [ %130, %141 ]
  %169 = phi i64 [ %166, %cleanup26.i.i.i43 ], [ %131, %141 ]
  %170 = phi i32 [ %165, %cleanup26.i.i.i43 ], [ %132, %141 ]
  %171 = phi i32 [ %165, %cleanup26.i.i.i43 ], [ %retval.sroa.8.0.i.i.i17, %141 ]
  %172 = phi i64 [ %166, %cleanup26.i.i.i43 ], [ %retval.sroa.2.0.i.i.i16, %141 ]
  %173 = phi i32 [ %167, %cleanup26.i.i.i43 ], [ %retval.sroa.6.0.i.i.i14, %141 ]
  %174 = phi i1 [ %cmp.i46.not.i.i.i35, %cleanup26.i.i.i43 ], [ true, %141 ]
  br i1 %174, label %for.inc.i.i.i48, label %Flow3867, !amdgpu.uniform !42

cleanup26.i.i.i43:                                ; preds = %Flow3865
  br label %Flow3866, !amdgpu.uniform !42

for.inc.i.i.i48:                                  ; preds = %Flow3866
  %inc.i.i.i53 = add i32 %spec.store.select.i.i.i20, 1
  br label %Flow3867, !amdgpu.uniform !42

Flow3867:                                         ; preds = %for.inc.i.i.i48, %Flow3866
  %175 = phi i32 [ %inc.i.i.i53, %for.inc.i.i.i48 ], [ poison, %Flow3866 ]
  %176 = phi i1 [ false, %for.inc.i.i.i48 ], [ true, %Flow3866 ]
  br i1 %176, label %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i54, label %for.cond.i.i1.i13, !amdgpu.uniform !42

_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i54: ; preds = %Flow3867
  %.lcssa4190 = phi i32 [ %168, %Flow3867 ]
  %.lcssa4189 = phi i64 [ %169, %Flow3867 ]
  %.lcssa4188 = phi i32 [ %170, %Flow3867 ]
  fence acquire
  %177 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %mul.i.i.i.i.i.i55 = shl i32 %.lcssa4190, 6
  %idxprom.i19.i.i.i.i.i56 = zext i32 %mul.i.i.i.i.i.i55 to i64
  %arrayidx.i20.i.i.i.i.i57 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %177, i64 %idxprom.i19.i.i.i.i.i56
  %arrayidx.i22.i.i.i.i.i58 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %arrayidx.i20.i.i.i.i.i57, i64 %sh_prom.i.i.i.i
  store i64 sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.6.10 to ptr), i64 1) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.6.10 to ptr) to i64)), ptr %arrayidx.i22.i.i.i.i.i58, align 8, !tbaa !82
  %spec.select.i.i.i.i.i.i.i59 = tail call i64 @llvm.umin.i64(i64 sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.6.10 to ptr), i64 1) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.6.10 to ptr) to i64)), i64 56)
  %178 = and i64 %spec.select.i.i.i.i.i.i.i59, 15
  %179 = sub i64 %spec.select.i.i.i.i.i.i.i59, %178
  %180 = icmp ne i64 %179, 0
  br i1 %180, label %loop-memcpy-expansion2320.preheader, label %Flow3863, !amdgpu.uniform !42

loop-memcpy-expansion2320.preheader:              ; preds = %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i54
  %181 = add nuw nsw i64 %75, 8
  %scevgep3216 = getelementptr i8, ptr %177, i64 %181
  %182 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i56, 6
  %scevgep3217 = getelementptr i8, ptr %scevgep3216, i64 %182
  br label %loop-memcpy-expansion2320, !amdgpu.uniform !42

Flow3863:                                         ; preds = %Flow3862, %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i54
  br label %loop-memcpy-residual-header2323, !amdgpu.uniform !42

loop-memcpy-expansion2320:                        ; preds = %loop-memcpy-expansion2320.preheader, %loop-memcpy-expansion2320
  %loop-index2321 = phi i64 [ %185, %loop-memcpy-expansion2320 ], [ 0, %loop-memcpy-expansion2320.preheader ]
  %183 = getelementptr i8, ptr addrspace(4) @.str.6.10, i64 %loop-index2321, !amdgpu.uniform !42
  %184 = load <4 x i32>, ptr addrspace(4) %183, align 1
  %scevgep3218 = getelementptr i8, ptr %scevgep3217, i64 %loop-index2321
  store <4 x i32> %184, ptr %scevgep3218, align 1
  %185 = add i64 %loop-index2321, 16
  %186 = icmp uge i64 %185, %179
  br i1 %186, label %Flow3862, label %loop-memcpy-expansion2320, !amdgpu.uniform !42

Flow3861:                                         ; preds = %Flow3859, %Flow3860
  %187 = phi i64 [ %2852, %Flow3859 ], [ %2849, %Flow3860 ]
  br label %post-loop-memcpy-expansion2319, !amdgpu.uniform !42

loop-memcpy-residual2322:                         ; preds = %loop-memcpy-residual2322.preheader, %loop-memcpy-residual2322
  %residual-loop-index2324 = phi i64 [ %189, %loop-memcpy-residual2322 ], [ 0, %loop-memcpy-residual2322.preheader ]
  %scevgep3215 = getelementptr i8, ptr addrspace(4) @.str.6.10, i64 %residual-loop-index2324, !amdgpu.uniform !42
  %188 = load i8, ptr addrspace(4) %scevgep3215, align 1
  %scevgep3214 = getelementptr i8, ptr %scevgep3213, i64 %residual-loop-index2324
  store i8 %188, ptr %scevgep3214, align 1
  %189 = add i64 %residual-loop-index2324, 1
  %190 = icmp uge i64 %189, %178
  br i1 %190, label %Flow3859, label %loop-memcpy-residual2322, !amdgpu.uniform !42

Flow3859:                                         ; preds = %loop-memcpy-residual2322
  br label %Flow3861, !amdgpu.uniform !42

post-loop-memcpy-expansion2319:                   ; preds = %Flow3861
  %tobool.not.i.i.i.i.i.i61 = icmp eq i32 %.lcssa4188, 0
  %conv.i.i.i.i.i.i62 = zext i1 %tobool.not.i.i.i.i.i.i61 to i32
  fence release
  %191 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %idxprom.i23.i.i.i.i.i63 = zext i32 %.lcssa4190 to i64
  %arrayidx.i24.i.i.i.i.i64 = getelementptr inbounds nuw i32, ptr %191, i64 %idxprom.i23.i.i.i.i.i63
  store atomic i32 %conv.i.i.i.i.i.i62, ptr %arrayidx.i24.i.i.i.i.i64 monotonic, align 4
  %192 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !84
  %arrayidx.i.i5.i.i65 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %192, i64 %idxprom.i23.i.i.i.i.i63, !amdgpu.uniform !42
  %193 = load i64, ptr %arrayidx.i.i5.i.i65, align 8, !tbaa !80
  br label %while.cond.i.i.i.i66, !amdgpu.uniform !42

while.cond.i.i.i.i66:                             ; preds = %post-loop-memcpy-expansion2319, %Flow3858
  %phi.broken3894 = phi i64 [ 0, %post-loop-memcpy-expansion2319 ], [ %210, %Flow3858 ]
  %lsr.iv3205 = phi i64 [ %208, %Flow3858 ], [ -55, %post-loop-memcpy-expansion2319 ]
  %lsr.iv3200 = phi ptr addrspace(4) [ %207, %Flow3858 ], [ getelementptr (i8, ptr addrspace(4) @.str.6.10, i64 56), %post-loop-memcpy-expansion2319 ]
  %port.sroa.37.0.i.i67 = phi i32 [ %206, %Flow3858 ], [ %conv.i.i.i.i.i.i62, %post-loop-memcpy-expansion2319 ]
  %idx.0.i.i.i.i68 = phi i64 [ %205, %Flow3858 ], [ 56, %post-loop-memcpy-expansion2319 ]
  %cmp.i.i.i.i69 = icmp ult i64 %idx.0.i.i.i.i68, sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.6.10 to ptr), i64 1) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.6.10 to ptr) to i64))
  %194 = tail call i64 @llvm.amdgcn.ballot.i64(i1 %cmp.i.i.i.i69)
  %and.i.i.i.i6.i.i70 = and i64 %194, %193
  %tobool.not.i.i.i.i71 = icmp ne i64 %and.i.i.i.i6.i.i70, 0
  %195 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i22.i.i.i72 = getelementptr inbounds nuw i32, ptr %195, i64 %idxprom.i23.i.i.i.i.i63, !amdgpu.uniform !42
  %196 = load atomic i32, ptr %arrayidx.i22.i.i.i72 monotonic, align 4
  %197 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %tobool.not.i.i.i.i71)
  %198 = extractvalue { i1, i64 } %197, 0
  %199 = extractvalue { i1, i64 } %197, 1
  br i1 %198, label %cond.false.i8.i.i.i.i73, label %Flow3858

cond.false.i8.i.i.i.i73:                          ; preds = %while.cond.i.i.i.i66
  %200 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %196)
  %201 = icmp ne i32 %200, %port.sroa.37.0.i.i67
  br i1 %201, label %while.body.i.i21.i.i.i.i74.preheader, label %Flow3857, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i74.preheader:             ; preds = %cond.false.i8.i.i.i.i73
  br label %while.body.i.i21.i.i.i.i74, !amdgpu.uniform !42

Flow3857:                                         ; preds = %Flow3856, %cond.false.i8.i.i.i.i73
  br label %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i77, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i74:                       ; preds = %while.body.i.i21.i.i.i.i74.preheader, %while.body.i.i21.i.i.i.i74
  tail call void @llvm.amdgcn.s.sleep(i32 2)
  %202 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i22.i.i.i.i75 = getelementptr inbounds nuw i32, ptr %202, i64 %idxprom.i23.i.i.i.i.i63, !amdgpu.uniform !42
  %203 = load atomic i32, ptr %arrayidx.i.i.i22.i.i.i.i75 monotonic, align 4
  %204 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %203)
  %cmp.i.not.i.i23.i.i.i.i76 = icmp eq i32 %204, %port.sroa.37.0.i.i67
  br i1 %cmp.i.not.i.i23.i.i.i.i76, label %Flow3856, label %while.body.i.i21.i.i.i.i74, !amdgpu.uniform !42

Flow3856:                                         ; preds = %while.body.i.i21.i.i.i.i74
  br label %Flow3857, !amdgpu.uniform !42

Flow3858:                                         ; preds = %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i78, %while.cond.i.i.i.i66
  %205 = phi i64 [ %add.i.i.i.i81, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i78 ], [ poison, %while.cond.i.i.i.i66 ]
  %206 = phi i32 [ %conv.i.i26.i.i.i.i79, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i78 ], [ poison, %while.cond.i.i.i.i66 ]
  %207 = phi ptr addrspace(4) [ %scevgep3201, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i78 ], [ poison, %while.cond.i.i.i.i66 ]
  %208 = phi i64 [ %lsr.iv.next3206, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i78 ], [ poison, %while.cond.i.i.i.i66 ]
  %209 = phi i1 [ false, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i78 ], [ true, %while.cond.i.i.i.i66 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %199)
  %210 = call i64 @llvm.amdgcn.if.break.i64(i1 %209, i64 %phi.broken3894)
  %211 = call i1 @llvm.amdgcn.loop.i64(i64 %210)
  br i1 %211, label %cond.false.i.i.i88, label %while.cond.i.i.i.i66

_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i77: ; preds = %Flow3857
  fence acquire
  br i1 %cmp.i.i.i.i69, label %if.then.i.i.i.i.i.i.i82, label %Flow3855, !amdgpu.uniform !42

if.then.i.i.i.i.i.i.i82:                          ; preds = %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i77
  %sub.i.i.i.i.i.i.i84 = sub nuw i64 sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.6.10 to ptr), i64 1) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.6.10 to ptr) to i64)), %idx.0.i.i.i.i68
  %spec.select.i.i.i36.i.i.i.i85 = tail call i64 @llvm.umin.i64(i64 %sub.i.i.i.i.i.i.i84, i64 64)
  %212 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %213 = and i64 %spec.select.i.i.i36.i.i.i.i85, 15
  %214 = sub i64 %spec.select.i.i.i36.i.i.i.i85, %213
  %215 = icmp ne i64 %214, 0
  br i1 %215, label %loop-memcpy-expansion2326.preheader, label %Flow3854, !amdgpu.uniform !42

loop-memcpy-expansion2326.preheader:              ; preds = %if.then.i.i.i.i.i.i.i82
  %arrayidx.i20.i35.i.i.i.i86 = getelementptr %"struct.rpc::Buffer", ptr %212, i64 %idxprom.i19.i.i.i.i.i56
  %arrayidx.i22.i38.i.i.i.i87 = getelementptr %"struct.rpc::Buffer", ptr %arrayidx.i20.i35.i.i.i.i86, i64 %sh_prom.i.i.i.i
  br label %loop-memcpy-expansion2326, !amdgpu.uniform !42

Flow3854:                                         ; preds = %Flow3853, %if.then.i.i.i.i.i.i.i82
  br label %loop-memcpy-residual-header2329, !amdgpu.uniform !42

loop-memcpy-expansion2326:                        ; preds = %loop-memcpy-expansion2326.preheader, %loop-memcpy-expansion2326
  %loop-index2327 = phi i64 [ %218, %loop-memcpy-expansion2326 ], [ 0, %loop-memcpy-expansion2326.preheader ]
  %scevgep3202 = getelementptr i8, ptr addrspace(4) %lsr.iv3200, i64 %loop-index2327, !amdgpu.uniform !42
  %216 = load <4 x i32>, ptr addrspace(4) %scevgep3202, align 1
  %217 = getelementptr i8, ptr %arrayidx.i22.i38.i.i.i.i87, i64 %loop-index2327
  store <4 x i32> %216, ptr %217, align 1
  %218 = add i64 %loop-index2327, 16
  %219 = icmp uge i64 %218, %214
  br i1 %219, label %Flow3853, label %loop-memcpy-expansion2326, !amdgpu.uniform !42

Flow3852:                                         ; preds = %Flow3851, %loop-memcpy-residual-header2329
  br label %Flow3855, !amdgpu.uniform !42

loop-memcpy-residual2328:                         ; preds = %loop-memcpy-residual2328.preheader, %loop-memcpy-residual2328
  %residual-loop-index2330 = phi i64 [ %221, %loop-memcpy-residual2328 ], [ 0, %loop-memcpy-residual2328.preheader ]
  %scevgep3211 = getelementptr i8, ptr addrspace(4) %scevgep3210, i64 %residual-loop-index2330, !amdgpu.uniform !42
  %220 = load i8, ptr addrspace(4) %scevgep3211, align 1
  %scevgep3209 = getelementptr i8, ptr %scevgep3208, i64 %residual-loop-index2330
  store i8 %220, ptr %scevgep3209, align 1
  %221 = add i64 %residual-loop-index2330, 1
  %222 = icmp uge i64 %221, %213
  br i1 %222, label %Flow3851, label %loop-memcpy-residual2328, !amdgpu.uniform !42

Flow3851:                                         ; preds = %loop-memcpy-residual2328
  br label %Flow3852, !amdgpu.uniform !42

_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i78: ; preds = %Flow3855
  %conv.i.i26.i.i.i.i79 = xor i32 %port.sroa.37.0.i.i67, 1
  fence release
  %223 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %arrayidx.i24.i29.i.i.i.i80 = getelementptr inbounds nuw i32, ptr %223, i64 %idxprom.i23.i.i.i.i.i63
  store atomic i32 %conv.i.i26.i.i.i.i79, ptr %arrayidx.i24.i29.i.i.i.i80 monotonic, align 4
  %add.i.i.i.i81 = add i64 %idx.0.i.i.i.i68, 64
  %scevgep3201 = getelementptr i8, ptr addrspace(4) %lsr.iv3200, i64 64
  %lsr.iv.next3206 = add i64 %lsr.iv3205, -64
  br label %Flow3858, !amdgpu.uniform !42

cond.false.i.i.i88:                               ; preds = %Flow3858
  %.lcssa4184 = phi i64 [ %210, %Flow3858 ]
  %port.sroa.37.0.i.i67.lcssa = phi i32 [ %port.sroa.37.0.i.i67, %Flow3858 ]
  %.lcssa4183 = phi i32 [ %196, %Flow3858 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa4184)
  %224 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %.lcssa4183)
  %225 = icmp ne i32 %224, %port.sroa.37.0.i.i67.lcssa
  %226 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %225)
  %227 = extractvalue { i1, i64 } %226, 0
  %228 = extractvalue { i1, i64 } %226, 1
  br i1 %227, label %while.body.i.i.i.i89.preheader, label %Flow3850

while.body.i.i.i.i89.preheader:                   ; preds = %cond.false.i.i.i88
  br label %while.body.i.i.i.i89, !amdgpu.uniform !42

Flow3850:                                         ; preds = %Flow3849, %cond.false.i.i.i88
  call void @llvm.amdgcn.end.cf.i64(i64 %228)
  br label %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit109, !amdgpu.uniform !42

while.body.i.i.i.i89:                             ; preds = %while.body.i.i.i.i89.preheader, %while.body.i.i.i.i89
  %phi.broken3895 = phi i64 [ 0, %while.body.i.i.i.i89.preheader ], [ %232, %while.body.i.i.i.i89 ]
  tail call void @llvm.amdgcn.s.sleep(i32 2)
  %229 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i8.i.i90 = getelementptr inbounds nuw i32, ptr %229, i64 %idxprom.i23.i.i.i.i.i63, !amdgpu.uniform !42
  %230 = load atomic i32, ptr %arrayidx.i.i.i8.i.i90 monotonic, align 4
  %231 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %230)
  %cmp.i.not.i.i.i.i91 = icmp eq i32 %231, %port.sroa.37.0.i.i67.lcssa
  %232 = call i64 @llvm.amdgcn.if.break.i64(i1 %cmp.i.not.i.i.i.i91, i64 %phi.broken3895)
  %233 = call i1 @llvm.amdgcn.loop.i64(i64 %232)
  br i1 %233, label %Flow3849, label %while.body.i.i.i.i89

Flow3849:                                         ; preds = %while.body.i.i.i.i89
  %.lcssa4182 = phi i64 [ %232, %while.body.i.i.i.i89 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa4182)
  br label %Flow3850, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit109: ; preds = %Flow3850
  fence acquire
  tail call void @llvm.amdgcn.wave.barrier()
  fence syncscope("agent") release
  %234 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %.lcssa4189, i1 true)
  %iszero.i.i.i.i.i13.i.i92 = icmp ne i64 %.lcssa4189, 0
  %cmp2.i.i.i.i14.i.i93 = icmp eq i64 %234, %sh_prom.i.i.i.i
  %cmp.i.i.i.i15.i.i94 = select i1 %iszero.i.i.i.i.i13.i.i92, i1 %cmp2.i.i.i.i14.i.i93, i1 false
  %235 = and i32 %.lcssa4190, 31
  %conv4.i.i.i19.i.i98 = zext i1 %cmp.i.i.i.i15.i.i94 to i32
  %shl.i.i.i20.i.i99 = shl nuw i32 %conv4.i.i.i19.i.i98, %235
  %xor.i.i.i21.i.i100 = xor i32 %shl.i.i.i20.i.i99, -1
  %236 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2444, !amdgpu.uniform !42

237:                                              ; preds = %ComputeEnd2445
  %div8.i.i.i16.i.i95 = lshr i32 %.lcssa4190, 5
  %idxprom.i.i.i17.i.i96 = zext nneg i32 %div8.i.i.i16.i.i95 to i64
  %238 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i17.i.i96
  %sunkaddr3276 = getelementptr inbounds i8, ptr addrspace(1) %238, i64 40
  %239 = atomicrmw and ptr addrspace(1) %sunkaddr3276, i32 %.lcssa4181 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %240, !amdgpu.uniform !42

240:                                              ; preds = %237, %ComputeEnd2445
  call void @llvm.amdgcn.end.cf.i64(i64 %3029)
  tail call void @llvm.amdgcn.wave.barrier()
  call void @llvm.lifetime.start.p5(ptr addrspace(5) %buffer.i.i.i) #34, !noalias !86
  br label %while.body.i.i.i.i.i.i, !amdgpu.uniform !42

while.body.i.i.i.i.i.i:                           ; preds = %240, %_ZN22__llvm_libc_22_0_0_git7details22StringBufferWriterImplILb0EE4pushEc.exit.i.i.i.i.i.i
  %phi.broken3896 = phi i64 [ 0, %240 ], [ %308, %_ZN22__llvm_libc_22_0_0_git7details22StringBufferWriterImplILb0EE4pushEc.exit.i.i.i.i.i.i ]
  %writer.sroa.10.0.i.i.i.i = phi i64 [ %writer.sroa.10.1.i.i.i.i, %_ZN22__llvm_libc_22_0_0_git7details22StringBufferWriterImplILb0EE4pushEc.exit.i.i.i.i.i.i ], [ 0, %240 ]
  %value.addr.09.i.i.i.i.i.i = phi i32 [ %div.i.i.i.i.i.i.i, %_ZN22__llvm_libc_22_0_0_git7details22StringBufferWriterImplILb0EE4pushEc.exit.i.i.i.i.i.i ], [ %Loc.coerce1, %240 ]
  br label %NodeBlock3350, !amdgpu.uniform !42

NodeBlock3350:                                    ; preds = %while.body.i.i.i.i.i.i
  %div.i.i.i.i.i.i.i = udiv i32 %value.addr.09.i.i.i.i.i.i, 10
  %.neg.i = call i32 @llvm.amdgcn.mul.u24.i32(i32 %div.i.i.i.i.i.i.i, i32 246)
  %rem.i.i.i.i.i.i.decomposed.i = add nuw nsw i32 %.neg.i, %value.addr.09.i.i.i.i.i.i
  %conv.i.i.i.i.i.i.i = trunc i32 %rem.i.i.i.i.i.i.decomposed.i to i8
  %241 = zext i8 %conv.i.i.i.i.i.i.i to i16
  %Pivot3351 = icmp sge i16 %241, 5
  %242 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %Pivot3351)
  %243 = extractvalue { i1, i64 } %242, 0
  %244 = extractvalue { i1, i64 } %242, 1
  br i1 %243, label %NodeBlock3348, label %Flow3848

NodeBlock3348:                                    ; preds = %NodeBlock3350
  %Pivot3349 = icmp sge i16 %241, 7
  %245 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %Pivot3349)
  %246 = extractvalue { i1, i64 } %245, 0
  %247 = extractvalue { i1, i64 } %245, 1
  br i1 %246, label %NodeBlock3346, label %Flow3839

NodeBlock3346:                                    ; preds = %NodeBlock3348
  %Pivot3347 = icmp sge i16 %241, 8
  %248 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %Pivot3347)
  %249 = extractvalue { i1, i64 } %248, 0
  %250 = extractvalue { i1, i64 } %248, 1
  br i1 %249, label %NodeBlock3344, label %Flow3835

NodeBlock3344:                                    ; preds = %NodeBlock3346
  %Pivot3345 = icmp sge i16 %241, 9
  %251 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %Pivot3345)
  %252 = extractvalue { i1, i64 } %251, 0
  %253 = extractvalue { i1, i64 } %251, 1
  br i1 %252, label %sw.bb9.i.i.i.i.i.i.i.i, label %Flow3833

Flow3839:                                         ; preds = %Flow3836, %NodeBlock3348
  %254 = phi i8 [ %298, %Flow3836 ], [ poison, %NodeBlock3348 ]
  %255 = call { i1, i64 } @llvm.amdgcn.else.i64.i64(i64 %247)
  %256 = extractvalue { i1, i64 } %255, 0
  %257 = extractvalue { i1, i64 } %255, 1
  br i1 %256, label %NodeBlock3342, label %Flow3840

NodeBlock3342:                                    ; preds = %Flow3839
  %Pivot3343 = icmp sge i16 %241, 6
  %258 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %Pivot3343)
  %259 = extractvalue { i1, i64 } %258, 0
  %260 = extractvalue { i1, i64 } %258, 1
  br i1 %259, label %sw.bb6.i.i.i.i.i.i.i.i, label %Flow3837

Flow3848:                                         ; preds = %Flow3840, %NodeBlock3350
  %261 = phi i8 [ %300, %Flow3840 ], [ poison, %NodeBlock3350 ]
  %262 = call { i1, i64 } @llvm.amdgcn.else.i64.i64(i64 %244)
  %263 = extractvalue { i1, i64 } %262, 0
  %264 = extractvalue { i1, i64 } %262, 1
  br i1 %263, label %NodeBlock3340, label %if.then.i.i.i.i.i.i.i110

NodeBlock3340:                                    ; preds = %Flow3848
  %Pivot3341 = icmp sge i16 %241, 3
  %265 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %Pivot3341)
  %266 = extractvalue { i1, i64 } %265, 0
  %267 = extractvalue { i1, i64 } %265, 1
  br i1 %266, label %NodeBlock3338, label %Flow3846

NodeBlock3338:                                    ; preds = %NodeBlock3340
  %Pivot3339 = icmp sge i16 %241, 4
  %268 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %Pivot3339)
  %269 = extractvalue { i1, i64 } %268, 0
  %270 = extractvalue { i1, i64 } %268, 1
  br i1 %269, label %sw.bb4.i.i.i.i.i.i.i.i, label %Flow3841

Flow3846:                                         ; preds = %Flow3842, %NodeBlock3340
  %271 = phi i8 [ %301, %Flow3842 ], [ poison, %NodeBlock3340 ]
  %272 = call { i1, i64 } @llvm.amdgcn.else.i64.i64(i64 %267)
  %273 = extractvalue { i1, i64 } %272, 0
  %274 = extractvalue { i1, i64 } %272, 1
  br i1 %273, label %NodeBlock, label %Flow3847

NodeBlock:                                        ; preds = %Flow3846
  %Pivot = icmp sge i16 %241, 2
  %275 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %Pivot)
  %276 = extractvalue { i1, i64 } %275, 0
  %277 = extractvalue { i1, i64 } %275, 1
  br i1 %276, label %sw.bb2.i.i.i.i.i.i.i.i, label %Flow3844

Flow3844:                                         ; preds = %sw.bb2.i.i.i.i.i.i.i.i, %NodeBlock
  %278 = call { i1, i64 } @llvm.amdgcn.else.i64.i64(i64 %277)
  %279 = extractvalue { i1, i64 } %278, 0
  %280 = extractvalue { i1, i64 } %278, 1
  br i1 %279, label %LeafBlock, label %Flow3845

LeafBlock:                                        ; preds = %Flow3844
  %SwitchLeaf = icmp eq i16 %241, 1
  %281 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %SwitchLeaf)
  %282 = extractvalue { i1, i64 } %281, 0
  %283 = extractvalue { i1, i64 } %281, 1
  br i1 %282, label %sw.bb1.i.i.i.i.i.i.i.i, label %Flow3843

sw.bb1.i.i.i.i.i.i.i.i:                           ; preds = %LeafBlock
  br label %Flow3843, !amdgpu.uniform !42

sw.bb2.i.i.i.i.i.i.i.i:                           ; preds = %NodeBlock
  br label %Flow3844, !amdgpu.uniform !42

Flow3841:                                         ; preds = %sw.bb4.i.i.i.i.i.i.i.i, %NodeBlock3338
  %284 = call { i1, i64 } @llvm.amdgcn.else.i64.i64(i64 %270)
  %285 = extractvalue { i1, i64 } %284, 0
  %286 = extractvalue { i1, i64 } %284, 1
  br i1 %285, label %sw.bb3.i.i.i.i.i.i.i.i, label %Flow3842

sw.bb3.i.i.i.i.i.i.i.i:                           ; preds = %Flow3841
  br label %Flow3842, !amdgpu.uniform !42

sw.bb4.i.i.i.i.i.i.i.i:                           ; preds = %NodeBlock3338
  br label %Flow3841, !amdgpu.uniform !42

Flow3837:                                         ; preds = %sw.bb6.i.i.i.i.i.i.i.i, %NodeBlock3342
  %287 = call { i1, i64 } @llvm.amdgcn.else.i64.i64(i64 %260)
  %288 = extractvalue { i1, i64 } %287, 0
  %289 = extractvalue { i1, i64 } %287, 1
  br i1 %288, label %sw.bb5.i.i.i.i.i.i.i.i, label %Flow3838

sw.bb5.i.i.i.i.i.i.i.i:                           ; preds = %Flow3837
  br label %Flow3838, !amdgpu.uniform !42

sw.bb6.i.i.i.i.i.i.i.i:                           ; preds = %NodeBlock3342
  br label %Flow3837, !amdgpu.uniform !42

Flow3835:                                         ; preds = %Flow3834, %NodeBlock3346
  %290 = phi i8 [ %297, %Flow3834 ], [ poison, %NodeBlock3346 ]
  %291 = call { i1, i64 } @llvm.amdgcn.else.i64.i64(i64 %250)
  %292 = extractvalue { i1, i64 } %291, 0
  %293 = extractvalue { i1, i64 } %291, 1
  br i1 %292, label %sw.bb7.i.i.i.i.i.i.i.i, label %Flow3836

sw.bb7.i.i.i.i.i.i.i.i:                           ; preds = %Flow3835
  br label %Flow3836, !amdgpu.uniform !42

Flow3833:                                         ; preds = %sw.bb9.i.i.i.i.i.i.i.i, %NodeBlock3344
  %294 = call { i1, i64 } @llvm.amdgcn.else.i64.i64(i64 %253)
  %295 = extractvalue { i1, i64 } %294, 0
  %296 = extractvalue { i1, i64 } %294, 1
  br i1 %295, label %sw.bb8.i.i.i.i.i.i.i.i, label %Flow3834

sw.bb8.i.i.i.i.i.i.i.i:                           ; preds = %Flow3833
  br label %Flow3834, !amdgpu.uniform !42

sw.bb9.i.i.i.i.i.i.i.i:                           ; preds = %NodeBlock3344
  br label %Flow3833, !amdgpu.uniform !42

Flow3834:                                         ; preds = %sw.bb8.i.i.i.i.i.i.i.i, %Flow3833
  %297 = phi i8 [ 56, %sw.bb8.i.i.i.i.i.i.i.i ], [ 57, %Flow3833 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %296)
  br label %Flow3835, !amdgpu.uniform !42

Flow3836:                                         ; preds = %sw.bb7.i.i.i.i.i.i.i.i, %Flow3835
  %298 = phi i8 [ 55, %sw.bb7.i.i.i.i.i.i.i.i ], [ %290, %Flow3835 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %293)
  br label %Flow3839, !amdgpu.uniform !42

Flow3838:                                         ; preds = %sw.bb5.i.i.i.i.i.i.i.i, %Flow3837
  %299 = phi i8 [ 53, %sw.bb5.i.i.i.i.i.i.i.i ], [ 54, %Flow3837 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %289)
  br label %Flow3840, !amdgpu.uniform !42

Flow3840:                                         ; preds = %Flow3838, %Flow3839
  %300 = phi i8 [ %299, %Flow3838 ], [ %254, %Flow3839 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %257)
  br label %Flow3848, !amdgpu.uniform !42

Flow3842:                                         ; preds = %sw.bb3.i.i.i.i.i.i.i.i, %Flow3841
  %301 = phi i8 [ 51, %sw.bb3.i.i.i.i.i.i.i.i ], [ 52, %Flow3841 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %286)
  br label %Flow3846, !amdgpu.uniform !42

Flow3843:                                         ; preds = %sw.bb1.i.i.i.i.i.i.i.i, %LeafBlock
  %302 = phi i8 [ 49, %sw.bb1.i.i.i.i.i.i.i.i ], [ 48, %LeafBlock ]
  call void @llvm.amdgcn.end.cf.i64(i64 %283)
  br label %Flow3845, !amdgpu.uniform !42

Flow3845:                                         ; preds = %Flow3843, %Flow3844
  %303 = phi i8 [ %302, %Flow3843 ], [ 50, %Flow3844 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %280)
  br label %Flow3847, !amdgpu.uniform !42

Flow3847:                                         ; preds = %Flow3845, %Flow3846
  %304 = phi i8 [ %303, %Flow3845 ], [ %271, %Flow3846 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %274)
  br label %if.then.i.i.i.i.i.i.i110, !amdgpu.uniform !42

if.then.i.i.i.i.i.i.i110:                         ; preds = %Flow3848, %Flow3847
  %retval.0.i.i.i.i.i.i.i.i = phi i8 [ %261, %Flow3848 ], [ %304, %Flow3847 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %264)
  %cmp.i.i.i.i.i.i.i.i = icmp eq i64 %writer.sroa.10.0.i.i.i.i, 11
  %cmp.i.i.i.i.i.i.i.i.inv = xor i1 %cmp.i.i.i.i.i.i.i.i, true
  br i1 %cmp.i.i.i.i.i.i.i.i.inv, label %if.then3.i.i.i.i.i.i.i, label %_ZN22__llvm_libc_22_0_0_git7details22StringBufferWriterImplILb0EE4pushEc.exit.i.i.i.i.i.i, !amdgpu.uniform !42

if.then3.i.i.i.i.i.i.i:                           ; preds = %if.then.i.i.i.i.i.i.i110
  %305 = trunc i64 %writer.sroa.10.0.i.i.i.i to i32
  %306 = xor i32 %305, -1
  %307 = getelementptr i8, ptr addrspace(5) %buffer.i.i.i, i32 %306
  %arrayidx.i.i.i.i.i.i.i.i1112699 = getelementptr i8, ptr addrspace(5) %307, i32 11
  %inc.i.i.i.i.i.i.i = add i64 %writer.sroa.10.0.i.i.i.i, 1
  store i8 %retval.0.i.i.i.i.i.i.i.i, ptr addrspace(5) %arrayidx.i.i.i.i.i.i.i.i1112699, align 1, !tbaa !57, !noalias !86
  br label %_ZN22__llvm_libc_22_0_0_git7details22StringBufferWriterImplILb0EE4pushEc.exit.i.i.i.i.i.i, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git7details22StringBufferWriterImplILb0EE4pushEc.exit.i.i.i.i.i.i: ; preds = %if.then3.i.i.i.i.i.i.i, %if.then.i.i.i.i.i.i.i110
  %writer.sroa.10.1.i.i.i.i = phi i64 [ %inc.i.i.i.i.i.i.i, %if.then3.i.i.i.i.i.i.i ], [ 11, %if.then.i.i.i.i.i.i.i110 ]
  %cmp.i.i.i.i.i.i112 = icmp samesign ult i32 %value.addr.09.i.i.i.i.i.i, 10
  %.not.i.i.i.i.i.i = or i1 %cmp.i.i.i.i.i.i.i.i, %cmp.i.i.i.i.i.i112
  %308 = call i64 @llvm.amdgcn.if.break.i64(i1 %.not.i.i.i.i.i.i, i64 %phi.broken3896)
  %309 = call i1 @llvm.amdgcn.loop.i64(i64 %308)
  br i1 %309, label %_ZN22__llvm_libc_22_0_0_git15IntegerToStringIiNS_7details3FmtILh10ELb0ELb0ELb0ELm1EEEE13IntegerWriter25write_unsigned_number_decEjRNS1_22StringBufferWriterImplILb0EEE.exit.i.i.i.i.i, label %while.body.i.i.i.i.i.i

_ZN22__llvm_libc_22_0_0_git15IntegerToStringIiNS_7details3FmtILh10ELb0ELb0ELb0ELm1EEEE13IntegerWriter25write_unsigned_number_decEjRNS1_22StringBufferWriterImplILb0EEE.exit.i.i.i.i.i: ; preds = %_ZN22__llvm_libc_22_0_0_git7details22StringBufferWriterImplILb0EE4pushEc.exit.i.i.i.i.i.i
  %writer.sroa.10.1.i.i.i.i.lcssa = phi i64 [ %writer.sroa.10.1.i.i.i.i, %_ZN22__llvm_libc_22_0_0_git7details22StringBufferWriterImplILb0EE4pushEc.exit.i.i.i.i.i.i ]
  %.lcssa4180 = phi i64 [ %308, %_ZN22__llvm_libc_22_0_0_git7details22StringBufferWriterImplILb0EE4pushEc.exit.i.i.i.i.i.i ]
  %cmp.i.i.i.i.i.i.i.i.lcssa = phi i1 [ %cmp.i.i.i.i.i.i.i.i, %_ZN22__llvm_libc_22_0_0_git7details22StringBufferWriterImplILb0EE4pushEc.exit.i.i.i.i.i.i ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa4180)
  %cmp.i19.i.i.i.i = icmp ne i64 %writer.sroa.10.1.i.i.i.i.lcssa, 0
  %or.cond.not.i.i.i.i = select i1 %cmp.i.i.i.i.i.i.i.i.lcssa, i1 true, i1 %cmp.i19.i.i.i.i
  %or.cond.not.i.i.i.i.inv = xor i1 %or.cond.not.i.i.i.i, true
  %310 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %or.cond.not.i.i.i.i.inv)
  %311 = extractvalue { i1, i64 } %310, 0
  %312 = extractvalue { i1, i64 } %310, 1
  br i1 %311, label %if.then3.i.i.lr.ph.i.i.i.i, label %while.end.i.i.i.i.i

if.then3.i.i.lr.ph.i.i.i.i:                       ; preds = %_ZN22__llvm_libc_22_0_0_git15IntegerToStringIiNS_7details3FmtILh10ELb0ELb0ELb0ELm1EEEE13IntegerWriter25write_unsigned_number_decEjRNS1_22StringBufferWriterImplILb0EEE.exit.i.i.i.i.i
  %arrayidx.i.i.i.i.i.i.i = getelementptr inbounds nuw i8, ptr addrspace(5) %buffer.i.i.i, i32 10
  store i8 48, ptr addrspace(5) %arrayidx.i.i.i.i.i.i.i, align 2, !tbaa !57, !noalias !86
  br label %while.end.i.i.i.i.i, !amdgpu.uniform !42

while.end.i.i.i.i.i:                              ; preds = %if.then3.i.i.lr.ph.i.i.i.i, %_ZN22__llvm_libc_22_0_0_git15IntegerToStringIiNS_7details3FmtILh10ELb0ELb0ELb0ELm1EEEE13IntegerWriter25write_unsigned_number_decEjRNS1_22StringBufferWriterImplILb0EEE.exit.i.i.i.i.i
  %writer.sroa.10.4.i.i.i.i = phi i64 [ 1, %if.then3.i.i.lr.ph.i.i.i.i ], [ %writer.sroa.10.1.i.i.i.i.lcssa, %_ZN22__llvm_libc_22_0_0_git15IntegerToStringIiNS_7details3FmtILh10ELb0ELb0ELb0ELm1EEEE13IntegerWriter25write_unsigned_number_decEjRNS1_22StringBufferWriterImplILb0EEE.exit.i.i.i.i.i ]
  call void @llvm.amdgcn.end.cf.i64(i64 %312)
  %written.i.i.i.i = getelementptr inbounds nuw i8, ptr addrspace(5) %buffer.i.i.i, i32 16
  store i64 %writer.sroa.10.4.i.i.i.i, ptr addrspace(5) %written.i.i.i.i, align 8, !tbaa !91, !noalias !86
  %cmp.i.not.i.i.i.i.i = icmp ne i64 %writer.sroa.10.4.i.i.i.i, 0
  %313 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.i.not.i.i.i.i.i)
  %314 = extractvalue { i1, i64 } %313, 0
  %315 = extractvalue { i1, i64 } %313, 1
  br i1 %314, label %if.then.i.i.i.i.i.i, label %Flow3832

if.then.i.i.i.i.i.i:                              ; preds = %while.end.i.i.i.i.i
  %316 = call { i64, i1 } @llvm.uadd.with.overflow.i64(i64 %writer.sroa.10.4.i.i.i.i, i64 1)
  %ov = extractvalue { i64, i1 } %316, 1
  %ov.inv = xor i1 %ov, true
  %317 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %ov.inv)
  %318 = extractvalue { i1, i64 } %317, 0
  %319 = extractvalue { i1, i64 } %317, 1
  br i1 %318, label %if.end.i.i.i.i.i.i.i113, label %Flow3831

if.end.i.i.i.i.i.i.i113:                          ; preds = %if.then.i.i.i.i.i.i
  %math = extractvalue { i64, i1 } %316, 0
  %cmp2.i.i.i.i.i.i.i = icmp ult i64 %math, 1676976733973595601
  %320 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp2.i.i.i.i.i.i.i)
  %321 = extractvalue { i1, i64 } %320, 0
  %322 = extractvalue { i1, i64 } %320, 1
  br i1 %321, label %if.then3.i.i.i.i4.i.i.i, label %if.end4.i.i.i.i.i.i.i

if.then3.i.i.i.i4.i.i.i:                          ; preds = %if.end.i.i.i.i.i.i.i113
  %mul.i.i.i.i.i.i.i = mul nuw i64 %math, 11
  %div18.i.i.i.i.i.i.i = lshr i64 %mul.i.i.i.i.i.i.i, 3
  br label %if.end4.i.i.i.i.i.i.i, !amdgpu.uniform !42

Flow3831:                                         ; preds = %if.end4.i.i.i.i.i.i.i, %if.then.i.i.i.i.i.i
  %323 = phi ptr [ %cond.i.i.i, %if.end4.i.i.i.i.i.i.i ], [ addrspacecast (ptr addrspace(4) @_ZN22__llvm_libc_22_0_0_git3cpp6string14NULL_CHARACTERE to ptr), %if.then.i.i.i.i.i.i ]
  call void @llvm.amdgcn.end.cf.i64(i64 %319)
  br label %_ZN22__llvm_libc_22_0_0_git3cpp6string7reserveEm.exit.i.i.i.i.i.i, !amdgpu.uniform !42

if.end4.i.i.i.i.i.i.i:                            ; preds = %if.then3.i.i.i.i4.i.i.i, %if.end.i.i.i.i.i.i.i113
  %new_capacity.addr.0.i.i.i.i.i.i.i = phi i64 [ %div18.i.i.i.i.i.i.i, %if.then3.i.i.i.i4.i.i.i ], [ %math, %if.end.i.i.i.i.i.i.i113 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %322)
  %sub.i.i.i = add i64 %new_capacity.addr.0.i.i.i.i.i.i.i, 7
  %div2.i.i.i = and i64 %sub.i.i.i, -8
  %324 = load ptr, ptr addrspace(1) @_ZN12_GLOBAL__N_13ptrE, align 8, !tbaa !55
  %add.ptr.i.i.i = getelementptr inbounds nuw i8, ptr %324, i64 %div2.i.i.i
  store ptr %add.ptr.i.i.i, ptr addrspace(1) @_ZN12_GLOBAL__N_13ptrE, align 8, !tbaa !55
  %sub.ptr.lhs.cast.i.i.i114 = ptrtoint ptr %add.ptr.i.i.i to i64
  %sub.ptr.sub.i.i.i115 = sub i64 %sub.ptr.lhs.cast.i.i.i114, ptrtoint (ptr addrspacecast (ptr addrspace(1) @_ZN12_GLOBAL__N_16memoryE to ptr) to i64)
  %cmp.i.i.i116 = icmp ugt i64 %sub.ptr.sub.i.i.i115, 65335
  %cond.i.i.i = select i1 %cmp.i.i.i116, ptr null, ptr %324
  br label %Flow3831, !amdgpu.uniform !42

Flow3832:                                         ; preds = %Flow3830, %while.end.i.i.i.i.i
  %325 = phi ptr [ %323, %Flow3830 ], [ addrspacecast (ptr addrspace(4) @_ZN22__llvm_libc_22_0_0_git3cpp6string14NULL_CHARACTERE to ptr), %while.end.i.i.i.i.i ]
  call void @llvm.amdgcn.end.cf.i64(i64 %315)
  br label %if.end.i.i.i.i.i.i, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git3cpp6string7reserveEm.exit.i.i.i.i.i.i: ; preds = %Flow3831
  br label %loadstoreloop, !amdgpu.uniform !42

loadstoreloop:                                    ; preds = %_ZN22__llvm_libc_22_0_0_git3cpp6string7reserveEm.exit.i.i.i.i.i.i, %loadstoreloop
  %phi.broken3897 = phi i64 [ 0, %_ZN22__llvm_libc_22_0_0_git3cpp6string7reserveEm.exit.i.i.i.i.i.i ], [ %330, %loadstoreloop ]
  %326 = phi i64 [ %328, %loadstoreloop ], [ 0, %_ZN22__llvm_libc_22_0_0_git3cpp6string7reserveEm.exit.i.i.i.i.i.i ]
  %327 = getelementptr i8, ptr %323, i64 %326
  store i8 0, ptr %327, align 1
  %328 = add i64 %326, 1
  %329 = icmp uge i64 %328, %writer.sroa.10.4.i.i.i.i
  %330 = call i64 @llvm.amdgcn.if.break.i64(i1 %329, i64 %phi.broken3897)
  %331 = call i1 @llvm.amdgcn.loop.i64(i64 %330)
  br i1 %331, label %Flow3830, label %loadstoreloop

Flow3830:                                         ; preds = %loadstoreloop
  %.lcssa4179 = phi i64 [ %330, %loadstoreloop ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa4179)
  br label %Flow3832, !amdgpu.uniform !42

if.end.i.i.i.i.i.i:                               ; preds = %Flow3832
  %cmp.not.i6.i.i.i.i.i.i = icmp ne ptr %325, addrspacecast (ptr addrspace(4) @_ZN22__llvm_libc_22_0_0_git3cpp6string14NULL_CHARACTERE to ptr)
  %332 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.not.i6.i.i.i.i.i.i)
  %333 = extractvalue { i1, i64 } %332, 0
  %334 = extractvalue { i1, i64 } %332, 1
  br i1 %333, label %if.then.i.i.i.i2.i.i.i, label %_ZN22__llvm_libc_22_0_0_git3cpp9to_stringEi.exit.i

if.then.i.i.i.i2.i.i.i:                           ; preds = %if.end.i.i.i.i.i.i
  %arrayidx.i.i.i.i3.i.i.i = getelementptr inbounds nuw i8, ptr %325, i64 %writer.sroa.10.4.i.i.i.i
  store i8 0, ptr %arrayidx.i.i.i.i3.i.i.i, align 1, !tbaa !57, !noalias !86
  br label %_ZN22__llvm_libc_22_0_0_git3cpp9to_stringEi.exit.i, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git3cpp9to_stringEi.exit.i: ; preds = %if.then.i.i.i.i2.i.i.i, %if.end.i.i.i.i.i.i
  call void @llvm.amdgcn.end.cf.i64(i64 %334)
  %335 = trunc i64 %writer.sroa.10.4.i.i.i.i to i32
  %336 = and i64 %writer.sroa.10.4.i.i.i.i, 15
  %337 = sub i64 %writer.sroa.10.4.i.i.i.i, %336
  %338 = icmp ne i64 %337, 0
  %339 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %338)
  %340 = extractvalue { i1, i64 } %339, 0
  %341 = extractvalue { i1, i64 } %339, 1
  br i1 %340, label %loop-memcpy-expansion2290.preheader, label %Flow3829

loop-memcpy-expansion2290.preheader:              ; preds = %_ZN22__llvm_libc_22_0_0_git3cpp9to_stringEi.exit.i
  %342 = sub i32 11, %335
  %scevgep3197 = getelementptr i8, ptr addrspace(5) %buffer.i.i.i, i32 %342
  br label %loop-memcpy-expansion2290, !amdgpu.uniform !42

Flow3829:                                         ; preds = %Flow3828, %_ZN22__llvm_libc_22_0_0_git3cpp9to_stringEi.exit.i
  call void @llvm.amdgcn.end.cf.i64(i64 %341)
  br label %loop-memcpy-residual-header2293, !amdgpu.uniform !42

loop-memcpy-expansion2290:                        ; preds = %loop-memcpy-expansion2290.preheader, %loop-memcpy-expansion2290
  %phi.broken3898 = phi i64 [ 0, %loop-memcpy-expansion2290.preheader ], [ %347, %loop-memcpy-expansion2290 ]
  %lsr.iv3198 = phi ptr addrspace(5) [ %scevgep3199, %loop-memcpy-expansion2290 ], [ %scevgep3197, %loop-memcpy-expansion2290.preheader ]
  %loop-index2291 = phi i64 [ %345, %loop-memcpy-expansion2290 ], [ 0, %loop-memcpy-expansion2290.preheader ]
  %343 = load <4 x i32>, ptr addrspace(5) %lsr.iv3198, align 1
  %344 = getelementptr i8, ptr %325, i64 %loop-index2291
  store <4 x i32> %343, ptr %344, align 1
  %345 = add i64 %loop-index2291, 16
  %scevgep3199 = getelementptr i8, ptr addrspace(5) %lsr.iv3198, i32 16
  %346 = icmp uge i64 %345, %337
  %347 = call i64 @llvm.amdgcn.if.break.i64(i1 %346, i64 %phi.broken3898)
  %348 = call i1 @llvm.amdgcn.loop.i64(i64 %347)
  br i1 %348, label %Flow3828, label %loop-memcpy-expansion2290

Flow3827:                                         ; preds = %Flow3826, %loop-memcpy-residual-header2293
  call void @llvm.amdgcn.end.cf.i64(i64 %2811)
  br label %post-loop-memcpy-expansion2289, !amdgpu.uniform !42

loop-memcpy-residual2292:                         ; preds = %loop-memcpy-residual2292.preheader, %loop-memcpy-residual2292
  %phi.broken3899 = phi i64 [ 0, %loop-memcpy-residual2292.preheader ], [ %352, %loop-memcpy-residual2292 ]
  %lsr.iv3193 = phi ptr addrspace(5) [ %scevgep3194, %loop-memcpy-residual2292 ], [ %scevgep3192, %loop-memcpy-residual2292.preheader ]
  %residual-loop-index2294 = phi i64 [ %350, %loop-memcpy-residual2292 ], [ 0, %loop-memcpy-residual2292.preheader ]
  %349 = load i8, ptr addrspace(5) %lsr.iv3193, align 1
  %scevgep3196 = getelementptr i8, ptr %scevgep3195, i64 %residual-loop-index2294
  store i8 %349, ptr %scevgep3196, align 1
  %350 = add i64 %residual-loop-index2294, 1
  %scevgep3194 = getelementptr i8, ptr addrspace(5) %lsr.iv3193, i32 1
  %351 = icmp uge i64 %350, %336
  %352 = call i64 @llvm.amdgcn.if.break.i64(i1 %351, i64 %phi.broken3899)
  %353 = call i1 @llvm.amdgcn.loop.i64(i64 %352)
  br i1 %353, label %Flow3826, label %loop-memcpy-residual2292

Flow3826:                                         ; preds = %loop-memcpy-residual2292
  %.lcssa4177 = phi i64 [ %352, %loop-memcpy-residual2292 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa4177)
  br label %Flow3827, !amdgpu.uniform !42

post-loop-memcpy-expansion2289:                   ; preds = %Flow3827
  call void @llvm.lifetime.end.p5(ptr addrspace(5) %buffer.i.i.i) #34, !noalias !86
  br label %for.cond.i.i.i118, !amdgpu.uniform !42

for.cond.i.i.i118:                                ; preds = %post-loop-memcpy-expansion2289, %Flow3825
  %354 = phi i32 [ %392, %Flow3825 ], [ poison, %post-loop-memcpy-expansion2289 ]
  %355 = phi i64 [ %393, %Flow3825 ], [ poison, %post-loop-memcpy-expansion2289 ]
  %356 = phi i32 [ %394, %Flow3825 ], [ poison, %post-loop-memcpy-expansion2289 ]
  %retval.sroa.6.0.i.i.i119 = phi i32 [ %397, %Flow3825 ], [ undef, %post-loop-memcpy-expansion2289 ]
  %retval.sroa.2.0.i.i.i121 = phi i64 [ %396, %Flow3825 ], [ undef, %post-loop-memcpy-expansion2289 ]
  %retval.sroa.8.0.i.i.i122 = phi i32 [ %395, %Flow3825 ], [ undef, %post-loop-memcpy-expansion2289 ]
  %index.0.i.i.i123 = phi i32 [ %399, %Flow3825 ], [ 0, %post-loop-memcpy-expansion2289 ]
  %357 = load i32, ptr addrspace(1) @__llvm_rpc_client, align 8, !tbaa !68
  %cmp.not.i.i.i124 = icmp ult i32 %index.0.i.i.i123, %357
  %spec.store.select.i.i.i125 = select i1 %cmp.not.i.i.i124, i32 %index.0.i.i.i123, i32 0
  %358 = tail call noundef i64 @llvm.amdgcn.ballot.i64(i1 true)
  %359 = and i32 %spec.store.select.i.i.i125, 31
  %360 = lshr i64 %358, %sh_prom.i.i.i.i
  %361 = trunc i64 %360 to i32
  %conv4.i.i.i.i.i129 = and i32 %361, 1
  %shl.i.i.i.i.i130 = shl nuw i32 %conv4.i.i.i.i.i129, %359
  %362 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2448, !amdgpu.uniform !42

363:                                              ; preds = %ComputeEnd2449
  %sunkaddr3277 = getelementptr inbounds i8, ptr addrspace(1) %3039, i64 40
  %364 = atomicrmw or ptr addrspace(1) %sunkaddr3277, i32 %.lcssa4171 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %365, !amdgpu.uniform !42

365:                                              ; preds = %363, %ComputeEnd2449
  %366 = phi i32 [ %364, %363 ], [ poison, %ComputeEnd2449 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %3048)
  %367 = call i32 @llvm.amdgcn.readfirstlane.i32(i32 %366)
  %368 = or i32 %367, %.lcssa4172
  %shl5.i.i.i.i.i131 = shl nuw i32 1, %359
  %and.i.i.i.i.i132 = and i32 %shl5.i.i.i.i.i131, %368
  %tobool3.i.i.i.i133 = icmp ne i32 %and.i.i.i.i.i132, 0
  %369 = tail call i64 @llvm.amdgcn.ballot.i64(i1 %tobool3.i.i.i.i133)
  %and.i.i.i.i.i.i134 = and i64 %369, %358
  %cmp.i.not.i.i.i135 = icmp ne i64 %358, %and.i.i.i.i.i.i134
  br i1 %cmp.i.not.i.i.i135, label %if.then.i.i.i.i136, label %Flow3824, !amdgpu.uniform !42

if.then.i.i.i.i136:                               ; preds = %365
  fence syncscope("agent") acquire
  %370 = load <2 x i64>, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !75
  %371 = extractelement <2 x i64> %370, i32 0
  %372 = inttoptr i64 %371 to ptr
  %373 = extractelement <2 x i64> %370, i32 1
  %374 = inttoptr i64 %373 to ptr
  %idxprom.i.i.i.i137 = zext i32 %spec.store.select.i.i.i125 to i64
  %arrayidx.i.i.i.i138 = getelementptr inbounds nuw i32, ptr %372, i64 %idxprom.i.i.i.i137, !amdgpu.uniform !42
  %375 = load atomic i32, ptr %arrayidx.i.i.i.i138 monotonic, align 4
  %376 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %375)
  %arrayidx.i45.i.i.i139 = getelementptr inbounds nuw i32, ptr %374, i64 %idxprom.i.i.i.i137, !amdgpu.uniform !42
  %377 = load atomic i32, ptr %arrayidx.i45.i.i.i139 monotonic, align 4
  %378 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %377)
  %cmp.i46.not.i.i.i140 = icmp ne i32 %376, %378
  br i1 %cmp.i46.not.i.i.i140, label %if.then12.i.i.i141, label %Flow3822, !amdgpu.uniform !42

if.then12.i.i.i141:                               ; preds = %if.then.i.i.i.i136
  fence syncscope("agent") release
  %379 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %358, i1 true)
  %iszero.i.i.i.i.i.i.i142 = icmp ne i64 %358, 0
  %cmp2.i.i.i.i.i.i143 = icmp eq i64 %379, %sh_prom.i.i.i.i
  %cmp.i.i.i.i.i2.i = select i1 %iszero.i.i.i.i.i.i.i142, i1 %cmp2.i.i.i.i.i.i143, i1 false
  %conv4.i.i51.i.i.i144 = zext i1 %cmp.i.i.i.i.i2.i to i32
  %shl.i.i52.i.i.i145 = shl nuw i32 %conv4.i.i51.i.i.i144, %359
  %xor.i.i.i.i.i146 = xor i32 %shl.i.i52.i.i.i145, -1
  %380 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2453, !amdgpu.uniform !42

381:                                              ; preds = %ComputeEnd2454
  %sunkaddr3278 = getelementptr inbounds i8, ptr addrspace(1) %3039, i64 40
  %382 = atomicrmw and ptr addrspace(1) %sunkaddr3278, i32 %.lcssa4173 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %383, !amdgpu.uniform !42

383:                                              ; preds = %381, %ComputeEnd2454
  call void @llvm.amdgcn.end.cf.i64(i64 %3066)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %Flow3822, !amdgpu.uniform !42

if.end14.i.i.i206:                                ; preds = %Flow3822
  %384 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %358, i1 true)
  %iszero.i.i.i.i.i.i207 = icmp ne i64 %358, 0
  %cmp2.i.i.i.i.i208 = icmp eq i64 %384, %sh_prom.i.i.i.i
  %cmp.i.i.i.i.i209 = select i1 %iszero.i.i.i.i.i.i207, i1 %cmp2.i.i.i.i.i208, i1 false
  %385 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.i.i.i.i.i209)
  %386 = extractvalue { i1, i64 } %385, 0
  %387 = extractvalue { i1, i64 } %385, 1
  br i1 %386, label %if.then16.i.i.i211, label %if.end22.i.i.i210

if.then16.i.i.i211:                               ; preds = %if.end14.i.i.i206
  %388 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !76
  %opcode.i.i.i212 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %388, i64 %idxprom.i.i.i.i137, i32 1
  store i32 1660944387, ptr %opcode.i.i.i212, align 8, !tbaa !77
  %arrayidx21.i.i.i213 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %388, i64 %idxprom.i.i.i.i137
  store i64 %358, ptr %arrayidx21.i.i.i213, align 8, !tbaa !80
  br label %if.end22.i.i.i210, !amdgpu.uniform !42

Flow3823:                                         ; preds = %if.end22.i.i.i210, %Flow3822
  %389 = phi i32 [ %376, %if.end22.i.i.i210 ], [ %retval.sroa.8.0.i.i.i122, %Flow3822 ]
  %390 = phi i64 [ %358, %if.end22.i.i.i210 ], [ %retval.sroa.2.0.i.i.i121, %Flow3822 ]
  %391 = phi i32 [ %spec.store.select.i.i.i125, %if.end22.i.i.i210 ], [ %retval.sroa.6.0.i.i.i119, %Flow3822 ]
  br label %cleanup26.i.i.i147, !amdgpu.uniform !42

if.end22.i.i.i210:                                ; preds = %if.then16.i.i.i211, %if.end14.i.i.i206
  call void @llvm.amdgcn.end.cf.i64(i64 %387)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %Flow3823, !amdgpu.uniform !42

Flow3824:                                         ; preds = %cleanup26.i.i.i147, %365
  %392 = phi i32 [ %391, %cleanup26.i.i.i147 ], [ %354, %365 ]
  %393 = phi i64 [ %390, %cleanup26.i.i.i147 ], [ %355, %365 ]
  %394 = phi i32 [ %389, %cleanup26.i.i.i147 ], [ %356, %365 ]
  %395 = phi i32 [ %389, %cleanup26.i.i.i147 ], [ %retval.sroa.8.0.i.i.i122, %365 ]
  %396 = phi i64 [ %390, %cleanup26.i.i.i147 ], [ %retval.sroa.2.0.i.i.i121, %365 ]
  %397 = phi i32 [ %391, %cleanup26.i.i.i147 ], [ %retval.sroa.6.0.i.i.i119, %365 ]
  %398 = phi i1 [ %cmp.i46.not.i.i.i140, %cleanup26.i.i.i147 ], [ true, %365 ]
  br i1 %398, label %for.inc.i.i.i152, label %Flow3825, !amdgpu.uniform !42

cleanup26.i.i.i147:                               ; preds = %Flow3823
  br label %Flow3824, !amdgpu.uniform !42

for.inc.i.i.i152:                                 ; preds = %Flow3824
  %inc.i.i.i157 = add i32 %spec.store.select.i.i.i125, 1
  br label %Flow3825, !amdgpu.uniform !42

Flow3825:                                         ; preds = %for.inc.i.i.i152, %Flow3824
  %399 = phi i32 [ %inc.i.i.i157, %for.inc.i.i.i152 ], [ poison, %Flow3824 ]
  %400 = phi i1 [ false, %for.inc.i.i.i152 ], [ true, %Flow3824 ]
  br i1 %400, label %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i158, label %for.cond.i.i.i118, !amdgpu.uniform !42

_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i158: ; preds = %Flow3825
  %.lcssa4176 = phi i32 [ %392, %Flow3825 ]
  %.lcssa4175 = phi i64 [ %393, %Flow3825 ]
  %.lcssa4174 = phi i32 [ %394, %Flow3825 ]
  fence acquire
  %401 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %mul.i.i.i.i.i.i159 = shl i32 %.lcssa4176, 6
  %idxprom.i19.i.i.i.i.i160 = zext i32 %mul.i.i.i.i.i.i159 to i64
  %arrayidx.i20.i.i.i.i.i161 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %401, i64 %idxprom.i19.i.i.i.i.i160
  %arrayidx.i22.i.i.i.i.i162 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %arrayidx.i20.i.i.i.i.i161, i64 %sh_prom.i.i.i.i
  store i64 %writer.sroa.10.4.i.i.i.i, ptr %arrayidx.i22.i.i.i.i.i162, align 8, !tbaa !82
  %spec.select.i.i.i.i.i.i.i163 = tail call i64 @llvm.umin.i64(i64 %writer.sroa.10.4.i.i.i.i, i64 56)
  %402 = and i64 %spec.select.i.i.i.i.i.i.i163, 15
  %403 = sub i64 %spec.select.i.i.i.i.i.i.i163, %402
  %404 = icmp ne i64 %403, 0
  %405 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %404)
  %406 = extractvalue { i1, i64 } %405, 0
  %407 = extractvalue { i1, i64 } %405, 1
  br i1 %406, label %loop-memcpy-expansion2266.preheader, label %Flow3821

loop-memcpy-expansion2266.preheader:              ; preds = %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i158
  %408 = add nuw nsw i64 %75, 8
  %scevgep3189 = getelementptr i8, ptr %401, i64 %408
  %409 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i160, 6
  %scevgep3190 = getelementptr i8, ptr %scevgep3189, i64 %409
  br label %loop-memcpy-expansion2266, !amdgpu.uniform !42

Flow3821:                                         ; preds = %Flow3820, %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i158
  call void @llvm.amdgcn.end.cf.i64(i64 %407)
  br label %loop-memcpy-residual-header2269, !amdgpu.uniform !42

loop-memcpy-expansion2266:                        ; preds = %loop-memcpy-expansion2266.preheader, %loop-memcpy-expansion2266
  %phi.broken3900 = phi i64 [ 0, %loop-memcpy-expansion2266.preheader ], [ %414, %loop-memcpy-expansion2266 ]
  %loop-index2267 = phi i64 [ %412, %loop-memcpy-expansion2266 ], [ 0, %loop-memcpy-expansion2266.preheader ]
  %410 = getelementptr i8, ptr %325, i64 %loop-index2267
  %411 = load <4 x i32>, ptr %410, align 1
  %scevgep3191 = getelementptr i8, ptr %scevgep3190, i64 %loop-index2267
  store <4 x i32> %411, ptr %scevgep3191, align 1
  %412 = add i64 %loop-index2267, 16
  %413 = icmp uge i64 %412, %403
  %414 = call i64 @llvm.amdgcn.if.break.i64(i1 %413, i64 %phi.broken3900)
  %415 = call i1 @llvm.amdgcn.loop.i64(i64 %414)
  br i1 %415, label %Flow3820, label %loop-memcpy-expansion2266

Flow3819:                                         ; preds = %Flow3817, %Flow3818
  %416 = phi i64 [ %2804, %Flow3817 ], [ %2796, %Flow3818 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %2799)
  br label %post-loop-memcpy-expansion2265, !amdgpu.uniform !42

loop-memcpy-residual2268:                         ; preds = %loop-memcpy-residual2268.preheader, %loop-memcpy-residual2268
  %phi.broken3901 = phi i64 [ 0, %loop-memcpy-residual2268.preheader ], [ %420, %loop-memcpy-residual2268 ]
  %residual-loop-index2270 = phi i64 [ %418, %loop-memcpy-residual2268 ], [ 0, %loop-memcpy-residual2268.preheader ]
  %scevgep3188 = getelementptr i8, ptr %scevgep3187, i64 %residual-loop-index2270
  %417 = load i8, ptr %scevgep3188, align 1
  %scevgep3186 = getelementptr i8, ptr %scevgep3185, i64 %residual-loop-index2270
  store i8 %417, ptr %scevgep3186, align 1
  %418 = add i64 %residual-loop-index2270, 1
  %419 = icmp uge i64 %418, %402
  %420 = call i64 @llvm.amdgcn.if.break.i64(i1 %419, i64 %phi.broken3901)
  %421 = call i1 @llvm.amdgcn.loop.i64(i64 %420)
  br i1 %421, label %Flow3817, label %loop-memcpy-residual2268

Flow3817:                                         ; preds = %loop-memcpy-residual2268
  %.lcssa4169 = phi i64 [ %420, %loop-memcpy-residual2268 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa4169)
  br label %Flow3819, !amdgpu.uniform !42

post-loop-memcpy-expansion2265:                   ; preds = %Flow3819
  %tobool.not.i.i.i.i.i.i165 = icmp eq i32 %.lcssa4174, 0
  %conv.i.i.i.i.i.i166 = zext i1 %tobool.not.i.i.i.i.i.i165 to i32
  fence release
  %422 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %idxprom.i23.i.i.i.i.i167 = zext i32 %.lcssa4176 to i64
  %arrayidx.i24.i.i.i.i.i168 = getelementptr inbounds nuw i32, ptr %422, i64 %idxprom.i23.i.i.i.i.i167
  store atomic i32 %conv.i.i.i.i.i.i166, ptr %arrayidx.i24.i.i.i.i.i168 monotonic, align 4
  %423 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !84
  %arrayidx.i.i5.i.i169 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %423, i64 %idxprom.i23.i.i.i.i.i167, !amdgpu.uniform !42
  %424 = load i64, ptr %arrayidx.i.i5.i.i169, align 8, !tbaa !80
  %scevgep3171 = getelementptr i8, ptr %325, i64 56
  %425 = add i64 %writer.sroa.10.4.i.i.i.i, -56
  br label %while.cond.i.i.i.i170, !amdgpu.uniform !42

while.cond.i.i.i.i170:                            ; preds = %post-loop-memcpy-expansion2265, %Flow3816
  %phi.broken3904 = phi i64 [ 0, %post-loop-memcpy-expansion2265 ], [ %442, %Flow3816 ]
  %lsr.iv3177 = phi i64 [ %440, %Flow3816 ], [ %425, %post-loop-memcpy-expansion2265 ]
  %lsr.iv3172 = phi ptr [ %439, %Flow3816 ], [ %scevgep3171, %post-loop-memcpy-expansion2265 ]
  %port.sroa.37.0.i.i171 = phi i32 [ %438, %Flow3816 ], [ %conv.i.i.i.i.i.i166, %post-loop-memcpy-expansion2265 ]
  %idx.0.i.i.i.i172 = phi i64 [ %437, %Flow3816 ], [ 56, %post-loop-memcpy-expansion2265 ]
  %cmp.i.i.i.i173 = icmp ugt i64 %writer.sroa.10.4.i.i.i.i, %idx.0.i.i.i.i172
  %426 = tail call i64 @llvm.amdgcn.ballot.i64(i1 %cmp.i.i.i.i173)
  %and.i.i.i.i6.i.i174 = and i64 %426, %424
  %tobool.not.i.i.i.i175 = icmp ne i64 %and.i.i.i.i6.i.i174, 0
  %427 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i22.i.i.i176 = getelementptr inbounds nuw i32, ptr %427, i64 %idxprom.i23.i.i.i.i.i167, !amdgpu.uniform !42
  %428 = load atomic i32, ptr %arrayidx.i22.i.i.i176 monotonic, align 4
  %429 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %tobool.not.i.i.i.i175)
  %430 = extractvalue { i1, i64 } %429, 0
  %431 = extractvalue { i1, i64 } %429, 1
  br i1 %430, label %cond.false.i8.i.i.i.i177, label %Flow3816

cond.false.i8.i.i.i.i177:                         ; preds = %while.cond.i.i.i.i170
  %432 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %428)
  %433 = icmp ne i32 %432, %port.sroa.37.0.i.i171
  br i1 %433, label %while.body.i.i21.i.i.i.i178.preheader, label %Flow3815, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i178.preheader:            ; preds = %cond.false.i8.i.i.i.i177
  br label %while.body.i.i21.i.i.i.i178, !amdgpu.uniform !42

Flow3815:                                         ; preds = %Flow3814, %cond.false.i8.i.i.i.i177
  br label %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i181, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i178:                      ; preds = %while.body.i.i21.i.i.i.i178.preheader, %while.body.i.i21.i.i.i.i178
  tail call void @llvm.amdgcn.s.sleep(i32 2)
  %434 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i22.i.i.i.i179 = getelementptr inbounds nuw i32, ptr %434, i64 %idxprom.i23.i.i.i.i.i167, !amdgpu.uniform !42
  %435 = load atomic i32, ptr %arrayidx.i.i.i22.i.i.i.i179 monotonic, align 4
  %436 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %435)
  %cmp.i.not.i.i23.i.i.i.i180 = icmp eq i32 %436, %port.sroa.37.0.i.i171
  br i1 %cmp.i.not.i.i23.i.i.i.i180, label %Flow3814, label %while.body.i.i21.i.i.i.i178, !amdgpu.uniform !42

Flow3814:                                         ; preds = %while.body.i.i21.i.i.i.i178
  br label %Flow3815, !amdgpu.uniform !42

Flow3816:                                         ; preds = %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i182, %while.cond.i.i.i.i170
  %437 = phi i64 [ %add.i.i.i.i185, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i182 ], [ poison, %while.cond.i.i.i.i170 ]
  %438 = phi i32 [ %conv.i.i26.i.i.i.i183, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i182 ], [ poison, %while.cond.i.i.i.i170 ]
  %439 = phi ptr [ %scevgep3173, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i182 ], [ poison, %while.cond.i.i.i.i170 ]
  %440 = phi i64 [ %lsr.iv.next3178, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i182 ], [ poison, %while.cond.i.i.i.i170 ]
  %441 = phi i1 [ false, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i182 ], [ true, %while.cond.i.i.i.i170 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %431)
  %442 = call i64 @llvm.amdgcn.if.break.i64(i1 %441, i64 %phi.broken3904)
  %443 = call i1 @llvm.amdgcn.loop.i64(i64 %442)
  br i1 %443, label %cond.false.i.i.i191, label %while.cond.i.i.i.i170

_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i181: ; preds = %Flow3815
  fence acquire
  %444 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.i.i.i.i173)
  %445 = extractvalue { i1, i64 } %444, 0
  %446 = extractvalue { i1, i64 } %444, 1
  br i1 %445, label %if.then.i.i.i.i.i.i3.i, label %Flow3813

if.then.i.i.i.i.i.i3.i:                           ; preds = %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i181
  %sub.i.i.i.i.i.i.i187 = sub nuw i64 %writer.sroa.10.4.i.i.i.i, %idx.0.i.i.i.i172
  %spec.select.i.i.i36.i.i.i.i188 = tail call i64 @llvm.umin.i64(i64 %sub.i.i.i.i.i.i.i187, i64 64)
  %447 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %448 = and i64 %spec.select.i.i.i36.i.i.i.i188, 15
  %449 = sub i64 %spec.select.i.i.i36.i.i.i.i188, %448
  %450 = icmp ne i64 %449, 0
  %451 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %450)
  %452 = extractvalue { i1, i64 } %451, 0
  %453 = extractvalue { i1, i64 } %451, 1
  br i1 %452, label %loop-memcpy-expansion2260.preheader, label %Flow3812

loop-memcpy-expansion2260.preheader:              ; preds = %if.then.i.i.i.i.i.i3.i
  %arrayidx.i20.i35.i.i.i.i189 = getelementptr %"struct.rpc::Buffer", ptr %447, i64 %idxprom.i19.i.i.i.i.i160
  %arrayidx.i22.i38.i.i.i.i190 = getelementptr %"struct.rpc::Buffer", ptr %arrayidx.i20.i35.i.i.i.i189, i64 %sh_prom.i.i.i.i
  br label %loop-memcpy-expansion2260, !amdgpu.uniform !42

Flow3812:                                         ; preds = %Flow3811, %if.then.i.i.i.i.i.i3.i
  call void @llvm.amdgcn.end.cf.i64(i64 %453)
  br label %loop-memcpy-residual-header2263, !amdgpu.uniform !42

loop-memcpy-expansion2260:                        ; preds = %loop-memcpy-expansion2260.preheader, %loop-memcpy-expansion2260
  %phi.broken3902 = phi i64 [ 0, %loop-memcpy-expansion2260.preheader ], [ %458, %loop-memcpy-expansion2260 ]
  %loop-index2261 = phi i64 [ %456, %loop-memcpy-expansion2260 ], [ 0, %loop-memcpy-expansion2260.preheader ]
  %scevgep3174 = getelementptr i8, ptr %lsr.iv3172, i64 %loop-index2261
  %454 = load <4 x i32>, ptr %scevgep3174, align 1
  %455 = getelementptr i8, ptr %arrayidx.i22.i38.i.i.i.i190, i64 %loop-index2261
  store <4 x i32> %454, ptr %455, align 1
  %456 = add i64 %loop-index2261, 16
  %457 = icmp uge i64 %456, %449
  %458 = call i64 @llvm.amdgcn.if.break.i64(i1 %457, i64 %phi.broken3902)
  %459 = call i1 @llvm.amdgcn.loop.i64(i64 %458)
  br i1 %459, label %Flow3811, label %loop-memcpy-expansion2260

Flow3810:                                         ; preds = %Flow3809, %loop-memcpy-residual-header2263
  call void @llvm.amdgcn.end.cf.i64(i64 %2789)
  br label %Flow3813, !amdgpu.uniform !42

loop-memcpy-residual2262:                         ; preds = %loop-memcpy-residual2262.preheader, %loop-memcpy-residual2262
  %phi.broken3903 = phi i64 [ 0, %loop-memcpy-residual2262.preheader ], [ %463, %loop-memcpy-residual2262 ]
  %residual-loop-index2264 = phi i64 [ %461, %loop-memcpy-residual2262 ], [ 0, %loop-memcpy-residual2262.preheader ]
  %scevgep3183 = getelementptr i8, ptr %scevgep3182, i64 %residual-loop-index2264
  %460 = load i8, ptr %scevgep3183, align 1
  %scevgep3181 = getelementptr i8, ptr %scevgep3180, i64 %residual-loop-index2264
  store i8 %460, ptr %scevgep3181, align 1
  %461 = add i64 %residual-loop-index2264, 1
  %462 = icmp uge i64 %461, %448
  %463 = call i64 @llvm.amdgcn.if.break.i64(i1 %462, i64 %phi.broken3903)
  %464 = call i1 @llvm.amdgcn.loop.i64(i64 %463)
  br i1 %464, label %Flow3809, label %loop-memcpy-residual2262

Flow3809:                                         ; preds = %loop-memcpy-residual2262
  %.lcssa4166 = phi i64 [ %463, %loop-memcpy-residual2262 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa4166)
  br label %Flow3810, !amdgpu.uniform !42

_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i182: ; preds = %Flow3813
  %conv.i.i26.i.i.i.i183 = xor i32 %port.sroa.37.0.i.i171, 1
  fence release
  %465 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %arrayidx.i24.i29.i.i.i.i184 = getelementptr inbounds nuw i32, ptr %465, i64 %idxprom.i23.i.i.i.i.i167
  store atomic i32 %conv.i.i26.i.i.i.i183, ptr %arrayidx.i24.i29.i.i.i.i184 monotonic, align 4
  %add.i.i.i.i185 = add i64 %idx.0.i.i.i.i172, 64
  %scevgep3173 = getelementptr i8, ptr %lsr.iv3172, i64 64
  %lsr.iv.next3178 = add i64 %lsr.iv3177, -64
  br label %Flow3816, !amdgpu.uniform !42

cond.false.i.i.i191:                              ; preds = %Flow3816
  %.lcssa4168 = phi i64 [ %442, %Flow3816 ]
  %port.sroa.37.0.i.i171.lcssa = phi i32 [ %port.sroa.37.0.i.i171, %Flow3816 ]
  %.lcssa4167 = phi i32 [ %428, %Flow3816 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa4168)
  %466 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %.lcssa4167)
  %467 = icmp ne i32 %466, %port.sroa.37.0.i.i171.lcssa
  %468 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %467)
  %469 = extractvalue { i1, i64 } %468, 0
  %470 = extractvalue { i1, i64 } %468, 1
  br i1 %469, label %while.body.i.i.i.i192.preheader, label %Flow3808

while.body.i.i.i.i192.preheader:                  ; preds = %cond.false.i.i.i191
  br label %while.body.i.i.i.i192, !amdgpu.uniform !42

Flow3808:                                         ; preds = %Flow3807, %cond.false.i.i.i191
  call void @llvm.amdgcn.end.cf.i64(i64 %470)
  br label %_ZN22__llvm_libc_22_0_0_git15write_to_stderrENS_3cpp11string_viewE.exit.i, !amdgpu.uniform !42

while.body.i.i.i.i192:                            ; preds = %while.body.i.i.i.i192.preheader, %while.body.i.i.i.i192
  %phi.broken3905 = phi i64 [ 0, %while.body.i.i.i.i192.preheader ], [ %474, %while.body.i.i.i.i192 ]
  tail call void @llvm.amdgcn.s.sleep(i32 2)
  %471 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i8.i.i193 = getelementptr inbounds nuw i32, ptr %471, i64 %idxprom.i23.i.i.i.i.i167, !amdgpu.uniform !42
  %472 = load atomic i32, ptr %arrayidx.i.i.i8.i.i193 monotonic, align 4
  %473 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %472)
  %cmp.i.not.i.i.i.i194 = icmp eq i32 %473, %port.sroa.37.0.i.i171.lcssa
  %474 = call i64 @llvm.amdgcn.if.break.i64(i1 %cmp.i.not.i.i.i.i194, i64 %phi.broken3905)
  %475 = call i1 @llvm.amdgcn.loop.i64(i64 %474)
  br i1 %475, label %Flow3807, label %while.body.i.i.i.i192

Flow3807:                                         ; preds = %while.body.i.i.i.i192
  %.lcssa4164 = phi i64 [ %474, %while.body.i.i.i.i192 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa4164)
  br label %Flow3808, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git15write_to_stderrENS_3cpp11string_viewE.exit.i: ; preds = %Flow3808
  fence acquire
  tail call void @llvm.amdgcn.wave.barrier()
  fence syncscope("agent") release
  %476 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %.lcssa4175, i1 true)
  %iszero.i.i.i.i.i13.i.i195 = icmp ne i64 %.lcssa4175, 0
  %cmp2.i.i.i.i14.i.i196 = icmp eq i64 %476, %sh_prom.i.i.i.i
  %cmp.i.i.i.i15.i.i197 = select i1 %iszero.i.i.i.i.i13.i.i195, i1 %cmp2.i.i.i.i14.i.i196, i1 false
  %477 = and i32 %.lcssa4176, 31
  %conv4.i.i.i19.i.i201 = zext i1 %cmp.i.i.i.i15.i.i197 to i32
  %shl.i.i.i20.i.i202 = shl nuw i32 %conv4.i.i.i19.i.i201, %477
  %xor.i.i.i21.i.i203 = xor i32 %shl.i.i.i20.i.i202, -1
  %478 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2457, !amdgpu.uniform !42

479:                                              ; preds = %ComputeEnd2458
  %div8.i.i.i16.i.i198 = lshr i32 %.lcssa4176, 5
  %idxprom.i.i.i17.i.i199 = zext nneg i32 %div8.i.i.i16.i.i198 to i64
  %480 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i17.i.i199
  %sunkaddr3279 = getelementptr inbounds i8, ptr addrspace(1) %480, i64 40
  %481 = atomicrmw and ptr addrspace(1) %sunkaddr3279, i32 %.lcssa4163 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %482, !amdgpu.uniform !42

482:                                              ; preds = %479, %ComputeEnd2458
  call void @llvm.amdgcn.end.cf.i64(i64 %3083)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %for.cond.i.i1.i222, !amdgpu.uniform !42

for.cond.i.i1.i222:                               ; preds = %482, %Flow3806
  %483 = phi i32 [ %521, %Flow3806 ], [ poison, %482 ]
  %484 = phi i64 [ %522, %Flow3806 ], [ poison, %482 ]
  %485 = phi i32 [ %523, %Flow3806 ], [ poison, %482 ]
  %retval.sroa.6.0.i.i.i223 = phi i32 [ %526, %Flow3806 ], [ undef, %482 ]
  %retval.sroa.2.0.i.i.i225 = phi i64 [ %525, %Flow3806 ], [ undef, %482 ]
  %retval.sroa.8.0.i.i.i226 = phi i32 [ %524, %Flow3806 ], [ undef, %482 ]
  %index.0.i.i.i227 = phi i32 [ %528, %Flow3806 ], [ 0, %482 ]
  %486 = load i32, ptr addrspace(1) @__llvm_rpc_client, align 8, !tbaa !68
  %cmp.not.i.i.i228 = icmp ult i32 %index.0.i.i.i227, %486
  %spec.store.select.i.i.i229 = select i1 %cmp.not.i.i.i228, i32 %index.0.i.i.i227, i32 0
  %487 = tail call noundef i64 @llvm.amdgcn.ballot.i64(i1 true)
  %488 = and i32 %spec.store.select.i.i.i229, 31
  %489 = lshr i64 %487, %sh_prom.i.i.i.i
  %490 = trunc i64 %489 to i32
  %conv4.i.i.i.i.i233 = and i32 %490, 1
  %shl.i.i.i.i.i234 = shl nuw i32 %conv4.i.i.i.i.i233, %488
  %491 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2461, !amdgpu.uniform !42

492:                                              ; preds = %ComputeEnd2462
  %sunkaddr3280 = getelementptr inbounds i8, ptr addrspace(1) %3093, i64 40
  %493 = atomicrmw or ptr addrspace(1) %sunkaddr3280, i32 %.lcssa4157 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %494, !amdgpu.uniform !42

494:                                              ; preds = %492, %ComputeEnd2462
  %495 = phi i32 [ %493, %492 ], [ poison, %ComputeEnd2462 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %3102)
  %496 = call i32 @llvm.amdgcn.readfirstlane.i32(i32 %495)
  %497 = or i32 %496, %.lcssa4158
  %shl5.i.i.i.i.i235 = shl nuw i32 1, %488
  %and.i.i.i.i.i236 = and i32 %shl5.i.i.i.i.i235, %497
  %tobool3.i.i.i.i237 = icmp ne i32 %and.i.i.i.i.i236, 0
  %498 = tail call i64 @llvm.amdgcn.ballot.i64(i1 %tobool3.i.i.i.i237)
  %and.i.i.i.i.i.i238 = and i64 %498, %487
  %cmp.i.not.i.i.i239 = icmp ne i64 %487, %and.i.i.i.i.i.i238
  br i1 %cmp.i.not.i.i.i239, label %if.then.i.i.i.i240, label %Flow3805, !amdgpu.uniform !42

if.then.i.i.i.i240:                               ; preds = %494
  fence syncscope("agent") acquire
  %499 = load <2 x i64>, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !75
  %500 = extractelement <2 x i64> %499, i32 0
  %501 = inttoptr i64 %500 to ptr
  %502 = extractelement <2 x i64> %499, i32 1
  %503 = inttoptr i64 %502 to ptr
  %idxprom.i.i.i.i241 = zext i32 %spec.store.select.i.i.i229 to i64
  %arrayidx.i.i.i.i242 = getelementptr inbounds nuw i32, ptr %501, i64 %idxprom.i.i.i.i241, !amdgpu.uniform !42
  %504 = load atomic i32, ptr %arrayidx.i.i.i.i242 monotonic, align 4
  %505 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %504)
  %arrayidx.i45.i.i.i243 = getelementptr inbounds nuw i32, ptr %503, i64 %idxprom.i.i.i.i241, !amdgpu.uniform !42
  %506 = load atomic i32, ptr %arrayidx.i45.i.i.i243 monotonic, align 4
  %507 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %506)
  %cmp.i46.not.i.i.i244 = icmp ne i32 %505, %507
  br i1 %cmp.i46.not.i.i.i244, label %if.then12.i.i.i245, label %Flow3803, !amdgpu.uniform !42

if.then12.i.i.i245:                               ; preds = %if.then.i.i.i.i240
  fence syncscope("agent") release
  %508 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %487, i1 true)
  %iszero.i.i.i.i.i.i.i246 = icmp ne i64 %487, 0
  %cmp2.i.i.i.i.i.i247 = icmp eq i64 %508, %sh_prom.i.i.i.i
  %cmp.i.i.i.i.i.i248 = select i1 %iszero.i.i.i.i.i.i.i246, i1 %cmp2.i.i.i.i.i.i247, i1 false
  %conv4.i.i51.i.i.i249 = zext i1 %cmp.i.i.i.i.i.i248 to i32
  %shl.i.i52.i.i.i250 = shl nuw i32 %conv4.i.i51.i.i.i249, %488
  %xor.i.i.i.i.i251 = xor i32 %shl.i.i52.i.i.i250, -1
  %509 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2466, !amdgpu.uniform !42

510:                                              ; preds = %ComputeEnd2467
  %sunkaddr3281 = getelementptr inbounds i8, ptr addrspace(1) %3093, i64 40
  %511 = atomicrmw and ptr addrspace(1) %sunkaddr3281, i32 %.lcssa4159 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %512, !amdgpu.uniform !42

512:                                              ; preds = %510, %ComputeEnd2467
  call void @llvm.amdgcn.end.cf.i64(i64 %3120)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %Flow3803, !amdgpu.uniform !42

if.end14.i.i.i311:                                ; preds = %Flow3803
  %513 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %487, i1 true)
  %iszero.i.i.i.i.i.i312 = icmp ne i64 %487, 0
  %cmp2.i.i.i.i.i313 = icmp eq i64 %513, %sh_prom.i.i.i.i
  %cmp.i.i.i.i.i314 = select i1 %iszero.i.i.i.i.i.i312, i1 %cmp2.i.i.i.i.i313, i1 false
  %514 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.i.i.i.i.i314)
  %515 = extractvalue { i1, i64 } %514, 0
  %516 = extractvalue { i1, i64 } %514, 1
  br i1 %515, label %if.then16.i.i.i316, label %if.end22.i.i.i315

if.then16.i.i.i316:                               ; preds = %if.end14.i.i.i311
  %517 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !76
  %opcode.i.i.i317 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %517, i64 %idxprom.i.i.i.i241, i32 1
  store i32 1660944387, ptr %opcode.i.i.i317, align 8, !tbaa !77
  %arrayidx21.i.i.i318 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %517, i64 %idxprom.i.i.i.i241
  store i64 %487, ptr %arrayidx21.i.i.i318, align 8, !tbaa !80
  br label %if.end22.i.i.i315, !amdgpu.uniform !42

Flow3804:                                         ; preds = %if.end22.i.i.i315, %Flow3803
  %518 = phi i32 [ %505, %if.end22.i.i.i315 ], [ %retval.sroa.8.0.i.i.i226, %Flow3803 ]
  %519 = phi i64 [ %487, %if.end22.i.i.i315 ], [ %retval.sroa.2.0.i.i.i225, %Flow3803 ]
  %520 = phi i32 [ %spec.store.select.i.i.i229, %if.end22.i.i.i315 ], [ %retval.sroa.6.0.i.i.i223, %Flow3803 ]
  br label %cleanup26.i.i.i252, !amdgpu.uniform !42

if.end22.i.i.i315:                                ; preds = %if.then16.i.i.i316, %if.end14.i.i.i311
  call void @llvm.amdgcn.end.cf.i64(i64 %516)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %Flow3804, !amdgpu.uniform !42

Flow3805:                                         ; preds = %cleanup26.i.i.i252, %494
  %521 = phi i32 [ %520, %cleanup26.i.i.i252 ], [ %483, %494 ]
  %522 = phi i64 [ %519, %cleanup26.i.i.i252 ], [ %484, %494 ]
  %523 = phi i32 [ %518, %cleanup26.i.i.i252 ], [ %485, %494 ]
  %524 = phi i32 [ %518, %cleanup26.i.i.i252 ], [ %retval.sroa.8.0.i.i.i226, %494 ]
  %525 = phi i64 [ %519, %cleanup26.i.i.i252 ], [ %retval.sroa.2.0.i.i.i225, %494 ]
  %526 = phi i32 [ %520, %cleanup26.i.i.i252 ], [ %retval.sroa.6.0.i.i.i223, %494 ]
  %527 = phi i1 [ %cmp.i46.not.i.i.i244, %cleanup26.i.i.i252 ], [ true, %494 ]
  br i1 %527, label %for.inc.i.i.i257, label %Flow3806, !amdgpu.uniform !42

cleanup26.i.i.i252:                               ; preds = %Flow3804
  br label %Flow3805, !amdgpu.uniform !42

for.inc.i.i.i257:                                 ; preds = %Flow3805
  %inc.i.i.i262 = add i32 %spec.store.select.i.i.i229, 1
  br label %Flow3806, !amdgpu.uniform !42

Flow3806:                                         ; preds = %for.inc.i.i.i257, %Flow3805
  %528 = phi i32 [ %inc.i.i.i262, %for.inc.i.i.i257 ], [ poison, %Flow3805 ]
  %529 = phi i1 [ false, %for.inc.i.i.i257 ], [ true, %Flow3805 ]
  br i1 %529, label %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i263, label %for.cond.i.i1.i222, !amdgpu.uniform !42

_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i263: ; preds = %Flow3806
  %.lcssa4162 = phi i32 [ %521, %Flow3806 ]
  %.lcssa4161 = phi i64 [ %522, %Flow3806 ]
  %.lcssa4160 = phi i32 [ %523, %Flow3806 ]
  fence acquire
  %530 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %mul.i.i.i.i.i.i264 = shl i32 %.lcssa4162, 6
  %idxprom.i19.i.i.i.i.i265 = zext i32 %mul.i.i.i.i.i.i264 to i64
  %arrayidx.i20.i.i.i.i.i266 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %530, i64 %idxprom.i19.i.i.i.i.i265
  %arrayidx.i22.i.i.i.i.i267 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %arrayidx.i20.i.i.i.i.i266, i64 %sh_prom.i.i.i.i
  store i64 sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.7.11 to ptr), i64 10) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.7.11 to ptr) to i64)), ptr %arrayidx.i22.i.i.i.i.i267, align 8, !tbaa !82
  %spec.select.i.i.i.i.i.i.i268 = tail call i64 @llvm.umin.i64(i64 sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.7.11 to ptr), i64 10) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.7.11 to ptr) to i64)), i64 56)
  %531 = and i64 %spec.select.i.i.i.i.i.i.i268, 15
  %532 = sub i64 %spec.select.i.i.i.i.i.i.i268, %531
  %533 = icmp ne i64 %532, 0
  br i1 %533, label %loop-memcpy-expansion2332.preheader, label %Flow3802, !amdgpu.uniform !42

loop-memcpy-expansion2332.preheader:              ; preds = %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i263
  %534 = add nuw nsw i64 %75, 8
  %scevgep3168 = getelementptr i8, ptr %530, i64 %534
  %535 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i265, 6
  %scevgep3169 = getelementptr i8, ptr %scevgep3168, i64 %535
  br label %loop-memcpy-expansion2332, !amdgpu.uniform !42

Flow3802:                                         ; preds = %Flow3801, %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i263
  br label %loop-memcpy-residual-header2335, !amdgpu.uniform !42

loop-memcpy-expansion2332:                        ; preds = %loop-memcpy-expansion2332.preheader, %loop-memcpy-expansion2332
  %loop-index2333 = phi i64 [ %538, %loop-memcpy-expansion2332 ], [ 0, %loop-memcpy-expansion2332.preheader ]
  %536 = getelementptr i8, ptr addrspace(4) @.str.7.11, i64 %loop-index2333, !amdgpu.uniform !42
  %537 = load <4 x i32>, ptr addrspace(4) %536, align 1
  %scevgep3170 = getelementptr i8, ptr %scevgep3169, i64 %loop-index2333
  store <4 x i32> %537, ptr %scevgep3170, align 1
  %538 = add i64 %loop-index2333, 16
  %539 = icmp uge i64 %538, %532
  br i1 %539, label %Flow3801, label %loop-memcpy-expansion2332, !amdgpu.uniform !42

Flow3800:                                         ; preds = %Flow3798, %Flow3799
  %540 = phi i64 [ %2860, %Flow3798 ], [ %2857, %Flow3799 ]
  br label %post-loop-memcpy-expansion2331, !amdgpu.uniform !42

loop-memcpy-residual2334:                         ; preds = %loop-memcpy-residual2334.preheader, %loop-memcpy-residual2334
  %residual-loop-index2336 = phi i64 [ %542, %loop-memcpy-residual2334 ], [ 0, %loop-memcpy-residual2334.preheader ]
  %scevgep3167 = getelementptr i8, ptr addrspace(4) @.str.7.11, i64 %residual-loop-index2336, !amdgpu.uniform !42
  %541 = load i8, ptr addrspace(4) %scevgep3167, align 1
  %scevgep3166 = getelementptr i8, ptr %scevgep3165, i64 %residual-loop-index2336
  store i8 %541, ptr %scevgep3166, align 1
  %542 = add i64 %residual-loop-index2336, 1
  %543 = icmp uge i64 %542, %531
  br i1 %543, label %Flow3798, label %loop-memcpy-residual2334, !amdgpu.uniform !42

Flow3798:                                         ; preds = %loop-memcpy-residual2334
  br label %Flow3800, !amdgpu.uniform !42

post-loop-memcpy-expansion2331:                   ; preds = %Flow3800
  %tobool.not.i.i.i.i.i.i270 = icmp eq i32 %.lcssa4160, 0
  %conv.i.i.i.i.i.i271 = zext i1 %tobool.not.i.i.i.i.i.i270 to i32
  fence release
  %544 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %idxprom.i23.i.i.i.i.i272 = zext i32 %.lcssa4162 to i64
  %arrayidx.i24.i.i.i.i.i273 = getelementptr inbounds nuw i32, ptr %544, i64 %idxprom.i23.i.i.i.i.i272
  store atomic i32 %conv.i.i.i.i.i.i271, ptr %arrayidx.i24.i.i.i.i.i273 monotonic, align 4
  %545 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !84
  %arrayidx.i.i5.i.i274 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %545, i64 %idxprom.i23.i.i.i.i.i272, !amdgpu.uniform !42
  %546 = load i64, ptr %arrayidx.i.i5.i.i274, align 8, !tbaa !80
  br label %while.cond.i.i.i.i275, !amdgpu.uniform !42

while.cond.i.i.i.i275:                            ; preds = %post-loop-memcpy-expansion2331, %Flow3797
  %phi.broken3906 = phi i64 [ 0, %post-loop-memcpy-expansion2331 ], [ %563, %Flow3797 ]
  %lsr.iv3157 = phi i64 [ %561, %Flow3797 ], [ -46, %post-loop-memcpy-expansion2331 ]
  %lsr.iv3152 = phi ptr addrspace(4) [ %560, %Flow3797 ], [ getelementptr (i8, ptr addrspace(4) @.str.7.11, i64 56), %post-loop-memcpy-expansion2331 ]
  %port.sroa.37.0.i.i276 = phi i32 [ %559, %Flow3797 ], [ %conv.i.i.i.i.i.i271, %post-loop-memcpy-expansion2331 ]
  %idx.0.i.i.i.i277 = phi i64 [ %558, %Flow3797 ], [ 56, %post-loop-memcpy-expansion2331 ]
  %cmp.i.i.i.i278 = icmp ult i64 %idx.0.i.i.i.i277, sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.7.11 to ptr), i64 10) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.7.11 to ptr) to i64))
  %547 = tail call i64 @llvm.amdgcn.ballot.i64(i1 %cmp.i.i.i.i278)
  %and.i.i.i.i6.i.i279 = and i64 %547, %546
  %tobool.not.i.i.i.i280 = icmp ne i64 %and.i.i.i.i6.i.i279, 0
  %548 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i22.i.i.i281 = getelementptr inbounds nuw i32, ptr %548, i64 %idxprom.i23.i.i.i.i.i272, !amdgpu.uniform !42
  %549 = load atomic i32, ptr %arrayidx.i22.i.i.i281 monotonic, align 4
  %550 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %tobool.not.i.i.i.i280)
  %551 = extractvalue { i1, i64 } %550, 0
  %552 = extractvalue { i1, i64 } %550, 1
  br i1 %551, label %cond.false.i8.i.i.i.i282, label %Flow3797

cond.false.i8.i.i.i.i282:                         ; preds = %while.cond.i.i.i.i275
  %553 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %549)
  %554 = icmp ne i32 %553, %port.sroa.37.0.i.i276
  br i1 %554, label %while.body.i.i21.i.i.i.i283.preheader, label %Flow3796, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i283.preheader:            ; preds = %cond.false.i8.i.i.i.i282
  br label %while.body.i.i21.i.i.i.i283, !amdgpu.uniform !42

Flow3796:                                         ; preds = %Flow3795, %cond.false.i8.i.i.i.i282
  br label %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i286, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i283:                      ; preds = %while.body.i.i21.i.i.i.i283.preheader, %while.body.i.i21.i.i.i.i283
  tail call void @llvm.amdgcn.s.sleep(i32 2)
  %555 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i22.i.i.i.i284 = getelementptr inbounds nuw i32, ptr %555, i64 %idxprom.i23.i.i.i.i.i272, !amdgpu.uniform !42
  %556 = load atomic i32, ptr %arrayidx.i.i.i22.i.i.i.i284 monotonic, align 4
  %557 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %556)
  %cmp.i.not.i.i23.i.i.i.i285 = icmp eq i32 %557, %port.sroa.37.0.i.i276
  br i1 %cmp.i.not.i.i23.i.i.i.i285, label %Flow3795, label %while.body.i.i21.i.i.i.i283, !amdgpu.uniform !42

Flow3795:                                         ; preds = %while.body.i.i21.i.i.i.i283
  br label %Flow3796, !amdgpu.uniform !42

Flow3797:                                         ; preds = %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i287, %while.cond.i.i.i.i275
  %558 = phi i64 [ %add.i.i.i.i290, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i287 ], [ poison, %while.cond.i.i.i.i275 ]
  %559 = phi i32 [ %conv.i.i26.i.i.i.i288, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i287 ], [ poison, %while.cond.i.i.i.i275 ]
  %560 = phi ptr addrspace(4) [ %scevgep3153, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i287 ], [ poison, %while.cond.i.i.i.i275 ]
  %561 = phi i64 [ %lsr.iv.next3158, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i287 ], [ poison, %while.cond.i.i.i.i275 ]
  %562 = phi i1 [ false, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i287 ], [ true, %while.cond.i.i.i.i275 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %552)
  %563 = call i64 @llvm.amdgcn.if.break.i64(i1 %562, i64 %phi.broken3906)
  %564 = call i1 @llvm.amdgcn.loop.i64(i64 %563)
  br i1 %564, label %cond.false.i.i.i297, label %while.cond.i.i.i.i275

_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i286: ; preds = %Flow3796
  fence acquire
  br i1 %cmp.i.i.i.i278, label %if.then.i.i.i.i.i.i.i291, label %Flow3794, !amdgpu.uniform !42

if.then.i.i.i.i.i.i.i291:                         ; preds = %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i286
  %sub.i.i.i.i.i.i.i293 = sub nuw i64 sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.7.11 to ptr), i64 10) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.7.11 to ptr) to i64)), %idx.0.i.i.i.i277
  %spec.select.i.i.i36.i.i.i.i294 = tail call i64 @llvm.umin.i64(i64 %sub.i.i.i.i.i.i.i293, i64 64)
  %565 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %566 = and i64 %spec.select.i.i.i36.i.i.i.i294, 15
  %567 = sub i64 %spec.select.i.i.i36.i.i.i.i294, %566
  %568 = icmp ne i64 %567, 0
  br i1 %568, label %loop-memcpy-expansion2338.preheader, label %Flow3793, !amdgpu.uniform !42

loop-memcpy-expansion2338.preheader:              ; preds = %if.then.i.i.i.i.i.i.i291
  %arrayidx.i20.i35.i.i.i.i295 = getelementptr %"struct.rpc::Buffer", ptr %565, i64 %idxprom.i19.i.i.i.i.i265
  %arrayidx.i22.i38.i.i.i.i296 = getelementptr %"struct.rpc::Buffer", ptr %arrayidx.i20.i35.i.i.i.i295, i64 %sh_prom.i.i.i.i
  br label %loop-memcpy-expansion2338, !amdgpu.uniform !42

Flow3793:                                         ; preds = %Flow3792, %if.then.i.i.i.i.i.i.i291
  br label %loop-memcpy-residual-header2341, !amdgpu.uniform !42

loop-memcpy-expansion2338:                        ; preds = %loop-memcpy-expansion2338.preheader, %loop-memcpy-expansion2338
  %loop-index2339 = phi i64 [ %571, %loop-memcpy-expansion2338 ], [ 0, %loop-memcpy-expansion2338.preheader ]
  %scevgep3154 = getelementptr i8, ptr addrspace(4) %lsr.iv3152, i64 %loop-index2339, !amdgpu.uniform !42
  %569 = load <4 x i32>, ptr addrspace(4) %scevgep3154, align 1
  %570 = getelementptr i8, ptr %arrayidx.i22.i38.i.i.i.i296, i64 %loop-index2339
  store <4 x i32> %569, ptr %570, align 1
  %571 = add i64 %loop-index2339, 16
  %572 = icmp uge i64 %571, %567
  br i1 %572, label %Flow3792, label %loop-memcpy-expansion2338, !amdgpu.uniform !42

Flow3791:                                         ; preds = %Flow3790, %loop-memcpy-residual-header2341
  br label %Flow3794, !amdgpu.uniform !42

loop-memcpy-residual2340:                         ; preds = %loop-memcpy-residual2340.preheader, %loop-memcpy-residual2340
  %residual-loop-index2342 = phi i64 [ %574, %loop-memcpy-residual2340 ], [ 0, %loop-memcpy-residual2340.preheader ]
  %scevgep3163 = getelementptr i8, ptr addrspace(4) %scevgep3162, i64 %residual-loop-index2342, !amdgpu.uniform !42
  %573 = load i8, ptr addrspace(4) %scevgep3163, align 1
  %scevgep3161 = getelementptr i8, ptr %scevgep3160, i64 %residual-loop-index2342
  store i8 %573, ptr %scevgep3161, align 1
  %574 = add i64 %residual-loop-index2342, 1
  %575 = icmp uge i64 %574, %566
  br i1 %575, label %Flow3790, label %loop-memcpy-residual2340, !amdgpu.uniform !42

Flow3790:                                         ; preds = %loop-memcpy-residual2340
  br label %Flow3791, !amdgpu.uniform !42

_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i287: ; preds = %Flow3794
  %conv.i.i26.i.i.i.i288 = xor i32 %port.sroa.37.0.i.i276, 1
  fence release
  %576 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %arrayidx.i24.i29.i.i.i.i289 = getelementptr inbounds nuw i32, ptr %576, i64 %idxprom.i23.i.i.i.i.i272
  store atomic i32 %conv.i.i26.i.i.i.i288, ptr %arrayidx.i24.i29.i.i.i.i289 monotonic, align 4
  %add.i.i.i.i290 = add i64 %idx.0.i.i.i.i277, 64
  %scevgep3153 = getelementptr i8, ptr addrspace(4) %lsr.iv3152, i64 64
  %lsr.iv.next3158 = add i64 %lsr.iv3157, -64
  br label %Flow3797, !amdgpu.uniform !42

cond.false.i.i.i297:                              ; preds = %Flow3797
  %.lcssa4156 = phi i64 [ %563, %Flow3797 ]
  %port.sroa.37.0.i.i276.lcssa = phi i32 [ %port.sroa.37.0.i.i276, %Flow3797 ]
  %.lcssa4155 = phi i32 [ %549, %Flow3797 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa4156)
  %577 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %.lcssa4155)
  %578 = icmp ne i32 %577, %port.sroa.37.0.i.i276.lcssa
  %579 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %578)
  %580 = extractvalue { i1, i64 } %579, 0
  %581 = extractvalue { i1, i64 } %579, 1
  br i1 %580, label %while.body.i.i.i.i298.preheader, label %Flow3789

while.body.i.i.i.i298.preheader:                  ; preds = %cond.false.i.i.i297
  br label %while.body.i.i.i.i298, !amdgpu.uniform !42

Flow3789:                                         ; preds = %Flow3788, %cond.false.i.i.i297
  call void @llvm.amdgcn.end.cf.i64(i64 %581)
  br label %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit319, !amdgpu.uniform !42

while.body.i.i.i.i298:                            ; preds = %while.body.i.i.i.i298.preheader, %while.body.i.i.i.i298
  %phi.broken3907 = phi i64 [ 0, %while.body.i.i.i.i298.preheader ], [ %585, %while.body.i.i.i.i298 ]
  tail call void @llvm.amdgcn.s.sleep(i32 2)
  %582 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i8.i.i299 = getelementptr inbounds nuw i32, ptr %582, i64 %idxprom.i23.i.i.i.i.i272, !amdgpu.uniform !42
  %583 = load atomic i32, ptr %arrayidx.i.i.i8.i.i299 monotonic, align 4
  %584 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %583)
  %cmp.i.not.i.i.i.i300 = icmp eq i32 %584, %port.sroa.37.0.i.i276.lcssa
  %585 = call i64 @llvm.amdgcn.if.break.i64(i1 %cmp.i.not.i.i.i.i300, i64 %phi.broken3907)
  %586 = call i1 @llvm.amdgcn.loop.i64(i64 %585)
  br i1 %586, label %Flow3788, label %while.body.i.i.i.i298

Flow3788:                                         ; preds = %while.body.i.i.i.i298
  %.lcssa4154 = phi i64 [ %585, %while.body.i.i.i.i298 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa4154)
  br label %Flow3789, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit319: ; preds = %Flow3789
  fence acquire
  tail call void @llvm.amdgcn.wave.barrier()
  fence syncscope("agent") release
  %587 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %.lcssa4161, i1 true)
  %iszero.i.i.i.i.i13.i.i302 = icmp ne i64 %.lcssa4161, 0
  %cmp2.i.i.i.i14.i.i303 = icmp eq i64 %587, %sh_prom.i.i.i.i
  %cmp.i.i.i.i15.i.i304 = select i1 %iszero.i.i.i.i.i13.i.i302, i1 %cmp2.i.i.i.i14.i.i303, i1 false
  %588 = and i32 %.lcssa4162, 31
  %conv4.i.i.i19.i.i308 = zext i1 %cmp.i.i.i.i15.i.i304 to i32
  %shl.i.i.i20.i.i309 = shl nuw i32 %conv4.i.i.i19.i.i308, %588
  %xor.i.i.i21.i.i310 = xor i32 %shl.i.i.i20.i.i309, -1
  %589 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2470, !amdgpu.uniform !42

590:                                              ; preds = %ComputeEnd2471
  %div8.i.i.i16.i.i305 = lshr i32 %.lcssa4162, 5
  %idxprom.i.i.i17.i.i306 = zext nneg i32 %div8.i.i.i16.i.i305 to i64
  %591 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i17.i.i306
  %sunkaddr3282 = getelementptr inbounds i8, ptr addrspace(1) %591, i64 40
  %592 = atomicrmw and ptr addrspace(1) %sunkaddr3282, i32 %.lcssa4153 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %593, !amdgpu.uniform !42

593:                                              ; preds = %590, %ComputeEnd2471
  call void @llvm.amdgcn.end.cf.i64(i64 %3137)
  tail call void @llvm.amdgcn.wave.barrier()
  %inc.i.i.i32.i = add nuw i64 %spec.select.i, 1
  %cmp2.i.i.i35.i = icmp ult i64 %spec.select.i, 1676976733973595600
  br i1 %cmp2.i.i.i35.i, label %if.then3.i.i.i44.i, label %if.end4.i.i.i36.i, !amdgpu.uniform !42

if.then3.i.i.i44.i:                               ; preds = %593
  %mul.i.i.i45.i = mul nuw i64 %inc.i.i.i32.i, 11
  %div18.i.i.i46.i = lshr i64 %mul.i.i.i45.i, 3
  br label %if.end4.i.i.i36.i, !amdgpu.uniform !42

if.end4.i.i.i36.i:                                ; preds = %if.then3.i.i.i44.i, %593
  %new_capacity.addr.0.i.i.i37.i = phi i64 [ %div18.i.i.i46.i, %if.then3.i.i.i44.i ], [ %inc.i.i.i32.i, %593 ]
  %sub.i.i320 = add i64 %new_capacity.addr.0.i.i.i37.i, 7
  %div2.i.i321 = and i64 %sub.i.i320, -8
  %594 = load ptr, ptr addrspace(1) @_ZN12_GLOBAL__N_13ptrE, align 8, !tbaa !55
  %add.ptr.i.i322 = getelementptr inbounds nuw i8, ptr %594, i64 %div2.i.i321
  store ptr %add.ptr.i.i322, ptr addrspace(1) @_ZN12_GLOBAL__N_13ptrE, align 8, !tbaa !55
  %595 = and i64 %spec.select.i, 15
  %596 = sub i64 %spec.select.i, %595
  %597 = icmp ne i64 %596, 0
  br i1 %597, label %loop-memcpy-expansion2284.preheader, label %Flow3787, !amdgpu.uniform !42

loop-memcpy-expansion2284.preheader:              ; preds = %if.end4.i.i.i36.i
  br label %loop-memcpy-expansion2284, !amdgpu.uniform !42

Flow3787:                                         ; preds = %Flow3786, %if.end4.i.i.i36.i
  br label %loop-memcpy-residual-header2287, !amdgpu.uniform !42

loop-memcpy-expansion2284:                        ; preds = %loop-memcpy-expansion2284.preheader, %loop-memcpy-expansion2284
  %loop-index2285 = phi i64 [ %601, %loop-memcpy-expansion2284 ], [ 0, %loop-memcpy-expansion2284.preheader ]
  %598 = getelementptr i8, ptr %3, i64 %loop-index2285, !amdgpu.uniform !42
  %599 = load <4 x i32>, ptr %598, align 1
  %600 = getelementptr i8, ptr %594, i64 %loop-index2285
  store <4 x i32> %599, ptr %600, align 1
  %601 = add i64 %loop-index2285, 16
  %602 = icmp uge i64 %601, %596
  br i1 %602, label %Flow3786, label %loop-memcpy-expansion2284, !amdgpu.uniform !42

Flow3785:                                         ; preds = %Flow3784, %loop-memcpy-residual-header2287
  br label %post-loop-memcpy-expansion2283, !amdgpu.uniform !42

loop-memcpy-residual2286:                         ; preds = %loop-memcpy-residual2286.preheader, %loop-memcpy-residual2286
  %residual-loop-index2288 = phi i64 [ %604, %loop-memcpy-residual2286 ], [ 0, %loop-memcpy-residual2286.preheader ]
  %scevgep3151 = getelementptr i8, ptr %scevgep3150, i64 %residual-loop-index2288, !amdgpu.uniform !42
  %603 = load i8, ptr %scevgep3151, align 1
  %scevgep3149 = getelementptr i8, ptr %scevgep3148, i64 %residual-loop-index2288
  store i8 %603, ptr %scevgep3149, align 1
  %604 = add i64 %residual-loop-index2288, 1
  %605 = icmp uge i64 %604, %595
  br i1 %605, label %Flow3784, label %loop-memcpy-residual2286, !amdgpu.uniform !42

Flow3784:                                         ; preds = %loop-memcpy-residual2286
  br label %Flow3785, !amdgpu.uniform !42

post-loop-memcpy-expansion2283:                   ; preds = %Flow3785
  %sub.ptr.lhs.cast.i.i323 = ptrtoint ptr %add.ptr.i.i322 to i64
  %sub.ptr.sub.i.i324 = sub i64 %sub.ptr.lhs.cast.i.i323, ptrtoint (ptr addrspacecast (ptr addrspace(1) @_ZN12_GLOBAL__N_16memoryE to ptr) to i64)
  %cmp.i.i325 = icmp ugt i64 %sub.ptr.sub.i.i324, 65335
  %cond.i.i326 = select i1 %cmp.i.i325, ptr null, ptr %594
  %cmp.not.i10.i.i.i = icmp ne ptr %cond.i.i326, addrspacecast (ptr addrspace(4) @_ZN22__llvm_libc_22_0_0_git3cpp6string14NULL_CHARACTERE to ptr)
  br i1 %cmp.not.i10.i.i.i, label %if.then.i.i.i42.i, label %for.cond.i.i.i328.preheader, !amdgpu.uniform !42

if.then.i.i.i42.i:                                ; preds = %post-loop-memcpy-expansion2283
  %arrayidx.i.i.i43.i = getelementptr inbounds nuw i8, ptr %594, i64 %spec.select.i
  store i8 0, ptr %arrayidx.i.i.i43.i, align 1, !tbaa !57
  br label %for.cond.i.i.i328.preheader, !amdgpu.uniform !42

for.cond.i.i.i328.preheader:                      ; preds = %if.then.i.i.i42.i, %post-loop-memcpy-expansion2283
  br label %for.cond.i.i.i328, !amdgpu.uniform !42

for.cond.i.i.i328:                                ; preds = %for.cond.i.i.i328.preheader, %Flow3783
  %606 = phi i32 [ %644, %Flow3783 ], [ poison, %for.cond.i.i.i328.preheader ]
  %607 = phi i64 [ %645, %Flow3783 ], [ poison, %for.cond.i.i.i328.preheader ]
  %608 = phi i32 [ %646, %Flow3783 ], [ poison, %for.cond.i.i.i328.preheader ]
  %retval.sroa.6.0.i.i.i329 = phi i32 [ %649, %Flow3783 ], [ undef, %for.cond.i.i.i328.preheader ]
  %retval.sroa.2.0.i.i.i331 = phi i64 [ %648, %Flow3783 ], [ undef, %for.cond.i.i.i328.preheader ]
  %retval.sroa.8.0.i.i.i332 = phi i32 [ %647, %Flow3783 ], [ undef, %for.cond.i.i.i328.preheader ]
  %index.0.i.i.i333 = phi i32 [ %651, %Flow3783 ], [ 0, %for.cond.i.i.i328.preheader ]
  %609 = load i32, ptr addrspace(1) @__llvm_rpc_client, align 8, !tbaa !68
  %cmp.not.i.i.i334 = icmp ult i32 %index.0.i.i.i333, %609
  %spec.store.select.i.i.i335 = select i1 %cmp.not.i.i.i334, i32 %index.0.i.i.i333, i32 0
  %610 = tail call noundef i64 @llvm.amdgcn.ballot.i64(i1 true)
  %611 = and i32 %spec.store.select.i.i.i335, 31
  %612 = lshr i64 %610, %sh_prom.i.i.i.i
  %613 = trunc i64 %612 to i32
  %conv4.i.i.i.i.i339 = and i32 %613, 1
  %shl.i.i.i.i.i340 = shl nuw i32 %conv4.i.i.i.i.i339, %611
  %614 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2474, !amdgpu.uniform !42

615:                                              ; preds = %ComputeEnd2475
  %sunkaddr3283 = getelementptr inbounds i8, ptr addrspace(1) %3147, i64 40
  %616 = atomicrmw or ptr addrspace(1) %sunkaddr3283, i32 %.lcssa4147 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %617, !amdgpu.uniform !42

617:                                              ; preds = %615, %ComputeEnd2475
  %618 = phi i32 [ %616, %615 ], [ poison, %ComputeEnd2475 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %3156)
  %619 = call i32 @llvm.amdgcn.readfirstlane.i32(i32 %618)
  %620 = or i32 %619, %.lcssa4148
  %shl5.i.i.i.i.i341 = shl nuw i32 1, %611
  %and.i.i.i.i.i342 = and i32 %shl5.i.i.i.i.i341, %620
  %tobool3.i.i.i.i343 = icmp ne i32 %and.i.i.i.i.i342, 0
  %621 = tail call i64 @llvm.amdgcn.ballot.i64(i1 %tobool3.i.i.i.i343)
  %and.i.i.i.i.i.i344 = and i64 %621, %610
  %cmp.i.not.i.i.i345 = icmp ne i64 %610, %and.i.i.i.i.i.i344
  br i1 %cmp.i.not.i.i.i345, label %if.then.i.i.i.i346, label %Flow3782, !amdgpu.uniform !42

if.then.i.i.i.i346:                               ; preds = %617
  fence syncscope("agent") acquire
  %622 = load <2 x i64>, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !75
  %623 = extractelement <2 x i64> %622, i32 0
  %624 = inttoptr i64 %623 to ptr
  %625 = extractelement <2 x i64> %622, i32 1
  %626 = inttoptr i64 %625 to ptr
  %idxprom.i.i.i.i347 = zext i32 %spec.store.select.i.i.i335 to i64
  %arrayidx.i.i.i.i348 = getelementptr inbounds nuw i32, ptr %624, i64 %idxprom.i.i.i.i347, !amdgpu.uniform !42
  %627 = load atomic i32, ptr %arrayidx.i.i.i.i348 monotonic, align 4
  %628 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %627)
  %arrayidx.i45.i.i.i349 = getelementptr inbounds nuw i32, ptr %626, i64 %idxprom.i.i.i.i347, !amdgpu.uniform !42
  %629 = load atomic i32, ptr %arrayidx.i45.i.i.i349 monotonic, align 4
  %630 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %629)
  %cmp.i46.not.i.i.i350 = icmp ne i32 %628, %630
  br i1 %cmp.i46.not.i.i.i350, label %if.then12.i.i.i351, label %Flow3780, !amdgpu.uniform !42

if.then12.i.i.i351:                               ; preds = %if.then.i.i.i.i346
  fence syncscope("agent") release
  %631 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %610, i1 true)
  %iszero.i.i.i.i.i.i.i352 = icmp ne i64 %610, 0
  %cmp2.i.i.i.i.i.i353 = icmp eq i64 %631, %sh_prom.i.i.i.i
  %cmp.i.i.i.i.i.i354 = select i1 %iszero.i.i.i.i.i.i.i352, i1 %cmp2.i.i.i.i.i.i353, i1 false
  %conv4.i.i51.i.i.i355 = zext i1 %cmp.i.i.i.i.i.i354 to i32
  %shl.i.i52.i.i.i356 = shl nuw i32 %conv4.i.i51.i.i.i355, %611
  %xor.i.i.i.i.i357 = xor i32 %shl.i.i52.i.i.i356, -1
  %632 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2479, !amdgpu.uniform !42

633:                                              ; preds = %ComputeEnd2480
  %sunkaddr3284 = getelementptr inbounds i8, ptr addrspace(1) %3147, i64 40
  %634 = atomicrmw and ptr addrspace(1) %sunkaddr3284, i32 %.lcssa4149 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %635, !amdgpu.uniform !42

635:                                              ; preds = %633, %ComputeEnd2480
  call void @llvm.amdgcn.end.cf.i64(i64 %3174)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %Flow3780, !amdgpu.uniform !42

if.end14.i.i.i417:                                ; preds = %Flow3780
  %636 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %610, i1 true)
  %iszero.i.i.i.i.i.i418 = icmp ne i64 %610, 0
  %cmp2.i.i.i.i.i419 = icmp eq i64 %636, %sh_prom.i.i.i.i
  %cmp.i.i.i.i.i420 = select i1 %iszero.i.i.i.i.i.i418, i1 %cmp2.i.i.i.i.i419, i1 false
  %637 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.i.i.i.i.i420)
  %638 = extractvalue { i1, i64 } %637, 0
  %639 = extractvalue { i1, i64 } %637, 1
  br i1 %638, label %if.then16.i.i.i422, label %if.end22.i.i.i421

if.then16.i.i.i422:                               ; preds = %if.end14.i.i.i417
  %640 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !76
  %opcode.i.i.i423 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %640, i64 %idxprom.i.i.i.i347, i32 1
  store i32 1660944387, ptr %opcode.i.i.i423, align 8, !tbaa !77
  %arrayidx21.i.i.i424 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %640, i64 %idxprom.i.i.i.i347
  store i64 %610, ptr %arrayidx21.i.i.i424, align 8, !tbaa !80
  br label %if.end22.i.i.i421, !amdgpu.uniform !42

Flow3781:                                         ; preds = %if.end22.i.i.i421, %Flow3780
  %641 = phi i32 [ %628, %if.end22.i.i.i421 ], [ %retval.sroa.8.0.i.i.i332, %Flow3780 ]
  %642 = phi i64 [ %610, %if.end22.i.i.i421 ], [ %retval.sroa.2.0.i.i.i331, %Flow3780 ]
  %643 = phi i32 [ %spec.store.select.i.i.i335, %if.end22.i.i.i421 ], [ %retval.sroa.6.0.i.i.i329, %Flow3780 ]
  br label %cleanup26.i.i.i358, !amdgpu.uniform !42

if.end22.i.i.i421:                                ; preds = %if.then16.i.i.i422, %if.end14.i.i.i417
  call void @llvm.amdgcn.end.cf.i64(i64 %639)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %Flow3781, !amdgpu.uniform !42

Flow3782:                                         ; preds = %cleanup26.i.i.i358, %617
  %644 = phi i32 [ %643, %cleanup26.i.i.i358 ], [ %606, %617 ]
  %645 = phi i64 [ %642, %cleanup26.i.i.i358 ], [ %607, %617 ]
  %646 = phi i32 [ %641, %cleanup26.i.i.i358 ], [ %608, %617 ]
  %647 = phi i32 [ %641, %cleanup26.i.i.i358 ], [ %retval.sroa.8.0.i.i.i332, %617 ]
  %648 = phi i64 [ %642, %cleanup26.i.i.i358 ], [ %retval.sroa.2.0.i.i.i331, %617 ]
  %649 = phi i32 [ %643, %cleanup26.i.i.i358 ], [ %retval.sroa.6.0.i.i.i329, %617 ]
  %650 = phi i1 [ %cmp.i46.not.i.i.i350, %cleanup26.i.i.i358 ], [ true, %617 ]
  br i1 %650, label %for.inc.i.i.i363, label %Flow3783, !amdgpu.uniform !42

cleanup26.i.i.i358:                               ; preds = %Flow3781
  br label %Flow3782, !amdgpu.uniform !42

for.inc.i.i.i363:                                 ; preds = %Flow3782
  %inc.i.i.i368 = add i32 %spec.store.select.i.i.i335, 1
  br label %Flow3783, !amdgpu.uniform !42

Flow3783:                                         ; preds = %for.inc.i.i.i363, %Flow3782
  %651 = phi i32 [ %inc.i.i.i368, %for.inc.i.i.i363 ], [ poison, %Flow3782 ]
  %652 = phi i1 [ false, %for.inc.i.i.i363 ], [ true, %Flow3782 ]
  br i1 %652, label %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i369, label %for.cond.i.i.i328, !amdgpu.uniform !42

_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i369: ; preds = %Flow3783
  %.lcssa4152 = phi i32 [ %644, %Flow3783 ]
  %.lcssa4151 = phi i64 [ %645, %Flow3783 ]
  %.lcssa4150 = phi i32 [ %646, %Flow3783 ]
  fence acquire
  %653 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %mul.i.i.i.i.i.i370 = shl i32 %.lcssa4152, 6
  %idxprom.i19.i.i.i.i.i371 = zext i32 %mul.i.i.i.i.i.i370 to i64
  %arrayidx.i20.i.i.i.i.i372 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %653, i64 %idxprom.i19.i.i.i.i.i371
  %arrayidx.i22.i.i.i.i.i373 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %arrayidx.i20.i.i.i.i.i372, i64 %sh_prom.i.i.i.i
  store i64 %spec.select.i, ptr %arrayidx.i22.i.i.i.i.i373, align 8, !tbaa !82
  %spec.select.i.i.i.i.i.i.i374 = tail call i64 @llvm.umin.i64(i64 %spec.select.i, i64 56)
  %654 = and i64 %spec.select.i.i.i.i.i.i.i374, 15
  %655 = sub i64 %spec.select.i.i.i.i.i.i.i374, %654
  %656 = icmp ne i64 %655, 0
  br i1 %656, label %loop-memcpy-expansion2254.preheader, label %Flow3779, !amdgpu.uniform !42

loop-memcpy-expansion2254.preheader:              ; preds = %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i369
  %657 = add nuw nsw i64 %75, 8
  %scevgep3145 = getelementptr i8, ptr %653, i64 %657
  %658 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i371, 6
  %scevgep3146 = getelementptr i8, ptr %scevgep3145, i64 %658
  br label %loop-memcpy-expansion2254, !amdgpu.uniform !42

Flow3779:                                         ; preds = %Flow3778, %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i369
  br label %loop-memcpy-residual-header2257, !amdgpu.uniform !42

loop-memcpy-expansion2254:                        ; preds = %loop-memcpy-expansion2254.preheader, %loop-memcpy-expansion2254
  %loop-index2255 = phi i64 [ %661, %loop-memcpy-expansion2254 ], [ 0, %loop-memcpy-expansion2254.preheader ]
  %659 = getelementptr i8, ptr %594, i64 %loop-index2255, !amdgpu.uniform !42
  %660 = load <4 x i32>, ptr %659, align 1
  %scevgep3147 = getelementptr i8, ptr %scevgep3146, i64 %loop-index2255
  store <4 x i32> %660, ptr %scevgep3147, align 1
  %661 = add i64 %loop-index2255, 16
  %662 = icmp uge i64 %661, %655
  br i1 %662, label %Flow3778, label %loop-memcpy-expansion2254, !amdgpu.uniform !42

Flow3777:                                         ; preds = %Flow3775, %Flow3776
  %663 = phi i64 [ %2785, %Flow3775 ], [ %2781, %Flow3776 ]
  br label %post-loop-memcpy-expansion2253, !amdgpu.uniform !42

loop-memcpy-residual2256:                         ; preds = %loop-memcpy-residual2256.preheader, %loop-memcpy-residual2256
  %residual-loop-index2258 = phi i64 [ %665, %loop-memcpy-residual2256 ], [ 0, %loop-memcpy-residual2256.preheader ]
  %scevgep3144 = getelementptr i8, ptr %scevgep3143, i64 %residual-loop-index2258, !amdgpu.uniform !42
  %664 = load i8, ptr %scevgep3144, align 1
  %scevgep3142 = getelementptr i8, ptr %scevgep3141, i64 %residual-loop-index2258
  store i8 %664, ptr %scevgep3142, align 1
  %665 = add i64 %residual-loop-index2258, 1
  %666 = icmp uge i64 %665, %654
  br i1 %666, label %Flow3775, label %loop-memcpy-residual2256, !amdgpu.uniform !42

Flow3775:                                         ; preds = %loop-memcpy-residual2256
  br label %Flow3777, !amdgpu.uniform !42

post-loop-memcpy-expansion2253:                   ; preds = %Flow3777
  %tobool.not.i.i.i.i.i.i376 = icmp eq i32 %.lcssa4150, 0
  %conv.i.i.i.i.i.i377 = zext i1 %tobool.not.i.i.i.i.i.i376 to i32
  fence release
  %667 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %idxprom.i23.i.i.i.i.i378 = zext i32 %.lcssa4152 to i64
  %arrayidx.i24.i.i.i.i.i379 = getelementptr inbounds nuw i32, ptr %667, i64 %idxprom.i23.i.i.i.i.i378
  store atomic i32 %conv.i.i.i.i.i.i377, ptr %arrayidx.i24.i.i.i.i.i379 monotonic, align 4
  %668 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !84
  %arrayidx.i.i5.i.i380 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %668, i64 %idxprom.i23.i.i.i.i.i378, !amdgpu.uniform !42
  %669 = load i64, ptr %arrayidx.i.i5.i.i380, align 8, !tbaa !80
  %scevgep3126 = getelementptr i8, ptr %594, i64 56
  %umin3132 = call i64 @llvm.umin.i64(i64 sub (i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) getelementptr inbounds nuw (i8, ptr addrspace(4) @.str.9, i64 8) to ptr) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.9 to ptr) to i64)), i64 2)
  %670 = sub i64 sub (i64 add (i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) getelementptr inbounds nuw (i8, ptr addrspace(4) @.str.9, i64 8) to ptr) to i64), i64 -56), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.9 to ptr) to i64)), %umin3132
  br label %while.cond.i.i.i.i381, !amdgpu.uniform !42

while.cond.i.i.i.i381:                            ; preds = %post-loop-memcpy-expansion2253, %Flow3774
  %phi.broken3908 = phi i64 [ 0, %post-loop-memcpy-expansion2253 ], [ %687, %Flow3774 ]
  %lsr.iv3133 = phi i64 [ %685, %Flow3774 ], [ %670, %post-loop-memcpy-expansion2253 ]
  %lsr.iv3127 = phi ptr [ %684, %Flow3774 ], [ %scevgep3126, %post-loop-memcpy-expansion2253 ]
  %port.sroa.37.0.i.i382 = phi i32 [ %683, %Flow3774 ], [ %conv.i.i.i.i.i.i377, %post-loop-memcpy-expansion2253 ]
  %idx.0.i.i.i.i383 = phi i64 [ %682, %Flow3774 ], [ 56, %post-loop-memcpy-expansion2253 ]
  %cmp.i.i.i.i384 = icmp ugt i64 %spec.select.i, %idx.0.i.i.i.i383
  %671 = tail call i64 @llvm.amdgcn.ballot.i64(i1 %cmp.i.i.i.i384)
  %and.i.i.i.i6.i.i385 = and i64 %671, %669
  %tobool.not.i.i.i.i386 = icmp ne i64 %and.i.i.i.i6.i.i385, 0
  %672 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i22.i.i.i387 = getelementptr inbounds nuw i32, ptr %672, i64 %idxprom.i23.i.i.i.i.i378, !amdgpu.uniform !42
  %673 = load atomic i32, ptr %arrayidx.i22.i.i.i387 monotonic, align 4
  %674 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %tobool.not.i.i.i.i386)
  %675 = extractvalue { i1, i64 } %674, 0
  %676 = extractvalue { i1, i64 } %674, 1
  br i1 %675, label %cond.false.i8.i.i.i.i388, label %Flow3774

cond.false.i8.i.i.i.i388:                         ; preds = %while.cond.i.i.i.i381
  %677 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %673)
  %678 = icmp ne i32 %677, %port.sroa.37.0.i.i382
  br i1 %678, label %while.body.i.i21.i.i.i.i389.preheader, label %Flow3773, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i389.preheader:            ; preds = %cond.false.i8.i.i.i.i388
  br label %while.body.i.i21.i.i.i.i389, !amdgpu.uniform !42

Flow3773:                                         ; preds = %Flow3772, %cond.false.i8.i.i.i.i388
  br label %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i392, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i389:                      ; preds = %while.body.i.i21.i.i.i.i389.preheader, %while.body.i.i21.i.i.i.i389
  tail call void @llvm.amdgcn.s.sleep(i32 2)
  %679 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i22.i.i.i.i390 = getelementptr inbounds nuw i32, ptr %679, i64 %idxprom.i23.i.i.i.i.i378, !amdgpu.uniform !42
  %680 = load atomic i32, ptr %arrayidx.i.i.i22.i.i.i.i390 monotonic, align 4
  %681 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %680)
  %cmp.i.not.i.i23.i.i.i.i391 = icmp eq i32 %681, %port.sroa.37.0.i.i382
  br i1 %cmp.i.not.i.i23.i.i.i.i391, label %Flow3772, label %while.body.i.i21.i.i.i.i389, !amdgpu.uniform !42

Flow3772:                                         ; preds = %while.body.i.i21.i.i.i.i389
  br label %Flow3773, !amdgpu.uniform !42

Flow3774:                                         ; preds = %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i393, %while.cond.i.i.i.i381
  %682 = phi i64 [ %add.i.i.i.i396, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i393 ], [ poison, %while.cond.i.i.i.i381 ]
  %683 = phi i32 [ %conv.i.i26.i.i.i.i394, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i393 ], [ poison, %while.cond.i.i.i.i381 ]
  %684 = phi ptr [ %scevgep3128, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i393 ], [ poison, %while.cond.i.i.i.i381 ]
  %685 = phi i64 [ %lsr.iv.next3134, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i393 ], [ poison, %while.cond.i.i.i.i381 ]
  %686 = phi i1 [ false, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i393 ], [ true, %while.cond.i.i.i.i381 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %676)
  %687 = call i64 @llvm.amdgcn.if.break.i64(i1 %686, i64 %phi.broken3908)
  %688 = call i1 @llvm.amdgcn.loop.i64(i64 %687)
  br i1 %688, label %cond.false.i.i.i403, label %while.cond.i.i.i.i381

_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i392: ; preds = %Flow3773
  fence acquire
  br i1 %cmp.i.i.i.i384, label %if.then.i.i.i.i.i.i.i397, label %Flow3771, !amdgpu.uniform !42

if.then.i.i.i.i.i.i.i397:                         ; preds = %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i392
  %sub.i.i.i.i.i.i.i399 = sub nuw i64 %spec.select.i, %idx.0.i.i.i.i383
  %spec.select.i.i.i36.i.i.i.i400 = tail call i64 @llvm.umin.i64(i64 %sub.i.i.i.i.i.i.i399, i64 64)
  %689 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %690 = and i64 %spec.select.i.i.i36.i.i.i.i400, 15
  %691 = sub i64 %spec.select.i.i.i36.i.i.i.i400, %690
  %692 = icmp ne i64 %691, 0
  br i1 %692, label %loop-memcpy-expansion2248.preheader, label %Flow3770, !amdgpu.uniform !42

loop-memcpy-expansion2248.preheader:              ; preds = %if.then.i.i.i.i.i.i.i397
  %arrayidx.i20.i35.i.i.i.i401 = getelementptr %"struct.rpc::Buffer", ptr %689, i64 %idxprom.i19.i.i.i.i.i371
  %arrayidx.i22.i38.i.i.i.i402 = getelementptr %"struct.rpc::Buffer", ptr %arrayidx.i20.i35.i.i.i.i401, i64 %sh_prom.i.i.i.i
  br label %loop-memcpy-expansion2248, !amdgpu.uniform !42

Flow3770:                                         ; preds = %Flow3769, %if.then.i.i.i.i.i.i.i397
  br label %loop-memcpy-residual-header2251, !amdgpu.uniform !42

loop-memcpy-expansion2248:                        ; preds = %loop-memcpy-expansion2248.preheader, %loop-memcpy-expansion2248
  %loop-index2249 = phi i64 [ %695, %loop-memcpy-expansion2248 ], [ 0, %loop-memcpy-expansion2248.preheader ]
  %scevgep3129 = getelementptr i8, ptr %lsr.iv3127, i64 %loop-index2249, !amdgpu.uniform !42
  %693 = load <4 x i32>, ptr %scevgep3129, align 1
  %694 = getelementptr i8, ptr %arrayidx.i22.i38.i.i.i.i402, i64 %loop-index2249
  store <4 x i32> %693, ptr %694, align 1
  %695 = add i64 %loop-index2249, 16
  %696 = icmp uge i64 %695, %691
  br i1 %696, label %Flow3769, label %loop-memcpy-expansion2248, !amdgpu.uniform !42

Flow3768:                                         ; preds = %Flow3767, %loop-memcpy-residual-header2251
  br label %Flow3771, !amdgpu.uniform !42

loop-memcpy-residual2250:                         ; preds = %loop-memcpy-residual2250.preheader, %loop-memcpy-residual2250
  %residual-loop-index2252 = phi i64 [ %698, %loop-memcpy-residual2250 ], [ 0, %loop-memcpy-residual2250.preheader ]
  %scevgep3139 = getelementptr i8, ptr %scevgep3138, i64 %residual-loop-index2252, !amdgpu.uniform !42
  %697 = load i8, ptr %scevgep3139, align 1
  %scevgep3137 = getelementptr i8, ptr %scevgep3136, i64 %residual-loop-index2252
  store i8 %697, ptr %scevgep3137, align 1
  %698 = add i64 %residual-loop-index2252, 1
  %699 = icmp uge i64 %698, %690
  br i1 %699, label %Flow3767, label %loop-memcpy-residual2250, !amdgpu.uniform !42

Flow3767:                                         ; preds = %loop-memcpy-residual2250
  br label %Flow3768, !amdgpu.uniform !42

_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i393: ; preds = %Flow3771
  %conv.i.i26.i.i.i.i394 = xor i32 %port.sroa.37.0.i.i382, 1
  fence release
  %700 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %arrayidx.i24.i29.i.i.i.i395 = getelementptr inbounds nuw i32, ptr %700, i64 %idxprom.i23.i.i.i.i.i378
  store atomic i32 %conv.i.i26.i.i.i.i394, ptr %arrayidx.i24.i29.i.i.i.i395 monotonic, align 4
  %add.i.i.i.i396 = add i64 %idx.0.i.i.i.i383, 64
  %scevgep3128 = getelementptr i8, ptr %lsr.iv3127, i64 64
  %lsr.iv.next3134 = add i64 %lsr.iv3133, -64
  br label %Flow3774, !amdgpu.uniform !42

cond.false.i.i.i403:                              ; preds = %Flow3774
  %.lcssa4146 = phi i64 [ %687, %Flow3774 ]
  %port.sroa.37.0.i.i382.lcssa = phi i32 [ %port.sroa.37.0.i.i382, %Flow3774 ]
  %.lcssa4145 = phi i32 [ %673, %Flow3774 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa4146)
  %701 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %.lcssa4145)
  %702 = icmp ne i32 %701, %port.sroa.37.0.i.i382.lcssa
  %703 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %702)
  %704 = extractvalue { i1, i64 } %703, 0
  %705 = extractvalue { i1, i64 } %703, 1
  br i1 %704, label %while.body.i.i.i.i404.preheader, label %Flow3766

while.body.i.i.i.i404.preheader:                  ; preds = %cond.false.i.i.i403
  br label %while.body.i.i.i.i404, !amdgpu.uniform !42

Flow3766:                                         ; preds = %Flow3765, %cond.false.i.i.i403
  call void @llvm.amdgcn.end.cf.i64(i64 %705)
  br label %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsINS_3cpp6stringEEERS1_T_.exit, !amdgpu.uniform !42

while.body.i.i.i.i404:                            ; preds = %while.body.i.i.i.i404.preheader, %while.body.i.i.i.i404
  %phi.broken3909 = phi i64 [ 0, %while.body.i.i.i.i404.preheader ], [ %709, %while.body.i.i.i.i404 ]
  tail call void @llvm.amdgcn.s.sleep(i32 2)
  %706 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i8.i.i405 = getelementptr inbounds nuw i32, ptr %706, i64 %idxprom.i23.i.i.i.i.i378, !amdgpu.uniform !42
  %707 = load atomic i32, ptr %arrayidx.i.i.i8.i.i405 monotonic, align 4
  %708 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %707)
  %cmp.i.not.i.i.i.i406 = icmp eq i32 %708, %port.sroa.37.0.i.i382.lcssa
  %709 = call i64 @llvm.amdgcn.if.break.i64(i1 %cmp.i.not.i.i.i.i406, i64 %phi.broken3909)
  %710 = call i1 @llvm.amdgcn.loop.i64(i64 %709)
  br i1 %710, label %Flow3765, label %while.body.i.i.i.i404

Flow3765:                                         ; preds = %while.body.i.i.i.i404
  %.lcssa4144 = phi i64 [ %709, %while.body.i.i.i.i404 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa4144)
  br label %Flow3766, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsINS_3cpp6stringEEERS1_T_.exit: ; preds = %Flow3766
  fence acquire
  tail call void @llvm.amdgcn.wave.barrier()
  fence syncscope("agent") release
  %711 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %.lcssa4151, i1 true)
  %iszero.i.i.i.i.i13.i.i408 = icmp ne i64 %.lcssa4151, 0
  %cmp2.i.i.i.i14.i.i409 = icmp eq i64 %711, %sh_prom.i.i.i.i
  %cmp.i.i.i.i15.i.i410 = select i1 %iszero.i.i.i.i.i13.i.i408, i1 %cmp2.i.i.i.i14.i.i409, i1 false
  %712 = and i32 %.lcssa4152, 31
  %conv4.i.i.i19.i.i414 = zext i1 %cmp.i.i.i.i15.i.i410 to i32
  %shl.i.i.i20.i.i415 = shl nuw i32 %conv4.i.i.i19.i.i414, %712
  %xor.i.i.i21.i.i416 = xor i32 %shl.i.i.i20.i.i415, -1
  %713 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2483, !amdgpu.uniform !42

714:                                              ; preds = %ComputeEnd2484
  %div8.i.i.i16.i.i411 = lshr i32 %.lcssa4152, 5
  %idxprom.i.i.i17.i.i412 = zext nneg i32 %div8.i.i.i16.i.i411 to i64
  %715 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i17.i.i412
  %sunkaddr3285 = getelementptr inbounds i8, ptr addrspace(1) %715, i64 40
  %716 = atomicrmw and ptr addrspace(1) %sunkaddr3285, i32 %.lcssa4143 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %717, !amdgpu.uniform !42

717:                                              ; preds = %714, %ComputeEnd2484
  call void @llvm.amdgcn.end.cf.i64(i64 %3191)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %for.cond.i.i1.i433, !amdgpu.uniform !42

for.cond.i.i1.i433:                               ; preds = %717, %Flow3764
  %718 = phi i32 [ %756, %Flow3764 ], [ poison, %717 ]
  %719 = phi i64 [ %757, %Flow3764 ], [ poison, %717 ]
  %720 = phi i32 [ %758, %Flow3764 ], [ poison, %717 ]
  %retval.sroa.6.0.i.i.i434 = phi i32 [ %761, %Flow3764 ], [ undef, %717 ]
  %retval.sroa.2.0.i.i.i436 = phi i64 [ %760, %Flow3764 ], [ undef, %717 ]
  %retval.sroa.8.0.i.i.i437 = phi i32 [ %759, %Flow3764 ], [ undef, %717 ]
  %index.0.i.i.i438 = phi i32 [ %763, %Flow3764 ], [ 0, %717 ]
  %721 = load i32, ptr addrspace(1) @__llvm_rpc_client, align 8, !tbaa !68
  %cmp.not.i.i.i439 = icmp ult i32 %index.0.i.i.i438, %721
  %spec.store.select.i.i.i440 = select i1 %cmp.not.i.i.i439, i32 %index.0.i.i.i438, i32 0
  %722 = tail call noundef i64 @llvm.amdgcn.ballot.i64(i1 true)
  %723 = and i32 %spec.store.select.i.i.i440, 31
  %724 = lshr i64 %722, %sh_prom.i.i.i.i
  %725 = trunc i64 %724 to i32
  %conv4.i.i.i.i.i444 = and i32 %725, 1
  %shl.i.i.i.i.i445 = shl nuw i32 %conv4.i.i.i.i.i444, %723
  %726 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2487, !amdgpu.uniform !42

727:                                              ; preds = %ComputeEnd2488
  %sunkaddr3286 = getelementptr inbounds i8, ptr addrspace(1) %3201, i64 40
  %728 = atomicrmw or ptr addrspace(1) %sunkaddr3286, i32 %.lcssa4137 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %729, !amdgpu.uniform !42

729:                                              ; preds = %727, %ComputeEnd2488
  %730 = phi i32 [ %728, %727 ], [ poison, %ComputeEnd2488 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %3210)
  %731 = call i32 @llvm.amdgcn.readfirstlane.i32(i32 %730)
  %732 = or i32 %731, %.lcssa4138
  %shl5.i.i.i.i.i446 = shl nuw i32 1, %723
  %and.i.i.i.i.i447 = and i32 %shl5.i.i.i.i.i446, %732
  %tobool3.i.i.i.i448 = icmp ne i32 %and.i.i.i.i.i447, 0
  %733 = tail call i64 @llvm.amdgcn.ballot.i64(i1 %tobool3.i.i.i.i448)
  %and.i.i.i.i.i.i449 = and i64 %733, %722
  %cmp.i.not.i.i.i450 = icmp ne i64 %722, %and.i.i.i.i.i.i449
  br i1 %cmp.i.not.i.i.i450, label %if.then.i.i.i.i451, label %Flow3763, !amdgpu.uniform !42

if.then.i.i.i.i451:                               ; preds = %729
  fence syncscope("agent") acquire
  %734 = load <2 x i64>, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !75
  %735 = extractelement <2 x i64> %734, i32 0
  %736 = inttoptr i64 %735 to ptr
  %737 = extractelement <2 x i64> %734, i32 1
  %738 = inttoptr i64 %737 to ptr
  %idxprom.i.i.i.i452 = zext i32 %spec.store.select.i.i.i440 to i64
  %arrayidx.i.i.i.i453 = getelementptr inbounds nuw i32, ptr %736, i64 %idxprom.i.i.i.i452, !amdgpu.uniform !42
  %739 = load atomic i32, ptr %arrayidx.i.i.i.i453 monotonic, align 4
  %740 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %739)
  %arrayidx.i45.i.i.i454 = getelementptr inbounds nuw i32, ptr %738, i64 %idxprom.i.i.i.i452, !amdgpu.uniform !42
  %741 = load atomic i32, ptr %arrayidx.i45.i.i.i454 monotonic, align 4
  %742 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %741)
  %cmp.i46.not.i.i.i455 = icmp ne i32 %740, %742
  br i1 %cmp.i46.not.i.i.i455, label %if.then12.i.i.i456, label %Flow3761, !amdgpu.uniform !42

if.then12.i.i.i456:                               ; preds = %if.then.i.i.i.i451
  fence syncscope("agent") release
  %743 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %722, i1 true)
  %iszero.i.i.i.i.i.i.i457 = icmp ne i64 %722, 0
  %cmp2.i.i.i.i.i.i458 = icmp eq i64 %743, %sh_prom.i.i.i.i
  %cmp.i.i.i.i.i.i459 = select i1 %iszero.i.i.i.i.i.i.i457, i1 %cmp2.i.i.i.i.i.i458, i1 false
  %conv4.i.i51.i.i.i460 = zext i1 %cmp.i.i.i.i.i.i459 to i32
  %shl.i.i52.i.i.i461 = shl nuw i32 %conv4.i.i51.i.i.i460, %723
  %xor.i.i.i.i.i462 = xor i32 %shl.i.i52.i.i.i461, -1
  %744 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2492, !amdgpu.uniform !42

745:                                              ; preds = %ComputeEnd2493
  %sunkaddr3287 = getelementptr inbounds i8, ptr addrspace(1) %3201, i64 40
  %746 = atomicrmw and ptr addrspace(1) %sunkaddr3287, i32 %.lcssa4139 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %747, !amdgpu.uniform !42

747:                                              ; preds = %745, %ComputeEnd2493
  call void @llvm.amdgcn.end.cf.i64(i64 %3228)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %Flow3761, !amdgpu.uniform !42

if.end14.i.i.i522:                                ; preds = %Flow3761
  %748 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %722, i1 true)
  %iszero.i.i.i.i.i.i523 = icmp ne i64 %722, 0
  %cmp2.i.i.i.i.i524 = icmp eq i64 %748, %sh_prom.i.i.i.i
  %cmp.i.i.i.i.i525 = select i1 %iszero.i.i.i.i.i.i523, i1 %cmp2.i.i.i.i.i524, i1 false
  %749 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.i.i.i.i.i525)
  %750 = extractvalue { i1, i64 } %749, 0
  %751 = extractvalue { i1, i64 } %749, 1
  br i1 %750, label %if.then16.i.i.i527, label %if.end22.i.i.i526

if.then16.i.i.i527:                               ; preds = %if.end14.i.i.i522
  %752 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !76
  %opcode.i.i.i528 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %752, i64 %idxprom.i.i.i.i452, i32 1
  store i32 1660944387, ptr %opcode.i.i.i528, align 8, !tbaa !77
  %arrayidx21.i.i.i529 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %752, i64 %idxprom.i.i.i.i452
  store i64 %722, ptr %arrayidx21.i.i.i529, align 8, !tbaa !80
  br label %if.end22.i.i.i526, !amdgpu.uniform !42

Flow3762:                                         ; preds = %if.end22.i.i.i526, %Flow3761
  %753 = phi i32 [ %740, %if.end22.i.i.i526 ], [ %retval.sroa.8.0.i.i.i437, %Flow3761 ]
  %754 = phi i64 [ %722, %if.end22.i.i.i526 ], [ %retval.sroa.2.0.i.i.i436, %Flow3761 ]
  %755 = phi i32 [ %spec.store.select.i.i.i440, %if.end22.i.i.i526 ], [ %retval.sroa.6.0.i.i.i434, %Flow3761 ]
  br label %cleanup26.i.i.i463, !amdgpu.uniform !42

if.end22.i.i.i526:                                ; preds = %if.then16.i.i.i527, %if.end14.i.i.i522
  call void @llvm.amdgcn.end.cf.i64(i64 %751)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %Flow3762, !amdgpu.uniform !42

Flow3763:                                         ; preds = %cleanup26.i.i.i463, %729
  %756 = phi i32 [ %755, %cleanup26.i.i.i463 ], [ %718, %729 ]
  %757 = phi i64 [ %754, %cleanup26.i.i.i463 ], [ %719, %729 ]
  %758 = phi i32 [ %753, %cleanup26.i.i.i463 ], [ %720, %729 ]
  %759 = phi i32 [ %753, %cleanup26.i.i.i463 ], [ %retval.sroa.8.0.i.i.i437, %729 ]
  %760 = phi i64 [ %754, %cleanup26.i.i.i463 ], [ %retval.sroa.2.0.i.i.i436, %729 ]
  %761 = phi i32 [ %755, %cleanup26.i.i.i463 ], [ %retval.sroa.6.0.i.i.i434, %729 ]
  %762 = phi i1 [ %cmp.i46.not.i.i.i455, %cleanup26.i.i.i463 ], [ true, %729 ]
  br i1 %762, label %for.inc.i.i.i468, label %Flow3764, !amdgpu.uniform !42

cleanup26.i.i.i463:                               ; preds = %Flow3762
  br label %Flow3763, !amdgpu.uniform !42

for.inc.i.i.i468:                                 ; preds = %Flow3763
  %inc.i.i.i473 = add i32 %spec.store.select.i.i.i440, 1
  br label %Flow3764, !amdgpu.uniform !42

Flow3764:                                         ; preds = %for.inc.i.i.i468, %Flow3763
  %763 = phi i32 [ %inc.i.i.i473, %for.inc.i.i.i468 ], [ poison, %Flow3763 ]
  %764 = phi i1 [ false, %for.inc.i.i.i468 ], [ true, %Flow3763 ]
  br i1 %764, label %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i474, label %for.cond.i.i1.i433, !amdgpu.uniform !42

_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i474: ; preds = %Flow3764
  %.lcssa4142 = phi i32 [ %756, %Flow3764 ]
  %.lcssa4141 = phi i64 [ %757, %Flow3764 ]
  %.lcssa4140 = phi i32 [ %758, %Flow3764 ]
  fence acquire
  %765 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %mul.i.i.i.i.i.i475 = shl i32 %.lcssa4142, 6
  %idxprom.i19.i.i.i.i.i476 = zext i32 %mul.i.i.i.i.i.i475 to i64
  %arrayidx.i20.i.i.i.i.i477 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %765, i64 %idxprom.i19.i.i.i.i.i476
  %arrayidx.i22.i.i.i.i.i478 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %arrayidx.i20.i.i.i.i.i477, i64 %sh_prom.i.i.i.i
  store i64 sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.34.12 to ptr), i64 10) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.34.12 to ptr) to i64)), ptr %arrayidx.i22.i.i.i.i.i478, align 8, !tbaa !82
  %spec.select.i.i.i.i.i.i.i479 = tail call i64 @llvm.umin.i64(i64 sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.34.12 to ptr), i64 10) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.34.12 to ptr) to i64)), i64 56)
  %766 = and i64 %spec.select.i.i.i.i.i.i.i479, 15
  %767 = sub i64 %spec.select.i.i.i.i.i.i.i479, %766
  %768 = icmp ne i64 %767, 0
  br i1 %768, label %loop-memcpy-expansion2344.preheader, label %Flow3760, !amdgpu.uniform !42

loop-memcpy-expansion2344.preheader:              ; preds = %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i474
  %769 = add nuw nsw i64 %75, 8
  %scevgep3123 = getelementptr i8, ptr %765, i64 %769
  %770 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i476, 6
  %scevgep3124 = getelementptr i8, ptr %scevgep3123, i64 %770
  br label %loop-memcpy-expansion2344, !amdgpu.uniform !42

Flow3760:                                         ; preds = %Flow3759, %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i474
  br label %loop-memcpy-residual-header2347, !amdgpu.uniform !42

loop-memcpy-expansion2344:                        ; preds = %loop-memcpy-expansion2344.preheader, %loop-memcpy-expansion2344
  %loop-index2345 = phi i64 [ %773, %loop-memcpy-expansion2344 ], [ 0, %loop-memcpy-expansion2344.preheader ]
  %771 = getelementptr i8, ptr addrspace(4) @.str.34.12, i64 %loop-index2345, !amdgpu.uniform !42
  %772 = load <4 x i32>, ptr addrspace(4) %771, align 1
  %scevgep3125 = getelementptr i8, ptr %scevgep3124, i64 %loop-index2345
  store <4 x i32> %772, ptr %scevgep3125, align 1
  %773 = add i64 %loop-index2345, 16
  %774 = icmp uge i64 %773, %767
  br i1 %774, label %Flow3759, label %loop-memcpy-expansion2344, !amdgpu.uniform !42

Flow3758:                                         ; preds = %Flow3756, %Flow3757
  %775 = phi i64 [ %2868, %Flow3756 ], [ %2865, %Flow3757 ]
  br label %post-loop-memcpy-expansion2343, !amdgpu.uniform !42

loop-memcpy-residual2346:                         ; preds = %loop-memcpy-residual2346.preheader, %loop-memcpy-residual2346
  %residual-loop-index2348 = phi i64 [ %777, %loop-memcpy-residual2346 ], [ 0, %loop-memcpy-residual2346.preheader ]
  %scevgep3122 = getelementptr i8, ptr addrspace(4) @.str.34.12, i64 %residual-loop-index2348, !amdgpu.uniform !42
  %776 = load i8, ptr addrspace(4) %scevgep3122, align 1
  %scevgep3121 = getelementptr i8, ptr %scevgep3120, i64 %residual-loop-index2348
  store i8 %776, ptr %scevgep3121, align 1
  %777 = add i64 %residual-loop-index2348, 1
  %778 = icmp uge i64 %777, %766
  br i1 %778, label %Flow3756, label %loop-memcpy-residual2346, !amdgpu.uniform !42

Flow3756:                                         ; preds = %loop-memcpy-residual2346
  br label %Flow3758, !amdgpu.uniform !42

post-loop-memcpy-expansion2343:                   ; preds = %Flow3758
  %tobool.not.i.i.i.i.i.i481 = icmp eq i32 %.lcssa4140, 0
  %conv.i.i.i.i.i.i482 = zext i1 %tobool.not.i.i.i.i.i.i481 to i32
  fence release
  %779 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %idxprom.i23.i.i.i.i.i483 = zext i32 %.lcssa4142 to i64
  %arrayidx.i24.i.i.i.i.i484 = getelementptr inbounds nuw i32, ptr %779, i64 %idxprom.i23.i.i.i.i.i483
  store atomic i32 %conv.i.i.i.i.i.i482, ptr %arrayidx.i24.i.i.i.i.i484 monotonic, align 4
  %780 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !84
  %arrayidx.i.i5.i.i485 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %780, i64 %idxprom.i23.i.i.i.i.i483, !amdgpu.uniform !42
  %781 = load i64, ptr %arrayidx.i.i5.i.i485, align 8, !tbaa !80
  br label %while.cond.i.i.i.i486, !amdgpu.uniform !42

while.cond.i.i.i.i486:                            ; preds = %post-loop-memcpy-expansion2343, %Flow3755
  %phi.broken3910 = phi i64 [ 0, %post-loop-memcpy-expansion2343 ], [ %798, %Flow3755 ]
  %lsr.iv3112 = phi i64 [ %796, %Flow3755 ], [ -46, %post-loop-memcpy-expansion2343 ]
  %lsr.iv3107 = phi ptr addrspace(4) [ %795, %Flow3755 ], [ getelementptr (i8, ptr addrspace(4) @.str.34.12, i64 56), %post-loop-memcpy-expansion2343 ]
  %port.sroa.37.0.i.i487 = phi i32 [ %794, %Flow3755 ], [ %conv.i.i.i.i.i.i482, %post-loop-memcpy-expansion2343 ]
  %idx.0.i.i.i.i488 = phi i64 [ %793, %Flow3755 ], [ 56, %post-loop-memcpy-expansion2343 ]
  %cmp.i.i.i.i489 = icmp ult i64 %idx.0.i.i.i.i488, sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.34.12 to ptr), i64 10) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.34.12 to ptr) to i64))
  %782 = tail call i64 @llvm.amdgcn.ballot.i64(i1 %cmp.i.i.i.i489)
  %and.i.i.i.i6.i.i490 = and i64 %782, %781
  %tobool.not.i.i.i.i491 = icmp ne i64 %and.i.i.i.i6.i.i490, 0
  %783 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i22.i.i.i492 = getelementptr inbounds nuw i32, ptr %783, i64 %idxprom.i23.i.i.i.i.i483, !amdgpu.uniform !42
  %784 = load atomic i32, ptr %arrayidx.i22.i.i.i492 monotonic, align 4
  %785 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %tobool.not.i.i.i.i491)
  %786 = extractvalue { i1, i64 } %785, 0
  %787 = extractvalue { i1, i64 } %785, 1
  br i1 %786, label %cond.false.i8.i.i.i.i493, label %Flow3755

cond.false.i8.i.i.i.i493:                         ; preds = %while.cond.i.i.i.i486
  %788 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %784)
  %789 = icmp ne i32 %788, %port.sroa.37.0.i.i487
  br i1 %789, label %while.body.i.i21.i.i.i.i494.preheader, label %Flow3754, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i494.preheader:            ; preds = %cond.false.i8.i.i.i.i493
  br label %while.body.i.i21.i.i.i.i494, !amdgpu.uniform !42

Flow3754:                                         ; preds = %Flow3753, %cond.false.i8.i.i.i.i493
  br label %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i497, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i494:                      ; preds = %while.body.i.i21.i.i.i.i494.preheader, %while.body.i.i21.i.i.i.i494
  tail call void @llvm.amdgcn.s.sleep(i32 2)
  %790 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i22.i.i.i.i495 = getelementptr inbounds nuw i32, ptr %790, i64 %idxprom.i23.i.i.i.i.i483, !amdgpu.uniform !42
  %791 = load atomic i32, ptr %arrayidx.i.i.i22.i.i.i.i495 monotonic, align 4
  %792 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %791)
  %cmp.i.not.i.i23.i.i.i.i496 = icmp eq i32 %792, %port.sroa.37.0.i.i487
  br i1 %cmp.i.not.i.i23.i.i.i.i496, label %Flow3753, label %while.body.i.i21.i.i.i.i494, !amdgpu.uniform !42

Flow3753:                                         ; preds = %while.body.i.i21.i.i.i.i494
  br label %Flow3754, !amdgpu.uniform !42

Flow3755:                                         ; preds = %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i498, %while.cond.i.i.i.i486
  %793 = phi i64 [ %add.i.i.i.i501, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i498 ], [ poison, %while.cond.i.i.i.i486 ]
  %794 = phi i32 [ %conv.i.i26.i.i.i.i499, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i498 ], [ poison, %while.cond.i.i.i.i486 ]
  %795 = phi ptr addrspace(4) [ %scevgep3108, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i498 ], [ poison, %while.cond.i.i.i.i486 ]
  %796 = phi i64 [ %lsr.iv.next3113, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i498 ], [ poison, %while.cond.i.i.i.i486 ]
  %797 = phi i1 [ false, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i498 ], [ true, %while.cond.i.i.i.i486 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %787)
  %798 = call i64 @llvm.amdgcn.if.break.i64(i1 %797, i64 %phi.broken3910)
  %799 = call i1 @llvm.amdgcn.loop.i64(i64 %798)
  br i1 %799, label %cond.false.i.i.i508, label %while.cond.i.i.i.i486

_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i497: ; preds = %Flow3754
  fence acquire
  br i1 %cmp.i.i.i.i489, label %if.then.i.i.i.i.i.i.i502, label %Flow3752, !amdgpu.uniform !42

if.then.i.i.i.i.i.i.i502:                         ; preds = %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i497
  %sub.i.i.i.i.i.i.i504 = sub nuw i64 sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.34.12 to ptr), i64 10) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.34.12 to ptr) to i64)), %idx.0.i.i.i.i488
  %spec.select.i.i.i36.i.i.i.i505 = tail call i64 @llvm.umin.i64(i64 %sub.i.i.i.i.i.i.i504, i64 64)
  %800 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %801 = and i64 %spec.select.i.i.i36.i.i.i.i505, 15
  %802 = sub i64 %spec.select.i.i.i36.i.i.i.i505, %801
  %803 = icmp ne i64 %802, 0
  br i1 %803, label %loop-memcpy-expansion2350.preheader, label %Flow3751, !amdgpu.uniform !42

loop-memcpy-expansion2350.preheader:              ; preds = %if.then.i.i.i.i.i.i.i502
  %arrayidx.i20.i35.i.i.i.i506 = getelementptr %"struct.rpc::Buffer", ptr %800, i64 %idxprom.i19.i.i.i.i.i476
  %arrayidx.i22.i38.i.i.i.i507 = getelementptr %"struct.rpc::Buffer", ptr %arrayidx.i20.i35.i.i.i.i506, i64 %sh_prom.i.i.i.i
  br label %loop-memcpy-expansion2350, !amdgpu.uniform !42

Flow3751:                                         ; preds = %Flow3750, %if.then.i.i.i.i.i.i.i502
  br label %loop-memcpy-residual-header2353, !amdgpu.uniform !42

loop-memcpy-expansion2350:                        ; preds = %loop-memcpy-expansion2350.preheader, %loop-memcpy-expansion2350
  %loop-index2351 = phi i64 [ %806, %loop-memcpy-expansion2350 ], [ 0, %loop-memcpy-expansion2350.preheader ]
  %scevgep3109 = getelementptr i8, ptr addrspace(4) %lsr.iv3107, i64 %loop-index2351, !amdgpu.uniform !42
  %804 = load <4 x i32>, ptr addrspace(4) %scevgep3109, align 1
  %805 = getelementptr i8, ptr %arrayidx.i22.i38.i.i.i.i507, i64 %loop-index2351
  store <4 x i32> %804, ptr %805, align 1
  %806 = add i64 %loop-index2351, 16
  %807 = icmp uge i64 %806, %802
  br i1 %807, label %Flow3750, label %loop-memcpy-expansion2350, !amdgpu.uniform !42

Flow3749:                                         ; preds = %Flow3748, %loop-memcpy-residual-header2353
  br label %Flow3752, !amdgpu.uniform !42

loop-memcpy-residual2352:                         ; preds = %loop-memcpy-residual2352.preheader, %loop-memcpy-residual2352
  %residual-loop-index2354 = phi i64 [ %809, %loop-memcpy-residual2352 ], [ 0, %loop-memcpy-residual2352.preheader ]
  %scevgep3118 = getelementptr i8, ptr addrspace(4) %scevgep3117, i64 %residual-loop-index2354, !amdgpu.uniform !42
  %808 = load i8, ptr addrspace(4) %scevgep3118, align 1
  %scevgep3116 = getelementptr i8, ptr %scevgep3115, i64 %residual-loop-index2354
  store i8 %808, ptr %scevgep3116, align 1
  %809 = add i64 %residual-loop-index2354, 1
  %810 = icmp uge i64 %809, %801
  br i1 %810, label %Flow3748, label %loop-memcpy-residual2352, !amdgpu.uniform !42

Flow3748:                                         ; preds = %loop-memcpy-residual2352
  br label %Flow3749, !amdgpu.uniform !42

_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i498: ; preds = %Flow3752
  %conv.i.i26.i.i.i.i499 = xor i32 %port.sroa.37.0.i.i487, 1
  fence release
  %811 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %arrayidx.i24.i29.i.i.i.i500 = getelementptr inbounds nuw i32, ptr %811, i64 %idxprom.i23.i.i.i.i.i483
  store atomic i32 %conv.i.i26.i.i.i.i499, ptr %arrayidx.i24.i29.i.i.i.i500 monotonic, align 4
  %add.i.i.i.i501 = add i64 %idx.0.i.i.i.i488, 64
  %scevgep3108 = getelementptr i8, ptr addrspace(4) %lsr.iv3107, i64 64
  %lsr.iv.next3113 = add i64 %lsr.iv3112, -64
  br label %Flow3755, !amdgpu.uniform !42

cond.false.i.i.i508:                              ; preds = %Flow3755
  %.lcssa4136 = phi i64 [ %798, %Flow3755 ]
  %port.sroa.37.0.i.i487.lcssa = phi i32 [ %port.sroa.37.0.i.i487, %Flow3755 ]
  %.lcssa4135 = phi i32 [ %784, %Flow3755 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa4136)
  %812 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %.lcssa4135)
  %813 = icmp ne i32 %812, %port.sroa.37.0.i.i487.lcssa
  %814 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %813)
  %815 = extractvalue { i1, i64 } %814, 0
  %816 = extractvalue { i1, i64 } %814, 1
  br i1 %815, label %while.body.i.i.i.i509.preheader, label %Flow3747

while.body.i.i.i.i509.preheader:                  ; preds = %cond.false.i.i.i508
  br label %while.body.i.i.i.i509, !amdgpu.uniform !42

Flow3747:                                         ; preds = %Flow3746, %cond.false.i.i.i508
  call void @llvm.amdgcn.end.cf.i64(i64 %816)
  br label %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit530, !amdgpu.uniform !42

while.body.i.i.i.i509:                            ; preds = %while.body.i.i.i.i509.preheader, %while.body.i.i.i.i509
  %phi.broken3911 = phi i64 [ 0, %while.body.i.i.i.i509.preheader ], [ %820, %while.body.i.i.i.i509 ]
  tail call void @llvm.amdgcn.s.sleep(i32 2)
  %817 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i8.i.i510 = getelementptr inbounds nuw i32, ptr %817, i64 %idxprom.i23.i.i.i.i.i483, !amdgpu.uniform !42
  %818 = load atomic i32, ptr %arrayidx.i.i.i8.i.i510 monotonic, align 4
  %819 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %818)
  %cmp.i.not.i.i.i.i511 = icmp eq i32 %819, %port.sroa.37.0.i.i487.lcssa
  %820 = call i64 @llvm.amdgcn.if.break.i64(i1 %cmp.i.not.i.i.i.i511, i64 %phi.broken3911)
  %821 = call i1 @llvm.amdgcn.loop.i64(i64 %820)
  br i1 %821, label %Flow3746, label %while.body.i.i.i.i509

Flow3746:                                         ; preds = %while.body.i.i.i.i509
  %.lcssa4134 = phi i64 [ %820, %while.body.i.i.i.i509 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa4134)
  br label %Flow3747, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit530: ; preds = %Flow3747
  fence acquire
  tail call void @llvm.amdgcn.wave.barrier()
  fence syncscope("agent") release
  %822 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %.lcssa4141, i1 true)
  %iszero.i.i.i.i.i13.i.i513 = icmp ne i64 %.lcssa4141, 0
  %cmp2.i.i.i.i14.i.i514 = icmp eq i64 %822, %sh_prom.i.i.i.i
  %cmp.i.i.i.i15.i.i515 = select i1 %iszero.i.i.i.i.i13.i.i513, i1 %cmp2.i.i.i.i14.i.i514, i1 false
  %823 = and i32 %.lcssa4142, 31
  %conv4.i.i.i19.i.i519 = zext i1 %cmp.i.i.i.i15.i.i515 to i32
  %shl.i.i.i20.i.i520 = shl nuw i32 %conv4.i.i.i19.i.i519, %823
  %xor.i.i.i21.i.i521 = xor i32 %shl.i.i.i20.i.i520, -1
  %824 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2496, !amdgpu.uniform !42

825:                                              ; preds = %ComputeEnd2497
  %div8.i.i.i16.i.i516 = lshr i32 %.lcssa4142, 5
  %idxprom.i.i.i17.i.i517 = zext nneg i32 %div8.i.i.i16.i.i516 to i64
  %826 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i17.i.i517
  %sunkaddr3288 = getelementptr inbounds i8, ptr addrspace(1) %826, i64 40
  %827 = atomicrmw and ptr addrspace(1) %sunkaddr3288, i32 %.lcssa4133 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %828, !amdgpu.uniform !42

828:                                              ; preds = %825, %ComputeEnd2497
  call void @llvm.amdgcn.end.cf.i64(i64 %3245)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %for.cond.i.i1.i539, !amdgpu.uniform !42

for.cond.i.i1.i539:                               ; preds = %828, %Flow3745
  %829 = phi i32 [ %867, %Flow3745 ], [ poison, %828 ]
  %830 = phi i64 [ %868, %Flow3745 ], [ poison, %828 ]
  %831 = phi i32 [ %869, %Flow3745 ], [ poison, %828 ]
  %retval.sroa.6.0.i.i.i540 = phi i32 [ %872, %Flow3745 ], [ undef, %828 ]
  %retval.sroa.2.0.i.i.i542 = phi i64 [ %871, %Flow3745 ], [ undef, %828 ]
  %retval.sroa.8.0.i.i.i543 = phi i32 [ %870, %Flow3745 ], [ undef, %828 ]
  %index.0.i.i.i544 = phi i32 [ %874, %Flow3745 ], [ 0, %828 ]
  %832 = load i32, ptr addrspace(1) @__llvm_rpc_client, align 8, !tbaa !68
  %cmp.not.i.i.i545 = icmp ult i32 %index.0.i.i.i544, %832
  %spec.store.select.i.i.i546 = select i1 %cmp.not.i.i.i545, i32 %index.0.i.i.i544, i32 0
  %833 = tail call noundef i64 @llvm.amdgcn.ballot.i64(i1 true)
  %834 = and i32 %spec.store.select.i.i.i546, 31
  %835 = lshr i64 %833, %sh_prom.i.i.i.i
  %836 = trunc i64 %835 to i32
  %conv4.i.i.i.i.i550 = and i32 %836, 1
  %shl.i.i.i.i.i551 = shl nuw i32 %conv4.i.i.i.i.i550, %834
  %837 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2500, !amdgpu.uniform !42

838:                                              ; preds = %ComputeEnd2501
  %sunkaddr3289 = getelementptr inbounds i8, ptr addrspace(1) %3255, i64 40
  %839 = atomicrmw or ptr addrspace(1) %sunkaddr3289, i32 %.lcssa4127 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %840, !amdgpu.uniform !42

840:                                              ; preds = %838, %ComputeEnd2501
  %841 = phi i32 [ %839, %838 ], [ poison, %ComputeEnd2501 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %3264)
  %842 = call i32 @llvm.amdgcn.readfirstlane.i32(i32 %841)
  %843 = or i32 %842, %.lcssa4128
  %shl5.i.i.i.i.i552 = shl nuw i32 1, %834
  %and.i.i.i.i.i553 = and i32 %shl5.i.i.i.i.i552, %843
  %tobool3.i.i.i.i554 = icmp ne i32 %and.i.i.i.i.i553, 0
  %844 = tail call i64 @llvm.amdgcn.ballot.i64(i1 %tobool3.i.i.i.i554)
  %and.i.i.i.i.i.i555 = and i64 %844, %833
  %cmp.i.not.i.i.i556 = icmp ne i64 %833, %and.i.i.i.i.i.i555
  br i1 %cmp.i.not.i.i.i556, label %if.then.i.i.i.i557, label %Flow3744, !amdgpu.uniform !42

if.then.i.i.i.i557:                               ; preds = %840
  fence syncscope("agent") acquire
  %845 = load <2 x i64>, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !75
  %846 = extractelement <2 x i64> %845, i32 0
  %847 = inttoptr i64 %846 to ptr
  %848 = extractelement <2 x i64> %845, i32 1
  %849 = inttoptr i64 %848 to ptr
  %idxprom.i.i.i.i558 = zext i32 %spec.store.select.i.i.i546 to i64
  %arrayidx.i.i.i.i559 = getelementptr inbounds nuw i32, ptr %847, i64 %idxprom.i.i.i.i558, !amdgpu.uniform !42
  %850 = load atomic i32, ptr %arrayidx.i.i.i.i559 monotonic, align 4
  %851 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %850)
  %arrayidx.i45.i.i.i560 = getelementptr inbounds nuw i32, ptr %849, i64 %idxprom.i.i.i.i558, !amdgpu.uniform !42
  %852 = load atomic i32, ptr %arrayidx.i45.i.i.i560 monotonic, align 4
  %853 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %852)
  %cmp.i46.not.i.i.i561 = icmp ne i32 %851, %853
  br i1 %cmp.i46.not.i.i.i561, label %if.then12.i.i.i562, label %Flow3742, !amdgpu.uniform !42

if.then12.i.i.i562:                               ; preds = %if.then.i.i.i.i557
  fence syncscope("agent") release
  %854 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %833, i1 true)
  %iszero.i.i.i.i.i.i.i563 = icmp ne i64 %833, 0
  %cmp2.i.i.i.i.i.i564 = icmp eq i64 %854, %sh_prom.i.i.i.i
  %cmp.i.i.i.i.i.i565 = select i1 %iszero.i.i.i.i.i.i.i563, i1 %cmp2.i.i.i.i.i.i564, i1 false
  %conv4.i.i51.i.i.i566 = zext i1 %cmp.i.i.i.i.i.i565 to i32
  %shl.i.i52.i.i.i567 = shl nuw i32 %conv4.i.i51.i.i.i566, %834
  %xor.i.i.i.i.i568 = xor i32 %shl.i.i52.i.i.i567, -1
  %855 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2505, !amdgpu.uniform !42

856:                                              ; preds = %ComputeEnd2506
  %sunkaddr3290 = getelementptr inbounds i8, ptr addrspace(1) %3255, i64 40
  %857 = atomicrmw and ptr addrspace(1) %sunkaddr3290, i32 %.lcssa4129 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %858, !amdgpu.uniform !42

858:                                              ; preds = %856, %ComputeEnd2506
  call void @llvm.amdgcn.end.cf.i64(i64 %3282)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %Flow3742, !amdgpu.uniform !42

if.end14.i.i.i628:                                ; preds = %Flow3742
  %859 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %833, i1 true)
  %iszero.i.i.i.i.i.i629 = icmp ne i64 %833, 0
  %cmp2.i.i.i.i.i630 = icmp eq i64 %859, %sh_prom.i.i.i.i
  %cmp.i.i.i.i.i631 = select i1 %iszero.i.i.i.i.i.i629, i1 %cmp2.i.i.i.i.i630, i1 false
  %860 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.i.i.i.i.i631)
  %861 = extractvalue { i1, i64 } %860, 0
  %862 = extractvalue { i1, i64 } %860, 1
  br i1 %861, label %if.then16.i.i.i633, label %if.end22.i.i.i632

if.then16.i.i.i633:                               ; preds = %if.end14.i.i.i628
  %863 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !76
  %opcode.i.i.i634 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %863, i64 %idxprom.i.i.i.i558, i32 1
  store i32 1660944387, ptr %opcode.i.i.i634, align 8, !tbaa !77
  %arrayidx21.i.i.i635 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %863, i64 %idxprom.i.i.i.i558
  store i64 %833, ptr %arrayidx21.i.i.i635, align 8, !tbaa !80
  br label %if.end22.i.i.i632, !amdgpu.uniform !42

Flow3743:                                         ; preds = %if.end22.i.i.i632, %Flow3742
  %864 = phi i32 [ %851, %if.end22.i.i.i632 ], [ %retval.sroa.8.0.i.i.i543, %Flow3742 ]
  %865 = phi i64 [ %833, %if.end22.i.i.i632 ], [ %retval.sroa.2.0.i.i.i542, %Flow3742 ]
  %866 = phi i32 [ %spec.store.select.i.i.i546, %if.end22.i.i.i632 ], [ %retval.sroa.6.0.i.i.i540, %Flow3742 ]
  br label %cleanup26.i.i.i569, !amdgpu.uniform !42

if.end22.i.i.i632:                                ; preds = %if.then16.i.i.i633, %if.end14.i.i.i628
  call void @llvm.amdgcn.end.cf.i64(i64 %862)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %Flow3743, !amdgpu.uniform !42

Flow3744:                                         ; preds = %cleanup26.i.i.i569, %840
  %867 = phi i32 [ %866, %cleanup26.i.i.i569 ], [ %829, %840 ]
  %868 = phi i64 [ %865, %cleanup26.i.i.i569 ], [ %830, %840 ]
  %869 = phi i32 [ %864, %cleanup26.i.i.i569 ], [ %831, %840 ]
  %870 = phi i32 [ %864, %cleanup26.i.i.i569 ], [ %retval.sroa.8.0.i.i.i543, %840 ]
  %871 = phi i64 [ %865, %cleanup26.i.i.i569 ], [ %retval.sroa.2.0.i.i.i542, %840 ]
  %872 = phi i32 [ %866, %cleanup26.i.i.i569 ], [ %retval.sroa.6.0.i.i.i540, %840 ]
  %873 = phi i1 [ %cmp.i46.not.i.i.i561, %cleanup26.i.i.i569 ], [ true, %840 ]
  br i1 %873, label %for.inc.i.i.i574, label %Flow3745, !amdgpu.uniform !42

cleanup26.i.i.i569:                               ; preds = %Flow3743
  br label %Flow3744, !amdgpu.uniform !42

for.inc.i.i.i574:                                 ; preds = %Flow3744
  %inc.i.i.i579 = add i32 %spec.store.select.i.i.i546, 1
  br label %Flow3745, !amdgpu.uniform !42

Flow3745:                                         ; preds = %for.inc.i.i.i574, %Flow3744
  %874 = phi i32 [ %inc.i.i.i579, %for.inc.i.i.i574 ], [ poison, %Flow3744 ]
  %875 = phi i1 [ false, %for.inc.i.i.i574 ], [ true, %Flow3744 ]
  br i1 %875, label %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i580, label %for.cond.i.i1.i539, !amdgpu.uniform !42

_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i580: ; preds = %Flow3745
  %.lcssa4132 = phi i32 [ %867, %Flow3745 ]
  %.lcssa4131 = phi i64 [ %868, %Flow3745 ]
  %.lcssa4130 = phi i32 [ %869, %Flow3745 ]
  fence acquire
  %876 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %mul.i.i.i.i.i.i581 = shl i32 %.lcssa4132, 6
  %idxprom.i19.i.i.i.i.i582 = zext i32 %mul.i.i.i.i.i.i581 to i64
  %arrayidx.i20.i.i.i.i.i583 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %876, i64 %idxprom.i19.i.i.i.i.i582
  %arrayidx.i22.i.i.i.i.i584 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %arrayidx.i20.i.i.i.i.i583, i64 %sh_prom.i.i.i.i
  store i64 sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.4 to ptr), i64 33) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.4 to ptr) to i64)), ptr %arrayidx.i22.i.i.i.i.i584, align 8, !tbaa !82
  %spec.select.i.i.i.i.i.i.i585 = tail call i64 @llvm.umin.i64(i64 sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.4 to ptr), i64 33) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.4 to ptr) to i64)), i64 56)
  %877 = and i64 %spec.select.i.i.i.i.i.i.i585, 15
  %878 = sub i64 %spec.select.i.i.i.i.i.i.i585, %877
  %879 = icmp ne i64 %878, 0
  br i1 %879, label %loop-memcpy-expansion2356.preheader, label %Flow3741, !amdgpu.uniform !42

loop-memcpy-expansion2356.preheader:              ; preds = %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i580
  %880 = add nuw nsw i64 %75, 8
  %scevgep3104 = getelementptr i8, ptr %876, i64 %880
  %881 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i582, 6
  %scevgep3105 = getelementptr i8, ptr %scevgep3104, i64 %881
  br label %loop-memcpy-expansion2356, !amdgpu.uniform !42

Flow3741:                                         ; preds = %Flow3740, %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i580
  br label %loop-memcpy-residual-header2359, !amdgpu.uniform !42

loop-memcpy-expansion2356:                        ; preds = %loop-memcpy-expansion2356.preheader, %loop-memcpy-expansion2356
  %loop-index2357 = phi i64 [ %884, %loop-memcpy-expansion2356 ], [ 0, %loop-memcpy-expansion2356.preheader ]
  %882 = getelementptr i8, ptr addrspace(4) @.str.4, i64 %loop-index2357, !amdgpu.uniform !42
  %883 = load <4 x i32>, ptr addrspace(4) %882, align 1
  %scevgep3106 = getelementptr i8, ptr %scevgep3105, i64 %loop-index2357
  store <4 x i32> %883, ptr %scevgep3106, align 1
  %884 = add i64 %loop-index2357, 16
  %885 = icmp uge i64 %884, %878
  br i1 %885, label %Flow3740, label %loop-memcpy-expansion2356, !amdgpu.uniform !42

Flow3739:                                         ; preds = %Flow3737, %Flow3738
  %886 = phi i64 [ %2876, %Flow3737 ], [ %2873, %Flow3738 ]
  br label %post-loop-memcpy-expansion2355, !amdgpu.uniform !42

loop-memcpy-residual2358:                         ; preds = %loop-memcpy-residual2358.preheader, %loop-memcpy-residual2358
  %residual-loop-index2360 = phi i64 [ %888, %loop-memcpy-residual2358 ], [ 0, %loop-memcpy-residual2358.preheader ]
  %scevgep3103 = getelementptr i8, ptr addrspace(4) getelementptr inbounds nuw (i8, ptr addrspace(4) @.str.4, i64 32), i64 %residual-loop-index2360, !amdgpu.uniform !42
  %887 = load i8, ptr addrspace(4) %scevgep3103, align 1
  %scevgep3102 = getelementptr i8, ptr %scevgep3101, i64 %residual-loop-index2360
  store i8 %887, ptr %scevgep3102, align 1
  %888 = add i64 %residual-loop-index2360, 1
  %889 = icmp uge i64 %888, %877
  br i1 %889, label %Flow3737, label %loop-memcpy-residual2358, !amdgpu.uniform !42

Flow3737:                                         ; preds = %loop-memcpy-residual2358
  br label %Flow3739, !amdgpu.uniform !42

post-loop-memcpy-expansion2355:                   ; preds = %Flow3739
  %tobool.not.i.i.i.i.i.i587 = icmp eq i32 %.lcssa4130, 0
  %conv.i.i.i.i.i.i588 = zext i1 %tobool.not.i.i.i.i.i.i587 to i32
  fence release
  %890 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %idxprom.i23.i.i.i.i.i589 = zext i32 %.lcssa4132 to i64
  %arrayidx.i24.i.i.i.i.i590 = getelementptr inbounds nuw i32, ptr %890, i64 %idxprom.i23.i.i.i.i.i589
  store atomic i32 %conv.i.i.i.i.i.i588, ptr %arrayidx.i24.i.i.i.i.i590 monotonic, align 4
  %891 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !84
  %arrayidx.i.i5.i.i591 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %891, i64 %idxprom.i23.i.i.i.i.i589, !amdgpu.uniform !42
  %892 = load i64, ptr %arrayidx.i.i5.i.i591, align 8, !tbaa !80
  br label %while.cond.i.i.i.i592, !amdgpu.uniform !42

while.cond.i.i.i.i592:                            ; preds = %post-loop-memcpy-expansion2355, %Flow3736
  %phi.broken3912 = phi i64 [ 0, %post-loop-memcpy-expansion2355 ], [ %909, %Flow3736 ]
  %lsr.iv3093 = phi i64 [ %907, %Flow3736 ], [ -23, %post-loop-memcpy-expansion2355 ]
  %lsr.iv3088 = phi ptr addrspace(4) [ %906, %Flow3736 ], [ getelementptr (i8, ptr addrspace(4) @.str.4, i64 56), %post-loop-memcpy-expansion2355 ]
  %port.sroa.37.0.i.i593 = phi i32 [ %905, %Flow3736 ], [ %conv.i.i.i.i.i.i588, %post-loop-memcpy-expansion2355 ]
  %idx.0.i.i.i.i594 = phi i64 [ %904, %Flow3736 ], [ 56, %post-loop-memcpy-expansion2355 ]
  %cmp.i.i.i.i595 = icmp ult i64 %idx.0.i.i.i.i594, sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.4 to ptr), i64 33) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.4 to ptr) to i64))
  %893 = tail call i64 @llvm.amdgcn.ballot.i64(i1 %cmp.i.i.i.i595)
  %and.i.i.i.i6.i.i596 = and i64 %893, %892
  %tobool.not.i.i.i.i597 = icmp ne i64 %and.i.i.i.i6.i.i596, 0
  %894 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i22.i.i.i598 = getelementptr inbounds nuw i32, ptr %894, i64 %idxprom.i23.i.i.i.i.i589, !amdgpu.uniform !42
  %895 = load atomic i32, ptr %arrayidx.i22.i.i.i598 monotonic, align 4
  %896 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %tobool.not.i.i.i.i597)
  %897 = extractvalue { i1, i64 } %896, 0
  %898 = extractvalue { i1, i64 } %896, 1
  br i1 %897, label %cond.false.i8.i.i.i.i599, label %Flow3736

cond.false.i8.i.i.i.i599:                         ; preds = %while.cond.i.i.i.i592
  %899 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %895)
  %900 = icmp ne i32 %899, %port.sroa.37.0.i.i593
  br i1 %900, label %while.body.i.i21.i.i.i.i600.preheader, label %Flow3735, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i600.preheader:            ; preds = %cond.false.i8.i.i.i.i599
  br label %while.body.i.i21.i.i.i.i600, !amdgpu.uniform !42

Flow3735:                                         ; preds = %Flow3734, %cond.false.i8.i.i.i.i599
  br label %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i603, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i600:                      ; preds = %while.body.i.i21.i.i.i.i600.preheader, %while.body.i.i21.i.i.i.i600
  tail call void @llvm.amdgcn.s.sleep(i32 2)
  %901 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i22.i.i.i.i601 = getelementptr inbounds nuw i32, ptr %901, i64 %idxprom.i23.i.i.i.i.i589, !amdgpu.uniform !42
  %902 = load atomic i32, ptr %arrayidx.i.i.i22.i.i.i.i601 monotonic, align 4
  %903 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %902)
  %cmp.i.not.i.i23.i.i.i.i602 = icmp eq i32 %903, %port.sroa.37.0.i.i593
  br i1 %cmp.i.not.i.i23.i.i.i.i602, label %Flow3734, label %while.body.i.i21.i.i.i.i600, !amdgpu.uniform !42

Flow3734:                                         ; preds = %while.body.i.i21.i.i.i.i600
  br label %Flow3735, !amdgpu.uniform !42

Flow3736:                                         ; preds = %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i604, %while.cond.i.i.i.i592
  %904 = phi i64 [ %add.i.i.i.i607, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i604 ], [ poison, %while.cond.i.i.i.i592 ]
  %905 = phi i32 [ %conv.i.i26.i.i.i.i605, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i604 ], [ poison, %while.cond.i.i.i.i592 ]
  %906 = phi ptr addrspace(4) [ %scevgep3089, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i604 ], [ poison, %while.cond.i.i.i.i592 ]
  %907 = phi i64 [ %lsr.iv.next3094, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i604 ], [ poison, %while.cond.i.i.i.i592 ]
  %908 = phi i1 [ false, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i604 ], [ true, %while.cond.i.i.i.i592 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %898)
  %909 = call i64 @llvm.amdgcn.if.break.i64(i1 %908, i64 %phi.broken3912)
  %910 = call i1 @llvm.amdgcn.loop.i64(i64 %909)
  br i1 %910, label %cond.false.i.i.i614, label %while.cond.i.i.i.i592

_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i603: ; preds = %Flow3735
  fence acquire
  br i1 %cmp.i.i.i.i595, label %if.then.i.i.i.i.i.i.i608, label %Flow3733, !amdgpu.uniform !42

if.then.i.i.i.i.i.i.i608:                         ; preds = %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i603
  %sub.i.i.i.i.i.i.i610 = sub nuw i64 sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.4 to ptr), i64 33) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.4 to ptr) to i64)), %idx.0.i.i.i.i594
  %spec.select.i.i.i36.i.i.i.i611 = tail call i64 @llvm.umin.i64(i64 %sub.i.i.i.i.i.i.i610, i64 64)
  %911 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %912 = and i64 %spec.select.i.i.i36.i.i.i.i611, 15
  %913 = sub i64 %spec.select.i.i.i36.i.i.i.i611, %912
  %914 = icmp ne i64 %913, 0
  br i1 %914, label %loop-memcpy-expansion2362.preheader, label %Flow3732, !amdgpu.uniform !42

loop-memcpy-expansion2362.preheader:              ; preds = %if.then.i.i.i.i.i.i.i608
  %arrayidx.i20.i35.i.i.i.i612 = getelementptr %"struct.rpc::Buffer", ptr %911, i64 %idxprom.i19.i.i.i.i.i582
  %arrayidx.i22.i38.i.i.i.i613 = getelementptr %"struct.rpc::Buffer", ptr %arrayidx.i20.i35.i.i.i.i612, i64 %sh_prom.i.i.i.i
  br label %loop-memcpy-expansion2362, !amdgpu.uniform !42

Flow3732:                                         ; preds = %Flow3731, %if.then.i.i.i.i.i.i.i608
  br label %loop-memcpy-residual-header2365, !amdgpu.uniform !42

loop-memcpy-expansion2362:                        ; preds = %loop-memcpy-expansion2362.preheader, %loop-memcpy-expansion2362
  %loop-index2363 = phi i64 [ %917, %loop-memcpy-expansion2362 ], [ 0, %loop-memcpy-expansion2362.preheader ]
  %scevgep3090 = getelementptr i8, ptr addrspace(4) %lsr.iv3088, i64 %loop-index2363, !amdgpu.uniform !42
  %915 = load <4 x i32>, ptr addrspace(4) %scevgep3090, align 1
  %916 = getelementptr i8, ptr %arrayidx.i22.i38.i.i.i.i613, i64 %loop-index2363
  store <4 x i32> %915, ptr %916, align 1
  %917 = add i64 %loop-index2363, 16
  %918 = icmp uge i64 %917, %913
  br i1 %918, label %Flow3731, label %loop-memcpy-expansion2362, !amdgpu.uniform !42

Flow3730:                                         ; preds = %Flow3729, %loop-memcpy-residual-header2365
  br label %Flow3733, !amdgpu.uniform !42

loop-memcpy-residual2364:                         ; preds = %loop-memcpy-residual2364.preheader, %loop-memcpy-residual2364
  %residual-loop-index2366 = phi i64 [ %920, %loop-memcpy-residual2364 ], [ 0, %loop-memcpy-residual2364.preheader ]
  %scevgep3099 = getelementptr i8, ptr addrspace(4) %scevgep3098, i64 %residual-loop-index2366, !amdgpu.uniform !42
  %919 = load i8, ptr addrspace(4) %scevgep3099, align 1
  %scevgep3097 = getelementptr i8, ptr %scevgep3096, i64 %residual-loop-index2366
  store i8 %919, ptr %scevgep3097, align 1
  %920 = add i64 %residual-loop-index2366, 1
  %921 = icmp uge i64 %920, %912
  br i1 %921, label %Flow3729, label %loop-memcpy-residual2364, !amdgpu.uniform !42

Flow3729:                                         ; preds = %loop-memcpy-residual2364
  br label %Flow3730, !amdgpu.uniform !42

_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i604: ; preds = %Flow3733
  %conv.i.i26.i.i.i.i605 = xor i32 %port.sroa.37.0.i.i593, 1
  fence release
  %922 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %arrayidx.i24.i29.i.i.i.i606 = getelementptr inbounds nuw i32, ptr %922, i64 %idxprom.i23.i.i.i.i.i589
  store atomic i32 %conv.i.i26.i.i.i.i605, ptr %arrayidx.i24.i29.i.i.i.i606 monotonic, align 4
  %add.i.i.i.i607 = add i64 %idx.0.i.i.i.i594, 64
  %scevgep3089 = getelementptr i8, ptr addrspace(4) %lsr.iv3088, i64 64
  %lsr.iv.next3094 = add i64 %lsr.iv3093, -64
  br label %Flow3736, !amdgpu.uniform !42

cond.false.i.i.i614:                              ; preds = %Flow3736
  %.lcssa4126 = phi i64 [ %909, %Flow3736 ]
  %port.sroa.37.0.i.i593.lcssa = phi i32 [ %port.sroa.37.0.i.i593, %Flow3736 ]
  %.lcssa4125 = phi i32 [ %895, %Flow3736 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa4126)
  %923 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %.lcssa4125)
  %924 = icmp ne i32 %923, %port.sroa.37.0.i.i593.lcssa
  %925 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %924)
  %926 = extractvalue { i1, i64 } %925, 0
  %927 = extractvalue { i1, i64 } %925, 1
  br i1 %926, label %while.body.i.i.i.i615.preheader, label %Flow3728

while.body.i.i.i.i615.preheader:                  ; preds = %cond.false.i.i.i614
  br label %while.body.i.i.i.i615, !amdgpu.uniform !42

Flow3728:                                         ; preds = %Flow3727, %cond.false.i.i.i614
  call void @llvm.amdgcn.end.cf.i64(i64 %927)
  br label %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit636, !amdgpu.uniform !42

while.body.i.i.i.i615:                            ; preds = %while.body.i.i.i.i615.preheader, %while.body.i.i.i.i615
  %phi.broken3913 = phi i64 [ 0, %while.body.i.i.i.i615.preheader ], [ %931, %while.body.i.i.i.i615 ]
  tail call void @llvm.amdgcn.s.sleep(i32 2)
  %928 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i8.i.i616 = getelementptr inbounds nuw i32, ptr %928, i64 %idxprom.i23.i.i.i.i.i589, !amdgpu.uniform !42
  %929 = load atomic i32, ptr %arrayidx.i.i.i8.i.i616 monotonic, align 4
  %930 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %929)
  %cmp.i.not.i.i.i.i617 = icmp eq i32 %930, %port.sroa.37.0.i.i593.lcssa
  %931 = call i64 @llvm.amdgcn.if.break.i64(i1 %cmp.i.not.i.i.i.i617, i64 %phi.broken3913)
  %932 = call i1 @llvm.amdgcn.loop.i64(i64 %931)
  br i1 %932, label %Flow3727, label %while.body.i.i.i.i615

Flow3727:                                         ; preds = %while.body.i.i.i.i615
  %.lcssa4124 = phi i64 [ %931, %while.body.i.i.i.i615 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa4124)
  br label %Flow3728, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit636: ; preds = %Flow3728
  fence acquire
  tail call void @llvm.amdgcn.wave.barrier()
  fence syncscope("agent") release
  %933 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %.lcssa4131, i1 true)
  %iszero.i.i.i.i.i13.i.i619 = icmp ne i64 %.lcssa4131, 0
  %cmp2.i.i.i.i14.i.i620 = icmp eq i64 %933, %sh_prom.i.i.i.i
  %cmp.i.i.i.i15.i.i621 = select i1 %iszero.i.i.i.i.i13.i.i619, i1 %cmp2.i.i.i.i14.i.i620, i1 false
  %934 = and i32 %.lcssa4132, 31
  %conv4.i.i.i19.i.i625 = zext i1 %cmp.i.i.i.i15.i.i621 to i32
  %shl.i.i.i20.i.i626 = shl nuw i32 %conv4.i.i.i19.i.i625, %934
  %xor.i.i.i21.i.i627 = xor i32 %shl.i.i.i20.i.i626, -1
  %935 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2509, !amdgpu.uniform !42

936:                                              ; preds = %ComputeEnd2510
  %div8.i.i.i16.i.i622 = lshr i32 %.lcssa4132, 5
  %idxprom.i.i.i17.i.i623 = zext nneg i32 %div8.i.i.i16.i.i622 to i64
  %937 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i17.i.i623
  %sunkaddr3291 = getelementptr inbounds i8, ptr addrspace(1) %937, i64 40
  %938 = atomicrmw and ptr addrspace(1) %sunkaddr3291, i32 %.lcssa4123 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %939, !amdgpu.uniform !42

939:                                              ; preds = %936, %ComputeEnd2510
  call void @llvm.amdgcn.end.cf.i64(i64 %3299)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %for.cond.i.i.i638, !amdgpu.uniform !42

for.cond.i.i.i638:                                ; preds = %939, %Flow3726
  %940 = phi i32 [ %978, %Flow3726 ], [ poison, %939 ]
  %941 = phi i64 [ %979, %Flow3726 ], [ poison, %939 ]
  %942 = phi i32 [ %980, %Flow3726 ], [ poison, %939 ]
  %retval.sroa.6.0.i.i.i639 = phi i32 [ %983, %Flow3726 ], [ undef, %939 ]
  %retval.sroa.2.0.i.i.i641 = phi i64 [ %982, %Flow3726 ], [ undef, %939 ]
  %retval.sroa.8.0.i.i.i642 = phi i32 [ %981, %Flow3726 ], [ undef, %939 ]
  %index.0.i.i.i643 = phi i32 [ %985, %Flow3726 ], [ 0, %939 ]
  %943 = load i32, ptr addrspace(1) @__llvm_rpc_client, align 8, !tbaa !68
  %cmp.not.i.i.i644 = icmp ult i32 %index.0.i.i.i643, %943
  %spec.store.select.i.i.i645 = select i1 %cmp.not.i.i.i644, i32 %index.0.i.i.i643, i32 0
  %944 = tail call noundef i64 @llvm.amdgcn.ballot.i64(i1 true)
  %945 = and i32 %spec.store.select.i.i.i645, 31
  %946 = lshr i64 %944, %sh_prom.i.i.i.i
  %947 = trunc i64 %946 to i32
  %conv4.i.i.i.i.i649 = and i32 %947, 1
  %shl.i.i.i.i.i650 = shl nuw i32 %conv4.i.i.i.i.i649, %945
  %948 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2513, !amdgpu.uniform !42

949:                                              ; preds = %ComputeEnd2514
  %sunkaddr3292 = getelementptr inbounds i8, ptr addrspace(1) %3309, i64 40
  %950 = atomicrmw or ptr addrspace(1) %sunkaddr3292, i32 %.lcssa4117 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %951, !amdgpu.uniform !42

951:                                              ; preds = %949, %ComputeEnd2514
  %952 = phi i32 [ %950, %949 ], [ poison, %ComputeEnd2514 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %3318)
  %953 = call i32 @llvm.amdgcn.readfirstlane.i32(i32 %952)
  %954 = or i32 %953, %.lcssa4118
  %shl5.i.i.i.i.i651 = shl nuw i32 1, %945
  %and.i.i.i.i.i652 = and i32 %shl5.i.i.i.i.i651, %954
  %tobool3.i.i.i.i653 = icmp ne i32 %and.i.i.i.i.i652, 0
  %955 = tail call i64 @llvm.amdgcn.ballot.i64(i1 %tobool3.i.i.i.i653)
  %and.i.i.i.i.i.i654 = and i64 %955, %944
  %cmp.i.not.i.i.i655 = icmp ne i64 %944, %and.i.i.i.i.i.i654
  br i1 %cmp.i.not.i.i.i655, label %if.then.i.i.i.i656, label %Flow3725, !amdgpu.uniform !42

if.then.i.i.i.i656:                               ; preds = %951
  fence syncscope("agent") acquire
  %956 = load <2 x i64>, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !75
  %957 = extractelement <2 x i64> %956, i32 0
  %958 = inttoptr i64 %957 to ptr
  %959 = extractelement <2 x i64> %956, i32 1
  %960 = inttoptr i64 %959 to ptr
  %idxprom.i.i.i.i657 = zext i32 %spec.store.select.i.i.i645 to i64
  %arrayidx.i.i.i.i658 = getelementptr inbounds nuw i32, ptr %958, i64 %idxprom.i.i.i.i657, !amdgpu.uniform !42
  %961 = load atomic i32, ptr %arrayidx.i.i.i.i658 monotonic, align 4
  %962 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %961)
  %arrayidx.i45.i.i.i659 = getelementptr inbounds nuw i32, ptr %960, i64 %idxprom.i.i.i.i657, !amdgpu.uniform !42
  %963 = load atomic i32, ptr %arrayidx.i45.i.i.i659 monotonic, align 4
  %964 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %963)
  %cmp.i46.not.i.i.i660 = icmp ne i32 %962, %964
  br i1 %cmp.i46.not.i.i.i660, label %if.then12.i.i.i661, label %Flow3723, !amdgpu.uniform !42

if.then12.i.i.i661:                               ; preds = %if.then.i.i.i.i656
  fence syncscope("agent") release
  %965 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %944, i1 true)
  %iszero.i.i.i.i.i.i.i662 = icmp ne i64 %944, 0
  %cmp2.i.i.i.i.i.i663 = icmp eq i64 %965, %sh_prom.i.i.i.i
  %cmp.i.i.i.i.i.i664 = select i1 %iszero.i.i.i.i.i.i.i662, i1 %cmp2.i.i.i.i.i.i663, i1 false
  %conv4.i.i51.i.i.i665 = zext i1 %cmp.i.i.i.i.i.i664 to i32
  %shl.i.i52.i.i.i666 = shl nuw i32 %conv4.i.i51.i.i.i665, %945
  %xor.i.i.i.i.i667 = xor i32 %shl.i.i52.i.i.i666, -1
  %966 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2518, !amdgpu.uniform !42

967:                                              ; preds = %ComputeEnd2519
  %sunkaddr3293 = getelementptr inbounds i8, ptr addrspace(1) %3309, i64 40
  %968 = atomicrmw and ptr addrspace(1) %sunkaddr3293, i32 %.lcssa4119 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %969, !amdgpu.uniform !42

969:                                              ; preds = %967, %ComputeEnd2519
  call void @llvm.amdgcn.end.cf.i64(i64 %3336)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %Flow3723, !amdgpu.uniform !42

if.end14.i.i.i725:                                ; preds = %Flow3723
  %970 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %944, i1 true)
  %iszero.i.i.i.i.i.i726 = icmp ne i64 %944, 0
  %cmp2.i.i.i.i.i727 = icmp eq i64 %970, %sh_prom.i.i.i.i
  %cmp.i.i.i.i.i728 = select i1 %iszero.i.i.i.i.i.i726, i1 %cmp2.i.i.i.i.i727, i1 false
  %971 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.i.i.i.i.i728)
  %972 = extractvalue { i1, i64 } %971, 0
  %973 = extractvalue { i1, i64 } %971, 1
  br i1 %972, label %if.then16.i.i.i730, label %if.end22.i.i.i729

if.then16.i.i.i730:                               ; preds = %if.end14.i.i.i725
  %974 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !76
  %opcode.i.i.i731 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %974, i64 %idxprom.i.i.i.i657, i32 1
  store i32 1660944387, ptr %opcode.i.i.i731, align 8, !tbaa !77
  %arrayidx21.i.i.i732 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %974, i64 %idxprom.i.i.i.i657
  store i64 %944, ptr %arrayidx21.i.i.i732, align 8, !tbaa !80
  br label %if.end22.i.i.i729, !amdgpu.uniform !42

Flow3724:                                         ; preds = %if.end22.i.i.i729, %Flow3723
  %975 = phi i32 [ %962, %if.end22.i.i.i729 ], [ %retval.sroa.8.0.i.i.i642, %Flow3723 ]
  %976 = phi i64 [ %944, %if.end22.i.i.i729 ], [ %retval.sroa.2.0.i.i.i641, %Flow3723 ]
  %977 = phi i32 [ %spec.store.select.i.i.i645, %if.end22.i.i.i729 ], [ %retval.sroa.6.0.i.i.i639, %Flow3723 ]
  br label %cleanup26.i.i.i668, !amdgpu.uniform !42

if.end22.i.i.i729:                                ; preds = %if.then16.i.i.i730, %if.end14.i.i.i725
  call void @llvm.amdgcn.end.cf.i64(i64 %973)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %Flow3724, !amdgpu.uniform !42

Flow3725:                                         ; preds = %cleanup26.i.i.i668, %951
  %978 = phi i32 [ %977, %cleanup26.i.i.i668 ], [ %940, %951 ]
  %979 = phi i64 [ %976, %cleanup26.i.i.i668 ], [ %941, %951 ]
  %980 = phi i32 [ %975, %cleanup26.i.i.i668 ], [ %942, %951 ]
  %981 = phi i32 [ %975, %cleanup26.i.i.i668 ], [ %retval.sroa.8.0.i.i.i642, %951 ]
  %982 = phi i64 [ %976, %cleanup26.i.i.i668 ], [ %retval.sroa.2.0.i.i.i641, %951 ]
  %983 = phi i32 [ %977, %cleanup26.i.i.i668 ], [ %retval.sroa.6.0.i.i.i639, %951 ]
  %984 = phi i1 [ %cmp.i46.not.i.i.i660, %cleanup26.i.i.i668 ], [ true, %951 ]
  br i1 %984, label %for.inc.i.i.i673, label %Flow3726, !amdgpu.uniform !42

cleanup26.i.i.i668:                               ; preds = %Flow3724
  br label %Flow3725, !amdgpu.uniform !42

for.inc.i.i.i673:                                 ; preds = %Flow3725
  %inc.i.i.i678 = add i32 %spec.store.select.i.i.i645, 1
  br label %Flow3726, !amdgpu.uniform !42

Flow3726:                                         ; preds = %for.inc.i.i.i673, %Flow3725
  %985 = phi i32 [ %inc.i.i.i678, %for.inc.i.i.i673 ], [ poison, %Flow3725 ]
  %986 = phi i1 [ false, %for.inc.i.i.i673 ], [ true, %Flow3725 ]
  br i1 %986, label %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i679, label %for.cond.i.i.i638, !amdgpu.uniform !42

_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i679: ; preds = %Flow3726
  %.lcssa4122 = phi i32 [ %978, %Flow3726 ]
  %.lcssa4121 = phi i64 [ %979, %Flow3726 ]
  %.lcssa4120 = phi i32 [ %980, %Flow3726 ]
  fence acquire
  %987 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %mul.i.i.i.i.i.i680 = shl i32 %.lcssa4122, 6
  %idxprom.i19.i.i.i.i.i681 = zext i32 %mul.i.i.i.i.i.i680 to i64
  %arrayidx.i20.i.i.i.i.i682 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %987, i64 %idxprom.i19.i.i.i.i.i681
  %arrayidx.i22.i.i.i.i.i683 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %arrayidx.i20.i.i.i.i.i682, i64 %sh_prom.i.i.i.i
  store i64 1, ptr %arrayidx.i22.i.i.i.i.i683, align 8, !tbaa !82
  %arrayidx6.i.i.i.i.i.i.i684 = getelementptr inbounds nuw i8, ptr %arrayidx.i22.i.i.i.i.i683, i64 8
  store i8 10, ptr %arrayidx6.i.i.i.i.i.i.i684, align 8
  %tobool.not.i.i.i.i.i.i685 = icmp eq i32 %.lcssa4120, 0
  %conv.i.i.i.i.i.i686 = zext i1 %tobool.not.i.i.i.i.i.i685 to i32
  fence release
  %988 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %idxprom.i23.i.i.i.i.i687 = zext i32 %.lcssa4122 to i64
  %arrayidx.i24.i.i.i.i.i688 = getelementptr inbounds nuw i32, ptr %988, i64 %idxprom.i23.i.i.i.i.i687
  store atomic i32 %conv.i.i.i.i.i.i686, ptr %arrayidx.i24.i.i.i.i.i688 monotonic, align 4
  %989 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !84
  %arrayidx.i.i5.i.i689 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %989, i64 %idxprom.i23.i.i.i.i.i687, !amdgpu.uniform !42
  %990 = load i64, ptr %arrayidx.i.i5.i.i689, align 8, !tbaa !80
  br label %while.cond.i.i.i.i690, !amdgpu.uniform !42

while.cond.i.i.i.i690:                            ; preds = %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i679, %Flow3722
  %phi.broken3914 = phi i64 [ 0, %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i679 ], [ %1005, %Flow3722 ]
  %port.sroa.37.0.i.i691 = phi i32 [ %1003, %Flow3722 ], [ %conv.i.i.i.i.i.i686, %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i679 ]
  %idx.0.i.i.i.i692 = phi i64 [ %1002, %Flow3722 ], [ 56, %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i679 ]
  %cmp.i.i.i.i693 = icmp eq i64 %idx.0.i.i.i.i692, 0
  %991 = tail call i64 @llvm.amdgcn.ballot.i64(i1 %cmp.i.i.i.i693)
  %and.i.i.i.i6.i.i694 = and i64 %991, %990
  %tobool.not.i.i.i.i695 = icmp ne i64 %and.i.i.i.i6.i.i694, 0
  %992 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i22.i.i.i696 = getelementptr inbounds nuw i32, ptr %992, i64 %idxprom.i23.i.i.i.i.i687, !amdgpu.uniform !42
  %993 = load atomic i32, ptr %arrayidx.i22.i.i.i696 monotonic, align 4
  %994 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %tobool.not.i.i.i.i695)
  %995 = extractvalue { i1, i64 } %994, 0
  %996 = extractvalue { i1, i64 } %994, 1
  br i1 %995, label %cond.false.i8.i.i.i.i697, label %Flow3722

cond.false.i8.i.i.i.i697:                         ; preds = %while.cond.i.i.i.i690
  %997 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %993)
  %998 = icmp ne i32 %997, %port.sroa.37.0.i.i691
  br i1 %998, label %while.body.i.i21.i.i.i.i698.preheader, label %Flow3721, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i698.preheader:            ; preds = %cond.false.i8.i.i.i.i697
  br label %while.body.i.i21.i.i.i.i698, !amdgpu.uniform !42

Flow3721:                                         ; preds = %Flow3720, %cond.false.i8.i.i.i.i697
  br label %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i701, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i698:                      ; preds = %while.body.i.i21.i.i.i.i698.preheader, %while.body.i.i21.i.i.i.i698
  tail call void @llvm.amdgcn.s.sleep(i32 2)
  %999 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i22.i.i.i.i699 = getelementptr inbounds nuw i32, ptr %999, i64 %idxprom.i23.i.i.i.i.i687, !amdgpu.uniform !42
  %1000 = load atomic i32, ptr %arrayidx.i.i.i22.i.i.i.i699 monotonic, align 4
  %1001 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %1000)
  %cmp.i.not.i.i23.i.i.i.i700 = icmp eq i32 %1001, %port.sroa.37.0.i.i691
  br i1 %cmp.i.not.i.i23.i.i.i.i700, label %Flow3720, label %while.body.i.i21.i.i.i.i698, !amdgpu.uniform !42

Flow3720:                                         ; preds = %while.body.i.i21.i.i.i.i698
  br label %Flow3721, !amdgpu.uniform !42

Flow3722:                                         ; preds = %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i702, %while.cond.i.i.i.i690
  %1002 = phi i64 [ %add.i.i.i.i705, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i702 ], [ poison, %while.cond.i.i.i.i690 ]
  %1003 = phi i32 [ %conv.i.i26.i.i.i.i703, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i702 ], [ poison, %while.cond.i.i.i.i690 ]
  %1004 = phi i1 [ false, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i702 ], [ true, %while.cond.i.i.i.i690 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %996)
  %1005 = call i64 @llvm.amdgcn.if.break.i64(i1 %1004, i64 %phi.broken3914)
  %1006 = call i1 @llvm.amdgcn.loop.i64(i64 %1005)
  br i1 %1006, label %cond.false.i.i.i711, label %while.cond.i.i.i.i690

_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i701: ; preds = %Flow3721
  fence acquire
  br i1 %cmp.i.i.i.i693, label %if.then.i.i.i.i.i.i.i706, label %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i702, !amdgpu.uniform !42

if.then.i.i.i.i.i.i.i706:                         ; preds = %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i701
  %1007 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %arrayidx.i20.i35.i.i.i.i709 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %1007, i64 %idxprom.i19.i.i.i.i.i681
  %arrayidx.i22.i38.i.i.i.i710 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %arrayidx.i20.i35.i.i.i.i709, i64 %sh_prom.i.i.i.i
  store i8 10, ptr %arrayidx.i22.i38.i.i.i.i710, align 1
  br label %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i702, !amdgpu.uniform !42

_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i702: ; preds = %if.then.i.i.i.i.i.i.i706, %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i701
  %conv.i.i26.i.i.i.i703 = xor i32 %port.sroa.37.0.i.i691, 1
  fence release
  %1008 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %arrayidx.i24.i29.i.i.i.i704 = getelementptr inbounds nuw i32, ptr %1008, i64 %idxprom.i23.i.i.i.i.i687
  store atomic i32 %conv.i.i26.i.i.i.i703, ptr %arrayidx.i24.i29.i.i.i.i704 monotonic, align 4
  %add.i.i.i.i705 = add i64 %idx.0.i.i.i.i692, 64
  br label %Flow3722, !amdgpu.uniform !42

cond.false.i.i.i711:                              ; preds = %Flow3722
  %.lcssa4116 = phi i64 [ %1005, %Flow3722 ]
  %port.sroa.37.0.i.i691.lcssa = phi i32 [ %port.sroa.37.0.i.i691, %Flow3722 ]
  %.lcssa4115 = phi i32 [ %993, %Flow3722 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa4116)
  %1009 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %.lcssa4115)
  %1010 = icmp ne i32 %1009, %port.sroa.37.0.i.i691.lcssa
  %1011 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %1010)
  %1012 = extractvalue { i1, i64 } %1011, 0
  %1013 = extractvalue { i1, i64 } %1011, 1
  br i1 %1012, label %while.body.i.i.i.i712.preheader, label %Flow3719

while.body.i.i.i.i712.preheader:                  ; preds = %cond.false.i.i.i711
  br label %while.body.i.i.i.i712, !amdgpu.uniform !42

Flow3719:                                         ; preds = %Flow3718, %cond.false.i.i.i711
  call void @llvm.amdgcn.end.cf.i64(i64 %1013)
  br label %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIcEERS1_T_.exit, !amdgpu.uniform !42

while.body.i.i.i.i712:                            ; preds = %while.body.i.i.i.i712.preheader, %while.body.i.i.i.i712
  %phi.broken3915 = phi i64 [ 0, %while.body.i.i.i.i712.preheader ], [ %1017, %while.body.i.i.i.i712 ]
  tail call void @llvm.amdgcn.s.sleep(i32 2)
  %1014 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i8.i.i713 = getelementptr inbounds nuw i32, ptr %1014, i64 %idxprom.i23.i.i.i.i.i687, !amdgpu.uniform !42
  %1015 = load atomic i32, ptr %arrayidx.i.i.i8.i.i713 monotonic, align 4
  %1016 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %1015)
  %cmp.i.not.i.i.i.i714 = icmp eq i32 %1016, %port.sroa.37.0.i.i691.lcssa
  %1017 = call i64 @llvm.amdgcn.if.break.i64(i1 %cmp.i.not.i.i.i.i714, i64 %phi.broken3915)
  %1018 = call i1 @llvm.amdgcn.loop.i64(i64 %1017)
  br i1 %1018, label %Flow3718, label %while.body.i.i.i.i712

Flow3718:                                         ; preds = %while.body.i.i.i.i712
  %.lcssa4114 = phi i64 [ %1017, %while.body.i.i.i.i712 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa4114)
  br label %Flow3719, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIcEERS1_T_.exit: ; preds = %Flow3719
  fence acquire
  tail call void @llvm.amdgcn.wave.barrier()
  fence syncscope("agent") release
  %1019 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %.lcssa4121, i1 true)
  %iszero.i.i.i.i.i13.i.i716 = icmp ne i64 %.lcssa4121, 0
  %cmp2.i.i.i.i14.i.i717 = icmp eq i64 %1019, %sh_prom.i.i.i.i
  %cmp.i.i.i.i15.i.i718 = select i1 %iszero.i.i.i.i.i13.i.i716, i1 %cmp2.i.i.i.i14.i.i717, i1 false
  %1020 = and i32 %.lcssa4122, 31
  %conv4.i.i.i19.i.i722 = zext i1 %cmp.i.i.i.i15.i.i718 to i32
  %shl.i.i.i20.i.i723 = shl nuw i32 %conv4.i.i.i19.i.i722, %1020
  %xor.i.i.i21.i.i724 = xor i32 %shl.i.i.i20.i.i723, -1
  %1021 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2522, !amdgpu.uniform !42

1022:                                             ; preds = %ComputeEnd2523
  %div8.i.i.i16.i.i719 = lshr i32 %.lcssa4122, 5
  %idxprom.i.i.i17.i.i720 = zext nneg i32 %div8.i.i.i16.i.i719 to i64
  %1023 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i17.i.i720
  %sunkaddr3294 = getelementptr inbounds i8, ptr addrspace(1) %1023, i64 40
  %1024 = atomicrmw and ptr addrspace(1) %sunkaddr3294, i32 %.lcssa4113 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %1025, !amdgpu.uniform !42

1025:                                             ; preds = %1022, %ComputeEnd2523
  call void @llvm.amdgcn.end.cf.i64(i64 %3353)
  tail call void @llvm.amdgcn.wave.barrier()
  br i1 %cmp2.i.i.i35.i, label %if.then3.i.i.i65.i, label %if.end4.i.i.i54.i, !amdgpu.uniform !42

if.then3.i.i.i65.i:                               ; preds = %1025
  %mul.i.i.i66.i = mul nuw i64 %inc.i.i.i32.i, 11
  %div18.i.i.i67.i = lshr i64 %mul.i.i.i66.i, 3
  br label %if.end4.i.i.i54.i, !amdgpu.uniform !42

if.end4.i.i.i54.i:                                ; preds = %if.then3.i.i.i65.i, %1025
  %new_capacity.addr.0.i.i.i55.i = phi i64 [ %div18.i.i.i67.i, %if.then3.i.i.i65.i ], [ %inc.i.i.i32.i, %1025 ]
  %sub.i.i733 = add i64 %new_capacity.addr.0.i.i.i55.i, 7
  %div2.i.i734 = and i64 %sub.i.i733, -8
  %1026 = load ptr, ptr addrspace(1) @_ZN12_GLOBAL__N_13ptrE, align 8, !tbaa !55
  %add.ptr.i.i735 = getelementptr inbounds nuw i8, ptr %1026, i64 %div2.i.i734
  store ptr %add.ptr.i.i735, ptr addrspace(1) @_ZN12_GLOBAL__N_13ptrE, align 8, !tbaa !55
  br i1 %597, label %loop-memcpy-expansion2278.preheader, label %Flow3717, !amdgpu.uniform !42

loop-memcpy-expansion2278.preheader:              ; preds = %if.end4.i.i.i54.i
  br label %loop-memcpy-expansion2278, !amdgpu.uniform !42

Flow3717:                                         ; preds = %Flow3716, %if.end4.i.i.i54.i
  br label %loop-memcpy-residual-header2281, !amdgpu.uniform !42

loop-memcpy-expansion2278:                        ; preds = %loop-memcpy-expansion2278.preheader, %loop-memcpy-expansion2278
  %loop-index2279 = phi i64 [ %1030, %loop-memcpy-expansion2278 ], [ 0, %loop-memcpy-expansion2278.preheader ]
  %1027 = getelementptr i8, ptr %3, i64 %loop-index2279, !amdgpu.uniform !42
  %1028 = load <4 x i32>, ptr %1027, align 1
  %1029 = getelementptr i8, ptr %1026, i64 %loop-index2279
  store <4 x i32> %1028, ptr %1029, align 1
  %1030 = add i64 %loop-index2279, 16
  %1031 = icmp uge i64 %1030, %596
  br i1 %1031, label %Flow3716, label %loop-memcpy-expansion2278, !amdgpu.uniform !42

Flow3715:                                         ; preds = %Flow3714, %loop-memcpy-residual-header2281
  br label %post-loop-memcpy-expansion2277, !amdgpu.uniform !42

loop-memcpy-residual2280:                         ; preds = %loop-memcpy-residual2280.preheader, %loop-memcpy-residual2280
  %residual-loop-index2282 = phi i64 [ %1033, %loop-memcpy-residual2280 ], [ 0, %loop-memcpy-residual2280.preheader ]
  %scevgep3087 = getelementptr i8, ptr %scevgep3086, i64 %residual-loop-index2282, !amdgpu.uniform !42
  %1032 = load i8, ptr %scevgep3087, align 1
  %scevgep3085 = getelementptr i8, ptr %scevgep3084, i64 %residual-loop-index2282
  store i8 %1032, ptr %scevgep3085, align 1
  %1033 = add i64 %residual-loop-index2282, 1
  %1034 = icmp uge i64 %1033, %595
  br i1 %1034, label %Flow3714, label %loop-memcpy-residual2280, !amdgpu.uniform !42

Flow3714:                                         ; preds = %loop-memcpy-residual2280
  br label %Flow3715, !amdgpu.uniform !42

post-loop-memcpy-expansion2277:                   ; preds = %Flow3715
  %sub.ptr.lhs.cast.i.i736 = ptrtoint ptr %add.ptr.i.i735 to i64
  %sub.ptr.sub.i.i737 = sub i64 %sub.ptr.lhs.cast.i.i736, ptrtoint (ptr addrspacecast (ptr addrspace(1) @_ZN12_GLOBAL__N_16memoryE to ptr) to i64)
  %cmp.i.i738 = icmp ugt i64 %sub.ptr.sub.i.i737, 65335
  %cond.i.i739 = select i1 %cmp.i.i738, ptr null, ptr %1026
  %cmp.not.i10.i.i62.i = icmp ne ptr %cond.i.i739, addrspacecast (ptr addrspace(4) @_ZN22__llvm_libc_22_0_0_git3cpp6string14NULL_CHARACTERE to ptr)
  br i1 %cmp.not.i10.i.i62.i, label %if.then.i.i.i63.i, label %for.cond.i.i.i741.preheader, !amdgpu.uniform !42

if.then.i.i.i63.i:                                ; preds = %post-loop-memcpy-expansion2277
  %arrayidx.i.i.i64.i = getelementptr inbounds nuw i8, ptr %1026, i64 %spec.select.i
  store i8 0, ptr %arrayidx.i.i.i64.i, align 1, !tbaa !57
  br label %for.cond.i.i.i741.preheader, !amdgpu.uniform !42

for.cond.i.i.i741.preheader:                      ; preds = %if.then.i.i.i63.i, %post-loop-memcpy-expansion2277
  br label %for.cond.i.i.i741, !amdgpu.uniform !42

for.cond.i.i.i741:                                ; preds = %for.cond.i.i.i741.preheader, %Flow3713
  %1035 = phi i32 [ %1073, %Flow3713 ], [ poison, %for.cond.i.i.i741.preheader ]
  %1036 = phi i64 [ %1074, %Flow3713 ], [ poison, %for.cond.i.i.i741.preheader ]
  %1037 = phi i32 [ %1075, %Flow3713 ], [ poison, %for.cond.i.i.i741.preheader ]
  %retval.sroa.6.0.i.i.i742 = phi i32 [ %1078, %Flow3713 ], [ undef, %for.cond.i.i.i741.preheader ]
  %retval.sroa.2.0.i.i.i744 = phi i64 [ %1077, %Flow3713 ], [ undef, %for.cond.i.i.i741.preheader ]
  %retval.sroa.8.0.i.i.i745 = phi i32 [ %1076, %Flow3713 ], [ undef, %for.cond.i.i.i741.preheader ]
  %index.0.i.i.i746 = phi i32 [ %1080, %Flow3713 ], [ 0, %for.cond.i.i.i741.preheader ]
  %1038 = load i32, ptr addrspace(1) @__llvm_rpc_client, align 8, !tbaa !68
  %cmp.not.i.i.i747 = icmp ult i32 %index.0.i.i.i746, %1038
  %spec.store.select.i.i.i748 = select i1 %cmp.not.i.i.i747, i32 %index.0.i.i.i746, i32 0
  %1039 = tail call noundef i64 @llvm.amdgcn.ballot.i64(i1 true)
  %1040 = and i32 %spec.store.select.i.i.i748, 31
  %1041 = lshr i64 %1039, %sh_prom.i.i.i.i
  %1042 = trunc i64 %1041 to i32
  %conv4.i.i.i.i.i752 = and i32 %1042, 1
  %shl.i.i.i.i.i753 = shl nuw i32 %conv4.i.i.i.i.i752, %1040
  %1043 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2526, !amdgpu.uniform !42

1044:                                             ; preds = %ComputeEnd2527
  %sunkaddr3295 = getelementptr inbounds i8, ptr addrspace(1) %3363, i64 40
  %1045 = atomicrmw or ptr addrspace(1) %sunkaddr3295, i32 %.lcssa4107 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %1046, !amdgpu.uniform !42

1046:                                             ; preds = %1044, %ComputeEnd2527
  %1047 = phi i32 [ %1045, %1044 ], [ poison, %ComputeEnd2527 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %3372)
  %1048 = call i32 @llvm.amdgcn.readfirstlane.i32(i32 %1047)
  %1049 = or i32 %1048, %.lcssa4108
  %shl5.i.i.i.i.i754 = shl nuw i32 1, %1040
  %and.i.i.i.i.i755 = and i32 %shl5.i.i.i.i.i754, %1049
  %tobool3.i.i.i.i756 = icmp ne i32 %and.i.i.i.i.i755, 0
  %1050 = tail call i64 @llvm.amdgcn.ballot.i64(i1 %tobool3.i.i.i.i756)
  %and.i.i.i.i.i.i757 = and i64 %1050, %1039
  %cmp.i.not.i.i.i758 = icmp ne i64 %1039, %and.i.i.i.i.i.i757
  br i1 %cmp.i.not.i.i.i758, label %if.then.i.i.i.i759, label %Flow3712, !amdgpu.uniform !42

if.then.i.i.i.i759:                               ; preds = %1046
  fence syncscope("agent") acquire
  %1051 = load <2 x i64>, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !75
  %1052 = extractelement <2 x i64> %1051, i32 0
  %1053 = inttoptr i64 %1052 to ptr
  %1054 = extractelement <2 x i64> %1051, i32 1
  %1055 = inttoptr i64 %1054 to ptr
  %idxprom.i.i.i.i760 = zext i32 %spec.store.select.i.i.i748 to i64
  %arrayidx.i.i.i.i761 = getelementptr inbounds nuw i32, ptr %1053, i64 %idxprom.i.i.i.i760, !amdgpu.uniform !42
  %1056 = load atomic i32, ptr %arrayidx.i.i.i.i761 monotonic, align 4
  %1057 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %1056)
  %arrayidx.i45.i.i.i762 = getelementptr inbounds nuw i32, ptr %1055, i64 %idxprom.i.i.i.i760, !amdgpu.uniform !42
  %1058 = load atomic i32, ptr %arrayidx.i45.i.i.i762 monotonic, align 4
  %1059 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %1058)
  %cmp.i46.not.i.i.i763 = icmp ne i32 %1057, %1059
  br i1 %cmp.i46.not.i.i.i763, label %if.then12.i.i.i764, label %Flow3710, !amdgpu.uniform !42

if.then12.i.i.i764:                               ; preds = %if.then.i.i.i.i759
  fence syncscope("agent") release
  %1060 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %1039, i1 true)
  %iszero.i.i.i.i.i.i.i765 = icmp ne i64 %1039, 0
  %cmp2.i.i.i.i.i.i766 = icmp eq i64 %1060, %sh_prom.i.i.i.i
  %cmp.i.i.i.i.i.i767 = select i1 %iszero.i.i.i.i.i.i.i765, i1 %cmp2.i.i.i.i.i.i766, i1 false
  %conv4.i.i51.i.i.i768 = zext i1 %cmp.i.i.i.i.i.i767 to i32
  %shl.i.i52.i.i.i769 = shl nuw i32 %conv4.i.i51.i.i.i768, %1040
  %xor.i.i.i.i.i770 = xor i32 %shl.i.i52.i.i.i769, -1
  %1061 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2531, !amdgpu.uniform !42

1062:                                             ; preds = %ComputeEnd2532
  %sunkaddr3296 = getelementptr inbounds i8, ptr addrspace(1) %3363, i64 40
  %1063 = atomicrmw and ptr addrspace(1) %sunkaddr3296, i32 %.lcssa4109 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %1064, !amdgpu.uniform !42

1064:                                             ; preds = %1062, %ComputeEnd2532
  call void @llvm.amdgcn.end.cf.i64(i64 %3390)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %Flow3710, !amdgpu.uniform !42

if.end14.i.i.i830:                                ; preds = %Flow3710
  %1065 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %1039, i1 true)
  %iszero.i.i.i.i.i.i831 = icmp ne i64 %1039, 0
  %cmp2.i.i.i.i.i832 = icmp eq i64 %1065, %sh_prom.i.i.i.i
  %cmp.i.i.i.i.i833 = select i1 %iszero.i.i.i.i.i.i831, i1 %cmp2.i.i.i.i.i832, i1 false
  %1066 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.i.i.i.i.i833)
  %1067 = extractvalue { i1, i64 } %1066, 0
  %1068 = extractvalue { i1, i64 } %1066, 1
  br i1 %1067, label %if.then16.i.i.i835, label %if.end22.i.i.i834

if.then16.i.i.i835:                               ; preds = %if.end14.i.i.i830
  %1069 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !76
  %opcode.i.i.i836 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %1069, i64 %idxprom.i.i.i.i760, i32 1
  store i32 1660944387, ptr %opcode.i.i.i836, align 8, !tbaa !77
  %arrayidx21.i.i.i837 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %1069, i64 %idxprom.i.i.i.i760
  store i64 %1039, ptr %arrayidx21.i.i.i837, align 8, !tbaa !80
  br label %if.end22.i.i.i834, !amdgpu.uniform !42

Flow3711:                                         ; preds = %if.end22.i.i.i834, %Flow3710
  %1070 = phi i32 [ %1057, %if.end22.i.i.i834 ], [ %retval.sroa.8.0.i.i.i745, %Flow3710 ]
  %1071 = phi i64 [ %1039, %if.end22.i.i.i834 ], [ %retval.sroa.2.0.i.i.i744, %Flow3710 ]
  %1072 = phi i32 [ %spec.store.select.i.i.i748, %if.end22.i.i.i834 ], [ %retval.sroa.6.0.i.i.i742, %Flow3710 ]
  br label %cleanup26.i.i.i771, !amdgpu.uniform !42

if.end22.i.i.i834:                                ; preds = %if.then16.i.i.i835, %if.end14.i.i.i830
  call void @llvm.amdgcn.end.cf.i64(i64 %1068)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %Flow3711, !amdgpu.uniform !42

Flow3712:                                         ; preds = %cleanup26.i.i.i771, %1046
  %1073 = phi i32 [ %1072, %cleanup26.i.i.i771 ], [ %1035, %1046 ]
  %1074 = phi i64 [ %1071, %cleanup26.i.i.i771 ], [ %1036, %1046 ]
  %1075 = phi i32 [ %1070, %cleanup26.i.i.i771 ], [ %1037, %1046 ]
  %1076 = phi i32 [ %1070, %cleanup26.i.i.i771 ], [ %retval.sroa.8.0.i.i.i745, %1046 ]
  %1077 = phi i64 [ %1071, %cleanup26.i.i.i771 ], [ %retval.sroa.2.0.i.i.i744, %1046 ]
  %1078 = phi i32 [ %1072, %cleanup26.i.i.i771 ], [ %retval.sroa.6.0.i.i.i742, %1046 ]
  %1079 = phi i1 [ %cmp.i46.not.i.i.i763, %cleanup26.i.i.i771 ], [ true, %1046 ]
  br i1 %1079, label %for.inc.i.i.i776, label %Flow3713, !amdgpu.uniform !42

cleanup26.i.i.i771:                               ; preds = %Flow3711
  br label %Flow3712, !amdgpu.uniform !42

for.inc.i.i.i776:                                 ; preds = %Flow3712
  %inc.i.i.i781 = add i32 %spec.store.select.i.i.i748, 1
  br label %Flow3713, !amdgpu.uniform !42

Flow3713:                                         ; preds = %for.inc.i.i.i776, %Flow3712
  %1080 = phi i32 [ %inc.i.i.i781, %for.inc.i.i.i776 ], [ poison, %Flow3712 ]
  %1081 = phi i1 [ false, %for.inc.i.i.i776 ], [ true, %Flow3712 ]
  br i1 %1081, label %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i782, label %for.cond.i.i.i741, !amdgpu.uniform !42

_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i782: ; preds = %Flow3713
  %.lcssa4112 = phi i32 [ %1073, %Flow3713 ]
  %.lcssa4111 = phi i64 [ %1074, %Flow3713 ]
  %.lcssa4110 = phi i32 [ %1075, %Flow3713 ]
  fence acquire
  %1082 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %mul.i.i.i.i.i.i783 = shl i32 %.lcssa4112, 6
  %idxprom.i19.i.i.i.i.i784 = zext i32 %mul.i.i.i.i.i.i783 to i64
  %arrayidx.i20.i.i.i.i.i785 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %1082, i64 %idxprom.i19.i.i.i.i.i784
  %arrayidx.i22.i.i.i.i.i786 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %arrayidx.i20.i.i.i.i.i785, i64 %sh_prom.i.i.i.i
  store i64 %spec.select.i, ptr %arrayidx.i22.i.i.i.i.i786, align 8, !tbaa !82
  br i1 %656, label %loop-memcpy-expansion2242.preheader, label %Flow3709, !amdgpu.uniform !42

loop-memcpy-expansion2242.preheader:              ; preds = %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i782
  %1083 = add nuw nsw i64 %75, 8
  %scevgep3081 = getelementptr i8, ptr %1082, i64 %1083
  %1084 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i784, 6
  %scevgep3082 = getelementptr i8, ptr %scevgep3081, i64 %1084
  br label %loop-memcpy-expansion2242, !amdgpu.uniform !42

Flow3709:                                         ; preds = %Flow3708, %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i782
  br label %loop-memcpy-residual-header2245, !amdgpu.uniform !42

loop-memcpy-expansion2242:                        ; preds = %loop-memcpy-expansion2242.preheader, %loop-memcpy-expansion2242
  %loop-index2243 = phi i64 [ %1087, %loop-memcpy-expansion2242 ], [ 0, %loop-memcpy-expansion2242.preheader ]
  %1085 = getelementptr i8, ptr %1026, i64 %loop-index2243, !amdgpu.uniform !42
  %1086 = load <4 x i32>, ptr %1085, align 1
  %scevgep3083 = getelementptr i8, ptr %scevgep3082, i64 %loop-index2243
  store <4 x i32> %1086, ptr %scevgep3083, align 1
  %1087 = add i64 %loop-index2243, 16
  %1088 = icmp uge i64 %1087, %655
  br i1 %1088, label %Flow3708, label %loop-memcpy-expansion2242, !amdgpu.uniform !42

Flow3707:                                         ; preds = %Flow3705, %Flow3706
  %1089 = phi i64 [ %2776, %Flow3705 ], [ %2772, %Flow3706 ]
  br label %post-loop-memcpy-expansion2241, !amdgpu.uniform !42

loop-memcpy-residual2244:                         ; preds = %loop-memcpy-residual2244.preheader, %loop-memcpy-residual2244
  %residual-loop-index2246 = phi i64 [ %1091, %loop-memcpy-residual2244 ], [ 0, %loop-memcpy-residual2244.preheader ]
  %scevgep3080 = getelementptr i8, ptr %scevgep3079, i64 %residual-loop-index2246, !amdgpu.uniform !42
  %1090 = load i8, ptr %scevgep3080, align 1
  %scevgep3078 = getelementptr i8, ptr %scevgep3077, i64 %residual-loop-index2246
  store i8 %1090, ptr %scevgep3078, align 1
  %1091 = add i64 %residual-loop-index2246, 1
  %1092 = icmp uge i64 %1091, %654
  br i1 %1092, label %Flow3705, label %loop-memcpy-residual2244, !amdgpu.uniform !42

Flow3705:                                         ; preds = %loop-memcpy-residual2244
  br label %Flow3707, !amdgpu.uniform !42

post-loop-memcpy-expansion2241:                   ; preds = %Flow3707
  %tobool.not.i.i.i.i.i.i789 = icmp eq i32 %.lcssa4110, 0
  %conv.i.i.i.i.i.i790 = zext i1 %tobool.not.i.i.i.i.i.i789 to i32
  fence release
  %1093 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %idxprom.i23.i.i.i.i.i791 = zext i32 %.lcssa4112 to i64
  %arrayidx.i24.i.i.i.i.i792 = getelementptr inbounds nuw i32, ptr %1093, i64 %idxprom.i23.i.i.i.i.i791
  store atomic i32 %conv.i.i.i.i.i.i790, ptr %arrayidx.i24.i.i.i.i.i792 monotonic, align 4
  %1094 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !84
  %arrayidx.i.i5.i.i793 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %1094, i64 %idxprom.i23.i.i.i.i.i791, !amdgpu.uniform !42
  %1095 = load i64, ptr %arrayidx.i.i5.i.i793, align 8, !tbaa !80
  %scevgep3062 = getelementptr i8, ptr %1026, i64 56
  br label %while.cond.i.i.i.i794, !amdgpu.uniform !42

while.cond.i.i.i.i794:                            ; preds = %post-loop-memcpy-expansion2241, %Flow3704
  %phi.broken3916 = phi i64 [ 0, %post-loop-memcpy-expansion2241 ], [ %1112, %Flow3704 ]
  %lsr.iv3069 = phi i64 [ %1110, %Flow3704 ], [ %670, %post-loop-memcpy-expansion2241 ]
  %lsr.iv3063 = phi ptr [ %1109, %Flow3704 ], [ %scevgep3062, %post-loop-memcpy-expansion2241 ]
  %port.sroa.37.0.i.i795 = phi i32 [ %1108, %Flow3704 ], [ %conv.i.i.i.i.i.i790, %post-loop-memcpy-expansion2241 ]
  %idx.0.i.i.i.i796 = phi i64 [ %1107, %Flow3704 ], [ 56, %post-loop-memcpy-expansion2241 ]
  %cmp.i.i.i.i797 = icmp ugt i64 %spec.select.i, %idx.0.i.i.i.i796
  %1096 = tail call i64 @llvm.amdgcn.ballot.i64(i1 %cmp.i.i.i.i797)
  %and.i.i.i.i6.i.i798 = and i64 %1096, %1095
  %tobool.not.i.i.i.i799 = icmp ne i64 %and.i.i.i.i6.i.i798, 0
  %1097 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i22.i.i.i800 = getelementptr inbounds nuw i32, ptr %1097, i64 %idxprom.i23.i.i.i.i.i791, !amdgpu.uniform !42
  %1098 = load atomic i32, ptr %arrayidx.i22.i.i.i800 monotonic, align 4
  %1099 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %tobool.not.i.i.i.i799)
  %1100 = extractvalue { i1, i64 } %1099, 0
  %1101 = extractvalue { i1, i64 } %1099, 1
  br i1 %1100, label %cond.false.i8.i.i.i.i801, label %Flow3704

cond.false.i8.i.i.i.i801:                         ; preds = %while.cond.i.i.i.i794
  %1102 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %1098)
  %1103 = icmp ne i32 %1102, %port.sroa.37.0.i.i795
  br i1 %1103, label %while.body.i.i21.i.i.i.i802.preheader, label %Flow3703, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i802.preheader:            ; preds = %cond.false.i8.i.i.i.i801
  br label %while.body.i.i21.i.i.i.i802, !amdgpu.uniform !42

Flow3703:                                         ; preds = %Flow3702, %cond.false.i8.i.i.i.i801
  br label %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i805, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i802:                      ; preds = %while.body.i.i21.i.i.i.i802.preheader, %while.body.i.i21.i.i.i.i802
  tail call void @llvm.amdgcn.s.sleep(i32 2)
  %1104 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i22.i.i.i.i803 = getelementptr inbounds nuw i32, ptr %1104, i64 %idxprom.i23.i.i.i.i.i791, !amdgpu.uniform !42
  %1105 = load atomic i32, ptr %arrayidx.i.i.i22.i.i.i.i803 monotonic, align 4
  %1106 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %1105)
  %cmp.i.not.i.i23.i.i.i.i804 = icmp eq i32 %1106, %port.sroa.37.0.i.i795
  br i1 %cmp.i.not.i.i23.i.i.i.i804, label %Flow3702, label %while.body.i.i21.i.i.i.i802, !amdgpu.uniform !42

Flow3702:                                         ; preds = %while.body.i.i21.i.i.i.i802
  br label %Flow3703, !amdgpu.uniform !42

Flow3704:                                         ; preds = %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i806, %while.cond.i.i.i.i794
  %1107 = phi i64 [ %add.i.i.i.i809, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i806 ], [ poison, %while.cond.i.i.i.i794 ]
  %1108 = phi i32 [ %conv.i.i26.i.i.i.i807, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i806 ], [ poison, %while.cond.i.i.i.i794 ]
  %1109 = phi ptr [ %scevgep3064, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i806 ], [ poison, %while.cond.i.i.i.i794 ]
  %1110 = phi i64 [ %lsr.iv.next3070, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i806 ], [ poison, %while.cond.i.i.i.i794 ]
  %1111 = phi i1 [ false, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i806 ], [ true, %while.cond.i.i.i.i794 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %1101)
  %1112 = call i64 @llvm.amdgcn.if.break.i64(i1 %1111, i64 %phi.broken3916)
  %1113 = call i1 @llvm.amdgcn.loop.i64(i64 %1112)
  br i1 %1113, label %cond.false.i.i.i816, label %while.cond.i.i.i.i794

_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i805: ; preds = %Flow3703
  fence acquire
  br i1 %cmp.i.i.i.i797, label %if.then.i.i.i.i.i.i.i810, label %Flow3701, !amdgpu.uniform !42

if.then.i.i.i.i.i.i.i810:                         ; preds = %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i805
  %sub.i.i.i.i.i.i.i812 = sub nuw i64 %spec.select.i, %idx.0.i.i.i.i796
  %spec.select.i.i.i36.i.i.i.i813 = tail call i64 @llvm.umin.i64(i64 %sub.i.i.i.i.i.i.i812, i64 64)
  %1114 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %1115 = and i64 %spec.select.i.i.i36.i.i.i.i813, 15
  %1116 = sub i64 %spec.select.i.i.i36.i.i.i.i813, %1115
  %1117 = icmp ne i64 %1116, 0
  br i1 %1117, label %loop-memcpy-expansion2236.preheader, label %Flow3700, !amdgpu.uniform !42

loop-memcpy-expansion2236.preheader:              ; preds = %if.then.i.i.i.i.i.i.i810
  %arrayidx.i20.i35.i.i.i.i814 = getelementptr %"struct.rpc::Buffer", ptr %1114, i64 %idxprom.i19.i.i.i.i.i784
  %arrayidx.i22.i38.i.i.i.i815 = getelementptr %"struct.rpc::Buffer", ptr %arrayidx.i20.i35.i.i.i.i814, i64 %sh_prom.i.i.i.i
  br label %loop-memcpy-expansion2236, !amdgpu.uniform !42

Flow3700:                                         ; preds = %Flow3699, %if.then.i.i.i.i.i.i.i810
  br label %loop-memcpy-residual-header2239, !amdgpu.uniform !42

loop-memcpy-expansion2236:                        ; preds = %loop-memcpy-expansion2236.preheader, %loop-memcpy-expansion2236
  %loop-index2237 = phi i64 [ %1120, %loop-memcpy-expansion2236 ], [ 0, %loop-memcpy-expansion2236.preheader ]
  %scevgep3065 = getelementptr i8, ptr %lsr.iv3063, i64 %loop-index2237, !amdgpu.uniform !42
  %1118 = load <4 x i32>, ptr %scevgep3065, align 1
  %1119 = getelementptr i8, ptr %arrayidx.i22.i38.i.i.i.i815, i64 %loop-index2237
  store <4 x i32> %1118, ptr %1119, align 1
  %1120 = add i64 %loop-index2237, 16
  %1121 = icmp uge i64 %1120, %1116
  br i1 %1121, label %Flow3699, label %loop-memcpy-expansion2236, !amdgpu.uniform !42

Flow3698:                                         ; preds = %Flow3697, %loop-memcpy-residual-header2239
  br label %Flow3701, !amdgpu.uniform !42

loop-memcpy-residual2238:                         ; preds = %loop-memcpy-residual2238.preheader, %loop-memcpy-residual2238
  %residual-loop-index2240 = phi i64 [ %1123, %loop-memcpy-residual2238 ], [ 0, %loop-memcpy-residual2238.preheader ]
  %scevgep3075 = getelementptr i8, ptr %scevgep3074, i64 %residual-loop-index2240, !amdgpu.uniform !42
  %1122 = load i8, ptr %scevgep3075, align 1
  %scevgep3073 = getelementptr i8, ptr %scevgep3072, i64 %residual-loop-index2240
  store i8 %1122, ptr %scevgep3073, align 1
  %1123 = add i64 %residual-loop-index2240, 1
  %1124 = icmp uge i64 %1123, %1115
  br i1 %1124, label %Flow3697, label %loop-memcpy-residual2238, !amdgpu.uniform !42

Flow3697:                                         ; preds = %loop-memcpy-residual2238
  br label %Flow3698, !amdgpu.uniform !42

_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i806: ; preds = %Flow3701
  %conv.i.i26.i.i.i.i807 = xor i32 %port.sroa.37.0.i.i795, 1
  fence release
  %1125 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %arrayidx.i24.i29.i.i.i.i808 = getelementptr inbounds nuw i32, ptr %1125, i64 %idxprom.i23.i.i.i.i.i791
  store atomic i32 %conv.i.i26.i.i.i.i807, ptr %arrayidx.i24.i29.i.i.i.i808 monotonic, align 4
  %add.i.i.i.i809 = add i64 %idx.0.i.i.i.i796, 64
  %scevgep3064 = getelementptr i8, ptr %lsr.iv3063, i64 64
  %lsr.iv.next3070 = add i64 %lsr.iv3069, -64
  br label %Flow3704, !amdgpu.uniform !42

cond.false.i.i.i816:                              ; preds = %Flow3704
  %.lcssa4106 = phi i64 [ %1112, %Flow3704 ]
  %port.sroa.37.0.i.i795.lcssa = phi i32 [ %port.sroa.37.0.i.i795, %Flow3704 ]
  %.lcssa4105 = phi i32 [ %1098, %Flow3704 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa4106)
  %1126 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %.lcssa4105)
  %1127 = icmp ne i32 %1126, %port.sroa.37.0.i.i795.lcssa
  %1128 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %1127)
  %1129 = extractvalue { i1, i64 } %1128, 0
  %1130 = extractvalue { i1, i64 } %1128, 1
  br i1 %1129, label %while.body.i.i.i.i817.preheader, label %Flow3696

while.body.i.i.i.i817.preheader:                  ; preds = %cond.false.i.i.i816
  br label %while.body.i.i.i.i817, !amdgpu.uniform !42

Flow3696:                                         ; preds = %Flow3695, %cond.false.i.i.i816
  call void @llvm.amdgcn.end.cf.i64(i64 %1130)
  br label %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsINS_3cpp6stringEEERS1_T_.exit838, !amdgpu.uniform !42

while.body.i.i.i.i817:                            ; preds = %while.body.i.i.i.i817.preheader, %while.body.i.i.i.i817
  %phi.broken3917 = phi i64 [ 0, %while.body.i.i.i.i817.preheader ], [ %1134, %while.body.i.i.i.i817 ]
  tail call void @llvm.amdgcn.s.sleep(i32 2)
  %1131 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i8.i.i818 = getelementptr inbounds nuw i32, ptr %1131, i64 %idxprom.i23.i.i.i.i.i791, !amdgpu.uniform !42
  %1132 = load atomic i32, ptr %arrayidx.i.i.i8.i.i818 monotonic, align 4
  %1133 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %1132)
  %cmp.i.not.i.i.i.i819 = icmp eq i32 %1133, %port.sroa.37.0.i.i795.lcssa
  %1134 = call i64 @llvm.amdgcn.if.break.i64(i1 %cmp.i.not.i.i.i.i819, i64 %phi.broken3917)
  %1135 = call i1 @llvm.amdgcn.loop.i64(i64 %1134)
  br i1 %1135, label %Flow3695, label %while.body.i.i.i.i817

Flow3695:                                         ; preds = %while.body.i.i.i.i817
  %.lcssa4104 = phi i64 [ %1134, %while.body.i.i.i.i817 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa4104)
  br label %Flow3696, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsINS_3cpp6stringEEERS1_T_.exit838: ; preds = %Flow3696
  fence acquire
  tail call void @llvm.amdgcn.wave.barrier()
  fence syncscope("agent") release
  %1136 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %.lcssa4111, i1 true)
  %iszero.i.i.i.i.i13.i.i821 = icmp ne i64 %.lcssa4111, 0
  %cmp2.i.i.i.i14.i.i822 = icmp eq i64 %1136, %sh_prom.i.i.i.i
  %cmp.i.i.i.i15.i.i823 = select i1 %iszero.i.i.i.i.i13.i.i821, i1 %cmp2.i.i.i.i14.i.i822, i1 false
  %1137 = and i32 %.lcssa4112, 31
  %conv4.i.i.i19.i.i827 = zext i1 %cmp.i.i.i.i15.i.i823 to i32
  %shl.i.i.i20.i.i828 = shl nuw i32 %conv4.i.i.i19.i.i827, %1137
  %xor.i.i.i21.i.i829 = xor i32 %shl.i.i.i20.i.i828, -1
  %1138 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2535, !amdgpu.uniform !42

1139:                                             ; preds = %ComputeEnd2536
  %div8.i.i.i16.i.i824 = lshr i32 %.lcssa4112, 5
  %idxprom.i.i.i17.i.i825 = zext nneg i32 %div8.i.i.i16.i.i824 to i64
  %1140 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i17.i.i825
  %sunkaddr3297 = getelementptr inbounds i8, ptr addrspace(1) %1140, i64 40
  %1141 = atomicrmw and ptr addrspace(1) %sunkaddr3297, i32 %.lcssa4103 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %1142, !amdgpu.uniform !42

1142:                                             ; preds = %1139, %ComputeEnd2536
  call void @llvm.amdgcn.end.cf.i64(i64 %3407)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %for.cond.i.i1.i847, !amdgpu.uniform !42

for.cond.i.i1.i847:                               ; preds = %1142, %Flow3694
  %1143 = phi i32 [ %1181, %Flow3694 ], [ poison, %1142 ]
  %1144 = phi i64 [ %1182, %Flow3694 ], [ poison, %1142 ]
  %1145 = phi i32 [ %1183, %Flow3694 ], [ poison, %1142 ]
  %retval.sroa.6.0.i.i.i848 = phi i32 [ %1186, %Flow3694 ], [ undef, %1142 ]
  %retval.sroa.2.0.i.i.i850 = phi i64 [ %1185, %Flow3694 ], [ undef, %1142 ]
  %retval.sroa.8.0.i.i.i851 = phi i32 [ %1184, %Flow3694 ], [ undef, %1142 ]
  %index.0.i.i.i852 = phi i32 [ %1188, %Flow3694 ], [ 0, %1142 ]
  %1146 = load i32, ptr addrspace(1) @__llvm_rpc_client, align 8, !tbaa !68
  %cmp.not.i.i.i853 = icmp ult i32 %index.0.i.i.i852, %1146
  %spec.store.select.i.i.i854 = select i1 %cmp.not.i.i.i853, i32 %index.0.i.i.i852, i32 0
  %1147 = tail call noundef i64 @llvm.amdgcn.ballot.i64(i1 true)
  %1148 = and i32 %spec.store.select.i.i.i854, 31
  %1149 = lshr i64 %1147, %sh_prom.i.i.i.i
  %1150 = trunc i64 %1149 to i32
  %conv4.i.i.i.i.i858 = and i32 %1150, 1
  %shl.i.i.i.i.i859 = shl nuw i32 %conv4.i.i.i.i.i858, %1148
  %1151 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2539, !amdgpu.uniform !42

1152:                                             ; preds = %ComputeEnd2540
  %sunkaddr3298 = getelementptr inbounds i8, ptr addrspace(1) %3417, i64 40
  %1153 = atomicrmw or ptr addrspace(1) %sunkaddr3298, i32 %.lcssa4097 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %1154, !amdgpu.uniform !42

1154:                                             ; preds = %1152, %ComputeEnd2540
  %1155 = phi i32 [ %1153, %1152 ], [ poison, %ComputeEnd2540 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %3426)
  %1156 = call i32 @llvm.amdgcn.readfirstlane.i32(i32 %1155)
  %1157 = or i32 %1156, %.lcssa4098
  %shl5.i.i.i.i.i860 = shl nuw i32 1, %1148
  %and.i.i.i.i.i861 = and i32 %shl5.i.i.i.i.i860, %1157
  %tobool3.i.i.i.i862 = icmp ne i32 %and.i.i.i.i.i861, 0
  %1158 = tail call i64 @llvm.amdgcn.ballot.i64(i1 %tobool3.i.i.i.i862)
  %and.i.i.i.i.i.i863 = and i64 %1158, %1147
  %cmp.i.not.i.i.i864 = icmp ne i64 %1147, %and.i.i.i.i.i.i863
  br i1 %cmp.i.not.i.i.i864, label %if.then.i.i.i.i865, label %Flow3693, !amdgpu.uniform !42

if.then.i.i.i.i865:                               ; preds = %1154
  fence syncscope("agent") acquire
  %1159 = load <2 x i64>, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !75
  %1160 = extractelement <2 x i64> %1159, i32 0
  %1161 = inttoptr i64 %1160 to ptr
  %1162 = extractelement <2 x i64> %1159, i32 1
  %1163 = inttoptr i64 %1162 to ptr
  %idxprom.i.i.i.i866 = zext i32 %spec.store.select.i.i.i854 to i64
  %arrayidx.i.i.i.i867 = getelementptr inbounds nuw i32, ptr %1161, i64 %idxprom.i.i.i.i866, !amdgpu.uniform !42
  %1164 = load atomic i32, ptr %arrayidx.i.i.i.i867 monotonic, align 4
  %1165 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %1164)
  %arrayidx.i45.i.i.i868 = getelementptr inbounds nuw i32, ptr %1163, i64 %idxprom.i.i.i.i866, !amdgpu.uniform !42
  %1166 = load atomic i32, ptr %arrayidx.i45.i.i.i868 monotonic, align 4
  %1167 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %1166)
  %cmp.i46.not.i.i.i869 = icmp ne i32 %1165, %1167
  br i1 %cmp.i46.not.i.i.i869, label %if.then12.i.i.i870, label %Flow3691, !amdgpu.uniform !42

if.then12.i.i.i870:                               ; preds = %if.then.i.i.i.i865
  fence syncscope("agent") release
  %1168 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %1147, i1 true)
  %iszero.i.i.i.i.i.i.i871 = icmp ne i64 %1147, 0
  %cmp2.i.i.i.i.i.i872 = icmp eq i64 %1168, %sh_prom.i.i.i.i
  %cmp.i.i.i.i.i.i873 = select i1 %iszero.i.i.i.i.i.i.i871, i1 %cmp2.i.i.i.i.i.i872, i1 false
  %conv4.i.i51.i.i.i874 = zext i1 %cmp.i.i.i.i.i.i873 to i32
  %shl.i.i52.i.i.i875 = shl nuw i32 %conv4.i.i51.i.i.i874, %1148
  %xor.i.i.i.i.i876 = xor i32 %shl.i.i52.i.i.i875, -1
  %1169 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2544, !amdgpu.uniform !42

1170:                                             ; preds = %ComputeEnd2545
  %sunkaddr3299 = getelementptr inbounds i8, ptr addrspace(1) %3417, i64 40
  %1171 = atomicrmw and ptr addrspace(1) %sunkaddr3299, i32 %.lcssa4099 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %1172, !amdgpu.uniform !42

1172:                                             ; preds = %1170, %ComputeEnd2545
  call void @llvm.amdgcn.end.cf.i64(i64 %3444)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %Flow3691, !amdgpu.uniform !42

if.end14.i.i.i936:                                ; preds = %Flow3691
  %1173 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %1147, i1 true)
  %iszero.i.i.i.i.i.i937 = icmp ne i64 %1147, 0
  %cmp2.i.i.i.i.i938 = icmp eq i64 %1173, %sh_prom.i.i.i.i
  %cmp.i.i.i.i.i939 = select i1 %iszero.i.i.i.i.i.i937, i1 %cmp2.i.i.i.i.i938, i1 false
  %1174 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.i.i.i.i.i939)
  %1175 = extractvalue { i1, i64 } %1174, 0
  %1176 = extractvalue { i1, i64 } %1174, 1
  br i1 %1175, label %if.then16.i.i.i941, label %if.end22.i.i.i940

if.then16.i.i.i941:                               ; preds = %if.end14.i.i.i936
  %1177 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !76
  %opcode.i.i.i942 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %1177, i64 %idxprom.i.i.i.i866, i32 1
  store i32 1660944387, ptr %opcode.i.i.i942, align 8, !tbaa !77
  %arrayidx21.i.i.i943 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %1177, i64 %idxprom.i.i.i.i866
  store i64 %1147, ptr %arrayidx21.i.i.i943, align 8, !tbaa !80
  br label %if.end22.i.i.i940, !amdgpu.uniform !42

Flow3692:                                         ; preds = %if.end22.i.i.i940, %Flow3691
  %1178 = phi i32 [ %1165, %if.end22.i.i.i940 ], [ %retval.sroa.8.0.i.i.i851, %Flow3691 ]
  %1179 = phi i64 [ %1147, %if.end22.i.i.i940 ], [ %retval.sroa.2.0.i.i.i850, %Flow3691 ]
  %1180 = phi i32 [ %spec.store.select.i.i.i854, %if.end22.i.i.i940 ], [ %retval.sroa.6.0.i.i.i848, %Flow3691 ]
  br label %cleanup26.i.i.i877, !amdgpu.uniform !42

if.end22.i.i.i940:                                ; preds = %if.then16.i.i.i941, %if.end14.i.i.i936
  call void @llvm.amdgcn.end.cf.i64(i64 %1176)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %Flow3692, !amdgpu.uniform !42

Flow3693:                                         ; preds = %cleanup26.i.i.i877, %1154
  %1181 = phi i32 [ %1180, %cleanup26.i.i.i877 ], [ %1143, %1154 ]
  %1182 = phi i64 [ %1179, %cleanup26.i.i.i877 ], [ %1144, %1154 ]
  %1183 = phi i32 [ %1178, %cleanup26.i.i.i877 ], [ %1145, %1154 ]
  %1184 = phi i32 [ %1178, %cleanup26.i.i.i877 ], [ %retval.sroa.8.0.i.i.i851, %1154 ]
  %1185 = phi i64 [ %1179, %cleanup26.i.i.i877 ], [ %retval.sroa.2.0.i.i.i850, %1154 ]
  %1186 = phi i32 [ %1180, %cleanup26.i.i.i877 ], [ %retval.sroa.6.0.i.i.i848, %1154 ]
  %1187 = phi i1 [ %cmp.i46.not.i.i.i869, %cleanup26.i.i.i877 ], [ true, %1154 ]
  br i1 %1187, label %for.inc.i.i.i882, label %Flow3694, !amdgpu.uniform !42

cleanup26.i.i.i877:                               ; preds = %Flow3692
  br label %Flow3693, !amdgpu.uniform !42

for.inc.i.i.i882:                                 ; preds = %Flow3693
  %inc.i.i.i887 = add i32 %spec.store.select.i.i.i854, 1
  br label %Flow3694, !amdgpu.uniform !42

Flow3694:                                         ; preds = %for.inc.i.i.i882, %Flow3693
  %1188 = phi i32 [ %inc.i.i.i887, %for.inc.i.i.i882 ], [ poison, %Flow3693 ]
  %1189 = phi i1 [ false, %for.inc.i.i.i882 ], [ true, %Flow3693 ]
  br i1 %1189, label %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i888, label %for.cond.i.i1.i847, !amdgpu.uniform !42

_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i888: ; preds = %Flow3694
  %.lcssa4102 = phi i32 [ %1181, %Flow3694 ]
  %.lcssa4101 = phi i64 [ %1182, %Flow3694 ]
  %.lcssa4100 = phi i32 [ %1183, %Flow3694 ]
  fence acquire
  %1190 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %mul.i.i.i.i.i.i889 = shl i32 %.lcssa4102, 6
  %idxprom.i19.i.i.i.i.i890 = zext i32 %mul.i.i.i.i.i.i889 to i64
  %arrayidx.i20.i.i.i.i.i891 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %1190, i64 %idxprom.i19.i.i.i.i.i890
  %arrayidx.i22.i.i.i.i.i892 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %arrayidx.i20.i.i.i.i.i891, i64 %sh_prom.i.i.i.i
  store i64 sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.35 to ptr), i64 10) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.35 to ptr) to i64)), ptr %arrayidx.i22.i.i.i.i.i892, align 8, !tbaa !82
  %spec.select.i.i.i.i.i.i.i893 = tail call i64 @llvm.umin.i64(i64 sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.35 to ptr), i64 10) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.35 to ptr) to i64)), i64 56)
  %1191 = and i64 %spec.select.i.i.i.i.i.i.i893, 15
  %1192 = sub i64 %spec.select.i.i.i.i.i.i.i893, %1191
  %1193 = icmp ne i64 %1192, 0
  br i1 %1193, label %loop-memcpy-expansion2368.preheader, label %Flow3690, !amdgpu.uniform !42

loop-memcpy-expansion2368.preheader:              ; preds = %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i888
  %1194 = add nuw nsw i64 %75, 8
  %scevgep3059 = getelementptr i8, ptr %1190, i64 %1194
  %1195 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i890, 6
  %scevgep3060 = getelementptr i8, ptr %scevgep3059, i64 %1195
  br label %loop-memcpy-expansion2368, !amdgpu.uniform !42

Flow3690:                                         ; preds = %Flow3689, %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i888
  br label %loop-memcpy-residual-header2371, !amdgpu.uniform !42

loop-memcpy-expansion2368:                        ; preds = %loop-memcpy-expansion2368.preheader, %loop-memcpy-expansion2368
  %loop-index2369 = phi i64 [ %1198, %loop-memcpy-expansion2368 ], [ 0, %loop-memcpy-expansion2368.preheader ]
  %1196 = getelementptr i8, ptr addrspace(4) @.str.35, i64 %loop-index2369, !amdgpu.uniform !42
  %1197 = load <4 x i32>, ptr addrspace(4) %1196, align 1
  %scevgep3061 = getelementptr i8, ptr %scevgep3060, i64 %loop-index2369
  store <4 x i32> %1197, ptr %scevgep3061, align 1
  %1198 = add i64 %loop-index2369, 16
  %1199 = icmp uge i64 %1198, %1192
  br i1 %1199, label %Flow3689, label %loop-memcpy-expansion2368, !amdgpu.uniform !42

Flow3688:                                         ; preds = %Flow3686, %Flow3687
  %1200 = phi i64 [ %2884, %Flow3686 ], [ %2881, %Flow3687 ]
  br label %post-loop-memcpy-expansion2367, !amdgpu.uniform !42

loop-memcpy-residual2370:                         ; preds = %loop-memcpy-residual2370.preheader, %loop-memcpy-residual2370
  %residual-loop-index2372 = phi i64 [ %1202, %loop-memcpy-residual2370 ], [ 0, %loop-memcpy-residual2370.preheader ]
  %scevgep3058 = getelementptr i8, ptr addrspace(4) @.str.35, i64 %residual-loop-index2372, !amdgpu.uniform !42
  %1201 = load i8, ptr addrspace(4) %scevgep3058, align 1
  %scevgep3057 = getelementptr i8, ptr %scevgep3056, i64 %residual-loop-index2372
  store i8 %1201, ptr %scevgep3057, align 1
  %1202 = add i64 %residual-loop-index2372, 1
  %1203 = icmp uge i64 %1202, %1191
  br i1 %1203, label %Flow3686, label %loop-memcpy-residual2370, !amdgpu.uniform !42

Flow3686:                                         ; preds = %loop-memcpy-residual2370
  br label %Flow3688, !amdgpu.uniform !42

post-loop-memcpy-expansion2367:                   ; preds = %Flow3688
  %tobool.not.i.i.i.i.i.i895 = icmp eq i32 %.lcssa4100, 0
  %conv.i.i.i.i.i.i896 = zext i1 %tobool.not.i.i.i.i.i.i895 to i32
  fence release
  %1204 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %idxprom.i23.i.i.i.i.i897 = zext i32 %.lcssa4102 to i64
  %arrayidx.i24.i.i.i.i.i898 = getelementptr inbounds nuw i32, ptr %1204, i64 %idxprom.i23.i.i.i.i.i897
  store atomic i32 %conv.i.i.i.i.i.i896, ptr %arrayidx.i24.i.i.i.i.i898 monotonic, align 4
  %1205 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !84
  %arrayidx.i.i5.i.i899 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %1205, i64 %idxprom.i23.i.i.i.i.i897, !amdgpu.uniform !42
  %1206 = load i64, ptr %arrayidx.i.i5.i.i899, align 8, !tbaa !80
  br label %while.cond.i.i.i.i900, !amdgpu.uniform !42

while.cond.i.i.i.i900:                            ; preds = %post-loop-memcpy-expansion2367, %Flow3684
  %phi.broken3918 = phi i64 [ 0, %post-loop-memcpy-expansion2367 ], [ %1223, %Flow3684 ]
  %lsr.iv3048 = phi i64 [ %1221, %Flow3684 ], [ -46, %post-loop-memcpy-expansion2367 ]
  %lsr.iv3043 = phi ptr addrspace(4) [ %1220, %Flow3684 ], [ getelementptr (i8, ptr addrspace(4) @.str.35, i64 56), %post-loop-memcpy-expansion2367 ]
  %port.sroa.37.0.i.i901 = phi i32 [ %1219, %Flow3684 ], [ %conv.i.i.i.i.i.i896, %post-loop-memcpy-expansion2367 ]
  %idx.0.i.i.i.i902 = phi i64 [ %1218, %Flow3684 ], [ 56, %post-loop-memcpy-expansion2367 ]
  %cmp.i.i.i.i903 = icmp ult i64 %idx.0.i.i.i.i902, sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.35 to ptr), i64 10) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.35 to ptr) to i64))
  %1207 = tail call i64 @llvm.amdgcn.ballot.i64(i1 %cmp.i.i.i.i903)
  %and.i.i.i.i6.i.i904 = and i64 %1207, %1206
  %tobool.not.i.i.i.i905 = icmp ne i64 %and.i.i.i.i6.i.i904, 0
  %1208 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i22.i.i.i906 = getelementptr inbounds nuw i32, ptr %1208, i64 %idxprom.i23.i.i.i.i.i897, !amdgpu.uniform !42
  %1209 = load atomic i32, ptr %arrayidx.i22.i.i.i906 monotonic, align 4
  %1210 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %tobool.not.i.i.i.i905)
  %1211 = extractvalue { i1, i64 } %1210, 0
  %1212 = extractvalue { i1, i64 } %1210, 1
  br i1 %1211, label %cond.false.i8.i.i.i.i907, label %Flow3684

cond.false.i8.i.i.i.i907:                         ; preds = %while.cond.i.i.i.i900
  %1213 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %1209)
  %1214 = icmp ne i32 %1213, %port.sroa.37.0.i.i901
  br i1 %1214, label %while.body.i.i21.i.i.i.i908.preheader, label %Flow3683, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i908.preheader:            ; preds = %cond.false.i8.i.i.i.i907
  br label %while.body.i.i21.i.i.i.i908, !amdgpu.uniform !42

Flow3683:                                         ; preds = %Flow3681, %cond.false.i8.i.i.i.i907
  br label %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i911, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i908:                      ; preds = %while.body.i.i21.i.i.i.i908.preheader, %while.body.i.i21.i.i.i.i908
  tail call void @llvm.amdgcn.s.sleep(i32 2)
  %1215 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i22.i.i.i.i909 = getelementptr inbounds nuw i32, ptr %1215, i64 %idxprom.i23.i.i.i.i.i897, !amdgpu.uniform !42
  %1216 = load atomic i32, ptr %arrayidx.i.i.i22.i.i.i.i909 monotonic, align 4
  %1217 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %1216)
  %cmp.i.not.i.i23.i.i.i.i910 = icmp eq i32 %1217, %port.sroa.37.0.i.i901
  br i1 %cmp.i.not.i.i23.i.i.i.i910, label %Flow3681, label %while.body.i.i21.i.i.i.i908, !amdgpu.uniform !42

Flow3681:                                         ; preds = %while.body.i.i21.i.i.i.i908
  br label %Flow3683, !amdgpu.uniform !42

Flow3684:                                         ; preds = %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i912, %while.cond.i.i.i.i900
  %1218 = phi i64 [ %add.i.i.i.i915, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i912 ], [ poison, %while.cond.i.i.i.i900 ]
  %1219 = phi i32 [ %conv.i.i26.i.i.i.i913, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i912 ], [ poison, %while.cond.i.i.i.i900 ]
  %1220 = phi ptr addrspace(4) [ %scevgep3044, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i912 ], [ poison, %while.cond.i.i.i.i900 ]
  %1221 = phi i64 [ %lsr.iv.next3049, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i912 ], [ poison, %while.cond.i.i.i.i900 ]
  %1222 = phi i1 [ false, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i912 ], [ true, %while.cond.i.i.i.i900 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %1212)
  %1223 = call i64 @llvm.amdgcn.if.break.i64(i1 %1222, i64 %phi.broken3918)
  %1224 = call i1 @llvm.amdgcn.loop.i64(i64 %1223)
  br i1 %1224, label %cond.false.i.i.i922, label %while.cond.i.i.i.i900

_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i911: ; preds = %Flow3683
  fence acquire
  br i1 %cmp.i.i.i.i903, label %if.then.i.i.i.i.i.i.i916, label %Flow3680, !amdgpu.uniform !42

if.then.i.i.i.i.i.i.i916:                         ; preds = %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i911
  %sub.i.i.i.i.i.i.i918 = sub nuw i64 sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.35 to ptr), i64 10) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.35 to ptr) to i64)), %idx.0.i.i.i.i902
  %spec.select.i.i.i36.i.i.i.i919 = tail call i64 @llvm.umin.i64(i64 %sub.i.i.i.i.i.i.i918, i64 64)
  %1225 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %1226 = and i64 %spec.select.i.i.i36.i.i.i.i919, 15
  %1227 = sub i64 %spec.select.i.i.i36.i.i.i.i919, %1226
  %1228 = icmp ne i64 %1227, 0
  br i1 %1228, label %loop-memcpy-expansion2380.preheader, label %Flow3679, !amdgpu.uniform !42

loop-memcpy-expansion2380.preheader:              ; preds = %if.then.i.i.i.i.i.i.i916
  %arrayidx.i20.i35.i.i.i.i920 = getelementptr %"struct.rpc::Buffer", ptr %1225, i64 %idxprom.i19.i.i.i.i.i890
  %arrayidx.i22.i38.i.i.i.i921 = getelementptr %"struct.rpc::Buffer", ptr %arrayidx.i20.i35.i.i.i.i920, i64 %sh_prom.i.i.i.i
  br label %loop-memcpy-expansion2380, !amdgpu.uniform !42

Flow3679:                                         ; preds = %Flow3678, %if.then.i.i.i.i.i.i.i916
  br label %loop-memcpy-residual-header2383, !amdgpu.uniform !42

loop-memcpy-expansion2380:                        ; preds = %loop-memcpy-expansion2380.preheader, %loop-memcpy-expansion2380
  %loop-index2381 = phi i64 [ %1231, %loop-memcpy-expansion2380 ], [ 0, %loop-memcpy-expansion2380.preheader ]
  %scevgep3045 = getelementptr i8, ptr addrspace(4) %lsr.iv3043, i64 %loop-index2381, !amdgpu.uniform !42
  %1229 = load <4 x i32>, ptr addrspace(4) %scevgep3045, align 1
  %1230 = getelementptr i8, ptr %arrayidx.i22.i38.i.i.i.i921, i64 %loop-index2381
  store <4 x i32> %1229, ptr %1230, align 1
  %1231 = add i64 %loop-index2381, 16
  %1232 = icmp uge i64 %1231, %1227
  br i1 %1232, label %Flow3678, label %loop-memcpy-expansion2380, !amdgpu.uniform !42

Flow3676:                                         ; preds = %Flow3675, %loop-memcpy-residual-header2383
  br label %Flow3680, !amdgpu.uniform !42

loop-memcpy-residual2382:                         ; preds = %loop-memcpy-residual2382.preheader, %loop-memcpy-residual2382
  %residual-loop-index2384 = phi i64 [ %1234, %loop-memcpy-residual2382 ], [ 0, %loop-memcpy-residual2382.preheader ]
  %scevgep3054 = getelementptr i8, ptr addrspace(4) %scevgep3053, i64 %residual-loop-index2384, !amdgpu.uniform !42
  %1233 = load i8, ptr addrspace(4) %scevgep3054, align 1
  %scevgep3052 = getelementptr i8, ptr %scevgep3051, i64 %residual-loop-index2384
  store i8 %1233, ptr %scevgep3052, align 1
  %1234 = add i64 %residual-loop-index2384, 1
  %1235 = icmp uge i64 %1234, %1226
  br i1 %1235, label %Flow3675, label %loop-memcpy-residual2382, !amdgpu.uniform !42

Flow3675:                                         ; preds = %loop-memcpy-residual2382
  br label %Flow3676, !amdgpu.uniform !42

_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i912: ; preds = %Flow3680
  %conv.i.i26.i.i.i.i913 = xor i32 %port.sroa.37.0.i.i901, 1
  fence release
  %1236 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %arrayidx.i24.i29.i.i.i.i914 = getelementptr inbounds nuw i32, ptr %1236, i64 %idxprom.i23.i.i.i.i.i897
  store atomic i32 %conv.i.i26.i.i.i.i913, ptr %arrayidx.i24.i29.i.i.i.i914 monotonic, align 4
  %add.i.i.i.i915 = add i64 %idx.0.i.i.i.i902, 64
  %scevgep3044 = getelementptr i8, ptr addrspace(4) %lsr.iv3043, i64 64
  %lsr.iv.next3049 = add i64 %lsr.iv3048, -64
  br label %Flow3684, !amdgpu.uniform !42

cond.false.i.i.i922:                              ; preds = %Flow3684
  %.lcssa4096 = phi i64 [ %1223, %Flow3684 ]
  %port.sroa.37.0.i.i901.lcssa = phi i32 [ %port.sroa.37.0.i.i901, %Flow3684 ]
  %.lcssa4095 = phi i32 [ %1209, %Flow3684 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa4096)
  %1237 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %.lcssa4095)
  %1238 = icmp ne i32 %1237, %port.sroa.37.0.i.i901.lcssa
  %1239 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %1238)
  %1240 = extractvalue { i1, i64 } %1239, 0
  %1241 = extractvalue { i1, i64 } %1239, 1
  br i1 %1240, label %while.body.i.i.i.i923.preheader, label %Flow3673

while.body.i.i.i.i923.preheader:                  ; preds = %cond.false.i.i.i922
  br label %while.body.i.i.i.i923, !amdgpu.uniform !42

Flow3673:                                         ; preds = %Flow3671, %cond.false.i.i.i922
  call void @llvm.amdgcn.end.cf.i64(i64 %1241)
  br label %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit944, !amdgpu.uniform !42

while.body.i.i.i.i923:                            ; preds = %while.body.i.i.i.i923.preheader, %while.body.i.i.i.i923
  %phi.broken3919 = phi i64 [ 0, %while.body.i.i.i.i923.preheader ], [ %1245, %while.body.i.i.i.i923 ]
  tail call void @llvm.amdgcn.s.sleep(i32 2)
  %1242 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i8.i.i924 = getelementptr inbounds nuw i32, ptr %1242, i64 %idxprom.i23.i.i.i.i.i897, !amdgpu.uniform !42
  %1243 = load atomic i32, ptr %arrayidx.i.i.i8.i.i924 monotonic, align 4
  %1244 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %1243)
  %cmp.i.not.i.i.i.i925 = icmp eq i32 %1244, %port.sroa.37.0.i.i901.lcssa
  %1245 = call i64 @llvm.amdgcn.if.break.i64(i1 %cmp.i.not.i.i.i.i925, i64 %phi.broken3919)
  %1246 = call i1 @llvm.amdgcn.loop.i64(i64 %1245)
  br i1 %1246, label %Flow3671, label %while.body.i.i.i.i923

Flow3671:                                         ; preds = %while.body.i.i.i.i923
  %.lcssa4094 = phi i64 [ %1245, %while.body.i.i.i.i923 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa4094)
  br label %Flow3673, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit944: ; preds = %Flow3673
  fence acquire
  tail call void @llvm.amdgcn.wave.barrier()
  fence syncscope("agent") release
  %1247 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %.lcssa4101, i1 true)
  %iszero.i.i.i.i.i13.i.i927 = icmp ne i64 %.lcssa4101, 0
  %cmp2.i.i.i.i14.i.i928 = icmp eq i64 %1247, %sh_prom.i.i.i.i
  %cmp.i.i.i.i15.i.i929 = select i1 %iszero.i.i.i.i.i13.i.i927, i1 %cmp2.i.i.i.i14.i.i928, i1 false
  %1248 = and i32 %.lcssa4102, 31
  %conv4.i.i.i19.i.i933 = zext i1 %cmp.i.i.i.i15.i.i929 to i32
  %shl.i.i.i20.i.i934 = shl nuw i32 %conv4.i.i.i19.i.i933, %1248
  %xor.i.i.i21.i.i935 = xor i32 %shl.i.i.i20.i.i934, -1
  %1249 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2548, !amdgpu.uniform !42

1250:                                             ; preds = %ComputeEnd2549
  %div8.i.i.i16.i.i930 = lshr i32 %.lcssa4102, 5
  %idxprom.i.i.i17.i.i931 = zext nneg i32 %div8.i.i.i16.i.i930 to i64
  %1251 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i17.i.i931
  %sunkaddr3300 = getelementptr inbounds i8, ptr addrspace(1) %1251, i64 40
  %1252 = atomicrmw and ptr addrspace(1) %sunkaddr3300, i32 %.lcssa4093 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %1253, !amdgpu.uniform !42

1253:                                             ; preds = %1250, %ComputeEnd2549
  call void @llvm.amdgcn.end.cf.i64(i64 %3461)
  tail call void @llvm.amdgcn.wave.barrier()
  call void @llvm.lifetime.start.p5(ptr addrspace(5) %buffer.i.i.i) #34, !noalias !94
  %cmp7.i.i.i.i.i.i.i = icmp ne i8 %LHS, 0
  %1254 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp7.i.i.i.i.i.i.i)
  %1255 = extractvalue { i1, i64 } %1254, 0
  %1256 = extractvalue { i1, i64 } %1254, 1
  br i1 %1255, label %while.body.lr.ph.i.i.i.i.i.i.i, label %Flow3670

while.body.lr.ph.i.i.i.i.i.i.i:                   ; preds = %1253
  %1257 = tail call i8 @llvm.abs.i8(i8 %LHS, i1 false)
  %retval.0.i.i.i.i.i.i.i = zext i8 %1257 to i32
  br label %while.body.i.i.i.i.i.i.i, !amdgpu.uniform !42

Flow3670:                                         ; preds = %Flow3669, %1253
  %1258 = phi i1 [ %cmp.i.i.i.i.i.i.i.i.i.lcssa, %Flow3669 ], [ false, %1253 ]
  %1259 = phi i64 [ %writer.sroa.10.1.i.i.i.i.i.lcssa, %Flow3669 ], [ 0, %1253 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %1256)
  br label %_ZN22__llvm_libc_22_0_0_git15IntegerToStringIiNS_7details3FmtILh10ELb0ELb0ELb0ELm1EEEE13IntegerWriter25write_unsigned_number_decEjRNS1_22StringBufferWriterImplILb0EEE.exit.i.i.i.i.i.i, !amdgpu.uniform !42

while.body.i.i.i.i.i.i.i:                         ; preds = %while.body.lr.ph.i.i.i.i.i.i.i, %_ZN22__llvm_libc_22_0_0_git7details22StringBufferWriterImplILb0EE4pushEc.exit.i.i.i.i.i.i.i
  %phi.broken3920 = phi i64 [ 0, %while.body.lr.ph.i.i.i.i.i.i.i ], [ %1327, %_ZN22__llvm_libc_22_0_0_git7details22StringBufferWriterImplILb0EE4pushEc.exit.i.i.i.i.i.i.i ]
  %writer.sroa.10.0.i.i.i.i.i = phi i64 [ %writer.sroa.10.1.i.i.i.i.i, %_ZN22__llvm_libc_22_0_0_git7details22StringBufferWriterImplILb0EE4pushEc.exit.i.i.i.i.i.i.i ], [ 0, %while.body.lr.ph.i.i.i.i.i.i.i ]
  %value.addr.09.i.i.i.i.i.i.i = phi i32 [ %div.i.i.i.i.i.i.i.i, %_ZN22__llvm_libc_22_0_0_git7details22StringBufferWriterImplILb0EE4pushEc.exit.i.i.i.i.i.i.i ], [ %retval.0.i.i.i.i.i.i.i, %while.body.lr.ph.i.i.i.i.i.i.i ]
  br label %NodeBlock3368, !amdgpu.uniform !42

NodeBlock3368:                                    ; preds = %while.body.i.i.i.i.i.i.i
  %div.i.i.i.i.i.i.i.i = udiv i32 %value.addr.09.i.i.i.i.i.i.i, 10
  %.neg = call i32 @llvm.amdgcn.mul.u24.i32(i32 %div.i.i.i.i.i.i.i.i, i32 246)
  %rem.i.i.i.i.i.i.i.i.decomposed = add nuw nsw i32 %.neg, %value.addr.09.i.i.i.i.i.i.i
  %conv.i.i.i.i.i.i.i.i = trunc i32 %rem.i.i.i.i.i.i.i.i.decomposed to i8
  %1260 = zext i8 %conv.i.i.i.i.i.i.i.i to i16
  %Pivot3369 = icmp sge i16 %1260, 5
  %1261 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %Pivot3369)
  %1262 = extractvalue { i1, i64 } %1261, 0
  %1263 = extractvalue { i1, i64 } %1261, 1
  br i1 %1262, label %NodeBlock3366, label %Flow3668

NodeBlock3366:                                    ; preds = %NodeBlock3368
  %Pivot3367 = icmp sge i16 %1260, 7
  %1264 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %Pivot3367)
  %1265 = extractvalue { i1, i64 } %1264, 0
  %1266 = extractvalue { i1, i64 } %1264, 1
  br i1 %1265, label %NodeBlock3364, label %Flow3659

NodeBlock3364:                                    ; preds = %NodeBlock3366
  %Pivot3365 = icmp sge i16 %1260, 8
  %1267 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %Pivot3365)
  %1268 = extractvalue { i1, i64 } %1267, 0
  %1269 = extractvalue { i1, i64 } %1267, 1
  br i1 %1268, label %NodeBlock3362, label %Flow3655

NodeBlock3362:                                    ; preds = %NodeBlock3364
  %Pivot3363 = icmp sge i16 %1260, 9
  %1270 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %Pivot3363)
  %1271 = extractvalue { i1, i64 } %1270, 0
  %1272 = extractvalue { i1, i64 } %1270, 1
  br i1 %1271, label %sw.bb9.i.i.i.i.i.i.i.i.i, label %Flow3653

Flow3659:                                         ; preds = %Flow3656, %NodeBlock3366
  %1273 = phi i8 [ %1317, %Flow3656 ], [ poison, %NodeBlock3366 ]
  %1274 = call { i1, i64 } @llvm.amdgcn.else.i64.i64(i64 %1266)
  %1275 = extractvalue { i1, i64 } %1274, 0
  %1276 = extractvalue { i1, i64 } %1274, 1
  br i1 %1275, label %NodeBlock3360, label %Flow3660

NodeBlock3360:                                    ; preds = %Flow3659
  %Pivot3361 = icmp sge i16 %1260, 6
  %1277 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %Pivot3361)
  %1278 = extractvalue { i1, i64 } %1277, 0
  %1279 = extractvalue { i1, i64 } %1277, 1
  br i1 %1278, label %sw.bb6.i.i.i.i.i.i.i.i.i, label %Flow3657

Flow3668:                                         ; preds = %Flow3660, %NodeBlock3368
  %1280 = phi i8 [ %1319, %Flow3660 ], [ poison, %NodeBlock3368 ]
  %1281 = call { i1, i64 } @llvm.amdgcn.else.i64.i64(i64 %1263)
  %1282 = extractvalue { i1, i64 } %1281, 0
  %1283 = extractvalue { i1, i64 } %1281, 1
  br i1 %1282, label %NodeBlock3358, label %if.then.i.i.i.i.i.i.i.i

NodeBlock3358:                                    ; preds = %Flow3668
  %Pivot3359 = icmp sge i16 %1260, 3
  %1284 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %Pivot3359)
  %1285 = extractvalue { i1, i64 } %1284, 0
  %1286 = extractvalue { i1, i64 } %1284, 1
  br i1 %1285, label %NodeBlock3356, label %Flow3666

NodeBlock3356:                                    ; preds = %NodeBlock3358
  %Pivot3357 = icmp sge i16 %1260, 4
  %1287 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %Pivot3357)
  %1288 = extractvalue { i1, i64 } %1287, 0
  %1289 = extractvalue { i1, i64 } %1287, 1
  br i1 %1288, label %sw.bb4.i.i.i.i.i.i.i.i.i, label %Flow3661

Flow3666:                                         ; preds = %Flow3662, %NodeBlock3358
  %1290 = phi i8 [ %1320, %Flow3662 ], [ poison, %NodeBlock3358 ]
  %1291 = call { i1, i64 } @llvm.amdgcn.else.i64.i64(i64 %1286)
  %1292 = extractvalue { i1, i64 } %1291, 0
  %1293 = extractvalue { i1, i64 } %1291, 1
  br i1 %1292, label %NodeBlock3354, label %Flow3667

NodeBlock3354:                                    ; preds = %Flow3666
  %Pivot3355 = icmp sge i16 %1260, 2
  %1294 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %Pivot3355)
  %1295 = extractvalue { i1, i64 } %1294, 0
  %1296 = extractvalue { i1, i64 } %1294, 1
  br i1 %1295, label %sw.bb2.i.i.i.i.i.i.i.i.i, label %Flow3664

Flow3664:                                         ; preds = %sw.bb2.i.i.i.i.i.i.i.i.i, %NodeBlock3354
  %1297 = call { i1, i64 } @llvm.amdgcn.else.i64.i64(i64 %1296)
  %1298 = extractvalue { i1, i64 } %1297, 0
  %1299 = extractvalue { i1, i64 } %1297, 1
  br i1 %1298, label %LeafBlock3352, label %Flow3665

LeafBlock3352:                                    ; preds = %Flow3664
  %SwitchLeaf3353 = icmp eq i16 %1260, 1
  %1300 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %SwitchLeaf3353)
  %1301 = extractvalue { i1, i64 } %1300, 0
  %1302 = extractvalue { i1, i64 } %1300, 1
  br i1 %1301, label %sw.bb1.i.i.i.i.i.i.i.i.i, label %Flow3663

sw.bb1.i.i.i.i.i.i.i.i.i:                         ; preds = %LeafBlock3352
  br label %Flow3663, !amdgpu.uniform !42

sw.bb2.i.i.i.i.i.i.i.i.i:                         ; preds = %NodeBlock3354
  br label %Flow3664, !amdgpu.uniform !42

Flow3661:                                         ; preds = %sw.bb4.i.i.i.i.i.i.i.i.i, %NodeBlock3356
  %1303 = call { i1, i64 } @llvm.amdgcn.else.i64.i64(i64 %1289)
  %1304 = extractvalue { i1, i64 } %1303, 0
  %1305 = extractvalue { i1, i64 } %1303, 1
  br i1 %1304, label %sw.bb3.i.i.i.i.i.i.i.i.i, label %Flow3662

sw.bb3.i.i.i.i.i.i.i.i.i:                         ; preds = %Flow3661
  br label %Flow3662, !amdgpu.uniform !42

sw.bb4.i.i.i.i.i.i.i.i.i:                         ; preds = %NodeBlock3356
  br label %Flow3661, !amdgpu.uniform !42

Flow3657:                                         ; preds = %sw.bb6.i.i.i.i.i.i.i.i.i, %NodeBlock3360
  %1306 = call { i1, i64 } @llvm.amdgcn.else.i64.i64(i64 %1279)
  %1307 = extractvalue { i1, i64 } %1306, 0
  %1308 = extractvalue { i1, i64 } %1306, 1
  br i1 %1307, label %sw.bb5.i.i.i.i.i.i.i.i.i, label %Flow3658

sw.bb5.i.i.i.i.i.i.i.i.i:                         ; preds = %Flow3657
  br label %Flow3658, !amdgpu.uniform !42

sw.bb6.i.i.i.i.i.i.i.i.i:                         ; preds = %NodeBlock3360
  br label %Flow3657, !amdgpu.uniform !42

Flow3655:                                         ; preds = %Flow3654, %NodeBlock3364
  %1309 = phi i8 [ %1316, %Flow3654 ], [ poison, %NodeBlock3364 ]
  %1310 = call { i1, i64 } @llvm.amdgcn.else.i64.i64(i64 %1269)
  %1311 = extractvalue { i1, i64 } %1310, 0
  %1312 = extractvalue { i1, i64 } %1310, 1
  br i1 %1311, label %sw.bb7.i.i.i.i.i.i.i.i.i, label %Flow3656

sw.bb7.i.i.i.i.i.i.i.i.i:                         ; preds = %Flow3655
  br label %Flow3656, !amdgpu.uniform !42

Flow3653:                                         ; preds = %sw.bb9.i.i.i.i.i.i.i.i.i, %NodeBlock3362
  %1313 = call { i1, i64 } @llvm.amdgcn.else.i64.i64(i64 %1272)
  %1314 = extractvalue { i1, i64 } %1313, 0
  %1315 = extractvalue { i1, i64 } %1313, 1
  br i1 %1314, label %sw.bb8.i.i.i.i.i.i.i.i.i, label %Flow3654

sw.bb8.i.i.i.i.i.i.i.i.i:                         ; preds = %Flow3653
  br label %Flow3654, !amdgpu.uniform !42

sw.bb9.i.i.i.i.i.i.i.i.i:                         ; preds = %NodeBlock3362
  br label %Flow3653, !amdgpu.uniform !42

Flow3654:                                         ; preds = %sw.bb8.i.i.i.i.i.i.i.i.i, %Flow3653
  %1316 = phi i8 [ 56, %sw.bb8.i.i.i.i.i.i.i.i.i ], [ 57, %Flow3653 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %1315)
  br label %Flow3655, !amdgpu.uniform !42

Flow3656:                                         ; preds = %sw.bb7.i.i.i.i.i.i.i.i.i, %Flow3655
  %1317 = phi i8 [ 55, %sw.bb7.i.i.i.i.i.i.i.i.i ], [ %1309, %Flow3655 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %1312)
  br label %Flow3659, !amdgpu.uniform !42

Flow3658:                                         ; preds = %sw.bb5.i.i.i.i.i.i.i.i.i, %Flow3657
  %1318 = phi i8 [ 53, %sw.bb5.i.i.i.i.i.i.i.i.i ], [ 54, %Flow3657 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %1308)
  br label %Flow3660, !amdgpu.uniform !42

Flow3660:                                         ; preds = %Flow3658, %Flow3659
  %1319 = phi i8 [ %1318, %Flow3658 ], [ %1273, %Flow3659 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %1276)
  br label %Flow3668, !amdgpu.uniform !42

Flow3662:                                         ; preds = %sw.bb3.i.i.i.i.i.i.i.i.i, %Flow3661
  %1320 = phi i8 [ 51, %sw.bb3.i.i.i.i.i.i.i.i.i ], [ 52, %Flow3661 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %1305)
  br label %Flow3666, !amdgpu.uniform !42

Flow3663:                                         ; preds = %sw.bb1.i.i.i.i.i.i.i.i.i, %LeafBlock3352
  %1321 = phi i8 [ 49, %sw.bb1.i.i.i.i.i.i.i.i.i ], [ 48, %LeafBlock3352 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %1302)
  br label %Flow3665, !amdgpu.uniform !42

Flow3665:                                         ; preds = %Flow3663, %Flow3664
  %1322 = phi i8 [ %1321, %Flow3663 ], [ 50, %Flow3664 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %1299)
  br label %Flow3667, !amdgpu.uniform !42

Flow3667:                                         ; preds = %Flow3665, %Flow3666
  %1323 = phi i8 [ %1322, %Flow3665 ], [ %1290, %Flow3666 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %1293)
  br label %if.then.i.i.i.i.i.i.i.i, !amdgpu.uniform !42

if.then.i.i.i.i.i.i.i.i:                          ; preds = %Flow3668, %Flow3667
  %retval.0.i.i.i.i.i.i.i.i.i = phi i8 [ %1280, %Flow3668 ], [ %1323, %Flow3667 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %1283)
  %cmp.i.i.i.i.i.i.i.i.i = icmp eq i64 %writer.sroa.10.0.i.i.i.i.i, 11
  %cmp.i.i.i.i.i.i.i.i.i.inv = xor i1 %cmp.i.i.i.i.i.i.i.i.i, true
  br i1 %cmp.i.i.i.i.i.i.i.i.i.inv, label %if.then3.i.i.i.i.i.i.i.i, label %_ZN22__llvm_libc_22_0_0_git7details22StringBufferWriterImplILb0EE4pushEc.exit.i.i.i.i.i.i.i, !amdgpu.uniform !42

if.then3.i.i.i.i.i.i.i.i:                         ; preds = %if.then.i.i.i.i.i.i.i.i
  %1324 = trunc i64 %writer.sroa.10.0.i.i.i.i.i to i32
  %1325 = xor i32 %1324, -1
  %1326 = getelementptr i8, ptr addrspace(5) %buffer.i.i.i, i32 %1325
  %arrayidx.i.i.i.i.i.i.i.i.i2718 = getelementptr i8, ptr addrspace(5) %1326, i32 11
  %inc.i.i.i.i.i.i.i.i = add i64 %writer.sroa.10.0.i.i.i.i.i, 1
  store i8 %retval.0.i.i.i.i.i.i.i.i.i, ptr addrspace(5) %arrayidx.i.i.i.i.i.i.i.i.i2718, align 1, !tbaa !57, !noalias !94
  br label %_ZN22__llvm_libc_22_0_0_git7details22StringBufferWriterImplILb0EE4pushEc.exit.i.i.i.i.i.i.i, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git7details22StringBufferWriterImplILb0EE4pushEc.exit.i.i.i.i.i.i.i: ; preds = %if.then3.i.i.i.i.i.i.i.i, %if.then.i.i.i.i.i.i.i.i
  %writer.sroa.10.1.i.i.i.i.i = phi i64 [ %inc.i.i.i.i.i.i.i.i, %if.then3.i.i.i.i.i.i.i.i ], [ 11, %if.then.i.i.i.i.i.i.i.i ]
  %cmp.i.i.i.i.i.i.i = icmp samesign ult i32 %value.addr.09.i.i.i.i.i.i.i, 10
  %.not.i.i.i.i.i.i.i = or i1 %cmp.i.i.i.i.i.i.i.i.i, %cmp.i.i.i.i.i.i.i
  %1327 = call i64 @llvm.amdgcn.if.break.i64(i1 %.not.i.i.i.i.i.i.i, i64 %phi.broken3920)
  %1328 = call i1 @llvm.amdgcn.loop.i64(i64 %1327)
  br i1 %1328, label %Flow3669, label %while.body.i.i.i.i.i.i.i

Flow3669:                                         ; preds = %_ZN22__llvm_libc_22_0_0_git7details22StringBufferWriterImplILb0EE4pushEc.exit.i.i.i.i.i.i.i
  %writer.sroa.10.1.i.i.i.i.i.lcssa = phi i64 [ %writer.sroa.10.1.i.i.i.i.i, %_ZN22__llvm_libc_22_0_0_git7details22StringBufferWriterImplILb0EE4pushEc.exit.i.i.i.i.i.i.i ]
  %.lcssa4092 = phi i64 [ %1327, %_ZN22__llvm_libc_22_0_0_git7details22StringBufferWriterImplILb0EE4pushEc.exit.i.i.i.i.i.i.i ]
  %cmp.i.i.i.i.i.i.i.i.i.lcssa = phi i1 [ %cmp.i.i.i.i.i.i.i.i.i, %_ZN22__llvm_libc_22_0_0_git7details22StringBufferWriterImplILb0EE4pushEc.exit.i.i.i.i.i.i.i ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa4092)
  br label %Flow3670, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git15IntegerToStringIiNS_7details3FmtILh10ELb0ELb0ELb0ELm1EEEE13IntegerWriter25write_unsigned_number_decEjRNS1_22StringBufferWriterImplILb0EEE.exit.i.i.i.i.i.i: ; preds = %Flow3670
  %cmp.i19.i.i.i.i.i = icmp ne i64 %1259, 0
  %or.cond.not.i.i.i.i.i = select i1 %1258, i1 true, i1 %cmp.i19.i.i.i.i.i
  %or.cond.not.i.i.i.i.i.inv = xor i1 %or.cond.not.i.i.i.i.i, true
  %1329 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %or.cond.not.i.i.i.i.i.inv)
  %1330 = extractvalue { i1, i64 } %1329, 0
  %1331 = extractvalue { i1, i64 } %1329, 1
  br i1 %1330, label %if.then3.i.i.lr.ph.i.i.i.i.i, label %while.end.i.i.i.i.i.i

if.then3.i.i.lr.ph.i.i.i.i.i:                     ; preds = %_ZN22__llvm_libc_22_0_0_git15IntegerToStringIiNS_7details3FmtILh10ELb0ELb0ELb0ELm1EEEE13IntegerWriter25write_unsigned_number_decEjRNS1_22StringBufferWriterImplILb0EEE.exit.i.i.i.i.i.i
  %arrayidx.i.i.i.i.i.i.i.i = getelementptr inbounds nuw i8, ptr addrspace(5) %buffer.i.i.i, i32 10
  store i8 48, ptr addrspace(5) %arrayidx.i.i.i.i.i.i.i.i, align 2, !tbaa !57, !noalias !94
  br label %while.end.i.i.i.i.i.i, !amdgpu.uniform !42

while.end.i.i.i.i.i.i:                            ; preds = %if.then3.i.i.lr.ph.i.i.i.i.i, %_ZN22__llvm_libc_22_0_0_git15IntegerToStringIiNS_7details3FmtILh10ELb0ELb0ELb0ELm1EEEE13IntegerWriter25write_unsigned_number_decEjRNS1_22StringBufferWriterImplILb0EEE.exit.i.i.i.i.i.i
  %writer.sroa.10.4.i.i.i.i.i = phi i64 [ 1, %if.then3.i.i.lr.ph.i.i.i.i.i ], [ %1259, %_ZN22__llvm_libc_22_0_0_git15IntegerToStringIiNS_7details3FmtILh10ELb0ELb0ELb0ELm1EEEE13IntegerWriter25write_unsigned_number_decEjRNS1_22StringBufferWriterImplILb0EEE.exit.i.i.i.i.i.i ]
  call void @llvm.amdgcn.end.cf.i64(i64 %1331)
  %cmp4.i.i.i.i.i.i = icmp sgt i8 %LHS, -1
  %brmerge.i.i.i.i.i.i = or i1 %cmp4.i.i.i.i.i.i, %1258
  %brmerge.i.i.i.i.i.i.inv = xor i1 %brmerge.i.i.i.i.i.i, true
  %1332 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %brmerge.i.i.i.i.i.i.inv)
  %1333 = extractvalue { i1, i64 } %1332, 0
  %1334 = extractvalue { i1, i64 } %1332, 1
  br i1 %1333, label %if.then.i14.i.i.i.i.i.i, label %_ZN22__llvm_libc_22_0_0_git15IntegerToStringIiNS_7details3FmtILh10ELb0ELb0ELb0ELm1EEEEC2Ei.exit.i.i.i.i

if.then.i14.i.i.i.i.i.i:                          ; preds = %while.end.i.i.i.i.i.i
  %cmp.i.i17.i.i.i.i.i.i = icmp ne i64 %writer.sroa.10.4.i.i.i.i.i, 11
  %1335 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.i.i17.i.i.i.i.i.i)
  %1336 = extractvalue { i1, i64 } %1335, 0
  %1337 = extractvalue { i1, i64 } %1335, 1
  br i1 %1336, label %if.then3.i18.i.i.i.i.i.i, label %Flow3652

if.then3.i18.i.i.i.i.i.i:                         ; preds = %if.then.i14.i.i.i.i.i.i
  %1338 = trunc i64 %writer.sroa.10.4.i.i.i.i.i to i32
  %1339 = xor i32 %1338, -1
  %1340 = getelementptr i8, ptr addrspace(5) %buffer.i.i.i, i32 %1339
  %arrayidx.i.i19.i.i.i.i.i.i2719 = getelementptr i8, ptr addrspace(5) %1340, i32 11
  %inc.i20.i.i.i.i.i.i = add i64 %writer.sroa.10.4.i.i.i.i.i, 1
  store i8 45, ptr addrspace(5) %arrayidx.i.i19.i.i.i.i.i.i2719, align 1, !tbaa !57, !noalias !94
  br label %Flow3652, !amdgpu.uniform !42

Flow3652:                                         ; preds = %if.then3.i18.i.i.i.i.i.i, %if.then.i14.i.i.i.i.i.i
  %1341 = phi i64 [ %inc.i20.i.i.i.i.i.i, %if.then3.i18.i.i.i.i.i.i ], [ 11, %if.then.i14.i.i.i.i.i.i ]
  call void @llvm.amdgcn.end.cf.i64(i64 %1337)
  br label %_ZN22__llvm_libc_22_0_0_git15IntegerToStringIiNS_7details3FmtILh10ELb0ELb0ELb0ELm1EEEEC2Ei.exit.i.i.i.i, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git15IntegerToStringIiNS_7details3FmtILh10ELb0ELb0ELb0ELm1EEEEC2Ei.exit.i.i.i.i: ; preds = %while.end.i.i.i.i.i.i, %Flow3652
  %writer.sroa.10.5.i.i.i.i.i = phi i64 [ %writer.sroa.10.4.i.i.i.i.i, %while.end.i.i.i.i.i.i ], [ %1341, %Flow3652 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %1334)
  %written.i.i.i.i.i = getelementptr inbounds nuw i8, ptr addrspace(5) %buffer.i.i.i, i32 16
  store i64 %writer.sroa.10.5.i.i.i.i.i, ptr addrspace(5) %written.i.i.i.i.i, align 8, !tbaa !91, !noalias !94
  %cmp.i.not.i.i.i.i.i.i = icmp ne i64 %writer.sroa.10.5.i.i.i.i.i, 0
  %1342 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.i.not.i.i.i.i.i.i)
  %1343 = extractvalue { i1, i64 } %1342, 0
  %1344 = extractvalue { i1, i64 } %1342, 1
  br i1 %1343, label %if.then.i.i.i.i.i.i.i, label %Flow3651

if.then.i.i.i.i.i.i.i:                            ; preds = %_ZN22__llvm_libc_22_0_0_git15IntegerToStringIiNS_7details3FmtILh10ELb0ELb0ELb0ELm1EEEEC2Ei.exit.i.i.i.i
  %1345 = call { i64, i1 } @llvm.uadd.with.overflow.i64(i64 %writer.sroa.10.5.i.i.i.i.i, i64 1)
  %ov3302 = extractvalue { i64, i1 } %1345, 1
  %ov3302.inv = xor i1 %ov3302, true
  %1346 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %ov3302.inv)
  %1347 = extractvalue { i1, i64 } %1346, 0
  %1348 = extractvalue { i1, i64 } %1346, 1
  br i1 %1347, label %if.end.i.i.i.i.i.i.i.i, label %Flow3650

if.end.i.i.i.i.i.i.i.i:                           ; preds = %if.then.i.i.i.i.i.i.i
  %math3301 = extractvalue { i64, i1 } %1345, 0
  %cmp2.i.i.i.i.i.i.i.i = icmp ult i64 %math3301, 1676976733973595601
  %1349 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp2.i.i.i.i.i.i.i.i)
  %1350 = extractvalue { i1, i64 } %1349, 0
  %1351 = extractvalue { i1, i64 } %1349, 1
  br i1 %1350, label %if.then3.i.i.i.i4.i.i.i.i, label %if.end4.i.i.i.i.i.i.i.i

if.then3.i.i.i.i4.i.i.i.i:                        ; preds = %if.end.i.i.i.i.i.i.i.i
  %mul.i.i.i.i.i.i.i.i = mul nuw i64 %math3301, 11
  %div18.i.i.i.i.i.i.i.i = lshr i64 %mul.i.i.i.i.i.i.i.i, 3
  br label %if.end4.i.i.i.i.i.i.i.i, !amdgpu.uniform !42

Flow3650:                                         ; preds = %if.end4.i.i.i.i.i.i.i.i, %if.then.i.i.i.i.i.i.i
  %1352 = phi ptr [ %cond.i.i951, %if.end4.i.i.i.i.i.i.i.i ], [ addrspacecast (ptr addrspace(4) @_ZN22__llvm_libc_22_0_0_git3cpp6string14NULL_CHARACTERE to ptr), %if.then.i.i.i.i.i.i.i ]
  call void @llvm.amdgcn.end.cf.i64(i64 %1348)
  br label %_ZN22__llvm_libc_22_0_0_git3cpp6string7reserveEm.exit.i.i.i.i.i.i.i, !amdgpu.uniform !42

if.end4.i.i.i.i.i.i.i.i:                          ; preds = %if.then3.i.i.i.i4.i.i.i.i, %if.end.i.i.i.i.i.i.i.i
  %new_capacity.addr.0.i.i.i.i.i.i.i.i = phi i64 [ %div18.i.i.i.i.i.i.i.i, %if.then3.i.i.i.i4.i.i.i.i ], [ %math3301, %if.end.i.i.i.i.i.i.i.i ]
  call void @llvm.amdgcn.end.cf.i64(i64 %1351)
  %sub.i.i945 = add i64 %new_capacity.addr.0.i.i.i.i.i.i.i.i, 7
  %div2.i.i946 = and i64 %sub.i.i945, -8
  %1353 = load ptr, ptr addrspace(1) @_ZN12_GLOBAL__N_13ptrE, align 8, !tbaa !55
  %add.ptr.i.i947 = getelementptr inbounds nuw i8, ptr %1353, i64 %div2.i.i946
  store ptr %add.ptr.i.i947, ptr addrspace(1) @_ZN12_GLOBAL__N_13ptrE, align 8, !tbaa !55
  %sub.ptr.lhs.cast.i.i948 = ptrtoint ptr %add.ptr.i.i947 to i64
  %sub.ptr.sub.i.i949 = sub i64 %sub.ptr.lhs.cast.i.i948, ptrtoint (ptr addrspacecast (ptr addrspace(1) @_ZN12_GLOBAL__N_16memoryE to ptr) to i64)
  %cmp.i.i950 = icmp ugt i64 %sub.ptr.sub.i.i949, 65335
  %cond.i.i951 = select i1 %cmp.i.i950, ptr null, ptr %1353
  br label %Flow3650, !amdgpu.uniform !42

Flow3651:                                         ; preds = %Flow3649, %_ZN22__llvm_libc_22_0_0_git15IntegerToStringIiNS_7details3FmtILh10ELb0ELb0ELb0ELm1EEEEC2Ei.exit.i.i.i.i
  %1354 = phi ptr [ %1352, %Flow3649 ], [ addrspacecast (ptr addrspace(4) @_ZN22__llvm_libc_22_0_0_git3cpp6string14NULL_CHARACTERE to ptr), %_ZN22__llvm_libc_22_0_0_git15IntegerToStringIiNS_7details3FmtILh10ELb0ELb0ELb0ELm1EEEEC2Ei.exit.i.i.i.i ]
  call void @llvm.amdgcn.end.cf.i64(i64 %1344)
  br label %if.end.i.i.i.i.i.i.i, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git3cpp6string7reserveEm.exit.i.i.i.i.i.i.i: ; preds = %Flow3650
  br label %loadstoreloop2188, !amdgpu.uniform !42

loadstoreloop2188:                                ; preds = %_ZN22__llvm_libc_22_0_0_git3cpp6string7reserveEm.exit.i.i.i.i.i.i.i, %loadstoreloop2188
  %phi.broken3921 = phi i64 [ 0, %_ZN22__llvm_libc_22_0_0_git3cpp6string7reserveEm.exit.i.i.i.i.i.i.i ], [ %1359, %loadstoreloop2188 ]
  %1355 = phi i64 [ %1357, %loadstoreloop2188 ], [ 0, %_ZN22__llvm_libc_22_0_0_git3cpp6string7reserveEm.exit.i.i.i.i.i.i.i ]
  %1356 = getelementptr i8, ptr %1352, i64 %1355
  store i8 0, ptr %1356, align 1
  %1357 = add i64 %1355, 1
  %1358 = icmp uge i64 %1357, %writer.sroa.10.5.i.i.i.i.i
  %1359 = call i64 @llvm.amdgcn.if.break.i64(i1 %1358, i64 %phi.broken3921)
  %1360 = call i1 @llvm.amdgcn.loop.i64(i64 %1359)
  br i1 %1360, label %Flow3649, label %loadstoreloop2188

Flow3649:                                         ; preds = %loadstoreloop2188
  %.lcssa4091 = phi i64 [ %1359, %loadstoreloop2188 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa4091)
  br label %Flow3651, !amdgpu.uniform !42

if.end.i.i.i.i.i.i.i:                             ; preds = %Flow3651
  %cmp.not.i6.i.i.i.i.i.i.i = icmp ne ptr %1354, addrspacecast (ptr addrspace(4) @_ZN22__llvm_libc_22_0_0_git3cpp6string14NULL_CHARACTERE to ptr)
  %1361 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.not.i6.i.i.i.i.i.i.i)
  %1362 = extractvalue { i1, i64 } %1361, 0
  %1363 = extractvalue { i1, i64 } %1361, 1
  br i1 %1362, label %if.then.i.i.i.i2.i.i.i.i, label %_ZN22__llvm_libc_22_0_0_git7testing8internal13describeValueIcEENS_3cpp9enable_ifIXaasr3cppE13is_integral_vIT_ElestS5_Lm8EENS3_6stringEE4typeES5_.exit.i

if.then.i.i.i.i2.i.i.i.i:                         ; preds = %if.end.i.i.i.i.i.i.i
  %arrayidx.i.i.i.i3.i.i.i.i = getelementptr inbounds nuw i8, ptr %1354, i64 %writer.sroa.10.5.i.i.i.i.i
  store i8 0, ptr %arrayidx.i.i.i.i3.i.i.i.i, align 1, !tbaa !57, !noalias !94
  br label %_ZN22__llvm_libc_22_0_0_git7testing8internal13describeValueIcEENS_3cpp9enable_ifIXaasr3cppE13is_integral_vIT_ElestS5_Lm8EENS3_6stringEE4typeES5_.exit.i, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git7testing8internal13describeValueIcEENS_3cpp9enable_ifIXaasr3cppE13is_integral_vIT_ElestS5_Lm8EENS3_6stringEE4typeES5_.exit.i: ; preds = %if.then.i.i.i.i2.i.i.i.i, %if.end.i.i.i.i.i.i.i
  call void @llvm.amdgcn.end.cf.i64(i64 %1363)
  %1364 = trunc i64 %writer.sroa.10.5.i.i.i.i.i to i32
  %1365 = and i64 %writer.sroa.10.5.i.i.i.i.i, 15
  %1366 = sub i64 %writer.sroa.10.5.i.i.i.i.i, %1365
  %1367 = icmp ne i64 %1366, 0
  %1368 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %1367)
  %1369 = extractvalue { i1, i64 } %1368, 0
  %1370 = extractvalue { i1, i64 } %1368, 1
  br i1 %1369, label %loop-memcpy-expansion2302.preheader, label %Flow3647

loop-memcpy-expansion2302.preheader:              ; preds = %_ZN22__llvm_libc_22_0_0_git7testing8internal13describeValueIcEENS_3cpp9enable_ifIXaasr3cppE13is_integral_vIT_ElestS5_Lm8EENS3_6stringEE4typeES5_.exit.i
  %1371 = sub i32 11, %1364
  %scevgep3040 = getelementptr i8, ptr addrspace(5) %buffer.i.i.i, i32 %1371
  br label %loop-memcpy-expansion2302, !amdgpu.uniform !42

Flow3647:                                         ; preds = %Flow3646, %_ZN22__llvm_libc_22_0_0_git7testing8internal13describeValueIcEENS_3cpp9enable_ifIXaasr3cppE13is_integral_vIT_ElestS5_Lm8EENS3_6stringEE4typeES5_.exit.i
  call void @llvm.amdgcn.end.cf.i64(i64 %1370)
  br label %loop-memcpy-residual-header2305, !amdgpu.uniform !42

loop-memcpy-expansion2302:                        ; preds = %loop-memcpy-expansion2302.preheader, %loop-memcpy-expansion2302
  %phi.broken3922 = phi i64 [ 0, %loop-memcpy-expansion2302.preheader ], [ %1376, %loop-memcpy-expansion2302 ]
  %lsr.iv3041 = phi ptr addrspace(5) [ %scevgep3042, %loop-memcpy-expansion2302 ], [ %scevgep3040, %loop-memcpy-expansion2302.preheader ]
  %loop-index2303 = phi i64 [ %1374, %loop-memcpy-expansion2302 ], [ 0, %loop-memcpy-expansion2302.preheader ]
  %1372 = load <4 x i32>, ptr addrspace(5) %lsr.iv3041, align 1
  %1373 = getelementptr i8, ptr %1354, i64 %loop-index2303
  store <4 x i32> %1372, ptr %1373, align 1
  %1374 = add i64 %loop-index2303, 16
  %scevgep3042 = getelementptr i8, ptr addrspace(5) %lsr.iv3041, i32 16
  %1375 = icmp uge i64 %1374, %1366
  %1376 = call i64 @llvm.amdgcn.if.break.i64(i1 %1375, i64 %phi.broken3922)
  %1377 = call i1 @llvm.amdgcn.loop.i64(i64 %1376)
  br i1 %1377, label %Flow3646, label %loop-memcpy-expansion2302

Flow3644:                                         ; preds = %Flow3643, %loop-memcpy-residual-header2305
  call void @llvm.amdgcn.end.cf.i64(i64 %2831)
  br label %post-loop-memcpy-expansion2301, !amdgpu.uniform !42

loop-memcpy-residual2304:                         ; preds = %loop-memcpy-residual2304.preheader, %loop-memcpy-residual2304
  %phi.broken3923 = phi i64 [ 0, %loop-memcpy-residual2304.preheader ], [ %1381, %loop-memcpy-residual2304 ]
  %lsr.iv3036 = phi ptr addrspace(5) [ %scevgep3037, %loop-memcpy-residual2304 ], [ %scevgep3035, %loop-memcpy-residual2304.preheader ]
  %residual-loop-index2306 = phi i64 [ %1379, %loop-memcpy-residual2304 ], [ 0, %loop-memcpy-residual2304.preheader ]
  %1378 = load i8, ptr addrspace(5) %lsr.iv3036, align 1
  %scevgep3039 = getelementptr i8, ptr %scevgep3038, i64 %residual-loop-index2306
  store i8 %1378, ptr %scevgep3039, align 1
  %1379 = add i64 %residual-loop-index2306, 1
  %scevgep3037 = getelementptr i8, ptr addrspace(5) %lsr.iv3036, i32 1
  %1380 = icmp uge i64 %1379, %1365
  %1381 = call i64 @llvm.amdgcn.if.break.i64(i1 %1380, i64 %phi.broken3923)
  %1382 = call i1 @llvm.amdgcn.loop.i64(i64 %1381)
  br i1 %1382, label %Flow3643, label %loop-memcpy-residual2304

Flow3643:                                         ; preds = %loop-memcpy-residual2304
  %.lcssa4089 = phi i64 [ %1381, %loop-memcpy-residual2304 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa4089)
  br label %Flow3644, !amdgpu.uniform !42

post-loop-memcpy-expansion2301:                   ; preds = %Flow3644
  call void @llvm.lifetime.end.p5(ptr addrspace(5) %buffer.i.i.i) #34, !noalias !94
  br label %for.cond.i.i.i953, !amdgpu.uniform !42

for.cond.i.i.i953:                                ; preds = %post-loop-memcpy-expansion2301, %Flow3641
  %1383 = phi i32 [ %1421, %Flow3641 ], [ poison, %post-loop-memcpy-expansion2301 ]
  %1384 = phi i64 [ %1422, %Flow3641 ], [ poison, %post-loop-memcpy-expansion2301 ]
  %1385 = phi i32 [ %1423, %Flow3641 ], [ poison, %post-loop-memcpy-expansion2301 ]
  %retval.sroa.6.0.i.i.i954 = phi i32 [ %1426, %Flow3641 ], [ undef, %post-loop-memcpy-expansion2301 ]
  %retval.sroa.2.0.i.i.i956 = phi i64 [ %1425, %Flow3641 ], [ undef, %post-loop-memcpy-expansion2301 ]
  %retval.sroa.8.0.i.i.i957 = phi i32 [ %1424, %Flow3641 ], [ undef, %post-loop-memcpy-expansion2301 ]
  %index.0.i.i.i958 = phi i32 [ %1428, %Flow3641 ], [ 0, %post-loop-memcpy-expansion2301 ]
  %1386 = load i32, ptr addrspace(1) @__llvm_rpc_client, align 8, !tbaa !68
  %cmp.not.i.i.i959 = icmp ult i32 %index.0.i.i.i958, %1386
  %spec.store.select.i.i.i960 = select i1 %cmp.not.i.i.i959, i32 %index.0.i.i.i958, i32 0
  %1387 = tail call noundef i64 @llvm.amdgcn.ballot.i64(i1 true)
  %1388 = and i32 %spec.store.select.i.i.i960, 31
  %1389 = lshr i64 %1387, %sh_prom.i.i.i.i
  %1390 = trunc i64 %1389 to i32
  %conv4.i.i.i.i.i964 = and i32 %1390, 1
  %shl.i.i.i.i.i965 = shl nuw i32 %conv4.i.i.i.i.i964, %1388
  %1391 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2552, !amdgpu.uniform !42

1392:                                             ; preds = %ComputeEnd2553
  %sunkaddr3303 = getelementptr inbounds i8, ptr addrspace(1) %3471, i64 40
  %1393 = atomicrmw or ptr addrspace(1) %sunkaddr3303, i32 %.lcssa4083 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %1394, !amdgpu.uniform !42

1394:                                             ; preds = %1392, %ComputeEnd2553
  %1395 = phi i32 [ %1393, %1392 ], [ poison, %ComputeEnd2553 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %3480)
  %1396 = call i32 @llvm.amdgcn.readfirstlane.i32(i32 %1395)
  %1397 = or i32 %1396, %.lcssa4084
  %shl5.i.i.i.i.i966 = shl nuw i32 1, %1388
  %and.i.i.i.i.i967 = and i32 %shl5.i.i.i.i.i966, %1397
  %tobool3.i.i.i.i968 = icmp ne i32 %and.i.i.i.i.i967, 0
  %1398 = tail call i64 @llvm.amdgcn.ballot.i64(i1 %tobool3.i.i.i.i968)
  %and.i.i.i.i.i.i969 = and i64 %1398, %1387
  %cmp.i.not.i.i.i970 = icmp ne i64 %1387, %and.i.i.i.i.i.i969
  br i1 %cmp.i.not.i.i.i970, label %if.then.i.i.i.i971, label %Flow3640, !amdgpu.uniform !42

if.then.i.i.i.i971:                               ; preds = %1394
  fence syncscope("agent") acquire
  %1399 = load <2 x i64>, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !75
  %1400 = extractelement <2 x i64> %1399, i32 0
  %1401 = inttoptr i64 %1400 to ptr
  %1402 = extractelement <2 x i64> %1399, i32 1
  %1403 = inttoptr i64 %1402 to ptr
  %idxprom.i.i.i.i972 = zext i32 %spec.store.select.i.i.i960 to i64
  %arrayidx.i.i.i.i973 = getelementptr inbounds nuw i32, ptr %1401, i64 %idxprom.i.i.i.i972, !amdgpu.uniform !42
  %1404 = load atomic i32, ptr %arrayidx.i.i.i.i973 monotonic, align 4
  %1405 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %1404)
  %arrayidx.i45.i.i.i974 = getelementptr inbounds nuw i32, ptr %1403, i64 %idxprom.i.i.i.i972, !amdgpu.uniform !42
  %1406 = load atomic i32, ptr %arrayidx.i45.i.i.i974 monotonic, align 4
  %1407 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %1406)
  %cmp.i46.not.i.i.i975 = icmp ne i32 %1405, %1407
  br i1 %cmp.i46.not.i.i.i975, label %if.then12.i.i.i976, label %Flow3638, !amdgpu.uniform !42

if.then12.i.i.i976:                               ; preds = %if.then.i.i.i.i971
  fence syncscope("agent") release
  %1408 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %1387, i1 true)
  %iszero.i.i.i.i.i.i.i977 = icmp ne i64 %1387, 0
  %cmp2.i.i.i.i.i.i978 = icmp eq i64 %1408, %sh_prom.i.i.i.i
  %cmp.i.i.i.i.i.i979 = select i1 %iszero.i.i.i.i.i.i.i977, i1 %cmp2.i.i.i.i.i.i978, i1 false
  %conv4.i.i51.i.i.i980 = zext i1 %cmp.i.i.i.i.i.i979 to i32
  %shl.i.i52.i.i.i981 = shl nuw i32 %conv4.i.i51.i.i.i980, %1388
  %xor.i.i.i.i.i982 = xor i32 %shl.i.i52.i.i.i981, -1
  %1409 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2557, !amdgpu.uniform !42

1410:                                             ; preds = %ComputeEnd2558
  %sunkaddr3304 = getelementptr inbounds i8, ptr addrspace(1) %3471, i64 40
  %1411 = atomicrmw and ptr addrspace(1) %sunkaddr3304, i32 %.lcssa4085 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %1412, !amdgpu.uniform !42

1412:                                             ; preds = %1410, %ComputeEnd2558
  call void @llvm.amdgcn.end.cf.i64(i64 %3498)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %Flow3638, !amdgpu.uniform !42

if.end14.i.i.i1042:                               ; preds = %Flow3638
  %1413 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %1387, i1 true)
  %iszero.i.i.i.i.i.i1043 = icmp ne i64 %1387, 0
  %cmp2.i.i.i.i.i1044 = icmp eq i64 %1413, %sh_prom.i.i.i.i
  %cmp.i.i.i.i.i1045 = select i1 %iszero.i.i.i.i.i.i1043, i1 %cmp2.i.i.i.i.i1044, i1 false
  %1414 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.i.i.i.i.i1045)
  %1415 = extractvalue { i1, i64 } %1414, 0
  %1416 = extractvalue { i1, i64 } %1414, 1
  br i1 %1415, label %if.then16.i.i.i1047, label %if.end22.i.i.i1046

if.then16.i.i.i1047:                              ; preds = %if.end14.i.i.i1042
  %1417 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !76
  %opcode.i.i.i1048 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %1417, i64 %idxprom.i.i.i.i972, i32 1
  store i32 1660944387, ptr %opcode.i.i.i1048, align 8, !tbaa !77
  %arrayidx21.i.i.i1049 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %1417, i64 %idxprom.i.i.i.i972
  store i64 %1387, ptr %arrayidx21.i.i.i1049, align 8, !tbaa !80
  br label %if.end22.i.i.i1046, !amdgpu.uniform !42

Flow3639:                                         ; preds = %if.end22.i.i.i1046, %Flow3638
  %1418 = phi i32 [ %1405, %if.end22.i.i.i1046 ], [ %retval.sroa.8.0.i.i.i957, %Flow3638 ]
  %1419 = phi i64 [ %1387, %if.end22.i.i.i1046 ], [ %retval.sroa.2.0.i.i.i956, %Flow3638 ]
  %1420 = phi i32 [ %spec.store.select.i.i.i960, %if.end22.i.i.i1046 ], [ %retval.sroa.6.0.i.i.i954, %Flow3638 ]
  br label %cleanup26.i.i.i983, !amdgpu.uniform !42

if.end22.i.i.i1046:                               ; preds = %if.then16.i.i.i1047, %if.end14.i.i.i1042
  call void @llvm.amdgcn.end.cf.i64(i64 %1416)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %Flow3639, !amdgpu.uniform !42

Flow3640:                                         ; preds = %cleanup26.i.i.i983, %1394
  %1421 = phi i32 [ %1420, %cleanup26.i.i.i983 ], [ %1383, %1394 ]
  %1422 = phi i64 [ %1419, %cleanup26.i.i.i983 ], [ %1384, %1394 ]
  %1423 = phi i32 [ %1418, %cleanup26.i.i.i983 ], [ %1385, %1394 ]
  %1424 = phi i32 [ %1418, %cleanup26.i.i.i983 ], [ %retval.sroa.8.0.i.i.i957, %1394 ]
  %1425 = phi i64 [ %1419, %cleanup26.i.i.i983 ], [ %retval.sroa.2.0.i.i.i956, %1394 ]
  %1426 = phi i32 [ %1420, %cleanup26.i.i.i983 ], [ %retval.sroa.6.0.i.i.i954, %1394 ]
  %1427 = phi i1 [ %cmp.i46.not.i.i.i975, %cleanup26.i.i.i983 ], [ true, %1394 ]
  br i1 %1427, label %for.inc.i.i.i988, label %Flow3641, !amdgpu.uniform !42

cleanup26.i.i.i983:                               ; preds = %Flow3639
  br label %Flow3640, !amdgpu.uniform !42

for.inc.i.i.i988:                                 ; preds = %Flow3640
  %inc.i.i.i993 = add i32 %spec.store.select.i.i.i960, 1
  br label %Flow3641, !amdgpu.uniform !42

Flow3641:                                         ; preds = %for.inc.i.i.i988, %Flow3640
  %1428 = phi i32 [ %inc.i.i.i993, %for.inc.i.i.i988 ], [ poison, %Flow3640 ]
  %1429 = phi i1 [ false, %for.inc.i.i.i988 ], [ true, %Flow3640 ]
  br i1 %1429, label %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i994, label %for.cond.i.i.i953, !amdgpu.uniform !42

_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i994: ; preds = %Flow3641
  %.lcssa4088 = phi i32 [ %1421, %Flow3641 ]
  %.lcssa4087 = phi i64 [ %1422, %Flow3641 ]
  %.lcssa4086 = phi i32 [ %1423, %Flow3641 ]
  fence acquire
  %1430 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %mul.i.i.i.i.i.i995 = shl i32 %.lcssa4088, 6
  %idxprom.i19.i.i.i.i.i996 = zext i32 %mul.i.i.i.i.i.i995 to i64
  %arrayidx.i20.i.i.i.i.i997 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %1430, i64 %idxprom.i19.i.i.i.i.i996
  %arrayidx.i22.i.i.i.i.i998 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %arrayidx.i20.i.i.i.i.i997, i64 %sh_prom.i.i.i.i
  store i64 %writer.sroa.10.5.i.i.i.i.i, ptr %arrayidx.i22.i.i.i.i.i998, align 8, !tbaa !82
  %spec.select.i.i.i.i.i.i.i999 = tail call i64 @llvm.umin.i64(i64 %writer.sroa.10.5.i.i.i.i.i, i64 56)
  %1431 = and i64 %spec.select.i.i.i.i.i.i.i999, 15
  %1432 = sub i64 %spec.select.i.i.i.i.i.i.i999, %1431
  %1433 = icmp ne i64 %1432, 0
  %1434 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %1433)
  %1435 = extractvalue { i1, i64 } %1434, 0
  %1436 = extractvalue { i1, i64 } %1434, 1
  br i1 %1435, label %loop-memcpy-expansion2230.preheader, label %Flow3637

loop-memcpy-expansion2230.preheader:              ; preds = %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i994
  %1437 = add nuw nsw i64 %75, 8
  %scevgep3032 = getelementptr i8, ptr %1430, i64 %1437
  %1438 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i996, 6
  %scevgep3033 = getelementptr i8, ptr %scevgep3032, i64 %1438
  br label %loop-memcpy-expansion2230, !amdgpu.uniform !42

Flow3637:                                         ; preds = %Flow3636, %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i994
  call void @llvm.amdgcn.end.cf.i64(i64 %1436)
  br label %loop-memcpy-residual-header2233, !amdgpu.uniform !42

loop-memcpy-expansion2230:                        ; preds = %loop-memcpy-expansion2230.preheader, %loop-memcpy-expansion2230
  %phi.broken3924 = phi i64 [ 0, %loop-memcpy-expansion2230.preheader ], [ %1443, %loop-memcpy-expansion2230 ]
  %loop-index2231 = phi i64 [ %1441, %loop-memcpy-expansion2230 ], [ 0, %loop-memcpy-expansion2230.preheader ]
  %1439 = getelementptr i8, ptr %1354, i64 %loop-index2231
  %1440 = load <4 x i32>, ptr %1439, align 1
  %scevgep3034 = getelementptr i8, ptr %scevgep3033, i64 %loop-index2231
  store <4 x i32> %1440, ptr %scevgep3034, align 1
  %1441 = add i64 %loop-index2231, 16
  %1442 = icmp uge i64 %1441, %1432
  %1443 = call i64 @llvm.amdgcn.if.break.i64(i1 %1442, i64 %phi.broken3924)
  %1444 = call i1 @llvm.amdgcn.loop.i64(i64 %1443)
  br i1 %1444, label %Flow3636, label %loop-memcpy-expansion2230

Flow3634:                                         ; preds = %Flow3631, %Flow3633
  %1445 = phi i64 [ %2768, %Flow3631 ], [ %2760, %Flow3633 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %2763)
  br label %post-loop-memcpy-expansion2229, !amdgpu.uniform !42

loop-memcpy-residual2232:                         ; preds = %loop-memcpy-residual2232.preheader, %loop-memcpy-residual2232
  %phi.broken3925 = phi i64 [ 0, %loop-memcpy-residual2232.preheader ], [ %1449, %loop-memcpy-residual2232 ]
  %residual-loop-index2234 = phi i64 [ %1447, %loop-memcpy-residual2232 ], [ 0, %loop-memcpy-residual2232.preheader ]
  %scevgep3031 = getelementptr i8, ptr %scevgep3030, i64 %residual-loop-index2234
  %1446 = load i8, ptr %scevgep3031, align 1
  %scevgep3029 = getelementptr i8, ptr %scevgep3028, i64 %residual-loop-index2234
  store i8 %1446, ptr %scevgep3029, align 1
  %1447 = add i64 %residual-loop-index2234, 1
  %1448 = icmp uge i64 %1447, %1431
  %1449 = call i64 @llvm.amdgcn.if.break.i64(i1 %1448, i64 %phi.broken3925)
  %1450 = call i1 @llvm.amdgcn.loop.i64(i64 %1449)
  br i1 %1450, label %Flow3631, label %loop-memcpy-residual2232

Flow3631:                                         ; preds = %loop-memcpy-residual2232
  %.lcssa4081 = phi i64 [ %1449, %loop-memcpy-residual2232 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa4081)
  br label %Flow3634, !amdgpu.uniform !42

post-loop-memcpy-expansion2229:                   ; preds = %Flow3634
  %tobool.not.i.i.i.i.i.i1001 = icmp eq i32 %.lcssa4086, 0
  %conv.i.i.i.i.i.i1002 = zext i1 %tobool.not.i.i.i.i.i.i1001 to i32
  fence release
  %1451 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %idxprom.i23.i.i.i.i.i1003 = zext i32 %.lcssa4088 to i64
  %arrayidx.i24.i.i.i.i.i1004 = getelementptr inbounds nuw i32, ptr %1451, i64 %idxprom.i23.i.i.i.i.i1003
  store atomic i32 %conv.i.i.i.i.i.i1002, ptr %arrayidx.i24.i.i.i.i.i1004 monotonic, align 4
  %1452 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !84
  %arrayidx.i.i5.i.i1005 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %1452, i64 %idxprom.i23.i.i.i.i.i1003, !amdgpu.uniform !42
  %1453 = load i64, ptr %arrayidx.i.i5.i.i1005, align 8, !tbaa !80
  %scevgep3014 = getelementptr i8, ptr %1354, i64 56
  %1454 = add i64 %writer.sroa.10.5.i.i.i.i.i, -56
  br label %while.cond.i.i.i.i1006, !amdgpu.uniform !42

while.cond.i.i.i.i1006:                           ; preds = %post-loop-memcpy-expansion2229, %Flow3629
  %phi.broken3928 = phi i64 [ 0, %post-loop-memcpy-expansion2229 ], [ %1471, %Flow3629 ]
  %lsr.iv3020 = phi i64 [ %1469, %Flow3629 ], [ %1454, %post-loop-memcpy-expansion2229 ]
  %lsr.iv3015 = phi ptr [ %1468, %Flow3629 ], [ %scevgep3014, %post-loop-memcpy-expansion2229 ]
  %port.sroa.37.0.i.i1007 = phi i32 [ %1467, %Flow3629 ], [ %conv.i.i.i.i.i.i1002, %post-loop-memcpy-expansion2229 ]
  %idx.0.i.i.i.i1008 = phi i64 [ %1466, %Flow3629 ], [ 56, %post-loop-memcpy-expansion2229 ]
  %cmp.i.i.i.i1009 = icmp ugt i64 %writer.sroa.10.5.i.i.i.i.i, %idx.0.i.i.i.i1008
  %1455 = tail call i64 @llvm.amdgcn.ballot.i64(i1 %cmp.i.i.i.i1009)
  %and.i.i.i.i6.i.i1010 = and i64 %1455, %1453
  %tobool.not.i.i.i.i1011 = icmp ne i64 %and.i.i.i.i6.i.i1010, 0
  %1456 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i22.i.i.i1012 = getelementptr inbounds nuw i32, ptr %1456, i64 %idxprom.i23.i.i.i.i.i1003, !amdgpu.uniform !42
  %1457 = load atomic i32, ptr %arrayidx.i22.i.i.i1012 monotonic, align 4
  %1458 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %tobool.not.i.i.i.i1011)
  %1459 = extractvalue { i1, i64 } %1458, 0
  %1460 = extractvalue { i1, i64 } %1458, 1
  br i1 %1459, label %cond.false.i8.i.i.i.i1013, label %Flow3629

cond.false.i8.i.i.i.i1013:                        ; preds = %while.cond.i.i.i.i1006
  %1461 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %1457)
  %1462 = icmp ne i32 %1461, %port.sroa.37.0.i.i1007
  br i1 %1462, label %while.body.i.i21.i.i.i.i1014.preheader, label %Flow3628, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i1014.preheader:           ; preds = %cond.false.i8.i.i.i.i1013
  br label %while.body.i.i21.i.i.i.i1014, !amdgpu.uniform !42

Flow3628:                                         ; preds = %Flow3626, %cond.false.i8.i.i.i.i1013
  br label %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i1017, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i1014:                     ; preds = %while.body.i.i21.i.i.i.i1014.preheader, %while.body.i.i21.i.i.i.i1014
  tail call void @llvm.amdgcn.s.sleep(i32 2)
  %1463 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i22.i.i.i.i1015 = getelementptr inbounds nuw i32, ptr %1463, i64 %idxprom.i23.i.i.i.i.i1003, !amdgpu.uniform !42
  %1464 = load atomic i32, ptr %arrayidx.i.i.i22.i.i.i.i1015 monotonic, align 4
  %1465 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %1464)
  %cmp.i.not.i.i23.i.i.i.i1016 = icmp eq i32 %1465, %port.sroa.37.0.i.i1007
  br i1 %cmp.i.not.i.i23.i.i.i.i1016, label %Flow3626, label %while.body.i.i21.i.i.i.i1014, !amdgpu.uniform !42

Flow3626:                                         ; preds = %while.body.i.i21.i.i.i.i1014
  br label %Flow3628, !amdgpu.uniform !42

Flow3629:                                         ; preds = %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1018, %while.cond.i.i.i.i1006
  %1466 = phi i64 [ %add.i.i.i.i1021, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1018 ], [ poison, %while.cond.i.i.i.i1006 ]
  %1467 = phi i32 [ %conv.i.i26.i.i.i.i1019, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1018 ], [ poison, %while.cond.i.i.i.i1006 ]
  %1468 = phi ptr [ %scevgep3016, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1018 ], [ poison, %while.cond.i.i.i.i1006 ]
  %1469 = phi i64 [ %lsr.iv.next3021, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1018 ], [ poison, %while.cond.i.i.i.i1006 ]
  %1470 = phi i1 [ false, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1018 ], [ true, %while.cond.i.i.i.i1006 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %1460)
  %1471 = call i64 @llvm.amdgcn.if.break.i64(i1 %1470, i64 %phi.broken3928)
  %1472 = call i1 @llvm.amdgcn.loop.i64(i64 %1471)
  br i1 %1472, label %cond.false.i.i.i1028, label %while.cond.i.i.i.i1006

_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i1017: ; preds = %Flow3628
  fence acquire
  %1473 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.i.i.i.i1009)
  %1474 = extractvalue { i1, i64 } %1473, 0
  %1475 = extractvalue { i1, i64 } %1473, 1
  br i1 %1474, label %if.then.i.i.i.i.i.i.i1022, label %Flow3625

if.then.i.i.i.i.i.i.i1022:                        ; preds = %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i1017
  %sub.i.i.i.i.i.i.i1024 = sub nuw i64 %writer.sroa.10.5.i.i.i.i.i, %idx.0.i.i.i.i1008
  %spec.select.i.i.i36.i.i.i.i1025 = tail call i64 @llvm.umin.i64(i64 %sub.i.i.i.i.i.i.i1024, i64 64)
  %1476 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %1477 = and i64 %spec.select.i.i.i36.i.i.i.i1025, 15
  %1478 = sub i64 %spec.select.i.i.i36.i.i.i.i1025, %1477
  %1479 = icmp ne i64 %1478, 0
  %1480 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %1479)
  %1481 = extractvalue { i1, i64 } %1480, 0
  %1482 = extractvalue { i1, i64 } %1480, 1
  br i1 %1481, label %loop-memcpy-expansion2224.preheader, label %Flow3624

loop-memcpy-expansion2224.preheader:              ; preds = %if.then.i.i.i.i.i.i.i1022
  %arrayidx.i20.i35.i.i.i.i1026 = getelementptr %"struct.rpc::Buffer", ptr %1476, i64 %idxprom.i19.i.i.i.i.i996
  %arrayidx.i22.i38.i.i.i.i1027 = getelementptr %"struct.rpc::Buffer", ptr %arrayidx.i20.i35.i.i.i.i1026, i64 %sh_prom.i.i.i.i
  br label %loop-memcpy-expansion2224, !amdgpu.uniform !42

Flow3624:                                         ; preds = %Flow3623, %if.then.i.i.i.i.i.i.i1022
  call void @llvm.amdgcn.end.cf.i64(i64 %1482)
  br label %loop-memcpy-residual-header2227, !amdgpu.uniform !42

loop-memcpy-expansion2224:                        ; preds = %loop-memcpy-expansion2224.preheader, %loop-memcpy-expansion2224
  %phi.broken3926 = phi i64 [ 0, %loop-memcpy-expansion2224.preheader ], [ %1487, %loop-memcpy-expansion2224 ]
  %loop-index2225 = phi i64 [ %1485, %loop-memcpy-expansion2224 ], [ 0, %loop-memcpy-expansion2224.preheader ]
  %scevgep3017 = getelementptr i8, ptr %lsr.iv3015, i64 %loop-index2225
  %1483 = load <4 x i32>, ptr %scevgep3017, align 1
  %1484 = getelementptr i8, ptr %arrayidx.i22.i38.i.i.i.i1027, i64 %loop-index2225
  store <4 x i32> %1483, ptr %1484, align 1
  %1485 = add i64 %loop-index2225, 16
  %1486 = icmp uge i64 %1485, %1478
  %1487 = call i64 @llvm.amdgcn.if.break.i64(i1 %1486, i64 %phi.broken3926)
  %1488 = call i1 @llvm.amdgcn.loop.i64(i64 %1487)
  br i1 %1488, label %Flow3623, label %loop-memcpy-expansion2224

Flow3621:                                         ; preds = %Flow3620, %loop-memcpy-residual-header2227
  call void @llvm.amdgcn.end.cf.i64(i64 %2753)
  br label %Flow3625, !amdgpu.uniform !42

loop-memcpy-residual2226:                         ; preds = %loop-memcpy-residual2226.preheader, %loop-memcpy-residual2226
  %phi.broken3927 = phi i64 [ 0, %loop-memcpy-residual2226.preheader ], [ %1492, %loop-memcpy-residual2226 ]
  %residual-loop-index2228 = phi i64 [ %1490, %loop-memcpy-residual2226 ], [ 0, %loop-memcpy-residual2226.preheader ]
  %scevgep3026 = getelementptr i8, ptr %scevgep3025, i64 %residual-loop-index2228
  %1489 = load i8, ptr %scevgep3026, align 1
  %scevgep3024 = getelementptr i8, ptr %scevgep3023, i64 %residual-loop-index2228
  store i8 %1489, ptr %scevgep3024, align 1
  %1490 = add i64 %residual-loop-index2228, 1
  %1491 = icmp uge i64 %1490, %1477
  %1492 = call i64 @llvm.amdgcn.if.break.i64(i1 %1491, i64 %phi.broken3927)
  %1493 = call i1 @llvm.amdgcn.loop.i64(i64 %1492)
  br i1 %1493, label %Flow3620, label %loop-memcpy-residual2226

Flow3620:                                         ; preds = %loop-memcpy-residual2226
  %.lcssa4078 = phi i64 [ %1492, %loop-memcpy-residual2226 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa4078)
  br label %Flow3621, !amdgpu.uniform !42

_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1018: ; preds = %Flow3625
  %conv.i.i26.i.i.i.i1019 = xor i32 %port.sroa.37.0.i.i1007, 1
  fence release
  %1494 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %arrayidx.i24.i29.i.i.i.i1020 = getelementptr inbounds nuw i32, ptr %1494, i64 %idxprom.i23.i.i.i.i.i1003
  store atomic i32 %conv.i.i26.i.i.i.i1019, ptr %arrayidx.i24.i29.i.i.i.i1020 monotonic, align 4
  %add.i.i.i.i1021 = add i64 %idx.0.i.i.i.i1008, 64
  %scevgep3016 = getelementptr i8, ptr %lsr.iv3015, i64 64
  %lsr.iv.next3021 = add i64 %lsr.iv3020, -64
  br label %Flow3629, !amdgpu.uniform !42

cond.false.i.i.i1028:                             ; preds = %Flow3629
  %.lcssa4080 = phi i64 [ %1471, %Flow3629 ]
  %port.sroa.37.0.i.i1007.lcssa = phi i32 [ %port.sroa.37.0.i.i1007, %Flow3629 ]
  %.lcssa4079 = phi i32 [ %1457, %Flow3629 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa4080)
  %1495 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %.lcssa4079)
  %1496 = icmp ne i32 %1495, %port.sroa.37.0.i.i1007.lcssa
  %1497 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %1496)
  %1498 = extractvalue { i1, i64 } %1497, 0
  %1499 = extractvalue { i1, i64 } %1497, 1
  br i1 %1498, label %while.body.i.i.i.i1029.preheader, label %Flow3618

while.body.i.i.i.i1029.preheader:                 ; preds = %cond.false.i.i.i1028
  br label %while.body.i.i.i.i1029, !amdgpu.uniform !42

Flow3618:                                         ; preds = %Flow3616, %cond.false.i.i.i1028
  call void @llvm.amdgcn.end.cf.i64(i64 %1499)
  br label %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsINS_3cpp6stringEEERS1_T_.exit1050, !amdgpu.uniform !42

while.body.i.i.i.i1029:                           ; preds = %while.body.i.i.i.i1029.preheader, %while.body.i.i.i.i1029
  %phi.broken3929 = phi i64 [ 0, %while.body.i.i.i.i1029.preheader ], [ %1503, %while.body.i.i.i.i1029 ]
  tail call void @llvm.amdgcn.s.sleep(i32 2)
  %1500 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i8.i.i1030 = getelementptr inbounds nuw i32, ptr %1500, i64 %idxprom.i23.i.i.i.i.i1003, !amdgpu.uniform !42
  %1501 = load atomic i32, ptr %arrayidx.i.i.i8.i.i1030 monotonic, align 4
  %1502 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %1501)
  %cmp.i.not.i.i.i.i1031 = icmp eq i32 %1502, %port.sroa.37.0.i.i1007.lcssa
  %1503 = call i64 @llvm.amdgcn.if.break.i64(i1 %cmp.i.not.i.i.i.i1031, i64 %phi.broken3929)
  %1504 = call i1 @llvm.amdgcn.loop.i64(i64 %1503)
  br i1 %1504, label %Flow3616, label %while.body.i.i.i.i1029

Flow3616:                                         ; preds = %while.body.i.i.i.i1029
  %.lcssa4076 = phi i64 [ %1503, %while.body.i.i.i.i1029 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa4076)
  br label %Flow3618, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsINS_3cpp6stringEEERS1_T_.exit1050: ; preds = %Flow3618
  fence acquire
  tail call void @llvm.amdgcn.wave.barrier()
  fence syncscope("agent") release
  %1505 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %.lcssa4087, i1 true)
  %iszero.i.i.i.i.i13.i.i1033 = icmp ne i64 %.lcssa4087, 0
  %cmp2.i.i.i.i14.i.i1034 = icmp eq i64 %1505, %sh_prom.i.i.i.i
  %cmp.i.i.i.i15.i.i1035 = select i1 %iszero.i.i.i.i.i13.i.i1033, i1 %cmp2.i.i.i.i14.i.i1034, i1 false
  %1506 = and i32 %.lcssa4088, 31
  %conv4.i.i.i19.i.i1039 = zext i1 %cmp.i.i.i.i15.i.i1035 to i32
  %shl.i.i.i20.i.i1040 = shl nuw i32 %conv4.i.i.i19.i.i1039, %1506
  %xor.i.i.i21.i.i1041 = xor i32 %shl.i.i.i20.i.i1040, -1
  %1507 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2561, !amdgpu.uniform !42

1508:                                             ; preds = %ComputeEnd2562
  %div8.i.i.i16.i.i1036 = lshr i32 %.lcssa4088, 5
  %idxprom.i.i.i17.i.i1037 = zext nneg i32 %div8.i.i.i16.i.i1036 to i64
  %1509 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i17.i.i1037
  %sunkaddr3305 = getelementptr inbounds i8, ptr addrspace(1) %1509, i64 40
  %1510 = atomicrmw and ptr addrspace(1) %sunkaddr3305, i32 %.lcssa4075 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %1511, !amdgpu.uniform !42

1511:                                             ; preds = %1508, %ComputeEnd2562
  call void @llvm.amdgcn.end.cf.i64(i64 %3515)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %for.cond.i.i.i1054, !amdgpu.uniform !42

for.cond.i.i.i1054:                               ; preds = %1511, %Flow3615
  %1512 = phi i32 [ %1550, %Flow3615 ], [ poison, %1511 ]
  %1513 = phi i64 [ %1551, %Flow3615 ], [ poison, %1511 ]
  %1514 = phi i32 [ %1552, %Flow3615 ], [ poison, %1511 ]
  %retval.sroa.6.0.i.i.i1055 = phi i32 [ %1555, %Flow3615 ], [ undef, %1511 ]
  %retval.sroa.2.0.i.i.i1057 = phi i64 [ %1554, %Flow3615 ], [ undef, %1511 ]
  %retval.sroa.8.0.i.i.i1058 = phi i32 [ %1553, %Flow3615 ], [ undef, %1511 ]
  %index.0.i.i.i1059 = phi i32 [ %1557, %Flow3615 ], [ 0, %1511 ]
  %1515 = load i32, ptr addrspace(1) @__llvm_rpc_client, align 8, !tbaa !68
  %cmp.not.i.i.i1060 = icmp ult i32 %index.0.i.i.i1059, %1515
  %spec.store.select.i.i.i1061 = select i1 %cmp.not.i.i.i1060, i32 %index.0.i.i.i1059, i32 0
  %1516 = tail call noundef i64 @llvm.amdgcn.ballot.i64(i1 true)
  %1517 = and i32 %spec.store.select.i.i.i1061, 31
  %1518 = lshr i64 %1516, %sh_prom.i.i.i.i
  %1519 = trunc i64 %1518 to i32
  %conv4.i.i.i.i.i1065 = and i32 %1519, 1
  %shl.i.i.i.i.i1066 = shl nuw i32 %conv4.i.i.i.i.i1065, %1517
  %1520 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2565, !amdgpu.uniform !42

1521:                                             ; preds = %ComputeEnd2566
  %sunkaddr3306 = getelementptr inbounds i8, ptr addrspace(1) %3525, i64 40
  %1522 = atomicrmw or ptr addrspace(1) %sunkaddr3306, i32 %.lcssa4069 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %1523, !amdgpu.uniform !42

1523:                                             ; preds = %1521, %ComputeEnd2566
  %1524 = phi i32 [ %1522, %1521 ], [ poison, %ComputeEnd2566 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %3534)
  %1525 = call i32 @llvm.amdgcn.readfirstlane.i32(i32 %1524)
  %1526 = or i32 %1525, %.lcssa4070
  %shl5.i.i.i.i.i1067 = shl nuw i32 1, %1517
  %and.i.i.i.i.i1068 = and i32 %shl5.i.i.i.i.i1067, %1526
  %tobool3.i.i.i.i1069 = icmp ne i32 %and.i.i.i.i.i1068, 0
  %1527 = tail call i64 @llvm.amdgcn.ballot.i64(i1 %tobool3.i.i.i.i1069)
  %and.i.i.i.i.i.i1070 = and i64 %1527, %1516
  %cmp.i.not.i.i.i1071 = icmp ne i64 %1516, %and.i.i.i.i.i.i1070
  br i1 %cmp.i.not.i.i.i1071, label %if.then.i.i.i.i1072, label %Flow3614, !amdgpu.uniform !42

if.then.i.i.i.i1072:                              ; preds = %1523
  fence syncscope("agent") acquire
  %1528 = load <2 x i64>, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !75
  %1529 = extractelement <2 x i64> %1528, i32 0
  %1530 = inttoptr i64 %1529 to ptr
  %1531 = extractelement <2 x i64> %1528, i32 1
  %1532 = inttoptr i64 %1531 to ptr
  %idxprom.i.i.i.i1073 = zext i32 %spec.store.select.i.i.i1061 to i64
  %arrayidx.i.i.i.i1074 = getelementptr inbounds nuw i32, ptr %1530, i64 %idxprom.i.i.i.i1073, !amdgpu.uniform !42
  %1533 = load atomic i32, ptr %arrayidx.i.i.i.i1074 monotonic, align 4
  %1534 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %1533)
  %arrayidx.i45.i.i.i1075 = getelementptr inbounds nuw i32, ptr %1532, i64 %idxprom.i.i.i.i1073, !amdgpu.uniform !42
  %1535 = load atomic i32, ptr %arrayidx.i45.i.i.i1075 monotonic, align 4
  %1536 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %1535)
  %cmp.i46.not.i.i.i1076 = icmp ne i32 %1534, %1536
  br i1 %cmp.i46.not.i.i.i1076, label %if.then12.i.i.i1077, label %Flow3612, !amdgpu.uniform !42

if.then12.i.i.i1077:                              ; preds = %if.then.i.i.i.i1072
  fence syncscope("agent") release
  %1537 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %1516, i1 true)
  %iszero.i.i.i.i.i.i.i1078 = icmp ne i64 %1516, 0
  %cmp2.i.i.i.i.i.i1079 = icmp eq i64 %1537, %sh_prom.i.i.i.i
  %cmp.i.i.i.i.i.i1080 = select i1 %iszero.i.i.i.i.i.i.i1078, i1 %cmp2.i.i.i.i.i.i1079, i1 false
  %conv4.i.i51.i.i.i1081 = zext i1 %cmp.i.i.i.i.i.i1080 to i32
  %shl.i.i52.i.i.i1082 = shl nuw i32 %conv4.i.i51.i.i.i1081, %1517
  %xor.i.i.i.i.i1083 = xor i32 %shl.i.i52.i.i.i1082, -1
  %1538 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2570, !amdgpu.uniform !42

1539:                                             ; preds = %ComputeEnd2571
  %sunkaddr3307 = getelementptr inbounds i8, ptr addrspace(1) %3525, i64 40
  %1540 = atomicrmw and ptr addrspace(1) %sunkaddr3307, i32 %.lcssa4071 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %1541, !amdgpu.uniform !42

1541:                                             ; preds = %1539, %ComputeEnd2571
  call void @llvm.amdgcn.end.cf.i64(i64 %3552)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %Flow3612, !amdgpu.uniform !42

if.end14.i.i.i1141:                               ; preds = %Flow3612
  %1542 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %1516, i1 true)
  %iszero.i.i.i.i.i.i1142 = icmp ne i64 %1516, 0
  %cmp2.i.i.i.i.i1143 = icmp eq i64 %1542, %sh_prom.i.i.i.i
  %cmp.i.i.i.i.i1144 = select i1 %iszero.i.i.i.i.i.i1142, i1 %cmp2.i.i.i.i.i1143, i1 false
  %1543 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.i.i.i.i.i1144)
  %1544 = extractvalue { i1, i64 } %1543, 0
  %1545 = extractvalue { i1, i64 } %1543, 1
  br i1 %1544, label %if.then16.i.i.i1146, label %if.end22.i.i.i1145

if.then16.i.i.i1146:                              ; preds = %if.end14.i.i.i1141
  %1546 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !76
  %opcode.i.i.i1147 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %1546, i64 %idxprom.i.i.i.i1073, i32 1
  store i32 1660944387, ptr %opcode.i.i.i1147, align 8, !tbaa !77
  %arrayidx21.i.i.i1148 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %1546, i64 %idxprom.i.i.i.i1073
  store i64 %1516, ptr %arrayidx21.i.i.i1148, align 8, !tbaa !80
  br label %if.end22.i.i.i1145, !amdgpu.uniform !42

Flow3613:                                         ; preds = %if.end22.i.i.i1145, %Flow3612
  %1547 = phi i32 [ %1534, %if.end22.i.i.i1145 ], [ %retval.sroa.8.0.i.i.i1058, %Flow3612 ]
  %1548 = phi i64 [ %1516, %if.end22.i.i.i1145 ], [ %retval.sroa.2.0.i.i.i1057, %Flow3612 ]
  %1549 = phi i32 [ %spec.store.select.i.i.i1061, %if.end22.i.i.i1145 ], [ %retval.sroa.6.0.i.i.i1055, %Flow3612 ]
  br label %cleanup26.i.i.i1084, !amdgpu.uniform !42

if.end22.i.i.i1145:                               ; preds = %if.then16.i.i.i1146, %if.end14.i.i.i1141
  call void @llvm.amdgcn.end.cf.i64(i64 %1545)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %Flow3613, !amdgpu.uniform !42

Flow3614:                                         ; preds = %cleanup26.i.i.i1084, %1523
  %1550 = phi i32 [ %1549, %cleanup26.i.i.i1084 ], [ %1512, %1523 ]
  %1551 = phi i64 [ %1548, %cleanup26.i.i.i1084 ], [ %1513, %1523 ]
  %1552 = phi i32 [ %1547, %cleanup26.i.i.i1084 ], [ %1514, %1523 ]
  %1553 = phi i32 [ %1547, %cleanup26.i.i.i1084 ], [ %retval.sroa.8.0.i.i.i1058, %1523 ]
  %1554 = phi i64 [ %1548, %cleanup26.i.i.i1084 ], [ %retval.sroa.2.0.i.i.i1057, %1523 ]
  %1555 = phi i32 [ %1549, %cleanup26.i.i.i1084 ], [ %retval.sroa.6.0.i.i.i1055, %1523 ]
  %1556 = phi i1 [ %cmp.i46.not.i.i.i1076, %cleanup26.i.i.i1084 ], [ true, %1523 ]
  br i1 %1556, label %for.inc.i.i.i1089, label %Flow3615, !amdgpu.uniform !42

cleanup26.i.i.i1084:                              ; preds = %Flow3613
  br label %Flow3614, !amdgpu.uniform !42

for.inc.i.i.i1089:                                ; preds = %Flow3614
  %inc.i.i.i1094 = add i32 %spec.store.select.i.i.i1061, 1
  br label %Flow3615, !amdgpu.uniform !42

Flow3615:                                         ; preds = %for.inc.i.i.i1089, %Flow3614
  %1557 = phi i32 [ %inc.i.i.i1094, %for.inc.i.i.i1089 ], [ poison, %Flow3614 ]
  %1558 = phi i1 [ false, %for.inc.i.i.i1089 ], [ true, %Flow3614 ]
  br i1 %1558, label %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1095, label %for.cond.i.i.i1054, !amdgpu.uniform !42

_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1095: ; preds = %Flow3615
  %.lcssa4074 = phi i32 [ %1550, %Flow3615 ]
  %.lcssa4073 = phi i64 [ %1551, %Flow3615 ]
  %.lcssa4072 = phi i32 [ %1552, %Flow3615 ]
  fence acquire
  %1559 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %mul.i.i.i.i.i.i1096 = shl i32 %.lcssa4074, 6
  %idxprom.i19.i.i.i.i.i1097 = zext i32 %mul.i.i.i.i.i.i1096 to i64
  %arrayidx.i20.i.i.i.i.i1098 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %1559, i64 %idxprom.i19.i.i.i.i.i1097
  %arrayidx.i22.i.i.i.i.i1099 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %arrayidx.i20.i.i.i.i.i1098, i64 %sh_prom.i.i.i.i
  store i64 1, ptr %arrayidx.i22.i.i.i.i.i1099, align 8, !tbaa !82
  %arrayidx6.i.i.i.i.i.i.i1100 = getelementptr inbounds nuw i8, ptr %arrayidx.i22.i.i.i.i.i1099, i64 8
  store i8 10, ptr %arrayidx6.i.i.i.i.i.i.i1100, align 8
  %tobool.not.i.i.i.i.i.i1101 = icmp eq i32 %.lcssa4072, 0
  %conv.i.i.i.i.i.i1102 = zext i1 %tobool.not.i.i.i.i.i.i1101 to i32
  fence release
  %1560 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %idxprom.i23.i.i.i.i.i1103 = zext i32 %.lcssa4074 to i64
  %arrayidx.i24.i.i.i.i.i1104 = getelementptr inbounds nuw i32, ptr %1560, i64 %idxprom.i23.i.i.i.i.i1103
  store atomic i32 %conv.i.i.i.i.i.i1102, ptr %arrayidx.i24.i.i.i.i.i1104 monotonic, align 4
  %1561 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !84
  %arrayidx.i.i5.i.i1105 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %1561, i64 %idxprom.i23.i.i.i.i.i1103, !amdgpu.uniform !42
  %1562 = load i64, ptr %arrayidx.i.i5.i.i1105, align 8, !tbaa !80
  br label %while.cond.i.i.i.i1106, !amdgpu.uniform !42

while.cond.i.i.i.i1106:                           ; preds = %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1095, %Flow3611
  %phi.broken3930 = phi i64 [ 0, %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1095 ], [ %1577, %Flow3611 ]
  %port.sroa.37.0.i.i1107 = phi i32 [ %1575, %Flow3611 ], [ %conv.i.i.i.i.i.i1102, %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1095 ]
  %idx.0.i.i.i.i1108 = phi i64 [ %1574, %Flow3611 ], [ 56, %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1095 ]
  %cmp.i.i.i.i1109 = icmp eq i64 %idx.0.i.i.i.i1108, 0
  %1563 = tail call i64 @llvm.amdgcn.ballot.i64(i1 %cmp.i.i.i.i1109)
  %and.i.i.i.i6.i.i1110 = and i64 %1563, %1562
  %tobool.not.i.i.i.i1111 = icmp ne i64 %and.i.i.i.i6.i.i1110, 0
  %1564 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i22.i.i.i1112 = getelementptr inbounds nuw i32, ptr %1564, i64 %idxprom.i23.i.i.i.i.i1103, !amdgpu.uniform !42
  %1565 = load atomic i32, ptr %arrayidx.i22.i.i.i1112 monotonic, align 4
  %1566 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %tobool.not.i.i.i.i1111)
  %1567 = extractvalue { i1, i64 } %1566, 0
  %1568 = extractvalue { i1, i64 } %1566, 1
  br i1 %1567, label %cond.false.i8.i.i.i.i1113, label %Flow3611

cond.false.i8.i.i.i.i1113:                        ; preds = %while.cond.i.i.i.i1106
  %1569 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %1565)
  %1570 = icmp ne i32 %1569, %port.sroa.37.0.i.i1107
  br i1 %1570, label %while.body.i.i21.i.i.i.i1114.preheader, label %Flow3610, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i1114.preheader:           ; preds = %cond.false.i8.i.i.i.i1113
  br label %while.body.i.i21.i.i.i.i1114, !amdgpu.uniform !42

Flow3610:                                         ; preds = %Flow3608, %cond.false.i8.i.i.i.i1113
  br label %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i1117, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i1114:                     ; preds = %while.body.i.i21.i.i.i.i1114.preheader, %while.body.i.i21.i.i.i.i1114
  tail call void @llvm.amdgcn.s.sleep(i32 2)
  %1571 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i22.i.i.i.i1115 = getelementptr inbounds nuw i32, ptr %1571, i64 %idxprom.i23.i.i.i.i.i1103, !amdgpu.uniform !42
  %1572 = load atomic i32, ptr %arrayidx.i.i.i22.i.i.i.i1115 monotonic, align 4
  %1573 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %1572)
  %cmp.i.not.i.i23.i.i.i.i1116 = icmp eq i32 %1573, %port.sroa.37.0.i.i1107
  br i1 %cmp.i.not.i.i23.i.i.i.i1116, label %Flow3608, label %while.body.i.i21.i.i.i.i1114, !amdgpu.uniform !42

Flow3608:                                         ; preds = %while.body.i.i21.i.i.i.i1114
  br label %Flow3610, !amdgpu.uniform !42

Flow3611:                                         ; preds = %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1118, %while.cond.i.i.i.i1106
  %1574 = phi i64 [ %add.i.i.i.i1121, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1118 ], [ poison, %while.cond.i.i.i.i1106 ]
  %1575 = phi i32 [ %conv.i.i26.i.i.i.i1119, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1118 ], [ poison, %while.cond.i.i.i.i1106 ]
  %1576 = phi i1 [ false, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1118 ], [ true, %while.cond.i.i.i.i1106 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %1568)
  %1577 = call i64 @llvm.amdgcn.if.break.i64(i1 %1576, i64 %phi.broken3930)
  %1578 = call i1 @llvm.amdgcn.loop.i64(i64 %1577)
  br i1 %1578, label %cond.false.i.i.i1127, label %while.cond.i.i.i.i1106

_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i1117: ; preds = %Flow3610
  fence acquire
  br i1 %cmp.i.i.i.i1109, label %if.then.i.i.i.i.i.i.i1122, label %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1118, !amdgpu.uniform !42

if.then.i.i.i.i.i.i.i1122:                        ; preds = %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i1117
  %1579 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %arrayidx.i20.i35.i.i.i.i1125 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %1579, i64 %idxprom.i19.i.i.i.i.i1097
  %arrayidx.i22.i38.i.i.i.i1126 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %arrayidx.i20.i35.i.i.i.i1125, i64 %sh_prom.i.i.i.i
  store i8 10, ptr %arrayidx.i22.i38.i.i.i.i1126, align 1
  br label %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1118, !amdgpu.uniform !42

_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1118: ; preds = %if.then.i.i.i.i.i.i.i1122, %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i1117
  %conv.i.i26.i.i.i.i1119 = xor i32 %port.sroa.37.0.i.i1107, 1
  fence release
  %1580 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %arrayidx.i24.i29.i.i.i.i1120 = getelementptr inbounds nuw i32, ptr %1580, i64 %idxprom.i23.i.i.i.i.i1103
  store atomic i32 %conv.i.i26.i.i.i.i1119, ptr %arrayidx.i24.i29.i.i.i.i1120 monotonic, align 4
  %add.i.i.i.i1121 = add i64 %idx.0.i.i.i.i1108, 64
  br label %Flow3611, !amdgpu.uniform !42

cond.false.i.i.i1127:                             ; preds = %Flow3611
  %.lcssa4068 = phi i64 [ %1577, %Flow3611 ]
  %port.sroa.37.0.i.i1107.lcssa = phi i32 [ %port.sroa.37.0.i.i1107, %Flow3611 ]
  %.lcssa4067 = phi i32 [ %1565, %Flow3611 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa4068)
  %1581 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %.lcssa4067)
  %1582 = icmp ne i32 %1581, %port.sroa.37.0.i.i1107.lcssa
  %1583 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %1582)
  %1584 = extractvalue { i1, i64 } %1583, 0
  %1585 = extractvalue { i1, i64 } %1583, 1
  br i1 %1584, label %while.body.i.i.i.i1128.preheader, label %Flow3607

while.body.i.i.i.i1128.preheader:                 ; preds = %cond.false.i.i.i1127
  br label %while.body.i.i.i.i1128, !amdgpu.uniform !42

Flow3607:                                         ; preds = %Flow3605, %cond.false.i.i.i1127
  call void @llvm.amdgcn.end.cf.i64(i64 %1585)
  br label %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIcEERS1_T_.exit1149, !amdgpu.uniform !42

while.body.i.i.i.i1128:                           ; preds = %while.body.i.i.i.i1128.preheader, %while.body.i.i.i.i1128
  %phi.broken3931 = phi i64 [ 0, %while.body.i.i.i.i1128.preheader ], [ %1589, %while.body.i.i.i.i1128 ]
  tail call void @llvm.amdgcn.s.sleep(i32 2)
  %1586 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i8.i.i1129 = getelementptr inbounds nuw i32, ptr %1586, i64 %idxprom.i23.i.i.i.i.i1103, !amdgpu.uniform !42
  %1587 = load atomic i32, ptr %arrayidx.i.i.i8.i.i1129 monotonic, align 4
  %1588 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %1587)
  %cmp.i.not.i.i.i.i1130 = icmp eq i32 %1588, %port.sroa.37.0.i.i1107.lcssa
  %1589 = call i64 @llvm.amdgcn.if.break.i64(i1 %cmp.i.not.i.i.i.i1130, i64 %phi.broken3931)
  %1590 = call i1 @llvm.amdgcn.loop.i64(i64 %1589)
  br i1 %1590, label %Flow3605, label %while.body.i.i.i.i1128

Flow3605:                                         ; preds = %while.body.i.i.i.i1128
  %.lcssa4066 = phi i64 [ %1589, %while.body.i.i.i.i1128 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa4066)
  br label %Flow3607, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIcEERS1_T_.exit1149: ; preds = %Flow3607
  fence acquire
  tail call void @llvm.amdgcn.wave.barrier()
  fence syncscope("agent") release
  %1591 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %.lcssa4073, i1 true)
  %iszero.i.i.i.i.i13.i.i1132 = icmp ne i64 %.lcssa4073, 0
  %cmp2.i.i.i.i14.i.i1133 = icmp eq i64 %1591, %sh_prom.i.i.i.i
  %cmp.i.i.i.i15.i.i1134 = select i1 %iszero.i.i.i.i.i13.i.i1132, i1 %cmp2.i.i.i.i14.i.i1133, i1 false
  %1592 = and i32 %.lcssa4074, 31
  %conv4.i.i.i19.i.i1138 = zext i1 %cmp.i.i.i.i15.i.i1134 to i32
  %shl.i.i.i20.i.i1139 = shl nuw i32 %conv4.i.i.i19.i.i1138, %1592
  %xor.i.i.i21.i.i1140 = xor i32 %shl.i.i.i20.i.i1139, -1
  %1593 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2574, !amdgpu.uniform !42

1594:                                             ; preds = %ComputeEnd2575
  %div8.i.i.i16.i.i1135 = lshr i32 %.lcssa4074, 5
  %idxprom.i.i.i17.i.i1136 = zext nneg i32 %div8.i.i.i16.i.i1135 to i64
  %1595 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i17.i.i1136
  %sunkaddr3308 = getelementptr inbounds i8, ptr addrspace(1) %1595, i64 40
  %1596 = atomicrmw and ptr addrspace(1) %sunkaddr3308, i32 %.lcssa4065 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %1597, !amdgpu.uniform !42

1597:                                             ; preds = %1594, %ComputeEnd2575
  call void @llvm.amdgcn.end.cf.i64(i64 %3569)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %for.cond.i.i1.i1158, !amdgpu.uniform !42

for.cond.i.i1.i1158:                              ; preds = %1597, %Flow3604
  %1598 = phi i32 [ %1636, %Flow3604 ], [ poison, %1597 ]
  %1599 = phi i64 [ %1637, %Flow3604 ], [ poison, %1597 ]
  %1600 = phi i32 [ %1638, %Flow3604 ], [ poison, %1597 ]
  %retval.sroa.6.0.i.i.i1159 = phi i32 [ %1641, %Flow3604 ], [ undef, %1597 ]
  %retval.sroa.2.0.i.i.i1161 = phi i64 [ %1640, %Flow3604 ], [ undef, %1597 ]
  %retval.sroa.8.0.i.i.i1162 = phi i32 [ %1639, %Flow3604 ], [ undef, %1597 ]
  %index.0.i.i.i1163 = phi i32 [ %1643, %Flow3604 ], [ 0, %1597 ]
  %1601 = load i32, ptr addrspace(1) @__llvm_rpc_client, align 8, !tbaa !68
  %cmp.not.i.i.i1164 = icmp ult i32 %index.0.i.i.i1163, %1601
  %spec.store.select.i.i.i1165 = select i1 %cmp.not.i.i.i1164, i32 %index.0.i.i.i1163, i32 0
  %1602 = tail call noundef i64 @llvm.amdgcn.ballot.i64(i1 true)
  %1603 = and i32 %spec.store.select.i.i.i1165, 31
  %1604 = lshr i64 %1602, %sh_prom.i.i.i.i
  %1605 = trunc i64 %1604 to i32
  %conv4.i.i.i.i.i1169 = and i32 %1605, 1
  %shl.i.i.i.i.i1170 = shl nuw i32 %conv4.i.i.i.i.i1169, %1603
  %1606 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2578, !amdgpu.uniform !42

1607:                                             ; preds = %ComputeEnd2579
  %sunkaddr3309 = getelementptr inbounds i8, ptr addrspace(1) %3579, i64 40
  %1608 = atomicrmw or ptr addrspace(1) %sunkaddr3309, i32 %.lcssa4059 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %1609, !amdgpu.uniform !42

1609:                                             ; preds = %1607, %ComputeEnd2579
  %1610 = phi i32 [ %1608, %1607 ], [ poison, %ComputeEnd2579 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %3588)
  %1611 = call i32 @llvm.amdgcn.readfirstlane.i32(i32 %1610)
  %1612 = or i32 %1611, %.lcssa4060
  %shl5.i.i.i.i.i1171 = shl nuw i32 1, %1603
  %and.i.i.i.i.i1172 = and i32 %shl5.i.i.i.i.i1171, %1612
  %tobool3.i.i.i.i1173 = icmp ne i32 %and.i.i.i.i.i1172, 0
  %1613 = tail call i64 @llvm.amdgcn.ballot.i64(i1 %tobool3.i.i.i.i1173)
  %and.i.i.i.i.i.i1174 = and i64 %1613, %1602
  %cmp.i.not.i.i.i1175 = icmp ne i64 %1602, %and.i.i.i.i.i.i1174
  br i1 %cmp.i.not.i.i.i1175, label %if.then.i.i.i.i1176, label %Flow3603, !amdgpu.uniform !42

if.then.i.i.i.i1176:                              ; preds = %1609
  fence syncscope("agent") acquire
  %1614 = load <2 x i64>, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !75
  %1615 = extractelement <2 x i64> %1614, i32 0
  %1616 = inttoptr i64 %1615 to ptr
  %1617 = extractelement <2 x i64> %1614, i32 1
  %1618 = inttoptr i64 %1617 to ptr
  %idxprom.i.i.i.i1177 = zext i32 %spec.store.select.i.i.i1165 to i64
  %arrayidx.i.i.i.i1178 = getelementptr inbounds nuw i32, ptr %1616, i64 %idxprom.i.i.i.i1177, !amdgpu.uniform !42
  %1619 = load atomic i32, ptr %arrayidx.i.i.i.i1178 monotonic, align 4
  %1620 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %1619)
  %arrayidx.i45.i.i.i1179 = getelementptr inbounds nuw i32, ptr %1618, i64 %idxprom.i.i.i.i1177, !amdgpu.uniform !42
  %1621 = load atomic i32, ptr %arrayidx.i45.i.i.i1179 monotonic, align 4
  %1622 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %1621)
  %cmp.i46.not.i.i.i1180 = icmp ne i32 %1620, %1622
  br i1 %cmp.i46.not.i.i.i1180, label %if.then12.i.i.i1181, label %Flow3601, !amdgpu.uniform !42

if.then12.i.i.i1181:                              ; preds = %if.then.i.i.i.i1176
  fence syncscope("agent") release
  %1623 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %1602, i1 true)
  %iszero.i.i.i.i.i.i.i1182 = icmp ne i64 %1602, 0
  %cmp2.i.i.i.i.i.i1183 = icmp eq i64 %1623, %sh_prom.i.i.i.i
  %cmp.i.i.i.i.i.i1184 = select i1 %iszero.i.i.i.i.i.i.i1182, i1 %cmp2.i.i.i.i.i.i1183, i1 false
  %conv4.i.i51.i.i.i1185 = zext i1 %cmp.i.i.i.i.i.i1184 to i32
  %shl.i.i52.i.i.i1186 = shl nuw i32 %conv4.i.i51.i.i.i1185, %1603
  %xor.i.i.i.i.i1187 = xor i32 %shl.i.i52.i.i.i1186, -1
  %1624 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2583, !amdgpu.uniform !42

1625:                                             ; preds = %ComputeEnd2584
  %sunkaddr3310 = getelementptr inbounds i8, ptr addrspace(1) %3579, i64 40
  %1626 = atomicrmw and ptr addrspace(1) %sunkaddr3310, i32 %.lcssa4061 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %1627, !amdgpu.uniform !42

1627:                                             ; preds = %1625, %ComputeEnd2584
  call void @llvm.amdgcn.end.cf.i64(i64 %3606)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %Flow3601, !amdgpu.uniform !42

if.end14.i.i.i1247:                               ; preds = %Flow3601
  %1628 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %1602, i1 true)
  %iszero.i.i.i.i.i.i1248 = icmp ne i64 %1602, 0
  %cmp2.i.i.i.i.i1249 = icmp eq i64 %1628, %sh_prom.i.i.i.i
  %cmp.i.i.i.i.i1250 = select i1 %iszero.i.i.i.i.i.i1248, i1 %cmp2.i.i.i.i.i1249, i1 false
  %1629 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.i.i.i.i.i1250)
  %1630 = extractvalue { i1, i64 } %1629, 0
  %1631 = extractvalue { i1, i64 } %1629, 1
  br i1 %1630, label %if.then16.i.i.i1252, label %if.end22.i.i.i1251

if.then16.i.i.i1252:                              ; preds = %if.end14.i.i.i1247
  %1632 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !76
  %opcode.i.i.i1253 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %1632, i64 %idxprom.i.i.i.i1177, i32 1
  store i32 1660944387, ptr %opcode.i.i.i1253, align 8, !tbaa !77
  %arrayidx21.i.i.i1254 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %1632, i64 %idxprom.i.i.i.i1177
  store i64 %1602, ptr %arrayidx21.i.i.i1254, align 8, !tbaa !80
  br label %if.end22.i.i.i1251, !amdgpu.uniform !42

Flow3602:                                         ; preds = %if.end22.i.i.i1251, %Flow3601
  %1633 = phi i32 [ %1620, %if.end22.i.i.i1251 ], [ %retval.sroa.8.0.i.i.i1162, %Flow3601 ]
  %1634 = phi i64 [ %1602, %if.end22.i.i.i1251 ], [ %retval.sroa.2.0.i.i.i1161, %Flow3601 ]
  %1635 = phi i32 [ %spec.store.select.i.i.i1165, %if.end22.i.i.i1251 ], [ %retval.sroa.6.0.i.i.i1159, %Flow3601 ]
  br label %cleanup26.i.i.i1188, !amdgpu.uniform !42

if.end22.i.i.i1251:                               ; preds = %if.then16.i.i.i1252, %if.end14.i.i.i1247
  call void @llvm.amdgcn.end.cf.i64(i64 %1631)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %Flow3602, !amdgpu.uniform !42

Flow3603:                                         ; preds = %cleanup26.i.i.i1188, %1609
  %1636 = phi i32 [ %1635, %cleanup26.i.i.i1188 ], [ %1598, %1609 ]
  %1637 = phi i64 [ %1634, %cleanup26.i.i.i1188 ], [ %1599, %1609 ]
  %1638 = phi i32 [ %1633, %cleanup26.i.i.i1188 ], [ %1600, %1609 ]
  %1639 = phi i32 [ %1633, %cleanup26.i.i.i1188 ], [ %retval.sroa.8.0.i.i.i1162, %1609 ]
  %1640 = phi i64 [ %1634, %cleanup26.i.i.i1188 ], [ %retval.sroa.2.0.i.i.i1161, %1609 ]
  %1641 = phi i32 [ %1635, %cleanup26.i.i.i1188 ], [ %retval.sroa.6.0.i.i.i1159, %1609 ]
  %1642 = phi i1 [ %cmp.i46.not.i.i.i1180, %cleanup26.i.i.i1188 ], [ true, %1609 ]
  br i1 %1642, label %for.inc.i.i.i1193, label %Flow3604, !amdgpu.uniform !42

cleanup26.i.i.i1188:                              ; preds = %Flow3602
  br label %Flow3603, !amdgpu.uniform !42

for.inc.i.i.i1193:                                ; preds = %Flow3603
  %inc.i.i.i1198 = add i32 %spec.store.select.i.i.i1165, 1
  br label %Flow3604, !amdgpu.uniform !42

Flow3604:                                         ; preds = %for.inc.i.i.i1193, %Flow3603
  %1643 = phi i32 [ %inc.i.i.i1198, %for.inc.i.i.i1193 ], [ poison, %Flow3603 ]
  %1644 = phi i1 [ false, %for.inc.i.i.i1193 ], [ true, %Flow3603 ]
  br i1 %1644, label %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1199, label %for.cond.i.i1.i1158, !amdgpu.uniform !42

_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1199: ; preds = %Flow3604
  %.lcssa4064 = phi i32 [ %1636, %Flow3604 ]
  %.lcssa4063 = phi i64 [ %1637, %Flow3604 ]
  %.lcssa4062 = phi i32 [ %1638, %Flow3604 ]
  fence acquire
  %1645 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %mul.i.i.i.i.i.i1200 = shl i32 %.lcssa4064, 6
  %idxprom.i19.i.i.i.i.i1201 = zext i32 %mul.i.i.i.i.i.i1200 to i64
  %arrayidx.i20.i.i.i.i.i1202 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %1645, i64 %idxprom.i19.i.i.i.i.i1201
  %arrayidx.i22.i.i.i.i.i1203 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %arrayidx.i20.i.i.i.i.i1202, i64 %sh_prom.i.i.i.i
  store i64 sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.36.13 to ptr), i64 6) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.36.13 to ptr) to i64)), ptr %arrayidx.i22.i.i.i.i.i1203, align 8, !tbaa !82
  %spec.select.i.i.i.i.i.i.i1204 = tail call i64 @llvm.umin.i64(i64 sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.36.13 to ptr), i64 6) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.36.13 to ptr) to i64)), i64 56)
  %1646 = and i64 %spec.select.i.i.i.i.i.i.i1204, 15
  %1647 = sub i64 %spec.select.i.i.i.i.i.i.i1204, %1646
  %1648 = icmp ne i64 %1647, 0
  br i1 %1648, label %loop-memcpy-expansion2386.preheader, label %Flow3600, !amdgpu.uniform !42

loop-memcpy-expansion2386.preheader:              ; preds = %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1199
  %1649 = add nuw nsw i64 %75, 8
  %scevgep3011 = getelementptr i8, ptr %1645, i64 %1649
  %1650 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i1201, 6
  %scevgep3012 = getelementptr i8, ptr %scevgep3011, i64 %1650
  br label %loop-memcpy-expansion2386, !amdgpu.uniform !42

Flow3600:                                         ; preds = %Flow3599, %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1199
  br label %loop-memcpy-residual-header2389, !amdgpu.uniform !42

loop-memcpy-expansion2386:                        ; preds = %loop-memcpy-expansion2386.preheader, %loop-memcpy-expansion2386
  %loop-index2387 = phi i64 [ %1653, %loop-memcpy-expansion2386 ], [ 0, %loop-memcpy-expansion2386.preheader ]
  %1651 = getelementptr i8, ptr addrspace(4) @.str.36.13, i64 %loop-index2387, !amdgpu.uniform !42
  %1652 = load <4 x i32>, ptr addrspace(4) %1651, align 1
  %scevgep3013 = getelementptr i8, ptr %scevgep3012, i64 %loop-index2387
  store <4 x i32> %1652, ptr %scevgep3013, align 1
  %1653 = add i64 %loop-index2387, 16
  %1654 = icmp uge i64 %1653, %1647
  br i1 %1654, label %Flow3599, label %loop-memcpy-expansion2386, !amdgpu.uniform !42

Flow3597:                                         ; preds = %Flow3594, %Flow3596
  %1655 = phi i64 [ %2896, %Flow3594 ], [ %2893, %Flow3596 ]
  br label %post-loop-memcpy-expansion2385, !amdgpu.uniform !42

loop-memcpy-residual2388:                         ; preds = %loop-memcpy-residual2388.preheader, %loop-memcpy-residual2388
  %residual-loop-index2390 = phi i64 [ %1657, %loop-memcpy-residual2388 ], [ 0, %loop-memcpy-residual2388.preheader ]
  %scevgep3010 = getelementptr i8, ptr addrspace(4) @.str.36.13, i64 %residual-loop-index2390, !amdgpu.uniform !42
  %1656 = load i8, ptr addrspace(4) %scevgep3010, align 1
  %scevgep3009 = getelementptr i8, ptr %scevgep3008, i64 %residual-loop-index2390
  store i8 %1656, ptr %scevgep3009, align 1
  %1657 = add i64 %residual-loop-index2390, 1
  %1658 = icmp uge i64 %1657, %1646
  br i1 %1658, label %Flow3594, label %loop-memcpy-residual2388, !amdgpu.uniform !42

Flow3594:                                         ; preds = %loop-memcpy-residual2388
  br label %Flow3597, !amdgpu.uniform !42

post-loop-memcpy-expansion2385:                   ; preds = %Flow3597
  %tobool.not.i.i.i.i.i.i1206 = icmp eq i32 %.lcssa4062, 0
  %conv.i.i.i.i.i.i1207 = zext i1 %tobool.not.i.i.i.i.i.i1206 to i32
  fence release
  %1659 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %idxprom.i23.i.i.i.i.i1208 = zext i32 %.lcssa4064 to i64
  %arrayidx.i24.i.i.i.i.i1209 = getelementptr inbounds nuw i32, ptr %1659, i64 %idxprom.i23.i.i.i.i.i1208
  store atomic i32 %conv.i.i.i.i.i.i1207, ptr %arrayidx.i24.i.i.i.i.i1209 monotonic, align 4
  %1660 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !84
  %arrayidx.i.i5.i.i1210 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %1660, i64 %idxprom.i23.i.i.i.i.i1208, !amdgpu.uniform !42
  %1661 = load i64, ptr %arrayidx.i.i5.i.i1210, align 8, !tbaa !80
  br label %while.cond.i.i.i.i1211, !amdgpu.uniform !42

while.cond.i.i.i.i1211:                           ; preds = %post-loop-memcpy-expansion2385, %Flow3592
  %phi.broken3932 = phi i64 [ 0, %post-loop-memcpy-expansion2385 ], [ %1678, %Flow3592 ]
  %lsr.iv3000 = phi i64 [ %1676, %Flow3592 ], [ -50, %post-loop-memcpy-expansion2385 ]
  %lsr.iv2995 = phi ptr addrspace(4) [ %1675, %Flow3592 ], [ getelementptr (i8, ptr addrspace(4) @.str.36.13, i64 56), %post-loop-memcpy-expansion2385 ]
  %port.sroa.37.0.i.i1212 = phi i32 [ %1674, %Flow3592 ], [ %conv.i.i.i.i.i.i1207, %post-loop-memcpy-expansion2385 ]
  %idx.0.i.i.i.i1213 = phi i64 [ %1673, %Flow3592 ], [ 56, %post-loop-memcpy-expansion2385 ]
  %cmp.i.i.i.i1214 = icmp ult i64 %idx.0.i.i.i.i1213, sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.36.13 to ptr), i64 6) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.36.13 to ptr) to i64))
  %1662 = tail call i64 @llvm.amdgcn.ballot.i64(i1 %cmp.i.i.i.i1214)
  %and.i.i.i.i6.i.i1215 = and i64 %1662, %1661
  %tobool.not.i.i.i.i1216 = icmp ne i64 %and.i.i.i.i6.i.i1215, 0
  %1663 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i22.i.i.i1217 = getelementptr inbounds nuw i32, ptr %1663, i64 %idxprom.i23.i.i.i.i.i1208, !amdgpu.uniform !42
  %1664 = load atomic i32, ptr %arrayidx.i22.i.i.i1217 monotonic, align 4
  %1665 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %tobool.not.i.i.i.i1216)
  %1666 = extractvalue { i1, i64 } %1665, 0
  %1667 = extractvalue { i1, i64 } %1665, 1
  br i1 %1666, label %cond.false.i8.i.i.i.i1218, label %Flow3592

cond.false.i8.i.i.i.i1218:                        ; preds = %while.cond.i.i.i.i1211
  %1668 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %1664)
  %1669 = icmp ne i32 %1668, %port.sroa.37.0.i.i1212
  br i1 %1669, label %while.body.i.i21.i.i.i.i1219.preheader, label %Flow3591, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i1219.preheader:           ; preds = %cond.false.i8.i.i.i.i1218
  br label %while.body.i.i21.i.i.i.i1219, !amdgpu.uniform !42

Flow3591:                                         ; preds = %Flow3589, %cond.false.i8.i.i.i.i1218
  br label %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i1222, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i1219:                     ; preds = %while.body.i.i21.i.i.i.i1219.preheader, %while.body.i.i21.i.i.i.i1219
  tail call void @llvm.amdgcn.s.sleep(i32 2)
  %1670 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i22.i.i.i.i1220 = getelementptr inbounds nuw i32, ptr %1670, i64 %idxprom.i23.i.i.i.i.i1208, !amdgpu.uniform !42
  %1671 = load atomic i32, ptr %arrayidx.i.i.i22.i.i.i.i1220 monotonic, align 4
  %1672 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %1671)
  %cmp.i.not.i.i23.i.i.i.i1221 = icmp eq i32 %1672, %port.sroa.37.0.i.i1212
  br i1 %cmp.i.not.i.i23.i.i.i.i1221, label %Flow3589, label %while.body.i.i21.i.i.i.i1219, !amdgpu.uniform !42

Flow3589:                                         ; preds = %while.body.i.i21.i.i.i.i1219
  br label %Flow3591, !amdgpu.uniform !42

Flow3592:                                         ; preds = %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1223, %while.cond.i.i.i.i1211
  %1673 = phi i64 [ %add.i.i.i.i1226, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1223 ], [ poison, %while.cond.i.i.i.i1211 ]
  %1674 = phi i32 [ %conv.i.i26.i.i.i.i1224, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1223 ], [ poison, %while.cond.i.i.i.i1211 ]
  %1675 = phi ptr addrspace(4) [ %scevgep2996, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1223 ], [ poison, %while.cond.i.i.i.i1211 ]
  %1676 = phi i64 [ %lsr.iv.next3001, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1223 ], [ poison, %while.cond.i.i.i.i1211 ]
  %1677 = phi i1 [ false, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1223 ], [ true, %while.cond.i.i.i.i1211 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %1667)
  %1678 = call i64 @llvm.amdgcn.if.break.i64(i1 %1677, i64 %phi.broken3932)
  %1679 = call i1 @llvm.amdgcn.loop.i64(i64 %1678)
  br i1 %1679, label %cond.false.i.i.i1233, label %while.cond.i.i.i.i1211

_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i1222: ; preds = %Flow3591
  fence acquire
  br i1 %cmp.i.i.i.i1214, label %if.then.i.i.i.i.i.i.i1227, label %Flow3588, !amdgpu.uniform !42

if.then.i.i.i.i.i.i.i1227:                        ; preds = %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i1222
  %sub.i.i.i.i.i.i.i1229 = sub nuw i64 sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.36.13 to ptr), i64 6) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.36.13 to ptr) to i64)), %idx.0.i.i.i.i1213
  %spec.select.i.i.i36.i.i.i.i1230 = tail call i64 @llvm.umin.i64(i64 %sub.i.i.i.i.i.i.i1229, i64 64)
  %1680 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %1681 = and i64 %spec.select.i.i.i36.i.i.i.i1230, 15
  %1682 = sub i64 %spec.select.i.i.i36.i.i.i.i1230, %1681
  %1683 = icmp ne i64 %1682, 0
  br i1 %1683, label %loop-memcpy-expansion2392.preheader, label %Flow3587, !amdgpu.uniform !42

loop-memcpy-expansion2392.preheader:              ; preds = %if.then.i.i.i.i.i.i.i1227
  %arrayidx.i20.i35.i.i.i.i1231 = getelementptr %"struct.rpc::Buffer", ptr %1680, i64 %idxprom.i19.i.i.i.i.i1201
  %arrayidx.i22.i38.i.i.i.i1232 = getelementptr %"struct.rpc::Buffer", ptr %arrayidx.i20.i35.i.i.i.i1231, i64 %sh_prom.i.i.i.i
  br label %loop-memcpy-expansion2392, !amdgpu.uniform !42

Flow3587:                                         ; preds = %Flow3586, %if.then.i.i.i.i.i.i.i1227
  br label %loop-memcpy-residual-header2395, !amdgpu.uniform !42

loop-memcpy-expansion2392:                        ; preds = %loop-memcpy-expansion2392.preheader, %loop-memcpy-expansion2392
  %loop-index2393 = phi i64 [ %1686, %loop-memcpy-expansion2392 ], [ 0, %loop-memcpy-expansion2392.preheader ]
  %scevgep2997 = getelementptr i8, ptr addrspace(4) %lsr.iv2995, i64 %loop-index2393, !amdgpu.uniform !42
  %1684 = load <4 x i32>, ptr addrspace(4) %scevgep2997, align 1
  %1685 = getelementptr i8, ptr %arrayidx.i22.i38.i.i.i.i1232, i64 %loop-index2393
  store <4 x i32> %1684, ptr %1685, align 1
  %1686 = add i64 %loop-index2393, 16
  %1687 = icmp uge i64 %1686, %1682
  br i1 %1687, label %Flow3586, label %loop-memcpy-expansion2392, !amdgpu.uniform !42

Flow3584:                                         ; preds = %Flow3583, %loop-memcpy-residual-header2395
  br label %Flow3588, !amdgpu.uniform !42

loop-memcpy-residual2394:                         ; preds = %loop-memcpy-residual2394.preheader, %loop-memcpy-residual2394
  %residual-loop-index2396 = phi i64 [ %1689, %loop-memcpy-residual2394 ], [ 0, %loop-memcpy-residual2394.preheader ]
  %scevgep3006 = getelementptr i8, ptr addrspace(4) %scevgep3005, i64 %residual-loop-index2396, !amdgpu.uniform !42
  %1688 = load i8, ptr addrspace(4) %scevgep3006, align 1
  %scevgep3004 = getelementptr i8, ptr %scevgep3003, i64 %residual-loop-index2396
  store i8 %1688, ptr %scevgep3004, align 1
  %1689 = add i64 %residual-loop-index2396, 1
  %1690 = icmp uge i64 %1689, %1681
  br i1 %1690, label %Flow3583, label %loop-memcpy-residual2394, !amdgpu.uniform !42

Flow3583:                                         ; preds = %loop-memcpy-residual2394
  br label %Flow3584, !amdgpu.uniform !42

_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1223: ; preds = %Flow3588
  %conv.i.i26.i.i.i.i1224 = xor i32 %port.sroa.37.0.i.i1212, 1
  fence release
  %1691 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %arrayidx.i24.i29.i.i.i.i1225 = getelementptr inbounds nuw i32, ptr %1691, i64 %idxprom.i23.i.i.i.i.i1208
  store atomic i32 %conv.i.i26.i.i.i.i1224, ptr %arrayidx.i24.i29.i.i.i.i1225 monotonic, align 4
  %add.i.i.i.i1226 = add i64 %idx.0.i.i.i.i1213, 64
  %scevgep2996 = getelementptr i8, ptr addrspace(4) %lsr.iv2995, i64 64
  %lsr.iv.next3001 = add i64 %lsr.iv3000, -64
  br label %Flow3592, !amdgpu.uniform !42

cond.false.i.i.i1233:                             ; preds = %Flow3592
  %.lcssa4058 = phi i64 [ %1678, %Flow3592 ]
  %port.sroa.37.0.i.i1212.lcssa = phi i32 [ %port.sroa.37.0.i.i1212, %Flow3592 ]
  %.lcssa4057 = phi i32 [ %1664, %Flow3592 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa4058)
  %1692 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %.lcssa4057)
  %1693 = icmp ne i32 %1692, %port.sroa.37.0.i.i1212.lcssa
  %1694 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %1693)
  %1695 = extractvalue { i1, i64 } %1694, 0
  %1696 = extractvalue { i1, i64 } %1694, 1
  br i1 %1695, label %while.body.i.i.i.i1234.preheader, label %Flow3581

while.body.i.i.i.i1234.preheader:                 ; preds = %cond.false.i.i.i1233
  br label %while.body.i.i.i.i1234, !amdgpu.uniform !42

Flow3581:                                         ; preds = %Flow3579, %cond.false.i.i.i1233
  call void @llvm.amdgcn.end.cf.i64(i64 %1696)
  br label %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit1255, !amdgpu.uniform !42

while.body.i.i.i.i1234:                           ; preds = %while.body.i.i.i.i1234.preheader, %while.body.i.i.i.i1234
  %phi.broken3933 = phi i64 [ 0, %while.body.i.i.i.i1234.preheader ], [ %1700, %while.body.i.i.i.i1234 ]
  tail call void @llvm.amdgcn.s.sleep(i32 2)
  %1697 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i8.i.i1235 = getelementptr inbounds nuw i32, ptr %1697, i64 %idxprom.i23.i.i.i.i.i1208, !amdgpu.uniform !42
  %1698 = load atomic i32, ptr %arrayidx.i.i.i8.i.i1235 monotonic, align 4
  %1699 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %1698)
  %cmp.i.not.i.i.i.i1236 = icmp eq i32 %1699, %port.sroa.37.0.i.i1212.lcssa
  %1700 = call i64 @llvm.amdgcn.if.break.i64(i1 %cmp.i.not.i.i.i.i1236, i64 %phi.broken3933)
  %1701 = call i1 @llvm.amdgcn.loop.i64(i64 %1700)
  br i1 %1701, label %Flow3579, label %while.body.i.i.i.i1234

Flow3579:                                         ; preds = %while.body.i.i.i.i1234
  %.lcssa4056 = phi i64 [ %1700, %while.body.i.i.i.i1234 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa4056)
  br label %Flow3581, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit1255: ; preds = %Flow3581
  fence acquire
  tail call void @llvm.amdgcn.wave.barrier()
  fence syncscope("agent") release
  %1702 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %.lcssa4063, i1 true)
  %iszero.i.i.i.i.i13.i.i1238 = icmp ne i64 %.lcssa4063, 0
  %cmp2.i.i.i.i14.i.i1239 = icmp eq i64 %1702, %sh_prom.i.i.i.i
  %cmp.i.i.i.i15.i.i1240 = select i1 %iszero.i.i.i.i.i13.i.i1238, i1 %cmp2.i.i.i.i14.i.i1239, i1 false
  %1703 = and i32 %.lcssa4064, 31
  %conv4.i.i.i19.i.i1244 = zext i1 %cmp.i.i.i.i15.i.i1240 to i32
  %shl.i.i.i20.i.i1245 = shl nuw i32 %conv4.i.i.i19.i.i1244, %1703
  %xor.i.i.i21.i.i1246 = xor i32 %shl.i.i.i20.i.i1245, -1
  %1704 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2587, !amdgpu.uniform !42

1705:                                             ; preds = %ComputeEnd2588
  %div8.i.i.i16.i.i1241 = lshr i32 %.lcssa4064, 5
  %idxprom.i.i.i17.i.i1242 = zext nneg i32 %div8.i.i.i16.i.i1241 to i64
  %1706 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i17.i.i1242
  %sunkaddr3311 = getelementptr inbounds i8, ptr addrspace(1) %1706, i64 40
  %1707 = atomicrmw and ptr addrspace(1) %sunkaddr3311, i32 %.lcssa4055 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %1708, !amdgpu.uniform !42

1708:                                             ; preds = %1705, %ComputeEnd2588
  call void @llvm.amdgcn.end.cf.i64(i64 %3623)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %for.cond.i.i.i1257, !amdgpu.uniform !42

for.cond.i.i.i1257:                               ; preds = %1708, %Flow3578
  %1709 = phi i32 [ %1747, %Flow3578 ], [ poison, %1708 ]
  %1710 = phi i64 [ %1748, %Flow3578 ], [ poison, %1708 ]
  %1711 = phi i32 [ %1749, %Flow3578 ], [ poison, %1708 ]
  %retval.sroa.6.0.i.i.i1258 = phi i32 [ %1752, %Flow3578 ], [ undef, %1708 ]
  %retval.sroa.2.0.i.i.i1260 = phi i64 [ %1751, %Flow3578 ], [ undef, %1708 ]
  %retval.sroa.8.0.i.i.i1261 = phi i32 [ %1750, %Flow3578 ], [ undef, %1708 ]
  %index.0.i.i.i1262 = phi i32 [ %1754, %Flow3578 ], [ 0, %1708 ]
  %1712 = load i32, ptr addrspace(1) @__llvm_rpc_client, align 8, !tbaa !68
  %cmp.not.i.i.i1263 = icmp ult i32 %index.0.i.i.i1262, %1712
  %spec.store.select.i.i.i1264 = select i1 %cmp.not.i.i.i1263, i32 %index.0.i.i.i1262, i32 0
  %1713 = tail call noundef i64 @llvm.amdgcn.ballot.i64(i1 true)
  %1714 = and i32 %spec.store.select.i.i.i1264, 31
  %1715 = lshr i64 %1713, %sh_prom.i.i.i.i
  %1716 = trunc i64 %1715 to i32
  %conv4.i.i.i.i.i1268 = and i32 %1716, 1
  %shl.i.i.i.i.i1269 = shl nuw i32 %conv4.i.i.i.i.i1268, %1714
  %1717 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2591, !amdgpu.uniform !42

1718:                                             ; preds = %ComputeEnd2592
  %sunkaddr3312 = getelementptr inbounds i8, ptr addrspace(1) %3633, i64 40
  %1719 = atomicrmw or ptr addrspace(1) %sunkaddr3312, i32 %.lcssa4049 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %1720, !amdgpu.uniform !42

1720:                                             ; preds = %1718, %ComputeEnd2592
  %1721 = phi i32 [ %1719, %1718 ], [ poison, %ComputeEnd2592 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %3642)
  %1722 = call i32 @llvm.amdgcn.readfirstlane.i32(i32 %1721)
  %1723 = or i32 %1722, %.lcssa4050
  %shl5.i.i.i.i.i1270 = shl nuw i32 1, %1714
  %and.i.i.i.i.i1271 = and i32 %shl5.i.i.i.i.i1270, %1723
  %tobool3.i.i.i.i1272 = icmp ne i32 %and.i.i.i.i.i1271, 0
  %1724 = tail call i64 @llvm.amdgcn.ballot.i64(i1 %tobool3.i.i.i.i1272)
  %and.i.i.i.i.i.i1273 = and i64 %1724, %1713
  %cmp.i.not.i.i.i1274 = icmp ne i64 %1713, %and.i.i.i.i.i.i1273
  br i1 %cmp.i.not.i.i.i1274, label %if.then.i.i.i.i1275, label %Flow3577, !amdgpu.uniform !42

if.then.i.i.i.i1275:                              ; preds = %1720
  fence syncscope("agent") acquire
  %1725 = load <2 x i64>, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !75
  %1726 = extractelement <2 x i64> %1725, i32 0
  %1727 = inttoptr i64 %1726 to ptr
  %1728 = extractelement <2 x i64> %1725, i32 1
  %1729 = inttoptr i64 %1728 to ptr
  %idxprom.i.i.i.i1276 = zext i32 %spec.store.select.i.i.i1264 to i64
  %arrayidx.i.i.i.i1277 = getelementptr inbounds nuw i32, ptr %1727, i64 %idxprom.i.i.i.i1276, !amdgpu.uniform !42
  %1730 = load atomic i32, ptr %arrayidx.i.i.i.i1277 monotonic, align 4
  %1731 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %1730)
  %arrayidx.i45.i.i.i1278 = getelementptr inbounds nuw i32, ptr %1729, i64 %idxprom.i.i.i.i1276, !amdgpu.uniform !42
  %1732 = load atomic i32, ptr %arrayidx.i45.i.i.i1278 monotonic, align 4
  %1733 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %1732)
  %cmp.i46.not.i.i.i1279 = icmp ne i32 %1731, %1733
  br i1 %cmp.i46.not.i.i.i1279, label %if.then12.i.i.i1280, label %Flow3575, !amdgpu.uniform !42

if.then12.i.i.i1280:                              ; preds = %if.then.i.i.i.i1275
  fence syncscope("agent") release
  %1734 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %1713, i1 true)
  %iszero.i.i.i.i.i.i.i1281 = icmp ne i64 %1713, 0
  %cmp2.i.i.i.i.i.i1282 = icmp eq i64 %1734, %sh_prom.i.i.i.i
  %cmp.i.i.i.i.i.i1283 = select i1 %iszero.i.i.i.i.i.i.i1281, i1 %cmp2.i.i.i.i.i.i1282, i1 false
  %conv4.i.i51.i.i.i1284 = zext i1 %cmp.i.i.i.i.i.i1283 to i32
  %shl.i.i52.i.i.i1285 = shl nuw i32 %conv4.i.i51.i.i.i1284, %1714
  %xor.i.i.i.i.i1286 = xor i32 %shl.i.i52.i.i.i1285, -1
  %1735 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2596, !amdgpu.uniform !42

1736:                                             ; preds = %ComputeEnd2597
  %sunkaddr3313 = getelementptr inbounds i8, ptr addrspace(1) %3633, i64 40
  %1737 = atomicrmw and ptr addrspace(1) %sunkaddr3313, i32 %.lcssa4051 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %1738, !amdgpu.uniform !42

1738:                                             ; preds = %1736, %ComputeEnd2597
  call void @llvm.amdgcn.end.cf.i64(i64 %3660)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %Flow3575, !amdgpu.uniform !42

if.end14.i.i.i1346:                               ; preds = %Flow3575
  %1739 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %1713, i1 true)
  %iszero.i.i.i.i.i.i1347 = icmp ne i64 %1713, 0
  %cmp2.i.i.i.i.i1348 = icmp eq i64 %1739, %sh_prom.i.i.i.i
  %cmp.i.i.i.i.i1349 = select i1 %iszero.i.i.i.i.i.i1347, i1 %cmp2.i.i.i.i.i1348, i1 false
  %1740 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.i.i.i.i.i1349)
  %1741 = extractvalue { i1, i64 } %1740, 0
  %1742 = extractvalue { i1, i64 } %1740, 1
  br i1 %1741, label %if.then16.i.i.i1351, label %if.end22.i.i.i1350

if.then16.i.i.i1351:                              ; preds = %if.end14.i.i.i1346
  %1743 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !76
  %opcode.i.i.i1352 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %1743, i64 %idxprom.i.i.i.i1276, i32 1
  store i32 1660944387, ptr %opcode.i.i.i1352, align 8, !tbaa !77
  %arrayidx21.i.i.i1353 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %1743, i64 %idxprom.i.i.i.i1276
  store i64 %1713, ptr %arrayidx21.i.i.i1353, align 8, !tbaa !80
  br label %if.end22.i.i.i1350, !amdgpu.uniform !42

Flow3576:                                         ; preds = %if.end22.i.i.i1350, %Flow3575
  %1744 = phi i32 [ %1731, %if.end22.i.i.i1350 ], [ %retval.sroa.8.0.i.i.i1261, %Flow3575 ]
  %1745 = phi i64 [ %1713, %if.end22.i.i.i1350 ], [ %retval.sroa.2.0.i.i.i1260, %Flow3575 ]
  %1746 = phi i32 [ %spec.store.select.i.i.i1264, %if.end22.i.i.i1350 ], [ %retval.sroa.6.0.i.i.i1258, %Flow3575 ]
  br label %cleanup26.i.i.i1287, !amdgpu.uniform !42

if.end22.i.i.i1350:                               ; preds = %if.then16.i.i.i1351, %if.end14.i.i.i1346
  call void @llvm.amdgcn.end.cf.i64(i64 %1742)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %Flow3576, !amdgpu.uniform !42

Flow3577:                                         ; preds = %cleanup26.i.i.i1287, %1720
  %1747 = phi i32 [ %1746, %cleanup26.i.i.i1287 ], [ %1709, %1720 ]
  %1748 = phi i64 [ %1745, %cleanup26.i.i.i1287 ], [ %1710, %1720 ]
  %1749 = phi i32 [ %1744, %cleanup26.i.i.i1287 ], [ %1711, %1720 ]
  %1750 = phi i32 [ %1744, %cleanup26.i.i.i1287 ], [ %retval.sroa.8.0.i.i.i1261, %1720 ]
  %1751 = phi i64 [ %1745, %cleanup26.i.i.i1287 ], [ %retval.sroa.2.0.i.i.i1260, %1720 ]
  %1752 = phi i32 [ %1746, %cleanup26.i.i.i1287 ], [ %retval.sroa.6.0.i.i.i1258, %1720 ]
  %1753 = phi i1 [ %cmp.i46.not.i.i.i1279, %cleanup26.i.i.i1287 ], [ true, %1720 ]
  br i1 %1753, label %for.inc.i.i.i1292, label %Flow3578, !amdgpu.uniform !42

cleanup26.i.i.i1287:                              ; preds = %Flow3576
  br label %Flow3577, !amdgpu.uniform !42

for.inc.i.i.i1292:                                ; preds = %Flow3577
  %inc.i.i.i1297 = add i32 %spec.store.select.i.i.i1264, 1
  br label %Flow3578, !amdgpu.uniform !42

Flow3578:                                         ; preds = %for.inc.i.i.i1292, %Flow3577
  %1754 = phi i32 [ %inc.i.i.i1297, %for.inc.i.i.i1292 ], [ poison, %Flow3577 ]
  %1755 = phi i1 [ false, %for.inc.i.i.i1292 ], [ true, %Flow3577 ]
  br i1 %1755, label %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1298, label %for.cond.i.i.i1257, !amdgpu.uniform !42

_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1298: ; preds = %Flow3578
  %.lcssa4054 = phi i32 [ %1747, %Flow3578 ]
  %.lcssa4053 = phi i64 [ %1748, %Flow3578 ]
  %.lcssa4052 = phi i32 [ %1749, %Flow3578 ]
  fence acquire
  %1756 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %mul.i.i.i.i.i.i1299 = shl i32 %.lcssa4054, 6
  %idxprom.i19.i.i.i.i.i1300 = zext i32 %mul.i.i.i.i.i.i1299 to i64
  %arrayidx.i20.i.i.i.i.i1301 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %1756, i64 %idxprom.i19.i.i.i.i.i1300
  %arrayidx.i22.i.i.i.i.i1302 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %arrayidx.i20.i.i.i.i.i1301, i64 %sh_prom.i.i.i.i
  store i64 sub (i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) getelementptr inbounds nuw (i8, ptr addrspace(4) @.str.9, i64 8) to ptr) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.9 to ptr) to i64)), ptr %arrayidx.i22.i.i.i.i.i1302, align 8, !tbaa !82
  %spec.select.i.i.i.i.i.i.i1303 = tail call i64 @llvm.umin.i64(i64 sub (i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) getelementptr inbounds nuw (i8, ptr addrspace(4) @.str.9, i64 8) to ptr) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.9 to ptr) to i64)), i64 56)
  %1757 = and i64 %spec.select.i.i.i.i.i.i.i1303, 15
  %1758 = sub i64 %spec.select.i.i.i.i.i.i.i1303, %1757
  %1759 = icmp ne i64 %1758, 0
  br i1 %1759, label %loop-memcpy-expansion2398.preheader, label %Flow3574, !amdgpu.uniform !42

loop-memcpy-expansion2398.preheader:              ; preds = %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1298
  %1760 = add nuw nsw i64 %75, 8
  %scevgep2992 = getelementptr i8, ptr %1756, i64 %1760
  %1761 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i1300, 6
  %scevgep2993 = getelementptr i8, ptr %scevgep2992, i64 %1761
  br label %loop-memcpy-expansion2398, !amdgpu.uniform !42

Flow3574:                                         ; preds = %Flow3573, %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1298
  br label %loop-memcpy-residual-header2401, !amdgpu.uniform !42

loop-memcpy-expansion2398:                        ; preds = %loop-memcpy-expansion2398.preheader, %loop-memcpy-expansion2398
  %loop-index2399 = phi i64 [ %1764, %loop-memcpy-expansion2398 ], [ 0, %loop-memcpy-expansion2398.preheader ]
  %1762 = getelementptr i8, ptr addrspace(4) @.str.9, i64 %loop-index2399, !amdgpu.uniform !42
  %1763 = load <4 x i32>, ptr addrspace(4) %1762, align 1
  %scevgep2994 = getelementptr i8, ptr %scevgep2993, i64 %loop-index2399
  store <4 x i32> %1763, ptr %scevgep2994, align 1
  %1764 = add i64 %loop-index2399, 16
  %1765 = icmp uge i64 %1764, %1758
  br i1 %1765, label %Flow3573, label %loop-memcpy-expansion2398, !amdgpu.uniform !42

Flow3571:                                         ; preds = %Flow3568, %Flow3570
  %1766 = phi i64 [ %2907, %Flow3568 ], [ %2901, %Flow3570 ]
  br label %post-loop-memcpy-expansion2397, !amdgpu.uniform !42

loop-memcpy-residual2400:                         ; preds = %loop-memcpy-residual2400.preheader, %loop-memcpy-residual2400
  %residual-loop-index2402 = phi i64 [ %1768, %loop-memcpy-residual2400 ], [ 0, %loop-memcpy-residual2400.preheader ]
  %scevgep2991 = getelementptr i8, ptr addrspace(4) %scevgep2990, i64 %residual-loop-index2402, !amdgpu.uniform !42
  %1767 = load i8, ptr addrspace(4) %scevgep2991, align 1
  %scevgep2989 = getelementptr i8, ptr %scevgep2988, i64 %residual-loop-index2402
  store i8 %1767, ptr %scevgep2989, align 1
  %1768 = add i64 %residual-loop-index2402, 1
  %1769 = icmp uge i64 %1768, %1757
  br i1 %1769, label %Flow3568, label %loop-memcpy-residual2400, !amdgpu.uniform !42

Flow3568:                                         ; preds = %loop-memcpy-residual2400
  br label %Flow3571, !amdgpu.uniform !42

post-loop-memcpy-expansion2397:                   ; preds = %Flow3571
  %tobool.not.i.i.i.i.i.i1305 = icmp eq i32 %.lcssa4052, 0
  %conv.i.i.i.i.i.i1306 = zext i1 %tobool.not.i.i.i.i.i.i1305 to i32
  fence release
  %1770 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %idxprom.i23.i.i.i.i.i1307 = zext i32 %.lcssa4054 to i64
  %arrayidx.i24.i.i.i.i.i1308 = getelementptr inbounds nuw i32, ptr %1770, i64 %idxprom.i23.i.i.i.i.i1307
  store atomic i32 %conv.i.i.i.i.i.i1306, ptr %arrayidx.i24.i.i.i.i.i1308 monotonic, align 4
  %1771 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !84
  %arrayidx.i.i5.i.i1309 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %1771, i64 %idxprom.i23.i.i.i.i.i1307, !amdgpu.uniform !42
  %1772 = load i64, ptr %arrayidx.i.i5.i.i1309, align 8, !tbaa !80
  br label %while.cond.i.i.i.i1310, !amdgpu.uniform !42

while.cond.i.i.i.i1310:                           ; preds = %post-loop-memcpy-expansion2397, %Flow3566
  %phi.broken3934 = phi i64 [ 0, %post-loop-memcpy-expansion2397 ], [ %1789, %Flow3566 ]
  %lsr.iv2980 = phi i64 [ %1787, %Flow3566 ], [ sub (i64 add (i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) getelementptr inbounds nuw (i8, ptr addrspace(4) @.str.9, i64 8) to ptr) to i64), i64 -56), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.9 to ptr) to i64)), %post-loop-memcpy-expansion2397 ]
  %lsr.iv2975 = phi ptr addrspace(4) [ %1786, %Flow3566 ], [ getelementptr (i8, ptr addrspace(4) @.str.9, i64 56), %post-loop-memcpy-expansion2397 ]
  %port.sroa.37.0.i.i1311 = phi i32 [ %1785, %Flow3566 ], [ %conv.i.i.i.i.i.i1306, %post-loop-memcpy-expansion2397 ]
  %idx.0.i.i.i.i1312 = phi i64 [ %1784, %Flow3566 ], [ 56, %post-loop-memcpy-expansion2397 ]
  %cmp.i.i.i.i1313 = icmp ult i64 %idx.0.i.i.i.i1312, sub (i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) getelementptr inbounds nuw (i8, ptr addrspace(4) @.str.9, i64 8) to ptr) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.9 to ptr) to i64))
  %1773 = tail call i64 @llvm.amdgcn.ballot.i64(i1 %cmp.i.i.i.i1313)
  %and.i.i.i.i6.i.i1314 = and i64 %1773, %1772
  %tobool.not.i.i.i.i1315 = icmp ne i64 %and.i.i.i.i6.i.i1314, 0
  %1774 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i22.i.i.i1316 = getelementptr inbounds nuw i32, ptr %1774, i64 %idxprom.i23.i.i.i.i.i1307, !amdgpu.uniform !42
  %1775 = load atomic i32, ptr %arrayidx.i22.i.i.i1316 monotonic, align 4
  %1776 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %tobool.not.i.i.i.i1315)
  %1777 = extractvalue { i1, i64 } %1776, 0
  %1778 = extractvalue { i1, i64 } %1776, 1
  br i1 %1777, label %cond.false.i8.i.i.i.i1317, label %Flow3566

cond.false.i8.i.i.i.i1317:                        ; preds = %while.cond.i.i.i.i1310
  %1779 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %1775)
  %1780 = icmp ne i32 %1779, %port.sroa.37.0.i.i1311
  br i1 %1780, label %while.body.i.i21.i.i.i.i1318.preheader, label %Flow3565, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i1318.preheader:           ; preds = %cond.false.i8.i.i.i.i1317
  br label %while.body.i.i21.i.i.i.i1318, !amdgpu.uniform !42

Flow3565:                                         ; preds = %Flow3563, %cond.false.i8.i.i.i.i1317
  br label %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i1321, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i1318:                     ; preds = %while.body.i.i21.i.i.i.i1318.preheader, %while.body.i.i21.i.i.i.i1318
  tail call void @llvm.amdgcn.s.sleep(i32 2)
  %1781 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i22.i.i.i.i1319 = getelementptr inbounds nuw i32, ptr %1781, i64 %idxprom.i23.i.i.i.i.i1307, !amdgpu.uniform !42
  %1782 = load atomic i32, ptr %arrayidx.i.i.i22.i.i.i.i1319 monotonic, align 4
  %1783 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %1782)
  %cmp.i.not.i.i23.i.i.i.i1320 = icmp eq i32 %1783, %port.sroa.37.0.i.i1311
  br i1 %cmp.i.not.i.i23.i.i.i.i1320, label %Flow3563, label %while.body.i.i21.i.i.i.i1318, !amdgpu.uniform !42

Flow3563:                                         ; preds = %while.body.i.i21.i.i.i.i1318
  br label %Flow3565, !amdgpu.uniform !42

Flow3566:                                         ; preds = %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1322, %while.cond.i.i.i.i1310
  %1784 = phi i64 [ %add.i.i.i.i1325, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1322 ], [ poison, %while.cond.i.i.i.i1310 ]
  %1785 = phi i32 [ %conv.i.i26.i.i.i.i1323, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1322 ], [ poison, %while.cond.i.i.i.i1310 ]
  %1786 = phi ptr addrspace(4) [ %scevgep2976, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1322 ], [ poison, %while.cond.i.i.i.i1310 ]
  %1787 = phi i64 [ %lsr.iv.next2981, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1322 ], [ poison, %while.cond.i.i.i.i1310 ]
  %1788 = phi i1 [ false, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1322 ], [ true, %while.cond.i.i.i.i1310 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %1778)
  %1789 = call i64 @llvm.amdgcn.if.break.i64(i1 %1788, i64 %phi.broken3934)
  %1790 = call i1 @llvm.amdgcn.loop.i64(i64 %1789)
  br i1 %1790, label %cond.false.i.i.i1332, label %while.cond.i.i.i.i1310

_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i1321: ; preds = %Flow3565
  fence acquire
  br i1 %cmp.i.i.i.i1313, label %if.then.i.i.i.i.i.i.i1326, label %Flow3562, !amdgpu.uniform !42

if.then.i.i.i.i.i.i.i1326:                        ; preds = %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i1321
  %sub.i.i.i.i.i.i.i1328 = sub nuw i64 sub (i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) getelementptr inbounds nuw (i8, ptr addrspace(4) @.str.9, i64 8) to ptr) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.9 to ptr) to i64)), %idx.0.i.i.i.i1312
  %spec.select.i.i.i36.i.i.i.i1329 = tail call i64 @llvm.umin.i64(i64 %sub.i.i.i.i.i.i.i1328, i64 64)
  %1791 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %1792 = and i64 %spec.select.i.i.i36.i.i.i.i1329, 15
  %1793 = sub i64 %spec.select.i.i.i36.i.i.i.i1329, %1792
  %1794 = icmp ne i64 %1793, 0
  br i1 %1794, label %loop-memcpy-expansion2404.preheader, label %Flow3561, !amdgpu.uniform !42

loop-memcpy-expansion2404.preheader:              ; preds = %if.then.i.i.i.i.i.i.i1326
  %arrayidx.i20.i35.i.i.i.i1330 = getelementptr %"struct.rpc::Buffer", ptr %1791, i64 %idxprom.i19.i.i.i.i.i1300
  %arrayidx.i22.i38.i.i.i.i1331 = getelementptr %"struct.rpc::Buffer", ptr %arrayidx.i20.i35.i.i.i.i1330, i64 %sh_prom.i.i.i.i
  br label %loop-memcpy-expansion2404, !amdgpu.uniform !42

Flow3561:                                         ; preds = %Flow3560, %if.then.i.i.i.i.i.i.i1326
  br label %loop-memcpy-residual-header2407, !amdgpu.uniform !42

loop-memcpy-expansion2404:                        ; preds = %loop-memcpy-expansion2404.preheader, %loop-memcpy-expansion2404
  %loop-index2405 = phi i64 [ %1797, %loop-memcpy-expansion2404 ], [ 0, %loop-memcpy-expansion2404.preheader ]
  %scevgep2977 = getelementptr i8, ptr addrspace(4) %lsr.iv2975, i64 %loop-index2405, !amdgpu.uniform !42
  %1795 = load <4 x i32>, ptr addrspace(4) %scevgep2977, align 1
  %1796 = getelementptr i8, ptr %arrayidx.i22.i38.i.i.i.i1331, i64 %loop-index2405
  store <4 x i32> %1795, ptr %1796, align 1
  %1797 = add i64 %loop-index2405, 16
  %1798 = icmp uge i64 %1797, %1793
  br i1 %1798, label %Flow3560, label %loop-memcpy-expansion2404, !amdgpu.uniform !42

Flow3558:                                         ; preds = %Flow3557, %loop-memcpy-residual-header2407
  br label %Flow3562, !amdgpu.uniform !42

loop-memcpy-residual2406:                         ; preds = %loop-memcpy-residual2406.preheader, %loop-memcpy-residual2406
  %residual-loop-index2408 = phi i64 [ %1800, %loop-memcpy-residual2406 ], [ 0, %loop-memcpy-residual2406.preheader ]
  %scevgep2986 = getelementptr i8, ptr addrspace(4) %scevgep2985, i64 %residual-loop-index2408, !amdgpu.uniform !42
  %1799 = load i8, ptr addrspace(4) %scevgep2986, align 1
  %scevgep2984 = getelementptr i8, ptr %scevgep2983, i64 %residual-loop-index2408
  store i8 %1799, ptr %scevgep2984, align 1
  %1800 = add i64 %residual-loop-index2408, 1
  %1801 = icmp uge i64 %1800, %1792
  br i1 %1801, label %Flow3557, label %loop-memcpy-residual2406, !amdgpu.uniform !42

Flow3557:                                         ; preds = %loop-memcpy-residual2406
  br label %Flow3558, !amdgpu.uniform !42

_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1322: ; preds = %Flow3562
  %conv.i.i26.i.i.i.i1323 = xor i32 %port.sroa.37.0.i.i1311, 1
  fence release
  %1802 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %arrayidx.i24.i29.i.i.i.i1324 = getelementptr inbounds nuw i32, ptr %1802, i64 %idxprom.i23.i.i.i.i.i1307
  store atomic i32 %conv.i.i26.i.i.i.i1323, ptr %arrayidx.i24.i29.i.i.i.i1324 monotonic, align 4
  %add.i.i.i.i1325 = add i64 %idx.0.i.i.i.i1312, 64
  %scevgep2976 = getelementptr i8, ptr addrspace(4) %lsr.iv2975, i64 64
  %lsr.iv.next2981 = add i64 %lsr.iv2980, -64
  br label %Flow3566, !amdgpu.uniform !42

cond.false.i.i.i1332:                             ; preds = %Flow3566
  %.lcssa4048 = phi i64 [ %1789, %Flow3566 ]
  %port.sroa.37.0.i.i1311.lcssa = phi i32 [ %port.sroa.37.0.i.i1311, %Flow3566 ]
  %.lcssa4047 = phi i32 [ %1775, %Flow3566 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa4048)
  %1803 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %.lcssa4047)
  %1804 = icmp ne i32 %1803, %port.sroa.37.0.i.i1311.lcssa
  %1805 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %1804)
  %1806 = extractvalue { i1, i64 } %1805, 0
  %1807 = extractvalue { i1, i64 } %1805, 1
  br i1 %1806, label %while.body.i.i.i.i1333.preheader, label %Flow3555

while.body.i.i.i.i1333.preheader:                 ; preds = %cond.false.i.i.i1332
  br label %while.body.i.i.i.i1333, !amdgpu.uniform !42

Flow3555:                                         ; preds = %Flow3553, %cond.false.i.i.i1332
  call void @llvm.amdgcn.end.cf.i64(i64 %1807)
  br label %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsINS_3cpp11string_viewEEERS1_T_.exit, !amdgpu.uniform !42

while.body.i.i.i.i1333:                           ; preds = %while.body.i.i.i.i1333.preheader, %while.body.i.i.i.i1333
  %phi.broken3935 = phi i64 [ 0, %while.body.i.i.i.i1333.preheader ], [ %1811, %while.body.i.i.i.i1333 ]
  tail call void @llvm.amdgcn.s.sleep(i32 2)
  %1808 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i8.i.i1334 = getelementptr inbounds nuw i32, ptr %1808, i64 %idxprom.i23.i.i.i.i.i1307, !amdgpu.uniform !42
  %1809 = load atomic i32, ptr %arrayidx.i.i.i8.i.i1334 monotonic, align 4
  %1810 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %1809)
  %cmp.i.not.i.i.i.i1335 = icmp eq i32 %1810, %port.sroa.37.0.i.i1311.lcssa
  %1811 = call i64 @llvm.amdgcn.if.break.i64(i1 %cmp.i.not.i.i.i.i1335, i64 %phi.broken3935)
  %1812 = call i1 @llvm.amdgcn.loop.i64(i64 %1811)
  br i1 %1812, label %Flow3553, label %while.body.i.i.i.i1333

Flow3553:                                         ; preds = %while.body.i.i.i.i1333
  %.lcssa4046 = phi i64 [ %1811, %while.body.i.i.i.i1333 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa4046)
  br label %Flow3555, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsINS_3cpp11string_viewEEERS1_T_.exit: ; preds = %Flow3555
  fence acquire
  tail call void @llvm.amdgcn.wave.barrier()
  fence syncscope("agent") release
  %1813 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %.lcssa4053, i1 true)
  %iszero.i.i.i.i.i13.i.i1337 = icmp ne i64 %.lcssa4053, 0
  %cmp2.i.i.i.i14.i.i1338 = icmp eq i64 %1813, %sh_prom.i.i.i.i
  %cmp.i.i.i.i15.i.i1339 = select i1 %iszero.i.i.i.i.i13.i.i1337, i1 %cmp2.i.i.i.i14.i.i1338, i1 false
  %1814 = and i32 %.lcssa4054, 31
  %conv4.i.i.i19.i.i1343 = zext i1 %cmp.i.i.i.i15.i.i1339 to i32
  %shl.i.i.i20.i.i1344 = shl nuw i32 %conv4.i.i.i19.i.i1343, %1814
  %xor.i.i.i21.i.i1345 = xor i32 %shl.i.i.i20.i.i1344, -1
  %1815 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2600, !amdgpu.uniform !42

1816:                                             ; preds = %ComputeEnd2601
  %div8.i.i.i16.i.i1340 = lshr i32 %.lcssa4054, 5
  %idxprom.i.i.i17.i.i1341 = zext nneg i32 %div8.i.i.i16.i.i1340 to i64
  %1817 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i17.i.i1341
  %sunkaddr3314 = getelementptr inbounds i8, ptr addrspace(1) %1817, i64 40
  %1818 = atomicrmw and ptr addrspace(1) %sunkaddr3314, i32 %.lcssa4045 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %1819, !amdgpu.uniform !42

1819:                                             ; preds = %1816, %ComputeEnd2601
  call void @llvm.amdgcn.end.cf.i64(i64 %3677)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %for.cond.i.i1.i1362, !amdgpu.uniform !42

for.cond.i.i1.i1362:                              ; preds = %1819, %Flow3552
  %1820 = phi i32 [ %1858, %Flow3552 ], [ poison, %1819 ]
  %1821 = phi i64 [ %1859, %Flow3552 ], [ poison, %1819 ]
  %1822 = phi i32 [ %1860, %Flow3552 ], [ poison, %1819 ]
  %retval.sroa.6.0.i.i.i1363 = phi i32 [ %1863, %Flow3552 ], [ undef, %1819 ]
  %retval.sroa.2.0.i.i.i1365 = phi i64 [ %1862, %Flow3552 ], [ undef, %1819 ]
  %retval.sroa.8.0.i.i.i1366 = phi i32 [ %1861, %Flow3552 ], [ undef, %1819 ]
  %index.0.i.i.i1367 = phi i32 [ %1865, %Flow3552 ], [ 0, %1819 ]
  %1823 = load i32, ptr addrspace(1) @__llvm_rpc_client, align 8, !tbaa !68
  %cmp.not.i.i.i1368 = icmp ult i32 %index.0.i.i.i1367, %1823
  %spec.store.select.i.i.i1369 = select i1 %cmp.not.i.i.i1368, i32 %index.0.i.i.i1367, i32 0
  %1824 = tail call noundef i64 @llvm.amdgcn.ballot.i64(i1 true)
  %1825 = and i32 %spec.store.select.i.i.i1369, 31
  %1826 = lshr i64 %1824, %sh_prom.i.i.i.i
  %1827 = trunc i64 %1826 to i32
  %conv4.i.i.i.i.i1373 = and i32 %1827, 1
  %shl.i.i.i.i.i1374 = shl nuw i32 %conv4.i.i.i.i.i1373, %1825
  %1828 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2604, !amdgpu.uniform !42

1829:                                             ; preds = %ComputeEnd2605
  %sunkaddr3315 = getelementptr inbounds i8, ptr addrspace(1) %3687, i64 40
  %1830 = atomicrmw or ptr addrspace(1) %sunkaddr3315, i32 %.lcssa4039 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %1831, !amdgpu.uniform !42

1831:                                             ; preds = %1829, %ComputeEnd2605
  %1832 = phi i32 [ %1830, %1829 ], [ poison, %ComputeEnd2605 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %3696)
  %1833 = call i32 @llvm.amdgcn.readfirstlane.i32(i32 %1832)
  %1834 = or i32 %1833, %.lcssa4040
  %shl5.i.i.i.i.i1375 = shl nuw i32 1, %1825
  %and.i.i.i.i.i1376 = and i32 %shl5.i.i.i.i.i1375, %1834
  %tobool3.i.i.i.i1377 = icmp ne i32 %and.i.i.i.i.i1376, 0
  %1835 = tail call i64 @llvm.amdgcn.ballot.i64(i1 %tobool3.i.i.i.i1377)
  %and.i.i.i.i.i.i1378 = and i64 %1835, %1824
  %cmp.i.not.i.i.i1379 = icmp ne i64 %1824, %and.i.i.i.i.i.i1378
  br i1 %cmp.i.not.i.i.i1379, label %if.then.i.i.i.i1380, label %Flow3551, !amdgpu.uniform !42

if.then.i.i.i.i1380:                              ; preds = %1831
  fence syncscope("agent") acquire
  %1836 = load <2 x i64>, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !75
  %1837 = extractelement <2 x i64> %1836, i32 0
  %1838 = inttoptr i64 %1837 to ptr
  %1839 = extractelement <2 x i64> %1836, i32 1
  %1840 = inttoptr i64 %1839 to ptr
  %idxprom.i.i.i.i1381 = zext i32 %spec.store.select.i.i.i1369 to i64
  %arrayidx.i.i.i.i1382 = getelementptr inbounds nuw i32, ptr %1838, i64 %idxprom.i.i.i.i1381, !amdgpu.uniform !42
  %1841 = load atomic i32, ptr %arrayidx.i.i.i.i1382 monotonic, align 4
  %1842 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %1841)
  %arrayidx.i45.i.i.i1383 = getelementptr inbounds nuw i32, ptr %1840, i64 %idxprom.i.i.i.i1381, !amdgpu.uniform !42
  %1843 = load atomic i32, ptr %arrayidx.i45.i.i.i1383 monotonic, align 4
  %1844 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %1843)
  %cmp.i46.not.i.i.i1384 = icmp ne i32 %1842, %1844
  br i1 %cmp.i46.not.i.i.i1384, label %if.then12.i.i.i1385, label %Flow3549, !amdgpu.uniform !42

if.then12.i.i.i1385:                              ; preds = %if.then.i.i.i.i1380
  fence syncscope("agent") release
  %1845 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %1824, i1 true)
  %iszero.i.i.i.i.i.i.i1386 = icmp ne i64 %1824, 0
  %cmp2.i.i.i.i.i.i1387 = icmp eq i64 %1845, %sh_prom.i.i.i.i
  %cmp.i.i.i.i.i.i1388 = select i1 %iszero.i.i.i.i.i.i.i1386, i1 %cmp2.i.i.i.i.i.i1387, i1 false
  %conv4.i.i51.i.i.i1389 = zext i1 %cmp.i.i.i.i.i.i1388 to i32
  %shl.i.i52.i.i.i1390 = shl nuw i32 %conv4.i.i51.i.i.i1389, %1825
  %xor.i.i.i.i.i1391 = xor i32 %shl.i.i52.i.i.i1390, -1
  %1846 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2609, !amdgpu.uniform !42

1847:                                             ; preds = %ComputeEnd2610
  %sunkaddr3316 = getelementptr inbounds i8, ptr addrspace(1) %3687, i64 40
  %1848 = atomicrmw and ptr addrspace(1) %sunkaddr3316, i32 %.lcssa4041 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %1849, !amdgpu.uniform !42

1849:                                             ; preds = %1847, %ComputeEnd2610
  call void @llvm.amdgcn.end.cf.i64(i64 %3714)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %Flow3549, !amdgpu.uniform !42

if.end14.i.i.i1451:                               ; preds = %Flow3549
  %1850 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %1824, i1 true)
  %iszero.i.i.i.i.i.i1452 = icmp ne i64 %1824, 0
  %cmp2.i.i.i.i.i1453 = icmp eq i64 %1850, %sh_prom.i.i.i.i
  %cmp.i.i.i.i.i1454 = select i1 %iszero.i.i.i.i.i.i1452, i1 %cmp2.i.i.i.i.i1453, i1 false
  %1851 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.i.i.i.i.i1454)
  %1852 = extractvalue { i1, i64 } %1851, 0
  %1853 = extractvalue { i1, i64 } %1851, 1
  br i1 %1852, label %if.then16.i.i.i1456, label %if.end22.i.i.i1455

if.then16.i.i.i1456:                              ; preds = %if.end14.i.i.i1451
  %1854 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !76
  %opcode.i.i.i1457 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %1854, i64 %idxprom.i.i.i.i1381, i32 1
  store i32 1660944387, ptr %opcode.i.i.i1457, align 8, !tbaa !77
  %arrayidx21.i.i.i1458 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %1854, i64 %idxprom.i.i.i.i1381
  store i64 %1824, ptr %arrayidx21.i.i.i1458, align 8, !tbaa !80
  br label %if.end22.i.i.i1455, !amdgpu.uniform !42

Flow3550:                                         ; preds = %if.end22.i.i.i1455, %Flow3549
  %1855 = phi i32 [ %1842, %if.end22.i.i.i1455 ], [ %retval.sroa.8.0.i.i.i1366, %Flow3549 ]
  %1856 = phi i64 [ %1824, %if.end22.i.i.i1455 ], [ %retval.sroa.2.0.i.i.i1365, %Flow3549 ]
  %1857 = phi i32 [ %spec.store.select.i.i.i1369, %if.end22.i.i.i1455 ], [ %retval.sroa.6.0.i.i.i1363, %Flow3549 ]
  br label %cleanup26.i.i.i1392, !amdgpu.uniform !42

if.end22.i.i.i1455:                               ; preds = %if.then16.i.i.i1456, %if.end14.i.i.i1451
  call void @llvm.amdgcn.end.cf.i64(i64 %1853)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %Flow3550, !amdgpu.uniform !42

Flow3551:                                         ; preds = %cleanup26.i.i.i1392, %1831
  %1858 = phi i32 [ %1857, %cleanup26.i.i.i1392 ], [ %1820, %1831 ]
  %1859 = phi i64 [ %1856, %cleanup26.i.i.i1392 ], [ %1821, %1831 ]
  %1860 = phi i32 [ %1855, %cleanup26.i.i.i1392 ], [ %1822, %1831 ]
  %1861 = phi i32 [ %1855, %cleanup26.i.i.i1392 ], [ %retval.sroa.8.0.i.i.i1366, %1831 ]
  %1862 = phi i64 [ %1856, %cleanup26.i.i.i1392 ], [ %retval.sroa.2.0.i.i.i1365, %1831 ]
  %1863 = phi i32 [ %1857, %cleanup26.i.i.i1392 ], [ %retval.sroa.6.0.i.i.i1363, %1831 ]
  %1864 = phi i1 [ %cmp.i46.not.i.i.i1384, %cleanup26.i.i.i1392 ], [ true, %1831 ]
  br i1 %1864, label %for.inc.i.i.i1397, label %Flow3552, !amdgpu.uniform !42

cleanup26.i.i.i1392:                              ; preds = %Flow3550
  br label %Flow3551, !amdgpu.uniform !42

for.inc.i.i.i1397:                                ; preds = %Flow3551
  %inc.i.i.i1402 = add i32 %spec.store.select.i.i.i1369, 1
  br label %Flow3552, !amdgpu.uniform !42

Flow3552:                                         ; preds = %for.inc.i.i.i1397, %Flow3551
  %1865 = phi i32 [ %inc.i.i.i1402, %for.inc.i.i.i1397 ], [ poison, %Flow3551 ]
  %1866 = phi i1 [ false, %for.inc.i.i.i1397 ], [ true, %Flow3551 ]
  br i1 %1866, label %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1403, label %for.cond.i.i1.i1362, !amdgpu.uniform !42

_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1403: ; preds = %Flow3552
  %.lcssa4044 = phi i32 [ %1858, %Flow3552 ]
  %.lcssa4043 = phi i64 [ %1859, %Flow3552 ]
  %.lcssa4042 = phi i32 [ %1860, %Flow3552 ]
  fence acquire
  %1867 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %mul.i.i.i.i.i.i1404 = shl i32 %.lcssa4044, 6
  %idxprom.i19.i.i.i.i.i1405 = zext i32 %mul.i.i.i.i.i.i1404 to i64
  %arrayidx.i20.i.i.i.i.i1406 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %1867, i64 %idxprom.i19.i.i.i.i.i1405
  %arrayidx.i22.i.i.i.i.i1407 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %arrayidx.i20.i.i.i.i.i1406, i64 %sh_prom.i.i.i.i
  store i64 sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.37.14 to ptr), i64 2) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.37.14 to ptr) to i64)), ptr %arrayidx.i22.i.i.i.i.i1407, align 8, !tbaa !82
  %spec.select.i.i.i.i.i.i.i1408 = tail call i64 @llvm.umin.i64(i64 sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.37.14 to ptr), i64 2) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.37.14 to ptr) to i64)), i64 56)
  %1868 = and i64 %spec.select.i.i.i.i.i.i.i1408, 15
  %1869 = sub i64 %spec.select.i.i.i.i.i.i.i1408, %1868
  %1870 = icmp ne i64 %1869, 0
  br i1 %1870, label %loop-memcpy-expansion2410.preheader, label %Flow3548, !amdgpu.uniform !42

loop-memcpy-expansion2410.preheader:              ; preds = %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1403
  %1871 = add nuw nsw i64 %75, 8
  %scevgep2972 = getelementptr i8, ptr %1867, i64 %1871
  %1872 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i1405, 6
  %scevgep2973 = getelementptr i8, ptr %scevgep2972, i64 %1872
  br label %loop-memcpy-expansion2410, !amdgpu.uniform !42

Flow3548:                                         ; preds = %Flow3547, %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1403
  br label %loop-memcpy-residual-header2413, !amdgpu.uniform !42

loop-memcpy-expansion2410:                        ; preds = %loop-memcpy-expansion2410.preheader, %loop-memcpy-expansion2410
  %loop-index2411 = phi i64 [ %1875, %loop-memcpy-expansion2410 ], [ 0, %loop-memcpy-expansion2410.preheader ]
  %1873 = getelementptr i8, ptr addrspace(4) @.str.37.14, i64 %loop-index2411, !amdgpu.uniform !42
  %1874 = load <4 x i32>, ptr addrspace(4) %1873, align 1
  %scevgep2974 = getelementptr i8, ptr %scevgep2973, i64 %loop-index2411
  store <4 x i32> %1874, ptr %scevgep2974, align 1
  %1875 = add i64 %loop-index2411, 16
  %1876 = icmp uge i64 %1875, %1869
  br i1 %1876, label %Flow3547, label %loop-memcpy-expansion2410, !amdgpu.uniform !42

Flow3545:                                         ; preds = %Flow3542, %Flow3544
  %1877 = phi i64 [ %2915, %Flow3542 ], [ %2912, %Flow3544 ]
  br label %post-loop-memcpy-expansion2409, !amdgpu.uniform !42

loop-memcpy-residual2412:                         ; preds = %loop-memcpy-residual2412.preheader, %loop-memcpy-residual2412
  %residual-loop-index2414 = phi i64 [ %1879, %loop-memcpy-residual2412 ], [ 0, %loop-memcpy-residual2412.preheader ]
  %scevgep2971 = getelementptr i8, ptr addrspace(4) @.str.37.14, i64 %residual-loop-index2414, !amdgpu.uniform !42
  %1878 = load i8, ptr addrspace(4) %scevgep2971, align 1
  %scevgep2970 = getelementptr i8, ptr %scevgep2969, i64 %residual-loop-index2414
  store i8 %1878, ptr %scevgep2970, align 1
  %1879 = add i64 %residual-loop-index2414, 1
  %1880 = icmp uge i64 %1879, %1868
  br i1 %1880, label %Flow3542, label %loop-memcpy-residual2412, !amdgpu.uniform !42

Flow3542:                                         ; preds = %loop-memcpy-residual2412
  br label %Flow3545, !amdgpu.uniform !42

post-loop-memcpy-expansion2409:                   ; preds = %Flow3545
  %tobool.not.i.i.i.i.i.i1410 = icmp eq i32 %.lcssa4042, 0
  %conv.i.i.i.i.i.i1411 = zext i1 %tobool.not.i.i.i.i.i.i1410 to i32
  fence release
  %1881 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %idxprom.i23.i.i.i.i.i1412 = zext i32 %.lcssa4044 to i64
  %arrayidx.i24.i.i.i.i.i1413 = getelementptr inbounds nuw i32, ptr %1881, i64 %idxprom.i23.i.i.i.i.i1412
  store atomic i32 %conv.i.i.i.i.i.i1411, ptr %arrayidx.i24.i.i.i.i.i1413 monotonic, align 4
  %1882 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !84
  %arrayidx.i.i5.i.i1414 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %1882, i64 %idxprom.i23.i.i.i.i.i1412, !amdgpu.uniform !42
  %1883 = load i64, ptr %arrayidx.i.i5.i.i1414, align 8, !tbaa !80
  br label %while.cond.i.i.i.i1415, !amdgpu.uniform !42

while.cond.i.i.i.i1415:                           ; preds = %post-loop-memcpy-expansion2409, %Flow3540
  %phi.broken3936 = phi i64 [ 0, %post-loop-memcpy-expansion2409 ], [ %1900, %Flow3540 ]
  %lsr.iv2961 = phi i64 [ %1898, %Flow3540 ], [ -54, %post-loop-memcpy-expansion2409 ]
  %lsr.iv2956 = phi ptr addrspace(4) [ %1897, %Flow3540 ], [ getelementptr (i8, ptr addrspace(4) @.str.37.14, i64 56), %post-loop-memcpy-expansion2409 ]
  %port.sroa.37.0.i.i1416 = phi i32 [ %1896, %Flow3540 ], [ %conv.i.i.i.i.i.i1411, %post-loop-memcpy-expansion2409 ]
  %idx.0.i.i.i.i1417 = phi i64 [ %1895, %Flow3540 ], [ 56, %post-loop-memcpy-expansion2409 ]
  %cmp.i.i.i.i1418 = icmp ult i64 %idx.0.i.i.i.i1417, sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.37.14 to ptr), i64 2) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.37.14 to ptr) to i64))
  %1884 = tail call i64 @llvm.amdgcn.ballot.i64(i1 %cmp.i.i.i.i1418)
  %and.i.i.i.i6.i.i1419 = and i64 %1884, %1883
  %tobool.not.i.i.i.i1420 = icmp ne i64 %and.i.i.i.i6.i.i1419, 0
  %1885 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i22.i.i.i1421 = getelementptr inbounds nuw i32, ptr %1885, i64 %idxprom.i23.i.i.i.i.i1412, !amdgpu.uniform !42
  %1886 = load atomic i32, ptr %arrayidx.i22.i.i.i1421 monotonic, align 4
  %1887 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %tobool.not.i.i.i.i1420)
  %1888 = extractvalue { i1, i64 } %1887, 0
  %1889 = extractvalue { i1, i64 } %1887, 1
  br i1 %1888, label %cond.false.i8.i.i.i.i1422, label %Flow3540

cond.false.i8.i.i.i.i1422:                        ; preds = %while.cond.i.i.i.i1415
  %1890 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %1886)
  %1891 = icmp ne i32 %1890, %port.sroa.37.0.i.i1416
  br i1 %1891, label %while.body.i.i21.i.i.i.i1423.preheader, label %Flow3539, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i1423.preheader:           ; preds = %cond.false.i8.i.i.i.i1422
  br label %while.body.i.i21.i.i.i.i1423, !amdgpu.uniform !42

Flow3539:                                         ; preds = %Flow3537, %cond.false.i8.i.i.i.i1422
  br label %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i1426, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i1423:                     ; preds = %while.body.i.i21.i.i.i.i1423.preheader, %while.body.i.i21.i.i.i.i1423
  tail call void @llvm.amdgcn.s.sleep(i32 2)
  %1892 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i22.i.i.i.i1424 = getelementptr inbounds nuw i32, ptr %1892, i64 %idxprom.i23.i.i.i.i.i1412, !amdgpu.uniform !42
  %1893 = load atomic i32, ptr %arrayidx.i.i.i22.i.i.i.i1424 monotonic, align 4
  %1894 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %1893)
  %cmp.i.not.i.i23.i.i.i.i1425 = icmp eq i32 %1894, %port.sroa.37.0.i.i1416
  br i1 %cmp.i.not.i.i23.i.i.i.i1425, label %Flow3537, label %while.body.i.i21.i.i.i.i1423, !amdgpu.uniform !42

Flow3537:                                         ; preds = %while.body.i.i21.i.i.i.i1423
  br label %Flow3539, !amdgpu.uniform !42

Flow3540:                                         ; preds = %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1427, %while.cond.i.i.i.i1415
  %1895 = phi i64 [ %add.i.i.i.i1430, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1427 ], [ poison, %while.cond.i.i.i.i1415 ]
  %1896 = phi i32 [ %conv.i.i26.i.i.i.i1428, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1427 ], [ poison, %while.cond.i.i.i.i1415 ]
  %1897 = phi ptr addrspace(4) [ %scevgep2957, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1427 ], [ poison, %while.cond.i.i.i.i1415 ]
  %1898 = phi i64 [ %lsr.iv.next2962, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1427 ], [ poison, %while.cond.i.i.i.i1415 ]
  %1899 = phi i1 [ false, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1427 ], [ true, %while.cond.i.i.i.i1415 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %1889)
  %1900 = call i64 @llvm.amdgcn.if.break.i64(i1 %1899, i64 %phi.broken3936)
  %1901 = call i1 @llvm.amdgcn.loop.i64(i64 %1900)
  br i1 %1901, label %cond.false.i.i.i1437, label %while.cond.i.i.i.i1415

_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i1426: ; preds = %Flow3539
  fence acquire
  br i1 %cmp.i.i.i.i1418, label %if.then.i.i.i.i.i.i.i1431, label %Flow3536, !amdgpu.uniform !42

if.then.i.i.i.i.i.i.i1431:                        ; preds = %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i1426
  %sub.i.i.i.i.i.i.i1433 = sub nuw i64 sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.37.14 to ptr), i64 2) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.37.14 to ptr) to i64)), %idx.0.i.i.i.i1417
  %spec.select.i.i.i36.i.i.i.i1434 = tail call i64 @llvm.umin.i64(i64 %sub.i.i.i.i.i.i.i1433, i64 64)
  %1902 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %1903 = and i64 %spec.select.i.i.i36.i.i.i.i1434, 15
  %1904 = sub i64 %spec.select.i.i.i36.i.i.i.i1434, %1903
  %1905 = icmp ne i64 %1904, 0
  br i1 %1905, label %loop-memcpy-expansion2416.preheader, label %Flow3535, !amdgpu.uniform !42

loop-memcpy-expansion2416.preheader:              ; preds = %if.then.i.i.i.i.i.i.i1431
  %arrayidx.i20.i35.i.i.i.i1435 = getelementptr %"struct.rpc::Buffer", ptr %1902, i64 %idxprom.i19.i.i.i.i.i1405
  %arrayidx.i22.i38.i.i.i.i1436 = getelementptr %"struct.rpc::Buffer", ptr %arrayidx.i20.i35.i.i.i.i1435, i64 %sh_prom.i.i.i.i
  br label %loop-memcpy-expansion2416, !amdgpu.uniform !42

Flow3535:                                         ; preds = %Flow3534, %if.then.i.i.i.i.i.i.i1431
  br label %loop-memcpy-residual-header2419, !amdgpu.uniform !42

loop-memcpy-expansion2416:                        ; preds = %loop-memcpy-expansion2416.preheader, %loop-memcpy-expansion2416
  %loop-index2417 = phi i64 [ %1908, %loop-memcpy-expansion2416 ], [ 0, %loop-memcpy-expansion2416.preheader ]
  %scevgep2958 = getelementptr i8, ptr addrspace(4) %lsr.iv2956, i64 %loop-index2417, !amdgpu.uniform !42
  %1906 = load <4 x i32>, ptr addrspace(4) %scevgep2958, align 1
  %1907 = getelementptr i8, ptr %arrayidx.i22.i38.i.i.i.i1436, i64 %loop-index2417
  store <4 x i32> %1906, ptr %1907, align 1
  %1908 = add i64 %loop-index2417, 16
  %1909 = icmp uge i64 %1908, %1904
  br i1 %1909, label %Flow3534, label %loop-memcpy-expansion2416, !amdgpu.uniform !42

Flow3532:                                         ; preds = %Flow3531, %loop-memcpy-residual-header2419
  br label %Flow3536, !amdgpu.uniform !42

loop-memcpy-residual2418:                         ; preds = %loop-memcpy-residual2418.preheader, %loop-memcpy-residual2418
  %residual-loop-index2420 = phi i64 [ %1911, %loop-memcpy-residual2418 ], [ 0, %loop-memcpy-residual2418.preheader ]
  %scevgep2967 = getelementptr i8, ptr addrspace(4) %scevgep2966, i64 %residual-loop-index2420, !amdgpu.uniform !42
  %1910 = load i8, ptr addrspace(4) %scevgep2967, align 1
  %scevgep2965 = getelementptr i8, ptr %scevgep2964, i64 %residual-loop-index2420
  store i8 %1910, ptr %scevgep2965, align 1
  %1911 = add i64 %residual-loop-index2420, 1
  %1912 = icmp uge i64 %1911, %1903
  br i1 %1912, label %Flow3531, label %loop-memcpy-residual2418, !amdgpu.uniform !42

Flow3531:                                         ; preds = %loop-memcpy-residual2418
  br label %Flow3532, !amdgpu.uniform !42

_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1427: ; preds = %Flow3536
  %conv.i.i26.i.i.i.i1428 = xor i32 %port.sroa.37.0.i.i1416, 1
  fence release
  %1913 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %arrayidx.i24.i29.i.i.i.i1429 = getelementptr inbounds nuw i32, ptr %1913, i64 %idxprom.i23.i.i.i.i.i1412
  store atomic i32 %conv.i.i26.i.i.i.i1428, ptr %arrayidx.i24.i29.i.i.i.i1429 monotonic, align 4
  %add.i.i.i.i1430 = add i64 %idx.0.i.i.i.i1417, 64
  %scevgep2957 = getelementptr i8, ptr addrspace(4) %lsr.iv2956, i64 64
  %lsr.iv.next2962 = add i64 %lsr.iv2961, -64
  br label %Flow3540, !amdgpu.uniform !42

cond.false.i.i.i1437:                             ; preds = %Flow3540
  %.lcssa4038 = phi i64 [ %1900, %Flow3540 ]
  %port.sroa.37.0.i.i1416.lcssa = phi i32 [ %port.sroa.37.0.i.i1416, %Flow3540 ]
  %.lcssa4037 = phi i32 [ %1886, %Flow3540 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa4038)
  %1914 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %.lcssa4037)
  %1915 = icmp ne i32 %1914, %port.sroa.37.0.i.i1416.lcssa
  %1916 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %1915)
  %1917 = extractvalue { i1, i64 } %1916, 0
  %1918 = extractvalue { i1, i64 } %1916, 1
  br i1 %1917, label %while.body.i.i.i.i1438.preheader, label %Flow3529

while.body.i.i.i.i1438.preheader:                 ; preds = %cond.false.i.i.i1437
  br label %while.body.i.i.i.i1438, !amdgpu.uniform !42

Flow3529:                                         ; preds = %Flow3527, %cond.false.i.i.i1437
  call void @llvm.amdgcn.end.cf.i64(i64 %1918)
  br label %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit1459, !amdgpu.uniform !42

while.body.i.i.i.i1438:                           ; preds = %while.body.i.i.i.i1438.preheader, %while.body.i.i.i.i1438
  %phi.broken3937 = phi i64 [ 0, %while.body.i.i.i.i1438.preheader ], [ %1922, %while.body.i.i.i.i1438 ]
  tail call void @llvm.amdgcn.s.sleep(i32 2)
  %1919 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i8.i.i1439 = getelementptr inbounds nuw i32, ptr %1919, i64 %idxprom.i23.i.i.i.i.i1412, !amdgpu.uniform !42
  %1920 = load atomic i32, ptr %arrayidx.i.i.i8.i.i1439 monotonic, align 4
  %1921 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %1920)
  %cmp.i.not.i.i.i.i1440 = icmp eq i32 %1921, %port.sroa.37.0.i.i1416.lcssa
  %1922 = call i64 @llvm.amdgcn.if.break.i64(i1 %cmp.i.not.i.i.i.i1440, i64 %phi.broken3937)
  %1923 = call i1 @llvm.amdgcn.loop.i64(i64 %1922)
  br i1 %1923, label %Flow3527, label %while.body.i.i.i.i1438

Flow3527:                                         ; preds = %while.body.i.i.i.i1438
  %.lcssa4036 = phi i64 [ %1922, %while.body.i.i.i.i1438 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa4036)
  br label %Flow3529, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit1459: ; preds = %Flow3529
  fence acquire
  tail call void @llvm.amdgcn.wave.barrier()
  fence syncscope("agent") release
  %1924 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %.lcssa4043, i1 true)
  %iszero.i.i.i.i.i13.i.i1442 = icmp ne i64 %.lcssa4043, 0
  %cmp2.i.i.i.i14.i.i1443 = icmp eq i64 %1924, %sh_prom.i.i.i.i
  %cmp.i.i.i.i15.i.i1444 = select i1 %iszero.i.i.i.i.i13.i.i1442, i1 %cmp2.i.i.i.i14.i.i1443, i1 false
  %1925 = and i32 %.lcssa4044, 31
  %conv4.i.i.i19.i.i1448 = zext i1 %cmp.i.i.i.i15.i.i1444 to i32
  %shl.i.i.i20.i.i1449 = shl nuw i32 %conv4.i.i.i19.i.i1448, %1925
  %xor.i.i.i21.i.i1450 = xor i32 %shl.i.i.i20.i.i1449, -1
  %1926 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2613, !amdgpu.uniform !42

1927:                                             ; preds = %ComputeEnd2614
  %div8.i.i.i16.i.i1445 = lshr i32 %.lcssa4044, 5
  %idxprom.i.i.i17.i.i1446 = zext nneg i32 %div8.i.i.i16.i.i1445 to i64
  %1928 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i17.i.i1446
  %sunkaddr3317 = getelementptr inbounds i8, ptr addrspace(1) %1928, i64 40
  %1929 = atomicrmw and ptr addrspace(1) %sunkaddr3317, i32 %.lcssa4035 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %1930, !amdgpu.uniform !42

1930:                                             ; preds = %1927, %ComputeEnd2614
  call void @llvm.amdgcn.end.cf.i64(i64 %3731)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %for.cond.i.i.i1460, !amdgpu.uniform !42

for.cond.i.i.i1460:                               ; preds = %1930, %for.cond.i.i.i1460
  %phi.broken3938 = phi i64 [ 0, %1930 ], [ %1933, %for.cond.i.i.i1460 ]
  %lsr.iv2942 = phi i64 [ %lsr.iv.next2943, %for.cond.i.i.i1460 ], [ -1, %1930 ]
  %scevgep2954 = getelementptr i8, ptr %RHSStr, i64 %lsr.iv2942
  %scevgep2955 = getelementptr i8, ptr %scevgep2954, i64 1
  %1931 = addrspacecast ptr %scevgep2955 to ptr addrspace(4)
  %1932 = load i8, ptr addrspace(4) %1931, align 1, !tbaa !57, !noalias.addrspace !101
  %cmp.i.i.i1462 = icmp eq i8 %1932, 0
  %lsr.iv.next2943 = add i64 %lsr.iv2942, 1
  %1933 = call i64 @llvm.amdgcn.if.break.i64(i1 %cmp.i.i.i1462, i64 %phi.broken3938)
  %1934 = call i1 @llvm.amdgcn.loop.i64(i64 %1933)
  br i1 %1934, label %for.cond.i.i1.i1468.preheader, label %for.cond.i.i.i1460

for.cond.i.i1.i1468.preheader:                    ; preds = %for.cond.i.i.i1460
  %lsr.iv.next2943.lcssa = phi i64 [ %lsr.iv.next2943, %for.cond.i.i.i1460 ]
  %.lcssa4034 = phi i64 [ %1933, %for.cond.i.i.i1460 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa4034)
  br label %for.cond.i.i1.i1468, !amdgpu.uniform !42

for.cond.i.i1.i1468:                              ; preds = %for.cond.i.i1.i1468.preheader, %Flow3526
  %1935 = phi i32 [ %1973, %Flow3526 ], [ poison, %for.cond.i.i1.i1468.preheader ]
  %1936 = phi i64 [ %1974, %Flow3526 ], [ poison, %for.cond.i.i1.i1468.preheader ]
  %1937 = phi i32 [ %1975, %Flow3526 ], [ poison, %for.cond.i.i1.i1468.preheader ]
  %retval.sroa.6.0.i.i.i1469 = phi i32 [ %1978, %Flow3526 ], [ undef, %for.cond.i.i1.i1468.preheader ]
  %retval.sroa.2.0.i.i.i1471 = phi i64 [ %1977, %Flow3526 ], [ undef, %for.cond.i.i1.i1468.preheader ]
  %retval.sroa.8.0.i.i.i1472 = phi i32 [ %1976, %Flow3526 ], [ undef, %for.cond.i.i1.i1468.preheader ]
  %index.0.i.i.i1473 = phi i32 [ %1980, %Flow3526 ], [ 0, %for.cond.i.i1.i1468.preheader ]
  %1938 = load i32, ptr addrspace(1) @__llvm_rpc_client, align 8, !tbaa !68
  %cmp.not.i.i.i1474 = icmp ult i32 %index.0.i.i.i1473, %1938
  %spec.store.select.i.i.i1475 = select i1 %cmp.not.i.i.i1474, i32 %index.0.i.i.i1473, i32 0
  %1939 = tail call noundef i64 @llvm.amdgcn.ballot.i64(i1 true)
  %1940 = and i32 %spec.store.select.i.i.i1475, 31
  %1941 = lshr i64 %1939, %sh_prom.i.i.i.i
  %1942 = trunc i64 %1941 to i32
  %conv4.i.i.i.i.i1479 = and i32 %1942, 1
  %shl.i.i.i.i.i1480 = shl nuw i32 %conv4.i.i.i.i.i1479, %1940
  %1943 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2617, !amdgpu.uniform !42

1944:                                             ; preds = %ComputeEnd2618
  %sunkaddr3318 = getelementptr inbounds i8, ptr addrspace(1) %3741, i64 40
  %1945 = atomicrmw or ptr addrspace(1) %sunkaddr3318, i32 %.lcssa4028 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %1946, !amdgpu.uniform !42

1946:                                             ; preds = %1944, %ComputeEnd2618
  %1947 = phi i32 [ %1945, %1944 ], [ poison, %ComputeEnd2618 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %3750)
  %1948 = call i32 @llvm.amdgcn.readfirstlane.i32(i32 %1947)
  %1949 = or i32 %1948, %.lcssa4029
  %shl5.i.i.i.i.i1481 = shl nuw i32 1, %1940
  %and.i.i.i.i.i1482 = and i32 %shl5.i.i.i.i.i1481, %1949
  %tobool3.i.i.i.i1483 = icmp ne i32 %and.i.i.i.i.i1482, 0
  %1950 = tail call i64 @llvm.amdgcn.ballot.i64(i1 %tobool3.i.i.i.i1483)
  %and.i.i.i.i.i.i1484 = and i64 %1950, %1939
  %cmp.i.not.i.i.i1485 = icmp ne i64 %1939, %and.i.i.i.i.i.i1484
  br i1 %cmp.i.not.i.i.i1485, label %if.then.i.i.i.i1486, label %Flow3525, !amdgpu.uniform !42

if.then.i.i.i.i1486:                              ; preds = %1946
  fence syncscope("agent") acquire
  %1951 = load <2 x i64>, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !75
  %1952 = extractelement <2 x i64> %1951, i32 0
  %1953 = inttoptr i64 %1952 to ptr
  %1954 = extractelement <2 x i64> %1951, i32 1
  %1955 = inttoptr i64 %1954 to ptr
  %idxprom.i.i.i.i1487 = zext i32 %spec.store.select.i.i.i1475 to i64
  %arrayidx.i.i.i.i1488 = getelementptr inbounds nuw i32, ptr %1953, i64 %idxprom.i.i.i.i1487, !amdgpu.uniform !42
  %1956 = load atomic i32, ptr %arrayidx.i.i.i.i1488 monotonic, align 4
  %1957 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %1956)
  %arrayidx.i45.i.i.i1489 = getelementptr inbounds nuw i32, ptr %1955, i64 %idxprom.i.i.i.i1487, !amdgpu.uniform !42
  %1958 = load atomic i32, ptr %arrayidx.i45.i.i.i1489 monotonic, align 4
  %1959 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %1958)
  %cmp.i46.not.i.i.i1490 = icmp ne i32 %1957, %1959
  br i1 %cmp.i46.not.i.i.i1490, label %if.then12.i.i.i1491, label %Flow3523, !amdgpu.uniform !42

if.then12.i.i.i1491:                              ; preds = %if.then.i.i.i.i1486
  fence syncscope("agent") release
  %1960 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %1939, i1 true)
  %iszero.i.i.i.i.i.i.i1492 = icmp ne i64 %1939, 0
  %cmp2.i.i.i.i.i.i1493 = icmp eq i64 %1960, %sh_prom.i.i.i.i
  %cmp.i.i.i.i.i.i1494 = select i1 %iszero.i.i.i.i.i.i.i1492, i1 %cmp2.i.i.i.i.i.i1493, i1 false
  %conv4.i.i51.i.i.i1495 = zext i1 %cmp.i.i.i.i.i.i1494 to i32
  %shl.i.i52.i.i.i1496 = shl nuw i32 %conv4.i.i51.i.i.i1495, %1940
  %xor.i.i.i.i.i1497 = xor i32 %shl.i.i52.i.i.i1496, -1
  %1961 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2622, !amdgpu.uniform !42

1962:                                             ; preds = %ComputeEnd2623
  %sunkaddr3319 = getelementptr inbounds i8, ptr addrspace(1) %3741, i64 40
  %1963 = atomicrmw and ptr addrspace(1) %sunkaddr3319, i32 %.lcssa4030 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %1964, !amdgpu.uniform !42

1964:                                             ; preds = %1962, %ComputeEnd2623
  call void @llvm.amdgcn.end.cf.i64(i64 %3768)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %Flow3523, !amdgpu.uniform !42

if.end14.i.i.i1557:                               ; preds = %Flow3523
  %1965 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %1939, i1 true)
  %iszero.i.i.i.i.i.i1558 = icmp ne i64 %1939, 0
  %cmp2.i.i.i.i.i1559 = icmp eq i64 %1965, %sh_prom.i.i.i.i
  %cmp.i.i.i.i.i1560 = select i1 %iszero.i.i.i.i.i.i1558, i1 %cmp2.i.i.i.i.i1559, i1 false
  %1966 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.i.i.i.i.i1560)
  %1967 = extractvalue { i1, i64 } %1966, 0
  %1968 = extractvalue { i1, i64 } %1966, 1
  br i1 %1967, label %if.then16.i.i.i1562, label %if.end22.i.i.i1561

if.then16.i.i.i1562:                              ; preds = %if.end14.i.i.i1557
  %1969 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !76
  %opcode.i.i.i1563 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %1969, i64 %idxprom.i.i.i.i1487, i32 1
  store i32 1660944387, ptr %opcode.i.i.i1563, align 8, !tbaa !77
  %arrayidx21.i.i.i1564 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %1969, i64 %idxprom.i.i.i.i1487
  store i64 %1939, ptr %arrayidx21.i.i.i1564, align 8, !tbaa !80
  br label %if.end22.i.i.i1561, !amdgpu.uniform !42

Flow3524:                                         ; preds = %if.end22.i.i.i1561, %Flow3523
  %1970 = phi i32 [ %1957, %if.end22.i.i.i1561 ], [ %retval.sroa.8.0.i.i.i1472, %Flow3523 ]
  %1971 = phi i64 [ %1939, %if.end22.i.i.i1561 ], [ %retval.sroa.2.0.i.i.i1471, %Flow3523 ]
  %1972 = phi i32 [ %spec.store.select.i.i.i1475, %if.end22.i.i.i1561 ], [ %retval.sroa.6.0.i.i.i1469, %Flow3523 ]
  br label %cleanup26.i.i.i1498, !amdgpu.uniform !42

if.end22.i.i.i1561:                               ; preds = %if.then16.i.i.i1562, %if.end14.i.i.i1557
  call void @llvm.amdgcn.end.cf.i64(i64 %1968)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %Flow3524, !amdgpu.uniform !42

Flow3525:                                         ; preds = %cleanup26.i.i.i1498, %1946
  %1973 = phi i32 [ %1972, %cleanup26.i.i.i1498 ], [ %1935, %1946 ]
  %1974 = phi i64 [ %1971, %cleanup26.i.i.i1498 ], [ %1936, %1946 ]
  %1975 = phi i32 [ %1970, %cleanup26.i.i.i1498 ], [ %1937, %1946 ]
  %1976 = phi i32 [ %1970, %cleanup26.i.i.i1498 ], [ %retval.sroa.8.0.i.i.i1472, %1946 ]
  %1977 = phi i64 [ %1971, %cleanup26.i.i.i1498 ], [ %retval.sroa.2.0.i.i.i1471, %1946 ]
  %1978 = phi i32 [ %1972, %cleanup26.i.i.i1498 ], [ %retval.sroa.6.0.i.i.i1469, %1946 ]
  %1979 = phi i1 [ %cmp.i46.not.i.i.i1490, %cleanup26.i.i.i1498 ], [ true, %1946 ]
  br i1 %1979, label %for.inc.i.i.i1503, label %Flow3526, !amdgpu.uniform !42

cleanup26.i.i.i1498:                              ; preds = %Flow3524
  br label %Flow3525, !amdgpu.uniform !42

for.inc.i.i.i1503:                                ; preds = %Flow3525
  %inc.i.i.i1508 = add i32 %spec.store.select.i.i.i1475, 1
  br label %Flow3526, !amdgpu.uniform !42

Flow3526:                                         ; preds = %for.inc.i.i.i1503, %Flow3525
  %1980 = phi i32 [ %inc.i.i.i1508, %for.inc.i.i.i1503 ], [ poison, %Flow3525 ]
  %1981 = phi i1 [ false, %for.inc.i.i.i1503 ], [ true, %Flow3525 ]
  br i1 %1981, label %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1509, label %for.cond.i.i1.i1468, !amdgpu.uniform !42

_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1509: ; preds = %Flow3526
  %.lcssa4033 = phi i32 [ %1973, %Flow3526 ]
  %.lcssa4032 = phi i64 [ %1974, %Flow3526 ]
  %.lcssa4031 = phi i32 [ %1975, %Flow3526 ]
  %scevgep2953 = getelementptr i8, ptr %RHSStr, i64 %lsr.iv.next2943.lcssa
  %sub.ptr.lhs.cast.i.i.i1465 = ptrtoint ptr %scevgep2953 to i64
  %sub.ptr.rhs.cast.i.i.i = ptrtoint ptr %RHSStr to i64
  %sub.ptr.sub.i.i.i1466 = sub i64 %sub.ptr.lhs.cast.i.i.i1465, %sub.ptr.rhs.cast.i.i.i
  fence acquire
  %1982 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %mul.i.i.i.i.i.i1510 = shl i32 %.lcssa4033, 6
  %idxprom.i19.i.i.i.i.i1511 = zext i32 %mul.i.i.i.i.i.i1510 to i64
  %arrayidx.i20.i.i.i.i.i1512 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %1982, i64 %idxprom.i19.i.i.i.i.i1511
  %arrayidx.i22.i.i.i.i.i1513 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %arrayidx.i20.i.i.i.i.i1512, i64 %sh_prom.i.i.i.i
  store i64 %sub.ptr.sub.i.i.i1466, ptr %arrayidx.i22.i.i.i.i.i1513, align 8, !tbaa !82
  %spec.select.i.i.i.i.i.i.i1514 = tail call i64 @llvm.umin.i64(i64 %sub.ptr.sub.i.i.i1466, i64 56)
  %1983 = and i64 %spec.select.i.i.i.i.i.i.i1514, 15
  %1984 = sub i64 %spec.select.i.i.i.i.i.i.i1514, %1983
  %1985 = icmp ne i64 %1984, 0
  %1986 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %1985)
  %1987 = extractvalue { i1, i64 } %1986, 0
  %1988 = extractvalue { i1, i64 } %1986, 1
  br i1 %1987, label %loop-memcpy-expansion2218.preheader, label %Flow3522

loop-memcpy-expansion2218.preheader:              ; preds = %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1509
  %1989 = add nuw nsw i64 %75, 8
  %scevgep2950 = getelementptr i8, ptr %1982, i64 %1989
  %1990 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i1511, 6
  %scevgep2951 = getelementptr i8, ptr %scevgep2950, i64 %1990
  br label %loop-memcpy-expansion2218, !amdgpu.uniform !42

Flow3522:                                         ; preds = %Flow3521, %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1509
  call void @llvm.amdgcn.end.cf.i64(i64 %1988)
  br label %loop-memcpy-residual-header2221, !amdgpu.uniform !42

loop-memcpy-expansion2218:                        ; preds = %loop-memcpy-expansion2218.preheader, %loop-memcpy-expansion2218
  %phi.broken3939 = phi i64 [ 0, %loop-memcpy-expansion2218.preheader ], [ %1995, %loop-memcpy-expansion2218 ]
  %loop-index2219 = phi i64 [ %1993, %loop-memcpy-expansion2218 ], [ 0, %loop-memcpy-expansion2218.preheader ]
  %1991 = getelementptr i8, ptr %RHSStr, i64 %loop-index2219
  %1992 = load <4 x i32>, ptr %1991, align 1
  %scevgep2952 = getelementptr i8, ptr %scevgep2951, i64 %loop-index2219
  store <4 x i32> %1992, ptr %scevgep2952, align 1
  %1993 = add i64 %loop-index2219, 16
  %1994 = icmp uge i64 %1993, %1984
  %1995 = call i64 @llvm.amdgcn.if.break.i64(i1 %1994, i64 %phi.broken3939)
  %1996 = call i1 @llvm.amdgcn.loop.i64(i64 %1995)
  br i1 %1996, label %Flow3521, label %loop-memcpy-expansion2218

Flow3519:                                         ; preds = %Flow3516, %Flow3518
  %1997 = phi i64 [ %2749, %Flow3516 ], [ %2742, %Flow3518 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %2745)
  br label %post-loop-memcpy-expansion2217, !amdgpu.uniform !42

loop-memcpy-residual2220:                         ; preds = %loop-memcpy-residual2220.preheader, %loop-memcpy-residual2220
  %phi.broken3940 = phi i64 [ 0, %loop-memcpy-residual2220.preheader ], [ %2001, %loop-memcpy-residual2220 ]
  %residual-loop-index2222 = phi i64 [ %1999, %loop-memcpy-residual2220 ], [ 0, %loop-memcpy-residual2220.preheader ]
  %scevgep2949 = getelementptr i8, ptr %scevgep2948, i64 %residual-loop-index2222
  %1998 = load i8, ptr %scevgep2949, align 1
  %scevgep2947 = getelementptr i8, ptr %scevgep2946, i64 %residual-loop-index2222
  store i8 %1998, ptr %scevgep2947, align 1
  %1999 = add i64 %residual-loop-index2222, 1
  %2000 = icmp uge i64 %1999, %1983
  %2001 = call i64 @llvm.amdgcn.if.break.i64(i1 %2000, i64 %phi.broken3940)
  %2002 = call i1 @llvm.amdgcn.loop.i64(i64 %2001)
  br i1 %2002, label %Flow3516, label %loop-memcpy-residual2220

Flow3516:                                         ; preds = %loop-memcpy-residual2220
  %.lcssa4026 = phi i64 [ %2001, %loop-memcpy-residual2220 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa4026)
  br label %Flow3519, !amdgpu.uniform !42

post-loop-memcpy-expansion2217:                   ; preds = %Flow3519
  %2003 = add i64 %lsr.iv.next2943.lcssa, -56
  %tobool.not.i.i.i.i.i.i1516 = icmp eq i32 %.lcssa4031, 0
  %conv.i.i.i.i.i.i1517 = zext i1 %tobool.not.i.i.i.i.i.i1516 to i32
  fence release
  %2004 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %idxprom.i23.i.i.i.i.i1518 = zext i32 %.lcssa4033 to i64
  %arrayidx.i24.i.i.i.i.i1519 = getelementptr inbounds nuw i32, ptr %2004, i64 %idxprom.i23.i.i.i.i.i1518
  store atomic i32 %conv.i.i.i.i.i.i1517, ptr %arrayidx.i24.i.i.i.i.i1519 monotonic, align 4
  %2005 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !84
  %arrayidx.i.i5.i.i1520 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %2005, i64 %idxprom.i23.i.i.i.i.i1518, !amdgpu.uniform !42
  %2006 = load i64, ptr %arrayidx.i.i5.i.i1520, align 8, !tbaa !80
  %scevgep2925 = getelementptr i8, ptr %RHSStr, i64 56
  br label %while.cond.i.i.i.i1521, !amdgpu.uniform !42

while.cond.i.i.i.i1521:                           ; preds = %post-loop-memcpy-expansion2217, %Flow3514
  %phi.broken3943 = phi i64 [ 0, %post-loop-memcpy-expansion2217 ], [ %2023, %Flow3514 ]
  %lsr.iv2933 = phi i64 [ %2021, %Flow3514 ], [ %2003, %post-loop-memcpy-expansion2217 ]
  %lsr.iv2926 = phi ptr [ %2020, %Flow3514 ], [ %scevgep2925, %post-loop-memcpy-expansion2217 ]
  %port.sroa.37.0.i.i1522 = phi i32 [ %2019, %Flow3514 ], [ %conv.i.i.i.i.i.i1517, %post-loop-memcpy-expansion2217 ]
  %idx.0.i.i.i.i1523 = phi i64 [ %2018, %Flow3514 ], [ 56, %post-loop-memcpy-expansion2217 ]
  %cmp.i.i.i.i1524 = icmp ugt i64 %sub.ptr.sub.i.i.i1466, %idx.0.i.i.i.i1523
  %2007 = tail call i64 @llvm.amdgcn.ballot.i64(i1 %cmp.i.i.i.i1524)
  %and.i.i.i.i6.i.i1525 = and i64 %2007, %2006
  %tobool.not.i.i.i.i1526 = icmp ne i64 %and.i.i.i.i6.i.i1525, 0
  %2008 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i22.i.i.i1527 = getelementptr inbounds nuw i32, ptr %2008, i64 %idxprom.i23.i.i.i.i.i1518, !amdgpu.uniform !42
  %2009 = load atomic i32, ptr %arrayidx.i22.i.i.i1527 monotonic, align 4
  %2010 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %tobool.not.i.i.i.i1526)
  %2011 = extractvalue { i1, i64 } %2010, 0
  %2012 = extractvalue { i1, i64 } %2010, 1
  br i1 %2011, label %cond.false.i8.i.i.i.i1528, label %Flow3514

cond.false.i8.i.i.i.i1528:                        ; preds = %while.cond.i.i.i.i1521
  %2013 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %2009)
  %2014 = icmp ne i32 %2013, %port.sroa.37.0.i.i1522
  br i1 %2014, label %while.body.i.i21.i.i.i.i1529.preheader, label %Flow3513, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i1529.preheader:           ; preds = %cond.false.i8.i.i.i.i1528
  br label %while.body.i.i21.i.i.i.i1529, !amdgpu.uniform !42

Flow3513:                                         ; preds = %Flow3511, %cond.false.i8.i.i.i.i1528
  br label %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i1532, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i1529:                     ; preds = %while.body.i.i21.i.i.i.i1529.preheader, %while.body.i.i21.i.i.i.i1529
  tail call void @llvm.amdgcn.s.sleep(i32 2)
  %2015 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i22.i.i.i.i1530 = getelementptr inbounds nuw i32, ptr %2015, i64 %idxprom.i23.i.i.i.i.i1518, !amdgpu.uniform !42
  %2016 = load atomic i32, ptr %arrayidx.i.i.i22.i.i.i.i1530 monotonic, align 4
  %2017 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %2016)
  %cmp.i.not.i.i23.i.i.i.i1531 = icmp eq i32 %2017, %port.sroa.37.0.i.i1522
  br i1 %cmp.i.not.i.i23.i.i.i.i1531, label %Flow3511, label %while.body.i.i21.i.i.i.i1529, !amdgpu.uniform !42

Flow3511:                                         ; preds = %while.body.i.i21.i.i.i.i1529
  br label %Flow3513, !amdgpu.uniform !42

Flow3514:                                         ; preds = %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1533, %while.cond.i.i.i.i1521
  %2018 = phi i64 [ %add.i.i.i.i1536, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1533 ], [ poison, %while.cond.i.i.i.i1521 ]
  %2019 = phi i32 [ %conv.i.i26.i.i.i.i1534, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1533 ], [ poison, %while.cond.i.i.i.i1521 ]
  %2020 = phi ptr [ %scevgep2927, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1533 ], [ poison, %while.cond.i.i.i.i1521 ]
  %2021 = phi i64 [ %lsr.iv.next2934, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1533 ], [ poison, %while.cond.i.i.i.i1521 ]
  %2022 = phi i1 [ false, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1533 ], [ true, %while.cond.i.i.i.i1521 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %2012)
  %2023 = call i64 @llvm.amdgcn.if.break.i64(i1 %2022, i64 %phi.broken3943)
  %2024 = call i1 @llvm.amdgcn.loop.i64(i64 %2023)
  br i1 %2024, label %cond.false.i.i.i1543, label %while.cond.i.i.i.i1521

_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i1532: ; preds = %Flow3513
  fence acquire
  %2025 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.i.i.i.i1524)
  %2026 = extractvalue { i1, i64 } %2025, 0
  %2027 = extractvalue { i1, i64 } %2025, 1
  br i1 %2026, label %if.then.i.i.i.i.i.i.i1537, label %Flow3510

if.then.i.i.i.i.i.i.i1537:                        ; preds = %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i1532
  %sub.i.i.i.i.i.i.i1539 = sub nuw i64 %sub.ptr.sub.i.i.i1466, %idx.0.i.i.i.i1523
  %spec.select.i.i.i36.i.i.i.i1540 = tail call i64 @llvm.umin.i64(i64 %sub.i.i.i.i.i.i.i1539, i64 64)
  %2028 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %2029 = and i64 %spec.select.i.i.i36.i.i.i.i1540, 15
  %2030 = sub i64 %spec.select.i.i.i36.i.i.i.i1540, %2029
  %2031 = icmp ne i64 %2030, 0
  %2032 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %2031)
  %2033 = extractvalue { i1, i64 } %2032, 0
  %2034 = extractvalue { i1, i64 } %2032, 1
  br i1 %2033, label %loop-memcpy-expansion2212.preheader, label %Flow3509

loop-memcpy-expansion2212.preheader:              ; preds = %if.then.i.i.i.i.i.i.i1537
  %arrayidx.i20.i35.i.i.i.i1541 = getelementptr %"struct.rpc::Buffer", ptr %2028, i64 %idxprom.i19.i.i.i.i.i1511
  %arrayidx.i22.i38.i.i.i.i1542 = getelementptr %"struct.rpc::Buffer", ptr %arrayidx.i20.i35.i.i.i.i1541, i64 %sh_prom.i.i.i.i
  br label %loop-memcpy-expansion2212, !amdgpu.uniform !42

Flow3509:                                         ; preds = %Flow3508, %if.then.i.i.i.i.i.i.i1537
  call void @llvm.amdgcn.end.cf.i64(i64 %2034)
  br label %loop-memcpy-residual-header2215, !amdgpu.uniform !42

loop-memcpy-expansion2212:                        ; preds = %loop-memcpy-expansion2212.preheader, %loop-memcpy-expansion2212
  %phi.broken3941 = phi i64 [ 0, %loop-memcpy-expansion2212.preheader ], [ %2039, %loop-memcpy-expansion2212 ]
  %loop-index2213 = phi i64 [ %2037, %loop-memcpy-expansion2212 ], [ 0, %loop-memcpy-expansion2212.preheader ]
  %scevgep2928 = getelementptr i8, ptr %lsr.iv2926, i64 %loop-index2213
  %2035 = load <4 x i32>, ptr %scevgep2928, align 1
  %2036 = getelementptr i8, ptr %arrayidx.i22.i38.i.i.i.i1542, i64 %loop-index2213
  store <4 x i32> %2035, ptr %2036, align 1
  %2037 = add i64 %loop-index2213, 16
  %2038 = icmp uge i64 %2037, %2030
  %2039 = call i64 @llvm.amdgcn.if.break.i64(i1 %2038, i64 %phi.broken3941)
  %2040 = call i1 @llvm.amdgcn.loop.i64(i64 %2039)
  br i1 %2040, label %Flow3508, label %loop-memcpy-expansion2212

Flow3506:                                         ; preds = %Flow3505, %loop-memcpy-residual-header2215
  call void @llvm.amdgcn.end.cf.i64(i64 %2735)
  br label %Flow3510, !amdgpu.uniform !42

loop-memcpy-residual2214:                         ; preds = %loop-memcpy-residual2214.preheader, %loop-memcpy-residual2214
  %phi.broken3942 = phi i64 [ 0, %loop-memcpy-residual2214.preheader ], [ %2044, %loop-memcpy-residual2214 ]
  %residual-loop-index2216 = phi i64 [ %2042, %loop-memcpy-residual2214 ], [ 0, %loop-memcpy-residual2214.preheader ]
  %scevgep2940 = getelementptr i8, ptr %scevgep2939, i64 %residual-loop-index2216
  %2041 = load i8, ptr %scevgep2940, align 1
  %scevgep2937 = getelementptr i8, ptr %scevgep2936, i64 %residual-loop-index2216
  store i8 %2041, ptr %scevgep2937, align 1
  %2042 = add i64 %residual-loop-index2216, 1
  %2043 = icmp uge i64 %2042, %2029
  %2044 = call i64 @llvm.amdgcn.if.break.i64(i1 %2043, i64 %phi.broken3942)
  %2045 = call i1 @llvm.amdgcn.loop.i64(i64 %2044)
  br i1 %2045, label %Flow3505, label %loop-memcpy-residual2214

Flow3505:                                         ; preds = %loop-memcpy-residual2214
  %.lcssa4023 = phi i64 [ %2044, %loop-memcpy-residual2214 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa4023)
  br label %Flow3506, !amdgpu.uniform !42

_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1533: ; preds = %Flow3510
  %conv.i.i26.i.i.i.i1534 = xor i32 %port.sroa.37.0.i.i1522, 1
  fence release
  %2046 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %arrayidx.i24.i29.i.i.i.i1535 = getelementptr inbounds nuw i32, ptr %2046, i64 %idxprom.i23.i.i.i.i.i1518
  store atomic i32 %conv.i.i26.i.i.i.i1534, ptr %arrayidx.i24.i29.i.i.i.i1535 monotonic, align 4
  %add.i.i.i.i1536 = add i64 %idx.0.i.i.i.i1523, 64
  %scevgep2927 = getelementptr i8, ptr %lsr.iv2926, i64 64
  %lsr.iv.next2934 = add i64 %lsr.iv2933, -64
  br label %Flow3514, !amdgpu.uniform !42

cond.false.i.i.i1543:                             ; preds = %Flow3514
  %.lcssa4025 = phi i64 [ %2023, %Flow3514 ]
  %port.sroa.37.0.i.i1522.lcssa = phi i32 [ %port.sroa.37.0.i.i1522, %Flow3514 ]
  %.lcssa4024 = phi i32 [ %2009, %Flow3514 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa4025)
  %2047 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %.lcssa4024)
  %2048 = icmp ne i32 %2047, %port.sroa.37.0.i.i1522.lcssa
  %2049 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %2048)
  %2050 = extractvalue { i1, i64 } %2049, 0
  %2051 = extractvalue { i1, i64 } %2049, 1
  br i1 %2050, label %while.body.i.i.i.i1544.preheader, label %Flow3503

while.body.i.i.i.i1544.preheader:                 ; preds = %cond.false.i.i.i1543
  br label %while.body.i.i.i.i1544, !amdgpu.uniform !42

Flow3503:                                         ; preds = %Flow3501, %cond.false.i.i.i1543
  call void @llvm.amdgcn.end.cf.i64(i64 %2051)
  br label %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit1565, !amdgpu.uniform !42

while.body.i.i.i.i1544:                           ; preds = %while.body.i.i.i.i1544.preheader, %while.body.i.i.i.i1544
  %phi.broken3944 = phi i64 [ 0, %while.body.i.i.i.i1544.preheader ], [ %2055, %while.body.i.i.i.i1544 ]
  tail call void @llvm.amdgcn.s.sleep(i32 2)
  %2052 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i8.i.i1545 = getelementptr inbounds nuw i32, ptr %2052, i64 %idxprom.i23.i.i.i.i.i1518, !amdgpu.uniform !42
  %2053 = load atomic i32, ptr %arrayidx.i.i.i8.i.i1545 monotonic, align 4
  %2054 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %2053)
  %cmp.i.not.i.i.i.i1546 = icmp eq i32 %2054, %port.sroa.37.0.i.i1522.lcssa
  %2055 = call i64 @llvm.amdgcn.if.break.i64(i1 %cmp.i.not.i.i.i.i1546, i64 %phi.broken3944)
  %2056 = call i1 @llvm.amdgcn.loop.i64(i64 %2055)
  br i1 %2056, label %Flow3501, label %while.body.i.i.i.i1544

Flow3501:                                         ; preds = %while.body.i.i.i.i1544
  %.lcssa4021 = phi i64 [ %2055, %while.body.i.i.i.i1544 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa4021)
  br label %Flow3503, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit1565: ; preds = %Flow3503
  fence acquire
  tail call void @llvm.amdgcn.wave.barrier()
  fence syncscope("agent") release
  %2057 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %.lcssa4032, i1 true)
  %iszero.i.i.i.i.i13.i.i1548 = icmp ne i64 %.lcssa4032, 0
  %cmp2.i.i.i.i14.i.i1549 = icmp eq i64 %2057, %sh_prom.i.i.i.i
  %cmp.i.i.i.i15.i.i1550 = select i1 %iszero.i.i.i.i.i13.i.i1548, i1 %cmp2.i.i.i.i14.i.i1549, i1 false
  %2058 = and i32 %.lcssa4033, 31
  %conv4.i.i.i19.i.i1554 = zext i1 %cmp.i.i.i.i15.i.i1550 to i32
  %shl.i.i.i20.i.i1555 = shl nuw i32 %conv4.i.i.i19.i.i1554, %2058
  %xor.i.i.i21.i.i1556 = xor i32 %shl.i.i.i20.i.i1555, -1
  %2059 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2626, !amdgpu.uniform !42

2060:                                             ; preds = %ComputeEnd2627
  %div8.i.i.i16.i.i1551 = lshr i32 %.lcssa4033, 5
  %idxprom.i.i.i17.i.i1552 = zext nneg i32 %div8.i.i.i16.i.i1551 to i64
  %2061 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i17.i.i1552
  %sunkaddr3320 = getelementptr inbounds i8, ptr addrspace(1) %2061, i64 40
  %2062 = atomicrmw and ptr addrspace(1) %sunkaddr3320, i32 %.lcssa4020 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %2063, !amdgpu.uniform !42

2063:                                             ; preds = %2060, %ComputeEnd2627
  call void @llvm.amdgcn.end.cf.i64(i64 %3785)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %for.cond.i.i.i1569, !amdgpu.uniform !42

for.cond.i.i.i1569:                               ; preds = %2063, %Flow3500
  %2064 = phi i32 [ %2102, %Flow3500 ], [ poison, %2063 ]
  %2065 = phi i64 [ %2103, %Flow3500 ], [ poison, %2063 ]
  %2066 = phi i32 [ %2104, %Flow3500 ], [ poison, %2063 ]
  %retval.sroa.6.0.i.i.i1570 = phi i32 [ %2107, %Flow3500 ], [ undef, %2063 ]
  %retval.sroa.2.0.i.i.i1572 = phi i64 [ %2106, %Flow3500 ], [ undef, %2063 ]
  %retval.sroa.8.0.i.i.i1573 = phi i32 [ %2105, %Flow3500 ], [ undef, %2063 ]
  %index.0.i.i.i1574 = phi i32 [ %2109, %Flow3500 ], [ 0, %2063 ]
  %2067 = load i32, ptr addrspace(1) @__llvm_rpc_client, align 8, !tbaa !68
  %cmp.not.i.i.i1575 = icmp ult i32 %index.0.i.i.i1574, %2067
  %spec.store.select.i.i.i1576 = select i1 %cmp.not.i.i.i1575, i32 %index.0.i.i.i1574, i32 0
  %2068 = tail call noundef i64 @llvm.amdgcn.ballot.i64(i1 true)
  %2069 = and i32 %spec.store.select.i.i.i1576, 31
  %2070 = lshr i64 %2068, %sh_prom.i.i.i.i
  %2071 = trunc i64 %2070 to i32
  %conv4.i.i.i.i.i1580 = and i32 %2071, 1
  %shl.i.i.i.i.i1581 = shl nuw i32 %conv4.i.i.i.i.i1580, %2069
  %2072 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2630, !amdgpu.uniform !42

2073:                                             ; preds = %ComputeEnd2631
  %sunkaddr3321 = getelementptr inbounds i8, ptr addrspace(1) %3795, i64 40
  %2074 = atomicrmw or ptr addrspace(1) %sunkaddr3321, i32 %.lcssa4014 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %2075, !amdgpu.uniform !42

2075:                                             ; preds = %2073, %ComputeEnd2631
  %2076 = phi i32 [ %2074, %2073 ], [ poison, %ComputeEnd2631 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %3804)
  %2077 = call i32 @llvm.amdgcn.readfirstlane.i32(i32 %2076)
  %2078 = or i32 %2077, %.lcssa4015
  %shl5.i.i.i.i.i1582 = shl nuw i32 1, %2069
  %and.i.i.i.i.i1583 = and i32 %shl5.i.i.i.i.i1582, %2078
  %tobool3.i.i.i.i1584 = icmp ne i32 %and.i.i.i.i.i1583, 0
  %2079 = tail call i64 @llvm.amdgcn.ballot.i64(i1 %tobool3.i.i.i.i1584)
  %and.i.i.i.i.i.i1585 = and i64 %2079, %2068
  %cmp.i.not.i.i.i1586 = icmp ne i64 %2068, %and.i.i.i.i.i.i1585
  br i1 %cmp.i.not.i.i.i1586, label %if.then.i.i.i.i1587, label %Flow3499, !amdgpu.uniform !42

if.then.i.i.i.i1587:                              ; preds = %2075
  fence syncscope("agent") acquire
  %2080 = load <2 x i64>, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !75
  %2081 = extractelement <2 x i64> %2080, i32 0
  %2082 = inttoptr i64 %2081 to ptr
  %2083 = extractelement <2 x i64> %2080, i32 1
  %2084 = inttoptr i64 %2083 to ptr
  %idxprom.i.i.i.i1588 = zext i32 %spec.store.select.i.i.i1576 to i64
  %arrayidx.i.i.i.i1589 = getelementptr inbounds nuw i32, ptr %2082, i64 %idxprom.i.i.i.i1588, !amdgpu.uniform !42
  %2085 = load atomic i32, ptr %arrayidx.i.i.i.i1589 monotonic, align 4
  %2086 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %2085)
  %arrayidx.i45.i.i.i1590 = getelementptr inbounds nuw i32, ptr %2084, i64 %idxprom.i.i.i.i1588, !amdgpu.uniform !42
  %2087 = load atomic i32, ptr %arrayidx.i45.i.i.i1590 monotonic, align 4
  %2088 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %2087)
  %cmp.i46.not.i.i.i1591 = icmp ne i32 %2086, %2088
  br i1 %cmp.i46.not.i.i.i1591, label %if.then12.i.i.i1592, label %Flow3497, !amdgpu.uniform !42

if.then12.i.i.i1592:                              ; preds = %if.then.i.i.i.i1587
  fence syncscope("agent") release
  %2089 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %2068, i1 true)
  %iszero.i.i.i.i.i.i.i1593 = icmp ne i64 %2068, 0
  %cmp2.i.i.i.i.i.i1594 = icmp eq i64 %2089, %sh_prom.i.i.i.i
  %cmp.i.i.i.i.i.i1595 = select i1 %iszero.i.i.i.i.i.i.i1593, i1 %cmp2.i.i.i.i.i.i1594, i1 false
  %conv4.i.i51.i.i.i1596 = zext i1 %cmp.i.i.i.i.i.i1595 to i32
  %shl.i.i52.i.i.i1597 = shl nuw i32 %conv4.i.i51.i.i.i1596, %2069
  %xor.i.i.i.i.i1598 = xor i32 %shl.i.i52.i.i.i1597, -1
  %2090 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2635, !amdgpu.uniform !42

2091:                                             ; preds = %ComputeEnd2636
  %sunkaddr3322 = getelementptr inbounds i8, ptr addrspace(1) %3795, i64 40
  %2092 = atomicrmw and ptr addrspace(1) %sunkaddr3322, i32 %.lcssa4016 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %2093, !amdgpu.uniform !42

2093:                                             ; preds = %2091, %ComputeEnd2636
  call void @llvm.amdgcn.end.cf.i64(i64 %3822)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %Flow3497, !amdgpu.uniform !42

if.end14.i.i.i1656:                               ; preds = %Flow3497
  %2094 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %2068, i1 true)
  %iszero.i.i.i.i.i.i1657 = icmp ne i64 %2068, 0
  %cmp2.i.i.i.i.i1658 = icmp eq i64 %2094, %sh_prom.i.i.i.i
  %cmp.i.i.i.i.i1659 = select i1 %iszero.i.i.i.i.i.i1657, i1 %cmp2.i.i.i.i.i1658, i1 false
  %2095 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.i.i.i.i.i1659)
  %2096 = extractvalue { i1, i64 } %2095, 0
  %2097 = extractvalue { i1, i64 } %2095, 1
  br i1 %2096, label %if.then16.i.i.i1661, label %if.end22.i.i.i1660

if.then16.i.i.i1661:                              ; preds = %if.end14.i.i.i1656
  %2098 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !76
  %opcode.i.i.i1662 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %2098, i64 %idxprom.i.i.i.i1588, i32 1
  store i32 1660944387, ptr %opcode.i.i.i1662, align 8, !tbaa !77
  %arrayidx21.i.i.i1663 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %2098, i64 %idxprom.i.i.i.i1588
  store i64 %2068, ptr %arrayidx21.i.i.i1663, align 8, !tbaa !80
  br label %if.end22.i.i.i1660, !amdgpu.uniform !42

Flow3498:                                         ; preds = %if.end22.i.i.i1660, %Flow3497
  %2099 = phi i32 [ %2086, %if.end22.i.i.i1660 ], [ %retval.sroa.8.0.i.i.i1573, %Flow3497 ]
  %2100 = phi i64 [ %2068, %if.end22.i.i.i1660 ], [ %retval.sroa.2.0.i.i.i1572, %Flow3497 ]
  %2101 = phi i32 [ %spec.store.select.i.i.i1576, %if.end22.i.i.i1660 ], [ %retval.sroa.6.0.i.i.i1570, %Flow3497 ]
  br label %cleanup26.i.i.i1599, !amdgpu.uniform !42

if.end22.i.i.i1660:                               ; preds = %if.then16.i.i.i1661, %if.end14.i.i.i1656
  call void @llvm.amdgcn.end.cf.i64(i64 %2097)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %Flow3498, !amdgpu.uniform !42

Flow3499:                                         ; preds = %cleanup26.i.i.i1599, %2075
  %2102 = phi i32 [ %2101, %cleanup26.i.i.i1599 ], [ %2064, %2075 ]
  %2103 = phi i64 [ %2100, %cleanup26.i.i.i1599 ], [ %2065, %2075 ]
  %2104 = phi i32 [ %2099, %cleanup26.i.i.i1599 ], [ %2066, %2075 ]
  %2105 = phi i32 [ %2099, %cleanup26.i.i.i1599 ], [ %retval.sroa.8.0.i.i.i1573, %2075 ]
  %2106 = phi i64 [ %2100, %cleanup26.i.i.i1599 ], [ %retval.sroa.2.0.i.i.i1572, %2075 ]
  %2107 = phi i32 [ %2101, %cleanup26.i.i.i1599 ], [ %retval.sroa.6.0.i.i.i1570, %2075 ]
  %2108 = phi i1 [ %cmp.i46.not.i.i.i1591, %cleanup26.i.i.i1599 ], [ true, %2075 ]
  br i1 %2108, label %for.inc.i.i.i1604, label %Flow3500, !amdgpu.uniform !42

cleanup26.i.i.i1599:                              ; preds = %Flow3498
  br label %Flow3499, !amdgpu.uniform !42

for.inc.i.i.i1604:                                ; preds = %Flow3499
  %inc.i.i.i1609 = add i32 %spec.store.select.i.i.i1576, 1
  br label %Flow3500, !amdgpu.uniform !42

Flow3500:                                         ; preds = %for.inc.i.i.i1604, %Flow3499
  %2109 = phi i32 [ %inc.i.i.i1609, %for.inc.i.i.i1604 ], [ poison, %Flow3499 ]
  %2110 = phi i1 [ false, %for.inc.i.i.i1604 ], [ true, %Flow3499 ]
  br i1 %2110, label %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1610, label %for.cond.i.i.i1569, !amdgpu.uniform !42

_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1610: ; preds = %Flow3500
  %.lcssa4019 = phi i32 [ %2102, %Flow3500 ]
  %.lcssa4018 = phi i64 [ %2103, %Flow3500 ]
  %.lcssa4017 = phi i32 [ %2104, %Flow3500 ]
  fence acquire
  %2111 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %mul.i.i.i.i.i.i1611 = shl i32 %.lcssa4019, 6
  %idxprom.i19.i.i.i.i.i1612 = zext i32 %mul.i.i.i.i.i.i1611 to i64
  %arrayidx.i20.i.i.i.i.i1613 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %2111, i64 %idxprom.i19.i.i.i.i.i1612
  %arrayidx.i22.i.i.i.i.i1614 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %arrayidx.i20.i.i.i.i.i1613, i64 %sh_prom.i.i.i.i
  store i64 1, ptr %arrayidx.i22.i.i.i.i.i1614, align 8, !tbaa !82
  %arrayidx6.i.i.i.i.i.i.i1615 = getelementptr inbounds nuw i8, ptr %arrayidx.i22.i.i.i.i.i1614, i64 8
  store i8 10, ptr %arrayidx6.i.i.i.i.i.i.i1615, align 8
  %tobool.not.i.i.i.i.i.i1616 = icmp eq i32 %.lcssa4017, 0
  %conv.i.i.i.i.i.i1617 = zext i1 %tobool.not.i.i.i.i.i.i1616 to i32
  fence release
  %2112 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %idxprom.i23.i.i.i.i.i1618 = zext i32 %.lcssa4019 to i64
  %arrayidx.i24.i.i.i.i.i1619 = getelementptr inbounds nuw i32, ptr %2112, i64 %idxprom.i23.i.i.i.i.i1618
  store atomic i32 %conv.i.i.i.i.i.i1617, ptr %arrayidx.i24.i.i.i.i.i1619 monotonic, align 4
  %2113 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !84
  %arrayidx.i.i5.i.i1620 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %2113, i64 %idxprom.i23.i.i.i.i.i1618, !amdgpu.uniform !42
  %2114 = load i64, ptr %arrayidx.i.i5.i.i1620, align 8, !tbaa !80
  br label %while.cond.i.i.i.i1621, !amdgpu.uniform !42

while.cond.i.i.i.i1621:                           ; preds = %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1610, %Flow3496
  %phi.broken3945 = phi i64 [ 0, %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1610 ], [ %2129, %Flow3496 ]
  %port.sroa.37.0.i.i1622 = phi i32 [ %2127, %Flow3496 ], [ %conv.i.i.i.i.i.i1617, %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1610 ]
  %idx.0.i.i.i.i1623 = phi i64 [ %2126, %Flow3496 ], [ 56, %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1610 ]
  %cmp.i.i.i.i1624 = icmp eq i64 %idx.0.i.i.i.i1623, 0
  %2115 = tail call i64 @llvm.amdgcn.ballot.i64(i1 %cmp.i.i.i.i1624)
  %and.i.i.i.i6.i.i1625 = and i64 %2115, %2114
  %tobool.not.i.i.i.i1626 = icmp ne i64 %and.i.i.i.i6.i.i1625, 0
  %2116 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i22.i.i.i1627 = getelementptr inbounds nuw i32, ptr %2116, i64 %idxprom.i23.i.i.i.i.i1618, !amdgpu.uniform !42
  %2117 = load atomic i32, ptr %arrayidx.i22.i.i.i1627 monotonic, align 4
  %2118 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %tobool.not.i.i.i.i1626)
  %2119 = extractvalue { i1, i64 } %2118, 0
  %2120 = extractvalue { i1, i64 } %2118, 1
  br i1 %2119, label %cond.false.i8.i.i.i.i1628, label %Flow3496

cond.false.i8.i.i.i.i1628:                        ; preds = %while.cond.i.i.i.i1621
  %2121 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %2117)
  %2122 = icmp ne i32 %2121, %port.sroa.37.0.i.i1622
  br i1 %2122, label %while.body.i.i21.i.i.i.i1629.preheader, label %Flow3495, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i1629.preheader:           ; preds = %cond.false.i8.i.i.i.i1628
  br label %while.body.i.i21.i.i.i.i1629, !amdgpu.uniform !42

Flow3495:                                         ; preds = %Flow3493, %cond.false.i8.i.i.i.i1628
  br label %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i1632, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i1629:                     ; preds = %while.body.i.i21.i.i.i.i1629.preheader, %while.body.i.i21.i.i.i.i1629
  tail call void @llvm.amdgcn.s.sleep(i32 2)
  %2123 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i22.i.i.i.i1630 = getelementptr inbounds nuw i32, ptr %2123, i64 %idxprom.i23.i.i.i.i.i1618, !amdgpu.uniform !42
  %2124 = load atomic i32, ptr %arrayidx.i.i.i22.i.i.i.i1630 monotonic, align 4
  %2125 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %2124)
  %cmp.i.not.i.i23.i.i.i.i1631 = icmp eq i32 %2125, %port.sroa.37.0.i.i1622
  br i1 %cmp.i.not.i.i23.i.i.i.i1631, label %Flow3493, label %while.body.i.i21.i.i.i.i1629, !amdgpu.uniform !42

Flow3493:                                         ; preds = %while.body.i.i21.i.i.i.i1629
  br label %Flow3495, !amdgpu.uniform !42

Flow3496:                                         ; preds = %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1633, %while.cond.i.i.i.i1621
  %2126 = phi i64 [ %add.i.i.i.i1636, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1633 ], [ poison, %while.cond.i.i.i.i1621 ]
  %2127 = phi i32 [ %conv.i.i26.i.i.i.i1634, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1633 ], [ poison, %while.cond.i.i.i.i1621 ]
  %2128 = phi i1 [ false, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1633 ], [ true, %while.cond.i.i.i.i1621 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %2120)
  %2129 = call i64 @llvm.amdgcn.if.break.i64(i1 %2128, i64 %phi.broken3945)
  %2130 = call i1 @llvm.amdgcn.loop.i64(i64 %2129)
  br i1 %2130, label %cond.false.i.i.i1642, label %while.cond.i.i.i.i1621

_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i1632: ; preds = %Flow3495
  fence acquire
  br i1 %cmp.i.i.i.i1624, label %if.then.i.i.i.i.i.i.i1637, label %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1633, !amdgpu.uniform !42

if.then.i.i.i.i.i.i.i1637:                        ; preds = %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i1632
  %2131 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %arrayidx.i20.i35.i.i.i.i1640 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %2131, i64 %idxprom.i19.i.i.i.i.i1612
  %arrayidx.i22.i38.i.i.i.i1641 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %arrayidx.i20.i35.i.i.i.i1640, i64 %sh_prom.i.i.i.i
  store i8 10, ptr %arrayidx.i22.i38.i.i.i.i1641, align 1
  br label %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1633, !amdgpu.uniform !42

_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1633: ; preds = %if.then.i.i.i.i.i.i.i1637, %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i1632
  %conv.i.i26.i.i.i.i1634 = xor i32 %port.sroa.37.0.i.i1622, 1
  fence release
  %2132 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %arrayidx.i24.i29.i.i.i.i1635 = getelementptr inbounds nuw i32, ptr %2132, i64 %idxprom.i23.i.i.i.i.i1618
  store atomic i32 %conv.i.i26.i.i.i.i1634, ptr %arrayidx.i24.i29.i.i.i.i1635 monotonic, align 4
  %add.i.i.i.i1636 = add i64 %idx.0.i.i.i.i1623, 64
  br label %Flow3496, !amdgpu.uniform !42

cond.false.i.i.i1642:                             ; preds = %Flow3496
  %.lcssa4013 = phi i64 [ %2129, %Flow3496 ]
  %port.sroa.37.0.i.i1622.lcssa = phi i32 [ %port.sroa.37.0.i.i1622, %Flow3496 ]
  %.lcssa4012 = phi i32 [ %2117, %Flow3496 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa4013)
  %2133 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %.lcssa4012)
  %2134 = icmp ne i32 %2133, %port.sroa.37.0.i.i1622.lcssa
  %2135 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %2134)
  %2136 = extractvalue { i1, i64 } %2135, 0
  %2137 = extractvalue { i1, i64 } %2135, 1
  br i1 %2136, label %while.body.i.i.i.i1643.preheader, label %Flow3492

while.body.i.i.i.i1643.preheader:                 ; preds = %cond.false.i.i.i1642
  br label %while.body.i.i.i.i1643, !amdgpu.uniform !42

Flow3492:                                         ; preds = %Flow3490, %cond.false.i.i.i1642
  call void @llvm.amdgcn.end.cf.i64(i64 %2137)
  br label %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIcEERS1_T_.exit1664, !amdgpu.uniform !42

while.body.i.i.i.i1643:                           ; preds = %while.body.i.i.i.i1643.preheader, %while.body.i.i.i.i1643
  %phi.broken3946 = phi i64 [ 0, %while.body.i.i.i.i1643.preheader ], [ %2141, %while.body.i.i.i.i1643 ]
  tail call void @llvm.amdgcn.s.sleep(i32 2)
  %2138 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i8.i.i1644 = getelementptr inbounds nuw i32, ptr %2138, i64 %idxprom.i23.i.i.i.i.i1618, !amdgpu.uniform !42
  %2139 = load atomic i32, ptr %arrayidx.i.i.i8.i.i1644 monotonic, align 4
  %2140 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %2139)
  %cmp.i.not.i.i.i.i1645 = icmp eq i32 %2140, %port.sroa.37.0.i.i1622.lcssa
  %2141 = call i64 @llvm.amdgcn.if.break.i64(i1 %cmp.i.not.i.i.i.i1645, i64 %phi.broken3946)
  %2142 = call i1 @llvm.amdgcn.loop.i64(i64 %2141)
  br i1 %2142, label %Flow3490, label %while.body.i.i.i.i1643

Flow3490:                                         ; preds = %while.body.i.i.i.i1643
  %.lcssa4011 = phi i64 [ %2141, %while.body.i.i.i.i1643 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa4011)
  br label %Flow3492, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIcEERS1_T_.exit1664: ; preds = %Flow3492
  fence acquire
  tail call void @llvm.amdgcn.wave.barrier()
  fence syncscope("agent") release
  %2143 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %.lcssa4018, i1 true)
  %iszero.i.i.i.i.i13.i.i1647 = icmp ne i64 %.lcssa4018, 0
  %cmp2.i.i.i.i14.i.i1648 = icmp eq i64 %2143, %sh_prom.i.i.i.i
  %cmp.i.i.i.i15.i.i1649 = select i1 %iszero.i.i.i.i.i13.i.i1647, i1 %cmp2.i.i.i.i14.i.i1648, i1 false
  %2144 = and i32 %.lcssa4019, 31
  %conv4.i.i.i19.i.i1653 = zext i1 %cmp.i.i.i.i15.i.i1649 to i32
  %shl.i.i.i20.i.i1654 = shl nuw i32 %conv4.i.i.i19.i.i1653, %2144
  %xor.i.i.i21.i.i1655 = xor i32 %shl.i.i.i20.i.i1654, -1
  %2145 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2639, !amdgpu.uniform !42

2146:                                             ; preds = %ComputeEnd2640
  %div8.i.i.i16.i.i1650 = lshr i32 %.lcssa4019, 5
  %idxprom.i.i.i17.i.i1651 = zext nneg i32 %div8.i.i.i16.i.i1650 to i64
  %2147 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i17.i.i1651
  %sunkaddr3323 = getelementptr inbounds i8, ptr addrspace(1) %2147, i64 40
  %2148 = atomicrmw and ptr addrspace(1) %sunkaddr3323, i32 %.lcssa4010 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %2149, !amdgpu.uniform !42

2149:                                             ; preds = %2146, %ComputeEnd2640
  call void @llvm.amdgcn.end.cf.i64(i64 %3839)
  tail call void @llvm.amdgcn.wave.barrier()
  br i1 %cmp2.i.i.i35.i, label %if.then3.i.i.i87.i, label %if.end4.i.i.i76.i, !amdgpu.uniform !42

if.then3.i.i.i87.i:                               ; preds = %2149
  %mul.i.i.i88.i = mul nuw i64 %inc.i.i.i32.i, 11
  %div18.i.i.i89.i = lshr i64 %mul.i.i.i88.i, 3
  br label %if.end4.i.i.i76.i, !amdgpu.uniform !42

if.end4.i.i.i76.i:                                ; preds = %if.then3.i.i.i87.i, %2149
  %new_capacity.addr.0.i.i.i77.i = phi i64 [ %div18.i.i.i89.i, %if.then3.i.i.i87.i ], [ %inc.i.i.i32.i, %2149 ]
  %sub.i.i1665 = add i64 %new_capacity.addr.0.i.i.i77.i, 7
  %div2.i.i1666 = and i64 %sub.i.i1665, -8
  %2150 = load ptr, ptr addrspace(1) @_ZN12_GLOBAL__N_13ptrE, align 8, !tbaa !55
  %add.ptr.i.i1667 = getelementptr inbounds nuw i8, ptr %2150, i64 %div2.i.i1666
  store ptr %add.ptr.i.i1667, ptr addrspace(1) @_ZN12_GLOBAL__N_13ptrE, align 8, !tbaa !55
  br i1 %597, label %loop-memcpy-expansion2272.preheader, label %Flow3489, !amdgpu.uniform !42

loop-memcpy-expansion2272.preheader:              ; preds = %if.end4.i.i.i76.i
  br label %loop-memcpy-expansion2272, !amdgpu.uniform !42

Flow3489:                                         ; preds = %Flow3488, %if.end4.i.i.i76.i
  br label %loop-memcpy-residual-header2275, !amdgpu.uniform !42

loop-memcpy-expansion2272:                        ; preds = %loop-memcpy-expansion2272.preheader, %loop-memcpy-expansion2272
  %loop-index2273 = phi i64 [ %2154, %loop-memcpy-expansion2272 ], [ 0, %loop-memcpy-expansion2272.preheader ]
  %2151 = getelementptr i8, ptr %3, i64 %loop-index2273, !amdgpu.uniform !42
  %2152 = load <4 x i32>, ptr %2151, align 1
  %2153 = getelementptr i8, ptr %2150, i64 %loop-index2273
  store <4 x i32> %2152, ptr %2153, align 1
  %2154 = add i64 %loop-index2273, 16
  %2155 = icmp uge i64 %2154, %596
  br i1 %2155, label %Flow3488, label %loop-memcpy-expansion2272, !amdgpu.uniform !42

Flow3486:                                         ; preds = %Flow3485, %loop-memcpy-residual-header2275
  br label %post-loop-memcpy-expansion2271, !amdgpu.uniform !42

loop-memcpy-residual2274:                         ; preds = %loop-memcpy-residual2274.preheader, %loop-memcpy-residual2274
  %residual-loop-index2276 = phi i64 [ %2157, %loop-memcpy-residual2274 ], [ 0, %loop-memcpy-residual2274.preheader ]
  %scevgep2924 = getelementptr i8, ptr %scevgep2923, i64 %residual-loop-index2276, !amdgpu.uniform !42
  %2156 = load i8, ptr %scevgep2924, align 1
  %scevgep2922 = getelementptr i8, ptr %scevgep2921, i64 %residual-loop-index2276
  store i8 %2156, ptr %scevgep2922, align 1
  %2157 = add i64 %residual-loop-index2276, 1
  %2158 = icmp uge i64 %2157, %595
  br i1 %2158, label %Flow3485, label %loop-memcpy-residual2274, !amdgpu.uniform !42

Flow3485:                                         ; preds = %loop-memcpy-residual2274
  br label %Flow3486, !amdgpu.uniform !42

post-loop-memcpy-expansion2271:                   ; preds = %Flow3486
  %sub.ptr.lhs.cast.i.i1668 = ptrtoint ptr %add.ptr.i.i1667 to i64
  %sub.ptr.sub.i.i1669 = sub i64 %sub.ptr.lhs.cast.i.i1668, ptrtoint (ptr addrspacecast (ptr addrspace(1) @_ZN12_GLOBAL__N_16memoryE to ptr) to i64)
  %cmp.i.i1670 = icmp ugt i64 %sub.ptr.sub.i.i1669, 65335
  %cond.i.i1671 = select i1 %cmp.i.i1670, ptr null, ptr %2150
  %cmp.not.i10.i.i84.i = icmp ne ptr %cond.i.i1671, addrspacecast (ptr addrspace(4) @_ZN22__llvm_libc_22_0_0_git3cpp6string14NULL_CHARACTERE to ptr)
  br i1 %cmp.not.i10.i.i84.i, label %if.then.i.i.i85.i, label %for.cond.i.i.i1673.preheader, !amdgpu.uniform !42

if.then.i.i.i85.i:                                ; preds = %post-loop-memcpy-expansion2271
  %arrayidx.i.i.i86.i = getelementptr inbounds nuw i8, ptr %2150, i64 %spec.select.i
  store i8 0, ptr %arrayidx.i.i.i86.i, align 1, !tbaa !57
  br label %for.cond.i.i.i1673.preheader, !amdgpu.uniform !42

for.cond.i.i.i1673.preheader:                     ; preds = %if.then.i.i.i85.i, %post-loop-memcpy-expansion2271
  br label %for.cond.i.i.i1673, !amdgpu.uniform !42

for.cond.i.i.i1673:                               ; preds = %for.cond.i.i.i1673.preheader, %Flow3483
  %2159 = phi i32 [ %2197, %Flow3483 ], [ poison, %for.cond.i.i.i1673.preheader ]
  %2160 = phi i64 [ %2198, %Flow3483 ], [ poison, %for.cond.i.i.i1673.preheader ]
  %2161 = phi i32 [ %2199, %Flow3483 ], [ poison, %for.cond.i.i.i1673.preheader ]
  %retval.sroa.6.0.i.i.i1674 = phi i32 [ %2202, %Flow3483 ], [ undef, %for.cond.i.i.i1673.preheader ]
  %retval.sroa.2.0.i.i.i1676 = phi i64 [ %2201, %Flow3483 ], [ undef, %for.cond.i.i.i1673.preheader ]
  %retval.sroa.8.0.i.i.i1677 = phi i32 [ %2200, %Flow3483 ], [ undef, %for.cond.i.i.i1673.preheader ]
  %index.0.i.i.i1678 = phi i32 [ %2204, %Flow3483 ], [ 0, %for.cond.i.i.i1673.preheader ]
  %2162 = load i32, ptr addrspace(1) @__llvm_rpc_client, align 8, !tbaa !68
  %cmp.not.i.i.i1679 = icmp ult i32 %index.0.i.i.i1678, %2162
  %spec.store.select.i.i.i1680 = select i1 %cmp.not.i.i.i1679, i32 %index.0.i.i.i1678, i32 0
  %2163 = tail call noundef i64 @llvm.amdgcn.ballot.i64(i1 true)
  %2164 = and i32 %spec.store.select.i.i.i1680, 31
  %2165 = lshr i64 %2163, %sh_prom.i.i.i.i
  %2166 = trunc i64 %2165 to i32
  %conv4.i.i.i.i.i1684 = and i32 %2166, 1
  %shl.i.i.i.i.i1685 = shl nuw i32 %conv4.i.i.i.i.i1684, %2164
  %2167 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2643, !amdgpu.uniform !42

2168:                                             ; preds = %ComputeEnd2644
  %sunkaddr3324 = getelementptr inbounds i8, ptr addrspace(1) %3849, i64 40
  %2169 = atomicrmw or ptr addrspace(1) %sunkaddr3324, i32 %.lcssa4004 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %2170, !amdgpu.uniform !42

2170:                                             ; preds = %2168, %ComputeEnd2644
  %2171 = phi i32 [ %2169, %2168 ], [ poison, %ComputeEnd2644 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %3858)
  %2172 = call i32 @llvm.amdgcn.readfirstlane.i32(i32 %2171)
  %2173 = or i32 %2172, %.lcssa4005
  %shl5.i.i.i.i.i1686 = shl nuw i32 1, %2164
  %and.i.i.i.i.i1687 = and i32 %shl5.i.i.i.i.i1686, %2173
  %tobool3.i.i.i.i1688 = icmp ne i32 %and.i.i.i.i.i1687, 0
  %2174 = tail call i64 @llvm.amdgcn.ballot.i64(i1 %tobool3.i.i.i.i1688)
  %and.i.i.i.i.i.i1689 = and i64 %2174, %2163
  %cmp.i.not.i.i.i1690 = icmp ne i64 %2163, %and.i.i.i.i.i.i1689
  br i1 %cmp.i.not.i.i.i1690, label %if.then.i.i.i.i1691, label %Flow3482, !amdgpu.uniform !42

if.then.i.i.i.i1691:                              ; preds = %2170
  fence syncscope("agent") acquire
  %2175 = load <2 x i64>, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !75
  %2176 = extractelement <2 x i64> %2175, i32 0
  %2177 = inttoptr i64 %2176 to ptr
  %2178 = extractelement <2 x i64> %2175, i32 1
  %2179 = inttoptr i64 %2178 to ptr
  %idxprom.i.i.i.i1692 = zext i32 %spec.store.select.i.i.i1680 to i64
  %arrayidx.i.i.i.i1693 = getelementptr inbounds nuw i32, ptr %2177, i64 %idxprom.i.i.i.i1692, !amdgpu.uniform !42
  %2180 = load atomic i32, ptr %arrayidx.i.i.i.i1693 monotonic, align 4
  %2181 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %2180)
  %arrayidx.i45.i.i.i1694 = getelementptr inbounds nuw i32, ptr %2179, i64 %idxprom.i.i.i.i1692, !amdgpu.uniform !42
  %2182 = load atomic i32, ptr %arrayidx.i45.i.i.i1694 monotonic, align 4
  %2183 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %2182)
  %cmp.i46.not.i.i.i1695 = icmp ne i32 %2181, %2183
  br i1 %cmp.i46.not.i.i.i1695, label %if.then12.i.i.i1696, label %Flow3480, !amdgpu.uniform !42

if.then12.i.i.i1696:                              ; preds = %if.then.i.i.i.i1691
  fence syncscope("agent") release
  %2184 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %2163, i1 true)
  %iszero.i.i.i.i.i.i.i1697 = icmp ne i64 %2163, 0
  %cmp2.i.i.i.i.i.i1698 = icmp eq i64 %2184, %sh_prom.i.i.i.i
  %cmp.i.i.i.i.i.i1699 = select i1 %iszero.i.i.i.i.i.i.i1697, i1 %cmp2.i.i.i.i.i.i1698, i1 false
  %conv4.i.i51.i.i.i1700 = zext i1 %cmp.i.i.i.i.i.i1699 to i32
  %shl.i.i52.i.i.i1701 = shl nuw i32 %conv4.i.i51.i.i.i1700, %2164
  %xor.i.i.i.i.i1702 = xor i32 %shl.i.i52.i.i.i1701, -1
  %2185 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2648, !amdgpu.uniform !42

2186:                                             ; preds = %ComputeEnd2649
  %sunkaddr3325 = getelementptr inbounds i8, ptr addrspace(1) %3849, i64 40
  %2187 = atomicrmw and ptr addrspace(1) %sunkaddr3325, i32 %.lcssa4006 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %2188, !amdgpu.uniform !42

2188:                                             ; preds = %2186, %ComputeEnd2649
  call void @llvm.amdgcn.end.cf.i64(i64 %3876)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %Flow3480, !amdgpu.uniform !42

if.end14.i.i.i1762:                               ; preds = %Flow3480
  %2189 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %2163, i1 true)
  %iszero.i.i.i.i.i.i1763 = icmp ne i64 %2163, 0
  %cmp2.i.i.i.i.i1764 = icmp eq i64 %2189, %sh_prom.i.i.i.i
  %cmp.i.i.i.i.i1765 = select i1 %iszero.i.i.i.i.i.i1763, i1 %cmp2.i.i.i.i.i1764, i1 false
  %2190 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.i.i.i.i.i1765)
  %2191 = extractvalue { i1, i64 } %2190, 0
  %2192 = extractvalue { i1, i64 } %2190, 1
  br i1 %2191, label %if.then16.i.i.i1767, label %if.end22.i.i.i1766

if.then16.i.i.i1767:                              ; preds = %if.end14.i.i.i1762
  %2193 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !76
  %opcode.i.i.i1768 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %2193, i64 %idxprom.i.i.i.i1692, i32 1
  store i32 1660944387, ptr %opcode.i.i.i1768, align 8, !tbaa !77
  %arrayidx21.i.i.i1769 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %2193, i64 %idxprom.i.i.i.i1692
  store i64 %2163, ptr %arrayidx21.i.i.i1769, align 8, !tbaa !80
  br label %if.end22.i.i.i1766, !amdgpu.uniform !42

Flow3481:                                         ; preds = %if.end22.i.i.i1766, %Flow3480
  %2194 = phi i32 [ %2181, %if.end22.i.i.i1766 ], [ %retval.sroa.8.0.i.i.i1677, %Flow3480 ]
  %2195 = phi i64 [ %2163, %if.end22.i.i.i1766 ], [ %retval.sroa.2.0.i.i.i1676, %Flow3480 ]
  %2196 = phi i32 [ %spec.store.select.i.i.i1680, %if.end22.i.i.i1766 ], [ %retval.sroa.6.0.i.i.i1674, %Flow3480 ]
  br label %cleanup26.i.i.i1703, !amdgpu.uniform !42

if.end22.i.i.i1766:                               ; preds = %if.then16.i.i.i1767, %if.end14.i.i.i1762
  call void @llvm.amdgcn.end.cf.i64(i64 %2192)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %Flow3481, !amdgpu.uniform !42

Flow3482:                                         ; preds = %cleanup26.i.i.i1703, %2170
  %2197 = phi i32 [ %2196, %cleanup26.i.i.i1703 ], [ %2159, %2170 ]
  %2198 = phi i64 [ %2195, %cleanup26.i.i.i1703 ], [ %2160, %2170 ]
  %2199 = phi i32 [ %2194, %cleanup26.i.i.i1703 ], [ %2161, %2170 ]
  %2200 = phi i32 [ %2194, %cleanup26.i.i.i1703 ], [ %retval.sroa.8.0.i.i.i1677, %2170 ]
  %2201 = phi i64 [ %2195, %cleanup26.i.i.i1703 ], [ %retval.sroa.2.0.i.i.i1676, %2170 ]
  %2202 = phi i32 [ %2196, %cleanup26.i.i.i1703 ], [ %retval.sroa.6.0.i.i.i1674, %2170 ]
  %2203 = phi i1 [ %cmp.i46.not.i.i.i1695, %cleanup26.i.i.i1703 ], [ true, %2170 ]
  br i1 %2203, label %for.inc.i.i.i1708, label %Flow3483, !amdgpu.uniform !42

cleanup26.i.i.i1703:                              ; preds = %Flow3481
  br label %Flow3482, !amdgpu.uniform !42

for.inc.i.i.i1708:                                ; preds = %Flow3482
  %inc.i.i.i1713 = add i32 %spec.store.select.i.i.i1680, 1
  br label %Flow3483, !amdgpu.uniform !42

Flow3483:                                         ; preds = %for.inc.i.i.i1708, %Flow3482
  %2204 = phi i32 [ %inc.i.i.i1713, %for.inc.i.i.i1708 ], [ poison, %Flow3482 ]
  %2205 = phi i1 [ false, %for.inc.i.i.i1708 ], [ true, %Flow3482 ]
  br i1 %2205, label %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1714, label %for.cond.i.i.i1673, !amdgpu.uniform !42

_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1714: ; preds = %Flow3483
  %.lcssa4009 = phi i32 [ %2197, %Flow3483 ]
  %.lcssa4008 = phi i64 [ %2198, %Flow3483 ]
  %.lcssa4007 = phi i32 [ %2199, %Flow3483 ]
  fence acquire
  %2206 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %mul.i.i.i.i.i.i1715 = shl i32 %.lcssa4009, 6
  %idxprom.i19.i.i.i.i.i1716 = zext i32 %mul.i.i.i.i.i.i1715 to i64
  %arrayidx.i20.i.i.i.i.i1717 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %2206, i64 %idxprom.i19.i.i.i.i.i1716
  %arrayidx.i22.i.i.i.i.i1718 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %arrayidx.i20.i.i.i.i.i1717, i64 %sh_prom.i.i.i.i
  store i64 %spec.select.i, ptr %arrayidx.i22.i.i.i.i.i1718, align 8, !tbaa !82
  br i1 %656, label %loop-memcpy-expansion2206.preheader, label %Flow3479, !amdgpu.uniform !42

loop-memcpy-expansion2206.preheader:              ; preds = %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1714
  %2207 = add nuw nsw i64 %75, 8
  %scevgep2917 = getelementptr i8, ptr %2206, i64 %2207
  %2208 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i1716, 6
  %scevgep2918 = getelementptr i8, ptr %scevgep2917, i64 %2208
  br label %loop-memcpy-expansion2206, !amdgpu.uniform !42

Flow3479:                                         ; preds = %Flow3478, %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1714
  br label %loop-memcpy-residual-header2209, !amdgpu.uniform !42

loop-memcpy-expansion2206:                        ; preds = %loop-memcpy-expansion2206.preheader, %loop-memcpy-expansion2206
  %loop-index2207 = phi i64 [ %2211, %loop-memcpy-expansion2206 ], [ 0, %loop-memcpy-expansion2206.preheader ]
  %2209 = getelementptr i8, ptr %2150, i64 %loop-index2207, !amdgpu.uniform !42
  %2210 = load <4 x i32>, ptr %2209, align 1
  %scevgep2919 = getelementptr i8, ptr %scevgep2918, i64 %loop-index2207
  store <4 x i32> %2210, ptr %scevgep2919, align 1
  %2211 = add i64 %loop-index2207, 16
  %2212 = icmp uge i64 %2211, %655
  br i1 %2212, label %Flow3478, label %loop-memcpy-expansion2206, !amdgpu.uniform !42

Flow3476:                                         ; preds = %Flow3473, %Flow3475
  %2213 = phi i64 [ %2731, %Flow3473 ], [ %2725, %Flow3475 ]
  br label %post-loop-memcpy-expansion2205, !amdgpu.uniform !42

loop-memcpy-residual2208:                         ; preds = %loop-memcpy-residual2208.preheader, %loop-memcpy-residual2208
  %residual-loop-index2210 = phi i64 [ %2215, %loop-memcpy-residual2208 ], [ 0, %loop-memcpy-residual2208.preheader ]
  %scevgep2916 = getelementptr i8, ptr %scevgep2915, i64 %residual-loop-index2210, !amdgpu.uniform !42
  %2214 = load i8, ptr %scevgep2916, align 1
  %scevgep2914 = getelementptr i8, ptr %scevgep2913, i64 %residual-loop-index2210
  store i8 %2214, ptr %scevgep2914, align 1
  %2215 = add i64 %residual-loop-index2210, 1
  %2216 = icmp uge i64 %2215, %654
  br i1 %2216, label %Flow3473, label %loop-memcpy-residual2208, !amdgpu.uniform !42

Flow3473:                                         ; preds = %loop-memcpy-residual2208
  br label %Flow3476, !amdgpu.uniform !42

post-loop-memcpy-expansion2205:                   ; preds = %Flow3476
  %tobool.not.i.i.i.i.i.i1721 = icmp eq i32 %.lcssa4007, 0
  %conv.i.i.i.i.i.i1722 = zext i1 %tobool.not.i.i.i.i.i.i1721 to i32
  fence release
  %2217 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %idxprom.i23.i.i.i.i.i1723 = zext i32 %.lcssa4009 to i64
  %arrayidx.i24.i.i.i.i.i1724 = getelementptr inbounds nuw i32, ptr %2217, i64 %idxprom.i23.i.i.i.i.i1723
  store atomic i32 %conv.i.i.i.i.i.i1722, ptr %arrayidx.i24.i.i.i.i.i1724 monotonic, align 4
  %2218 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !84
  %arrayidx.i.i5.i.i1725 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %2218, i64 %idxprom.i23.i.i.i.i.i1723, !amdgpu.uniform !42
  %2219 = load i64, ptr %arrayidx.i.i5.i.i1725, align 8, !tbaa !80
  %scevgep2898 = getelementptr i8, ptr %2150, i64 56
  br label %while.cond.i.i.i.i1726, !amdgpu.uniform !42

while.cond.i.i.i.i1726:                           ; preds = %post-loop-memcpy-expansion2205, %Flow3471
  %phi.broken3947 = phi i64 [ 0, %post-loop-memcpy-expansion2205 ], [ %2236, %Flow3471 ]
  %lsr.iv2905 = phi i64 [ %2234, %Flow3471 ], [ %670, %post-loop-memcpy-expansion2205 ]
  %lsr.iv2899 = phi ptr [ %2233, %Flow3471 ], [ %scevgep2898, %post-loop-memcpy-expansion2205 ]
  %port.sroa.37.0.i.i1727 = phi i32 [ %2232, %Flow3471 ], [ %conv.i.i.i.i.i.i1722, %post-loop-memcpy-expansion2205 ]
  %idx.0.i.i.i.i1728 = phi i64 [ %2231, %Flow3471 ], [ 56, %post-loop-memcpy-expansion2205 ]
  %cmp.i.i.i.i1729 = icmp ugt i64 %spec.select.i, %idx.0.i.i.i.i1728
  %2220 = tail call i64 @llvm.amdgcn.ballot.i64(i1 %cmp.i.i.i.i1729)
  %and.i.i.i.i6.i.i1730 = and i64 %2220, %2219
  %tobool.not.i.i.i.i1731 = icmp ne i64 %and.i.i.i.i6.i.i1730, 0
  %2221 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i22.i.i.i1732 = getelementptr inbounds nuw i32, ptr %2221, i64 %idxprom.i23.i.i.i.i.i1723, !amdgpu.uniform !42
  %2222 = load atomic i32, ptr %arrayidx.i22.i.i.i1732 monotonic, align 4
  %2223 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %tobool.not.i.i.i.i1731)
  %2224 = extractvalue { i1, i64 } %2223, 0
  %2225 = extractvalue { i1, i64 } %2223, 1
  br i1 %2224, label %cond.false.i8.i.i.i.i1733, label %Flow3471

cond.false.i8.i.i.i.i1733:                        ; preds = %while.cond.i.i.i.i1726
  %2226 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %2222)
  %2227 = icmp ne i32 %2226, %port.sroa.37.0.i.i1727
  br i1 %2227, label %while.body.i.i21.i.i.i.i1734.preheader, label %Flow3470, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i1734.preheader:           ; preds = %cond.false.i8.i.i.i.i1733
  br label %while.body.i.i21.i.i.i.i1734, !amdgpu.uniform !42

Flow3470:                                         ; preds = %Flow3468, %cond.false.i8.i.i.i.i1733
  br label %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i1737, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i1734:                     ; preds = %while.body.i.i21.i.i.i.i1734.preheader, %while.body.i.i21.i.i.i.i1734
  tail call void @llvm.amdgcn.s.sleep(i32 2)
  %2228 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i22.i.i.i.i1735 = getelementptr inbounds nuw i32, ptr %2228, i64 %idxprom.i23.i.i.i.i.i1723, !amdgpu.uniform !42
  %2229 = load atomic i32, ptr %arrayidx.i.i.i22.i.i.i.i1735 monotonic, align 4
  %2230 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %2229)
  %cmp.i.not.i.i23.i.i.i.i1736 = icmp eq i32 %2230, %port.sroa.37.0.i.i1727
  br i1 %cmp.i.not.i.i23.i.i.i.i1736, label %Flow3468, label %while.body.i.i21.i.i.i.i1734, !amdgpu.uniform !42

Flow3468:                                         ; preds = %while.body.i.i21.i.i.i.i1734
  br label %Flow3470, !amdgpu.uniform !42

Flow3471:                                         ; preds = %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1738, %while.cond.i.i.i.i1726
  %2231 = phi i64 [ %add.i.i.i.i1741, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1738 ], [ poison, %while.cond.i.i.i.i1726 ]
  %2232 = phi i32 [ %conv.i.i26.i.i.i.i1739, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1738 ], [ poison, %while.cond.i.i.i.i1726 ]
  %2233 = phi ptr [ %scevgep2900, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1738 ], [ poison, %while.cond.i.i.i.i1726 ]
  %2234 = phi i64 [ %lsr.iv.next2906, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1738 ], [ poison, %while.cond.i.i.i.i1726 ]
  %2235 = phi i1 [ false, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1738 ], [ true, %while.cond.i.i.i.i1726 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %2225)
  %2236 = call i64 @llvm.amdgcn.if.break.i64(i1 %2235, i64 %phi.broken3947)
  %2237 = call i1 @llvm.amdgcn.loop.i64(i64 %2236)
  br i1 %2237, label %cond.false.i.i.i1748, label %while.cond.i.i.i.i1726

_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i1737: ; preds = %Flow3470
  fence acquire
  br i1 %cmp.i.i.i.i1729, label %if.then.i.i.i.i.i.i.i1742, label %Flow3467, !amdgpu.uniform !42

if.then.i.i.i.i.i.i.i1742:                        ; preds = %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i1737
  %sub.i.i.i.i.i.i.i1744 = sub nuw i64 %spec.select.i, %idx.0.i.i.i.i1728
  %spec.select.i.i.i36.i.i.i.i1745 = tail call i64 @llvm.umin.i64(i64 %sub.i.i.i.i.i.i.i1744, i64 64)
  %2238 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %2239 = and i64 %spec.select.i.i.i36.i.i.i.i1745, 15
  %2240 = sub i64 %spec.select.i.i.i36.i.i.i.i1745, %2239
  %2241 = icmp ne i64 %2240, 0
  br i1 %2241, label %loop-memcpy-expansion2200.preheader, label %Flow3466, !amdgpu.uniform !42

loop-memcpy-expansion2200.preheader:              ; preds = %if.then.i.i.i.i.i.i.i1742
  %arrayidx.i20.i35.i.i.i.i1746 = getelementptr %"struct.rpc::Buffer", ptr %2238, i64 %idxprom.i19.i.i.i.i.i1716
  %arrayidx.i22.i38.i.i.i.i1747 = getelementptr %"struct.rpc::Buffer", ptr %arrayidx.i20.i35.i.i.i.i1746, i64 %sh_prom.i.i.i.i
  br label %loop-memcpy-expansion2200, !amdgpu.uniform !42

Flow3466:                                         ; preds = %Flow3465, %if.then.i.i.i.i.i.i.i1742
  br label %loop-memcpy-residual-header2203, !amdgpu.uniform !42

loop-memcpy-expansion2200:                        ; preds = %loop-memcpy-expansion2200.preheader, %loop-memcpy-expansion2200
  %loop-index2201 = phi i64 [ %2244, %loop-memcpy-expansion2200 ], [ 0, %loop-memcpy-expansion2200.preheader ]
  %scevgep2901 = getelementptr i8, ptr %lsr.iv2899, i64 %loop-index2201, !amdgpu.uniform !42
  %2242 = load <4 x i32>, ptr %scevgep2901, align 1
  %2243 = getelementptr i8, ptr %arrayidx.i22.i38.i.i.i.i1747, i64 %loop-index2201
  store <4 x i32> %2242, ptr %2243, align 1
  %2244 = add i64 %loop-index2201, 16
  %2245 = icmp uge i64 %2244, %2240
  br i1 %2245, label %Flow3465, label %loop-memcpy-expansion2200, !amdgpu.uniform !42

Flow3463:                                         ; preds = %Flow3462, %loop-memcpy-residual-header2203
  br label %Flow3467, !amdgpu.uniform !42

loop-memcpy-residual2202:                         ; preds = %loop-memcpy-residual2202.preheader, %loop-memcpy-residual2202
  %residual-loop-index2204 = phi i64 [ %2247, %loop-memcpy-residual2202 ], [ 0, %loop-memcpy-residual2202.preheader ]
  %scevgep2911 = getelementptr i8, ptr %scevgep2910, i64 %residual-loop-index2204, !amdgpu.uniform !42
  %2246 = load i8, ptr %scevgep2911, align 1
  %scevgep2909 = getelementptr i8, ptr %scevgep2908, i64 %residual-loop-index2204
  store i8 %2246, ptr %scevgep2909, align 1
  %2247 = add i64 %residual-loop-index2204, 1
  %2248 = icmp uge i64 %2247, %2239
  br i1 %2248, label %Flow3462, label %loop-memcpy-residual2202, !amdgpu.uniform !42

Flow3462:                                         ; preds = %loop-memcpy-residual2202
  br label %Flow3463, !amdgpu.uniform !42

_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1738: ; preds = %Flow3467
  %conv.i.i26.i.i.i.i1739 = xor i32 %port.sroa.37.0.i.i1727, 1
  fence release
  %2249 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %arrayidx.i24.i29.i.i.i.i1740 = getelementptr inbounds nuw i32, ptr %2249, i64 %idxprom.i23.i.i.i.i.i1723
  store atomic i32 %conv.i.i26.i.i.i.i1739, ptr %arrayidx.i24.i29.i.i.i.i1740 monotonic, align 4
  %add.i.i.i.i1741 = add i64 %idx.0.i.i.i.i1728, 64
  %scevgep2900 = getelementptr i8, ptr %lsr.iv2899, i64 64
  %lsr.iv.next2906 = add i64 %lsr.iv2905, -64
  br label %Flow3471, !amdgpu.uniform !42

cond.false.i.i.i1748:                             ; preds = %Flow3471
  %.lcssa4003 = phi i64 [ %2236, %Flow3471 ]
  %port.sroa.37.0.i.i1727.lcssa = phi i32 [ %port.sroa.37.0.i.i1727, %Flow3471 ]
  %.lcssa4002 = phi i32 [ %2222, %Flow3471 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa4003)
  %2250 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %.lcssa4002)
  %2251 = icmp ne i32 %2250, %port.sroa.37.0.i.i1727.lcssa
  %2252 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %2251)
  %2253 = extractvalue { i1, i64 } %2252, 0
  %2254 = extractvalue { i1, i64 } %2252, 1
  br i1 %2253, label %while.body.i.i.i.i1749.preheader, label %Flow3460

while.body.i.i.i.i1749.preheader:                 ; preds = %cond.false.i.i.i1748
  br label %while.body.i.i.i.i1749, !amdgpu.uniform !42

Flow3460:                                         ; preds = %Flow3458, %cond.false.i.i.i1748
  call void @llvm.amdgcn.end.cf.i64(i64 %2254)
  br label %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsINS_3cpp6stringEEERS1_T_.exit1770, !amdgpu.uniform !42

while.body.i.i.i.i1749:                           ; preds = %while.body.i.i.i.i1749.preheader, %while.body.i.i.i.i1749
  %phi.broken3948 = phi i64 [ 0, %while.body.i.i.i.i1749.preheader ], [ %2258, %while.body.i.i.i.i1749 ]
  tail call void @llvm.amdgcn.s.sleep(i32 2)
  %2255 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i8.i.i1750 = getelementptr inbounds nuw i32, ptr %2255, i64 %idxprom.i23.i.i.i.i.i1723, !amdgpu.uniform !42
  %2256 = load atomic i32, ptr %arrayidx.i.i.i8.i.i1750 monotonic, align 4
  %2257 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %2256)
  %cmp.i.not.i.i.i.i1751 = icmp eq i32 %2257, %port.sroa.37.0.i.i1727.lcssa
  %2258 = call i64 @llvm.amdgcn.if.break.i64(i1 %cmp.i.not.i.i.i.i1751, i64 %phi.broken3948)
  %2259 = call i1 @llvm.amdgcn.loop.i64(i64 %2258)
  br i1 %2259, label %Flow3458, label %while.body.i.i.i.i1749

Flow3458:                                         ; preds = %while.body.i.i.i.i1749
  %.lcssa4001 = phi i64 [ %2258, %while.body.i.i.i.i1749 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa4001)
  br label %Flow3460, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsINS_3cpp6stringEEERS1_T_.exit1770: ; preds = %Flow3460
  fence acquire
  tail call void @llvm.amdgcn.wave.barrier()
  fence syncscope("agent") release
  %2260 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %.lcssa4008, i1 true)
  %iszero.i.i.i.i.i13.i.i1753 = icmp ne i64 %.lcssa4008, 0
  %cmp2.i.i.i.i14.i.i1754 = icmp eq i64 %2260, %sh_prom.i.i.i.i
  %cmp.i.i.i.i15.i.i1755 = select i1 %iszero.i.i.i.i.i13.i.i1753, i1 %cmp2.i.i.i.i14.i.i1754, i1 false
  %2261 = and i32 %.lcssa4009, 31
  %conv4.i.i.i19.i.i1759 = zext i1 %cmp.i.i.i.i15.i.i1755 to i32
  %shl.i.i.i20.i.i1760 = shl nuw i32 %conv4.i.i.i19.i.i1759, %2261
  %xor.i.i.i21.i.i1761 = xor i32 %shl.i.i.i20.i.i1760, -1
  %2262 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2652, !amdgpu.uniform !42

2263:                                             ; preds = %ComputeEnd2653
  %div8.i.i.i16.i.i1756 = lshr i32 %.lcssa4009, 5
  %idxprom.i.i.i17.i.i1757 = zext nneg i32 %div8.i.i.i16.i.i1756 to i64
  %2264 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i17.i.i1757
  %sunkaddr3326 = getelementptr inbounds i8, ptr addrspace(1) %2264, i64 40
  %2265 = atomicrmw and ptr addrspace(1) %sunkaddr3326, i32 %.lcssa4000 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %2266, !amdgpu.uniform !42

2266:                                             ; preds = %2263, %ComputeEnd2653
  call void @llvm.amdgcn.end.cf.i64(i64 %3893)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %for.cond.i.i1.i1779, !amdgpu.uniform !42

for.cond.i.i1.i1779:                              ; preds = %2266, %Flow3457
  %2267 = phi i32 [ %2305, %Flow3457 ], [ poison, %2266 ]
  %2268 = phi i64 [ %2306, %Flow3457 ], [ poison, %2266 ]
  %2269 = phi i32 [ %2307, %Flow3457 ], [ poison, %2266 ]
  %retval.sroa.6.0.i.i.i1780 = phi i32 [ %2310, %Flow3457 ], [ undef, %2266 ]
  %retval.sroa.2.0.i.i.i1782 = phi i64 [ %2309, %Flow3457 ], [ undef, %2266 ]
  %retval.sroa.8.0.i.i.i1783 = phi i32 [ %2308, %Flow3457 ], [ undef, %2266 ]
  %index.0.i.i.i1784 = phi i32 [ %2312, %Flow3457 ], [ 0, %2266 ]
  %2270 = load i32, ptr addrspace(1) @__llvm_rpc_client, align 8, !tbaa !68
  %cmp.not.i.i.i1785 = icmp ult i32 %index.0.i.i.i1784, %2270
  %spec.store.select.i.i.i1786 = select i1 %cmp.not.i.i.i1785, i32 %index.0.i.i.i1784, i32 0
  %2271 = tail call noundef i64 @llvm.amdgcn.ballot.i64(i1 true)
  %2272 = and i32 %spec.store.select.i.i.i1786, 31
  %2273 = lshr i64 %2271, %sh_prom.i.i.i.i
  %2274 = trunc i64 %2273 to i32
  %conv4.i.i.i.i.i1790 = and i32 %2274, 1
  %shl.i.i.i.i.i1791 = shl nuw i32 %conv4.i.i.i.i.i1790, %2272
  %2275 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2656, !amdgpu.uniform !42

2276:                                             ; preds = %ComputeEnd2657
  %sunkaddr3327 = getelementptr inbounds i8, ptr addrspace(1) %3903, i64 40
  %2277 = atomicrmw or ptr addrspace(1) %sunkaddr3327, i32 %.lcssa3994 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %2278, !amdgpu.uniform !42

2278:                                             ; preds = %2276, %ComputeEnd2657
  %2279 = phi i32 [ %2277, %2276 ], [ poison, %ComputeEnd2657 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %3912)
  %2280 = call i32 @llvm.amdgcn.readfirstlane.i32(i32 %2279)
  %2281 = or i32 %2280, %.lcssa3995
  %shl5.i.i.i.i.i1792 = shl nuw i32 1, %2272
  %and.i.i.i.i.i1793 = and i32 %shl5.i.i.i.i.i1792, %2281
  %tobool3.i.i.i.i1794 = icmp ne i32 %and.i.i.i.i.i1793, 0
  %2282 = tail call i64 @llvm.amdgcn.ballot.i64(i1 %tobool3.i.i.i.i1794)
  %and.i.i.i.i.i.i1795 = and i64 %2282, %2271
  %cmp.i.not.i.i.i1796 = icmp ne i64 %2271, %and.i.i.i.i.i.i1795
  br i1 %cmp.i.not.i.i.i1796, label %if.then.i.i.i.i1797, label %Flow3456, !amdgpu.uniform !42

if.then.i.i.i.i1797:                              ; preds = %2278
  fence syncscope("agent") acquire
  %2283 = load <2 x i64>, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !75
  %2284 = extractelement <2 x i64> %2283, i32 0
  %2285 = inttoptr i64 %2284 to ptr
  %2286 = extractelement <2 x i64> %2283, i32 1
  %2287 = inttoptr i64 %2286 to ptr
  %idxprom.i.i.i.i1798 = zext i32 %spec.store.select.i.i.i1786 to i64
  %arrayidx.i.i.i.i1799 = getelementptr inbounds nuw i32, ptr %2285, i64 %idxprom.i.i.i.i1798, !amdgpu.uniform !42
  %2288 = load atomic i32, ptr %arrayidx.i.i.i.i1799 monotonic, align 4
  %2289 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %2288)
  %arrayidx.i45.i.i.i1800 = getelementptr inbounds nuw i32, ptr %2287, i64 %idxprom.i.i.i.i1798, !amdgpu.uniform !42
  %2290 = load atomic i32, ptr %arrayidx.i45.i.i.i1800 monotonic, align 4
  %2291 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %2290)
  %cmp.i46.not.i.i.i1801 = icmp ne i32 %2289, %2291
  br i1 %cmp.i46.not.i.i.i1801, label %if.then12.i.i.i1802, label %Flow3454, !amdgpu.uniform !42

if.then12.i.i.i1802:                              ; preds = %if.then.i.i.i.i1797
  fence syncscope("agent") release
  %2292 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %2271, i1 true)
  %iszero.i.i.i.i.i.i.i1803 = icmp ne i64 %2271, 0
  %cmp2.i.i.i.i.i.i1804 = icmp eq i64 %2292, %sh_prom.i.i.i.i
  %cmp.i.i.i.i.i.i1805 = select i1 %iszero.i.i.i.i.i.i.i1803, i1 %cmp2.i.i.i.i.i.i1804, i1 false
  %conv4.i.i51.i.i.i1806 = zext i1 %cmp.i.i.i.i.i.i1805 to i32
  %shl.i.i52.i.i.i1807 = shl nuw i32 %conv4.i.i51.i.i.i1806, %2272
  %xor.i.i.i.i.i1808 = xor i32 %shl.i.i52.i.i.i1807, -1
  %2293 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2661, !amdgpu.uniform !42

2294:                                             ; preds = %ComputeEnd2662
  %sunkaddr3328 = getelementptr inbounds i8, ptr addrspace(1) %3903, i64 40
  %2295 = atomicrmw and ptr addrspace(1) %sunkaddr3328, i32 %.lcssa3996 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %2296, !amdgpu.uniform !42

2296:                                             ; preds = %2294, %ComputeEnd2662
  call void @llvm.amdgcn.end.cf.i64(i64 %3930)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %Flow3454, !amdgpu.uniform !42

if.end14.i.i.i1868:                               ; preds = %Flow3454
  %2297 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %2271, i1 true)
  %iszero.i.i.i.i.i.i1869 = icmp ne i64 %2271, 0
  %cmp2.i.i.i.i.i1870 = icmp eq i64 %2297, %sh_prom.i.i.i.i
  %cmp.i.i.i.i.i1871 = select i1 %iszero.i.i.i.i.i.i1869, i1 %cmp2.i.i.i.i.i1870, i1 false
  %2298 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.i.i.i.i.i1871)
  %2299 = extractvalue { i1, i64 } %2298, 0
  %2300 = extractvalue { i1, i64 } %2298, 1
  br i1 %2299, label %if.then16.i.i.i1873, label %if.end22.i.i.i1872

if.then16.i.i.i1873:                              ; preds = %if.end14.i.i.i1868
  %2301 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !76
  %opcode.i.i.i1874 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %2301, i64 %idxprom.i.i.i.i1798, i32 1
  store i32 1660944387, ptr %opcode.i.i.i1874, align 8, !tbaa !77
  %arrayidx21.i.i.i1875 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %2301, i64 %idxprom.i.i.i.i1798
  store i64 %2271, ptr %arrayidx21.i.i.i1875, align 8, !tbaa !80
  br label %if.end22.i.i.i1872, !amdgpu.uniform !42

Flow3455:                                         ; preds = %if.end22.i.i.i1872, %Flow3454
  %2302 = phi i32 [ %2289, %if.end22.i.i.i1872 ], [ %retval.sroa.8.0.i.i.i1783, %Flow3454 ]
  %2303 = phi i64 [ %2271, %if.end22.i.i.i1872 ], [ %retval.sroa.2.0.i.i.i1782, %Flow3454 ]
  %2304 = phi i32 [ %spec.store.select.i.i.i1786, %if.end22.i.i.i1872 ], [ %retval.sroa.6.0.i.i.i1780, %Flow3454 ]
  br label %cleanup26.i.i.i1809, !amdgpu.uniform !42

if.end22.i.i.i1872:                               ; preds = %if.then16.i.i.i1873, %if.end14.i.i.i1868
  call void @llvm.amdgcn.end.cf.i64(i64 %2300)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %Flow3455, !amdgpu.uniform !42

Flow3456:                                         ; preds = %cleanup26.i.i.i1809, %2278
  %2305 = phi i32 [ %2304, %cleanup26.i.i.i1809 ], [ %2267, %2278 ]
  %2306 = phi i64 [ %2303, %cleanup26.i.i.i1809 ], [ %2268, %2278 ]
  %2307 = phi i32 [ %2302, %cleanup26.i.i.i1809 ], [ %2269, %2278 ]
  %2308 = phi i32 [ %2302, %cleanup26.i.i.i1809 ], [ %retval.sroa.8.0.i.i.i1783, %2278 ]
  %2309 = phi i64 [ %2303, %cleanup26.i.i.i1809 ], [ %retval.sroa.2.0.i.i.i1782, %2278 ]
  %2310 = phi i32 [ %2304, %cleanup26.i.i.i1809 ], [ %retval.sroa.6.0.i.i.i1780, %2278 ]
  %2311 = phi i1 [ %cmp.i46.not.i.i.i1801, %cleanup26.i.i.i1809 ], [ true, %2278 ]
  br i1 %2311, label %for.inc.i.i.i1814, label %Flow3457, !amdgpu.uniform !42

cleanup26.i.i.i1809:                              ; preds = %Flow3455
  br label %Flow3456, !amdgpu.uniform !42

for.inc.i.i.i1814:                                ; preds = %Flow3456
  %inc.i.i.i1819 = add i32 %spec.store.select.i.i.i1786, 1
  br label %Flow3457, !amdgpu.uniform !42

Flow3457:                                         ; preds = %for.inc.i.i.i1814, %Flow3456
  %2312 = phi i32 [ %inc.i.i.i1819, %for.inc.i.i.i1814 ], [ poison, %Flow3456 ]
  %2313 = phi i1 [ false, %for.inc.i.i.i1814 ], [ true, %Flow3456 ]
  br i1 %2313, label %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1820, label %for.cond.i.i1.i1779, !amdgpu.uniform !42

_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1820: ; preds = %Flow3457
  %.lcssa3999 = phi i32 [ %2305, %Flow3457 ]
  %.lcssa3998 = phi i64 [ %2306, %Flow3457 ]
  %.lcssa3997 = phi i32 [ %2307, %Flow3457 ]
  fence acquire
  %2314 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %mul.i.i.i.i.i.i1821 = shl i32 %.lcssa3999, 6
  %idxprom.i19.i.i.i.i.i1822 = zext i32 %mul.i.i.i.i.i.i1821 to i64
  %arrayidx.i20.i.i.i.i.i1823 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %2314, i64 %idxprom.i19.i.i.i.i.i1822
  %arrayidx.i22.i.i.i.i.i1824 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %arrayidx.i20.i.i.i.i.i1823, i64 %sh_prom.i.i.i.i
  store i64 sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.35 to ptr), i64 10) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.35 to ptr) to i64)), ptr %arrayidx.i22.i.i.i.i.i1824, align 8, !tbaa !82
  br i1 %1193, label %loop-memcpy-expansion2374.preheader, label %Flow3453, !amdgpu.uniform !42

loop-memcpy-expansion2374.preheader:              ; preds = %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1820
  %2315 = add nuw nsw i64 %75, 8
  %scevgep2895 = getelementptr i8, ptr %2314, i64 %2315
  %2316 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i1822, 6
  %scevgep2896 = getelementptr i8, ptr %scevgep2895, i64 %2316
  br label %loop-memcpy-expansion2374, !amdgpu.uniform !42

Flow3453:                                         ; preds = %Flow3452, %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1820
  br label %loop-memcpy-residual-header2377, !amdgpu.uniform !42

loop-memcpy-expansion2374:                        ; preds = %loop-memcpy-expansion2374.preheader, %loop-memcpy-expansion2374
  %loop-index2375 = phi i64 [ %2319, %loop-memcpy-expansion2374 ], [ 0, %loop-memcpy-expansion2374.preheader ]
  %2317 = getelementptr i8, ptr addrspace(4) @.str.35, i64 %loop-index2375, !amdgpu.uniform !42
  %2318 = load <4 x i32>, ptr addrspace(4) %2317, align 1
  %scevgep2897 = getelementptr i8, ptr %scevgep2896, i64 %loop-index2375
  store <4 x i32> %2318, ptr %scevgep2897, align 1
  %2319 = add i64 %loop-index2375, 16
  %2320 = icmp uge i64 %2319, %1192
  br i1 %2320, label %Flow3452, label %loop-memcpy-expansion2374, !amdgpu.uniform !42

Flow3450:                                         ; preds = %Flow3448, %Flow3449
  %2321 = phi i64 [ %2888, %Flow3448 ], [ %2885, %Flow3449 ]
  br label %post-loop-memcpy-expansion2373, !amdgpu.uniform !42

loop-memcpy-residual2376:                         ; preds = %loop-memcpy-residual2376.preheader, %loop-memcpy-residual2376
  %residual-loop-index2378 = phi i64 [ %2323, %loop-memcpy-residual2376 ], [ 0, %loop-memcpy-residual2376.preheader ]
  %scevgep2894 = getelementptr i8, ptr addrspace(4) @.str.35, i64 %residual-loop-index2378, !amdgpu.uniform !42
  %2322 = load i8, ptr addrspace(4) %scevgep2894, align 1
  %scevgep2893 = getelementptr i8, ptr %scevgep2892, i64 %residual-loop-index2378
  store i8 %2322, ptr %scevgep2893, align 1
  %2323 = add i64 %residual-loop-index2378, 1
  %2324 = icmp uge i64 %2323, %1191
  br i1 %2324, label %Flow3448, label %loop-memcpy-residual2376, !amdgpu.uniform !42

Flow3448:                                         ; preds = %loop-memcpy-residual2376
  br label %Flow3450, !amdgpu.uniform !42

post-loop-memcpy-expansion2373:                   ; preds = %Flow3450
  %tobool.not.i.i.i.i.i.i1827 = icmp eq i32 %.lcssa3997, 0
  %conv.i.i.i.i.i.i1828 = zext i1 %tobool.not.i.i.i.i.i.i1827 to i32
  fence release
  %2325 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %idxprom.i23.i.i.i.i.i1829 = zext i32 %.lcssa3999 to i64
  %arrayidx.i24.i.i.i.i.i1830 = getelementptr inbounds nuw i32, ptr %2325, i64 %idxprom.i23.i.i.i.i.i1829
  store atomic i32 %conv.i.i.i.i.i.i1828, ptr %arrayidx.i24.i.i.i.i.i1830 monotonic, align 4
  %2326 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !84
  %arrayidx.i.i5.i.i1831 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %2326, i64 %idxprom.i23.i.i.i.i.i1829, !amdgpu.uniform !42
  %2327 = load i64, ptr %arrayidx.i.i5.i.i1831, align 8, !tbaa !80
  br label %while.cond.i.i.i.i1832, !amdgpu.uniform !42

while.cond.i.i.i.i1832:                           ; preds = %post-loop-memcpy-expansion2373, %Flow3447
  %phi.broken3949 = phi i64 [ 0, %post-loop-memcpy-expansion2373 ], [ %2344, %Flow3447 ]
  %lsr.iv2884 = phi i64 [ %2342, %Flow3447 ], [ -46, %post-loop-memcpy-expansion2373 ]
  %lsr.iv2879 = phi ptr addrspace(4) [ %2341, %Flow3447 ], [ getelementptr (i8, ptr addrspace(4) @.str.35, i64 56), %post-loop-memcpy-expansion2373 ]
  %port.sroa.37.0.i.i1833 = phi i32 [ %2340, %Flow3447 ], [ %conv.i.i.i.i.i.i1828, %post-loop-memcpy-expansion2373 ]
  %idx.0.i.i.i.i1834 = phi i64 [ %2339, %Flow3447 ], [ 56, %post-loop-memcpy-expansion2373 ]
  %cmp.i.i.i.i1835 = icmp ult i64 %idx.0.i.i.i.i1834, sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.35 to ptr), i64 10) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.35 to ptr) to i64))
  %2328 = tail call i64 @llvm.amdgcn.ballot.i64(i1 %cmp.i.i.i.i1835)
  %and.i.i.i.i6.i.i1836 = and i64 %2328, %2327
  %tobool.not.i.i.i.i1837 = icmp ne i64 %and.i.i.i.i6.i.i1836, 0
  %2329 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i22.i.i.i1838 = getelementptr inbounds nuw i32, ptr %2329, i64 %idxprom.i23.i.i.i.i.i1829, !amdgpu.uniform !42
  %2330 = load atomic i32, ptr %arrayidx.i22.i.i.i1838 monotonic, align 4
  %2331 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %tobool.not.i.i.i.i1837)
  %2332 = extractvalue { i1, i64 } %2331, 0
  %2333 = extractvalue { i1, i64 } %2331, 1
  br i1 %2332, label %cond.false.i8.i.i.i.i1839, label %Flow3447

cond.false.i8.i.i.i.i1839:                        ; preds = %while.cond.i.i.i.i1832
  %2334 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %2330)
  %2335 = icmp ne i32 %2334, %port.sroa.37.0.i.i1833
  br i1 %2335, label %while.body.i.i21.i.i.i.i1840.preheader, label %Flow3446, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i1840.preheader:           ; preds = %cond.false.i8.i.i.i.i1839
  br label %while.body.i.i21.i.i.i.i1840, !amdgpu.uniform !42

Flow3446:                                         ; preds = %Flow3445, %cond.false.i8.i.i.i.i1839
  br label %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i1843, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i1840:                     ; preds = %while.body.i.i21.i.i.i.i1840.preheader, %while.body.i.i21.i.i.i.i1840
  tail call void @llvm.amdgcn.s.sleep(i32 2)
  %2336 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i22.i.i.i.i1841 = getelementptr inbounds nuw i32, ptr %2336, i64 %idxprom.i23.i.i.i.i.i1829, !amdgpu.uniform !42
  %2337 = load atomic i32, ptr %arrayidx.i.i.i22.i.i.i.i1841 monotonic, align 4
  %2338 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %2337)
  %cmp.i.not.i.i23.i.i.i.i1842 = icmp eq i32 %2338, %port.sroa.37.0.i.i1833
  br i1 %cmp.i.not.i.i23.i.i.i.i1842, label %Flow3445, label %while.body.i.i21.i.i.i.i1840, !amdgpu.uniform !42

Flow3445:                                         ; preds = %while.body.i.i21.i.i.i.i1840
  br label %Flow3446, !amdgpu.uniform !42

Flow3447:                                         ; preds = %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1844, %while.cond.i.i.i.i1832
  %2339 = phi i64 [ %add.i.i.i.i1847, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1844 ], [ poison, %while.cond.i.i.i.i1832 ]
  %2340 = phi i32 [ %conv.i.i26.i.i.i.i1845, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1844 ], [ poison, %while.cond.i.i.i.i1832 ]
  %2341 = phi ptr addrspace(4) [ %scevgep2880, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1844 ], [ poison, %while.cond.i.i.i.i1832 ]
  %2342 = phi i64 [ %lsr.iv.next2885, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1844 ], [ poison, %while.cond.i.i.i.i1832 ]
  %2343 = phi i1 [ false, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1844 ], [ true, %while.cond.i.i.i.i1832 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %2333)
  %2344 = call i64 @llvm.amdgcn.if.break.i64(i1 %2343, i64 %phi.broken3949)
  %2345 = call i1 @llvm.amdgcn.loop.i64(i64 %2344)
  br i1 %2345, label %cond.false.i.i.i1854, label %while.cond.i.i.i.i1832

_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i1843: ; preds = %Flow3446
  fence acquire
  br i1 %cmp.i.i.i.i1835, label %if.then.i.i.i.i.i.i.i1848, label %Flow3444, !amdgpu.uniform !42

if.then.i.i.i.i.i.i.i1848:                        ; preds = %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i1843
  %sub.i.i.i.i.i.i.i1850 = sub nuw i64 sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.35 to ptr), i64 10) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.35 to ptr) to i64)), %idx.0.i.i.i.i1834
  %spec.select.i.i.i36.i.i.i.i1851 = tail call i64 @llvm.umin.i64(i64 %sub.i.i.i.i.i.i.i1850, i64 64)
  %2346 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %2347 = and i64 %spec.select.i.i.i36.i.i.i.i1851, 15
  %2348 = sub i64 %spec.select.i.i.i36.i.i.i.i1851, %2347
  %2349 = icmp ne i64 %2348, 0
  br i1 %2349, label %loop-memcpy-expansion2422.preheader, label %Flow3443, !amdgpu.uniform !42

loop-memcpy-expansion2422.preheader:              ; preds = %if.then.i.i.i.i.i.i.i1848
  %arrayidx.i20.i35.i.i.i.i1852 = getelementptr %"struct.rpc::Buffer", ptr %2346, i64 %idxprom.i19.i.i.i.i.i1822
  %arrayidx.i22.i38.i.i.i.i1853 = getelementptr %"struct.rpc::Buffer", ptr %arrayidx.i20.i35.i.i.i.i1852, i64 %sh_prom.i.i.i.i
  br label %loop-memcpy-expansion2422, !amdgpu.uniform !42

Flow3443:                                         ; preds = %Flow3442, %if.then.i.i.i.i.i.i.i1848
  br label %loop-memcpy-residual-header2425, !amdgpu.uniform !42

loop-memcpy-expansion2422:                        ; preds = %loop-memcpy-expansion2422.preheader, %loop-memcpy-expansion2422
  %loop-index2423 = phi i64 [ %2352, %loop-memcpy-expansion2422 ], [ 0, %loop-memcpy-expansion2422.preheader ]
  %scevgep2881 = getelementptr i8, ptr addrspace(4) %lsr.iv2879, i64 %loop-index2423, !amdgpu.uniform !42
  %2350 = load <4 x i32>, ptr addrspace(4) %scevgep2881, align 1
  %2351 = getelementptr i8, ptr %arrayidx.i22.i38.i.i.i.i1853, i64 %loop-index2423
  store <4 x i32> %2350, ptr %2351, align 1
  %2352 = add i64 %loop-index2423, 16
  %2353 = icmp uge i64 %2352, %2348
  br i1 %2353, label %Flow3442, label %loop-memcpy-expansion2422, !amdgpu.uniform !42

Flow3441:                                         ; preds = %Flow3440, %loop-memcpy-residual-header2425
  br label %Flow3444, !amdgpu.uniform !42

loop-memcpy-residual2424:                         ; preds = %loop-memcpy-residual2424.preheader, %loop-memcpy-residual2424
  %residual-loop-index2426 = phi i64 [ %2355, %loop-memcpy-residual2424 ], [ 0, %loop-memcpy-residual2424.preheader ]
  %scevgep2890 = getelementptr i8, ptr addrspace(4) %scevgep2889, i64 %residual-loop-index2426, !amdgpu.uniform !42
  %2354 = load i8, ptr addrspace(4) %scevgep2890, align 1
  %scevgep2888 = getelementptr i8, ptr %scevgep2887, i64 %residual-loop-index2426
  store i8 %2354, ptr %scevgep2888, align 1
  %2355 = add i64 %residual-loop-index2426, 1
  %2356 = icmp uge i64 %2355, %2347
  br i1 %2356, label %Flow3440, label %loop-memcpy-residual2424, !amdgpu.uniform !42

Flow3440:                                         ; preds = %loop-memcpy-residual2424
  br label %Flow3441, !amdgpu.uniform !42

_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1844: ; preds = %Flow3444
  %conv.i.i26.i.i.i.i1845 = xor i32 %port.sroa.37.0.i.i1833, 1
  fence release
  %2357 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %arrayidx.i24.i29.i.i.i.i1846 = getelementptr inbounds nuw i32, ptr %2357, i64 %idxprom.i23.i.i.i.i.i1829
  store atomic i32 %conv.i.i26.i.i.i.i1845, ptr %arrayidx.i24.i29.i.i.i.i1846 monotonic, align 4
  %add.i.i.i.i1847 = add i64 %idx.0.i.i.i.i1834, 64
  %scevgep2880 = getelementptr i8, ptr addrspace(4) %lsr.iv2879, i64 64
  %lsr.iv.next2885 = add i64 %lsr.iv2884, -64
  br label %Flow3447, !amdgpu.uniform !42

cond.false.i.i.i1854:                             ; preds = %Flow3447
  %.lcssa3993 = phi i64 [ %2344, %Flow3447 ]
  %port.sroa.37.0.i.i1833.lcssa = phi i32 [ %port.sroa.37.0.i.i1833, %Flow3447 ]
  %.lcssa3992 = phi i32 [ %2330, %Flow3447 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa3993)
  %2358 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %.lcssa3992)
  %2359 = icmp ne i32 %2358, %port.sroa.37.0.i.i1833.lcssa
  %2360 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %2359)
  %2361 = extractvalue { i1, i64 } %2360, 0
  %2362 = extractvalue { i1, i64 } %2360, 1
  br i1 %2361, label %while.body.i.i.i.i1855.preheader, label %Flow3439

while.body.i.i.i.i1855.preheader:                 ; preds = %cond.false.i.i.i1854
  br label %while.body.i.i.i.i1855, !amdgpu.uniform !42

Flow3439:                                         ; preds = %Flow3438, %cond.false.i.i.i1854
  call void @llvm.amdgcn.end.cf.i64(i64 %2362)
  br label %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit1876, !amdgpu.uniform !42

while.body.i.i.i.i1855:                           ; preds = %while.body.i.i.i.i1855.preheader, %while.body.i.i.i.i1855
  %phi.broken3950 = phi i64 [ 0, %while.body.i.i.i.i1855.preheader ], [ %2366, %while.body.i.i.i.i1855 ]
  tail call void @llvm.amdgcn.s.sleep(i32 2)
  %2363 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i8.i.i1856 = getelementptr inbounds nuw i32, ptr %2363, i64 %idxprom.i23.i.i.i.i.i1829, !amdgpu.uniform !42
  %2364 = load atomic i32, ptr %arrayidx.i.i.i8.i.i1856 monotonic, align 4
  %2365 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %2364)
  %cmp.i.not.i.i.i.i1857 = icmp eq i32 %2365, %port.sroa.37.0.i.i1833.lcssa
  %2366 = call i64 @llvm.amdgcn.if.break.i64(i1 %cmp.i.not.i.i.i.i1857, i64 %phi.broken3950)
  %2367 = call i1 @llvm.amdgcn.loop.i64(i64 %2366)
  br i1 %2367, label %Flow3438, label %while.body.i.i.i.i1855

Flow3438:                                         ; preds = %while.body.i.i.i.i1855
  %.lcssa3991 = phi i64 [ %2366, %while.body.i.i.i.i1855 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa3991)
  br label %Flow3439, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit1876: ; preds = %Flow3439
  fence acquire
  tail call void @llvm.amdgcn.wave.barrier()
  fence syncscope("agent") release
  %2368 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %.lcssa3998, i1 true)
  %iszero.i.i.i.i.i13.i.i1859 = icmp ne i64 %.lcssa3998, 0
  %cmp2.i.i.i.i14.i.i1860 = icmp eq i64 %2368, %sh_prom.i.i.i.i
  %cmp.i.i.i.i15.i.i1861 = select i1 %iszero.i.i.i.i.i13.i.i1859, i1 %cmp2.i.i.i.i14.i.i1860, i1 false
  %2369 = and i32 %.lcssa3999, 31
  %conv4.i.i.i19.i.i1865 = zext i1 %cmp.i.i.i.i15.i.i1861 to i32
  %shl.i.i.i20.i.i1866 = shl nuw i32 %conv4.i.i.i19.i.i1865, %2369
  %xor.i.i.i21.i.i1867 = xor i32 %shl.i.i.i20.i.i1866, -1
  %2370 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2665, !amdgpu.uniform !42

2371:                                             ; preds = %ComputeEnd2666
  %div8.i.i.i16.i.i1862 = lshr i32 %.lcssa3999, 5
  %idxprom.i.i.i17.i.i1863 = zext nneg i32 %div8.i.i.i16.i.i1862 to i64
  %2372 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i17.i.i1863
  %sunkaddr3329 = getelementptr inbounds i8, ptr addrspace(1) %2372, i64 40
  %2373 = atomicrmw and ptr addrspace(1) %sunkaddr3329, i32 %.lcssa3990 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %2374, !amdgpu.uniform !42

2374:                                             ; preds = %2371, %ComputeEnd2666
  call void @llvm.amdgcn.end.cf.i64(i64 %3947)
  tail call void @llvm.amdgcn.wave.barrier()
  call void @llvm.lifetime.start.p5(ptr addrspace(5) %buffer.i.i.i) #34, !noalias !102
  %retval.0.i.i.i.i.i.i94.i = zext nneg i8 %RHS to i32
  br label %while.body.i.i.i.i.i.i95.i, !amdgpu.uniform !42

while.body.i.i.i.i.i.i95.i:                       ; preds = %2374, %_ZN22__llvm_libc_22_0_0_git7details22StringBufferWriterImplILb0EE4pushEc.exit.i.i.i.i.i.i108.i
  %phi.broken3951 = phi i64 [ 0, %2374 ], [ %2442, %_ZN22__llvm_libc_22_0_0_git7details22StringBufferWriterImplILb0EE4pushEc.exit.i.i.i.i.i.i108.i ]
  %writer.sroa.10.0.i.i.i.i96.i = phi i64 [ %writer.sroa.10.1.i.i.i.i109.i, %_ZN22__llvm_libc_22_0_0_git7details22StringBufferWriterImplILb0EE4pushEc.exit.i.i.i.i.i.i108.i ], [ 0, %2374 ]
  %value.addr.09.i.i.i.i.i.i97.i = phi i32 [ %div.i.i.i.i.i.i.i100.i, %_ZN22__llvm_libc_22_0_0_git7details22StringBufferWriterImplILb0EE4pushEc.exit.i.i.i.i.i.i108.i ], [ %retval.0.i.i.i.i.i.i94.i, %2374 ]
  br label %NodeBlock3386, !amdgpu.uniform !42

NodeBlock3386:                                    ; preds = %while.body.i.i.i.i.i.i95.i
  %div.i.i.i.i.i.i.i100.i = udiv i32 %value.addr.09.i.i.i.i.i.i97.i, 10
  %.neg1 = call i32 @llvm.amdgcn.mul.u24.i32(i32 %div.i.i.i.i.i.i.i100.i, i32 246)
  %rem.i.i.i.i.i.i.i98.i.decomposed = add nuw nsw i32 %.neg1, %value.addr.09.i.i.i.i.i.i97.i
  %conv.i.i.i.i.i.i.i99.i = trunc i32 %rem.i.i.i.i.i.i.i98.i.decomposed to i8
  %2375 = zext i8 %conv.i.i.i.i.i.i.i99.i to i16
  %Pivot3387 = icmp sge i16 %2375, 5
  %2376 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %Pivot3387)
  %2377 = extractvalue { i1, i64 } %2376, 0
  %2378 = extractvalue { i1, i64 } %2376, 1
  br i1 %2377, label %NodeBlock3384, label %Flow3437

NodeBlock3384:                                    ; preds = %NodeBlock3386
  %Pivot3385 = icmp sge i16 %2375, 7
  %2379 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %Pivot3385)
  %2380 = extractvalue { i1, i64 } %2379, 0
  %2381 = extractvalue { i1, i64 } %2379, 1
  br i1 %2380, label %NodeBlock3382, label %Flow3428

NodeBlock3382:                                    ; preds = %NodeBlock3384
  %Pivot3383 = icmp sge i16 %2375, 8
  %2382 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %Pivot3383)
  %2383 = extractvalue { i1, i64 } %2382, 0
  %2384 = extractvalue { i1, i64 } %2382, 1
  br i1 %2383, label %NodeBlock3380, label %Flow3424

NodeBlock3380:                                    ; preds = %NodeBlock3382
  %Pivot3381 = icmp sge i16 %2375, 9
  %2385 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %Pivot3381)
  %2386 = extractvalue { i1, i64 } %2385, 0
  %2387 = extractvalue { i1, i64 } %2385, 1
  br i1 %2386, label %sw.bb9.i.i.i.i.i.i.i.i101.i, label %Flow3422

Flow3428:                                         ; preds = %Flow3425, %NodeBlock3384
  %2388 = phi i8 [ %2432, %Flow3425 ], [ poison, %NodeBlock3384 ]
  %2389 = call { i1, i64 } @llvm.amdgcn.else.i64.i64(i64 %2381)
  %2390 = extractvalue { i1, i64 } %2389, 0
  %2391 = extractvalue { i1, i64 } %2389, 1
  br i1 %2390, label %NodeBlock3378, label %Flow3429

NodeBlock3378:                                    ; preds = %Flow3428
  %Pivot3379 = icmp sge i16 %2375, 6
  %2392 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %Pivot3379)
  %2393 = extractvalue { i1, i64 } %2392, 0
  %2394 = extractvalue { i1, i64 } %2392, 1
  br i1 %2393, label %sw.bb6.i.i.i.i.i.i.i.i157.i, label %Flow3426

Flow3437:                                         ; preds = %Flow3429, %NodeBlock3386
  %2395 = phi i8 [ %2434, %Flow3429 ], [ poison, %NodeBlock3386 ]
  %2396 = call { i1, i64 } @llvm.amdgcn.else.i64.i64(i64 %2378)
  %2397 = extractvalue { i1, i64 } %2396, 0
  %2398 = extractvalue { i1, i64 } %2396, 1
  br i1 %2397, label %NodeBlock3376, label %if.then.i.i.i.i.i.i.i102.i

NodeBlock3376:                                    ; preds = %Flow3437
  %Pivot3377 = icmp sge i16 %2375, 3
  %2399 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %Pivot3377)
  %2400 = extractvalue { i1, i64 } %2399, 0
  %2401 = extractvalue { i1, i64 } %2399, 1
  br i1 %2400, label %NodeBlock3374, label %Flow3435

NodeBlock3374:                                    ; preds = %NodeBlock3376
  %Pivot3375 = icmp sge i16 %2375, 4
  %2402 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %Pivot3375)
  %2403 = extractvalue { i1, i64 } %2402, 0
  %2404 = extractvalue { i1, i64 } %2402, 1
  br i1 %2403, label %sw.bb4.i.i.i.i.i.i.i.i159.i, label %Flow3430

Flow3435:                                         ; preds = %Flow3431, %NodeBlock3376
  %2405 = phi i8 [ %2435, %Flow3431 ], [ poison, %NodeBlock3376 ]
  %2406 = call { i1, i64 } @llvm.amdgcn.else.i64.i64(i64 %2401)
  %2407 = extractvalue { i1, i64 } %2406, 0
  %2408 = extractvalue { i1, i64 } %2406, 1
  br i1 %2407, label %NodeBlock3372, label %Flow3436

NodeBlock3372:                                    ; preds = %Flow3435
  %Pivot3373 = icmp sge i16 %2375, 2
  %2409 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %Pivot3373)
  %2410 = extractvalue { i1, i64 } %2409, 0
  %2411 = extractvalue { i1, i64 } %2409, 1
  br i1 %2410, label %sw.bb2.i.i.i.i.i.i.i.i161.i, label %Flow3433

Flow3433:                                         ; preds = %sw.bb2.i.i.i.i.i.i.i.i161.i, %NodeBlock3372
  %2412 = call { i1, i64 } @llvm.amdgcn.else.i64.i64(i64 %2411)
  %2413 = extractvalue { i1, i64 } %2412, 0
  %2414 = extractvalue { i1, i64 } %2412, 1
  br i1 %2413, label %LeafBlock3370, label %Flow3434

LeafBlock3370:                                    ; preds = %Flow3433
  %SwitchLeaf3371 = icmp eq i16 %2375, 1
  %2415 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %SwitchLeaf3371)
  %2416 = extractvalue { i1, i64 } %2415, 0
  %2417 = extractvalue { i1, i64 } %2415, 1
  br i1 %2416, label %sw.bb1.i.i.i.i.i.i.i.i162.i, label %Flow3432

sw.bb1.i.i.i.i.i.i.i.i162.i:                      ; preds = %LeafBlock3370
  br label %Flow3432, !amdgpu.uniform !42

sw.bb2.i.i.i.i.i.i.i.i161.i:                      ; preds = %NodeBlock3372
  br label %Flow3433, !amdgpu.uniform !42

Flow3430:                                         ; preds = %sw.bb4.i.i.i.i.i.i.i.i159.i, %NodeBlock3374
  %2418 = call { i1, i64 } @llvm.amdgcn.else.i64.i64(i64 %2404)
  %2419 = extractvalue { i1, i64 } %2418, 0
  %2420 = extractvalue { i1, i64 } %2418, 1
  br i1 %2419, label %sw.bb3.i.i.i.i.i.i.i.i160.i, label %Flow3431

sw.bb3.i.i.i.i.i.i.i.i160.i:                      ; preds = %Flow3430
  br label %Flow3431, !amdgpu.uniform !42

sw.bb4.i.i.i.i.i.i.i.i159.i:                      ; preds = %NodeBlock3374
  br label %Flow3430, !amdgpu.uniform !42

Flow3426:                                         ; preds = %sw.bb6.i.i.i.i.i.i.i.i157.i, %NodeBlock3378
  %2421 = call { i1, i64 } @llvm.amdgcn.else.i64.i64(i64 %2394)
  %2422 = extractvalue { i1, i64 } %2421, 0
  %2423 = extractvalue { i1, i64 } %2421, 1
  br i1 %2422, label %sw.bb5.i.i.i.i.i.i.i.i158.i, label %Flow3427

sw.bb5.i.i.i.i.i.i.i.i158.i:                      ; preds = %Flow3426
  br label %Flow3427, !amdgpu.uniform !42

sw.bb6.i.i.i.i.i.i.i.i157.i:                      ; preds = %NodeBlock3378
  br label %Flow3426, !amdgpu.uniform !42

Flow3424:                                         ; preds = %Flow3423, %NodeBlock3382
  %2424 = phi i8 [ %2431, %Flow3423 ], [ poison, %NodeBlock3382 ]
  %2425 = call { i1, i64 } @llvm.amdgcn.else.i64.i64(i64 %2384)
  %2426 = extractvalue { i1, i64 } %2425, 0
  %2427 = extractvalue { i1, i64 } %2425, 1
  br i1 %2426, label %sw.bb7.i.i.i.i.i.i.i.i156.i, label %Flow3425

sw.bb7.i.i.i.i.i.i.i.i156.i:                      ; preds = %Flow3424
  br label %Flow3425, !amdgpu.uniform !42

Flow3422:                                         ; preds = %sw.bb9.i.i.i.i.i.i.i.i101.i, %NodeBlock3380
  %2428 = call { i1, i64 } @llvm.amdgcn.else.i64.i64(i64 %2387)
  %2429 = extractvalue { i1, i64 } %2428, 0
  %2430 = extractvalue { i1, i64 } %2428, 1
  br i1 %2429, label %sw.bb8.i.i.i.i.i.i.i.i155.i, label %Flow3423

sw.bb8.i.i.i.i.i.i.i.i155.i:                      ; preds = %Flow3422
  br label %Flow3423, !amdgpu.uniform !42

sw.bb9.i.i.i.i.i.i.i.i101.i:                      ; preds = %NodeBlock3380
  br label %Flow3422, !amdgpu.uniform !42

Flow3423:                                         ; preds = %sw.bb8.i.i.i.i.i.i.i.i155.i, %Flow3422
  %2431 = phi i8 [ 56, %sw.bb8.i.i.i.i.i.i.i.i155.i ], [ 57, %Flow3422 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %2430)
  br label %Flow3424, !amdgpu.uniform !42

Flow3425:                                         ; preds = %sw.bb7.i.i.i.i.i.i.i.i156.i, %Flow3424
  %2432 = phi i8 [ 55, %sw.bb7.i.i.i.i.i.i.i.i156.i ], [ %2424, %Flow3424 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %2427)
  br label %Flow3428, !amdgpu.uniform !42

Flow3427:                                         ; preds = %sw.bb5.i.i.i.i.i.i.i.i158.i, %Flow3426
  %2433 = phi i8 [ 53, %sw.bb5.i.i.i.i.i.i.i.i158.i ], [ 54, %Flow3426 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %2423)
  br label %Flow3429, !amdgpu.uniform !42

Flow3429:                                         ; preds = %Flow3427, %Flow3428
  %2434 = phi i8 [ %2433, %Flow3427 ], [ %2388, %Flow3428 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %2391)
  br label %Flow3437, !amdgpu.uniform !42

Flow3431:                                         ; preds = %sw.bb3.i.i.i.i.i.i.i.i160.i, %Flow3430
  %2435 = phi i8 [ 51, %sw.bb3.i.i.i.i.i.i.i.i160.i ], [ 52, %Flow3430 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %2420)
  br label %Flow3435, !amdgpu.uniform !42

Flow3432:                                         ; preds = %sw.bb1.i.i.i.i.i.i.i.i162.i, %LeafBlock3370
  %2436 = phi i8 [ 49, %sw.bb1.i.i.i.i.i.i.i.i162.i ], [ 48, %LeafBlock3370 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %2417)
  br label %Flow3434, !amdgpu.uniform !42

Flow3434:                                         ; preds = %Flow3432, %Flow3433
  %2437 = phi i8 [ %2436, %Flow3432 ], [ 50, %Flow3433 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %2414)
  br label %Flow3436, !amdgpu.uniform !42

Flow3436:                                         ; preds = %Flow3434, %Flow3435
  %2438 = phi i8 [ %2437, %Flow3434 ], [ %2405, %Flow3435 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %2408)
  br label %if.then.i.i.i.i.i.i.i102.i, !amdgpu.uniform !42

if.then.i.i.i.i.i.i.i102.i:                       ; preds = %Flow3437, %Flow3436
  %retval.0.i.i.i.i.i.i.i.i103.i = phi i8 [ %2395, %Flow3437 ], [ %2438, %Flow3436 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %2398)
  %cmp.i.i.i.i.i.i.i.i104.i = icmp eq i64 %writer.sroa.10.0.i.i.i.i96.i, 11
  %cmp.i.i.i.i.i.i.i.i104.i.inv = xor i1 %cmp.i.i.i.i.i.i.i.i104.i, true
  br i1 %cmp.i.i.i.i.i.i.i.i104.i.inv, label %if.then3.i.i.i.i.i.i.i105.i, label %_ZN22__llvm_libc_22_0_0_git7details22StringBufferWriterImplILb0EE4pushEc.exit.i.i.i.i.i.i108.i, !amdgpu.uniform !42

if.then3.i.i.i.i.i.i.i105.i:                      ; preds = %if.then.i.i.i.i.i.i.i102.i
  %2439 = trunc i64 %writer.sroa.10.0.i.i.i.i96.i to i32
  %2440 = xor i32 %2439, -1
  %2441 = getelementptr i8, ptr addrspace(5) %buffer.i.i.i, i32 %2440
  %arrayidx.i.i.i.i.i.i.i.i106.i2741 = getelementptr i8, ptr addrspace(5) %2441, i32 11
  %inc.i.i.i.i.i.i.i107.i = add i64 %writer.sroa.10.0.i.i.i.i96.i, 1
  store i8 %retval.0.i.i.i.i.i.i.i.i103.i, ptr addrspace(5) %arrayidx.i.i.i.i.i.i.i.i106.i2741, align 1, !tbaa !57, !noalias !102
  br label %_ZN22__llvm_libc_22_0_0_git7details22StringBufferWriterImplILb0EE4pushEc.exit.i.i.i.i.i.i108.i, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git7details22StringBufferWriterImplILb0EE4pushEc.exit.i.i.i.i.i.i108.i: ; preds = %if.then3.i.i.i.i.i.i.i105.i, %if.then.i.i.i.i.i.i.i102.i
  %writer.sroa.10.1.i.i.i.i109.i = phi i64 [ %inc.i.i.i.i.i.i.i107.i, %if.then3.i.i.i.i.i.i.i105.i ], [ 11, %if.then.i.i.i.i.i.i.i102.i ]
  %cmp.i.i.i.i.i.i110.i = icmp samesign ult i32 %value.addr.09.i.i.i.i.i.i97.i, 10
  %.not.i.i.i.i.i.i111.i = or i1 %cmp.i.i.i.i.i.i.i.i104.i, %cmp.i.i.i.i.i.i110.i
  %2442 = call i64 @llvm.amdgcn.if.break.i64(i1 %.not.i.i.i.i.i.i111.i, i64 %phi.broken3951)
  %2443 = call i1 @llvm.amdgcn.loop.i64(i64 %2442)
  br i1 %2443, label %_ZN22__llvm_libc_22_0_0_git15IntegerToStringIiNS_7details3FmtILh10ELb0ELb0ELb0ELm1EEEE13IntegerWriter25write_unsigned_number_decEjRNS1_22StringBufferWriterImplILb0EEE.exit.i.i.i.i.i112.i, label %while.body.i.i.i.i.i.i95.i

_ZN22__llvm_libc_22_0_0_git15IntegerToStringIiNS_7details3FmtILh10ELb0ELb0ELb0ELm1EEEE13IntegerWriter25write_unsigned_number_decEjRNS1_22StringBufferWriterImplILb0EEE.exit.i.i.i.i.i112.i: ; preds = %_ZN22__llvm_libc_22_0_0_git7details22StringBufferWriterImplILb0EE4pushEc.exit.i.i.i.i.i.i108.i
  %writer.sroa.10.1.i.i.i.i109.i.lcssa = phi i64 [ %writer.sroa.10.1.i.i.i.i109.i, %_ZN22__llvm_libc_22_0_0_git7details22StringBufferWriterImplILb0EE4pushEc.exit.i.i.i.i.i.i108.i ]
  %.lcssa3989 = phi i64 [ %2442, %_ZN22__llvm_libc_22_0_0_git7details22StringBufferWriterImplILb0EE4pushEc.exit.i.i.i.i.i.i108.i ]
  %cmp.i.i.i.i.i.i.i.i104.i.lcssa = phi i1 [ %cmp.i.i.i.i.i.i.i.i104.i, %_ZN22__llvm_libc_22_0_0_git7details22StringBufferWriterImplILb0EE4pushEc.exit.i.i.i.i.i.i108.i ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa3989)
  %cmp.i19.i.i.i.i115.i = icmp ne i64 %writer.sroa.10.1.i.i.i.i109.i.lcssa, 0
  %or.cond.not.i.i.i.i116.i = select i1 %cmp.i.i.i.i.i.i.i.i104.i.lcssa, i1 true, i1 %cmp.i19.i.i.i.i115.i
  %or.cond.not.i.i.i.i116.i.inv = xor i1 %or.cond.not.i.i.i.i116.i, true
  %2444 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %or.cond.not.i.i.i.i116.i.inv)
  %2445 = extractvalue { i1, i64 } %2444, 0
  %2446 = extractvalue { i1, i64 } %2444, 1
  br i1 %2445, label %if.then3.i.i.lr.ph.i.i.i.i117.i, label %while.end.i.i.i.i.i119.i

if.then3.i.i.lr.ph.i.i.i.i117.i:                  ; preds = %_ZN22__llvm_libc_22_0_0_git15IntegerToStringIiNS_7details3FmtILh10ELb0ELb0ELb0ELm1EEEE13IntegerWriter25write_unsigned_number_decEjRNS1_22StringBufferWriterImplILb0EEE.exit.i.i.i.i.i112.i
  %arrayidx.i.i.i.i.i.i.i118.i = getelementptr inbounds nuw i8, ptr addrspace(5) %buffer.i.i.i, i32 10
  store i8 48, ptr addrspace(5) %arrayidx.i.i.i.i.i.i.i118.i, align 2, !tbaa !57, !noalias !102
  br label %while.end.i.i.i.i.i119.i, !amdgpu.uniform !42

while.end.i.i.i.i.i119.i:                         ; preds = %if.then3.i.i.lr.ph.i.i.i.i117.i, %_ZN22__llvm_libc_22_0_0_git15IntegerToStringIiNS_7details3FmtILh10ELb0ELb0ELb0ELm1EEEE13IntegerWriter25write_unsigned_number_decEjRNS1_22StringBufferWriterImplILb0EEE.exit.i.i.i.i.i112.i
  %writer.sroa.10.4.i.i.i.i120.i = phi i64 [ 1, %if.then3.i.i.lr.ph.i.i.i.i117.i ], [ %writer.sroa.10.1.i.i.i.i109.i.lcssa, %_ZN22__llvm_libc_22_0_0_git15IntegerToStringIiNS_7details3FmtILh10ELb0ELb0ELb0ELm1EEEE13IntegerWriter25write_unsigned_number_decEjRNS1_22StringBufferWriterImplILb0EEE.exit.i.i.i.i.i112.i ]
  call void @llvm.amdgcn.end.cf.i64(i64 %2446)
  %written.i.i.i.i130.i = getelementptr inbounds nuw i8, ptr addrspace(5) %buffer.i.i.i, i32 16
  store i64 %writer.sroa.10.4.i.i.i.i120.i, ptr addrspace(5) %written.i.i.i.i130.i, align 8, !tbaa !91, !noalias !102
  %cmp.i.not.i.i.i.i.i133.i = icmp ne i64 %writer.sroa.10.4.i.i.i.i120.i, 0
  %2447 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.i.not.i.i.i.i.i133.i)
  %2448 = extractvalue { i1, i64 } %2447, 0
  %2449 = extractvalue { i1, i64 } %2447, 1
  br i1 %2448, label %if.then.i.i.i.i.i.i134.i, label %Flow3421

if.then.i.i.i.i.i.i134.i:                         ; preds = %while.end.i.i.i.i.i119.i
  %2450 = call { i64, i1 } @llvm.uadd.with.overflow.i64(i64 %writer.sroa.10.4.i.i.i.i120.i, i64 1)
  %ov3331 = extractvalue { i64, i1 } %2450, 1
  %ov3331.inv = xor i1 %ov3331, true
  %2451 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %ov3331.inv)
  %2452 = extractvalue { i1, i64 } %2451, 0
  %2453 = extractvalue { i1, i64 } %2451, 1
  br i1 %2452, label %if.end.i.i.i.i.i.i.i137.i, label %Flow3420

if.end.i.i.i.i.i.i.i137.i:                        ; preds = %if.then.i.i.i.i.i.i134.i
  %math3330 = extractvalue { i64, i1 } %2450, 0
  %cmp2.i.i.i.i.i.i.i138.i = icmp ult i64 %math3330, 1676976733973595601
  %2454 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp2.i.i.i.i.i.i.i138.i)
  %2455 = extractvalue { i1, i64 } %2454, 0
  %2456 = extractvalue { i1, i64 } %2454, 1
  br i1 %2455, label %if.then3.i.i.i.i4.i.i.i152.i, label %if.end4.i.i.i.i.i.i.i139.i

if.then3.i.i.i.i4.i.i.i152.i:                     ; preds = %if.end.i.i.i.i.i.i.i137.i
  %mul.i.i.i.i.i.i.i153.i = mul nuw i64 %math3330, 11
  %div18.i.i.i.i.i.i.i154.i = lshr i64 %mul.i.i.i.i.i.i.i153.i, 3
  br label %if.end4.i.i.i.i.i.i.i139.i, !amdgpu.uniform !42

Flow3420:                                         ; preds = %if.end4.i.i.i.i.i.i.i139.i, %if.then.i.i.i.i.i.i134.i
  %2457 = phi ptr [ %cond.i.i1883, %if.end4.i.i.i.i.i.i.i139.i ], [ addrspacecast (ptr addrspace(4) @_ZN22__llvm_libc_22_0_0_git3cpp6string14NULL_CHARACTERE to ptr), %if.then.i.i.i.i.i.i134.i ]
  call void @llvm.amdgcn.end.cf.i64(i64 %2453)
  br label %_ZN22__llvm_libc_22_0_0_git3cpp6string7reserveEm.exit.i.i.i.i.i.i143.i, !amdgpu.uniform !42

if.end4.i.i.i.i.i.i.i139.i:                       ; preds = %if.then3.i.i.i.i4.i.i.i152.i, %if.end.i.i.i.i.i.i.i137.i
  %new_capacity.addr.0.i.i.i.i.i.i.i140.i = phi i64 [ %div18.i.i.i.i.i.i.i154.i, %if.then3.i.i.i.i4.i.i.i152.i ], [ %math3330, %if.end.i.i.i.i.i.i.i137.i ]
  call void @llvm.amdgcn.end.cf.i64(i64 %2456)
  %sub.i.i1877 = add i64 %new_capacity.addr.0.i.i.i.i.i.i.i140.i, 7
  %div2.i.i1878 = and i64 %sub.i.i1877, -8
  %2458 = load ptr, ptr addrspace(1) @_ZN12_GLOBAL__N_13ptrE, align 8, !tbaa !55
  %add.ptr.i.i1879 = getelementptr inbounds nuw i8, ptr %2458, i64 %div2.i.i1878
  store ptr %add.ptr.i.i1879, ptr addrspace(1) @_ZN12_GLOBAL__N_13ptrE, align 8, !tbaa !55
  %sub.ptr.lhs.cast.i.i1880 = ptrtoint ptr %add.ptr.i.i1879 to i64
  %sub.ptr.sub.i.i1881 = sub i64 %sub.ptr.lhs.cast.i.i1880, ptrtoint (ptr addrspacecast (ptr addrspace(1) @_ZN12_GLOBAL__N_16memoryE to ptr) to i64)
  %cmp.i.i1882 = icmp ugt i64 %sub.ptr.sub.i.i1881, 65335
  %cond.i.i1883 = select i1 %cmp.i.i1882, ptr null, ptr %2458
  br label %Flow3420, !amdgpu.uniform !42

Flow3421:                                         ; preds = %Flow3419, %while.end.i.i.i.i.i119.i
  %2459 = phi ptr [ %2457, %Flow3419 ], [ addrspacecast (ptr addrspace(4) @_ZN22__llvm_libc_22_0_0_git3cpp6string14NULL_CHARACTERE to ptr), %while.end.i.i.i.i.i119.i ]
  call void @llvm.amdgcn.end.cf.i64(i64 %2449)
  br label %if.end.i.i.i.i.i.i145.i, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git3cpp6string7reserveEm.exit.i.i.i.i.i.i143.i: ; preds = %Flow3420
  br label %loadstoreloop2186, !amdgpu.uniform !42

loadstoreloop2186:                                ; preds = %_ZN22__llvm_libc_22_0_0_git3cpp6string7reserveEm.exit.i.i.i.i.i.i143.i, %loadstoreloop2186
  %phi.broken3952 = phi i64 [ 0, %_ZN22__llvm_libc_22_0_0_git3cpp6string7reserveEm.exit.i.i.i.i.i.i143.i ], [ %2464, %loadstoreloop2186 ]
  %2460 = phi i64 [ %2462, %loadstoreloop2186 ], [ 0, %_ZN22__llvm_libc_22_0_0_git3cpp6string7reserveEm.exit.i.i.i.i.i.i143.i ]
  %2461 = getelementptr i8, ptr %2457, i64 %2460
  store i8 0, ptr %2461, align 1
  %2462 = add i64 %2460, 1
  %2463 = icmp uge i64 %2462, %writer.sroa.10.4.i.i.i.i120.i
  %2464 = call i64 @llvm.amdgcn.if.break.i64(i1 %2463, i64 %phi.broken3952)
  %2465 = call i1 @llvm.amdgcn.loop.i64(i64 %2464)
  br i1 %2465, label %Flow3419, label %loadstoreloop2186

Flow3419:                                         ; preds = %loadstoreloop2186
  %.lcssa3988 = phi i64 [ %2464, %loadstoreloop2186 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa3988)
  br label %Flow3421, !amdgpu.uniform !42

if.end.i.i.i.i.i.i145.i:                          ; preds = %Flow3421
  %cmp.not.i6.i.i.i.i.i.i147.i = icmp ne ptr %2459, addrspacecast (ptr addrspace(4) @_ZN22__llvm_libc_22_0_0_git3cpp6string14NULL_CHARACTERE to ptr)
  %2466 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.not.i6.i.i.i.i.i.i147.i)
  %2467 = extractvalue { i1, i64 } %2466, 0
  %2468 = extractvalue { i1, i64 } %2466, 1
  br i1 %2467, label %if.then.i.i.i.i2.i.i.i148.i, label %_ZN22__llvm_libc_22_0_0_git7testing8internal13describeValueIcEENS_3cpp9enable_ifIXaasr3cppE13is_integral_vIT_ElestS5_Lm8EENS3_6stringEE4typeES5_.exit164.i

if.then.i.i.i.i2.i.i.i148.i:                      ; preds = %if.end.i.i.i.i.i.i145.i
  %arrayidx.i.i.i.i3.i.i.i149.i = getelementptr inbounds nuw i8, ptr %2459, i64 %writer.sroa.10.4.i.i.i.i120.i
  store i8 0, ptr %arrayidx.i.i.i.i3.i.i.i149.i, align 1, !tbaa !57, !noalias !102
  br label %_ZN22__llvm_libc_22_0_0_git7testing8internal13describeValueIcEENS_3cpp9enable_ifIXaasr3cppE13is_integral_vIT_ElestS5_Lm8EENS3_6stringEE4typeES5_.exit164.i, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git7testing8internal13describeValueIcEENS_3cpp9enable_ifIXaasr3cppE13is_integral_vIT_ElestS5_Lm8EENS3_6stringEE4typeES5_.exit164.i: ; preds = %if.then.i.i.i.i2.i.i.i148.i, %if.end.i.i.i.i.i.i145.i
  call void @llvm.amdgcn.end.cf.i64(i64 %2468)
  %2469 = trunc i64 %writer.sroa.10.4.i.i.i.i120.i to i32
  %2470 = and i64 %writer.sroa.10.4.i.i.i.i120.i, 15
  %2471 = sub i64 %writer.sroa.10.4.i.i.i.i120.i, %2470
  %2472 = icmp ne i64 %2471, 0
  %2473 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %2472)
  %2474 = extractvalue { i1, i64 } %2473, 0
  %2475 = extractvalue { i1, i64 } %2473, 1
  br i1 %2474, label %loop-memcpy-expansion2296.preheader, label %Flow3418

loop-memcpy-expansion2296.preheader:              ; preds = %_ZN22__llvm_libc_22_0_0_git7testing8internal13describeValueIcEENS_3cpp9enable_ifIXaasr3cppE13is_integral_vIT_ElestS5_Lm8EENS3_6stringEE4typeES5_.exit164.i
  %2476 = sub i32 11, %2469
  %scevgep2876 = getelementptr i8, ptr addrspace(5) %buffer.i.i.i, i32 %2476
  br label %loop-memcpy-expansion2296, !amdgpu.uniform !42

Flow3418:                                         ; preds = %Flow3417, %_ZN22__llvm_libc_22_0_0_git7testing8internal13describeValueIcEENS_3cpp9enable_ifIXaasr3cppE13is_integral_vIT_ElestS5_Lm8EENS3_6stringEE4typeES5_.exit164.i
  call void @llvm.amdgcn.end.cf.i64(i64 %2475)
  br label %loop-memcpy-residual-header2299, !amdgpu.uniform !42

loop-memcpy-expansion2296:                        ; preds = %loop-memcpy-expansion2296.preheader, %loop-memcpy-expansion2296
  %phi.broken3953 = phi i64 [ 0, %loop-memcpy-expansion2296.preheader ], [ %2481, %loop-memcpy-expansion2296 ]
  %lsr.iv2877 = phi ptr addrspace(5) [ %scevgep2878, %loop-memcpy-expansion2296 ], [ %scevgep2876, %loop-memcpy-expansion2296.preheader ]
  %loop-index2297 = phi i64 [ %2479, %loop-memcpy-expansion2296 ], [ 0, %loop-memcpy-expansion2296.preheader ]
  %2477 = load <4 x i32>, ptr addrspace(5) %lsr.iv2877, align 1
  %2478 = getelementptr i8, ptr %2459, i64 %loop-index2297
  store <4 x i32> %2477, ptr %2478, align 1
  %2479 = add i64 %loop-index2297, 16
  %scevgep2878 = getelementptr i8, ptr addrspace(5) %lsr.iv2877, i32 16
  %2480 = icmp uge i64 %2479, %2471
  %2481 = call i64 @llvm.amdgcn.if.break.i64(i1 %2480, i64 %phi.broken3953)
  %2482 = call i1 @llvm.amdgcn.loop.i64(i64 %2481)
  br i1 %2482, label %Flow3417, label %loop-memcpy-expansion2296

Flow3416:                                         ; preds = %Flow3415, %loop-memcpy-residual-header2299
  call void @llvm.amdgcn.end.cf.i64(i64 %2821)
  br label %post-loop-memcpy-expansion2295, !amdgpu.uniform !42

loop-memcpy-residual2298:                         ; preds = %loop-memcpy-residual2298.preheader, %loop-memcpy-residual2298
  %phi.broken3954 = phi i64 [ 0, %loop-memcpy-residual2298.preheader ], [ %2486, %loop-memcpy-residual2298 ]
  %lsr.iv2872 = phi ptr addrspace(5) [ %scevgep2873, %loop-memcpy-residual2298 ], [ %scevgep2871, %loop-memcpy-residual2298.preheader ]
  %residual-loop-index2300 = phi i64 [ %2484, %loop-memcpy-residual2298 ], [ 0, %loop-memcpy-residual2298.preheader ]
  %2483 = load i8, ptr addrspace(5) %lsr.iv2872, align 1
  %scevgep2875 = getelementptr i8, ptr %scevgep2874, i64 %residual-loop-index2300
  store i8 %2483, ptr %scevgep2875, align 1
  %2484 = add i64 %residual-loop-index2300, 1
  %scevgep2873 = getelementptr i8, ptr addrspace(5) %lsr.iv2872, i32 1
  %2485 = icmp uge i64 %2484, %2470
  %2486 = call i64 @llvm.amdgcn.if.break.i64(i1 %2485, i64 %phi.broken3954)
  %2487 = call i1 @llvm.amdgcn.loop.i64(i64 %2486)
  br i1 %2487, label %Flow3415, label %loop-memcpy-residual2298

Flow3415:                                         ; preds = %loop-memcpy-residual2298
  %.lcssa3986 = phi i64 [ %2486, %loop-memcpy-residual2298 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa3986)
  br label %Flow3416, !amdgpu.uniform !42

post-loop-memcpy-expansion2295:                   ; preds = %Flow3416
  call void @llvm.lifetime.end.p5(ptr addrspace(5) %buffer.i.i.i) #34, !noalias !102
  br label %for.cond.i.i.i1885, !amdgpu.uniform !42

for.cond.i.i.i1885:                               ; preds = %post-loop-memcpy-expansion2295, %Flow3414
  %2488 = phi i32 [ %2526, %Flow3414 ], [ poison, %post-loop-memcpy-expansion2295 ]
  %2489 = phi i64 [ %2527, %Flow3414 ], [ poison, %post-loop-memcpy-expansion2295 ]
  %2490 = phi i32 [ %2528, %Flow3414 ], [ poison, %post-loop-memcpy-expansion2295 ]
  %retval.sroa.6.0.i.i.i1886 = phi i32 [ %2531, %Flow3414 ], [ undef, %post-loop-memcpy-expansion2295 ]
  %retval.sroa.2.0.i.i.i1888 = phi i64 [ %2530, %Flow3414 ], [ undef, %post-loop-memcpy-expansion2295 ]
  %retval.sroa.8.0.i.i.i1889 = phi i32 [ %2529, %Flow3414 ], [ undef, %post-loop-memcpy-expansion2295 ]
  %index.0.i.i.i1890 = phi i32 [ %2533, %Flow3414 ], [ 0, %post-loop-memcpy-expansion2295 ]
  %2491 = load i32, ptr addrspace(1) @__llvm_rpc_client, align 8, !tbaa !68
  %cmp.not.i.i.i1891 = icmp ult i32 %index.0.i.i.i1890, %2491
  %spec.store.select.i.i.i1892 = select i1 %cmp.not.i.i.i1891, i32 %index.0.i.i.i1890, i32 0
  %2492 = tail call noundef i64 @llvm.amdgcn.ballot.i64(i1 true)
  %2493 = and i32 %spec.store.select.i.i.i1892, 31
  %2494 = lshr i64 %2492, %sh_prom.i.i.i.i
  %2495 = trunc i64 %2494 to i32
  %conv4.i.i.i.i.i1896 = and i32 %2495, 1
  %shl.i.i.i.i.i1897 = shl nuw i32 %conv4.i.i.i.i.i1896, %2493
  %2496 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2669, !amdgpu.uniform !42

2497:                                             ; preds = %ComputeEnd2670
  %sunkaddr3332 = getelementptr inbounds i8, ptr addrspace(1) %3957, i64 40
  %2498 = atomicrmw or ptr addrspace(1) %sunkaddr3332, i32 %.lcssa3980 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %2499, !amdgpu.uniform !42

2499:                                             ; preds = %2497, %ComputeEnd2670
  %2500 = phi i32 [ %2498, %2497 ], [ poison, %ComputeEnd2670 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %3966)
  %2501 = call i32 @llvm.amdgcn.readfirstlane.i32(i32 %2500)
  %2502 = or i32 %2501, %.lcssa3981
  %shl5.i.i.i.i.i1898 = shl nuw i32 1, %2493
  %and.i.i.i.i.i1899 = and i32 %shl5.i.i.i.i.i1898, %2502
  %tobool3.i.i.i.i1900 = icmp ne i32 %and.i.i.i.i.i1899, 0
  %2503 = tail call i64 @llvm.amdgcn.ballot.i64(i1 %tobool3.i.i.i.i1900)
  %and.i.i.i.i.i.i1901 = and i64 %2503, %2492
  %cmp.i.not.i.i.i1902 = icmp ne i64 %2492, %and.i.i.i.i.i.i1901
  br i1 %cmp.i.not.i.i.i1902, label %if.then.i.i.i.i1903, label %Flow3413, !amdgpu.uniform !42

if.then.i.i.i.i1903:                              ; preds = %2499
  fence syncscope("agent") acquire
  %2504 = load <2 x i64>, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !75
  %2505 = extractelement <2 x i64> %2504, i32 0
  %2506 = inttoptr i64 %2505 to ptr
  %2507 = extractelement <2 x i64> %2504, i32 1
  %2508 = inttoptr i64 %2507 to ptr
  %idxprom.i.i.i.i1904 = zext i32 %spec.store.select.i.i.i1892 to i64
  %arrayidx.i.i.i.i1905 = getelementptr inbounds nuw i32, ptr %2506, i64 %idxprom.i.i.i.i1904, !amdgpu.uniform !42
  %2509 = load atomic i32, ptr %arrayidx.i.i.i.i1905 monotonic, align 4
  %2510 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %2509)
  %arrayidx.i45.i.i.i1906 = getelementptr inbounds nuw i32, ptr %2508, i64 %idxprom.i.i.i.i1904, !amdgpu.uniform !42
  %2511 = load atomic i32, ptr %arrayidx.i45.i.i.i1906 monotonic, align 4
  %2512 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %2511)
  %cmp.i46.not.i.i.i1907 = icmp ne i32 %2510, %2512
  br i1 %cmp.i46.not.i.i.i1907, label %if.then12.i.i.i1908, label %Flow3411, !amdgpu.uniform !42

if.then12.i.i.i1908:                              ; preds = %if.then.i.i.i.i1903
  fence syncscope("agent") release
  %2513 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %2492, i1 true)
  %iszero.i.i.i.i.i.i.i1909 = icmp ne i64 %2492, 0
  %cmp2.i.i.i.i.i.i1910 = icmp eq i64 %2513, %sh_prom.i.i.i.i
  %cmp.i.i.i.i.i.i1911 = select i1 %iszero.i.i.i.i.i.i.i1909, i1 %cmp2.i.i.i.i.i.i1910, i1 false
  %conv4.i.i51.i.i.i1912 = zext i1 %cmp.i.i.i.i.i.i1911 to i32
  %shl.i.i52.i.i.i1913 = shl nuw i32 %conv4.i.i51.i.i.i1912, %2493
  %xor.i.i.i.i.i1914 = xor i32 %shl.i.i52.i.i.i1913, -1
  %2514 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2674, !amdgpu.uniform !42

2515:                                             ; preds = %ComputeEnd2675
  %sunkaddr3333 = getelementptr inbounds i8, ptr addrspace(1) %3957, i64 40
  %2516 = atomicrmw and ptr addrspace(1) %sunkaddr3333, i32 %.lcssa3982 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %2517, !amdgpu.uniform !42

2517:                                             ; preds = %2515, %ComputeEnd2675
  call void @llvm.amdgcn.end.cf.i64(i64 %3984)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %Flow3411, !amdgpu.uniform !42

if.end14.i.i.i1974:                               ; preds = %Flow3411
  %2518 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %2492, i1 true)
  %iszero.i.i.i.i.i.i1975 = icmp ne i64 %2492, 0
  %cmp2.i.i.i.i.i1976 = icmp eq i64 %2518, %sh_prom.i.i.i.i
  %cmp.i.i.i.i.i1977 = select i1 %iszero.i.i.i.i.i.i1975, i1 %cmp2.i.i.i.i.i1976, i1 false
  %2519 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.i.i.i.i.i1977)
  %2520 = extractvalue { i1, i64 } %2519, 0
  %2521 = extractvalue { i1, i64 } %2519, 1
  br i1 %2520, label %if.then16.i.i.i1979, label %if.end22.i.i.i1978

if.then16.i.i.i1979:                              ; preds = %if.end14.i.i.i1974
  %2522 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !76
  %opcode.i.i.i1980 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %2522, i64 %idxprom.i.i.i.i1904, i32 1
  store i32 1660944387, ptr %opcode.i.i.i1980, align 8, !tbaa !77
  %arrayidx21.i.i.i1981 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %2522, i64 %idxprom.i.i.i.i1904
  store i64 %2492, ptr %arrayidx21.i.i.i1981, align 8, !tbaa !80
  br label %if.end22.i.i.i1978, !amdgpu.uniform !42

Flow3412:                                         ; preds = %if.end22.i.i.i1978, %Flow3411
  %2523 = phi i32 [ %2510, %if.end22.i.i.i1978 ], [ %retval.sroa.8.0.i.i.i1889, %Flow3411 ]
  %2524 = phi i64 [ %2492, %if.end22.i.i.i1978 ], [ %retval.sroa.2.0.i.i.i1888, %Flow3411 ]
  %2525 = phi i32 [ %spec.store.select.i.i.i1892, %if.end22.i.i.i1978 ], [ %retval.sroa.6.0.i.i.i1886, %Flow3411 ]
  br label %cleanup26.i.i.i1915, !amdgpu.uniform !42

if.end22.i.i.i1978:                               ; preds = %if.then16.i.i.i1979, %if.end14.i.i.i1974
  call void @llvm.amdgcn.end.cf.i64(i64 %2521)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %Flow3412, !amdgpu.uniform !42

Flow3413:                                         ; preds = %cleanup26.i.i.i1915, %2499
  %2526 = phi i32 [ %2525, %cleanup26.i.i.i1915 ], [ %2488, %2499 ]
  %2527 = phi i64 [ %2524, %cleanup26.i.i.i1915 ], [ %2489, %2499 ]
  %2528 = phi i32 [ %2523, %cleanup26.i.i.i1915 ], [ %2490, %2499 ]
  %2529 = phi i32 [ %2523, %cleanup26.i.i.i1915 ], [ %retval.sroa.8.0.i.i.i1889, %2499 ]
  %2530 = phi i64 [ %2524, %cleanup26.i.i.i1915 ], [ %retval.sroa.2.0.i.i.i1888, %2499 ]
  %2531 = phi i32 [ %2525, %cleanup26.i.i.i1915 ], [ %retval.sroa.6.0.i.i.i1886, %2499 ]
  %2532 = phi i1 [ %cmp.i46.not.i.i.i1907, %cleanup26.i.i.i1915 ], [ true, %2499 ]
  br i1 %2532, label %for.inc.i.i.i1920, label %Flow3414, !amdgpu.uniform !42

cleanup26.i.i.i1915:                              ; preds = %Flow3412
  br label %Flow3413, !amdgpu.uniform !42

for.inc.i.i.i1920:                                ; preds = %Flow3413
  %inc.i.i.i1925 = add i32 %spec.store.select.i.i.i1892, 1
  br label %Flow3414, !amdgpu.uniform !42

Flow3414:                                         ; preds = %for.inc.i.i.i1920, %Flow3413
  %2533 = phi i32 [ %inc.i.i.i1925, %for.inc.i.i.i1920 ], [ poison, %Flow3413 ]
  %2534 = phi i1 [ false, %for.inc.i.i.i1920 ], [ true, %Flow3413 ]
  br i1 %2534, label %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1926, label %for.cond.i.i.i1885, !amdgpu.uniform !42

_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1926: ; preds = %Flow3414
  %.lcssa3985 = phi i32 [ %2526, %Flow3414 ]
  %.lcssa3984 = phi i64 [ %2527, %Flow3414 ]
  %.lcssa3983 = phi i32 [ %2528, %Flow3414 ]
  fence acquire
  %2535 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %mul.i.i.i.i.i.i1927 = shl i32 %.lcssa3985, 6
  %idxprom.i19.i.i.i.i.i1928 = zext i32 %mul.i.i.i.i.i.i1927 to i64
  %arrayidx.i20.i.i.i.i.i1929 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %2535, i64 %idxprom.i19.i.i.i.i.i1928
  %arrayidx.i22.i.i.i.i.i1930 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %arrayidx.i20.i.i.i.i.i1929, i64 %sh_prom.i.i.i.i
  store i64 %writer.sroa.10.4.i.i.i.i120.i, ptr %arrayidx.i22.i.i.i.i.i1930, align 8, !tbaa !82
  %spec.select.i.i.i.i.i.i.i1931 = tail call i64 @llvm.umin.i64(i64 %writer.sroa.10.4.i.i.i.i120.i, i64 56)
  %2536 = and i64 %spec.select.i.i.i.i.i.i.i1931, 15
  %2537 = sub i64 %spec.select.i.i.i.i.i.i.i1931, %2536
  %2538 = icmp ne i64 %2537, 0
  %2539 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %2538)
  %2540 = extractvalue { i1, i64 } %2539, 0
  %2541 = extractvalue { i1, i64 } %2539, 1
  br i1 %2540, label %loop-memcpy-expansion2194.preheader, label %Flow3410

loop-memcpy-expansion2194.preheader:              ; preds = %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1926
  %2542 = add nuw nsw i64 %75, 8
  %scevgep2868 = getelementptr i8, ptr %2535, i64 %2542
  %2543 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i1928, 6
  %scevgep2869 = getelementptr i8, ptr %scevgep2868, i64 %2543
  br label %loop-memcpy-expansion2194, !amdgpu.uniform !42

Flow3410:                                         ; preds = %Flow3409, %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1926
  call void @llvm.amdgcn.end.cf.i64(i64 %2541)
  br label %loop-memcpy-residual-header2197, !amdgpu.uniform !42

loop-memcpy-expansion2194:                        ; preds = %loop-memcpy-expansion2194.preheader, %loop-memcpy-expansion2194
  %phi.broken3955 = phi i64 [ 0, %loop-memcpy-expansion2194.preheader ], [ %2548, %loop-memcpy-expansion2194 ]
  %loop-index2195 = phi i64 [ %2546, %loop-memcpy-expansion2194 ], [ 0, %loop-memcpy-expansion2194.preheader ]
  %2544 = getelementptr i8, ptr %2459, i64 %loop-index2195
  %2545 = load <4 x i32>, ptr %2544, align 1
  %scevgep2870 = getelementptr i8, ptr %scevgep2869, i64 %loop-index2195
  store <4 x i32> %2545, ptr %scevgep2870, align 1
  %2546 = add i64 %loop-index2195, 16
  %2547 = icmp uge i64 %2546, %2537
  %2548 = call i64 @llvm.amdgcn.if.break.i64(i1 %2547, i64 %phi.broken3955)
  %2549 = call i1 @llvm.amdgcn.loop.i64(i64 %2548)
  br i1 %2549, label %Flow3409, label %loop-memcpy-expansion2194

Flow3408:                                         ; preds = %Flow3406, %Flow3407
  %2550 = phi i64 [ %2721, %Flow3406 ], [ %2713, %Flow3407 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %2716)
  br label %post-loop-memcpy-expansion2193, !amdgpu.uniform !42

loop-memcpy-residual2196:                         ; preds = %loop-memcpy-residual2196.preheader, %loop-memcpy-residual2196
  %phi.broken3956 = phi i64 [ 0, %loop-memcpy-residual2196.preheader ], [ %2554, %loop-memcpy-residual2196 ]
  %residual-loop-index2198 = phi i64 [ %2552, %loop-memcpy-residual2196 ], [ 0, %loop-memcpy-residual2196.preheader ]
  %scevgep2867 = getelementptr i8, ptr %scevgep2866, i64 %residual-loop-index2198
  %2551 = load i8, ptr %scevgep2867, align 1
  %scevgep2865 = getelementptr i8, ptr %scevgep2864, i64 %residual-loop-index2198
  store i8 %2551, ptr %scevgep2865, align 1
  %2552 = add i64 %residual-loop-index2198, 1
  %2553 = icmp uge i64 %2552, %2536
  %2554 = call i64 @llvm.amdgcn.if.break.i64(i1 %2553, i64 %phi.broken3956)
  %2555 = call i1 @llvm.amdgcn.loop.i64(i64 %2554)
  br i1 %2555, label %Flow3406, label %loop-memcpy-residual2196

Flow3406:                                         ; preds = %loop-memcpy-residual2196
  %.lcssa3978 = phi i64 [ %2554, %loop-memcpy-residual2196 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa3978)
  br label %Flow3408, !amdgpu.uniform !42

post-loop-memcpy-expansion2193:                   ; preds = %Flow3408
  %tobool.not.i.i.i.i.i.i1933 = icmp eq i32 %.lcssa3983, 0
  %conv.i.i.i.i.i.i1934 = zext i1 %tobool.not.i.i.i.i.i.i1933 to i32
  fence release
  %2556 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %idxprom.i23.i.i.i.i.i1935 = zext i32 %.lcssa3985 to i64
  %arrayidx.i24.i.i.i.i.i1936 = getelementptr inbounds nuw i32, ptr %2556, i64 %idxprom.i23.i.i.i.i.i1935
  store atomic i32 %conv.i.i.i.i.i.i1934, ptr %arrayidx.i24.i.i.i.i.i1936 monotonic, align 4
  %2557 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !84
  %arrayidx.i.i5.i.i1937 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %2557, i64 %idxprom.i23.i.i.i.i.i1935, !amdgpu.uniform !42
  %2558 = load i64, ptr %arrayidx.i.i5.i.i1937, align 8, !tbaa !80
  %scevgep = getelementptr i8, ptr %2459, i64 56
  %2559 = add i64 %writer.sroa.10.4.i.i.i.i120.i, -56
  br label %while.cond.i.i.i.i1938, !amdgpu.uniform !42

while.cond.i.i.i.i1938:                           ; preds = %post-loop-memcpy-expansion2193, %Flow3405
  %phi.broken3959 = phi i64 [ 0, %post-loop-memcpy-expansion2193 ], [ %2576, %Flow3405 ]
  %lsr.iv2858 = phi i64 [ %2574, %Flow3405 ], [ %2559, %post-loop-memcpy-expansion2193 ]
  %lsr.iv = phi ptr [ %2573, %Flow3405 ], [ %scevgep, %post-loop-memcpy-expansion2193 ]
  %port.sroa.37.0.i.i1939 = phi i32 [ %2572, %Flow3405 ], [ %conv.i.i.i.i.i.i1934, %post-loop-memcpy-expansion2193 ]
  %idx.0.i.i.i.i1940 = phi i64 [ %2571, %Flow3405 ], [ 56, %post-loop-memcpy-expansion2193 ]
  %cmp.i.i.i.i1941 = icmp ugt i64 %writer.sroa.10.4.i.i.i.i120.i, %idx.0.i.i.i.i1940
  %2560 = tail call i64 @llvm.amdgcn.ballot.i64(i1 %cmp.i.i.i.i1941)
  %and.i.i.i.i6.i.i1942 = and i64 %2560, %2558
  %tobool.not.i.i.i.i1943 = icmp ne i64 %and.i.i.i.i6.i.i1942, 0
  %2561 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i22.i.i.i1944 = getelementptr inbounds nuw i32, ptr %2561, i64 %idxprom.i23.i.i.i.i.i1935, !amdgpu.uniform !42
  %2562 = load atomic i32, ptr %arrayidx.i22.i.i.i1944 monotonic, align 4
  %2563 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %tobool.not.i.i.i.i1943)
  %2564 = extractvalue { i1, i64 } %2563, 0
  %2565 = extractvalue { i1, i64 } %2563, 1
  br i1 %2564, label %cond.false.i8.i.i.i.i1945, label %Flow3405

cond.false.i8.i.i.i.i1945:                        ; preds = %while.cond.i.i.i.i1938
  %2566 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %2562)
  %2567 = icmp ne i32 %2566, %port.sroa.37.0.i.i1939
  br i1 %2567, label %while.body.i.i21.i.i.i.i1946.preheader, label %Flow3404, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i1946.preheader:           ; preds = %cond.false.i8.i.i.i.i1945
  br label %while.body.i.i21.i.i.i.i1946, !amdgpu.uniform !42

Flow3404:                                         ; preds = %Flow3403, %cond.false.i8.i.i.i.i1945
  br label %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i1949, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i1946:                     ; preds = %while.body.i.i21.i.i.i.i1946.preheader, %while.body.i.i21.i.i.i.i1946
  tail call void @llvm.amdgcn.s.sleep(i32 2)
  %2568 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i22.i.i.i.i1947 = getelementptr inbounds nuw i32, ptr %2568, i64 %idxprom.i23.i.i.i.i.i1935, !amdgpu.uniform !42
  %2569 = load atomic i32, ptr %arrayidx.i.i.i22.i.i.i.i1947 monotonic, align 4
  %2570 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %2569)
  %cmp.i.not.i.i23.i.i.i.i1948 = icmp eq i32 %2570, %port.sroa.37.0.i.i1939
  br i1 %cmp.i.not.i.i23.i.i.i.i1948, label %Flow3403, label %while.body.i.i21.i.i.i.i1946, !amdgpu.uniform !42

Flow3403:                                         ; preds = %while.body.i.i21.i.i.i.i1946
  br label %Flow3404, !amdgpu.uniform !42

Flow3405:                                         ; preds = %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1950, %while.cond.i.i.i.i1938
  %2571 = phi i64 [ %add.i.i.i.i1953, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1950 ], [ poison, %while.cond.i.i.i.i1938 ]
  %2572 = phi i32 [ %conv.i.i26.i.i.i.i1951, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1950 ], [ poison, %while.cond.i.i.i.i1938 ]
  %2573 = phi ptr [ %scevgep2854, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1950 ], [ poison, %while.cond.i.i.i.i1938 ]
  %2574 = phi i64 [ %lsr.iv.next, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1950 ], [ poison, %while.cond.i.i.i.i1938 ]
  %2575 = phi i1 [ false, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1950 ], [ true, %while.cond.i.i.i.i1938 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %2565)
  %2576 = call i64 @llvm.amdgcn.if.break.i64(i1 %2575, i64 %phi.broken3959)
  %2577 = call i1 @llvm.amdgcn.loop.i64(i64 %2576)
  br i1 %2577, label %cond.false.i.i.i1960, label %while.cond.i.i.i.i1938

_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i1949: ; preds = %Flow3404
  fence acquire
  %2578 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.i.i.i.i1941)
  %2579 = extractvalue { i1, i64 } %2578, 0
  %2580 = extractvalue { i1, i64 } %2578, 1
  br i1 %2579, label %if.then.i.i.i.i.i.i.i1954, label %Flow3402

if.then.i.i.i.i.i.i.i1954:                        ; preds = %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i1949
  %sub.i.i.i.i.i.i.i1956 = sub nuw i64 %writer.sroa.10.4.i.i.i.i120.i, %idx.0.i.i.i.i1940
  %spec.select.i.i.i36.i.i.i.i1957 = tail call i64 @llvm.umin.i64(i64 %sub.i.i.i.i.i.i.i1956, i64 64)
  %2581 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %2582 = and i64 %spec.select.i.i.i36.i.i.i.i1957, 15
  %2583 = sub i64 %spec.select.i.i.i36.i.i.i.i1957, %2582
  %2584 = icmp ne i64 %2583, 0
  %2585 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %2584)
  %2586 = extractvalue { i1, i64 } %2585, 0
  %2587 = extractvalue { i1, i64 } %2585, 1
  br i1 %2586, label %loop-memcpy-expansion.preheader, label %Flow3401

loop-memcpy-expansion.preheader:                  ; preds = %if.then.i.i.i.i.i.i.i1954
  %arrayidx.i20.i35.i.i.i.i1958 = getelementptr %"struct.rpc::Buffer", ptr %2581, i64 %idxprom.i19.i.i.i.i.i1928
  %arrayidx.i22.i38.i.i.i.i1959 = getelementptr %"struct.rpc::Buffer", ptr %arrayidx.i20.i35.i.i.i.i1958, i64 %sh_prom.i.i.i.i
  br label %loop-memcpy-expansion, !amdgpu.uniform !42

Flow3401:                                         ; preds = %Flow3400, %if.then.i.i.i.i.i.i.i1954
  call void @llvm.amdgcn.end.cf.i64(i64 %2587)
  br label %loop-memcpy-residual-header, !amdgpu.uniform !42

loop-memcpy-expansion:                            ; preds = %loop-memcpy-expansion.preheader, %loop-memcpy-expansion
  %phi.broken3957 = phi i64 [ 0, %loop-memcpy-expansion.preheader ], [ %2592, %loop-memcpy-expansion ]
  %loop-index = phi i64 [ %2590, %loop-memcpy-expansion ], [ 0, %loop-memcpy-expansion.preheader ]
  %scevgep2855 = getelementptr i8, ptr %lsr.iv, i64 %loop-index
  %2588 = load <4 x i32>, ptr %scevgep2855, align 1
  %2589 = getelementptr i8, ptr %arrayidx.i22.i38.i.i.i.i1959, i64 %loop-index
  store <4 x i32> %2588, ptr %2589, align 1
  %2590 = add i64 %loop-index, 16
  %2591 = icmp uge i64 %2590, %2583
  %2592 = call i64 @llvm.amdgcn.if.break.i64(i1 %2591, i64 %phi.broken3957)
  %2593 = call i1 @llvm.amdgcn.loop.i64(i64 %2592)
  br i1 %2593, label %Flow3400, label %loop-memcpy-expansion

Flow3399:                                         ; preds = %Flow3398, %loop-memcpy-residual-header
  call void @llvm.amdgcn.end.cf.i64(i64 %2706)
  br label %Flow3402, !amdgpu.uniform !42

loop-memcpy-residual:                             ; preds = %loop-memcpy-residual.preheader, %loop-memcpy-residual
  %phi.broken3958 = phi i64 [ 0, %loop-memcpy-residual.preheader ], [ %2597, %loop-memcpy-residual ]
  %residual-loop-index = phi i64 [ %2595, %loop-memcpy-residual ], [ 0, %loop-memcpy-residual.preheader ]
  %scevgep2862 = getelementptr i8, ptr %scevgep2861, i64 %residual-loop-index
  %2594 = load i8, ptr %scevgep2862, align 1
  %scevgep2860 = getelementptr i8, ptr %scevgep2859, i64 %residual-loop-index
  store i8 %2594, ptr %scevgep2860, align 1
  %2595 = add i64 %residual-loop-index, 1
  %2596 = icmp uge i64 %2595, %2582
  %2597 = call i64 @llvm.amdgcn.if.break.i64(i1 %2596, i64 %phi.broken3958)
  %2598 = call i1 @llvm.amdgcn.loop.i64(i64 %2597)
  br i1 %2598, label %Flow3398, label %loop-memcpy-residual

Flow3398:                                         ; preds = %loop-memcpy-residual
  %.lcssa3975 = phi i64 [ %2597, %loop-memcpy-residual ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa3975)
  br label %Flow3399, !amdgpu.uniform !42

_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1950: ; preds = %Flow3402
  %conv.i.i26.i.i.i.i1951 = xor i32 %port.sroa.37.0.i.i1939, 1
  fence release
  %2599 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %arrayidx.i24.i29.i.i.i.i1952 = getelementptr inbounds nuw i32, ptr %2599, i64 %idxprom.i23.i.i.i.i.i1935
  store atomic i32 %conv.i.i26.i.i.i.i1951, ptr %arrayidx.i24.i29.i.i.i.i1952 monotonic, align 4
  %add.i.i.i.i1953 = add i64 %idx.0.i.i.i.i1940, 64
  %scevgep2854 = getelementptr i8, ptr %lsr.iv, i64 64
  %lsr.iv.next = add i64 %lsr.iv2858, -64
  br label %Flow3405, !amdgpu.uniform !42

cond.false.i.i.i1960:                             ; preds = %Flow3405
  %.lcssa3977 = phi i64 [ %2576, %Flow3405 ]
  %port.sroa.37.0.i.i1939.lcssa = phi i32 [ %port.sroa.37.0.i.i1939, %Flow3405 ]
  %.lcssa3976 = phi i32 [ %2562, %Flow3405 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa3977)
  %2600 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %.lcssa3976)
  %2601 = icmp ne i32 %2600, %port.sroa.37.0.i.i1939.lcssa
  %2602 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %2601)
  %2603 = extractvalue { i1, i64 } %2602, 0
  %2604 = extractvalue { i1, i64 } %2602, 1
  br i1 %2603, label %while.body.i.i.i.i1961.preheader, label %Flow3397

while.body.i.i.i.i1961.preheader:                 ; preds = %cond.false.i.i.i1960
  br label %while.body.i.i.i.i1961, !amdgpu.uniform !42

Flow3397:                                         ; preds = %Flow3396, %cond.false.i.i.i1960
  call void @llvm.amdgcn.end.cf.i64(i64 %2604)
  br label %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsINS_3cpp6stringEEERS1_T_.exit1982, !amdgpu.uniform !42

while.body.i.i.i.i1961:                           ; preds = %while.body.i.i.i.i1961.preheader, %while.body.i.i.i.i1961
  %phi.broken3960 = phi i64 [ 0, %while.body.i.i.i.i1961.preheader ], [ %2608, %while.body.i.i.i.i1961 ]
  tail call void @llvm.amdgcn.s.sleep(i32 2)
  %2605 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i8.i.i1962 = getelementptr inbounds nuw i32, ptr %2605, i64 %idxprom.i23.i.i.i.i.i1935, !amdgpu.uniform !42
  %2606 = load atomic i32, ptr %arrayidx.i.i.i8.i.i1962 monotonic, align 4
  %2607 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %2606)
  %cmp.i.not.i.i.i.i1963 = icmp eq i32 %2607, %port.sroa.37.0.i.i1939.lcssa
  %2608 = call i64 @llvm.amdgcn.if.break.i64(i1 %cmp.i.not.i.i.i.i1963, i64 %phi.broken3960)
  %2609 = call i1 @llvm.amdgcn.loop.i64(i64 %2608)
  br i1 %2609, label %Flow3396, label %while.body.i.i.i.i1961

Flow3396:                                         ; preds = %while.body.i.i.i.i1961
  %.lcssa3973 = phi i64 [ %2608, %while.body.i.i.i.i1961 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa3973)
  br label %Flow3397, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsINS_3cpp6stringEEERS1_T_.exit1982: ; preds = %Flow3397
  fence acquire
  tail call void @llvm.amdgcn.wave.barrier()
  fence syncscope("agent") release
  %2610 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %.lcssa3984, i1 true)
  %iszero.i.i.i.i.i13.i.i1965 = icmp ne i64 %.lcssa3984, 0
  %cmp2.i.i.i.i14.i.i1966 = icmp eq i64 %2610, %sh_prom.i.i.i.i
  %cmp.i.i.i.i15.i.i1967 = select i1 %iszero.i.i.i.i.i13.i.i1965, i1 %cmp2.i.i.i.i14.i.i1966, i1 false
  %2611 = and i32 %.lcssa3985, 31
  %conv4.i.i.i19.i.i1971 = zext i1 %cmp.i.i.i.i15.i.i1967 to i32
  %shl.i.i.i20.i.i1972 = shl nuw i32 %conv4.i.i.i19.i.i1971, %2611
  %xor.i.i.i21.i.i1973 = xor i32 %shl.i.i.i20.i.i1972, -1
  %2612 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2678, !amdgpu.uniform !42

2613:                                             ; preds = %ComputeEnd2679
  %div8.i.i.i16.i.i1968 = lshr i32 %.lcssa3985, 5
  %idxprom.i.i.i17.i.i1969 = zext nneg i32 %div8.i.i.i16.i.i1968 to i64
  %2614 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i17.i.i1969
  %sunkaddr3334 = getelementptr inbounds i8, ptr addrspace(1) %2614, i64 40
  %2615 = atomicrmw and ptr addrspace(1) %sunkaddr3334, i32 %.lcssa3972 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %2616, !amdgpu.uniform !42

2616:                                             ; preds = %2613, %ComputeEnd2679
  call void @llvm.amdgcn.end.cf.i64(i64 %4001)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %for.cond.i.i.i1986, !amdgpu.uniform !42

for.cond.i.i.i1986:                               ; preds = %2616, %Flow3395
  %2617 = phi i32 [ %2655, %Flow3395 ], [ poison, %2616 ]
  %2618 = phi i64 [ %2656, %Flow3395 ], [ poison, %2616 ]
  %2619 = phi i32 [ %2657, %Flow3395 ], [ poison, %2616 ]
  %retval.sroa.6.0.i.i.i1987 = phi i32 [ %2660, %Flow3395 ], [ undef, %2616 ]
  %retval.sroa.2.0.i.i.i1989 = phi i64 [ %2659, %Flow3395 ], [ undef, %2616 ]
  %retval.sroa.8.0.i.i.i1990 = phi i32 [ %2658, %Flow3395 ], [ undef, %2616 ]
  %index.0.i.i.i1991 = phi i32 [ %2662, %Flow3395 ], [ 0, %2616 ]
  %2620 = load i32, ptr addrspace(1) @__llvm_rpc_client, align 8, !tbaa !68
  %cmp.not.i.i.i1992 = icmp ult i32 %index.0.i.i.i1991, %2620
  %spec.store.select.i.i.i1993 = select i1 %cmp.not.i.i.i1992, i32 %index.0.i.i.i1991, i32 0
  %2621 = tail call noundef i64 @llvm.amdgcn.ballot.i64(i1 true)
  %2622 = and i32 %spec.store.select.i.i.i1993, 31
  %2623 = lshr i64 %2621, %sh_prom.i.i.i.i
  %2624 = trunc i64 %2623 to i32
  %conv4.i.i.i.i.i1997 = and i32 %2624, 1
  %shl.i.i.i.i.i1998 = shl nuw i32 %conv4.i.i.i.i.i1997, %2622
  %2625 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2682, !amdgpu.uniform !42

2626:                                             ; preds = %ComputeEnd2683
  %sunkaddr3335 = getelementptr inbounds i8, ptr addrspace(1) %4011, i64 40
  %2627 = atomicrmw or ptr addrspace(1) %sunkaddr3335, i32 %.lcssa3966 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %2628, !amdgpu.uniform !42

2628:                                             ; preds = %2626, %ComputeEnd2683
  %2629 = phi i32 [ %2627, %2626 ], [ poison, %ComputeEnd2683 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %4020)
  %2630 = call i32 @llvm.amdgcn.readfirstlane.i32(i32 %2629)
  %2631 = or i32 %2630, %.lcssa3967
  %shl5.i.i.i.i.i1999 = shl nuw i32 1, %2622
  %and.i.i.i.i.i2000 = and i32 %shl5.i.i.i.i.i1999, %2631
  %tobool3.i.i.i.i2001 = icmp ne i32 %and.i.i.i.i.i2000, 0
  %2632 = tail call i64 @llvm.amdgcn.ballot.i64(i1 %tobool3.i.i.i.i2001)
  %and.i.i.i.i.i.i2002 = and i64 %2632, %2621
  %cmp.i.not.i.i.i2003 = icmp ne i64 %2621, %and.i.i.i.i.i.i2002
  br i1 %cmp.i.not.i.i.i2003, label %if.then.i.i.i.i2004, label %Flow3394, !amdgpu.uniform !42

if.then.i.i.i.i2004:                              ; preds = %2628
  fence syncscope("agent") acquire
  %2633 = load <2 x i64>, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !75
  %2634 = extractelement <2 x i64> %2633, i32 0
  %2635 = inttoptr i64 %2634 to ptr
  %2636 = extractelement <2 x i64> %2633, i32 1
  %2637 = inttoptr i64 %2636 to ptr
  %idxprom.i.i.i.i2005 = zext i32 %spec.store.select.i.i.i1993 to i64
  %arrayidx.i.i.i.i2006 = getelementptr inbounds nuw i32, ptr %2635, i64 %idxprom.i.i.i.i2005, !amdgpu.uniform !42
  %2638 = load atomic i32, ptr %arrayidx.i.i.i.i2006 monotonic, align 4
  %2639 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %2638)
  %arrayidx.i45.i.i.i2007 = getelementptr inbounds nuw i32, ptr %2637, i64 %idxprom.i.i.i.i2005, !amdgpu.uniform !42
  %2640 = load atomic i32, ptr %arrayidx.i45.i.i.i2007 monotonic, align 4
  %2641 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %2640)
  %cmp.i46.not.i.i.i2008 = icmp ne i32 %2639, %2641
  br i1 %cmp.i46.not.i.i.i2008, label %if.then12.i.i.i2009, label %Flow3392, !amdgpu.uniform !42

if.then12.i.i.i2009:                              ; preds = %if.then.i.i.i.i2004
  fence syncscope("agent") release
  %2642 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %2621, i1 true)
  %iszero.i.i.i.i.i.i.i2010 = icmp ne i64 %2621, 0
  %cmp2.i.i.i.i.i.i2011 = icmp eq i64 %2642, %sh_prom.i.i.i.i
  %cmp.i.i.i.i.i.i2012 = select i1 %iszero.i.i.i.i.i.i.i2010, i1 %cmp2.i.i.i.i.i.i2011, i1 false
  %conv4.i.i51.i.i.i2013 = zext i1 %cmp.i.i.i.i.i.i2012 to i32
  %shl.i.i52.i.i.i2014 = shl nuw i32 %conv4.i.i51.i.i.i2013, %2622
  %xor.i.i.i.i.i2015 = xor i32 %shl.i.i52.i.i.i2014, -1
  %2643 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2687, !amdgpu.uniform !42

2644:                                             ; preds = %ComputeEnd2688
  %sunkaddr3336 = getelementptr inbounds i8, ptr addrspace(1) %4011, i64 40
  %2645 = atomicrmw and ptr addrspace(1) %sunkaddr3336, i32 %.lcssa3968 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %2646, !amdgpu.uniform !42

2646:                                             ; preds = %2644, %ComputeEnd2688
  call void @llvm.amdgcn.end.cf.i64(i64 %4038)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %Flow3392, !amdgpu.uniform !42

if.end14.i.i.i2073:                               ; preds = %Flow3392
  %2647 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %2621, i1 true)
  %iszero.i.i.i.i.i.i2074 = icmp ne i64 %2621, 0
  %cmp2.i.i.i.i.i2075 = icmp eq i64 %2647, %sh_prom.i.i.i.i
  %cmp.i.i.i.i.i2076 = select i1 %iszero.i.i.i.i.i.i2074, i1 %cmp2.i.i.i.i.i2075, i1 false
  %2648 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.i.i.i.i.i2076)
  %2649 = extractvalue { i1, i64 } %2648, 0
  %2650 = extractvalue { i1, i64 } %2648, 1
  br i1 %2649, label %if.then16.i.i.i2078, label %if.end22.i.i.i2077

if.then16.i.i.i2078:                              ; preds = %if.end14.i.i.i2073
  %2651 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !76
  %opcode.i.i.i2079 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %2651, i64 %idxprom.i.i.i.i2005, i32 1
  store i32 1660944387, ptr %opcode.i.i.i2079, align 8, !tbaa !77
  %arrayidx21.i.i.i2080 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %2651, i64 %idxprom.i.i.i.i2005
  store i64 %2621, ptr %arrayidx21.i.i.i2080, align 8, !tbaa !80
  br label %if.end22.i.i.i2077, !amdgpu.uniform !42

Flow3393:                                         ; preds = %if.end22.i.i.i2077, %Flow3392
  %2652 = phi i32 [ %2639, %if.end22.i.i.i2077 ], [ %retval.sroa.8.0.i.i.i1990, %Flow3392 ]
  %2653 = phi i64 [ %2621, %if.end22.i.i.i2077 ], [ %retval.sroa.2.0.i.i.i1989, %Flow3392 ]
  %2654 = phi i32 [ %spec.store.select.i.i.i1993, %if.end22.i.i.i2077 ], [ %retval.sroa.6.0.i.i.i1987, %Flow3392 ]
  br label %cleanup26.i.i.i2016, !amdgpu.uniform !42

if.end22.i.i.i2077:                               ; preds = %if.then16.i.i.i2078, %if.end14.i.i.i2073
  call void @llvm.amdgcn.end.cf.i64(i64 %2650)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %Flow3393, !amdgpu.uniform !42

Flow3394:                                         ; preds = %cleanup26.i.i.i2016, %2628
  %2655 = phi i32 [ %2654, %cleanup26.i.i.i2016 ], [ %2617, %2628 ]
  %2656 = phi i64 [ %2653, %cleanup26.i.i.i2016 ], [ %2618, %2628 ]
  %2657 = phi i32 [ %2652, %cleanup26.i.i.i2016 ], [ %2619, %2628 ]
  %2658 = phi i32 [ %2652, %cleanup26.i.i.i2016 ], [ %retval.sroa.8.0.i.i.i1990, %2628 ]
  %2659 = phi i64 [ %2653, %cleanup26.i.i.i2016 ], [ %retval.sroa.2.0.i.i.i1989, %2628 ]
  %2660 = phi i32 [ %2654, %cleanup26.i.i.i2016 ], [ %retval.sroa.6.0.i.i.i1987, %2628 ]
  %2661 = phi i1 [ %cmp.i46.not.i.i.i2008, %cleanup26.i.i.i2016 ], [ true, %2628 ]
  br i1 %2661, label %for.inc.i.i.i2021, label %Flow3395, !amdgpu.uniform !42

cleanup26.i.i.i2016:                              ; preds = %Flow3393
  br label %Flow3394, !amdgpu.uniform !42

for.inc.i.i.i2021:                                ; preds = %Flow3394
  %inc.i.i.i2026 = add i32 %spec.store.select.i.i.i1993, 1
  br label %Flow3395, !amdgpu.uniform !42

Flow3395:                                         ; preds = %for.inc.i.i.i2021, %Flow3394
  %2662 = phi i32 [ %inc.i.i.i2026, %for.inc.i.i.i2021 ], [ poison, %Flow3394 ]
  %2663 = phi i1 [ false, %for.inc.i.i.i2021 ], [ true, %Flow3394 ]
  br i1 %2663, label %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i2027, label %for.cond.i.i.i1986, !amdgpu.uniform !42

_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i2027: ; preds = %Flow3395
  %.lcssa3971 = phi i32 [ %2655, %Flow3395 ]
  %.lcssa3970 = phi i64 [ %2656, %Flow3395 ]
  %.lcssa3969 = phi i32 [ %2657, %Flow3395 ]
  fence acquire
  %2664 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %mul.i.i.i.i.i.i2028 = shl i32 %.lcssa3971, 6
  %idxprom.i19.i.i.i.i.i2029 = zext i32 %mul.i.i.i.i.i.i2028 to i64
  %arrayidx.i20.i.i.i.i.i2030 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %2664, i64 %idxprom.i19.i.i.i.i.i2029
  %arrayidx.i22.i.i.i.i.i2031 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %arrayidx.i20.i.i.i.i.i2030, i64 %sh_prom.i.i.i.i
  store i64 1, ptr %arrayidx.i22.i.i.i.i.i2031, align 8, !tbaa !82
  %arrayidx6.i.i.i.i.i.i.i2032 = getelementptr inbounds nuw i8, ptr %arrayidx.i22.i.i.i.i.i2031, i64 8
  store i8 10, ptr %arrayidx6.i.i.i.i.i.i.i2032, align 8
  %tobool.not.i.i.i.i.i.i2033 = icmp eq i32 %.lcssa3969, 0
  %conv.i.i.i.i.i.i2034 = zext i1 %tobool.not.i.i.i.i.i.i2033 to i32
  fence release
  %2665 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %idxprom.i23.i.i.i.i.i2035 = zext i32 %.lcssa3971 to i64
  %arrayidx.i24.i.i.i.i.i2036 = getelementptr inbounds nuw i32, ptr %2665, i64 %idxprom.i23.i.i.i.i.i2035
  store atomic i32 %conv.i.i.i.i.i.i2034, ptr %arrayidx.i24.i.i.i.i.i2036 monotonic, align 4
  %2666 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !84
  %arrayidx.i.i5.i.i2037 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %2666, i64 %idxprom.i23.i.i.i.i.i2035, !amdgpu.uniform !42
  %2667 = load i64, ptr %arrayidx.i.i5.i.i2037, align 8, !tbaa !80
  br label %while.cond.i.i.i.i2038, !amdgpu.uniform !42

while.cond.i.i.i.i2038:                           ; preds = %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i2027, %Flow3391
  %phi.broken3961 = phi i64 [ 0, %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i2027 ], [ %2682, %Flow3391 ]
  %port.sroa.37.0.i.i2039 = phi i32 [ %2680, %Flow3391 ], [ %conv.i.i.i.i.i.i2034, %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i2027 ]
  %idx.0.i.i.i.i2040 = phi i64 [ %2679, %Flow3391 ], [ 56, %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i2027 ]
  %cmp.i.i.i.i2041 = icmp eq i64 %idx.0.i.i.i.i2040, 0
  %2668 = tail call i64 @llvm.amdgcn.ballot.i64(i1 %cmp.i.i.i.i2041)
  %and.i.i.i.i6.i.i2042 = and i64 %2668, %2667
  %tobool.not.i.i.i.i2043 = icmp ne i64 %and.i.i.i.i6.i.i2042, 0
  %2669 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i22.i.i.i2044 = getelementptr inbounds nuw i32, ptr %2669, i64 %idxprom.i23.i.i.i.i.i2035, !amdgpu.uniform !42
  %2670 = load atomic i32, ptr %arrayidx.i22.i.i.i2044 monotonic, align 4
  %2671 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %tobool.not.i.i.i.i2043)
  %2672 = extractvalue { i1, i64 } %2671, 0
  %2673 = extractvalue { i1, i64 } %2671, 1
  br i1 %2672, label %cond.false.i8.i.i.i.i2045, label %Flow3391

cond.false.i8.i.i.i.i2045:                        ; preds = %while.cond.i.i.i.i2038
  %2674 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %2670)
  %2675 = icmp ne i32 %2674, %port.sroa.37.0.i.i2039
  br i1 %2675, label %while.body.i.i21.i.i.i.i2046.preheader, label %Flow3390, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i2046.preheader:           ; preds = %cond.false.i8.i.i.i.i2045
  br label %while.body.i.i21.i.i.i.i2046, !amdgpu.uniform !42

Flow3390:                                         ; preds = %Flow3389, %cond.false.i8.i.i.i.i2045
  br label %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i2049, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i2046:                     ; preds = %while.body.i.i21.i.i.i.i2046.preheader, %while.body.i.i21.i.i.i.i2046
  tail call void @llvm.amdgcn.s.sleep(i32 2)
  %2676 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i22.i.i.i.i2047 = getelementptr inbounds nuw i32, ptr %2676, i64 %idxprom.i23.i.i.i.i.i2035, !amdgpu.uniform !42
  %2677 = load atomic i32, ptr %arrayidx.i.i.i22.i.i.i.i2047 monotonic, align 4
  %2678 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %2677)
  %cmp.i.not.i.i23.i.i.i.i2048 = icmp eq i32 %2678, %port.sroa.37.0.i.i2039
  br i1 %cmp.i.not.i.i23.i.i.i.i2048, label %Flow3389, label %while.body.i.i21.i.i.i.i2046, !amdgpu.uniform !42

Flow3389:                                         ; preds = %while.body.i.i21.i.i.i.i2046
  br label %Flow3390, !amdgpu.uniform !42

Flow3391:                                         ; preds = %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i2050, %while.cond.i.i.i.i2038
  %2679 = phi i64 [ %add.i.i.i.i2053, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i2050 ], [ poison, %while.cond.i.i.i.i2038 ]
  %2680 = phi i32 [ %conv.i.i26.i.i.i.i2051, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i2050 ], [ poison, %while.cond.i.i.i.i2038 ]
  %2681 = phi i1 [ false, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i2050 ], [ true, %while.cond.i.i.i.i2038 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %2673)
  %2682 = call i64 @llvm.amdgcn.if.break.i64(i1 %2681, i64 %phi.broken3961)
  %2683 = call i1 @llvm.amdgcn.loop.i64(i64 %2682)
  br i1 %2683, label %cond.false.i.i.i2059, label %while.cond.i.i.i.i2038

_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i2049: ; preds = %Flow3390
  fence acquire
  br i1 %cmp.i.i.i.i2041, label %if.then.i.i.i.i.i.i.i2054, label %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i2050, !amdgpu.uniform !42

if.then.i.i.i.i.i.i.i2054:                        ; preds = %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i2049
  %2684 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %arrayidx.i20.i35.i.i.i.i2057 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %2684, i64 %idxprom.i19.i.i.i.i.i2029
  %arrayidx.i22.i38.i.i.i.i2058 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %arrayidx.i20.i35.i.i.i.i2057, i64 %sh_prom.i.i.i.i
  store i8 10, ptr %arrayidx.i22.i38.i.i.i.i2058, align 1
  br label %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i2050, !amdgpu.uniform !42

_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i2050: ; preds = %if.then.i.i.i.i.i.i.i2054, %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i2049
  %conv.i.i26.i.i.i.i2051 = xor i32 %port.sroa.37.0.i.i2039, 1
  fence release
  %2685 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %arrayidx.i24.i29.i.i.i.i2052 = getelementptr inbounds nuw i32, ptr %2685, i64 %idxprom.i23.i.i.i.i.i2035
  store atomic i32 %conv.i.i26.i.i.i.i2051, ptr %arrayidx.i24.i29.i.i.i.i2052 monotonic, align 4
  %add.i.i.i.i2053 = add i64 %idx.0.i.i.i.i2040, 64
  br label %Flow3391, !amdgpu.uniform !42

cond.false.i.i.i2059:                             ; preds = %Flow3391
  %.lcssa3965 = phi i64 [ %2682, %Flow3391 ]
  %port.sroa.37.0.i.i2039.lcssa = phi i32 [ %port.sroa.37.0.i.i2039, %Flow3391 ]
  %.lcssa3964 = phi i32 [ %2670, %Flow3391 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa3965)
  %2686 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %.lcssa3964)
  %2687 = icmp ne i32 %2686, %port.sroa.37.0.i.i2039.lcssa
  %2688 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %2687)
  %2689 = extractvalue { i1, i64 } %2688, 0
  %2690 = extractvalue { i1, i64 } %2688, 1
  br i1 %2689, label %while.body.i.i.i.i2060.preheader, label %Flow3388

while.body.i.i.i.i2060.preheader:                 ; preds = %cond.false.i.i.i2059
  br label %while.body.i.i.i.i2060, !amdgpu.uniform !42

Flow3388:                                         ; preds = %Flow, %cond.false.i.i.i2059
  call void @llvm.amdgcn.end.cf.i64(i64 %2690)
  br label %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIcEERS1_T_.exit2081, !amdgpu.uniform !42

while.body.i.i.i.i2060:                           ; preds = %while.body.i.i.i.i2060.preheader, %while.body.i.i.i.i2060
  %phi.broken3962 = phi i64 [ 0, %while.body.i.i.i.i2060.preheader ], [ %2694, %while.body.i.i.i.i2060 ]
  tail call void @llvm.amdgcn.s.sleep(i32 2)
  %2691 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i8.i.i2061 = getelementptr inbounds nuw i32, ptr %2691, i64 %idxprom.i23.i.i.i.i.i2035, !amdgpu.uniform !42
  %2692 = load atomic i32, ptr %arrayidx.i.i.i8.i.i2061 monotonic, align 4
  %2693 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %2692)
  %cmp.i.not.i.i.i.i2062 = icmp eq i32 %2693, %port.sroa.37.0.i.i2039.lcssa
  %2694 = call i64 @llvm.amdgcn.if.break.i64(i1 %cmp.i.not.i.i.i.i2062, i64 %phi.broken3962)
  %2695 = call i1 @llvm.amdgcn.loop.i64(i64 %2694)
  br i1 %2695, label %Flow, label %while.body.i.i.i.i2060

Flow:                                             ; preds = %while.body.i.i.i.i2060
  %.lcssa3963 = phi i64 [ %2694, %while.body.i.i.i.i2060 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa3963)
  br label %Flow3388, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIcEERS1_T_.exit2081: ; preds = %Flow3388
  fence acquire
  tail call void @llvm.amdgcn.wave.barrier()
  fence syncscope("agent") release
  %2696 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %.lcssa3970, i1 true)
  %iszero.i.i.i.i.i13.i.i2064 = icmp ne i64 %.lcssa3970, 0
  %cmp2.i.i.i.i14.i.i2065 = icmp eq i64 %2696, %sh_prom.i.i.i.i
  %cmp.i.i.i.i15.i.i2066 = select i1 %iszero.i.i.i.i.i13.i.i2064, i1 %cmp2.i.i.i.i14.i.i2065, i1 false
  %2697 = and i32 %.lcssa3971, 31
  %conv4.i.i.i19.i.i2070 = zext i1 %cmp.i.i.i.i15.i.i2066 to i32
  %shl.i.i.i20.i.i2071 = shl nuw i32 %conv4.i.i.i19.i.i2070, %2697
  %xor.i.i.i21.i.i2072 = xor i32 %shl.i.i.i20.i.i2071, -1
  %2698 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2691, !amdgpu.uniform !42

2699:                                             ; preds = %ComputeEnd2692
  %div8.i.i.i16.i.i2067 = lshr i32 %.lcssa3971, 5
  %idxprom.i.i.i17.i.i2068 = zext nneg i32 %div8.i.i.i16.i.i2067 to i64
  %2700 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i17.i.i2068
  %sunkaddr3337 = getelementptr inbounds i8, ptr addrspace(1) %2700, i64 40
  %2701 = atomicrmw and ptr addrspace(1) %sunkaddr3337, i32 %.lcssa syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %2702, !amdgpu.uniform !42

2702:                                             ; preds = %2699, %ComputeEnd2692
  call void @llvm.amdgcn.end.cf.i64(i64 %4055)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %Flow3892, !amdgpu.uniform !42

_ZZN22__llvm_libc_22_0_0_git7testing8internal4testIcEEbPNS1_10RunContextENS0_8TestCondET_S6_PKcS8_NS1_8LocationEENKUlbNS_3cpp11string_viewEE_clEbSB_.exit: ; preds = %Flow3892
  ret i1 %cmp

Flow3400:                                         ; preds = %loop-memcpy-expansion
  %.lcssa3974 = phi i64 [ %2592, %loop-memcpy-expansion ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa3974)
  br label %Flow3401, !amdgpu.uniform !42

Flow3402:                                         ; preds = %Flow3399, %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i1949
  call void @llvm.amdgcn.end.cf.i64(i64 %2580)
  br label %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1950, !amdgpu.uniform !42

loop-memcpy-residual-header:                      ; preds = %Flow3401
  %2703 = icmp ne i64 %2582, 0
  %2704 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %2703)
  %2705 = extractvalue { i1, i64 } %2704, 0
  %2706 = extractvalue { i1, i64 } %2704, 1
  br i1 %2705, label %loop-memcpy-residual.preheader, label %Flow3399

loop-memcpy-residual.preheader:                   ; preds = %loop-memcpy-residual-header
  %umin = call i64 @llvm.umin.i64(i64 %lsr.iv2858, i64 64)
  %2707 = lshr i64 %umin, 4
  %2708 = shl nuw nsw i64 %2707, 4
  %scevgep2861 = getelementptr i8, ptr %lsr.iv, i64 %2708
  %scevgep2856 = getelementptr i8, ptr %2581, i64 %75
  %scevgep2857 = getelementptr i8, ptr %scevgep2856, i64 %2550
  %scevgep2859 = getelementptr i8, ptr %scevgep2857, i64 %2708
  br label %loop-memcpy-residual, !amdgpu.uniform !42

Flow3409:                                         ; preds = %loop-memcpy-expansion2194
  %.lcssa3979 = phi i64 [ %2548, %loop-memcpy-expansion2194 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa3979)
  br label %Flow3410, !amdgpu.uniform !42

loop-memcpy-residual-header2197:                  ; preds = %Flow3410
  %2709 = icmp eq i64 %2536, 0
  %2710 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %2709)
  %2711 = extractvalue { i1, i64 } %2710, 0
  %2712 = extractvalue { i1, i64 } %2710, 1
  br i1 %2711, label %loop-memcpy-residual-header2197.post-loop-memcpy-expansion2193_crit_edge, label %Flow3407

loop-memcpy-residual-header2197.post-loop-memcpy-expansion2193_crit_edge: ; preds = %loop-memcpy-residual-header2197
  %.pre3270 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i1928, 6
  br label %Flow3407, !amdgpu.uniform !42

Flow3407:                                         ; preds = %loop-memcpy-residual-header2197.post-loop-memcpy-expansion2193_crit_edge, %loop-memcpy-residual-header2197
  %2713 = phi i64 [ %.pre3270, %loop-memcpy-residual-header2197.post-loop-memcpy-expansion2193_crit_edge ], [ poison, %loop-memcpy-residual-header2197 ]
  %2714 = call { i1, i64 } @llvm.amdgcn.else.i64.i64(i64 %2712)
  %2715 = extractvalue { i1, i64 } %2714, 0
  %2716 = extractvalue { i1, i64 } %2714, 1
  br i1 %2715, label %loop-memcpy-residual2196.preheader, label %Flow3408

loop-memcpy-residual2196.preheader:               ; preds = %Flow3407
  %2717 = lshr i64 %spec.select.i.i.i.i.i.i.i1931, 4
  %2718 = shl nuw nsw i64 %2717, 4
  %2719 = add i64 %75, %2718
  %2720 = add i64 %2719, 8
  %scevgep2863 = getelementptr i8, ptr %2535, i64 %2720
  %2721 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i1928, 6
  %scevgep2864 = getelementptr i8, ptr %scevgep2863, i64 %2721
  %scevgep2866 = getelementptr i8, ptr %2459, i64 %2718
  br label %loop-memcpy-residual2196, !amdgpu.uniform !42

Flow3465:                                         ; preds = %loop-memcpy-expansion2200
  br label %Flow3466, !amdgpu.uniform !42

Flow3467:                                         ; preds = %Flow3463, %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i1737
  br label %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1738, !amdgpu.uniform !42

loop-memcpy-residual-header2203:                  ; preds = %Flow3466
  %2722 = icmp ne i64 %2239, 0
  br i1 %2722, label %loop-memcpy-residual2202.preheader, label %Flow3463, !amdgpu.uniform !42

loop-memcpy-residual2202.preheader:               ; preds = %loop-memcpy-residual-header2203
  %umin2907 = call i64 @llvm.umin.i64(i64 %lsr.iv2905, i64 64)
  %2723 = lshr i64 %umin2907, 4
  %2724 = shl nuw nsw i64 %2723, 4
  %scevgep2910 = getelementptr i8, ptr %lsr.iv2899, i64 %2724
  %scevgep2902 = getelementptr i8, ptr %2238, i64 %75
  %scevgep2903 = getelementptr i8, ptr %scevgep2902, i64 %2213
  %scevgep2908 = getelementptr i8, ptr %scevgep2903, i64 %2724
  br label %loop-memcpy-residual2202, !amdgpu.uniform !42

Flow3478:                                         ; preds = %loop-memcpy-expansion2206
  br label %Flow3479, !amdgpu.uniform !42

loop-memcpy-residual-header2209:                  ; preds = %Flow3479
  br i1 %2780, label %loop-memcpy-residual-header2209.post-loop-memcpy-expansion2205_crit_edge, label %Flow3475, !amdgpu.uniform !42

loop-memcpy-residual-header2209.post-loop-memcpy-expansion2205_crit_edge: ; preds = %loop-memcpy-residual-header2209
  %.pre3266 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i1716, 6
  br label %Flow3475, !amdgpu.uniform !42

Flow3475:                                         ; preds = %loop-memcpy-residual-header2209.post-loop-memcpy-expansion2205_crit_edge, %loop-memcpy-residual-header2209
  %2725 = phi i64 [ %.pre3266, %loop-memcpy-residual-header2209.post-loop-memcpy-expansion2205_crit_edge ], [ poison, %loop-memcpy-residual-header2209 ]
  %2726 = phi i1 [ false, %loop-memcpy-residual-header2209.post-loop-memcpy-expansion2205_crit_edge ], [ true, %loop-memcpy-residual-header2209 ]
  br i1 %2726, label %loop-memcpy-residual2208.preheader, label %Flow3476, !amdgpu.uniform !42

loop-memcpy-residual2208.preheader:               ; preds = %Flow3475
  %2727 = lshr i64 %spec.select.i.i.i.i.i.i.i374, 4
  %2728 = shl nuw nsw i64 %2727, 4
  %2729 = add i64 %75, %2728
  %2730 = add i64 %2729, 8
  %scevgep2912 = getelementptr i8, ptr %2206, i64 %2730
  %2731 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i1716, 6
  %scevgep2913 = getelementptr i8, ptr %scevgep2912, i64 %2731
  %scevgep2915 = getelementptr i8, ptr %2150, i64 %2728
  br label %loop-memcpy-residual2208, !amdgpu.uniform !42

Flow3508:                                         ; preds = %loop-memcpy-expansion2212
  %.lcssa4022 = phi i64 [ %2039, %loop-memcpy-expansion2212 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa4022)
  br label %Flow3509, !amdgpu.uniform !42

Flow3510:                                         ; preds = %Flow3506, %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i1532
  call void @llvm.amdgcn.end.cf.i64(i64 %2027)
  br label %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1533, !amdgpu.uniform !42

loop-memcpy-residual-header2215:                  ; preds = %Flow3509
  %2732 = icmp ne i64 %2029, 0
  %2733 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %2732)
  %2734 = extractvalue { i1, i64 } %2733, 0
  %2735 = extractvalue { i1, i64 } %2733, 1
  br i1 %2734, label %loop-memcpy-residual2214.preheader, label %Flow3506

loop-memcpy-residual2214.preheader:               ; preds = %loop-memcpy-residual-header2215
  %umin2938 = call i64 @llvm.umin.i64(i64 %lsr.iv2933, i64 64)
  %2736 = lshr i64 %umin2938, 4
  %2737 = shl nuw nsw i64 %2736, 4
  %scevgep2939 = getelementptr i8, ptr %lsr.iv2926, i64 %2737
  %scevgep2929 = getelementptr i8, ptr %2028, i64 %75
  %scevgep2930 = getelementptr i8, ptr %scevgep2929, i64 %1997
  %scevgep2936 = getelementptr i8, ptr %scevgep2930, i64 %2737
  br label %loop-memcpy-residual2214, !amdgpu.uniform !42

Flow3521:                                         ; preds = %loop-memcpy-expansion2218
  %.lcssa4027 = phi i64 [ %1995, %loop-memcpy-expansion2218 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa4027)
  br label %Flow3522, !amdgpu.uniform !42

loop-memcpy-residual-header2221:                  ; preds = %Flow3522
  %2738 = icmp eq i64 %1983, 0
  %2739 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %2738)
  %2740 = extractvalue { i1, i64 } %2739, 0
  %2741 = extractvalue { i1, i64 } %2739, 1
  br i1 %2740, label %loop-memcpy-residual-header2221.post-loop-memcpy-expansion2217_crit_edge, label %Flow3518

loop-memcpy-residual-header2221.post-loop-memcpy-expansion2217_crit_edge: ; preds = %loop-memcpy-residual-header2221
  %.pre3264 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i1511, 6
  br label %Flow3518, !amdgpu.uniform !42

Flow3518:                                         ; preds = %loop-memcpy-residual-header2221.post-loop-memcpy-expansion2217_crit_edge, %loop-memcpy-residual-header2221
  %2742 = phi i64 [ %.pre3264, %loop-memcpy-residual-header2221.post-loop-memcpy-expansion2217_crit_edge ], [ poison, %loop-memcpy-residual-header2221 ]
  %2743 = call { i1, i64 } @llvm.amdgcn.else.i64.i64(i64 %2741)
  %2744 = extractvalue { i1, i64 } %2743, 0
  %2745 = extractvalue { i1, i64 } %2743, 1
  br i1 %2744, label %loop-memcpy-residual2220.preheader, label %Flow3519

loop-memcpy-residual2220.preheader:               ; preds = %Flow3518
  %2746 = add nuw nsw i64 %75, 8
  %scevgep2941 = getelementptr i8, ptr %1982, i64 %2746
  %umin2944 = call i64 @llvm.umin.i64(i64 %lsr.iv.next2943.lcssa, i64 56)
  %2747 = lshr i64 %umin2944, 4
  %2748 = shl nuw nsw i64 %2747, 4
  %scevgep2945 = getelementptr i8, ptr %scevgep2941, i64 %2748
  %2749 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i1511, 6
  %scevgep2946 = getelementptr i8, ptr %scevgep2945, i64 %2749
  %scevgep2948 = getelementptr i8, ptr %RHSStr, i64 %2748
  br label %loop-memcpy-residual2220, !amdgpu.uniform !42

Flow3623:                                         ; preds = %loop-memcpy-expansion2224
  %.lcssa4077 = phi i64 [ %1487, %loop-memcpy-expansion2224 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa4077)
  br label %Flow3624, !amdgpu.uniform !42

Flow3625:                                         ; preds = %Flow3621, %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i1017
  call void @llvm.amdgcn.end.cf.i64(i64 %1475)
  br label %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1018, !amdgpu.uniform !42

loop-memcpy-residual-header2227:                  ; preds = %Flow3624
  %2750 = icmp ne i64 %1477, 0
  %2751 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %2750)
  %2752 = extractvalue { i1, i64 } %2751, 0
  %2753 = extractvalue { i1, i64 } %2751, 1
  br i1 %2752, label %loop-memcpy-residual2226.preheader, label %Flow3621

loop-memcpy-residual2226.preheader:               ; preds = %loop-memcpy-residual-header2227
  %umin3022 = call i64 @llvm.umin.i64(i64 %lsr.iv3020, i64 64)
  %2754 = lshr i64 %umin3022, 4
  %2755 = shl nuw nsw i64 %2754, 4
  %scevgep3025 = getelementptr i8, ptr %lsr.iv3015, i64 %2755
  %scevgep3018 = getelementptr i8, ptr %1476, i64 %75
  %scevgep3019 = getelementptr i8, ptr %scevgep3018, i64 %1445
  %scevgep3023 = getelementptr i8, ptr %scevgep3019, i64 %2755
  br label %loop-memcpy-residual2226, !amdgpu.uniform !42

Flow3636:                                         ; preds = %loop-memcpy-expansion2230
  %.lcssa4082 = phi i64 [ %1443, %loop-memcpy-expansion2230 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa4082)
  br label %Flow3637, !amdgpu.uniform !42

loop-memcpy-residual-header2233:                  ; preds = %Flow3637
  %2756 = icmp eq i64 %1431, 0
  %2757 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %2756)
  %2758 = extractvalue { i1, i64 } %2757, 0
  %2759 = extractvalue { i1, i64 } %2757, 1
  br i1 %2758, label %loop-memcpy-residual-header2233.post-loop-memcpy-expansion2229_crit_edge, label %Flow3633

loop-memcpy-residual-header2233.post-loop-memcpy-expansion2229_crit_edge: ; preds = %loop-memcpy-residual-header2233
  %.pre3256 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i996, 6
  br label %Flow3633, !amdgpu.uniform !42

Flow3633:                                         ; preds = %loop-memcpy-residual-header2233.post-loop-memcpy-expansion2229_crit_edge, %loop-memcpy-residual-header2233
  %2760 = phi i64 [ %.pre3256, %loop-memcpy-residual-header2233.post-loop-memcpy-expansion2229_crit_edge ], [ poison, %loop-memcpy-residual-header2233 ]
  %2761 = call { i1, i64 } @llvm.amdgcn.else.i64.i64(i64 %2759)
  %2762 = extractvalue { i1, i64 } %2761, 0
  %2763 = extractvalue { i1, i64 } %2761, 1
  br i1 %2762, label %loop-memcpy-residual2232.preheader, label %Flow3634

loop-memcpy-residual2232.preheader:               ; preds = %Flow3633
  %2764 = lshr i64 %spec.select.i.i.i.i.i.i.i999, 4
  %2765 = shl nuw nsw i64 %2764, 4
  %2766 = add i64 %75, %2765
  %2767 = add i64 %2766, 8
  %scevgep3027 = getelementptr i8, ptr %1430, i64 %2767
  %2768 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i996, 6
  %scevgep3028 = getelementptr i8, ptr %scevgep3027, i64 %2768
  %scevgep3030 = getelementptr i8, ptr %1354, i64 %2765
  br label %loop-memcpy-residual2232, !amdgpu.uniform !42

Flow3699:                                         ; preds = %loop-memcpy-expansion2236
  br label %Flow3700, !amdgpu.uniform !42

Flow3701:                                         ; preds = %Flow3698, %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i805
  br label %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i806, !amdgpu.uniform !42

loop-memcpy-residual-header2239:                  ; preds = %Flow3700
  %2769 = icmp ne i64 %1115, 0
  br i1 %2769, label %loop-memcpy-residual2238.preheader, label %Flow3698, !amdgpu.uniform !42

loop-memcpy-residual2238.preheader:               ; preds = %loop-memcpy-residual-header2239
  %umin3071 = call i64 @llvm.umin.i64(i64 %lsr.iv3069, i64 64)
  %2770 = lshr i64 %umin3071, 4
  %2771 = shl nuw nsw i64 %2770, 4
  %scevgep3074 = getelementptr i8, ptr %lsr.iv3063, i64 %2771
  %scevgep3066 = getelementptr i8, ptr %1114, i64 %75
  %scevgep3067 = getelementptr i8, ptr %scevgep3066, i64 %1089
  %scevgep3072 = getelementptr i8, ptr %scevgep3067, i64 %2771
  br label %loop-memcpy-residual2238, !amdgpu.uniform !42

Flow3708:                                         ; preds = %loop-memcpy-expansion2242
  br label %Flow3709, !amdgpu.uniform !42

loop-memcpy-residual-header2245:                  ; preds = %Flow3709
  br i1 %2780, label %loop-memcpy-residual-header2245.post-loop-memcpy-expansion2241_crit_edge, label %Flow3706, !amdgpu.uniform !42

loop-memcpy-residual-header2245.post-loop-memcpy-expansion2241_crit_edge: ; preds = %loop-memcpy-residual-header2245
  %.pre3252 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i784, 6
  br label %Flow3706, !amdgpu.uniform !42

Flow3706:                                         ; preds = %loop-memcpy-residual-header2245.post-loop-memcpy-expansion2241_crit_edge, %loop-memcpy-residual-header2245
  %2772 = phi i64 [ %.pre3252, %loop-memcpy-residual-header2245.post-loop-memcpy-expansion2241_crit_edge ], [ poison, %loop-memcpy-residual-header2245 ]
  %2773 = phi i1 [ false, %loop-memcpy-residual-header2245.post-loop-memcpy-expansion2241_crit_edge ], [ true, %loop-memcpy-residual-header2245 ]
  br i1 %2773, label %loop-memcpy-residual2244.preheader, label %Flow3707, !amdgpu.uniform !42

loop-memcpy-residual2244.preheader:               ; preds = %Flow3706
  %2774 = add i64 %75, %655
  %2775 = add i64 %2774, 8
  %scevgep3076 = getelementptr i8, ptr %1082, i64 %2775
  %2776 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i784, 6
  %scevgep3077 = getelementptr i8, ptr %scevgep3076, i64 %2776
  %scevgep3079 = getelementptr i8, ptr %1026, i64 %655
  br label %loop-memcpy-residual2244, !amdgpu.uniform !42

Flow3769:                                         ; preds = %loop-memcpy-expansion2248
  br label %Flow3770, !amdgpu.uniform !42

Flow3771:                                         ; preds = %Flow3768, %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i392
  br label %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i393, !amdgpu.uniform !42

loop-memcpy-residual-header2251:                  ; preds = %Flow3770
  %2777 = icmp ne i64 %690, 0
  br i1 %2777, label %loop-memcpy-residual2250.preheader, label %Flow3768, !amdgpu.uniform !42

loop-memcpy-residual2250.preheader:               ; preds = %loop-memcpy-residual-header2251
  %umin3135 = call i64 @llvm.umin.i64(i64 %lsr.iv3133, i64 64)
  %2778 = lshr i64 %umin3135, 4
  %2779 = shl nuw nsw i64 %2778, 4
  %scevgep3138 = getelementptr i8, ptr %lsr.iv3127, i64 %2779
  %scevgep3130 = getelementptr i8, ptr %689, i64 %75
  %scevgep3131 = getelementptr i8, ptr %scevgep3130, i64 %663
  %scevgep3136 = getelementptr i8, ptr %scevgep3131, i64 %2779
  br label %loop-memcpy-residual2250, !amdgpu.uniform !42

Flow3778:                                         ; preds = %loop-memcpy-expansion2254
  br label %Flow3779, !amdgpu.uniform !42

loop-memcpy-residual-header2257:                  ; preds = %Flow3779
  %2780 = icmp eq i64 %654, 0
  br i1 %2780, label %loop-memcpy-residual-header2257.post-loop-memcpy-expansion2253_crit_edge, label %Flow3776, !amdgpu.uniform !42

loop-memcpy-residual-header2257.post-loop-memcpy-expansion2253_crit_edge: ; preds = %loop-memcpy-residual-header2257
  %.pre3246 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i371, 6
  br label %Flow3776, !amdgpu.uniform !42

Flow3776:                                         ; preds = %loop-memcpy-residual-header2257.post-loop-memcpy-expansion2253_crit_edge, %loop-memcpy-residual-header2257
  %2781 = phi i64 [ %.pre3246, %loop-memcpy-residual-header2257.post-loop-memcpy-expansion2253_crit_edge ], [ poison, %loop-memcpy-residual-header2257 ]
  %2782 = phi i1 [ false, %loop-memcpy-residual-header2257.post-loop-memcpy-expansion2253_crit_edge ], [ true, %loop-memcpy-residual-header2257 ]
  br i1 %2782, label %loop-memcpy-residual2256.preheader, label %Flow3777, !amdgpu.uniform !42

loop-memcpy-residual2256.preheader:               ; preds = %Flow3776
  %2783 = add i64 %75, %655
  %2784 = add i64 %2783, 8
  %scevgep3140 = getelementptr i8, ptr %653, i64 %2784
  %2785 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i371, 6
  %scevgep3141 = getelementptr i8, ptr %scevgep3140, i64 %2785
  %scevgep3143 = getelementptr i8, ptr %594, i64 %655
  br label %loop-memcpy-residual2256, !amdgpu.uniform !42

Flow3811:                                         ; preds = %loop-memcpy-expansion2260
  %.lcssa4165 = phi i64 [ %458, %loop-memcpy-expansion2260 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa4165)
  br label %Flow3812, !amdgpu.uniform !42

Flow3813:                                         ; preds = %Flow3810, %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i181
  call void @llvm.amdgcn.end.cf.i64(i64 %446)
  br label %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i182, !amdgpu.uniform !42

loop-memcpy-residual-header2263:                  ; preds = %Flow3812
  %2786 = icmp ne i64 %448, 0
  %2787 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %2786)
  %2788 = extractvalue { i1, i64 } %2787, 0
  %2789 = extractvalue { i1, i64 } %2787, 1
  br i1 %2788, label %loop-memcpy-residual2262.preheader, label %Flow3810

loop-memcpy-residual2262.preheader:               ; preds = %loop-memcpy-residual-header2263
  %umin3179 = call i64 @llvm.umin.i64(i64 %lsr.iv3177, i64 64)
  %2790 = lshr i64 %umin3179, 4
  %2791 = shl nuw nsw i64 %2790, 4
  %scevgep3182 = getelementptr i8, ptr %lsr.iv3172, i64 %2791
  %scevgep3175 = getelementptr i8, ptr %447, i64 %75
  %scevgep3176 = getelementptr i8, ptr %scevgep3175, i64 %416
  %scevgep3180 = getelementptr i8, ptr %scevgep3176, i64 %2791
  br label %loop-memcpy-residual2262, !amdgpu.uniform !42

Flow3820:                                         ; preds = %loop-memcpy-expansion2266
  %.lcssa4170 = phi i64 [ %414, %loop-memcpy-expansion2266 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa4170)
  br label %Flow3821, !amdgpu.uniform !42

loop-memcpy-residual-header2269:                  ; preds = %Flow3821
  %2792 = icmp eq i64 %402, 0
  %2793 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %2792)
  %2794 = extractvalue { i1, i64 } %2793, 0
  %2795 = extractvalue { i1, i64 } %2793, 1
  br i1 %2794, label %loop-memcpy-residual-header2269.post-loop-memcpy-expansion2265_crit_edge, label %Flow3818

loop-memcpy-residual-header2269.post-loop-memcpy-expansion2265_crit_edge: ; preds = %loop-memcpy-residual-header2269
  %.pre3242 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i160, 6
  br label %Flow3818, !amdgpu.uniform !42

Flow3818:                                         ; preds = %loop-memcpy-residual-header2269.post-loop-memcpy-expansion2265_crit_edge, %loop-memcpy-residual-header2269
  %2796 = phi i64 [ %.pre3242, %loop-memcpy-residual-header2269.post-loop-memcpy-expansion2265_crit_edge ], [ poison, %loop-memcpy-residual-header2269 ]
  %2797 = call { i1, i64 } @llvm.amdgcn.else.i64.i64(i64 %2795)
  %2798 = extractvalue { i1, i64 } %2797, 0
  %2799 = extractvalue { i1, i64 } %2797, 1
  br i1 %2798, label %loop-memcpy-residual2268.preheader, label %Flow3819

loop-memcpy-residual2268.preheader:               ; preds = %Flow3818
  %2800 = lshr i64 %spec.select.i.i.i.i.i.i.i163, 4
  %2801 = shl nuw nsw i64 %2800, 4
  %2802 = add i64 %75, %2801
  %2803 = add i64 %2802, 8
  %scevgep3184 = getelementptr i8, ptr %401, i64 %2803
  %2804 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i160, 6
  %scevgep3185 = getelementptr i8, ptr %scevgep3184, i64 %2804
  %scevgep3187 = getelementptr i8, ptr %325, i64 %2801
  br label %loop-memcpy-residual2268, !amdgpu.uniform !42

Flow3488:                                         ; preds = %loop-memcpy-expansion2272
  br label %Flow3489, !amdgpu.uniform !42

loop-memcpy-residual-header2275:                  ; preds = %Flow3489
  br i1 %2807, label %loop-memcpy-residual2274.preheader, label %Flow3486, !amdgpu.uniform !42

loop-memcpy-residual2274.preheader:               ; preds = %loop-memcpy-residual-header2275
  %2805 = sub i64 sub (i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) getelementptr inbounds nuw (i8, ptr addrspace(4) @.str.9, i64 8) to ptr) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.9 to ptr) to i64)), %umin3132
  %2806 = sub i64 %2805, %595
  %scevgep2921 = getelementptr i8, ptr %2150, i64 %2806
  %scevgep2923 = getelementptr i8, ptr %3, i64 %2806
  br label %loop-memcpy-residual2274, !amdgpu.uniform !42

Flow3716:                                         ; preds = %loop-memcpy-expansion2278
  br label %Flow3717, !amdgpu.uniform !42

loop-memcpy-residual-header2281:                  ; preds = %Flow3717
  br i1 %2807, label %loop-memcpy-residual2280.preheader, label %Flow3715, !amdgpu.uniform !42

loop-memcpy-residual2280.preheader:               ; preds = %loop-memcpy-residual-header2281
  %scevgep3084 = getelementptr i8, ptr %1026, i64 %596
  %scevgep3086 = getelementptr i8, ptr %3, i64 %596
  br label %loop-memcpy-residual2280, !amdgpu.uniform !42

Flow3786:                                         ; preds = %loop-memcpy-expansion2284
  br label %Flow3787, !amdgpu.uniform !42

loop-memcpy-residual-header2287:                  ; preds = %Flow3787
  %2807 = icmp ne i64 %595, 0
  br i1 %2807, label %loop-memcpy-residual2286.preheader, label %Flow3785, !amdgpu.uniform !42

loop-memcpy-residual2286.preheader:               ; preds = %loop-memcpy-residual-header2287
  %scevgep3148 = getelementptr i8, ptr %594, i64 %596
  %scevgep3150 = getelementptr i8, ptr %3, i64 %596
  br label %loop-memcpy-residual2286, !amdgpu.uniform !42

Flow3828:                                         ; preds = %loop-memcpy-expansion2290
  %.lcssa4178 = phi i64 [ %347, %loop-memcpy-expansion2290 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa4178)
  br label %Flow3829, !amdgpu.uniform !42

loop-memcpy-residual-header2293:                  ; preds = %Flow3829
  %2808 = icmp ne i64 %336, 0
  %2809 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %2808)
  %2810 = extractvalue { i1, i64 } %2809, 0
  %2811 = extractvalue { i1, i64 } %2809, 1
  br i1 %2810, label %loop-memcpy-residual2292.preheader, label %Flow3827

loop-memcpy-residual2292.preheader:               ; preds = %loop-memcpy-residual-header2293
  %2812 = lshr i64 %writer.sroa.10.4.i.i.i.i, 4
  %2813 = trunc i64 %2812 to i32
  %2814 = shl i32 %2813, 4
  %2815 = add i32 %2814, 11
  %2816 = sub i32 %2815, %335
  %scevgep3192 = getelementptr i8, ptr addrspace(5) %buffer.i.i.i, i32 %2816
  %2817 = shl nuw i64 %2812, 4
  %scevgep3195 = getelementptr i8, ptr %325, i64 %2817
  br label %loop-memcpy-residual2292, !amdgpu.uniform !42

Flow3417:                                         ; preds = %loop-memcpy-expansion2296
  %.lcssa3987 = phi i64 [ %2481, %loop-memcpy-expansion2296 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa3987)
  br label %Flow3418, !amdgpu.uniform !42

loop-memcpy-residual-header2299:                  ; preds = %Flow3418
  %2818 = icmp ne i64 %2470, 0
  %2819 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %2818)
  %2820 = extractvalue { i1, i64 } %2819, 0
  %2821 = extractvalue { i1, i64 } %2819, 1
  br i1 %2820, label %loop-memcpy-residual2298.preheader, label %Flow3416

loop-memcpy-residual2298.preheader:               ; preds = %loop-memcpy-residual-header2299
  %2822 = lshr i64 %writer.sroa.10.4.i.i.i.i120.i, 4
  %2823 = trunc i64 %2822 to i32
  %2824 = shl i32 %2823, 4
  %2825 = add i32 %2824, 11
  %2826 = sub i32 %2825, %2469
  %scevgep2871 = getelementptr i8, ptr addrspace(5) %buffer.i.i.i, i32 %2826
  %2827 = shl nuw i64 %2822, 4
  %scevgep2874 = getelementptr i8, ptr %2459, i64 %2827
  br label %loop-memcpy-residual2298, !amdgpu.uniform !42

Flow3646:                                         ; preds = %loop-memcpy-expansion2302
  %.lcssa4090 = phi i64 [ %1376, %loop-memcpy-expansion2302 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa4090)
  br label %Flow3647, !amdgpu.uniform !42

loop-memcpy-residual-header2305:                  ; preds = %Flow3647
  %2828 = icmp ne i64 %1365, 0
  %2829 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %2828)
  %2830 = extractvalue { i1, i64 } %2829, 0
  %2831 = extractvalue { i1, i64 } %2829, 1
  br i1 %2830, label %loop-memcpy-residual2304.preheader, label %Flow3644

loop-memcpy-residual2304.preheader:               ; preds = %loop-memcpy-residual-header2305
  %2832 = lshr i64 %writer.sroa.10.5.i.i.i.i.i, 4
  %2833 = trunc i64 %2832 to i32
  %2834 = shl i32 %2833, 4
  %2835 = add i32 %2834, 11
  %2836 = sub i32 %2835, %1364
  %scevgep3035 = getelementptr i8, ptr addrspace(5) %buffer.i.i.i, i32 %2836
  %2837 = shl nuw i64 %2832, 4
  %scevgep3038 = getelementptr i8, ptr %1354, i64 %2837
  br label %loop-memcpy-residual2304, !amdgpu.uniform !42

Flow3881:                                         ; preds = %loop-memcpy-expansion2308
  br label %Flow3882, !amdgpu.uniform !42

loop-memcpy-residual-header2311:                  ; preds = %Flow3882
  %2838 = icmp eq i64 %65, 0
  br i1 %2838, label %loop-memcpy-residual-header2311.post-loop-memcpy-expansion2307_crit_edge, label %Flow3879, !amdgpu.uniform !42

loop-memcpy-residual-header2311.post-loop-memcpy-expansion2307_crit_edge: ; preds = %loop-memcpy-residual-header2311
  %.pre = shl nuw nsw i64 %sh_prom.i.i.i.i, 6
  %.pre3238 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i, 6
  br label %Flow3879, !amdgpu.uniform !42

Flow3879:                                         ; preds = %loop-memcpy-residual-header2311.post-loop-memcpy-expansion2307_crit_edge, %loop-memcpy-residual-header2311
  %2839 = phi i64 [ %.pre, %loop-memcpy-residual-header2311.post-loop-memcpy-expansion2307_crit_edge ], [ poison, %loop-memcpy-residual-header2311 ]
  %2840 = phi i64 [ %.pre3238, %loop-memcpy-residual-header2311.post-loop-memcpy-expansion2307_crit_edge ], [ poison, %loop-memcpy-residual-header2311 ]
  %2841 = phi i1 [ false, %loop-memcpy-residual-header2311.post-loop-memcpy-expansion2307_crit_edge ], [ true, %loop-memcpy-residual-header2311 ]
  br i1 %2841, label %loop-memcpy-residual2310.preheader, label %Flow3880, !amdgpu.uniform !42

loop-memcpy-residual2310.preheader:               ; preds = %Flow3879
  %2842 = shl nuw nsw i64 %sh_prom.i.i.i.i, 6
  %2843 = add nuw nsw i64 %2842, 56
  %scevgep3231 = getelementptr i8, ptr %64, i64 %2843
  %2844 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i, 6
  %scevgep3232 = getelementptr i8, ptr %scevgep3231, i64 %2844
  br label %loop-memcpy-residual2310, !amdgpu.uniform !42

Flow3872:                                         ; preds = %loop-memcpy-expansion2314
  br label %Flow3873, !amdgpu.uniform !42

Flow3874:                                         ; preds = %Flow3871, %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i
  br label %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i, !amdgpu.uniform !42

loop-memcpy-residual-header2317:                  ; preds = %Flow3873
  %2845 = icmp ne i64 %102, 0
  br i1 %2845, label %loop-memcpy-residual2316.preheader, label %Flow3871, !amdgpu.uniform !42

loop-memcpy-residual2316.preheader:               ; preds = %loop-memcpy-residual-header2317
  %umin3226 = call i64 @llvm.umin.i64(i64 %lsr.iv3224, i64 64)
  %2846 = lshr i64 %umin3226, 4
  %2847 = shl nuw nsw i64 %2846, 4
  %scevgep3229 = getelementptr i8, ptr addrspace(4) %lsr.iv3219, i64 %2847
  %scevgep3222 = getelementptr i8, ptr %101, i64 %75
  %scevgep3223 = getelementptr i8, ptr %scevgep3222, i64 %76
  %scevgep3227 = getelementptr i8, ptr %scevgep3223, i64 %2847
  br label %loop-memcpy-residual2316, !amdgpu.uniform !42

Flow3862:                                         ; preds = %loop-memcpy-expansion2320
  br label %Flow3863, !amdgpu.uniform !42

loop-memcpy-residual-header2323:                  ; preds = %Flow3863
  %2848 = icmp eq i64 %178, 0
  br i1 %2848, label %loop-memcpy-residual-header2323.post-loop-memcpy-expansion2319_crit_edge, label %Flow3860, !amdgpu.uniform !42

loop-memcpy-residual-header2323.post-loop-memcpy-expansion2319_crit_edge: ; preds = %loop-memcpy-residual-header2323
  %.pre3240 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i56, 6
  br label %Flow3860, !amdgpu.uniform !42

Flow3860:                                         ; preds = %loop-memcpy-residual-header2323.post-loop-memcpy-expansion2319_crit_edge, %loop-memcpy-residual-header2323
  %2849 = phi i64 [ %.pre3240, %loop-memcpy-residual-header2323.post-loop-memcpy-expansion2319_crit_edge ], [ poison, %loop-memcpy-residual-header2323 ]
  %2850 = phi i1 [ false, %loop-memcpy-residual-header2323.post-loop-memcpy-expansion2319_crit_edge ], [ true, %loop-memcpy-residual-header2323 ]
  br i1 %2850, label %loop-memcpy-residual2322.preheader, label %Flow3861, !amdgpu.uniform !42

loop-memcpy-residual2322.preheader:               ; preds = %Flow3860
  %2851 = add nuw nsw i64 %75, 8
  %scevgep3212 = getelementptr i8, ptr %177, i64 %2851
  %2852 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i56, 6
  %scevgep3213 = getelementptr i8, ptr %scevgep3212, i64 %2852
  br label %loop-memcpy-residual2322, !amdgpu.uniform !42

Flow3853:                                         ; preds = %loop-memcpy-expansion2326
  br label %Flow3854, !amdgpu.uniform !42

Flow3855:                                         ; preds = %Flow3852, %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i77
  br label %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i78, !amdgpu.uniform !42

loop-memcpy-residual-header2329:                  ; preds = %Flow3854
  %2853 = icmp ne i64 %213, 0
  br i1 %2853, label %loop-memcpy-residual2328.preheader, label %Flow3852, !amdgpu.uniform !42

loop-memcpy-residual2328.preheader:               ; preds = %loop-memcpy-residual-header2329
  %umin3207 = call i64 @llvm.umin.i64(i64 %lsr.iv3205, i64 64)
  %2854 = lshr i64 %umin3207, 4
  %2855 = shl nuw nsw i64 %2854, 4
  %scevgep3210 = getelementptr i8, ptr addrspace(4) %lsr.iv3200, i64 %2855
  %scevgep3203 = getelementptr i8, ptr %212, i64 %75
  %scevgep3204 = getelementptr i8, ptr %scevgep3203, i64 %187
  %scevgep3208 = getelementptr i8, ptr %scevgep3204, i64 %2855
  br label %loop-memcpy-residual2328, !amdgpu.uniform !42

Flow3801:                                         ; preds = %loop-memcpy-expansion2332
  br label %Flow3802, !amdgpu.uniform !42

loop-memcpy-residual-header2335:                  ; preds = %Flow3802
  %2856 = icmp eq i64 %531, 0
  br i1 %2856, label %loop-memcpy-residual-header2335.post-loop-memcpy-expansion2331_crit_edge, label %Flow3799, !amdgpu.uniform !42

loop-memcpy-residual-header2335.post-loop-memcpy-expansion2331_crit_edge: ; preds = %loop-memcpy-residual-header2335
  %.pre3244 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i265, 6
  br label %Flow3799, !amdgpu.uniform !42

Flow3799:                                         ; preds = %loop-memcpy-residual-header2335.post-loop-memcpy-expansion2331_crit_edge, %loop-memcpy-residual-header2335
  %2857 = phi i64 [ %.pre3244, %loop-memcpy-residual-header2335.post-loop-memcpy-expansion2331_crit_edge ], [ poison, %loop-memcpy-residual-header2335 ]
  %2858 = phi i1 [ false, %loop-memcpy-residual-header2335.post-loop-memcpy-expansion2331_crit_edge ], [ true, %loop-memcpy-residual-header2335 ]
  br i1 %2858, label %loop-memcpy-residual2334.preheader, label %Flow3800, !amdgpu.uniform !42

loop-memcpy-residual2334.preheader:               ; preds = %Flow3799
  %2859 = add nuw nsw i64 %75, 8
  %scevgep3164 = getelementptr i8, ptr %530, i64 %2859
  %2860 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i265, 6
  %scevgep3165 = getelementptr i8, ptr %scevgep3164, i64 %2860
  br label %loop-memcpy-residual2334, !amdgpu.uniform !42

Flow3792:                                         ; preds = %loop-memcpy-expansion2338
  br label %Flow3793, !amdgpu.uniform !42

Flow3794:                                         ; preds = %Flow3791, %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i286
  br label %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i287, !amdgpu.uniform !42

loop-memcpy-residual-header2341:                  ; preds = %Flow3793
  %2861 = icmp ne i64 %566, 0
  br i1 %2861, label %loop-memcpy-residual2340.preheader, label %Flow3791, !amdgpu.uniform !42

loop-memcpy-residual2340.preheader:               ; preds = %loop-memcpy-residual-header2341
  %umin3159 = call i64 @llvm.umin.i64(i64 %lsr.iv3157, i64 64)
  %2862 = lshr i64 %umin3159, 4
  %2863 = shl nuw nsw i64 %2862, 4
  %scevgep3162 = getelementptr i8, ptr addrspace(4) %lsr.iv3152, i64 %2863
  %scevgep3155 = getelementptr i8, ptr %565, i64 %75
  %scevgep3156 = getelementptr i8, ptr %scevgep3155, i64 %540
  %scevgep3160 = getelementptr i8, ptr %scevgep3156, i64 %2863
  br label %loop-memcpy-residual2340, !amdgpu.uniform !42

Flow3759:                                         ; preds = %loop-memcpy-expansion2344
  br label %Flow3760, !amdgpu.uniform !42

loop-memcpy-residual-header2347:                  ; preds = %Flow3760
  %2864 = icmp eq i64 %766, 0
  br i1 %2864, label %loop-memcpy-residual-header2347.post-loop-memcpy-expansion2343_crit_edge, label %Flow3757, !amdgpu.uniform !42

loop-memcpy-residual-header2347.post-loop-memcpy-expansion2343_crit_edge: ; preds = %loop-memcpy-residual-header2347
  %.pre3248 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i476, 6
  br label %Flow3757, !amdgpu.uniform !42

Flow3757:                                         ; preds = %loop-memcpy-residual-header2347.post-loop-memcpy-expansion2343_crit_edge, %loop-memcpy-residual-header2347
  %2865 = phi i64 [ %.pre3248, %loop-memcpy-residual-header2347.post-loop-memcpy-expansion2343_crit_edge ], [ poison, %loop-memcpy-residual-header2347 ]
  %2866 = phi i1 [ false, %loop-memcpy-residual-header2347.post-loop-memcpy-expansion2343_crit_edge ], [ true, %loop-memcpy-residual-header2347 ]
  br i1 %2866, label %loop-memcpy-residual2346.preheader, label %Flow3758, !amdgpu.uniform !42

loop-memcpy-residual2346.preheader:               ; preds = %Flow3757
  %2867 = add nuw nsw i64 %75, 8
  %scevgep3119 = getelementptr i8, ptr %765, i64 %2867
  %2868 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i476, 6
  %scevgep3120 = getelementptr i8, ptr %scevgep3119, i64 %2868
  br label %loop-memcpy-residual2346, !amdgpu.uniform !42

Flow3750:                                         ; preds = %loop-memcpy-expansion2350
  br label %Flow3751, !amdgpu.uniform !42

Flow3752:                                         ; preds = %Flow3749, %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i497
  br label %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i498, !amdgpu.uniform !42

loop-memcpy-residual-header2353:                  ; preds = %Flow3751
  %2869 = icmp ne i64 %801, 0
  br i1 %2869, label %loop-memcpy-residual2352.preheader, label %Flow3749, !amdgpu.uniform !42

loop-memcpy-residual2352.preheader:               ; preds = %loop-memcpy-residual-header2353
  %umin3114 = call i64 @llvm.umin.i64(i64 %lsr.iv3112, i64 64)
  %2870 = lshr i64 %umin3114, 4
  %2871 = shl nuw nsw i64 %2870, 4
  %scevgep3117 = getelementptr i8, ptr addrspace(4) %lsr.iv3107, i64 %2871
  %scevgep3110 = getelementptr i8, ptr %800, i64 %75
  %scevgep3111 = getelementptr i8, ptr %scevgep3110, i64 %775
  %scevgep3115 = getelementptr i8, ptr %scevgep3111, i64 %2871
  br label %loop-memcpy-residual2352, !amdgpu.uniform !42

Flow3740:                                         ; preds = %loop-memcpy-expansion2356
  br label %Flow3741, !amdgpu.uniform !42

loop-memcpy-residual-header2359:                  ; preds = %Flow3741
  %2872 = icmp eq i64 %877, 0
  br i1 %2872, label %loop-memcpy-residual-header2359.post-loop-memcpy-expansion2355_crit_edge, label %Flow3738, !amdgpu.uniform !42

loop-memcpy-residual-header2359.post-loop-memcpy-expansion2355_crit_edge: ; preds = %loop-memcpy-residual-header2359
  %.pre3250 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i582, 6
  br label %Flow3738, !amdgpu.uniform !42

Flow3738:                                         ; preds = %loop-memcpy-residual-header2359.post-loop-memcpy-expansion2355_crit_edge, %loop-memcpy-residual-header2359
  %2873 = phi i64 [ %.pre3250, %loop-memcpy-residual-header2359.post-loop-memcpy-expansion2355_crit_edge ], [ poison, %loop-memcpy-residual-header2359 ]
  %2874 = phi i1 [ false, %loop-memcpy-residual-header2359.post-loop-memcpy-expansion2355_crit_edge ], [ true, %loop-memcpy-residual-header2359 ]
  br i1 %2874, label %loop-memcpy-residual2358.preheader, label %Flow3739, !amdgpu.uniform !42

loop-memcpy-residual2358.preheader:               ; preds = %Flow3738
  %2875 = add nuw nsw i64 %75, 40
  %scevgep3100 = getelementptr i8, ptr %876, i64 %2875
  %2876 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i582, 6
  %scevgep3101 = getelementptr i8, ptr %scevgep3100, i64 %2876
  br label %loop-memcpy-residual2358, !amdgpu.uniform !42

Flow3731:                                         ; preds = %loop-memcpy-expansion2362
  br label %Flow3732, !amdgpu.uniform !42

Flow3733:                                         ; preds = %Flow3730, %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i603
  br label %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i604, !amdgpu.uniform !42

loop-memcpy-residual-header2365:                  ; preds = %Flow3732
  %2877 = icmp ne i64 %912, 0
  br i1 %2877, label %loop-memcpy-residual2364.preheader, label %Flow3730, !amdgpu.uniform !42

loop-memcpy-residual2364.preheader:               ; preds = %loop-memcpy-residual-header2365
  %umin3095 = call i64 @llvm.umin.i64(i64 %lsr.iv3093, i64 64)
  %2878 = lshr i64 %umin3095, 4
  %2879 = shl nuw nsw i64 %2878, 4
  %scevgep3098 = getelementptr i8, ptr addrspace(4) %lsr.iv3088, i64 %2879
  %scevgep3091 = getelementptr i8, ptr %911, i64 %75
  %scevgep3092 = getelementptr i8, ptr %scevgep3091, i64 %886
  %scevgep3096 = getelementptr i8, ptr %scevgep3092, i64 %2879
  br label %loop-memcpy-residual2364, !amdgpu.uniform !42

Flow3689:                                         ; preds = %loop-memcpy-expansion2368
  br label %Flow3690, !amdgpu.uniform !42

loop-memcpy-residual-header2371:                  ; preds = %Flow3690
  %2880 = icmp eq i64 %1191, 0
  br i1 %2880, label %loop-memcpy-residual-header2371.post-loop-memcpy-expansion2367_crit_edge, label %Flow3687, !amdgpu.uniform !42

loop-memcpy-residual-header2371.post-loop-memcpy-expansion2367_crit_edge: ; preds = %loop-memcpy-residual-header2371
  %.pre3254 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i890, 6
  br label %Flow3687, !amdgpu.uniform !42

Flow3687:                                         ; preds = %loop-memcpy-residual-header2371.post-loop-memcpy-expansion2367_crit_edge, %loop-memcpy-residual-header2371
  %2881 = phi i64 [ %.pre3254, %loop-memcpy-residual-header2371.post-loop-memcpy-expansion2367_crit_edge ], [ poison, %loop-memcpy-residual-header2371 ]
  %2882 = phi i1 [ false, %loop-memcpy-residual-header2371.post-loop-memcpy-expansion2367_crit_edge ], [ true, %loop-memcpy-residual-header2371 ]
  br i1 %2882, label %loop-memcpy-residual2370.preheader, label %Flow3688, !amdgpu.uniform !42

loop-memcpy-residual2370.preheader:               ; preds = %Flow3687
  %2883 = add nuw nsw i64 %75, 8
  %scevgep3055 = getelementptr i8, ptr %1190, i64 %2883
  %2884 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i890, 6
  %scevgep3056 = getelementptr i8, ptr %scevgep3055, i64 %2884
  br label %loop-memcpy-residual2370, !amdgpu.uniform !42

Flow3452:                                         ; preds = %loop-memcpy-expansion2374
  br label %Flow3453, !amdgpu.uniform !42

loop-memcpy-residual-header2377:                  ; preds = %Flow3453
  br i1 %2880, label %loop-memcpy-residual-header2377.post-loop-memcpy-expansion2373_crit_edge, label %Flow3449, !amdgpu.uniform !42

loop-memcpy-residual-header2377.post-loop-memcpy-expansion2373_crit_edge: ; preds = %loop-memcpy-residual-header2377
  %.pre3268 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i1822, 6
  br label %Flow3449, !amdgpu.uniform !42

Flow3449:                                         ; preds = %loop-memcpy-residual-header2377.post-loop-memcpy-expansion2373_crit_edge, %loop-memcpy-residual-header2377
  %2885 = phi i64 [ %.pre3268, %loop-memcpy-residual-header2377.post-loop-memcpy-expansion2373_crit_edge ], [ poison, %loop-memcpy-residual-header2377 ]
  %2886 = phi i1 [ false, %loop-memcpy-residual-header2377.post-loop-memcpy-expansion2373_crit_edge ], [ true, %loop-memcpy-residual-header2377 ]
  br i1 %2886, label %loop-memcpy-residual2376.preheader, label %Flow3450, !amdgpu.uniform !42

loop-memcpy-residual2376.preheader:               ; preds = %Flow3449
  %2887 = add nuw nsw i64 %75, 8
  %scevgep2891 = getelementptr i8, ptr %2314, i64 %2887
  %2888 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i1822, 6
  %scevgep2892 = getelementptr i8, ptr %scevgep2891, i64 %2888
  br label %loop-memcpy-residual2376, !amdgpu.uniform !42

Flow3678:                                         ; preds = %loop-memcpy-expansion2380
  br label %Flow3679, !amdgpu.uniform !42

Flow3680:                                         ; preds = %Flow3676, %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i911
  br label %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i912, !amdgpu.uniform !42

loop-memcpy-residual-header2383:                  ; preds = %Flow3679
  %2889 = icmp ne i64 %1226, 0
  br i1 %2889, label %loop-memcpy-residual2382.preheader, label %Flow3676, !amdgpu.uniform !42

loop-memcpy-residual2382.preheader:               ; preds = %loop-memcpy-residual-header2383
  %umin3050 = call i64 @llvm.umin.i64(i64 %lsr.iv3048, i64 64)
  %2890 = lshr i64 %umin3050, 4
  %2891 = shl nuw nsw i64 %2890, 4
  %scevgep3053 = getelementptr i8, ptr addrspace(4) %lsr.iv3043, i64 %2891
  %scevgep3046 = getelementptr i8, ptr %1225, i64 %75
  %scevgep3047 = getelementptr i8, ptr %scevgep3046, i64 %1200
  %scevgep3051 = getelementptr i8, ptr %scevgep3047, i64 %2891
  br label %loop-memcpy-residual2382, !amdgpu.uniform !42

Flow3599:                                         ; preds = %loop-memcpy-expansion2386
  br label %Flow3600, !amdgpu.uniform !42

loop-memcpy-residual-header2389:                  ; preds = %Flow3600
  %2892 = icmp eq i64 %1646, 0
  br i1 %2892, label %loop-memcpy-residual-header2389.post-loop-memcpy-expansion2385_crit_edge, label %Flow3596, !amdgpu.uniform !42

loop-memcpy-residual-header2389.post-loop-memcpy-expansion2385_crit_edge: ; preds = %loop-memcpy-residual-header2389
  %.pre3258 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i1201, 6
  br label %Flow3596, !amdgpu.uniform !42

Flow3596:                                         ; preds = %loop-memcpy-residual-header2389.post-loop-memcpy-expansion2385_crit_edge, %loop-memcpy-residual-header2389
  %2893 = phi i64 [ %.pre3258, %loop-memcpy-residual-header2389.post-loop-memcpy-expansion2385_crit_edge ], [ poison, %loop-memcpy-residual-header2389 ]
  %2894 = phi i1 [ false, %loop-memcpy-residual-header2389.post-loop-memcpy-expansion2385_crit_edge ], [ true, %loop-memcpy-residual-header2389 ]
  br i1 %2894, label %loop-memcpy-residual2388.preheader, label %Flow3597, !amdgpu.uniform !42

loop-memcpy-residual2388.preheader:               ; preds = %Flow3596
  %2895 = add nuw nsw i64 %75, 8
  %scevgep3007 = getelementptr i8, ptr %1645, i64 %2895
  %2896 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i1201, 6
  %scevgep3008 = getelementptr i8, ptr %scevgep3007, i64 %2896
  br label %loop-memcpy-residual2388, !amdgpu.uniform !42

Flow3586:                                         ; preds = %loop-memcpy-expansion2392
  br label %Flow3587, !amdgpu.uniform !42

Flow3588:                                         ; preds = %Flow3584, %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i1222
  br label %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1223, !amdgpu.uniform !42

loop-memcpy-residual-header2395:                  ; preds = %Flow3587
  %2897 = icmp ne i64 %1681, 0
  br i1 %2897, label %loop-memcpy-residual2394.preheader, label %Flow3584, !amdgpu.uniform !42

loop-memcpy-residual2394.preheader:               ; preds = %loop-memcpy-residual-header2395
  %umin3002 = call i64 @llvm.umin.i64(i64 %lsr.iv3000, i64 64)
  %2898 = lshr i64 %umin3002, 4
  %2899 = shl nuw nsw i64 %2898, 4
  %scevgep3005 = getelementptr i8, ptr addrspace(4) %lsr.iv2995, i64 %2899
  %scevgep2998 = getelementptr i8, ptr %1680, i64 %75
  %scevgep2999 = getelementptr i8, ptr %scevgep2998, i64 %1655
  %scevgep3003 = getelementptr i8, ptr %scevgep2999, i64 %2899
  br label %loop-memcpy-residual2394, !amdgpu.uniform !42

Flow3573:                                         ; preds = %loop-memcpy-expansion2398
  br label %Flow3574, !amdgpu.uniform !42

loop-memcpy-residual-header2401:                  ; preds = %Flow3574
  %2900 = icmp eq i64 %1757, 0
  br i1 %2900, label %loop-memcpy-residual-header2401.post-loop-memcpy-expansion2397_crit_edge, label %Flow3570, !amdgpu.uniform !42

loop-memcpy-residual-header2401.post-loop-memcpy-expansion2397_crit_edge: ; preds = %loop-memcpy-residual-header2401
  %.pre3260 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i1300, 6
  br label %Flow3570, !amdgpu.uniform !42

Flow3570:                                         ; preds = %loop-memcpy-residual-header2401.post-loop-memcpy-expansion2397_crit_edge, %loop-memcpy-residual-header2401
  %2901 = phi i64 [ %.pre3260, %loop-memcpy-residual-header2401.post-loop-memcpy-expansion2397_crit_edge ], [ poison, %loop-memcpy-residual-header2401 ]
  %2902 = phi i1 [ false, %loop-memcpy-residual-header2401.post-loop-memcpy-expansion2397_crit_edge ], [ true, %loop-memcpy-residual-header2401 ]
  br i1 %2902, label %loop-memcpy-residual2400.preheader, label %Flow3571, !amdgpu.uniform !42

loop-memcpy-residual2400.preheader:               ; preds = %Flow3570
  %2903 = lshr i64 %spec.select.i.i.i.i.i.i.i1303, 4
  %2904 = shl nuw nsw i64 %2903, 4
  %2905 = add i64 %75, %2904
  %2906 = add i64 %2905, 8
  %scevgep2987 = getelementptr i8, ptr %1756, i64 %2906
  %2907 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i1300, 6
  %scevgep2988 = getelementptr i8, ptr %scevgep2987, i64 %2907
  %scevgep2990 = getelementptr i8, ptr addrspace(4) @.str.9, i64 %2904
  br label %loop-memcpy-residual2400, !amdgpu.uniform !42

Flow3560:                                         ; preds = %loop-memcpy-expansion2404
  br label %Flow3561, !amdgpu.uniform !42

Flow3562:                                         ; preds = %Flow3558, %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i1321
  br label %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1322, !amdgpu.uniform !42

loop-memcpy-residual-header2407:                  ; preds = %Flow3561
  %2908 = icmp ne i64 %1792, 0
  br i1 %2908, label %loop-memcpy-residual2406.preheader, label %Flow3558, !amdgpu.uniform !42

loop-memcpy-residual2406.preheader:               ; preds = %loop-memcpy-residual-header2407
  %umin2982 = call i64 @llvm.umin.i64(i64 %lsr.iv2980, i64 64)
  %2909 = lshr i64 %umin2982, 4
  %2910 = shl nuw nsw i64 %2909, 4
  %scevgep2985 = getelementptr i8, ptr addrspace(4) %lsr.iv2975, i64 %2910
  %scevgep2978 = getelementptr i8, ptr %1791, i64 %75
  %scevgep2979 = getelementptr i8, ptr %scevgep2978, i64 %1766
  %scevgep2983 = getelementptr i8, ptr %scevgep2979, i64 %2910
  br label %loop-memcpy-residual2406, !amdgpu.uniform !42

Flow3547:                                         ; preds = %loop-memcpy-expansion2410
  br label %Flow3548, !amdgpu.uniform !42

loop-memcpy-residual-header2413:                  ; preds = %Flow3548
  %2911 = icmp eq i64 %1868, 0
  br i1 %2911, label %loop-memcpy-residual-header2413.post-loop-memcpy-expansion2409_crit_edge, label %Flow3544, !amdgpu.uniform !42

loop-memcpy-residual-header2413.post-loop-memcpy-expansion2409_crit_edge: ; preds = %loop-memcpy-residual-header2413
  %.pre3262 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i1405, 6
  br label %Flow3544, !amdgpu.uniform !42

Flow3544:                                         ; preds = %loop-memcpy-residual-header2413.post-loop-memcpy-expansion2409_crit_edge, %loop-memcpy-residual-header2413
  %2912 = phi i64 [ %.pre3262, %loop-memcpy-residual-header2413.post-loop-memcpy-expansion2409_crit_edge ], [ poison, %loop-memcpy-residual-header2413 ]
  %2913 = phi i1 [ false, %loop-memcpy-residual-header2413.post-loop-memcpy-expansion2409_crit_edge ], [ true, %loop-memcpy-residual-header2413 ]
  br i1 %2913, label %loop-memcpy-residual2412.preheader, label %Flow3545, !amdgpu.uniform !42

loop-memcpy-residual2412.preheader:               ; preds = %Flow3544
  %2914 = add nuw nsw i64 %75, 8
  %scevgep2968 = getelementptr i8, ptr %1867, i64 %2914
  %2915 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i1405, 6
  %scevgep2969 = getelementptr i8, ptr %scevgep2968, i64 %2915
  br label %loop-memcpy-residual2412, !amdgpu.uniform !42

Flow3534:                                         ; preds = %loop-memcpy-expansion2416
  br label %Flow3535, !amdgpu.uniform !42

Flow3536:                                         ; preds = %Flow3532, %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i1426
  br label %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1427, !amdgpu.uniform !42

loop-memcpy-residual-header2419:                  ; preds = %Flow3535
  %2916 = icmp ne i64 %1903, 0
  br i1 %2916, label %loop-memcpy-residual2418.preheader, label %Flow3532, !amdgpu.uniform !42

loop-memcpy-residual2418.preheader:               ; preds = %loop-memcpy-residual-header2419
  %umin2963 = call i64 @llvm.umin.i64(i64 %lsr.iv2961, i64 64)
  %2917 = lshr i64 %umin2963, 4
  %2918 = shl nuw nsw i64 %2917, 4
  %scevgep2966 = getelementptr i8, ptr addrspace(4) %lsr.iv2956, i64 %2918
  %scevgep2959 = getelementptr i8, ptr %1902, i64 %75
  %scevgep2960 = getelementptr i8, ptr %scevgep2959, i64 %1877
  %scevgep2964 = getelementptr i8, ptr %scevgep2960, i64 %2918
  br label %loop-memcpy-residual2418, !amdgpu.uniform !42

Flow3442:                                         ; preds = %loop-memcpy-expansion2422
  br label %Flow3443, !amdgpu.uniform !42

Flow3444:                                         ; preds = %Flow3441, %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i1843
  br label %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1844, !amdgpu.uniform !42

loop-memcpy-residual-header2425:                  ; preds = %Flow3443
  %2919 = icmp ne i64 %2347, 0
  br i1 %2919, label %loop-memcpy-residual2424.preheader, label %Flow3441, !amdgpu.uniform !42

loop-memcpy-residual2424.preheader:               ; preds = %loop-memcpy-residual-header2425
  %umin2886 = call i64 @llvm.umin.i64(i64 %lsr.iv2884, i64 64)
  %2920 = lshr i64 %umin2886, 4
  %2921 = shl nuw nsw i64 %2920, 4
  %scevgep2889 = getelementptr i8, ptr addrspace(4) %lsr.iv2879, i64 %2921
  %scevgep2882 = getelementptr i8, ptr %2346, i64 %75
  %scevgep2883 = getelementptr i8, ptr %scevgep2882, i64 %2321
  %scevgep2887 = getelementptr i8, ptr %scevgep2883, i64 %2921
  br label %loop-memcpy-residual2424, !amdgpu.uniform !42

ComputeLoop:                                      ; preds = %for.cond.i.i1.i, %ComputeLoop
  %Accumulator = phi i32 [ %2926, %ComputeLoop ], [ 0, %for.cond.i.i1.i ]
  %OldValuePhi = phi i32 [ %2925, %ComputeLoop ], [ poison, %for.cond.i.i1.i ]
  %ActiveBits = phi i64 [ %2929, %ComputeLoop ], [ %25, %for.cond.i.i1.i ]
  %2922 = call i64 @llvm.cttz.i64(i64 %ActiveBits, i1 true)
  %2923 = trunc i64 %2922 to i32
  %2924 = call i32 @llvm.amdgcn.readlane.i32(i32 %shl.i.i.i.i.i, i32 %2923)
  %2925 = call i32 @llvm.amdgcn.writelane.i32(i32 %Accumulator, i32 %2923, i32 %OldValuePhi)
  %2926 = or i32 %Accumulator, %2924
  %2927 = shl i64 1, %2922
  %2928 = xor i64 %2927, -1
  %2929 = and i64 %ActiveBits, %2928
  %2930 = icmp eq i64 %2929, 0
  br i1 %2930, label %ComputeEnd, label %ComputeLoop, !amdgpu.uniform !42

ComputeEnd:                                       ; preds = %ComputeLoop
  %.lcssa4196 = phi i32 [ %2925, %ComputeLoop ]
  %.lcssa4195 = phi i32 [ %2926, %ComputeLoop ]
  %div8.i.i.i.i.i = lshr i32 %spec.store.select.i.i.i, 5
  %idxprom.i.i.i.i.i = zext nneg i32 %div8.i.i.i.i.i to i64
  %2931 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i.i.i
  %2932 = trunc i64 %25 to i32
  %2933 = lshr i64 %25, 32
  %2934 = trunc i64 %2933 to i32
  %2935 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %2932, i32 0)
  %2936 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %2934, i32 %2935)
  %2937 = icmp eq i32 %2936, 0
  %2938 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %2937)
  %2939 = extractvalue { i1, i64 } %2938, 0
  %2940 = extractvalue { i1, i64 } %2938, 1
  br i1 %2939, label %26, label %28

Flow3883:                                         ; preds = %46, %if.then.i.i.i.i2
  %2941 = phi i1 [ false, %46 ], [ true, %if.then.i.i.i.i2 ]
  br i1 %2941, label %if.end14.i.i.i, label %Flow3884, !amdgpu.uniform !42

ComputeLoop2427:                                  ; preds = %if.then12.i.i.i, %ComputeLoop2427
  %Accumulator2429 = phi i32 [ %2945, %ComputeLoop2427 ], [ -1, %if.then12.i.i.i ]
  %ActiveBits2430 = phi i64 [ %2948, %ComputeLoop2427 ], [ %43, %if.then12.i.i.i ]
  %2942 = call i64 @llvm.cttz.i64(i64 %ActiveBits2430, i1 true)
  %2943 = trunc i64 %2942 to i32
  %2944 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i.i.i, i32 %2943)
  %2945 = and i32 %Accumulator2429, %2944
  %2946 = shl i64 1, %2942
  %2947 = xor i64 %2946, -1
  %2948 = and i64 %ActiveBits2430, %2947
  %2949 = icmp eq i64 %2948, 0
  br i1 %2949, label %ComputeEnd2428, label %ComputeLoop2427, !amdgpu.uniform !42

ComputeEnd2428:                                   ; preds = %ComputeLoop2427
  %.lcssa4197 = phi i32 [ %2945, %ComputeLoop2427 ]
  %2950 = trunc i64 %43 to i32
  %2951 = lshr i64 %43, 32
  %2952 = trunc i64 %2951 to i32
  %2953 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %2950, i32 0)
  %2954 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %2952, i32 %2953)
  %2955 = icmp eq i32 %2954, 0
  %2956 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %2955)
  %2957 = extractvalue { i1, i64 } %2956, 0
  %2958 = extractvalue { i1, i64 } %2956, 1
  br i1 %2957, label %44, label %46

ComputeLoop2431:                                  ; preds = %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit, %ComputeLoop2431
  %Accumulator2433 = phi i32 [ %2962, %ComputeLoop2431 ], [ -1, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit ]
  %ActiveBits2434 = phi i64 [ %2965, %ComputeLoop2431 ], [ %125, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit ]
  %2959 = call i64 @llvm.cttz.i64(i64 %ActiveBits2434, i1 true)
  %2960 = trunc i64 %2959 to i32
  %2961 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i21.i.i, i32 %2960)
  %2962 = and i32 %Accumulator2433, %2961
  %2963 = shl i64 1, %2959
  %2964 = xor i64 %2963, -1
  %2965 = and i64 %ActiveBits2434, %2964
  %2966 = icmp eq i64 %2965, 0
  br i1 %2966, label %ComputeEnd2432, label %ComputeLoop2431, !amdgpu.uniform !42

ComputeEnd2432:                                   ; preds = %ComputeLoop2431
  %.lcssa4191 = phi i32 [ %2962, %ComputeLoop2431 ]
  %2967 = trunc i64 %125 to i32
  %2968 = lshr i64 %125, 32
  %2969 = trunc i64 %2968 to i32
  %2970 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %2967, i32 0)
  %2971 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %2969, i32 %2970)
  %2972 = icmp eq i32 %2971, 0
  %2973 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %2972)
  %2974 = extractvalue { i1, i64 } %2973, 0
  %2975 = extractvalue { i1, i64 } %2973, 1
  br i1 %2974, label %126, label %129

ComputeLoop2435:                                  ; preds = %for.cond.i.i1.i13, %ComputeLoop2435
  %Accumulator2437 = phi i32 [ %2980, %ComputeLoop2435 ], [ 0, %for.cond.i.i1.i13 ]
  %OldValuePhi2438 = phi i32 [ %2979, %ComputeLoop2435 ], [ poison, %for.cond.i.i1.i13 ]
  %ActiveBits2439 = phi i64 [ %2983, %ComputeLoop2435 ], [ %138, %for.cond.i.i1.i13 ]
  %2976 = call i64 @llvm.cttz.i64(i64 %ActiveBits2439, i1 true)
  %2977 = trunc i64 %2976 to i32
  %2978 = call i32 @llvm.amdgcn.readlane.i32(i32 %shl.i.i.i.i.i25, i32 %2977)
  %2979 = call i32 @llvm.amdgcn.writelane.i32(i32 %Accumulator2437, i32 %2977, i32 %OldValuePhi2438)
  %2980 = or i32 %Accumulator2437, %2978
  %2981 = shl i64 1, %2976
  %2982 = xor i64 %2981, -1
  %2983 = and i64 %ActiveBits2439, %2982
  %2984 = icmp eq i64 %2983, 0
  br i1 %2984, label %ComputeEnd2436, label %ComputeLoop2435, !amdgpu.uniform !42

ComputeEnd2436:                                   ; preds = %ComputeLoop2435
  %.lcssa4186 = phi i32 [ %2979, %ComputeLoop2435 ]
  %.lcssa4185 = phi i32 [ %2980, %ComputeLoop2435 ]
  %div8.i.i.i.i.i21 = lshr i32 %spec.store.select.i.i.i20, 5
  %idxprom.i.i.i.i.i22 = zext nneg i32 %div8.i.i.i.i.i21 to i64
  %2985 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i.i.i22
  %2986 = trunc i64 %138 to i32
  %2987 = lshr i64 %138, 32
  %2988 = trunc i64 %2987 to i32
  %2989 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %2986, i32 0)
  %2990 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %2988, i32 %2989)
  %2991 = icmp eq i32 %2990, 0
  %2992 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %2991)
  %2993 = extractvalue { i1, i64 } %2992, 0
  %2994 = extractvalue { i1, i64 } %2992, 1
  br i1 %2993, label %139, label %141

Flow3864:                                         ; preds = %159, %if.then.i.i.i.i31
  %2995 = phi i1 [ false, %159 ], [ true, %if.then.i.i.i.i31 ]
  br i1 %2995, label %if.end14.i.i.i101, label %Flow3865, !amdgpu.uniform !42

ComputeLoop2440:                                  ; preds = %if.then12.i.i.i36, %ComputeLoop2440
  %Accumulator2442 = phi i32 [ %2999, %ComputeLoop2440 ], [ -1, %if.then12.i.i.i36 ]
  %ActiveBits2443 = phi i64 [ %3002, %ComputeLoop2440 ], [ %156, %if.then12.i.i.i36 ]
  %2996 = call i64 @llvm.cttz.i64(i64 %ActiveBits2443, i1 true)
  %2997 = trunc i64 %2996 to i32
  %2998 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i.i.i42, i32 %2997)
  %2999 = and i32 %Accumulator2442, %2998
  %3000 = shl i64 1, %2996
  %3001 = xor i64 %3000, -1
  %3002 = and i64 %ActiveBits2443, %3001
  %3003 = icmp eq i64 %3002, 0
  br i1 %3003, label %ComputeEnd2441, label %ComputeLoop2440, !amdgpu.uniform !42

ComputeEnd2441:                                   ; preds = %ComputeLoop2440
  %.lcssa4187 = phi i32 [ %2999, %ComputeLoop2440 ]
  %3004 = trunc i64 %156 to i32
  %3005 = lshr i64 %156, 32
  %3006 = trunc i64 %3005 to i32
  %3007 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %3004, i32 0)
  %3008 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %3006, i32 %3007)
  %3009 = icmp eq i32 %3008, 0
  %3010 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3009)
  %3011 = extractvalue { i1, i64 } %3010, 0
  %3012 = extractvalue { i1, i64 } %3010, 1
  br i1 %3011, label %157, label %159

ComputeLoop2444:                                  ; preds = %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit109, %ComputeLoop2444
  %Accumulator2446 = phi i32 [ %3016, %ComputeLoop2444 ], [ -1, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit109 ]
  %ActiveBits2447 = phi i64 [ %3019, %ComputeLoop2444 ], [ %236, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit109 ]
  %3013 = call i64 @llvm.cttz.i64(i64 %ActiveBits2447, i1 true)
  %3014 = trunc i64 %3013 to i32
  %3015 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i21.i.i100, i32 %3014)
  %3016 = and i32 %Accumulator2446, %3015
  %3017 = shl i64 1, %3013
  %3018 = xor i64 %3017, -1
  %3019 = and i64 %ActiveBits2447, %3018
  %3020 = icmp eq i64 %3019, 0
  br i1 %3020, label %ComputeEnd2445, label %ComputeLoop2444, !amdgpu.uniform !42

ComputeEnd2445:                                   ; preds = %ComputeLoop2444
  %.lcssa4181 = phi i32 [ %3016, %ComputeLoop2444 ]
  %3021 = trunc i64 %236 to i32
  %3022 = lshr i64 %236, 32
  %3023 = trunc i64 %3022 to i32
  %3024 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %3021, i32 0)
  %3025 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %3023, i32 %3024)
  %3026 = icmp eq i32 %3025, 0
  %3027 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3026)
  %3028 = extractvalue { i1, i64 } %3027, 0
  %3029 = extractvalue { i1, i64 } %3027, 1
  br i1 %3028, label %237, label %240

ComputeLoop2448:                                  ; preds = %for.cond.i.i.i118, %ComputeLoop2448
  %Accumulator2450 = phi i32 [ %3034, %ComputeLoop2448 ], [ 0, %for.cond.i.i.i118 ]
  %OldValuePhi2451 = phi i32 [ %3033, %ComputeLoop2448 ], [ poison, %for.cond.i.i.i118 ]
  %ActiveBits2452 = phi i64 [ %3037, %ComputeLoop2448 ], [ %362, %for.cond.i.i.i118 ]
  %3030 = call i64 @llvm.cttz.i64(i64 %ActiveBits2452, i1 true)
  %3031 = trunc i64 %3030 to i32
  %3032 = call i32 @llvm.amdgcn.readlane.i32(i32 %shl.i.i.i.i.i130, i32 %3031)
  %3033 = call i32 @llvm.amdgcn.writelane.i32(i32 %Accumulator2450, i32 %3031, i32 %OldValuePhi2451)
  %3034 = or i32 %Accumulator2450, %3032
  %3035 = shl i64 1, %3030
  %3036 = xor i64 %3035, -1
  %3037 = and i64 %ActiveBits2452, %3036
  %3038 = icmp eq i64 %3037, 0
  br i1 %3038, label %ComputeEnd2449, label %ComputeLoop2448, !amdgpu.uniform !42

ComputeEnd2449:                                   ; preds = %ComputeLoop2448
  %.lcssa4172 = phi i32 [ %3033, %ComputeLoop2448 ]
  %.lcssa4171 = phi i32 [ %3034, %ComputeLoop2448 ]
  %div8.i.i.i.i.i126 = lshr i32 %spec.store.select.i.i.i125, 5
  %idxprom.i.i.i.i.i127 = zext nneg i32 %div8.i.i.i.i.i126 to i64
  %3039 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i.i.i127
  %3040 = trunc i64 %362 to i32
  %3041 = lshr i64 %362, 32
  %3042 = trunc i64 %3041 to i32
  %3043 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %3040, i32 0)
  %3044 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %3042, i32 %3043)
  %3045 = icmp eq i32 %3044, 0
  %3046 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3045)
  %3047 = extractvalue { i1, i64 } %3046, 0
  %3048 = extractvalue { i1, i64 } %3046, 1
  br i1 %3047, label %363, label %365

Flow3822:                                         ; preds = %383, %if.then.i.i.i.i136
  %3049 = phi i1 [ false, %383 ], [ true, %if.then.i.i.i.i136 ]
  br i1 %3049, label %if.end14.i.i.i206, label %Flow3823, !amdgpu.uniform !42

ComputeLoop2453:                                  ; preds = %if.then12.i.i.i141, %ComputeLoop2453
  %Accumulator2455 = phi i32 [ %3053, %ComputeLoop2453 ], [ -1, %if.then12.i.i.i141 ]
  %ActiveBits2456 = phi i64 [ %3056, %ComputeLoop2453 ], [ %380, %if.then12.i.i.i141 ]
  %3050 = call i64 @llvm.cttz.i64(i64 %ActiveBits2456, i1 true)
  %3051 = trunc i64 %3050 to i32
  %3052 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i.i.i146, i32 %3051)
  %3053 = and i32 %Accumulator2455, %3052
  %3054 = shl i64 1, %3050
  %3055 = xor i64 %3054, -1
  %3056 = and i64 %ActiveBits2456, %3055
  %3057 = icmp eq i64 %3056, 0
  br i1 %3057, label %ComputeEnd2454, label %ComputeLoop2453, !amdgpu.uniform !42

ComputeEnd2454:                                   ; preds = %ComputeLoop2453
  %.lcssa4173 = phi i32 [ %3053, %ComputeLoop2453 ]
  %3058 = trunc i64 %380 to i32
  %3059 = lshr i64 %380, 32
  %3060 = trunc i64 %3059 to i32
  %3061 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %3058, i32 0)
  %3062 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %3060, i32 %3061)
  %3063 = icmp eq i32 %3062, 0
  %3064 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3063)
  %3065 = extractvalue { i1, i64 } %3064, 0
  %3066 = extractvalue { i1, i64 } %3064, 1
  br i1 %3065, label %381, label %383

ComputeLoop2457:                                  ; preds = %_ZN22__llvm_libc_22_0_0_git15write_to_stderrENS_3cpp11string_viewE.exit.i, %ComputeLoop2457
  %Accumulator2459 = phi i32 [ %3070, %ComputeLoop2457 ], [ -1, %_ZN22__llvm_libc_22_0_0_git15write_to_stderrENS_3cpp11string_viewE.exit.i ]
  %ActiveBits2460 = phi i64 [ %3073, %ComputeLoop2457 ], [ %478, %_ZN22__llvm_libc_22_0_0_git15write_to_stderrENS_3cpp11string_viewE.exit.i ]
  %3067 = call i64 @llvm.cttz.i64(i64 %ActiveBits2460, i1 true)
  %3068 = trunc i64 %3067 to i32
  %3069 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i21.i.i203, i32 %3068)
  %3070 = and i32 %Accumulator2459, %3069
  %3071 = shl i64 1, %3067
  %3072 = xor i64 %3071, -1
  %3073 = and i64 %ActiveBits2460, %3072
  %3074 = icmp eq i64 %3073, 0
  br i1 %3074, label %ComputeEnd2458, label %ComputeLoop2457, !amdgpu.uniform !42

ComputeEnd2458:                                   ; preds = %ComputeLoop2457
  %.lcssa4163 = phi i32 [ %3070, %ComputeLoop2457 ]
  %3075 = trunc i64 %478 to i32
  %3076 = lshr i64 %478, 32
  %3077 = trunc i64 %3076 to i32
  %3078 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %3075, i32 0)
  %3079 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %3077, i32 %3078)
  %3080 = icmp eq i32 %3079, 0
  %3081 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3080)
  %3082 = extractvalue { i1, i64 } %3081, 0
  %3083 = extractvalue { i1, i64 } %3081, 1
  br i1 %3082, label %479, label %482

ComputeLoop2461:                                  ; preds = %for.cond.i.i1.i222, %ComputeLoop2461
  %Accumulator2463 = phi i32 [ %3088, %ComputeLoop2461 ], [ 0, %for.cond.i.i1.i222 ]
  %OldValuePhi2464 = phi i32 [ %3087, %ComputeLoop2461 ], [ poison, %for.cond.i.i1.i222 ]
  %ActiveBits2465 = phi i64 [ %3091, %ComputeLoop2461 ], [ %491, %for.cond.i.i1.i222 ]
  %3084 = call i64 @llvm.cttz.i64(i64 %ActiveBits2465, i1 true)
  %3085 = trunc i64 %3084 to i32
  %3086 = call i32 @llvm.amdgcn.readlane.i32(i32 %shl.i.i.i.i.i234, i32 %3085)
  %3087 = call i32 @llvm.amdgcn.writelane.i32(i32 %Accumulator2463, i32 %3085, i32 %OldValuePhi2464)
  %3088 = or i32 %Accumulator2463, %3086
  %3089 = shl i64 1, %3084
  %3090 = xor i64 %3089, -1
  %3091 = and i64 %ActiveBits2465, %3090
  %3092 = icmp eq i64 %3091, 0
  br i1 %3092, label %ComputeEnd2462, label %ComputeLoop2461, !amdgpu.uniform !42

ComputeEnd2462:                                   ; preds = %ComputeLoop2461
  %.lcssa4158 = phi i32 [ %3087, %ComputeLoop2461 ]
  %.lcssa4157 = phi i32 [ %3088, %ComputeLoop2461 ]
  %div8.i.i.i.i.i230 = lshr i32 %spec.store.select.i.i.i229, 5
  %idxprom.i.i.i.i.i231 = zext nneg i32 %div8.i.i.i.i.i230 to i64
  %3093 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i.i.i231
  %3094 = trunc i64 %491 to i32
  %3095 = lshr i64 %491, 32
  %3096 = trunc i64 %3095 to i32
  %3097 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %3094, i32 0)
  %3098 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %3096, i32 %3097)
  %3099 = icmp eq i32 %3098, 0
  %3100 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3099)
  %3101 = extractvalue { i1, i64 } %3100, 0
  %3102 = extractvalue { i1, i64 } %3100, 1
  br i1 %3101, label %492, label %494

Flow3803:                                         ; preds = %512, %if.then.i.i.i.i240
  %3103 = phi i1 [ false, %512 ], [ true, %if.then.i.i.i.i240 ]
  br i1 %3103, label %if.end14.i.i.i311, label %Flow3804, !amdgpu.uniform !42

ComputeLoop2466:                                  ; preds = %if.then12.i.i.i245, %ComputeLoop2466
  %Accumulator2468 = phi i32 [ %3107, %ComputeLoop2466 ], [ -1, %if.then12.i.i.i245 ]
  %ActiveBits2469 = phi i64 [ %3110, %ComputeLoop2466 ], [ %509, %if.then12.i.i.i245 ]
  %3104 = call i64 @llvm.cttz.i64(i64 %ActiveBits2469, i1 true)
  %3105 = trunc i64 %3104 to i32
  %3106 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i.i.i251, i32 %3105)
  %3107 = and i32 %Accumulator2468, %3106
  %3108 = shl i64 1, %3104
  %3109 = xor i64 %3108, -1
  %3110 = and i64 %ActiveBits2469, %3109
  %3111 = icmp eq i64 %3110, 0
  br i1 %3111, label %ComputeEnd2467, label %ComputeLoop2466, !amdgpu.uniform !42

ComputeEnd2467:                                   ; preds = %ComputeLoop2466
  %.lcssa4159 = phi i32 [ %3107, %ComputeLoop2466 ]
  %3112 = trunc i64 %509 to i32
  %3113 = lshr i64 %509, 32
  %3114 = trunc i64 %3113 to i32
  %3115 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %3112, i32 0)
  %3116 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %3114, i32 %3115)
  %3117 = icmp eq i32 %3116, 0
  %3118 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3117)
  %3119 = extractvalue { i1, i64 } %3118, 0
  %3120 = extractvalue { i1, i64 } %3118, 1
  br i1 %3119, label %510, label %512

ComputeLoop2470:                                  ; preds = %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit319, %ComputeLoop2470
  %Accumulator2472 = phi i32 [ %3124, %ComputeLoop2470 ], [ -1, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit319 ]
  %ActiveBits2473 = phi i64 [ %3127, %ComputeLoop2470 ], [ %589, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit319 ]
  %3121 = call i64 @llvm.cttz.i64(i64 %ActiveBits2473, i1 true)
  %3122 = trunc i64 %3121 to i32
  %3123 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i21.i.i310, i32 %3122)
  %3124 = and i32 %Accumulator2472, %3123
  %3125 = shl i64 1, %3121
  %3126 = xor i64 %3125, -1
  %3127 = and i64 %ActiveBits2473, %3126
  %3128 = icmp eq i64 %3127, 0
  br i1 %3128, label %ComputeEnd2471, label %ComputeLoop2470, !amdgpu.uniform !42

ComputeEnd2471:                                   ; preds = %ComputeLoop2470
  %.lcssa4153 = phi i32 [ %3124, %ComputeLoop2470 ]
  %3129 = trunc i64 %589 to i32
  %3130 = lshr i64 %589, 32
  %3131 = trunc i64 %3130 to i32
  %3132 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %3129, i32 0)
  %3133 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %3131, i32 %3132)
  %3134 = icmp eq i32 %3133, 0
  %3135 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3134)
  %3136 = extractvalue { i1, i64 } %3135, 0
  %3137 = extractvalue { i1, i64 } %3135, 1
  br i1 %3136, label %590, label %593

ComputeLoop2474:                                  ; preds = %for.cond.i.i.i328, %ComputeLoop2474
  %Accumulator2476 = phi i32 [ %3142, %ComputeLoop2474 ], [ 0, %for.cond.i.i.i328 ]
  %OldValuePhi2477 = phi i32 [ %3141, %ComputeLoop2474 ], [ poison, %for.cond.i.i.i328 ]
  %ActiveBits2478 = phi i64 [ %3145, %ComputeLoop2474 ], [ %614, %for.cond.i.i.i328 ]
  %3138 = call i64 @llvm.cttz.i64(i64 %ActiveBits2478, i1 true)
  %3139 = trunc i64 %3138 to i32
  %3140 = call i32 @llvm.amdgcn.readlane.i32(i32 %shl.i.i.i.i.i340, i32 %3139)
  %3141 = call i32 @llvm.amdgcn.writelane.i32(i32 %Accumulator2476, i32 %3139, i32 %OldValuePhi2477)
  %3142 = or i32 %Accumulator2476, %3140
  %3143 = shl i64 1, %3138
  %3144 = xor i64 %3143, -1
  %3145 = and i64 %ActiveBits2478, %3144
  %3146 = icmp eq i64 %3145, 0
  br i1 %3146, label %ComputeEnd2475, label %ComputeLoop2474, !amdgpu.uniform !42

ComputeEnd2475:                                   ; preds = %ComputeLoop2474
  %.lcssa4148 = phi i32 [ %3141, %ComputeLoop2474 ]
  %.lcssa4147 = phi i32 [ %3142, %ComputeLoop2474 ]
  %div8.i.i.i.i.i336 = lshr i32 %spec.store.select.i.i.i335, 5
  %idxprom.i.i.i.i.i337 = zext nneg i32 %div8.i.i.i.i.i336 to i64
  %3147 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i.i.i337
  %3148 = trunc i64 %614 to i32
  %3149 = lshr i64 %614, 32
  %3150 = trunc i64 %3149 to i32
  %3151 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %3148, i32 0)
  %3152 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %3150, i32 %3151)
  %3153 = icmp eq i32 %3152, 0
  %3154 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3153)
  %3155 = extractvalue { i1, i64 } %3154, 0
  %3156 = extractvalue { i1, i64 } %3154, 1
  br i1 %3155, label %615, label %617

Flow3780:                                         ; preds = %635, %if.then.i.i.i.i346
  %3157 = phi i1 [ false, %635 ], [ true, %if.then.i.i.i.i346 ]
  br i1 %3157, label %if.end14.i.i.i417, label %Flow3781, !amdgpu.uniform !42

ComputeLoop2479:                                  ; preds = %if.then12.i.i.i351, %ComputeLoop2479
  %Accumulator2481 = phi i32 [ %3161, %ComputeLoop2479 ], [ -1, %if.then12.i.i.i351 ]
  %ActiveBits2482 = phi i64 [ %3164, %ComputeLoop2479 ], [ %632, %if.then12.i.i.i351 ]
  %3158 = call i64 @llvm.cttz.i64(i64 %ActiveBits2482, i1 true)
  %3159 = trunc i64 %3158 to i32
  %3160 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i.i.i357, i32 %3159)
  %3161 = and i32 %Accumulator2481, %3160
  %3162 = shl i64 1, %3158
  %3163 = xor i64 %3162, -1
  %3164 = and i64 %ActiveBits2482, %3163
  %3165 = icmp eq i64 %3164, 0
  br i1 %3165, label %ComputeEnd2480, label %ComputeLoop2479, !amdgpu.uniform !42

ComputeEnd2480:                                   ; preds = %ComputeLoop2479
  %.lcssa4149 = phi i32 [ %3161, %ComputeLoop2479 ]
  %3166 = trunc i64 %632 to i32
  %3167 = lshr i64 %632, 32
  %3168 = trunc i64 %3167 to i32
  %3169 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %3166, i32 0)
  %3170 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %3168, i32 %3169)
  %3171 = icmp eq i32 %3170, 0
  %3172 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3171)
  %3173 = extractvalue { i1, i64 } %3172, 0
  %3174 = extractvalue { i1, i64 } %3172, 1
  br i1 %3173, label %633, label %635

ComputeLoop2483:                                  ; preds = %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsINS_3cpp6stringEEERS1_T_.exit, %ComputeLoop2483
  %Accumulator2485 = phi i32 [ %3178, %ComputeLoop2483 ], [ -1, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsINS_3cpp6stringEEERS1_T_.exit ]
  %ActiveBits2486 = phi i64 [ %3181, %ComputeLoop2483 ], [ %713, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsINS_3cpp6stringEEERS1_T_.exit ]
  %3175 = call i64 @llvm.cttz.i64(i64 %ActiveBits2486, i1 true)
  %3176 = trunc i64 %3175 to i32
  %3177 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i21.i.i416, i32 %3176)
  %3178 = and i32 %Accumulator2485, %3177
  %3179 = shl i64 1, %3175
  %3180 = xor i64 %3179, -1
  %3181 = and i64 %ActiveBits2486, %3180
  %3182 = icmp eq i64 %3181, 0
  br i1 %3182, label %ComputeEnd2484, label %ComputeLoop2483, !amdgpu.uniform !42

ComputeEnd2484:                                   ; preds = %ComputeLoop2483
  %.lcssa4143 = phi i32 [ %3178, %ComputeLoop2483 ]
  %3183 = trunc i64 %713 to i32
  %3184 = lshr i64 %713, 32
  %3185 = trunc i64 %3184 to i32
  %3186 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %3183, i32 0)
  %3187 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %3185, i32 %3186)
  %3188 = icmp eq i32 %3187, 0
  %3189 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3188)
  %3190 = extractvalue { i1, i64 } %3189, 0
  %3191 = extractvalue { i1, i64 } %3189, 1
  br i1 %3190, label %714, label %717

ComputeLoop2487:                                  ; preds = %for.cond.i.i1.i433, %ComputeLoop2487
  %Accumulator2489 = phi i32 [ %3196, %ComputeLoop2487 ], [ 0, %for.cond.i.i1.i433 ]
  %OldValuePhi2490 = phi i32 [ %3195, %ComputeLoop2487 ], [ poison, %for.cond.i.i1.i433 ]
  %ActiveBits2491 = phi i64 [ %3199, %ComputeLoop2487 ], [ %726, %for.cond.i.i1.i433 ]
  %3192 = call i64 @llvm.cttz.i64(i64 %ActiveBits2491, i1 true)
  %3193 = trunc i64 %3192 to i32
  %3194 = call i32 @llvm.amdgcn.readlane.i32(i32 %shl.i.i.i.i.i445, i32 %3193)
  %3195 = call i32 @llvm.amdgcn.writelane.i32(i32 %Accumulator2489, i32 %3193, i32 %OldValuePhi2490)
  %3196 = or i32 %Accumulator2489, %3194
  %3197 = shl i64 1, %3192
  %3198 = xor i64 %3197, -1
  %3199 = and i64 %ActiveBits2491, %3198
  %3200 = icmp eq i64 %3199, 0
  br i1 %3200, label %ComputeEnd2488, label %ComputeLoop2487, !amdgpu.uniform !42

ComputeEnd2488:                                   ; preds = %ComputeLoop2487
  %.lcssa4138 = phi i32 [ %3195, %ComputeLoop2487 ]
  %.lcssa4137 = phi i32 [ %3196, %ComputeLoop2487 ]
  %div8.i.i.i.i.i441 = lshr i32 %spec.store.select.i.i.i440, 5
  %idxprom.i.i.i.i.i442 = zext nneg i32 %div8.i.i.i.i.i441 to i64
  %3201 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i.i.i442
  %3202 = trunc i64 %726 to i32
  %3203 = lshr i64 %726, 32
  %3204 = trunc i64 %3203 to i32
  %3205 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %3202, i32 0)
  %3206 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %3204, i32 %3205)
  %3207 = icmp eq i32 %3206, 0
  %3208 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3207)
  %3209 = extractvalue { i1, i64 } %3208, 0
  %3210 = extractvalue { i1, i64 } %3208, 1
  br i1 %3209, label %727, label %729

Flow3761:                                         ; preds = %747, %if.then.i.i.i.i451
  %3211 = phi i1 [ false, %747 ], [ true, %if.then.i.i.i.i451 ]
  br i1 %3211, label %if.end14.i.i.i522, label %Flow3762, !amdgpu.uniform !42

ComputeLoop2492:                                  ; preds = %if.then12.i.i.i456, %ComputeLoop2492
  %Accumulator2494 = phi i32 [ %3215, %ComputeLoop2492 ], [ -1, %if.then12.i.i.i456 ]
  %ActiveBits2495 = phi i64 [ %3218, %ComputeLoop2492 ], [ %744, %if.then12.i.i.i456 ]
  %3212 = call i64 @llvm.cttz.i64(i64 %ActiveBits2495, i1 true)
  %3213 = trunc i64 %3212 to i32
  %3214 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i.i.i462, i32 %3213)
  %3215 = and i32 %Accumulator2494, %3214
  %3216 = shl i64 1, %3212
  %3217 = xor i64 %3216, -1
  %3218 = and i64 %ActiveBits2495, %3217
  %3219 = icmp eq i64 %3218, 0
  br i1 %3219, label %ComputeEnd2493, label %ComputeLoop2492, !amdgpu.uniform !42

ComputeEnd2493:                                   ; preds = %ComputeLoop2492
  %.lcssa4139 = phi i32 [ %3215, %ComputeLoop2492 ]
  %3220 = trunc i64 %744 to i32
  %3221 = lshr i64 %744, 32
  %3222 = trunc i64 %3221 to i32
  %3223 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %3220, i32 0)
  %3224 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %3222, i32 %3223)
  %3225 = icmp eq i32 %3224, 0
  %3226 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3225)
  %3227 = extractvalue { i1, i64 } %3226, 0
  %3228 = extractvalue { i1, i64 } %3226, 1
  br i1 %3227, label %745, label %747

ComputeLoop2496:                                  ; preds = %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit530, %ComputeLoop2496
  %Accumulator2498 = phi i32 [ %3232, %ComputeLoop2496 ], [ -1, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit530 ]
  %ActiveBits2499 = phi i64 [ %3235, %ComputeLoop2496 ], [ %824, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit530 ]
  %3229 = call i64 @llvm.cttz.i64(i64 %ActiveBits2499, i1 true)
  %3230 = trunc i64 %3229 to i32
  %3231 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i21.i.i521, i32 %3230)
  %3232 = and i32 %Accumulator2498, %3231
  %3233 = shl i64 1, %3229
  %3234 = xor i64 %3233, -1
  %3235 = and i64 %ActiveBits2499, %3234
  %3236 = icmp eq i64 %3235, 0
  br i1 %3236, label %ComputeEnd2497, label %ComputeLoop2496, !amdgpu.uniform !42

ComputeEnd2497:                                   ; preds = %ComputeLoop2496
  %.lcssa4133 = phi i32 [ %3232, %ComputeLoop2496 ]
  %3237 = trunc i64 %824 to i32
  %3238 = lshr i64 %824, 32
  %3239 = trunc i64 %3238 to i32
  %3240 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %3237, i32 0)
  %3241 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %3239, i32 %3240)
  %3242 = icmp eq i32 %3241, 0
  %3243 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3242)
  %3244 = extractvalue { i1, i64 } %3243, 0
  %3245 = extractvalue { i1, i64 } %3243, 1
  br i1 %3244, label %825, label %828

ComputeLoop2500:                                  ; preds = %for.cond.i.i1.i539, %ComputeLoop2500
  %Accumulator2502 = phi i32 [ %3250, %ComputeLoop2500 ], [ 0, %for.cond.i.i1.i539 ]
  %OldValuePhi2503 = phi i32 [ %3249, %ComputeLoop2500 ], [ poison, %for.cond.i.i1.i539 ]
  %ActiveBits2504 = phi i64 [ %3253, %ComputeLoop2500 ], [ %837, %for.cond.i.i1.i539 ]
  %3246 = call i64 @llvm.cttz.i64(i64 %ActiveBits2504, i1 true)
  %3247 = trunc i64 %3246 to i32
  %3248 = call i32 @llvm.amdgcn.readlane.i32(i32 %shl.i.i.i.i.i551, i32 %3247)
  %3249 = call i32 @llvm.amdgcn.writelane.i32(i32 %Accumulator2502, i32 %3247, i32 %OldValuePhi2503)
  %3250 = or i32 %Accumulator2502, %3248
  %3251 = shl i64 1, %3246
  %3252 = xor i64 %3251, -1
  %3253 = and i64 %ActiveBits2504, %3252
  %3254 = icmp eq i64 %3253, 0
  br i1 %3254, label %ComputeEnd2501, label %ComputeLoop2500, !amdgpu.uniform !42

ComputeEnd2501:                                   ; preds = %ComputeLoop2500
  %.lcssa4128 = phi i32 [ %3249, %ComputeLoop2500 ]
  %.lcssa4127 = phi i32 [ %3250, %ComputeLoop2500 ]
  %div8.i.i.i.i.i547 = lshr i32 %spec.store.select.i.i.i546, 5
  %idxprom.i.i.i.i.i548 = zext nneg i32 %div8.i.i.i.i.i547 to i64
  %3255 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i.i.i548
  %3256 = trunc i64 %837 to i32
  %3257 = lshr i64 %837, 32
  %3258 = trunc i64 %3257 to i32
  %3259 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %3256, i32 0)
  %3260 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %3258, i32 %3259)
  %3261 = icmp eq i32 %3260, 0
  %3262 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3261)
  %3263 = extractvalue { i1, i64 } %3262, 0
  %3264 = extractvalue { i1, i64 } %3262, 1
  br i1 %3263, label %838, label %840

Flow3742:                                         ; preds = %858, %if.then.i.i.i.i557
  %3265 = phi i1 [ false, %858 ], [ true, %if.then.i.i.i.i557 ]
  br i1 %3265, label %if.end14.i.i.i628, label %Flow3743, !amdgpu.uniform !42

ComputeLoop2505:                                  ; preds = %if.then12.i.i.i562, %ComputeLoop2505
  %Accumulator2507 = phi i32 [ %3269, %ComputeLoop2505 ], [ -1, %if.then12.i.i.i562 ]
  %ActiveBits2508 = phi i64 [ %3272, %ComputeLoop2505 ], [ %855, %if.then12.i.i.i562 ]
  %3266 = call i64 @llvm.cttz.i64(i64 %ActiveBits2508, i1 true)
  %3267 = trunc i64 %3266 to i32
  %3268 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i.i.i568, i32 %3267)
  %3269 = and i32 %Accumulator2507, %3268
  %3270 = shl i64 1, %3266
  %3271 = xor i64 %3270, -1
  %3272 = and i64 %ActiveBits2508, %3271
  %3273 = icmp eq i64 %3272, 0
  br i1 %3273, label %ComputeEnd2506, label %ComputeLoop2505, !amdgpu.uniform !42

ComputeEnd2506:                                   ; preds = %ComputeLoop2505
  %.lcssa4129 = phi i32 [ %3269, %ComputeLoop2505 ]
  %3274 = trunc i64 %855 to i32
  %3275 = lshr i64 %855, 32
  %3276 = trunc i64 %3275 to i32
  %3277 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %3274, i32 0)
  %3278 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %3276, i32 %3277)
  %3279 = icmp eq i32 %3278, 0
  %3280 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3279)
  %3281 = extractvalue { i1, i64 } %3280, 0
  %3282 = extractvalue { i1, i64 } %3280, 1
  br i1 %3281, label %856, label %858

ComputeLoop2509:                                  ; preds = %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit636, %ComputeLoop2509
  %Accumulator2511 = phi i32 [ %3286, %ComputeLoop2509 ], [ -1, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit636 ]
  %ActiveBits2512 = phi i64 [ %3289, %ComputeLoop2509 ], [ %935, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit636 ]
  %3283 = call i64 @llvm.cttz.i64(i64 %ActiveBits2512, i1 true)
  %3284 = trunc i64 %3283 to i32
  %3285 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i21.i.i627, i32 %3284)
  %3286 = and i32 %Accumulator2511, %3285
  %3287 = shl i64 1, %3283
  %3288 = xor i64 %3287, -1
  %3289 = and i64 %ActiveBits2512, %3288
  %3290 = icmp eq i64 %3289, 0
  br i1 %3290, label %ComputeEnd2510, label %ComputeLoop2509, !amdgpu.uniform !42

ComputeEnd2510:                                   ; preds = %ComputeLoop2509
  %.lcssa4123 = phi i32 [ %3286, %ComputeLoop2509 ]
  %3291 = trunc i64 %935 to i32
  %3292 = lshr i64 %935, 32
  %3293 = trunc i64 %3292 to i32
  %3294 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %3291, i32 0)
  %3295 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %3293, i32 %3294)
  %3296 = icmp eq i32 %3295, 0
  %3297 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3296)
  %3298 = extractvalue { i1, i64 } %3297, 0
  %3299 = extractvalue { i1, i64 } %3297, 1
  br i1 %3298, label %936, label %939

ComputeLoop2513:                                  ; preds = %for.cond.i.i.i638, %ComputeLoop2513
  %Accumulator2515 = phi i32 [ %3304, %ComputeLoop2513 ], [ 0, %for.cond.i.i.i638 ]
  %OldValuePhi2516 = phi i32 [ %3303, %ComputeLoop2513 ], [ poison, %for.cond.i.i.i638 ]
  %ActiveBits2517 = phi i64 [ %3307, %ComputeLoop2513 ], [ %948, %for.cond.i.i.i638 ]
  %3300 = call i64 @llvm.cttz.i64(i64 %ActiveBits2517, i1 true)
  %3301 = trunc i64 %3300 to i32
  %3302 = call i32 @llvm.amdgcn.readlane.i32(i32 %shl.i.i.i.i.i650, i32 %3301)
  %3303 = call i32 @llvm.amdgcn.writelane.i32(i32 %Accumulator2515, i32 %3301, i32 %OldValuePhi2516)
  %3304 = or i32 %Accumulator2515, %3302
  %3305 = shl i64 1, %3300
  %3306 = xor i64 %3305, -1
  %3307 = and i64 %ActiveBits2517, %3306
  %3308 = icmp eq i64 %3307, 0
  br i1 %3308, label %ComputeEnd2514, label %ComputeLoop2513, !amdgpu.uniform !42

ComputeEnd2514:                                   ; preds = %ComputeLoop2513
  %.lcssa4118 = phi i32 [ %3303, %ComputeLoop2513 ]
  %.lcssa4117 = phi i32 [ %3304, %ComputeLoop2513 ]
  %div8.i.i.i.i.i646 = lshr i32 %spec.store.select.i.i.i645, 5
  %idxprom.i.i.i.i.i647 = zext nneg i32 %div8.i.i.i.i.i646 to i64
  %3309 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i.i.i647
  %3310 = trunc i64 %948 to i32
  %3311 = lshr i64 %948, 32
  %3312 = trunc i64 %3311 to i32
  %3313 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %3310, i32 0)
  %3314 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %3312, i32 %3313)
  %3315 = icmp eq i32 %3314, 0
  %3316 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3315)
  %3317 = extractvalue { i1, i64 } %3316, 0
  %3318 = extractvalue { i1, i64 } %3316, 1
  br i1 %3317, label %949, label %951

Flow3723:                                         ; preds = %969, %if.then.i.i.i.i656
  %3319 = phi i1 [ false, %969 ], [ true, %if.then.i.i.i.i656 ]
  br i1 %3319, label %if.end14.i.i.i725, label %Flow3724, !amdgpu.uniform !42

ComputeLoop2518:                                  ; preds = %if.then12.i.i.i661, %ComputeLoop2518
  %Accumulator2520 = phi i32 [ %3323, %ComputeLoop2518 ], [ -1, %if.then12.i.i.i661 ]
  %ActiveBits2521 = phi i64 [ %3326, %ComputeLoop2518 ], [ %966, %if.then12.i.i.i661 ]
  %3320 = call i64 @llvm.cttz.i64(i64 %ActiveBits2521, i1 true)
  %3321 = trunc i64 %3320 to i32
  %3322 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i.i.i667, i32 %3321)
  %3323 = and i32 %Accumulator2520, %3322
  %3324 = shl i64 1, %3320
  %3325 = xor i64 %3324, -1
  %3326 = and i64 %ActiveBits2521, %3325
  %3327 = icmp eq i64 %3326, 0
  br i1 %3327, label %ComputeEnd2519, label %ComputeLoop2518, !amdgpu.uniform !42

ComputeEnd2519:                                   ; preds = %ComputeLoop2518
  %.lcssa4119 = phi i32 [ %3323, %ComputeLoop2518 ]
  %3328 = trunc i64 %966 to i32
  %3329 = lshr i64 %966, 32
  %3330 = trunc i64 %3329 to i32
  %3331 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %3328, i32 0)
  %3332 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %3330, i32 %3331)
  %3333 = icmp eq i32 %3332, 0
  %3334 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3333)
  %3335 = extractvalue { i1, i64 } %3334, 0
  %3336 = extractvalue { i1, i64 } %3334, 1
  br i1 %3335, label %967, label %969

ComputeLoop2522:                                  ; preds = %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIcEERS1_T_.exit, %ComputeLoop2522
  %Accumulator2524 = phi i32 [ %3340, %ComputeLoop2522 ], [ -1, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIcEERS1_T_.exit ]
  %ActiveBits2525 = phi i64 [ %3343, %ComputeLoop2522 ], [ %1021, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIcEERS1_T_.exit ]
  %3337 = call i64 @llvm.cttz.i64(i64 %ActiveBits2525, i1 true)
  %3338 = trunc i64 %3337 to i32
  %3339 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i21.i.i724, i32 %3338)
  %3340 = and i32 %Accumulator2524, %3339
  %3341 = shl i64 1, %3337
  %3342 = xor i64 %3341, -1
  %3343 = and i64 %ActiveBits2525, %3342
  %3344 = icmp eq i64 %3343, 0
  br i1 %3344, label %ComputeEnd2523, label %ComputeLoop2522, !amdgpu.uniform !42

ComputeEnd2523:                                   ; preds = %ComputeLoop2522
  %.lcssa4113 = phi i32 [ %3340, %ComputeLoop2522 ]
  %3345 = trunc i64 %1021 to i32
  %3346 = lshr i64 %1021, 32
  %3347 = trunc i64 %3346 to i32
  %3348 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %3345, i32 0)
  %3349 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %3347, i32 %3348)
  %3350 = icmp eq i32 %3349, 0
  %3351 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3350)
  %3352 = extractvalue { i1, i64 } %3351, 0
  %3353 = extractvalue { i1, i64 } %3351, 1
  br i1 %3352, label %1022, label %1025

ComputeLoop2526:                                  ; preds = %for.cond.i.i.i741, %ComputeLoop2526
  %Accumulator2528 = phi i32 [ %3358, %ComputeLoop2526 ], [ 0, %for.cond.i.i.i741 ]
  %OldValuePhi2529 = phi i32 [ %3357, %ComputeLoop2526 ], [ poison, %for.cond.i.i.i741 ]
  %ActiveBits2530 = phi i64 [ %3361, %ComputeLoop2526 ], [ %1043, %for.cond.i.i.i741 ]
  %3354 = call i64 @llvm.cttz.i64(i64 %ActiveBits2530, i1 true)
  %3355 = trunc i64 %3354 to i32
  %3356 = call i32 @llvm.amdgcn.readlane.i32(i32 %shl.i.i.i.i.i753, i32 %3355)
  %3357 = call i32 @llvm.amdgcn.writelane.i32(i32 %Accumulator2528, i32 %3355, i32 %OldValuePhi2529)
  %3358 = or i32 %Accumulator2528, %3356
  %3359 = shl i64 1, %3354
  %3360 = xor i64 %3359, -1
  %3361 = and i64 %ActiveBits2530, %3360
  %3362 = icmp eq i64 %3361, 0
  br i1 %3362, label %ComputeEnd2527, label %ComputeLoop2526, !amdgpu.uniform !42

ComputeEnd2527:                                   ; preds = %ComputeLoop2526
  %.lcssa4108 = phi i32 [ %3357, %ComputeLoop2526 ]
  %.lcssa4107 = phi i32 [ %3358, %ComputeLoop2526 ]
  %div8.i.i.i.i.i749 = lshr i32 %spec.store.select.i.i.i748, 5
  %idxprom.i.i.i.i.i750 = zext nneg i32 %div8.i.i.i.i.i749 to i64
  %3363 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i.i.i750
  %3364 = trunc i64 %1043 to i32
  %3365 = lshr i64 %1043, 32
  %3366 = trunc i64 %3365 to i32
  %3367 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %3364, i32 0)
  %3368 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %3366, i32 %3367)
  %3369 = icmp eq i32 %3368, 0
  %3370 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3369)
  %3371 = extractvalue { i1, i64 } %3370, 0
  %3372 = extractvalue { i1, i64 } %3370, 1
  br i1 %3371, label %1044, label %1046

Flow3710:                                         ; preds = %1064, %if.then.i.i.i.i759
  %3373 = phi i1 [ false, %1064 ], [ true, %if.then.i.i.i.i759 ]
  br i1 %3373, label %if.end14.i.i.i830, label %Flow3711, !amdgpu.uniform !42

ComputeLoop2531:                                  ; preds = %if.then12.i.i.i764, %ComputeLoop2531
  %Accumulator2533 = phi i32 [ %3377, %ComputeLoop2531 ], [ -1, %if.then12.i.i.i764 ]
  %ActiveBits2534 = phi i64 [ %3380, %ComputeLoop2531 ], [ %1061, %if.then12.i.i.i764 ]
  %3374 = call i64 @llvm.cttz.i64(i64 %ActiveBits2534, i1 true)
  %3375 = trunc i64 %3374 to i32
  %3376 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i.i.i770, i32 %3375)
  %3377 = and i32 %Accumulator2533, %3376
  %3378 = shl i64 1, %3374
  %3379 = xor i64 %3378, -1
  %3380 = and i64 %ActiveBits2534, %3379
  %3381 = icmp eq i64 %3380, 0
  br i1 %3381, label %ComputeEnd2532, label %ComputeLoop2531, !amdgpu.uniform !42

ComputeEnd2532:                                   ; preds = %ComputeLoop2531
  %.lcssa4109 = phi i32 [ %3377, %ComputeLoop2531 ]
  %3382 = trunc i64 %1061 to i32
  %3383 = lshr i64 %1061, 32
  %3384 = trunc i64 %3383 to i32
  %3385 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %3382, i32 0)
  %3386 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %3384, i32 %3385)
  %3387 = icmp eq i32 %3386, 0
  %3388 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3387)
  %3389 = extractvalue { i1, i64 } %3388, 0
  %3390 = extractvalue { i1, i64 } %3388, 1
  br i1 %3389, label %1062, label %1064

ComputeLoop2535:                                  ; preds = %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsINS_3cpp6stringEEERS1_T_.exit838, %ComputeLoop2535
  %Accumulator2537 = phi i32 [ %3394, %ComputeLoop2535 ], [ -1, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsINS_3cpp6stringEEERS1_T_.exit838 ]
  %ActiveBits2538 = phi i64 [ %3397, %ComputeLoop2535 ], [ %1138, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsINS_3cpp6stringEEERS1_T_.exit838 ]
  %3391 = call i64 @llvm.cttz.i64(i64 %ActiveBits2538, i1 true)
  %3392 = trunc i64 %3391 to i32
  %3393 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i21.i.i829, i32 %3392)
  %3394 = and i32 %Accumulator2537, %3393
  %3395 = shl i64 1, %3391
  %3396 = xor i64 %3395, -1
  %3397 = and i64 %ActiveBits2538, %3396
  %3398 = icmp eq i64 %3397, 0
  br i1 %3398, label %ComputeEnd2536, label %ComputeLoop2535, !amdgpu.uniform !42

ComputeEnd2536:                                   ; preds = %ComputeLoop2535
  %.lcssa4103 = phi i32 [ %3394, %ComputeLoop2535 ]
  %3399 = trunc i64 %1138 to i32
  %3400 = lshr i64 %1138, 32
  %3401 = trunc i64 %3400 to i32
  %3402 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %3399, i32 0)
  %3403 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %3401, i32 %3402)
  %3404 = icmp eq i32 %3403, 0
  %3405 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3404)
  %3406 = extractvalue { i1, i64 } %3405, 0
  %3407 = extractvalue { i1, i64 } %3405, 1
  br i1 %3406, label %1139, label %1142

ComputeLoop2539:                                  ; preds = %for.cond.i.i1.i847, %ComputeLoop2539
  %Accumulator2541 = phi i32 [ %3412, %ComputeLoop2539 ], [ 0, %for.cond.i.i1.i847 ]
  %OldValuePhi2542 = phi i32 [ %3411, %ComputeLoop2539 ], [ poison, %for.cond.i.i1.i847 ]
  %ActiveBits2543 = phi i64 [ %3415, %ComputeLoop2539 ], [ %1151, %for.cond.i.i1.i847 ]
  %3408 = call i64 @llvm.cttz.i64(i64 %ActiveBits2543, i1 true)
  %3409 = trunc i64 %3408 to i32
  %3410 = call i32 @llvm.amdgcn.readlane.i32(i32 %shl.i.i.i.i.i859, i32 %3409)
  %3411 = call i32 @llvm.amdgcn.writelane.i32(i32 %Accumulator2541, i32 %3409, i32 %OldValuePhi2542)
  %3412 = or i32 %Accumulator2541, %3410
  %3413 = shl i64 1, %3408
  %3414 = xor i64 %3413, -1
  %3415 = and i64 %ActiveBits2543, %3414
  %3416 = icmp eq i64 %3415, 0
  br i1 %3416, label %ComputeEnd2540, label %ComputeLoop2539, !amdgpu.uniform !42

ComputeEnd2540:                                   ; preds = %ComputeLoop2539
  %.lcssa4098 = phi i32 [ %3411, %ComputeLoop2539 ]
  %.lcssa4097 = phi i32 [ %3412, %ComputeLoop2539 ]
  %div8.i.i.i.i.i855 = lshr i32 %spec.store.select.i.i.i854, 5
  %idxprom.i.i.i.i.i856 = zext nneg i32 %div8.i.i.i.i.i855 to i64
  %3417 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i.i.i856
  %3418 = trunc i64 %1151 to i32
  %3419 = lshr i64 %1151, 32
  %3420 = trunc i64 %3419 to i32
  %3421 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %3418, i32 0)
  %3422 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %3420, i32 %3421)
  %3423 = icmp eq i32 %3422, 0
  %3424 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3423)
  %3425 = extractvalue { i1, i64 } %3424, 0
  %3426 = extractvalue { i1, i64 } %3424, 1
  br i1 %3425, label %1152, label %1154

Flow3691:                                         ; preds = %1172, %if.then.i.i.i.i865
  %3427 = phi i1 [ false, %1172 ], [ true, %if.then.i.i.i.i865 ]
  br i1 %3427, label %if.end14.i.i.i936, label %Flow3692, !amdgpu.uniform !42

ComputeLoop2544:                                  ; preds = %if.then12.i.i.i870, %ComputeLoop2544
  %Accumulator2546 = phi i32 [ %3431, %ComputeLoop2544 ], [ -1, %if.then12.i.i.i870 ]
  %ActiveBits2547 = phi i64 [ %3434, %ComputeLoop2544 ], [ %1169, %if.then12.i.i.i870 ]
  %3428 = call i64 @llvm.cttz.i64(i64 %ActiveBits2547, i1 true)
  %3429 = trunc i64 %3428 to i32
  %3430 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i.i.i876, i32 %3429)
  %3431 = and i32 %Accumulator2546, %3430
  %3432 = shl i64 1, %3428
  %3433 = xor i64 %3432, -1
  %3434 = and i64 %ActiveBits2547, %3433
  %3435 = icmp eq i64 %3434, 0
  br i1 %3435, label %ComputeEnd2545, label %ComputeLoop2544, !amdgpu.uniform !42

ComputeEnd2545:                                   ; preds = %ComputeLoop2544
  %.lcssa4099 = phi i32 [ %3431, %ComputeLoop2544 ]
  %3436 = trunc i64 %1169 to i32
  %3437 = lshr i64 %1169, 32
  %3438 = trunc i64 %3437 to i32
  %3439 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %3436, i32 0)
  %3440 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %3438, i32 %3439)
  %3441 = icmp eq i32 %3440, 0
  %3442 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3441)
  %3443 = extractvalue { i1, i64 } %3442, 0
  %3444 = extractvalue { i1, i64 } %3442, 1
  br i1 %3443, label %1170, label %1172

ComputeLoop2548:                                  ; preds = %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit944, %ComputeLoop2548
  %Accumulator2550 = phi i32 [ %3448, %ComputeLoop2548 ], [ -1, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit944 ]
  %ActiveBits2551 = phi i64 [ %3451, %ComputeLoop2548 ], [ %1249, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit944 ]
  %3445 = call i64 @llvm.cttz.i64(i64 %ActiveBits2551, i1 true)
  %3446 = trunc i64 %3445 to i32
  %3447 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i21.i.i935, i32 %3446)
  %3448 = and i32 %Accumulator2550, %3447
  %3449 = shl i64 1, %3445
  %3450 = xor i64 %3449, -1
  %3451 = and i64 %ActiveBits2551, %3450
  %3452 = icmp eq i64 %3451, 0
  br i1 %3452, label %ComputeEnd2549, label %ComputeLoop2548, !amdgpu.uniform !42

ComputeEnd2549:                                   ; preds = %ComputeLoop2548
  %.lcssa4093 = phi i32 [ %3448, %ComputeLoop2548 ]
  %3453 = trunc i64 %1249 to i32
  %3454 = lshr i64 %1249, 32
  %3455 = trunc i64 %3454 to i32
  %3456 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %3453, i32 0)
  %3457 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %3455, i32 %3456)
  %3458 = icmp eq i32 %3457, 0
  %3459 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3458)
  %3460 = extractvalue { i1, i64 } %3459, 0
  %3461 = extractvalue { i1, i64 } %3459, 1
  br i1 %3460, label %1250, label %1253

ComputeLoop2552:                                  ; preds = %for.cond.i.i.i953, %ComputeLoop2552
  %Accumulator2554 = phi i32 [ %3466, %ComputeLoop2552 ], [ 0, %for.cond.i.i.i953 ]
  %OldValuePhi2555 = phi i32 [ %3465, %ComputeLoop2552 ], [ poison, %for.cond.i.i.i953 ]
  %ActiveBits2556 = phi i64 [ %3469, %ComputeLoop2552 ], [ %1391, %for.cond.i.i.i953 ]
  %3462 = call i64 @llvm.cttz.i64(i64 %ActiveBits2556, i1 true)
  %3463 = trunc i64 %3462 to i32
  %3464 = call i32 @llvm.amdgcn.readlane.i32(i32 %shl.i.i.i.i.i965, i32 %3463)
  %3465 = call i32 @llvm.amdgcn.writelane.i32(i32 %Accumulator2554, i32 %3463, i32 %OldValuePhi2555)
  %3466 = or i32 %Accumulator2554, %3464
  %3467 = shl i64 1, %3462
  %3468 = xor i64 %3467, -1
  %3469 = and i64 %ActiveBits2556, %3468
  %3470 = icmp eq i64 %3469, 0
  br i1 %3470, label %ComputeEnd2553, label %ComputeLoop2552, !amdgpu.uniform !42

ComputeEnd2553:                                   ; preds = %ComputeLoop2552
  %.lcssa4084 = phi i32 [ %3465, %ComputeLoop2552 ]
  %.lcssa4083 = phi i32 [ %3466, %ComputeLoop2552 ]
  %div8.i.i.i.i.i961 = lshr i32 %spec.store.select.i.i.i960, 5
  %idxprom.i.i.i.i.i962 = zext nneg i32 %div8.i.i.i.i.i961 to i64
  %3471 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i.i.i962
  %3472 = trunc i64 %1391 to i32
  %3473 = lshr i64 %1391, 32
  %3474 = trunc i64 %3473 to i32
  %3475 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %3472, i32 0)
  %3476 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %3474, i32 %3475)
  %3477 = icmp eq i32 %3476, 0
  %3478 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3477)
  %3479 = extractvalue { i1, i64 } %3478, 0
  %3480 = extractvalue { i1, i64 } %3478, 1
  br i1 %3479, label %1392, label %1394

Flow3638:                                         ; preds = %1412, %if.then.i.i.i.i971
  %3481 = phi i1 [ false, %1412 ], [ true, %if.then.i.i.i.i971 ]
  br i1 %3481, label %if.end14.i.i.i1042, label %Flow3639, !amdgpu.uniform !42

ComputeLoop2557:                                  ; preds = %if.then12.i.i.i976, %ComputeLoop2557
  %Accumulator2559 = phi i32 [ %3485, %ComputeLoop2557 ], [ -1, %if.then12.i.i.i976 ]
  %ActiveBits2560 = phi i64 [ %3488, %ComputeLoop2557 ], [ %1409, %if.then12.i.i.i976 ]
  %3482 = call i64 @llvm.cttz.i64(i64 %ActiveBits2560, i1 true)
  %3483 = trunc i64 %3482 to i32
  %3484 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i.i.i982, i32 %3483)
  %3485 = and i32 %Accumulator2559, %3484
  %3486 = shl i64 1, %3482
  %3487 = xor i64 %3486, -1
  %3488 = and i64 %ActiveBits2560, %3487
  %3489 = icmp eq i64 %3488, 0
  br i1 %3489, label %ComputeEnd2558, label %ComputeLoop2557, !amdgpu.uniform !42

ComputeEnd2558:                                   ; preds = %ComputeLoop2557
  %.lcssa4085 = phi i32 [ %3485, %ComputeLoop2557 ]
  %3490 = trunc i64 %1409 to i32
  %3491 = lshr i64 %1409, 32
  %3492 = trunc i64 %3491 to i32
  %3493 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %3490, i32 0)
  %3494 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %3492, i32 %3493)
  %3495 = icmp eq i32 %3494, 0
  %3496 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3495)
  %3497 = extractvalue { i1, i64 } %3496, 0
  %3498 = extractvalue { i1, i64 } %3496, 1
  br i1 %3497, label %1410, label %1412

ComputeLoop2561:                                  ; preds = %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsINS_3cpp6stringEEERS1_T_.exit1050, %ComputeLoop2561
  %Accumulator2563 = phi i32 [ %3502, %ComputeLoop2561 ], [ -1, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsINS_3cpp6stringEEERS1_T_.exit1050 ]
  %ActiveBits2564 = phi i64 [ %3505, %ComputeLoop2561 ], [ %1507, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsINS_3cpp6stringEEERS1_T_.exit1050 ]
  %3499 = call i64 @llvm.cttz.i64(i64 %ActiveBits2564, i1 true)
  %3500 = trunc i64 %3499 to i32
  %3501 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i21.i.i1041, i32 %3500)
  %3502 = and i32 %Accumulator2563, %3501
  %3503 = shl i64 1, %3499
  %3504 = xor i64 %3503, -1
  %3505 = and i64 %ActiveBits2564, %3504
  %3506 = icmp eq i64 %3505, 0
  br i1 %3506, label %ComputeEnd2562, label %ComputeLoop2561, !amdgpu.uniform !42

ComputeEnd2562:                                   ; preds = %ComputeLoop2561
  %.lcssa4075 = phi i32 [ %3502, %ComputeLoop2561 ]
  %3507 = trunc i64 %1507 to i32
  %3508 = lshr i64 %1507, 32
  %3509 = trunc i64 %3508 to i32
  %3510 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %3507, i32 0)
  %3511 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %3509, i32 %3510)
  %3512 = icmp eq i32 %3511, 0
  %3513 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3512)
  %3514 = extractvalue { i1, i64 } %3513, 0
  %3515 = extractvalue { i1, i64 } %3513, 1
  br i1 %3514, label %1508, label %1511

ComputeLoop2565:                                  ; preds = %for.cond.i.i.i1054, %ComputeLoop2565
  %Accumulator2567 = phi i32 [ %3520, %ComputeLoop2565 ], [ 0, %for.cond.i.i.i1054 ]
  %OldValuePhi2568 = phi i32 [ %3519, %ComputeLoop2565 ], [ poison, %for.cond.i.i.i1054 ]
  %ActiveBits2569 = phi i64 [ %3523, %ComputeLoop2565 ], [ %1520, %for.cond.i.i.i1054 ]
  %3516 = call i64 @llvm.cttz.i64(i64 %ActiveBits2569, i1 true)
  %3517 = trunc i64 %3516 to i32
  %3518 = call i32 @llvm.amdgcn.readlane.i32(i32 %shl.i.i.i.i.i1066, i32 %3517)
  %3519 = call i32 @llvm.amdgcn.writelane.i32(i32 %Accumulator2567, i32 %3517, i32 %OldValuePhi2568)
  %3520 = or i32 %Accumulator2567, %3518
  %3521 = shl i64 1, %3516
  %3522 = xor i64 %3521, -1
  %3523 = and i64 %ActiveBits2569, %3522
  %3524 = icmp eq i64 %3523, 0
  br i1 %3524, label %ComputeEnd2566, label %ComputeLoop2565, !amdgpu.uniform !42

ComputeEnd2566:                                   ; preds = %ComputeLoop2565
  %.lcssa4070 = phi i32 [ %3519, %ComputeLoop2565 ]
  %.lcssa4069 = phi i32 [ %3520, %ComputeLoop2565 ]
  %div8.i.i.i.i.i1062 = lshr i32 %spec.store.select.i.i.i1061, 5
  %idxprom.i.i.i.i.i1063 = zext nneg i32 %div8.i.i.i.i.i1062 to i64
  %3525 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i.i.i1063
  %3526 = trunc i64 %1520 to i32
  %3527 = lshr i64 %1520, 32
  %3528 = trunc i64 %3527 to i32
  %3529 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %3526, i32 0)
  %3530 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %3528, i32 %3529)
  %3531 = icmp eq i32 %3530, 0
  %3532 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3531)
  %3533 = extractvalue { i1, i64 } %3532, 0
  %3534 = extractvalue { i1, i64 } %3532, 1
  br i1 %3533, label %1521, label %1523

Flow3612:                                         ; preds = %1541, %if.then.i.i.i.i1072
  %3535 = phi i1 [ false, %1541 ], [ true, %if.then.i.i.i.i1072 ]
  br i1 %3535, label %if.end14.i.i.i1141, label %Flow3613, !amdgpu.uniform !42

ComputeLoop2570:                                  ; preds = %if.then12.i.i.i1077, %ComputeLoop2570
  %Accumulator2572 = phi i32 [ %3539, %ComputeLoop2570 ], [ -1, %if.then12.i.i.i1077 ]
  %ActiveBits2573 = phi i64 [ %3542, %ComputeLoop2570 ], [ %1538, %if.then12.i.i.i1077 ]
  %3536 = call i64 @llvm.cttz.i64(i64 %ActiveBits2573, i1 true)
  %3537 = trunc i64 %3536 to i32
  %3538 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i.i.i1083, i32 %3537)
  %3539 = and i32 %Accumulator2572, %3538
  %3540 = shl i64 1, %3536
  %3541 = xor i64 %3540, -1
  %3542 = and i64 %ActiveBits2573, %3541
  %3543 = icmp eq i64 %3542, 0
  br i1 %3543, label %ComputeEnd2571, label %ComputeLoop2570, !amdgpu.uniform !42

ComputeEnd2571:                                   ; preds = %ComputeLoop2570
  %.lcssa4071 = phi i32 [ %3539, %ComputeLoop2570 ]
  %3544 = trunc i64 %1538 to i32
  %3545 = lshr i64 %1538, 32
  %3546 = trunc i64 %3545 to i32
  %3547 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %3544, i32 0)
  %3548 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %3546, i32 %3547)
  %3549 = icmp eq i32 %3548, 0
  %3550 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3549)
  %3551 = extractvalue { i1, i64 } %3550, 0
  %3552 = extractvalue { i1, i64 } %3550, 1
  br i1 %3551, label %1539, label %1541

ComputeLoop2574:                                  ; preds = %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIcEERS1_T_.exit1149, %ComputeLoop2574
  %Accumulator2576 = phi i32 [ %3556, %ComputeLoop2574 ], [ -1, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIcEERS1_T_.exit1149 ]
  %ActiveBits2577 = phi i64 [ %3559, %ComputeLoop2574 ], [ %1593, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIcEERS1_T_.exit1149 ]
  %3553 = call i64 @llvm.cttz.i64(i64 %ActiveBits2577, i1 true)
  %3554 = trunc i64 %3553 to i32
  %3555 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i21.i.i1140, i32 %3554)
  %3556 = and i32 %Accumulator2576, %3555
  %3557 = shl i64 1, %3553
  %3558 = xor i64 %3557, -1
  %3559 = and i64 %ActiveBits2577, %3558
  %3560 = icmp eq i64 %3559, 0
  br i1 %3560, label %ComputeEnd2575, label %ComputeLoop2574, !amdgpu.uniform !42

ComputeEnd2575:                                   ; preds = %ComputeLoop2574
  %.lcssa4065 = phi i32 [ %3556, %ComputeLoop2574 ]
  %3561 = trunc i64 %1593 to i32
  %3562 = lshr i64 %1593, 32
  %3563 = trunc i64 %3562 to i32
  %3564 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %3561, i32 0)
  %3565 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %3563, i32 %3564)
  %3566 = icmp eq i32 %3565, 0
  %3567 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3566)
  %3568 = extractvalue { i1, i64 } %3567, 0
  %3569 = extractvalue { i1, i64 } %3567, 1
  br i1 %3568, label %1594, label %1597

ComputeLoop2578:                                  ; preds = %for.cond.i.i1.i1158, %ComputeLoop2578
  %Accumulator2580 = phi i32 [ %3574, %ComputeLoop2578 ], [ 0, %for.cond.i.i1.i1158 ]
  %OldValuePhi2581 = phi i32 [ %3573, %ComputeLoop2578 ], [ poison, %for.cond.i.i1.i1158 ]
  %ActiveBits2582 = phi i64 [ %3577, %ComputeLoop2578 ], [ %1606, %for.cond.i.i1.i1158 ]
  %3570 = call i64 @llvm.cttz.i64(i64 %ActiveBits2582, i1 true)
  %3571 = trunc i64 %3570 to i32
  %3572 = call i32 @llvm.amdgcn.readlane.i32(i32 %shl.i.i.i.i.i1170, i32 %3571)
  %3573 = call i32 @llvm.amdgcn.writelane.i32(i32 %Accumulator2580, i32 %3571, i32 %OldValuePhi2581)
  %3574 = or i32 %Accumulator2580, %3572
  %3575 = shl i64 1, %3570
  %3576 = xor i64 %3575, -1
  %3577 = and i64 %ActiveBits2582, %3576
  %3578 = icmp eq i64 %3577, 0
  br i1 %3578, label %ComputeEnd2579, label %ComputeLoop2578, !amdgpu.uniform !42

ComputeEnd2579:                                   ; preds = %ComputeLoop2578
  %.lcssa4060 = phi i32 [ %3573, %ComputeLoop2578 ]
  %.lcssa4059 = phi i32 [ %3574, %ComputeLoop2578 ]
  %div8.i.i.i.i.i1166 = lshr i32 %spec.store.select.i.i.i1165, 5
  %idxprom.i.i.i.i.i1167 = zext nneg i32 %div8.i.i.i.i.i1166 to i64
  %3579 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i.i.i1167
  %3580 = trunc i64 %1606 to i32
  %3581 = lshr i64 %1606, 32
  %3582 = trunc i64 %3581 to i32
  %3583 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %3580, i32 0)
  %3584 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %3582, i32 %3583)
  %3585 = icmp eq i32 %3584, 0
  %3586 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3585)
  %3587 = extractvalue { i1, i64 } %3586, 0
  %3588 = extractvalue { i1, i64 } %3586, 1
  br i1 %3587, label %1607, label %1609

Flow3601:                                         ; preds = %1627, %if.then.i.i.i.i1176
  %3589 = phi i1 [ false, %1627 ], [ true, %if.then.i.i.i.i1176 ]
  br i1 %3589, label %if.end14.i.i.i1247, label %Flow3602, !amdgpu.uniform !42

ComputeLoop2583:                                  ; preds = %if.then12.i.i.i1181, %ComputeLoop2583
  %Accumulator2585 = phi i32 [ %3593, %ComputeLoop2583 ], [ -1, %if.then12.i.i.i1181 ]
  %ActiveBits2586 = phi i64 [ %3596, %ComputeLoop2583 ], [ %1624, %if.then12.i.i.i1181 ]
  %3590 = call i64 @llvm.cttz.i64(i64 %ActiveBits2586, i1 true)
  %3591 = trunc i64 %3590 to i32
  %3592 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i.i.i1187, i32 %3591)
  %3593 = and i32 %Accumulator2585, %3592
  %3594 = shl i64 1, %3590
  %3595 = xor i64 %3594, -1
  %3596 = and i64 %ActiveBits2586, %3595
  %3597 = icmp eq i64 %3596, 0
  br i1 %3597, label %ComputeEnd2584, label %ComputeLoop2583, !amdgpu.uniform !42

ComputeEnd2584:                                   ; preds = %ComputeLoop2583
  %.lcssa4061 = phi i32 [ %3593, %ComputeLoop2583 ]
  %3598 = trunc i64 %1624 to i32
  %3599 = lshr i64 %1624, 32
  %3600 = trunc i64 %3599 to i32
  %3601 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %3598, i32 0)
  %3602 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %3600, i32 %3601)
  %3603 = icmp eq i32 %3602, 0
  %3604 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3603)
  %3605 = extractvalue { i1, i64 } %3604, 0
  %3606 = extractvalue { i1, i64 } %3604, 1
  br i1 %3605, label %1625, label %1627

ComputeLoop2587:                                  ; preds = %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit1255, %ComputeLoop2587
  %Accumulator2589 = phi i32 [ %3610, %ComputeLoop2587 ], [ -1, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit1255 ]
  %ActiveBits2590 = phi i64 [ %3613, %ComputeLoop2587 ], [ %1704, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit1255 ]
  %3607 = call i64 @llvm.cttz.i64(i64 %ActiveBits2590, i1 true)
  %3608 = trunc i64 %3607 to i32
  %3609 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i21.i.i1246, i32 %3608)
  %3610 = and i32 %Accumulator2589, %3609
  %3611 = shl i64 1, %3607
  %3612 = xor i64 %3611, -1
  %3613 = and i64 %ActiveBits2590, %3612
  %3614 = icmp eq i64 %3613, 0
  br i1 %3614, label %ComputeEnd2588, label %ComputeLoop2587, !amdgpu.uniform !42

ComputeEnd2588:                                   ; preds = %ComputeLoop2587
  %.lcssa4055 = phi i32 [ %3610, %ComputeLoop2587 ]
  %3615 = trunc i64 %1704 to i32
  %3616 = lshr i64 %1704, 32
  %3617 = trunc i64 %3616 to i32
  %3618 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %3615, i32 0)
  %3619 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %3617, i32 %3618)
  %3620 = icmp eq i32 %3619, 0
  %3621 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3620)
  %3622 = extractvalue { i1, i64 } %3621, 0
  %3623 = extractvalue { i1, i64 } %3621, 1
  br i1 %3622, label %1705, label %1708

ComputeLoop2591:                                  ; preds = %for.cond.i.i.i1257, %ComputeLoop2591
  %Accumulator2593 = phi i32 [ %3628, %ComputeLoop2591 ], [ 0, %for.cond.i.i.i1257 ]
  %OldValuePhi2594 = phi i32 [ %3627, %ComputeLoop2591 ], [ poison, %for.cond.i.i.i1257 ]
  %ActiveBits2595 = phi i64 [ %3631, %ComputeLoop2591 ], [ %1717, %for.cond.i.i.i1257 ]
  %3624 = call i64 @llvm.cttz.i64(i64 %ActiveBits2595, i1 true)
  %3625 = trunc i64 %3624 to i32
  %3626 = call i32 @llvm.amdgcn.readlane.i32(i32 %shl.i.i.i.i.i1269, i32 %3625)
  %3627 = call i32 @llvm.amdgcn.writelane.i32(i32 %Accumulator2593, i32 %3625, i32 %OldValuePhi2594)
  %3628 = or i32 %Accumulator2593, %3626
  %3629 = shl i64 1, %3624
  %3630 = xor i64 %3629, -1
  %3631 = and i64 %ActiveBits2595, %3630
  %3632 = icmp eq i64 %3631, 0
  br i1 %3632, label %ComputeEnd2592, label %ComputeLoop2591, !amdgpu.uniform !42

ComputeEnd2592:                                   ; preds = %ComputeLoop2591
  %.lcssa4050 = phi i32 [ %3627, %ComputeLoop2591 ]
  %.lcssa4049 = phi i32 [ %3628, %ComputeLoop2591 ]
  %div8.i.i.i.i.i1265 = lshr i32 %spec.store.select.i.i.i1264, 5
  %idxprom.i.i.i.i.i1266 = zext nneg i32 %div8.i.i.i.i.i1265 to i64
  %3633 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i.i.i1266
  %3634 = trunc i64 %1717 to i32
  %3635 = lshr i64 %1717, 32
  %3636 = trunc i64 %3635 to i32
  %3637 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %3634, i32 0)
  %3638 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %3636, i32 %3637)
  %3639 = icmp eq i32 %3638, 0
  %3640 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3639)
  %3641 = extractvalue { i1, i64 } %3640, 0
  %3642 = extractvalue { i1, i64 } %3640, 1
  br i1 %3641, label %1718, label %1720

Flow3575:                                         ; preds = %1738, %if.then.i.i.i.i1275
  %3643 = phi i1 [ false, %1738 ], [ true, %if.then.i.i.i.i1275 ]
  br i1 %3643, label %if.end14.i.i.i1346, label %Flow3576, !amdgpu.uniform !42

ComputeLoop2596:                                  ; preds = %if.then12.i.i.i1280, %ComputeLoop2596
  %Accumulator2598 = phi i32 [ %3647, %ComputeLoop2596 ], [ -1, %if.then12.i.i.i1280 ]
  %ActiveBits2599 = phi i64 [ %3650, %ComputeLoop2596 ], [ %1735, %if.then12.i.i.i1280 ]
  %3644 = call i64 @llvm.cttz.i64(i64 %ActiveBits2599, i1 true)
  %3645 = trunc i64 %3644 to i32
  %3646 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i.i.i1286, i32 %3645)
  %3647 = and i32 %Accumulator2598, %3646
  %3648 = shl i64 1, %3644
  %3649 = xor i64 %3648, -1
  %3650 = and i64 %ActiveBits2599, %3649
  %3651 = icmp eq i64 %3650, 0
  br i1 %3651, label %ComputeEnd2597, label %ComputeLoop2596, !amdgpu.uniform !42

ComputeEnd2597:                                   ; preds = %ComputeLoop2596
  %.lcssa4051 = phi i32 [ %3647, %ComputeLoop2596 ]
  %3652 = trunc i64 %1735 to i32
  %3653 = lshr i64 %1735, 32
  %3654 = trunc i64 %3653 to i32
  %3655 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %3652, i32 0)
  %3656 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %3654, i32 %3655)
  %3657 = icmp eq i32 %3656, 0
  %3658 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3657)
  %3659 = extractvalue { i1, i64 } %3658, 0
  %3660 = extractvalue { i1, i64 } %3658, 1
  br i1 %3659, label %1736, label %1738

ComputeLoop2600:                                  ; preds = %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsINS_3cpp11string_viewEEERS1_T_.exit, %ComputeLoop2600
  %Accumulator2602 = phi i32 [ %3664, %ComputeLoop2600 ], [ -1, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsINS_3cpp11string_viewEEERS1_T_.exit ]
  %ActiveBits2603 = phi i64 [ %3667, %ComputeLoop2600 ], [ %1815, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsINS_3cpp11string_viewEEERS1_T_.exit ]
  %3661 = call i64 @llvm.cttz.i64(i64 %ActiveBits2603, i1 true)
  %3662 = trunc i64 %3661 to i32
  %3663 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i21.i.i1345, i32 %3662)
  %3664 = and i32 %Accumulator2602, %3663
  %3665 = shl i64 1, %3661
  %3666 = xor i64 %3665, -1
  %3667 = and i64 %ActiveBits2603, %3666
  %3668 = icmp eq i64 %3667, 0
  br i1 %3668, label %ComputeEnd2601, label %ComputeLoop2600, !amdgpu.uniform !42

ComputeEnd2601:                                   ; preds = %ComputeLoop2600
  %.lcssa4045 = phi i32 [ %3664, %ComputeLoop2600 ]
  %3669 = trunc i64 %1815 to i32
  %3670 = lshr i64 %1815, 32
  %3671 = trunc i64 %3670 to i32
  %3672 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %3669, i32 0)
  %3673 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %3671, i32 %3672)
  %3674 = icmp eq i32 %3673, 0
  %3675 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3674)
  %3676 = extractvalue { i1, i64 } %3675, 0
  %3677 = extractvalue { i1, i64 } %3675, 1
  br i1 %3676, label %1816, label %1819

ComputeLoop2604:                                  ; preds = %for.cond.i.i1.i1362, %ComputeLoop2604
  %Accumulator2606 = phi i32 [ %3682, %ComputeLoop2604 ], [ 0, %for.cond.i.i1.i1362 ]
  %OldValuePhi2607 = phi i32 [ %3681, %ComputeLoop2604 ], [ poison, %for.cond.i.i1.i1362 ]
  %ActiveBits2608 = phi i64 [ %3685, %ComputeLoop2604 ], [ %1828, %for.cond.i.i1.i1362 ]
  %3678 = call i64 @llvm.cttz.i64(i64 %ActiveBits2608, i1 true)
  %3679 = trunc i64 %3678 to i32
  %3680 = call i32 @llvm.amdgcn.readlane.i32(i32 %shl.i.i.i.i.i1374, i32 %3679)
  %3681 = call i32 @llvm.amdgcn.writelane.i32(i32 %Accumulator2606, i32 %3679, i32 %OldValuePhi2607)
  %3682 = or i32 %Accumulator2606, %3680
  %3683 = shl i64 1, %3678
  %3684 = xor i64 %3683, -1
  %3685 = and i64 %ActiveBits2608, %3684
  %3686 = icmp eq i64 %3685, 0
  br i1 %3686, label %ComputeEnd2605, label %ComputeLoop2604, !amdgpu.uniform !42

ComputeEnd2605:                                   ; preds = %ComputeLoop2604
  %.lcssa4040 = phi i32 [ %3681, %ComputeLoop2604 ]
  %.lcssa4039 = phi i32 [ %3682, %ComputeLoop2604 ]
  %div8.i.i.i.i.i1370 = lshr i32 %spec.store.select.i.i.i1369, 5
  %idxprom.i.i.i.i.i1371 = zext nneg i32 %div8.i.i.i.i.i1370 to i64
  %3687 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i.i.i1371
  %3688 = trunc i64 %1828 to i32
  %3689 = lshr i64 %1828, 32
  %3690 = trunc i64 %3689 to i32
  %3691 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %3688, i32 0)
  %3692 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %3690, i32 %3691)
  %3693 = icmp eq i32 %3692, 0
  %3694 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3693)
  %3695 = extractvalue { i1, i64 } %3694, 0
  %3696 = extractvalue { i1, i64 } %3694, 1
  br i1 %3695, label %1829, label %1831

Flow3549:                                         ; preds = %1849, %if.then.i.i.i.i1380
  %3697 = phi i1 [ false, %1849 ], [ true, %if.then.i.i.i.i1380 ]
  br i1 %3697, label %if.end14.i.i.i1451, label %Flow3550, !amdgpu.uniform !42

ComputeLoop2609:                                  ; preds = %if.then12.i.i.i1385, %ComputeLoop2609
  %Accumulator2611 = phi i32 [ %3701, %ComputeLoop2609 ], [ -1, %if.then12.i.i.i1385 ]
  %ActiveBits2612 = phi i64 [ %3704, %ComputeLoop2609 ], [ %1846, %if.then12.i.i.i1385 ]
  %3698 = call i64 @llvm.cttz.i64(i64 %ActiveBits2612, i1 true)
  %3699 = trunc i64 %3698 to i32
  %3700 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i.i.i1391, i32 %3699)
  %3701 = and i32 %Accumulator2611, %3700
  %3702 = shl i64 1, %3698
  %3703 = xor i64 %3702, -1
  %3704 = and i64 %ActiveBits2612, %3703
  %3705 = icmp eq i64 %3704, 0
  br i1 %3705, label %ComputeEnd2610, label %ComputeLoop2609, !amdgpu.uniform !42

ComputeEnd2610:                                   ; preds = %ComputeLoop2609
  %.lcssa4041 = phi i32 [ %3701, %ComputeLoop2609 ]
  %3706 = trunc i64 %1846 to i32
  %3707 = lshr i64 %1846, 32
  %3708 = trunc i64 %3707 to i32
  %3709 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %3706, i32 0)
  %3710 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %3708, i32 %3709)
  %3711 = icmp eq i32 %3710, 0
  %3712 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3711)
  %3713 = extractvalue { i1, i64 } %3712, 0
  %3714 = extractvalue { i1, i64 } %3712, 1
  br i1 %3713, label %1847, label %1849

ComputeLoop2613:                                  ; preds = %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit1459, %ComputeLoop2613
  %Accumulator2615 = phi i32 [ %3718, %ComputeLoop2613 ], [ -1, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit1459 ]
  %ActiveBits2616 = phi i64 [ %3721, %ComputeLoop2613 ], [ %1926, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit1459 ]
  %3715 = call i64 @llvm.cttz.i64(i64 %ActiveBits2616, i1 true)
  %3716 = trunc i64 %3715 to i32
  %3717 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i21.i.i1450, i32 %3716)
  %3718 = and i32 %Accumulator2615, %3717
  %3719 = shl i64 1, %3715
  %3720 = xor i64 %3719, -1
  %3721 = and i64 %ActiveBits2616, %3720
  %3722 = icmp eq i64 %3721, 0
  br i1 %3722, label %ComputeEnd2614, label %ComputeLoop2613, !amdgpu.uniform !42

ComputeEnd2614:                                   ; preds = %ComputeLoop2613
  %.lcssa4035 = phi i32 [ %3718, %ComputeLoop2613 ]
  %3723 = trunc i64 %1926 to i32
  %3724 = lshr i64 %1926, 32
  %3725 = trunc i64 %3724 to i32
  %3726 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %3723, i32 0)
  %3727 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %3725, i32 %3726)
  %3728 = icmp eq i32 %3727, 0
  %3729 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3728)
  %3730 = extractvalue { i1, i64 } %3729, 0
  %3731 = extractvalue { i1, i64 } %3729, 1
  br i1 %3730, label %1927, label %1930

ComputeLoop2617:                                  ; preds = %for.cond.i.i1.i1468, %ComputeLoop2617
  %Accumulator2619 = phi i32 [ %3736, %ComputeLoop2617 ], [ 0, %for.cond.i.i1.i1468 ]
  %OldValuePhi2620 = phi i32 [ %3735, %ComputeLoop2617 ], [ poison, %for.cond.i.i1.i1468 ]
  %ActiveBits2621 = phi i64 [ %3739, %ComputeLoop2617 ], [ %1943, %for.cond.i.i1.i1468 ]
  %3732 = call i64 @llvm.cttz.i64(i64 %ActiveBits2621, i1 true)
  %3733 = trunc i64 %3732 to i32
  %3734 = call i32 @llvm.amdgcn.readlane.i32(i32 %shl.i.i.i.i.i1480, i32 %3733)
  %3735 = call i32 @llvm.amdgcn.writelane.i32(i32 %Accumulator2619, i32 %3733, i32 %OldValuePhi2620)
  %3736 = or i32 %Accumulator2619, %3734
  %3737 = shl i64 1, %3732
  %3738 = xor i64 %3737, -1
  %3739 = and i64 %ActiveBits2621, %3738
  %3740 = icmp eq i64 %3739, 0
  br i1 %3740, label %ComputeEnd2618, label %ComputeLoop2617, !amdgpu.uniform !42

ComputeEnd2618:                                   ; preds = %ComputeLoop2617
  %.lcssa4029 = phi i32 [ %3735, %ComputeLoop2617 ]
  %.lcssa4028 = phi i32 [ %3736, %ComputeLoop2617 ]
  %div8.i.i.i.i.i1476 = lshr i32 %spec.store.select.i.i.i1475, 5
  %idxprom.i.i.i.i.i1477 = zext nneg i32 %div8.i.i.i.i.i1476 to i64
  %3741 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i.i.i1477
  %3742 = trunc i64 %1943 to i32
  %3743 = lshr i64 %1943, 32
  %3744 = trunc i64 %3743 to i32
  %3745 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %3742, i32 0)
  %3746 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %3744, i32 %3745)
  %3747 = icmp eq i32 %3746, 0
  %3748 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3747)
  %3749 = extractvalue { i1, i64 } %3748, 0
  %3750 = extractvalue { i1, i64 } %3748, 1
  br i1 %3749, label %1944, label %1946

Flow3523:                                         ; preds = %1964, %if.then.i.i.i.i1486
  %3751 = phi i1 [ false, %1964 ], [ true, %if.then.i.i.i.i1486 ]
  br i1 %3751, label %if.end14.i.i.i1557, label %Flow3524, !amdgpu.uniform !42

ComputeLoop2622:                                  ; preds = %if.then12.i.i.i1491, %ComputeLoop2622
  %Accumulator2624 = phi i32 [ %3755, %ComputeLoop2622 ], [ -1, %if.then12.i.i.i1491 ]
  %ActiveBits2625 = phi i64 [ %3758, %ComputeLoop2622 ], [ %1961, %if.then12.i.i.i1491 ]
  %3752 = call i64 @llvm.cttz.i64(i64 %ActiveBits2625, i1 true)
  %3753 = trunc i64 %3752 to i32
  %3754 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i.i.i1497, i32 %3753)
  %3755 = and i32 %Accumulator2624, %3754
  %3756 = shl i64 1, %3752
  %3757 = xor i64 %3756, -1
  %3758 = and i64 %ActiveBits2625, %3757
  %3759 = icmp eq i64 %3758, 0
  br i1 %3759, label %ComputeEnd2623, label %ComputeLoop2622, !amdgpu.uniform !42

ComputeEnd2623:                                   ; preds = %ComputeLoop2622
  %.lcssa4030 = phi i32 [ %3755, %ComputeLoop2622 ]
  %3760 = trunc i64 %1961 to i32
  %3761 = lshr i64 %1961, 32
  %3762 = trunc i64 %3761 to i32
  %3763 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %3760, i32 0)
  %3764 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %3762, i32 %3763)
  %3765 = icmp eq i32 %3764, 0
  %3766 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3765)
  %3767 = extractvalue { i1, i64 } %3766, 0
  %3768 = extractvalue { i1, i64 } %3766, 1
  br i1 %3767, label %1962, label %1964

ComputeLoop2626:                                  ; preds = %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit1565, %ComputeLoop2626
  %Accumulator2628 = phi i32 [ %3772, %ComputeLoop2626 ], [ -1, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit1565 ]
  %ActiveBits2629 = phi i64 [ %3775, %ComputeLoop2626 ], [ %2059, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit1565 ]
  %3769 = call i64 @llvm.cttz.i64(i64 %ActiveBits2629, i1 true)
  %3770 = trunc i64 %3769 to i32
  %3771 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i21.i.i1556, i32 %3770)
  %3772 = and i32 %Accumulator2628, %3771
  %3773 = shl i64 1, %3769
  %3774 = xor i64 %3773, -1
  %3775 = and i64 %ActiveBits2629, %3774
  %3776 = icmp eq i64 %3775, 0
  br i1 %3776, label %ComputeEnd2627, label %ComputeLoop2626, !amdgpu.uniform !42

ComputeEnd2627:                                   ; preds = %ComputeLoop2626
  %.lcssa4020 = phi i32 [ %3772, %ComputeLoop2626 ]
  %3777 = trunc i64 %2059 to i32
  %3778 = lshr i64 %2059, 32
  %3779 = trunc i64 %3778 to i32
  %3780 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %3777, i32 0)
  %3781 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %3779, i32 %3780)
  %3782 = icmp eq i32 %3781, 0
  %3783 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3782)
  %3784 = extractvalue { i1, i64 } %3783, 0
  %3785 = extractvalue { i1, i64 } %3783, 1
  br i1 %3784, label %2060, label %2063

ComputeLoop2630:                                  ; preds = %for.cond.i.i.i1569, %ComputeLoop2630
  %Accumulator2632 = phi i32 [ %3790, %ComputeLoop2630 ], [ 0, %for.cond.i.i.i1569 ]
  %OldValuePhi2633 = phi i32 [ %3789, %ComputeLoop2630 ], [ poison, %for.cond.i.i.i1569 ]
  %ActiveBits2634 = phi i64 [ %3793, %ComputeLoop2630 ], [ %2072, %for.cond.i.i.i1569 ]
  %3786 = call i64 @llvm.cttz.i64(i64 %ActiveBits2634, i1 true)
  %3787 = trunc i64 %3786 to i32
  %3788 = call i32 @llvm.amdgcn.readlane.i32(i32 %shl.i.i.i.i.i1581, i32 %3787)
  %3789 = call i32 @llvm.amdgcn.writelane.i32(i32 %Accumulator2632, i32 %3787, i32 %OldValuePhi2633)
  %3790 = or i32 %Accumulator2632, %3788
  %3791 = shl i64 1, %3786
  %3792 = xor i64 %3791, -1
  %3793 = and i64 %ActiveBits2634, %3792
  %3794 = icmp eq i64 %3793, 0
  br i1 %3794, label %ComputeEnd2631, label %ComputeLoop2630, !amdgpu.uniform !42

ComputeEnd2631:                                   ; preds = %ComputeLoop2630
  %.lcssa4015 = phi i32 [ %3789, %ComputeLoop2630 ]
  %.lcssa4014 = phi i32 [ %3790, %ComputeLoop2630 ]
  %div8.i.i.i.i.i1577 = lshr i32 %spec.store.select.i.i.i1576, 5
  %idxprom.i.i.i.i.i1578 = zext nneg i32 %div8.i.i.i.i.i1577 to i64
  %3795 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i.i.i1578
  %3796 = trunc i64 %2072 to i32
  %3797 = lshr i64 %2072, 32
  %3798 = trunc i64 %3797 to i32
  %3799 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %3796, i32 0)
  %3800 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %3798, i32 %3799)
  %3801 = icmp eq i32 %3800, 0
  %3802 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3801)
  %3803 = extractvalue { i1, i64 } %3802, 0
  %3804 = extractvalue { i1, i64 } %3802, 1
  br i1 %3803, label %2073, label %2075

Flow3497:                                         ; preds = %2093, %if.then.i.i.i.i1587
  %3805 = phi i1 [ false, %2093 ], [ true, %if.then.i.i.i.i1587 ]
  br i1 %3805, label %if.end14.i.i.i1656, label %Flow3498, !amdgpu.uniform !42

ComputeLoop2635:                                  ; preds = %if.then12.i.i.i1592, %ComputeLoop2635
  %Accumulator2637 = phi i32 [ %3809, %ComputeLoop2635 ], [ -1, %if.then12.i.i.i1592 ]
  %ActiveBits2638 = phi i64 [ %3812, %ComputeLoop2635 ], [ %2090, %if.then12.i.i.i1592 ]
  %3806 = call i64 @llvm.cttz.i64(i64 %ActiveBits2638, i1 true)
  %3807 = trunc i64 %3806 to i32
  %3808 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i.i.i1598, i32 %3807)
  %3809 = and i32 %Accumulator2637, %3808
  %3810 = shl i64 1, %3806
  %3811 = xor i64 %3810, -1
  %3812 = and i64 %ActiveBits2638, %3811
  %3813 = icmp eq i64 %3812, 0
  br i1 %3813, label %ComputeEnd2636, label %ComputeLoop2635, !amdgpu.uniform !42

ComputeEnd2636:                                   ; preds = %ComputeLoop2635
  %.lcssa4016 = phi i32 [ %3809, %ComputeLoop2635 ]
  %3814 = trunc i64 %2090 to i32
  %3815 = lshr i64 %2090, 32
  %3816 = trunc i64 %3815 to i32
  %3817 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %3814, i32 0)
  %3818 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %3816, i32 %3817)
  %3819 = icmp eq i32 %3818, 0
  %3820 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3819)
  %3821 = extractvalue { i1, i64 } %3820, 0
  %3822 = extractvalue { i1, i64 } %3820, 1
  br i1 %3821, label %2091, label %2093

ComputeLoop2639:                                  ; preds = %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIcEERS1_T_.exit1664, %ComputeLoop2639
  %Accumulator2641 = phi i32 [ %3826, %ComputeLoop2639 ], [ -1, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIcEERS1_T_.exit1664 ]
  %ActiveBits2642 = phi i64 [ %3829, %ComputeLoop2639 ], [ %2145, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIcEERS1_T_.exit1664 ]
  %3823 = call i64 @llvm.cttz.i64(i64 %ActiveBits2642, i1 true)
  %3824 = trunc i64 %3823 to i32
  %3825 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i21.i.i1655, i32 %3824)
  %3826 = and i32 %Accumulator2641, %3825
  %3827 = shl i64 1, %3823
  %3828 = xor i64 %3827, -1
  %3829 = and i64 %ActiveBits2642, %3828
  %3830 = icmp eq i64 %3829, 0
  br i1 %3830, label %ComputeEnd2640, label %ComputeLoop2639, !amdgpu.uniform !42

ComputeEnd2640:                                   ; preds = %ComputeLoop2639
  %.lcssa4010 = phi i32 [ %3826, %ComputeLoop2639 ]
  %3831 = trunc i64 %2145 to i32
  %3832 = lshr i64 %2145, 32
  %3833 = trunc i64 %3832 to i32
  %3834 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %3831, i32 0)
  %3835 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %3833, i32 %3834)
  %3836 = icmp eq i32 %3835, 0
  %3837 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3836)
  %3838 = extractvalue { i1, i64 } %3837, 0
  %3839 = extractvalue { i1, i64 } %3837, 1
  br i1 %3838, label %2146, label %2149

ComputeLoop2643:                                  ; preds = %for.cond.i.i.i1673, %ComputeLoop2643
  %Accumulator2645 = phi i32 [ %3844, %ComputeLoop2643 ], [ 0, %for.cond.i.i.i1673 ]
  %OldValuePhi2646 = phi i32 [ %3843, %ComputeLoop2643 ], [ poison, %for.cond.i.i.i1673 ]
  %ActiveBits2647 = phi i64 [ %3847, %ComputeLoop2643 ], [ %2167, %for.cond.i.i.i1673 ]
  %3840 = call i64 @llvm.cttz.i64(i64 %ActiveBits2647, i1 true)
  %3841 = trunc i64 %3840 to i32
  %3842 = call i32 @llvm.amdgcn.readlane.i32(i32 %shl.i.i.i.i.i1685, i32 %3841)
  %3843 = call i32 @llvm.amdgcn.writelane.i32(i32 %Accumulator2645, i32 %3841, i32 %OldValuePhi2646)
  %3844 = or i32 %Accumulator2645, %3842
  %3845 = shl i64 1, %3840
  %3846 = xor i64 %3845, -1
  %3847 = and i64 %ActiveBits2647, %3846
  %3848 = icmp eq i64 %3847, 0
  br i1 %3848, label %ComputeEnd2644, label %ComputeLoop2643, !amdgpu.uniform !42

ComputeEnd2644:                                   ; preds = %ComputeLoop2643
  %.lcssa4005 = phi i32 [ %3843, %ComputeLoop2643 ]
  %.lcssa4004 = phi i32 [ %3844, %ComputeLoop2643 ]
  %div8.i.i.i.i.i1681 = lshr i32 %spec.store.select.i.i.i1680, 5
  %idxprom.i.i.i.i.i1682 = zext nneg i32 %div8.i.i.i.i.i1681 to i64
  %3849 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i.i.i1682
  %3850 = trunc i64 %2167 to i32
  %3851 = lshr i64 %2167, 32
  %3852 = trunc i64 %3851 to i32
  %3853 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %3850, i32 0)
  %3854 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %3852, i32 %3853)
  %3855 = icmp eq i32 %3854, 0
  %3856 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3855)
  %3857 = extractvalue { i1, i64 } %3856, 0
  %3858 = extractvalue { i1, i64 } %3856, 1
  br i1 %3857, label %2168, label %2170

Flow3480:                                         ; preds = %2188, %if.then.i.i.i.i1691
  %3859 = phi i1 [ false, %2188 ], [ true, %if.then.i.i.i.i1691 ]
  br i1 %3859, label %if.end14.i.i.i1762, label %Flow3481, !amdgpu.uniform !42

ComputeLoop2648:                                  ; preds = %if.then12.i.i.i1696, %ComputeLoop2648
  %Accumulator2650 = phi i32 [ %3863, %ComputeLoop2648 ], [ -1, %if.then12.i.i.i1696 ]
  %ActiveBits2651 = phi i64 [ %3866, %ComputeLoop2648 ], [ %2185, %if.then12.i.i.i1696 ]
  %3860 = call i64 @llvm.cttz.i64(i64 %ActiveBits2651, i1 true)
  %3861 = trunc i64 %3860 to i32
  %3862 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i.i.i1702, i32 %3861)
  %3863 = and i32 %Accumulator2650, %3862
  %3864 = shl i64 1, %3860
  %3865 = xor i64 %3864, -1
  %3866 = and i64 %ActiveBits2651, %3865
  %3867 = icmp eq i64 %3866, 0
  br i1 %3867, label %ComputeEnd2649, label %ComputeLoop2648, !amdgpu.uniform !42

ComputeEnd2649:                                   ; preds = %ComputeLoop2648
  %.lcssa4006 = phi i32 [ %3863, %ComputeLoop2648 ]
  %3868 = trunc i64 %2185 to i32
  %3869 = lshr i64 %2185, 32
  %3870 = trunc i64 %3869 to i32
  %3871 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %3868, i32 0)
  %3872 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %3870, i32 %3871)
  %3873 = icmp eq i32 %3872, 0
  %3874 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3873)
  %3875 = extractvalue { i1, i64 } %3874, 0
  %3876 = extractvalue { i1, i64 } %3874, 1
  br i1 %3875, label %2186, label %2188

ComputeLoop2652:                                  ; preds = %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsINS_3cpp6stringEEERS1_T_.exit1770, %ComputeLoop2652
  %Accumulator2654 = phi i32 [ %3880, %ComputeLoop2652 ], [ -1, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsINS_3cpp6stringEEERS1_T_.exit1770 ]
  %ActiveBits2655 = phi i64 [ %3883, %ComputeLoop2652 ], [ %2262, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsINS_3cpp6stringEEERS1_T_.exit1770 ]
  %3877 = call i64 @llvm.cttz.i64(i64 %ActiveBits2655, i1 true)
  %3878 = trunc i64 %3877 to i32
  %3879 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i21.i.i1761, i32 %3878)
  %3880 = and i32 %Accumulator2654, %3879
  %3881 = shl i64 1, %3877
  %3882 = xor i64 %3881, -1
  %3883 = and i64 %ActiveBits2655, %3882
  %3884 = icmp eq i64 %3883, 0
  br i1 %3884, label %ComputeEnd2653, label %ComputeLoop2652, !amdgpu.uniform !42

ComputeEnd2653:                                   ; preds = %ComputeLoop2652
  %.lcssa4000 = phi i32 [ %3880, %ComputeLoop2652 ]
  %3885 = trunc i64 %2262 to i32
  %3886 = lshr i64 %2262, 32
  %3887 = trunc i64 %3886 to i32
  %3888 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %3885, i32 0)
  %3889 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %3887, i32 %3888)
  %3890 = icmp eq i32 %3889, 0
  %3891 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3890)
  %3892 = extractvalue { i1, i64 } %3891, 0
  %3893 = extractvalue { i1, i64 } %3891, 1
  br i1 %3892, label %2263, label %2266

ComputeLoop2656:                                  ; preds = %for.cond.i.i1.i1779, %ComputeLoop2656
  %Accumulator2658 = phi i32 [ %3898, %ComputeLoop2656 ], [ 0, %for.cond.i.i1.i1779 ]
  %OldValuePhi2659 = phi i32 [ %3897, %ComputeLoop2656 ], [ poison, %for.cond.i.i1.i1779 ]
  %ActiveBits2660 = phi i64 [ %3901, %ComputeLoop2656 ], [ %2275, %for.cond.i.i1.i1779 ]
  %3894 = call i64 @llvm.cttz.i64(i64 %ActiveBits2660, i1 true)
  %3895 = trunc i64 %3894 to i32
  %3896 = call i32 @llvm.amdgcn.readlane.i32(i32 %shl.i.i.i.i.i1791, i32 %3895)
  %3897 = call i32 @llvm.amdgcn.writelane.i32(i32 %Accumulator2658, i32 %3895, i32 %OldValuePhi2659)
  %3898 = or i32 %Accumulator2658, %3896
  %3899 = shl i64 1, %3894
  %3900 = xor i64 %3899, -1
  %3901 = and i64 %ActiveBits2660, %3900
  %3902 = icmp eq i64 %3901, 0
  br i1 %3902, label %ComputeEnd2657, label %ComputeLoop2656, !amdgpu.uniform !42

ComputeEnd2657:                                   ; preds = %ComputeLoop2656
  %.lcssa3995 = phi i32 [ %3897, %ComputeLoop2656 ]
  %.lcssa3994 = phi i32 [ %3898, %ComputeLoop2656 ]
  %div8.i.i.i.i.i1787 = lshr i32 %spec.store.select.i.i.i1786, 5
  %idxprom.i.i.i.i.i1788 = zext nneg i32 %div8.i.i.i.i.i1787 to i64
  %3903 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i.i.i1788
  %3904 = trunc i64 %2275 to i32
  %3905 = lshr i64 %2275, 32
  %3906 = trunc i64 %3905 to i32
  %3907 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %3904, i32 0)
  %3908 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %3906, i32 %3907)
  %3909 = icmp eq i32 %3908, 0
  %3910 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3909)
  %3911 = extractvalue { i1, i64 } %3910, 0
  %3912 = extractvalue { i1, i64 } %3910, 1
  br i1 %3911, label %2276, label %2278

Flow3454:                                         ; preds = %2296, %if.then.i.i.i.i1797
  %3913 = phi i1 [ false, %2296 ], [ true, %if.then.i.i.i.i1797 ]
  br i1 %3913, label %if.end14.i.i.i1868, label %Flow3455, !amdgpu.uniform !42

ComputeLoop2661:                                  ; preds = %if.then12.i.i.i1802, %ComputeLoop2661
  %Accumulator2663 = phi i32 [ %3917, %ComputeLoop2661 ], [ -1, %if.then12.i.i.i1802 ]
  %ActiveBits2664 = phi i64 [ %3920, %ComputeLoop2661 ], [ %2293, %if.then12.i.i.i1802 ]
  %3914 = call i64 @llvm.cttz.i64(i64 %ActiveBits2664, i1 true)
  %3915 = trunc i64 %3914 to i32
  %3916 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i.i.i1808, i32 %3915)
  %3917 = and i32 %Accumulator2663, %3916
  %3918 = shl i64 1, %3914
  %3919 = xor i64 %3918, -1
  %3920 = and i64 %ActiveBits2664, %3919
  %3921 = icmp eq i64 %3920, 0
  br i1 %3921, label %ComputeEnd2662, label %ComputeLoop2661, !amdgpu.uniform !42

ComputeEnd2662:                                   ; preds = %ComputeLoop2661
  %.lcssa3996 = phi i32 [ %3917, %ComputeLoop2661 ]
  %3922 = trunc i64 %2293 to i32
  %3923 = lshr i64 %2293, 32
  %3924 = trunc i64 %3923 to i32
  %3925 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %3922, i32 0)
  %3926 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %3924, i32 %3925)
  %3927 = icmp eq i32 %3926, 0
  %3928 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3927)
  %3929 = extractvalue { i1, i64 } %3928, 0
  %3930 = extractvalue { i1, i64 } %3928, 1
  br i1 %3929, label %2294, label %2296

ComputeLoop2665:                                  ; preds = %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit1876, %ComputeLoop2665
  %Accumulator2667 = phi i32 [ %3934, %ComputeLoop2665 ], [ -1, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit1876 ]
  %ActiveBits2668 = phi i64 [ %3937, %ComputeLoop2665 ], [ %2370, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit1876 ]
  %3931 = call i64 @llvm.cttz.i64(i64 %ActiveBits2668, i1 true)
  %3932 = trunc i64 %3931 to i32
  %3933 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i21.i.i1867, i32 %3932)
  %3934 = and i32 %Accumulator2667, %3933
  %3935 = shl i64 1, %3931
  %3936 = xor i64 %3935, -1
  %3937 = and i64 %ActiveBits2668, %3936
  %3938 = icmp eq i64 %3937, 0
  br i1 %3938, label %ComputeEnd2666, label %ComputeLoop2665, !amdgpu.uniform !42

ComputeEnd2666:                                   ; preds = %ComputeLoop2665
  %.lcssa3990 = phi i32 [ %3934, %ComputeLoop2665 ]
  %3939 = trunc i64 %2370 to i32
  %3940 = lshr i64 %2370, 32
  %3941 = trunc i64 %3940 to i32
  %3942 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %3939, i32 0)
  %3943 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %3941, i32 %3942)
  %3944 = icmp eq i32 %3943, 0
  %3945 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3944)
  %3946 = extractvalue { i1, i64 } %3945, 0
  %3947 = extractvalue { i1, i64 } %3945, 1
  br i1 %3946, label %2371, label %2374

ComputeLoop2669:                                  ; preds = %for.cond.i.i.i1885, %ComputeLoop2669
  %Accumulator2671 = phi i32 [ %3952, %ComputeLoop2669 ], [ 0, %for.cond.i.i.i1885 ]
  %OldValuePhi2672 = phi i32 [ %3951, %ComputeLoop2669 ], [ poison, %for.cond.i.i.i1885 ]
  %ActiveBits2673 = phi i64 [ %3955, %ComputeLoop2669 ], [ %2496, %for.cond.i.i.i1885 ]
  %3948 = call i64 @llvm.cttz.i64(i64 %ActiveBits2673, i1 true)
  %3949 = trunc i64 %3948 to i32
  %3950 = call i32 @llvm.amdgcn.readlane.i32(i32 %shl.i.i.i.i.i1897, i32 %3949)
  %3951 = call i32 @llvm.amdgcn.writelane.i32(i32 %Accumulator2671, i32 %3949, i32 %OldValuePhi2672)
  %3952 = or i32 %Accumulator2671, %3950
  %3953 = shl i64 1, %3948
  %3954 = xor i64 %3953, -1
  %3955 = and i64 %ActiveBits2673, %3954
  %3956 = icmp eq i64 %3955, 0
  br i1 %3956, label %ComputeEnd2670, label %ComputeLoop2669, !amdgpu.uniform !42

ComputeEnd2670:                                   ; preds = %ComputeLoop2669
  %.lcssa3981 = phi i32 [ %3951, %ComputeLoop2669 ]
  %.lcssa3980 = phi i32 [ %3952, %ComputeLoop2669 ]
  %div8.i.i.i.i.i1893 = lshr i32 %spec.store.select.i.i.i1892, 5
  %idxprom.i.i.i.i.i1894 = zext nneg i32 %div8.i.i.i.i.i1893 to i64
  %3957 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i.i.i1894
  %3958 = trunc i64 %2496 to i32
  %3959 = lshr i64 %2496, 32
  %3960 = trunc i64 %3959 to i32
  %3961 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %3958, i32 0)
  %3962 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %3960, i32 %3961)
  %3963 = icmp eq i32 %3962, 0
  %3964 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3963)
  %3965 = extractvalue { i1, i64 } %3964, 0
  %3966 = extractvalue { i1, i64 } %3964, 1
  br i1 %3965, label %2497, label %2499

Flow3411:                                         ; preds = %2517, %if.then.i.i.i.i1903
  %3967 = phi i1 [ false, %2517 ], [ true, %if.then.i.i.i.i1903 ]
  br i1 %3967, label %if.end14.i.i.i1974, label %Flow3412, !amdgpu.uniform !42

ComputeLoop2674:                                  ; preds = %if.then12.i.i.i1908, %ComputeLoop2674
  %Accumulator2676 = phi i32 [ %3971, %ComputeLoop2674 ], [ -1, %if.then12.i.i.i1908 ]
  %ActiveBits2677 = phi i64 [ %3974, %ComputeLoop2674 ], [ %2514, %if.then12.i.i.i1908 ]
  %3968 = call i64 @llvm.cttz.i64(i64 %ActiveBits2677, i1 true)
  %3969 = trunc i64 %3968 to i32
  %3970 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i.i.i1914, i32 %3969)
  %3971 = and i32 %Accumulator2676, %3970
  %3972 = shl i64 1, %3968
  %3973 = xor i64 %3972, -1
  %3974 = and i64 %ActiveBits2677, %3973
  %3975 = icmp eq i64 %3974, 0
  br i1 %3975, label %ComputeEnd2675, label %ComputeLoop2674, !amdgpu.uniform !42

ComputeEnd2675:                                   ; preds = %ComputeLoop2674
  %.lcssa3982 = phi i32 [ %3971, %ComputeLoop2674 ]
  %3976 = trunc i64 %2514 to i32
  %3977 = lshr i64 %2514, 32
  %3978 = trunc i64 %3977 to i32
  %3979 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %3976, i32 0)
  %3980 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %3978, i32 %3979)
  %3981 = icmp eq i32 %3980, 0
  %3982 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3981)
  %3983 = extractvalue { i1, i64 } %3982, 0
  %3984 = extractvalue { i1, i64 } %3982, 1
  br i1 %3983, label %2515, label %2517

ComputeLoop2678:                                  ; preds = %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsINS_3cpp6stringEEERS1_T_.exit1982, %ComputeLoop2678
  %Accumulator2680 = phi i32 [ %3988, %ComputeLoop2678 ], [ -1, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsINS_3cpp6stringEEERS1_T_.exit1982 ]
  %ActiveBits2681 = phi i64 [ %3991, %ComputeLoop2678 ], [ %2612, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsINS_3cpp6stringEEERS1_T_.exit1982 ]
  %3985 = call i64 @llvm.cttz.i64(i64 %ActiveBits2681, i1 true)
  %3986 = trunc i64 %3985 to i32
  %3987 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i21.i.i1973, i32 %3986)
  %3988 = and i32 %Accumulator2680, %3987
  %3989 = shl i64 1, %3985
  %3990 = xor i64 %3989, -1
  %3991 = and i64 %ActiveBits2681, %3990
  %3992 = icmp eq i64 %3991, 0
  br i1 %3992, label %ComputeEnd2679, label %ComputeLoop2678, !amdgpu.uniform !42

ComputeEnd2679:                                   ; preds = %ComputeLoop2678
  %.lcssa3972 = phi i32 [ %3988, %ComputeLoop2678 ]
  %3993 = trunc i64 %2612 to i32
  %3994 = lshr i64 %2612, 32
  %3995 = trunc i64 %3994 to i32
  %3996 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %3993, i32 0)
  %3997 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %3995, i32 %3996)
  %3998 = icmp eq i32 %3997, 0
  %3999 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3998)
  %4000 = extractvalue { i1, i64 } %3999, 0
  %4001 = extractvalue { i1, i64 } %3999, 1
  br i1 %4000, label %2613, label %2616

ComputeLoop2682:                                  ; preds = %for.cond.i.i.i1986, %ComputeLoop2682
  %Accumulator2684 = phi i32 [ %4006, %ComputeLoop2682 ], [ 0, %for.cond.i.i.i1986 ]
  %OldValuePhi2685 = phi i32 [ %4005, %ComputeLoop2682 ], [ poison, %for.cond.i.i.i1986 ]
  %ActiveBits2686 = phi i64 [ %4009, %ComputeLoop2682 ], [ %2625, %for.cond.i.i.i1986 ]
  %4002 = call i64 @llvm.cttz.i64(i64 %ActiveBits2686, i1 true)
  %4003 = trunc i64 %4002 to i32
  %4004 = call i32 @llvm.amdgcn.readlane.i32(i32 %shl.i.i.i.i.i1998, i32 %4003)
  %4005 = call i32 @llvm.amdgcn.writelane.i32(i32 %Accumulator2684, i32 %4003, i32 %OldValuePhi2685)
  %4006 = or i32 %Accumulator2684, %4004
  %4007 = shl i64 1, %4002
  %4008 = xor i64 %4007, -1
  %4009 = and i64 %ActiveBits2686, %4008
  %4010 = icmp eq i64 %4009, 0
  br i1 %4010, label %ComputeEnd2683, label %ComputeLoop2682, !amdgpu.uniform !42

ComputeEnd2683:                                   ; preds = %ComputeLoop2682
  %.lcssa3967 = phi i32 [ %4005, %ComputeLoop2682 ]
  %.lcssa3966 = phi i32 [ %4006, %ComputeLoop2682 ]
  %div8.i.i.i.i.i1994 = lshr i32 %spec.store.select.i.i.i1993, 5
  %idxprom.i.i.i.i.i1995 = zext nneg i32 %div8.i.i.i.i.i1994 to i64
  %4011 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i.i.i1995
  %4012 = trunc i64 %2625 to i32
  %4013 = lshr i64 %2625, 32
  %4014 = trunc i64 %4013 to i32
  %4015 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %4012, i32 0)
  %4016 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %4014, i32 %4015)
  %4017 = icmp eq i32 %4016, 0
  %4018 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %4017)
  %4019 = extractvalue { i1, i64 } %4018, 0
  %4020 = extractvalue { i1, i64 } %4018, 1
  br i1 %4019, label %2626, label %2628

Flow3392:                                         ; preds = %2646, %if.then.i.i.i.i2004
  %4021 = phi i1 [ false, %2646 ], [ true, %if.then.i.i.i.i2004 ]
  br i1 %4021, label %if.end14.i.i.i2073, label %Flow3393, !amdgpu.uniform !42

ComputeLoop2687:                                  ; preds = %if.then12.i.i.i2009, %ComputeLoop2687
  %Accumulator2689 = phi i32 [ %4025, %ComputeLoop2687 ], [ -1, %if.then12.i.i.i2009 ]
  %ActiveBits2690 = phi i64 [ %4028, %ComputeLoop2687 ], [ %2643, %if.then12.i.i.i2009 ]
  %4022 = call i64 @llvm.cttz.i64(i64 %ActiveBits2690, i1 true)
  %4023 = trunc i64 %4022 to i32
  %4024 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i.i.i2015, i32 %4023)
  %4025 = and i32 %Accumulator2689, %4024
  %4026 = shl i64 1, %4022
  %4027 = xor i64 %4026, -1
  %4028 = and i64 %ActiveBits2690, %4027
  %4029 = icmp eq i64 %4028, 0
  br i1 %4029, label %ComputeEnd2688, label %ComputeLoop2687, !amdgpu.uniform !42

ComputeEnd2688:                                   ; preds = %ComputeLoop2687
  %.lcssa3968 = phi i32 [ %4025, %ComputeLoop2687 ]
  %4030 = trunc i64 %2643 to i32
  %4031 = lshr i64 %2643, 32
  %4032 = trunc i64 %4031 to i32
  %4033 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %4030, i32 0)
  %4034 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %4032, i32 %4033)
  %4035 = icmp eq i32 %4034, 0
  %4036 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %4035)
  %4037 = extractvalue { i1, i64 } %4036, 0
  %4038 = extractvalue { i1, i64 } %4036, 1
  br i1 %4037, label %2644, label %2646

ComputeLoop2691:                                  ; preds = %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIcEERS1_T_.exit2081, %ComputeLoop2691
  %Accumulator2693 = phi i32 [ %4042, %ComputeLoop2691 ], [ -1, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIcEERS1_T_.exit2081 ]
  %ActiveBits2694 = phi i64 [ %4045, %ComputeLoop2691 ], [ %2698, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIcEERS1_T_.exit2081 ]
  %4039 = call i64 @llvm.cttz.i64(i64 %ActiveBits2694, i1 true)
  %4040 = trunc i64 %4039 to i32
  %4041 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i21.i.i2072, i32 %4040)
  %4042 = and i32 %Accumulator2693, %4041
  %4043 = shl i64 1, %4039
  %4044 = xor i64 %4043, -1
  %4045 = and i64 %ActiveBits2694, %4044
  %4046 = icmp eq i64 %4045, 0
  br i1 %4046, label %ComputeEnd2692, label %ComputeLoop2691, !amdgpu.uniform !42

ComputeEnd2692:                                   ; preds = %ComputeLoop2691
  %.lcssa = phi i32 [ %4042, %ComputeLoop2691 ]
  %4047 = trunc i64 %2698 to i32
  %4048 = lshr i64 %2698, 32
  %4049 = trunc i64 %4048 to i32
  %4050 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %4047, i32 0)
  %4051 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %4049, i32 %4050)
  %4052 = icmp eq i32 %4051, 0
  %4053 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %4052)
  %4054 = extractvalue { i1, i64 } %4053, 0
  %4055 = extractvalue { i1, i64 } %4053, 1
  br i1 %4054, label %2699, label %2702
}

; Function Attrs: mustprogress nocallback nofree nosync nounwind speculatable willreturn memory(none)
declare i64 @llvm.usub.sat.i64(i64, i64) #2

; Function Attrs: mustprogress nocallback nofree nosync nounwind willreturn memory(inaccessiblemem: write)
declare void @llvm.assume(i1 noundef) #10

; Function Attrs: mustprogress nocallback nofree nounwind willreturn memory(argmem: write)
declare void @llvm.memset.p0.i64(ptr writeonly captures(none), i8, i64, i1 immarg) #11

; Function Attrs: mustprogress nocallback nofree nounwind willreturn memory(argmem: readwrite)
declare void @llvm.memcpy.p0.p0.i64(ptr noalias writeonly captures(none), ptr noalias readonly captures(none), i64, i1 immarg) #12

; Function Attrs: mustprogress nocallback nofree nosync nounwind speculatable willreturn memory(none)
declare i8 @llvm.abs.i8(i8, i1 immarg) #2

; Function Attrs: mustprogress nocallback nofree nounwind willreturn memory(argmem: readwrite)
declare void @llvm.memcpy.p0.p5.i64(ptr noalias writeonly captures(none), ptr addrspace(5) noalias readonly captures(none), i64, i1 immarg) #12

; Function Attrs: convergent mustprogress nofree norecurse nounwind
define internal fastcc noundef zeroext i1 @_ZN22__llvm_libc_22_0_0_git7testing8internal4testIiEEbPNS1_10RunContextENS0_8TestCondET_S6_PKcS8_NS1_8LocationE(ptr noundef writeonly captures(none) %Ctx, i32 noundef range(i32 -1, 256) %LHS, i32 noundef range(i32 -1, 241) %RHS, ptr noundef %LHSStr, ptr noundef %RHSStr, i32 range(i32 53, 188) %Loc.coerce1) unnamed_addr #8 comdat {
entry:
  %buffer.i.i.i = alloca %"class.__llvm_libc_22_0_0_git::IntegerToString", align 8, addrspace(5)
  %buffer.i.i.i91.i = alloca %"class.__llvm_libc_22_0_0_git::IntegerToString", align 8, addrspace(5)
  %buffer.i.i.i.i = alloca %"class.__llvm_libc_22_0_0_git::IntegerToString", align 8, addrspace(5)
  %cmp = icmp eq i32 %LHS, %RHS
  %cmp.inv = xor i1 %cmp, true
  %0 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.inv)
  %1 = extractvalue { i1, i64 } %0, 0
  %2 = extractvalue { i1, i64 } %0, 1
  br i1 %1, label %if.end.i, label %Flow3909

Flow3908:                                         ; preds = %Flow3907, %if.end.i
  %3 = phi ptr [ %cond.i.i, %Flow3907 ], [ addrspacecast (ptr addrspace(4) @_ZN22__llvm_libc_22_0_0_git3cpp6string14NULL_CHARACTERE to ptr), %if.end.i ]
  br label %if.end.i.i.i, !amdgpu.uniform !42

loadstoreloop2193.preheader:                      ; preds = %if.end4.i.i.i.i
  br label %loadstoreloop2193, !amdgpu.uniform !42

Flow3907:                                         ; preds = %Flow3906, %if.end4.i.i.i.i
  br label %Flow3908, !amdgpu.uniform !42

loadstoreloop2193:                                ; preds = %loadstoreloop2193.preheader, %loadstoreloop2193
  %6 = phi i64 [ %8, %loadstoreloop2193 ], [ 0, %loadstoreloop2193.preheader ]
  %7 = getelementptr i8, ptr %4, i64 %6
  store i8 0, ptr %7, align 1
  %8 = add i64 %6, 1
  %9 = icmp uge i64 %8, %spec.select.i
  br i1 %9, label %Flow3906, label %loadstoreloop2193, !amdgpu.uniform !42

Flow3906:                                         ; preds = %loadstoreloop2193
  br label %Flow3907, !amdgpu.uniform !42

Flow3909:                                         ; preds = %2738, %entry
  call void @llvm.amdgcn.end.cf.i64(i64 %2)
  br label %_ZZN22__llvm_libc_22_0_0_git7testing8internal4testIiEEbPNS1_10RunContextENS0_8TestCondET_S6_PKcS8_NS1_8LocationEENKUlbNS_3cpp11string_viewEE_clEbSB_.exit, !amdgpu.uniform !42

if.end.i.i.i:                                     ; preds = %Flow3908
  %cmp.not.i6.i.i.i = icmp ne ptr %3, addrspacecast (ptr addrspace(4) @_ZN22__llvm_libc_22_0_0_git3cpp6string14NULL_CHARACTERE to ptr)
  br i1 %cmp.not.i6.i.i.i, label %if.then.i.i.i.i, label %_ZN22__llvm_libc_22_0_0_git3cpp6stringC2Emc.exit.i, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git3cpp6stringC2Emc.exit.i: ; preds = %if.then.i.i.i.i, %if.end.i.i.i
  %10 = icmp ne i64 0, %spec.select.i
  br i1 %10, label %loadstoreloop2191.preheader, label %Flow3905, !amdgpu.uniform !42

loadstoreloop2191.preheader:                      ; preds = %_ZN22__llvm_libc_22_0_0_git3cpp6stringC2Emc.exit.i
  br label %loadstoreloop2191, !amdgpu.uniform !42

Flow3905:                                         ; preds = %Flow3904, %_ZN22__llvm_libc_22_0_0_git3cpp6stringC2Emc.exit.i
  br label %split2190, !amdgpu.uniform !42

loadstoreloop2191:                                ; preds = %loadstoreloop2191.preheader, %loadstoreloop2191
  %11 = phi i64 [ %13, %loadstoreloop2191 ], [ 0, %loadstoreloop2191.preheader ]
  %12 = getelementptr i8, ptr %3, i64 %11
  store i8 32, ptr %12, align 1
  %13 = add i64 %11, 1
  %14 = icmp uge i64 %13, %spec.select.i
  br i1 %14, label %Flow3904, label %loadstoreloop2191, !amdgpu.uniform !42

Flow3904:                                         ; preds = %loadstoreloop2191
  br label %Flow3905, !amdgpu.uniform !42

split2190:                                        ; preds = %Flow3905
  %15 = tail call i32 @llvm.amdgcn.mbcnt.lo(i32 -1, i32 0)
  %16 = tail call noundef i32 @llvm.amdgcn.mbcnt.hi(i32 -1, i32 %15)
  %sh_prom.i.i.i.i = zext i32 %16 to i64
  br label %for.cond.i.i1.i, !amdgpu.uniform !42

for.cond.i.i1.i:                                  ; preds = %split2190, %Flow3903
  %17 = phi i32 [ %55, %Flow3903 ], [ poison, %split2190 ]
  %18 = phi i64 [ %56, %Flow3903 ], [ poison, %split2190 ]
  %19 = phi i32 [ %57, %Flow3903 ], [ poison, %split2190 ]
  %retval.sroa.6.0.i.i.i = phi i32 [ %60, %Flow3903 ], [ undef, %split2190 ]
  %retval.sroa.2.0.i.i.i = phi i64 [ %59, %Flow3903 ], [ undef, %split2190 ]
  %retval.sroa.8.0.i.i.i = phi i32 [ %58, %Flow3903 ], [ undef, %split2190 ]
  %index.0.i.i.i = phi i32 [ %62, %Flow3903 ], [ 0, %split2190 ]
  %20 = load i32, ptr addrspace(1) @__llvm_rpc_client, align 8, !tbaa !68
  %cmp.not.i.i.i = icmp ult i32 %index.0.i.i.i, %20
  %spec.store.select.i.i.i = select i1 %cmp.not.i.i.i, i32 %index.0.i.i.i, i32 0
  %21 = tail call noundef i64 @llvm.amdgcn.ballot.i64(i1 true)
  %22 = and i32 %spec.store.select.i.i.i, 31
  %23 = lshr i64 %21, %sh_prom.i.i.i.i
  %24 = trunc i64 %23 to i32
  %conv4.i.i.i.i.i = and i32 %24, 1
  %shl.i.i.i.i.i = shl nuw i32 %conv4.i.i.i.i.i, %22
  %25 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop, !amdgpu.uniform !42

26:                                               ; preds = %ComputeEnd
  %sunkaddr = getelementptr inbounds i8, ptr addrspace(1) %2977, i64 40
  %27 = atomicrmw or ptr addrspace(1) %sunkaddr, i32 %.lcssa4222 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %28, !amdgpu.uniform !42

28:                                               ; preds = %26, %ComputeEnd
  %29 = phi i32 [ %27, %26 ], [ poison, %ComputeEnd ]
  call void @llvm.amdgcn.end.cf.i64(i64 %2986)
  %30 = call i32 @llvm.amdgcn.readfirstlane.i32(i32 %29)
  %31 = or i32 %30, %.lcssa4223
  %shl5.i.i.i.i.i = shl nuw i32 1, %22
  %and.i.i.i.i.i = and i32 %shl5.i.i.i.i.i, %31
  %tobool3.i.i.i.i = icmp ne i32 %and.i.i.i.i.i, 0
  %32 = tail call i64 @llvm.amdgcn.ballot.i64(i1 %tobool3.i.i.i.i)
  %and.i.i.i.i.i.i = and i64 %32, %21
  %cmp.i.not.i.i.i = icmp ne i64 %21, %and.i.i.i.i.i.i
  br i1 %cmp.i.not.i.i.i, label %if.then.i.i.i.i2, label %Flow3902, !amdgpu.uniform !42

44:                                               ; preds = %ComputeEnd2429
  %sunkaddr3286 = getelementptr inbounds i8, ptr addrspace(1) %2977, i64 40
  %45 = atomicrmw and ptr addrspace(1) %sunkaddr3286, i32 %.lcssa4224 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %46, !amdgpu.uniform !42

46:                                               ; preds = %44, %ComputeEnd2429
  call void @llvm.amdgcn.end.cf.i64(i64 %3004)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %Flow3900, !amdgpu.uniform !42

if.end14.i.i.i:                                   ; preds = %Flow3900
  %47 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %21, i1 true)
  %iszero.i.i.i.i.i.i = icmp ne i64 %21, 0
  %cmp2.i.i.i.i.i = icmp eq i64 %47, %sh_prom.i.i.i.i
  %cmp.i.i.i.i.i = select i1 %iszero.i.i.i.i.i.i, i1 %cmp2.i.i.i.i.i, i1 false
  %48 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.i.i.i.i.i)
  %49 = extractvalue { i1, i64 } %48, 0
  %50 = extractvalue { i1, i64 } %48, 1
  br i1 %49, label %if.then16.i.i.i, label %if.end22.i.i.i

Flow3901:                                         ; preds = %if.end22.i.i.i, %Flow3900
  %52 = phi i32 [ %39, %if.end22.i.i.i ], [ %retval.sroa.8.0.i.i.i, %Flow3900 ]
  %53 = phi i64 [ %21, %if.end22.i.i.i ], [ %retval.sroa.2.0.i.i.i, %Flow3900 ]
  %54 = phi i32 [ %spec.store.select.i.i.i, %if.end22.i.i.i ], [ %retval.sroa.6.0.i.i.i, %Flow3900 ]
  br label %cleanup26.i.i.i, !amdgpu.uniform !42

if.end22.i.i.i:                                   ; preds = %if.then16.i.i.i, %if.end14.i.i.i
  call void @llvm.amdgcn.end.cf.i64(i64 %50)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %Flow3901, !amdgpu.uniform !42

Flow3902:                                         ; preds = %cleanup26.i.i.i, %28
  %55 = phi i32 [ %54, %cleanup26.i.i.i ], [ %17, %28 ]
  %56 = phi i64 [ %53, %cleanup26.i.i.i ], [ %18, %28 ]
  %57 = phi i32 [ %52, %cleanup26.i.i.i ], [ %19, %28 ]
  %58 = phi i32 [ %52, %cleanup26.i.i.i ], [ %retval.sroa.8.0.i.i.i, %28 ]
  %59 = phi i64 [ %53, %cleanup26.i.i.i ], [ %retval.sroa.2.0.i.i.i, %28 ]
  %60 = phi i32 [ %54, %cleanup26.i.i.i ], [ %retval.sroa.6.0.i.i.i, %28 ]
  %61 = phi i1 [ %cmp.i46.not.i.i.i, %cleanup26.i.i.i ], [ true, %28 ]
  br i1 %61, label %for.inc.i.i.i, label %Flow3903, !amdgpu.uniform !42

cleanup26.i.i.i:                                  ; preds = %Flow3901
  br label %Flow3902, !amdgpu.uniform !42

for.inc.i.i.i:                                    ; preds = %Flow3902
  %inc.i.i.i = add i32 %spec.store.select.i.i.i, 1
  br label %Flow3903, !amdgpu.uniform !42

Flow3903:                                         ; preds = %for.inc.i.i.i, %Flow3902
  %62 = phi i32 [ %inc.i.i.i, %for.inc.i.i.i ], [ poison, %Flow3902 ]
  %63 = phi i1 [ false, %for.inc.i.i.i ], [ true, %Flow3902 ]
  br i1 %63, label %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i, label %for.cond.i.i1.i, !amdgpu.uniform !42

_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i: ; preds = %Flow3903
  %.lcssa4227 = phi i32 [ %55, %Flow3903 ]
  %.lcssa4226 = phi i64 [ %56, %Flow3903 ]
  %.lcssa4225 = phi i32 [ %57, %Flow3903 ]
  fence acquire
  %64 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %mul.i.i.i.i.i.i = shl i32 %.lcssa4227, 6
  %idxprom.i19.i.i.i.i.i = zext i32 %mul.i.i.i.i.i.i to i64
  %arrayidx.i20.i.i.i.i.i = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %64, i64 %idxprom.i19.i.i.i.i.i
  %arrayidx.i22.i.i.i.i.i = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %arrayidx.i20.i.i.i.i.i, i64 %sh_prom.i.i.i.i
  store i64 sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.2 to ptr), i64 81) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.2 to ptr) to i64)), ptr %arrayidx.i22.i.i.i.i.i, align 8, !tbaa !82
  %spec.select.i.i.i.i.i.i.i = tail call i64 @llvm.umin.i64(i64 sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.2 to ptr), i64 81) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.2 to ptr) to i64)), i64 56)
  %65 = and i64 %spec.select.i.i.i.i.i.i.i, 15
  %66 = sub i64 %spec.select.i.i.i.i.i.i.i, %65
  %67 = icmp ne i64 %66, 0
  br i1 %67, label %loop-memcpy-expansion2321.preheader, label %Flow3899, !amdgpu.uniform !42

loop-memcpy-expansion2321.preheader:              ; preds = %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i
  %68 = shl nuw nsw i64 %sh_prom.i.i.i.i, 6
  %69 = add nuw nsw i64 %68, 8
  %scevgep3249 = getelementptr i8, ptr %64, i64 %69
  %70 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i, 6
  %scevgep3250 = getelementptr i8, ptr %scevgep3249, i64 %70
  br label %loop-memcpy-expansion2321, !amdgpu.uniform !42

Flow3899:                                         ; preds = %Flow3898, %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i
  br label %loop-memcpy-residual-header2324, !amdgpu.uniform !42

loop-memcpy-expansion2321:                        ; preds = %loop-memcpy-expansion2321.preheader, %loop-memcpy-expansion2321
  %loop-index2322 = phi i64 [ %73, %loop-memcpy-expansion2321 ], [ 0, %loop-memcpy-expansion2321.preheader ]
  %71 = getelementptr i8, ptr addrspace(4) @.str.2, i64 %loop-index2322, !amdgpu.uniform !42
  %72 = load <4 x i32>, ptr addrspace(4) %71, align 1
  %scevgep3251 = getelementptr i8, ptr %scevgep3250, i64 %loop-index2322
  store <4 x i32> %72, ptr %scevgep3251, align 1
  %73 = add i64 %loop-index2322, 16
  %74 = icmp uge i64 %73, %66
  br i1 %74, label %Flow3898, label %loop-memcpy-expansion2321, !amdgpu.uniform !42

Flow3897:                                         ; preds = %Flow3895, %Flow3896
  %75 = phi i64 [ %2896, %Flow3895 ], [ %2893, %Flow3896 ]
  %76 = phi i64 [ %2898, %Flow3895 ], [ %2894, %Flow3896 ]
  br label %post-loop-memcpy-expansion2320, !amdgpu.uniform !42

loop-memcpy-residual2323:                         ; preds = %loop-memcpy-residual2323.preheader, %loop-memcpy-residual2323
  %residual-loop-index2325 = phi i64 [ %78, %loop-memcpy-residual2323 ], [ 0, %loop-memcpy-residual2323.preheader ]
  %scevgep3248 = getelementptr i8, ptr addrspace(4) getelementptr inbounds nuw (i8, ptr addrspace(4) @.str.2, i64 48), i64 %residual-loop-index2325, !amdgpu.uniform !42
  %77 = load i8, ptr addrspace(4) %scevgep3248, align 1
  %scevgep3247 = getelementptr i8, ptr %scevgep3246, i64 %residual-loop-index2325
  store i8 %77, ptr %scevgep3247, align 1
  %78 = add i64 %residual-loop-index2325, 1
  %79 = icmp uge i64 %78, %65
  br i1 %79, label %Flow3895, label %loop-memcpy-residual2323, !amdgpu.uniform !42

Flow3895:                                         ; preds = %loop-memcpy-residual2323
  br label %Flow3897, !amdgpu.uniform !42

post-loop-memcpy-expansion2320:                   ; preds = %Flow3897
  %tobool.not.i.i.i.i.i.i = icmp eq i32 %.lcssa4225, 0
  %conv.i.i.i.i.i.i = zext i1 %tobool.not.i.i.i.i.i.i to i32
  fence release
  %80 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %idxprom.i23.i.i.i.i.i = zext i32 %.lcssa4227 to i64
  %arrayidx.i24.i.i.i.i.i = getelementptr inbounds nuw i32, ptr %80, i64 %idxprom.i23.i.i.i.i.i
  store atomic i32 %conv.i.i.i.i.i.i, ptr %arrayidx.i24.i.i.i.i.i monotonic, align 4
  %81 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !84
  %arrayidx.i.i5.i.i = getelementptr inbounds nuw %"struct.rpc::Header", ptr %81, i64 %idxprom.i23.i.i.i.i.i, !amdgpu.uniform !42
  %82 = load i64, ptr %arrayidx.i.i5.i.i, align 8, !tbaa !80
  br label %while.cond.i.i.i.i, !amdgpu.uniform !42

while.cond.i.i.i.i:                               ; preds = %post-loop-memcpy-expansion2320, %Flow3894
  %phi.broken = phi i64 [ 0, %post-loop-memcpy-expansion2320 ], [ %99, %Flow3894 ]
  %lsr.iv3238 = phi i64 [ %97, %Flow3894 ], [ 25, %post-loop-memcpy-expansion2320 ]
  %lsr.iv3233 = phi ptr addrspace(4) [ %96, %Flow3894 ], [ getelementptr inbounds nuw (i8, ptr addrspace(4) @.str.2, i64 56), %post-loop-memcpy-expansion2320 ]
  %port.sroa.37.0.i.i = phi i32 [ %95, %Flow3894 ], [ %conv.i.i.i.i.i.i, %post-loop-memcpy-expansion2320 ]
  %idx.0.i.i.i.i = phi i64 [ %94, %Flow3894 ], [ 56, %post-loop-memcpy-expansion2320 ]
  %cmp.i.i.i.i = icmp ult i64 %idx.0.i.i.i.i, sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.2 to ptr), i64 81) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.2 to ptr) to i64))
  %83 = tail call i64 @llvm.amdgcn.ballot.i64(i1 %cmp.i.i.i.i)
  %and.i.i.i.i6.i.i = and i64 %83, %82
  %tobool.not.i.i.i.i = icmp ne i64 %and.i.i.i.i6.i.i, 0
  %84 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i22.i.i.i = getelementptr inbounds nuw i32, ptr %84, i64 %idxprom.i23.i.i.i.i.i, !amdgpu.uniform !42
  %85 = load atomic i32, ptr %arrayidx.i22.i.i.i monotonic, align 4
  %86 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %tobool.not.i.i.i.i)
  %87 = extractvalue { i1, i64 } %86, 0
  %88 = extractvalue { i1, i64 } %86, 1
  br i1 %87, label %cond.false.i8.i.i.i.i, label %Flow3894

while.body.i.i21.i.i.i.i.preheader:               ; preds = %cond.false.i8.i.i.i.i
  br label %while.body.i.i21.i.i.i.i, !amdgpu.uniform !42

Flow3893:                                         ; preds = %Flow3892, %cond.false.i8.i.i.i.i
  br label %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i:                         ; preds = %while.body.i.i21.i.i.i.i.preheader, %while.body.i.i21.i.i.i.i
  tail call void @llvm.amdgcn.s.sleep(i32 2)
  %91 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i22.i.i.i.i = getelementptr inbounds nuw i32, ptr %91, i64 %idxprom.i23.i.i.i.i.i, !amdgpu.uniform !42
  %92 = load atomic i32, ptr %arrayidx.i.i.i22.i.i.i.i monotonic, align 4
  %93 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %92)
  %cmp.i.not.i.i23.i.i.i.i = icmp eq i32 %93, %port.sroa.37.0.i.i
  br i1 %cmp.i.not.i.i23.i.i.i.i, label %Flow3892, label %while.body.i.i21.i.i.i.i, !amdgpu.uniform !42

Flow3892:                                         ; preds = %while.body.i.i21.i.i.i.i
  br label %Flow3893, !amdgpu.uniform !42

Flow3894:                                         ; preds = %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i, %while.cond.i.i.i.i
  %94 = phi i64 [ %add.i.i.i.i, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i ], [ poison, %while.cond.i.i.i.i ]
  %95 = phi i32 [ %conv.i.i26.i.i.i.i, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i ], [ poison, %while.cond.i.i.i.i ]
  %96 = phi ptr addrspace(4) [ %scevgep3234, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i ], [ poison, %while.cond.i.i.i.i ]
  %97 = phi i64 [ %lsr.iv.next3239, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i ], [ poison, %while.cond.i.i.i.i ]
  %98 = phi i1 [ false, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i ], [ true, %while.cond.i.i.i.i ]
  call void @llvm.amdgcn.end.cf.i64(i64 %88)
  %99 = call i64 @llvm.amdgcn.if.break.i64(i1 %98, i64 %phi.broken)
  %100 = call i1 @llvm.amdgcn.loop.i64(i64 %99)
  br i1 %100, label %cond.false.i.i.i, label %while.cond.i.i.i.i

_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i: ; preds = %Flow3893
  fence acquire
  br i1 %cmp.i.i.i.i, label %if.then.i.i.i.i.i.i.i4, label %Flow3891, !amdgpu.uniform !42

loop-memcpy-expansion2327.preheader:              ; preds = %if.then.i.i.i.i.i.i.i4
  %arrayidx.i20.i35.i.i.i.i = getelementptr %"struct.rpc::Buffer", ptr %101, i64 %idxprom.i19.i.i.i.i.i
  %arrayidx.i22.i38.i.i.i.i = getelementptr %"struct.rpc::Buffer", ptr %arrayidx.i20.i35.i.i.i.i, i64 %sh_prom.i.i.i.i
  br label %loop-memcpy-expansion2327, !amdgpu.uniform !42

Flow3890:                                         ; preds = %Flow3889, %if.then.i.i.i.i.i.i.i4
  br label %loop-memcpy-residual-header2330, !amdgpu.uniform !42

loop-memcpy-expansion2327:                        ; preds = %loop-memcpy-expansion2327.preheader, %loop-memcpy-expansion2327
  %loop-index2328 = phi i64 [ %107, %loop-memcpy-expansion2327 ], [ 0, %loop-memcpy-expansion2327.preheader ]
  %scevgep3235 = getelementptr i8, ptr addrspace(4) %lsr.iv3233, i64 %loop-index2328, !amdgpu.uniform !42
  %105 = load <4 x i32>, ptr addrspace(4) %scevgep3235, align 1
  %106 = getelementptr i8, ptr %arrayidx.i22.i38.i.i.i.i, i64 %loop-index2328
  store <4 x i32> %105, ptr %106, align 1
  %107 = add i64 %loop-index2328, 16
  %108 = icmp uge i64 %107, %103
  br i1 %108, label %Flow3889, label %loop-memcpy-expansion2327, !amdgpu.uniform !42

Flow3888:                                         ; preds = %Flow3887, %loop-memcpy-residual-header2330
  br label %Flow3891, !amdgpu.uniform !42

loop-memcpy-residual2329:                         ; preds = %loop-memcpy-residual2329.preheader, %loop-memcpy-residual2329
  %residual-loop-index2331 = phi i64 [ %110, %loop-memcpy-residual2329 ], [ 0, %loop-memcpy-residual2329.preheader ]
  %scevgep3244 = getelementptr i8, ptr addrspace(4) %scevgep3243, i64 %residual-loop-index2331, !amdgpu.uniform !42
  %109 = load i8, ptr addrspace(4) %scevgep3244, align 1
  %scevgep3242 = getelementptr i8, ptr %scevgep3241, i64 %residual-loop-index2331
  store i8 %109, ptr %scevgep3242, align 1
  %110 = add i64 %residual-loop-index2331, 1
  %111 = icmp uge i64 %110, %102
  br i1 %111, label %Flow3887, label %loop-memcpy-residual2329, !amdgpu.uniform !42

Flow3887:                                         ; preds = %loop-memcpy-residual2329
  br label %Flow3888, !amdgpu.uniform !42

_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i: ; preds = %Flow3891
  %conv.i.i26.i.i.i.i = xor i32 %port.sroa.37.0.i.i, 1
  fence release
  %112 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %arrayidx.i24.i29.i.i.i.i = getelementptr inbounds nuw i32, ptr %112, i64 %idxprom.i23.i.i.i.i.i
  store atomic i32 %conv.i.i26.i.i.i.i, ptr %arrayidx.i24.i29.i.i.i.i monotonic, align 4
  %add.i.i.i.i = add i64 %idx.0.i.i.i.i, 64
  %scevgep3234 = getelementptr i8, ptr addrspace(4) %lsr.iv3233, i64 64
  %lsr.iv.next3239 = add i64 %lsr.iv3238, -64
  br label %Flow3894, !amdgpu.uniform !42

cond.false.i.i.i:                                 ; preds = %Flow3894
  %.lcssa4221 = phi i64 [ %99, %Flow3894 ]
  %port.sroa.37.0.i.i.lcssa = phi i32 [ %port.sroa.37.0.i.i, %Flow3894 ]
  %.lcssa4220 = phi i32 [ %85, %Flow3894 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa4221)
  %113 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %.lcssa4220)
  %114 = icmp ne i32 %113, %port.sroa.37.0.i.i.lcssa
  %115 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %114)
  %116 = extractvalue { i1, i64 } %115, 0
  %117 = extractvalue { i1, i64 } %115, 1
  br i1 %116, label %while.body.i.i.i.i.preheader, label %Flow3886

while.body.i.i.i.i.preheader:                     ; preds = %cond.false.i.i.i
  br label %while.body.i.i.i.i, !amdgpu.uniform !42

Flow3886:                                         ; preds = %Flow3885, %cond.false.i.i.i
  call void @llvm.amdgcn.end.cf.i64(i64 %117)
  br label %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit, !amdgpu.uniform !42

while.body.i.i.i.i:                               ; preds = %while.body.i.i.i.i.preheader, %while.body.i.i.i.i
  %phi.broken3910 = phi i64 [ 0, %while.body.i.i.i.i.preheader ], [ %121, %while.body.i.i.i.i ]
  tail call void @llvm.amdgcn.s.sleep(i32 2)
  %118 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i8.i.i = getelementptr inbounds nuw i32, ptr %118, i64 %idxprom.i23.i.i.i.i.i, !amdgpu.uniform !42
  %119 = load atomic i32, ptr %arrayidx.i.i.i8.i.i monotonic, align 4
  %120 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %119)
  %cmp.i.not.i.i.i.i = icmp eq i32 %120, %port.sroa.37.0.i.i.lcssa
  %121 = call i64 @llvm.amdgcn.if.break.i64(i1 %cmp.i.not.i.i.i.i, i64 %phi.broken3910)
  %122 = call i1 @llvm.amdgcn.loop.i64(i64 %121)
  br i1 %122, label %Flow3885, label %while.body.i.i.i.i

Flow3885:                                         ; preds = %while.body.i.i.i.i
  %.lcssa4219 = phi i64 [ %121, %while.body.i.i.i.i ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa4219)
  br label %Flow3886, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit: ; preds = %Flow3886
  fence acquire
  tail call void @llvm.amdgcn.wave.barrier()
  fence syncscope("agent") release
  %123 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %.lcssa4226, i1 true)
  %iszero.i.i.i.i.i13.i.i = icmp ne i64 %.lcssa4226, 0
  %cmp2.i.i.i.i14.i.i = icmp eq i64 %123, %sh_prom.i.i.i.i
  %cmp.i.i.i.i15.i.i = select i1 %iszero.i.i.i.i.i13.i.i, i1 %cmp2.i.i.i.i14.i.i, i1 false
  %124 = and i32 %.lcssa4227, 31
  %conv4.i.i.i19.i.i = zext i1 %cmp.i.i.i.i15.i.i to i32
  %shl.i.i.i20.i.i = shl nuw i32 %conv4.i.i.i19.i.i, %124
  %xor.i.i.i21.i.i = xor i32 %shl.i.i.i20.i.i, -1
  %125 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2432, !amdgpu.uniform !42

126:                                              ; preds = %ComputeEnd2433
  %div8.i.i.i16.i.i = lshr i32 %.lcssa4227, 5
  %idxprom.i.i.i17.i.i = zext nneg i32 %div8.i.i.i16.i.i to i64
  %127 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i17.i.i
  %sunkaddr3287 = getelementptr inbounds i8, ptr addrspace(1) %127, i64 40
  %128 = atomicrmw and ptr addrspace(1) %sunkaddr3287, i32 %.lcssa4218 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %129, !amdgpu.uniform !42

129:                                              ; preds = %126, %ComputeEnd2433
  call void @llvm.amdgcn.end.cf.i64(i64 %3021)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %for.cond.i.i1.i13, !amdgpu.uniform !42

for.cond.i.i1.i13:                                ; preds = %129, %Flow3884
  %130 = phi i32 [ %168, %Flow3884 ], [ poison, %129 ]
  %131 = phi i64 [ %169, %Flow3884 ], [ poison, %129 ]
  %132 = phi i32 [ %170, %Flow3884 ], [ poison, %129 ]
  %retval.sroa.6.0.i.i.i14 = phi i32 [ %173, %Flow3884 ], [ undef, %129 ]
  %retval.sroa.2.0.i.i.i16 = phi i64 [ %172, %Flow3884 ], [ undef, %129 ]
  %retval.sroa.8.0.i.i.i17 = phi i32 [ %171, %Flow3884 ], [ undef, %129 ]
  %index.0.i.i.i18 = phi i32 [ %175, %Flow3884 ], [ 0, %129 ]
  %133 = load i32, ptr addrspace(1) @__llvm_rpc_client, align 8, !tbaa !68
  %cmp.not.i.i.i19 = icmp ult i32 %index.0.i.i.i18, %133
  %spec.store.select.i.i.i20 = select i1 %cmp.not.i.i.i19, i32 %index.0.i.i.i18, i32 0
  %134 = tail call noundef i64 @llvm.amdgcn.ballot.i64(i1 true)
  %135 = and i32 %spec.store.select.i.i.i20, 31
  %136 = lshr i64 %134, %sh_prom.i.i.i.i
  %137 = trunc i64 %136 to i32
  %conv4.i.i.i.i.i24 = and i32 %137, 1
  %shl.i.i.i.i.i25 = shl nuw i32 %conv4.i.i.i.i.i24, %135
  %138 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2436, !amdgpu.uniform !42

139:                                              ; preds = %ComputeEnd2437
  %sunkaddr3288 = getelementptr inbounds i8, ptr addrspace(1) %3031, i64 40
  %140 = atomicrmw or ptr addrspace(1) %sunkaddr3288, i32 %.lcssa4212 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %141, !amdgpu.uniform !42

141:                                              ; preds = %139, %ComputeEnd2437
  %142 = phi i32 [ %140, %139 ], [ poison, %ComputeEnd2437 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %3040)
  %143 = call i32 @llvm.amdgcn.readfirstlane.i32(i32 %142)
  %144 = or i32 %143, %.lcssa4213
  %shl5.i.i.i.i.i26 = shl nuw i32 1, %135
  %and.i.i.i.i.i27 = and i32 %shl5.i.i.i.i.i26, %144
  %tobool3.i.i.i.i28 = icmp ne i32 %and.i.i.i.i.i27, 0
  %145 = tail call i64 @llvm.amdgcn.ballot.i64(i1 %tobool3.i.i.i.i28)
  %and.i.i.i.i.i.i29 = and i64 %145, %134
  %cmp.i.not.i.i.i30 = icmp ne i64 %134, %and.i.i.i.i.i.i29
  br i1 %cmp.i.not.i.i.i30, label %if.then.i.i.i.i31, label %Flow3883, !amdgpu.uniform !42

157:                                              ; preds = %ComputeEnd2442
  %sunkaddr3289 = getelementptr inbounds i8, ptr addrspace(1) %3031, i64 40
  %158 = atomicrmw and ptr addrspace(1) %sunkaddr3289, i32 %.lcssa4214 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %159, !amdgpu.uniform !42

159:                                              ; preds = %157, %ComputeEnd2442
  call void @llvm.amdgcn.end.cf.i64(i64 %3058)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %Flow3881, !amdgpu.uniform !42

if.end14.i.i.i101:                                ; preds = %Flow3881
  %160 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %134, i1 true)
  %iszero.i.i.i.i.i.i102 = icmp ne i64 %134, 0
  %cmp2.i.i.i.i.i103 = icmp eq i64 %160, %sh_prom.i.i.i.i
  %cmp.i.i.i.i.i104 = select i1 %iszero.i.i.i.i.i.i102, i1 %cmp2.i.i.i.i.i103, i1 false
  %161 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.i.i.i.i.i104)
  %162 = extractvalue { i1, i64 } %161, 0
  %163 = extractvalue { i1, i64 } %161, 1
  br i1 %162, label %if.then16.i.i.i106, label %if.end22.i.i.i105

Flow3882:                                         ; preds = %if.end22.i.i.i105, %Flow3881
  %165 = phi i32 [ %152, %if.end22.i.i.i105 ], [ %retval.sroa.8.0.i.i.i17, %Flow3881 ]
  %166 = phi i64 [ %134, %if.end22.i.i.i105 ], [ %retval.sroa.2.0.i.i.i16, %Flow3881 ]
  %167 = phi i32 [ %spec.store.select.i.i.i20, %if.end22.i.i.i105 ], [ %retval.sroa.6.0.i.i.i14, %Flow3881 ]
  br label %cleanup26.i.i.i43, !amdgpu.uniform !42

Flow3883:                                         ; preds = %cleanup26.i.i.i43, %141
  %168 = phi i32 [ %167, %cleanup26.i.i.i43 ], [ %130, %141 ]
  %169 = phi i64 [ %166, %cleanup26.i.i.i43 ], [ %131, %141 ]
  %170 = phi i32 [ %165, %cleanup26.i.i.i43 ], [ %132, %141 ]
  %171 = phi i32 [ %165, %cleanup26.i.i.i43 ], [ %retval.sroa.8.0.i.i.i17, %141 ]
  %172 = phi i64 [ %166, %cleanup26.i.i.i43 ], [ %retval.sroa.2.0.i.i.i16, %141 ]
  %173 = phi i32 [ %167, %cleanup26.i.i.i43 ], [ %retval.sroa.6.0.i.i.i14, %141 ]
  %174 = phi i1 [ %cmp.i46.not.i.i.i35, %cleanup26.i.i.i43 ], [ true, %141 ]
  br i1 %174, label %for.inc.i.i.i48, label %Flow3884, !amdgpu.uniform !42

cleanup26.i.i.i43:                                ; preds = %Flow3882
  br label %Flow3883, !amdgpu.uniform !42

for.inc.i.i.i48:                                  ; preds = %Flow3883
  %inc.i.i.i53 = add i32 %spec.store.select.i.i.i20, 1
  br label %Flow3884, !amdgpu.uniform !42

Flow3884:                                         ; preds = %for.inc.i.i.i48, %Flow3883
  %175 = phi i32 [ %inc.i.i.i53, %for.inc.i.i.i48 ], [ poison, %Flow3883 ]
  %176 = phi i1 [ false, %for.inc.i.i.i48 ], [ true, %Flow3883 ]
  br i1 %176, label %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i54, label %for.cond.i.i1.i13, !amdgpu.uniform !42

_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i54: ; preds = %Flow3884
  %.lcssa4217 = phi i32 [ %168, %Flow3884 ]
  %.lcssa4216 = phi i64 [ %169, %Flow3884 ]
  %.lcssa4215 = phi i32 [ %170, %Flow3884 ]
  fence acquire
  %177 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %mul.i.i.i.i.i.i55 = shl i32 %.lcssa4217, 6
  %idxprom.i19.i.i.i.i.i56 = zext i32 %mul.i.i.i.i.i.i55 to i64
  %arrayidx.i20.i.i.i.i.i57 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %177, i64 %idxprom.i19.i.i.i.i.i56
  %arrayidx.i22.i.i.i.i.i58 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %arrayidx.i20.i.i.i.i.i57, i64 %sh_prom.i.i.i.i
  store i64 sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.6.10 to ptr), i64 1) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.6.10 to ptr) to i64)), ptr %arrayidx.i22.i.i.i.i.i58, align 8, !tbaa !82
  %spec.select.i.i.i.i.i.i.i59 = tail call i64 @llvm.umin.i64(i64 sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.6.10 to ptr), i64 1) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.6.10 to ptr) to i64)), i64 56)
  %178 = and i64 %spec.select.i.i.i.i.i.i.i59, 15
  %179 = sub i64 %spec.select.i.i.i.i.i.i.i59, %178
  %180 = icmp ne i64 %179, 0
  br i1 %180, label %loop-memcpy-expansion2333.preheader, label %Flow3880, !amdgpu.uniform !42

loop-memcpy-expansion2333.preheader:              ; preds = %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i54
  %181 = add nuw nsw i64 %75, 8
  %scevgep3230 = getelementptr i8, ptr %177, i64 %181
  %182 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i56, 6
  %scevgep3231 = getelementptr i8, ptr %scevgep3230, i64 %182
  br label %loop-memcpy-expansion2333, !amdgpu.uniform !42

Flow3880:                                         ; preds = %Flow3879, %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i54
  br label %loop-memcpy-residual-header2336, !amdgpu.uniform !42

loop-memcpy-expansion2333:                        ; preds = %loop-memcpy-expansion2333.preheader, %loop-memcpy-expansion2333
  %loop-index2334 = phi i64 [ %185, %loop-memcpy-expansion2333 ], [ 0, %loop-memcpy-expansion2333.preheader ]
  %183 = getelementptr i8, ptr addrspace(4) @.str.6.10, i64 %loop-index2334, !amdgpu.uniform !42
  %184 = load <4 x i32>, ptr addrspace(4) %183, align 1
  %scevgep3232 = getelementptr i8, ptr %scevgep3231, i64 %loop-index2334
  store <4 x i32> %184, ptr %scevgep3232, align 1
  %185 = add i64 %loop-index2334, 16
  %186 = icmp uge i64 %185, %179
  br i1 %186, label %Flow3879, label %loop-memcpy-expansion2333, !amdgpu.uniform !42

Flow3878:                                         ; preds = %Flow3876, %Flow3877
  %187 = phi i64 [ %2906, %Flow3876 ], [ %2903, %Flow3877 ]
  br label %post-loop-memcpy-expansion2332, !amdgpu.uniform !42

loop-memcpy-residual2335:                         ; preds = %loop-memcpy-residual2335.preheader, %loop-memcpy-residual2335
  %residual-loop-index2337 = phi i64 [ %189, %loop-memcpy-residual2335 ], [ 0, %loop-memcpy-residual2335.preheader ]
  %scevgep3229 = getelementptr i8, ptr addrspace(4) @.str.6.10, i64 %residual-loop-index2337, !amdgpu.uniform !42
  %188 = load i8, ptr addrspace(4) %scevgep3229, align 1
  %scevgep3228 = getelementptr i8, ptr %scevgep3227, i64 %residual-loop-index2337
  store i8 %188, ptr %scevgep3228, align 1
  %189 = add i64 %residual-loop-index2337, 1
  %190 = icmp uge i64 %189, %178
  br i1 %190, label %Flow3876, label %loop-memcpy-residual2335, !amdgpu.uniform !42

Flow3876:                                         ; preds = %loop-memcpy-residual2335
  br label %Flow3878, !amdgpu.uniform !42

post-loop-memcpy-expansion2332:                   ; preds = %Flow3878
  %tobool.not.i.i.i.i.i.i61 = icmp eq i32 %.lcssa4215, 0
  %conv.i.i.i.i.i.i62 = zext i1 %tobool.not.i.i.i.i.i.i61 to i32
  fence release
  %191 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %idxprom.i23.i.i.i.i.i63 = zext i32 %.lcssa4217 to i64
  %arrayidx.i24.i.i.i.i.i64 = getelementptr inbounds nuw i32, ptr %191, i64 %idxprom.i23.i.i.i.i.i63
  store atomic i32 %conv.i.i.i.i.i.i62, ptr %arrayidx.i24.i.i.i.i.i64 monotonic, align 4
  %192 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !84
  %arrayidx.i.i5.i.i65 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %192, i64 %idxprom.i23.i.i.i.i.i63, !amdgpu.uniform !42
  %193 = load i64, ptr %arrayidx.i.i5.i.i65, align 8, !tbaa !80
  br label %while.cond.i.i.i.i66, !amdgpu.uniform !42

while.cond.i.i.i.i66:                             ; preds = %post-loop-memcpy-expansion2332, %Flow3875
  %phi.broken3911 = phi i64 [ 0, %post-loop-memcpy-expansion2332 ], [ %210, %Flow3875 ]
  %lsr.iv3219 = phi i64 [ %208, %Flow3875 ], [ -55, %post-loop-memcpy-expansion2332 ]
  %lsr.iv3214 = phi ptr addrspace(4) [ %207, %Flow3875 ], [ getelementptr (i8, ptr addrspace(4) @.str.6.10, i64 56), %post-loop-memcpy-expansion2332 ]
  %port.sroa.37.0.i.i67 = phi i32 [ %206, %Flow3875 ], [ %conv.i.i.i.i.i.i62, %post-loop-memcpy-expansion2332 ]
  %idx.0.i.i.i.i68 = phi i64 [ %205, %Flow3875 ], [ 56, %post-loop-memcpy-expansion2332 ]
  %cmp.i.i.i.i69 = icmp ult i64 %idx.0.i.i.i.i68, sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.6.10 to ptr), i64 1) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.6.10 to ptr) to i64))
  %194 = tail call i64 @llvm.amdgcn.ballot.i64(i1 %cmp.i.i.i.i69)
  %and.i.i.i.i6.i.i70 = and i64 %194, %193
  %tobool.not.i.i.i.i71 = icmp ne i64 %and.i.i.i.i6.i.i70, 0
  %195 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i22.i.i.i72 = getelementptr inbounds nuw i32, ptr %195, i64 %idxprom.i23.i.i.i.i.i63, !amdgpu.uniform !42
  %196 = load atomic i32, ptr %arrayidx.i22.i.i.i72 monotonic, align 4
  %197 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %tobool.not.i.i.i.i71)
  %198 = extractvalue { i1, i64 } %197, 0
  %199 = extractvalue { i1, i64 } %197, 1
  br i1 %198, label %cond.false.i8.i.i.i.i73, label %Flow3875

while.body.i.i21.i.i.i.i74.preheader:             ; preds = %cond.false.i8.i.i.i.i73
  br label %while.body.i.i21.i.i.i.i74, !amdgpu.uniform !42

Flow3874:                                         ; preds = %Flow3873, %cond.false.i8.i.i.i.i73
  br label %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i77, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i74:                       ; preds = %while.body.i.i21.i.i.i.i74.preheader, %while.body.i.i21.i.i.i.i74
  tail call void @llvm.amdgcn.s.sleep(i32 2)
  %202 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i22.i.i.i.i75 = getelementptr inbounds nuw i32, ptr %202, i64 %idxprom.i23.i.i.i.i.i63, !amdgpu.uniform !42
  %203 = load atomic i32, ptr %arrayidx.i.i.i22.i.i.i.i75 monotonic, align 4
  %204 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %203)
  %cmp.i.not.i.i23.i.i.i.i76 = icmp eq i32 %204, %port.sroa.37.0.i.i67
  br i1 %cmp.i.not.i.i23.i.i.i.i76, label %Flow3873, label %while.body.i.i21.i.i.i.i74, !amdgpu.uniform !42

Flow3873:                                         ; preds = %while.body.i.i21.i.i.i.i74
  br label %Flow3874, !amdgpu.uniform !42

Flow3875:                                         ; preds = %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i78, %while.cond.i.i.i.i66
  %205 = phi i64 [ %add.i.i.i.i81, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i78 ], [ poison, %while.cond.i.i.i.i66 ]
  %206 = phi i32 [ %conv.i.i26.i.i.i.i79, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i78 ], [ poison, %while.cond.i.i.i.i66 ]
  %207 = phi ptr addrspace(4) [ %scevgep3215, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i78 ], [ poison, %while.cond.i.i.i.i66 ]
  %208 = phi i64 [ %lsr.iv.next3220, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i78 ], [ poison, %while.cond.i.i.i.i66 ]
  %209 = phi i1 [ false, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i78 ], [ true, %while.cond.i.i.i.i66 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %199)
  %210 = call i64 @llvm.amdgcn.if.break.i64(i1 %209, i64 %phi.broken3911)
  %211 = call i1 @llvm.amdgcn.loop.i64(i64 %210)
  br i1 %211, label %cond.false.i.i.i88, label %while.cond.i.i.i.i66

_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i77: ; preds = %Flow3874
  fence acquire
  br i1 %cmp.i.i.i.i69, label %if.then.i.i.i.i.i.i.i82, label %Flow3872, !amdgpu.uniform !42

loop-memcpy-expansion2339.preheader:              ; preds = %if.then.i.i.i.i.i.i.i82
  %arrayidx.i20.i35.i.i.i.i86 = getelementptr %"struct.rpc::Buffer", ptr %212, i64 %idxprom.i19.i.i.i.i.i56
  %arrayidx.i22.i38.i.i.i.i87 = getelementptr %"struct.rpc::Buffer", ptr %arrayidx.i20.i35.i.i.i.i86, i64 %sh_prom.i.i.i.i
  br label %loop-memcpy-expansion2339, !amdgpu.uniform !42

Flow3871:                                         ; preds = %Flow3870, %if.then.i.i.i.i.i.i.i82
  br label %loop-memcpy-residual-header2342, !amdgpu.uniform !42

loop-memcpy-expansion2339:                        ; preds = %loop-memcpy-expansion2339.preheader, %loop-memcpy-expansion2339
  %loop-index2340 = phi i64 [ %218, %loop-memcpy-expansion2339 ], [ 0, %loop-memcpy-expansion2339.preheader ]
  %scevgep3216 = getelementptr i8, ptr addrspace(4) %lsr.iv3214, i64 %loop-index2340, !amdgpu.uniform !42
  %216 = load <4 x i32>, ptr addrspace(4) %scevgep3216, align 1
  %217 = getelementptr i8, ptr %arrayidx.i22.i38.i.i.i.i87, i64 %loop-index2340
  store <4 x i32> %216, ptr %217, align 1
  %218 = add i64 %loop-index2340, 16
  %219 = icmp uge i64 %218, %214
  br i1 %219, label %Flow3870, label %loop-memcpy-expansion2339, !amdgpu.uniform !42

Flow3869:                                         ; preds = %Flow3868, %loop-memcpy-residual-header2342
  br label %Flow3872, !amdgpu.uniform !42

loop-memcpy-residual2341:                         ; preds = %loop-memcpy-residual2341.preheader, %loop-memcpy-residual2341
  %residual-loop-index2343 = phi i64 [ %221, %loop-memcpy-residual2341 ], [ 0, %loop-memcpy-residual2341.preheader ]
  %scevgep3225 = getelementptr i8, ptr addrspace(4) %scevgep3224, i64 %residual-loop-index2343, !amdgpu.uniform !42
  %220 = load i8, ptr addrspace(4) %scevgep3225, align 1
  %scevgep3223 = getelementptr i8, ptr %scevgep3222, i64 %residual-loop-index2343
  store i8 %220, ptr %scevgep3223, align 1
  %221 = add i64 %residual-loop-index2343, 1
  %222 = icmp uge i64 %221, %213
  br i1 %222, label %Flow3868, label %loop-memcpy-residual2341, !amdgpu.uniform !42

Flow3868:                                         ; preds = %loop-memcpy-residual2341
  br label %Flow3869, !amdgpu.uniform !42

_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i78: ; preds = %Flow3872
  %conv.i.i26.i.i.i.i79 = xor i32 %port.sroa.37.0.i.i67, 1
  fence release
  %223 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %arrayidx.i24.i29.i.i.i.i80 = getelementptr inbounds nuw i32, ptr %223, i64 %idxprom.i23.i.i.i.i.i63
  store atomic i32 %conv.i.i26.i.i.i.i79, ptr %arrayidx.i24.i29.i.i.i.i80 monotonic, align 4
  %add.i.i.i.i81 = add i64 %idx.0.i.i.i.i68, 64
  %scevgep3215 = getelementptr i8, ptr addrspace(4) %lsr.iv3214, i64 64
  %lsr.iv.next3220 = add i64 %lsr.iv3219, -64
  br label %Flow3875, !amdgpu.uniform !42

cond.false.i.i.i88:                               ; preds = %Flow3875
  %.lcssa4211 = phi i64 [ %210, %Flow3875 ]
  %port.sroa.37.0.i.i67.lcssa = phi i32 [ %port.sroa.37.0.i.i67, %Flow3875 ]
  %.lcssa4210 = phi i32 [ %196, %Flow3875 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa4211)
  %224 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %.lcssa4210)
  %225 = icmp ne i32 %224, %port.sroa.37.0.i.i67.lcssa
  %226 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %225)
  %227 = extractvalue { i1, i64 } %226, 0
  %228 = extractvalue { i1, i64 } %226, 1
  br i1 %227, label %while.body.i.i.i.i89.preheader, label %Flow3867

while.body.i.i.i.i89.preheader:                   ; preds = %cond.false.i.i.i88
  br label %while.body.i.i.i.i89, !amdgpu.uniform !42

Flow3867:                                         ; preds = %Flow3866, %cond.false.i.i.i88
  call void @llvm.amdgcn.end.cf.i64(i64 %228)
  br label %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit109, !amdgpu.uniform !42

while.body.i.i.i.i89:                             ; preds = %while.body.i.i.i.i89.preheader, %while.body.i.i.i.i89
  %phi.broken3912 = phi i64 [ 0, %while.body.i.i.i.i89.preheader ], [ %232, %while.body.i.i.i.i89 ]
  tail call void @llvm.amdgcn.s.sleep(i32 2)
  %229 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i8.i.i90 = getelementptr inbounds nuw i32, ptr %229, i64 %idxprom.i23.i.i.i.i.i63, !amdgpu.uniform !42
  %230 = load atomic i32, ptr %arrayidx.i.i.i8.i.i90 monotonic, align 4
  %231 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %230)
  %cmp.i.not.i.i.i.i91 = icmp eq i32 %231, %port.sroa.37.0.i.i67.lcssa
  %232 = call i64 @llvm.amdgcn.if.break.i64(i1 %cmp.i.not.i.i.i.i91, i64 %phi.broken3912)
  %233 = call i1 @llvm.amdgcn.loop.i64(i64 %232)
  br i1 %233, label %Flow3866, label %while.body.i.i.i.i89

Flow3866:                                         ; preds = %while.body.i.i.i.i89
  %.lcssa4209 = phi i64 [ %232, %while.body.i.i.i.i89 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa4209)
  br label %Flow3867, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit109: ; preds = %Flow3867
  fence acquire
  tail call void @llvm.amdgcn.wave.barrier()
  fence syncscope("agent") release
  %234 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %.lcssa4216, i1 true)
  %iszero.i.i.i.i.i13.i.i92 = icmp ne i64 %.lcssa4216, 0
  %cmp2.i.i.i.i14.i.i93 = icmp eq i64 %234, %sh_prom.i.i.i.i
  %cmp.i.i.i.i15.i.i94 = select i1 %iszero.i.i.i.i.i13.i.i92, i1 %cmp2.i.i.i.i14.i.i93, i1 false
  %235 = and i32 %.lcssa4217, 31
  %conv4.i.i.i19.i.i98 = zext i1 %cmp.i.i.i.i15.i.i94 to i32
  %shl.i.i.i20.i.i99 = shl nuw i32 %conv4.i.i.i19.i.i98, %235
  %xor.i.i.i21.i.i100 = xor i32 %shl.i.i.i20.i.i99, -1
  %236 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2445, !amdgpu.uniform !42

237:                                              ; preds = %ComputeEnd2446
  %div8.i.i.i16.i.i95 = lshr i32 %.lcssa4217, 5
  %idxprom.i.i.i17.i.i96 = zext nneg i32 %div8.i.i.i16.i.i95 to i64
  %238 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i17.i.i96
  %sunkaddr3290 = getelementptr inbounds i8, ptr addrspace(1) %238, i64 40
  %239 = atomicrmw and ptr addrspace(1) %sunkaddr3290, i32 %.lcssa4208 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %240, !amdgpu.uniform !42

240:                                              ; preds = %237, %ComputeEnd2446
  call void @llvm.amdgcn.end.cf.i64(i64 %3075)
  tail call void @llvm.amdgcn.wave.barrier()
  call void @llvm.lifetime.start.p5(ptr addrspace(5) %buffer.i.i.i) #34, !noalias !109
  br label %while.body.i.i.i.i.i.i, !amdgpu.uniform !42

while.body.i.i.i.i.i.i:                           ; preds = %240, %_ZN22__llvm_libc_22_0_0_git7details22StringBufferWriterImplILb0EE4pushEc.exit.i.i.i.i.i.i
  %phi.broken3913 = phi i64 [ 0, %240 ], [ %308, %_ZN22__llvm_libc_22_0_0_git7details22StringBufferWriterImplILb0EE4pushEc.exit.i.i.i.i.i.i ]
  %writer.sroa.10.0.i.i.i.i = phi i64 [ %writer.sroa.10.1.i.i.i.i, %_ZN22__llvm_libc_22_0_0_git7details22StringBufferWriterImplILb0EE4pushEc.exit.i.i.i.i.i.i ], [ 0, %240 ]
  %value.addr.09.i.i.i.i.i.i = phi i32 [ %div.i.i.i.i.i.i.i, %_ZN22__llvm_libc_22_0_0_git7details22StringBufferWriterImplILb0EE4pushEc.exit.i.i.i.i.i.i ], [ %Loc.coerce1, %240 ]
  br label %NodeBlock3364, !amdgpu.uniform !42

NodeBlock3364:                                    ; preds = %while.body.i.i.i.i.i.i
  %div.i.i.i.i.i.i.i = udiv i32 %value.addr.09.i.i.i.i.i.i, 10
  %.neg.i = call i32 @llvm.amdgcn.mul.u24.i32(i32 %div.i.i.i.i.i.i.i, i32 246)
  %rem.i.i.i.i.i.i.decomposed.i = add nuw nsw i32 %.neg.i, %value.addr.09.i.i.i.i.i.i
  %conv.i.i.i.i.i.i.i = trunc i32 %rem.i.i.i.i.i.i.decomposed.i to i8
  %241 = zext i8 %conv.i.i.i.i.i.i.i to i16
  %Pivot3365 = icmp sge i16 %241, 5
  %242 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %Pivot3365)
  %243 = extractvalue { i1, i64 } %242, 0
  %244 = extractvalue { i1, i64 } %242, 1
  br i1 %243, label %NodeBlock3362, label %Flow3865

NodeBlock3362:                                    ; preds = %NodeBlock3364
  %Pivot3363 = icmp sge i16 %241, 7
  %245 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %Pivot3363)
  %246 = extractvalue { i1, i64 } %245, 0
  %247 = extractvalue { i1, i64 } %245, 1
  br i1 %246, label %NodeBlock3360, label %Flow3856

NodeBlock3360:                                    ; preds = %NodeBlock3362
  %Pivot3361 = icmp sge i16 %241, 8
  %248 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %Pivot3361)
  %249 = extractvalue { i1, i64 } %248, 0
  %250 = extractvalue { i1, i64 } %248, 1
  br i1 %249, label %NodeBlock3358, label %Flow3852

NodeBlock3358:                                    ; preds = %NodeBlock3360
  %Pivot3359 = icmp sge i16 %241, 9
  %251 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %Pivot3359)
  %252 = extractvalue { i1, i64 } %251, 0
  %253 = extractvalue { i1, i64 } %251, 1
  br i1 %252, label %sw.bb9.i.i.i.i.i.i.i.i, label %Flow3850

Flow3856:                                         ; preds = %Flow3853, %NodeBlock3362
  %254 = phi i8 [ %298, %Flow3853 ], [ poison, %NodeBlock3362 ]
  %255 = call { i1, i64 } @llvm.amdgcn.else.i64.i64(i64 %247)
  %256 = extractvalue { i1, i64 } %255, 0
  %257 = extractvalue { i1, i64 } %255, 1
  br i1 %256, label %NodeBlock3356, label %Flow3857

NodeBlock3356:                                    ; preds = %Flow3856
  %Pivot3357 = icmp sge i16 %241, 6
  %258 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %Pivot3357)
  %259 = extractvalue { i1, i64 } %258, 0
  %260 = extractvalue { i1, i64 } %258, 1
  br i1 %259, label %sw.bb6.i.i.i.i.i.i.i.i, label %Flow3854

Flow3865:                                         ; preds = %Flow3857, %NodeBlock3364
  %261 = phi i8 [ %300, %Flow3857 ], [ poison, %NodeBlock3364 ]
  %262 = call { i1, i64 } @llvm.amdgcn.else.i64.i64(i64 %244)
  %263 = extractvalue { i1, i64 } %262, 0
  %264 = extractvalue { i1, i64 } %262, 1
  br i1 %263, label %NodeBlock3354, label %if.then.i.i.i.i.i.i.i110

NodeBlock3354:                                    ; preds = %Flow3865
  %Pivot3355 = icmp sge i16 %241, 3
  %265 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %Pivot3355)
  %266 = extractvalue { i1, i64 } %265, 0
  %267 = extractvalue { i1, i64 } %265, 1
  br i1 %266, label %NodeBlock3352, label %Flow3863

NodeBlock3352:                                    ; preds = %NodeBlock3354
  %Pivot3353 = icmp sge i16 %241, 4
  %268 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %Pivot3353)
  %269 = extractvalue { i1, i64 } %268, 0
  %270 = extractvalue { i1, i64 } %268, 1
  br i1 %269, label %sw.bb4.i.i.i.i.i.i.i.i, label %Flow3858

Flow3863:                                         ; preds = %Flow3859, %NodeBlock3354
  %271 = phi i8 [ %301, %Flow3859 ], [ poison, %NodeBlock3354 ]
  %272 = call { i1, i64 } @llvm.amdgcn.else.i64.i64(i64 %267)
  %273 = extractvalue { i1, i64 } %272, 0
  %274 = extractvalue { i1, i64 } %272, 1
  br i1 %273, label %NodeBlock, label %Flow3864

NodeBlock:                                        ; preds = %Flow3863
  %Pivot = icmp sge i16 %241, 2
  %275 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %Pivot)
  %276 = extractvalue { i1, i64 } %275, 0
  %277 = extractvalue { i1, i64 } %275, 1
  br i1 %276, label %sw.bb2.i.i.i.i.i.i.i.i, label %Flow3861

Flow3861:                                         ; preds = %sw.bb2.i.i.i.i.i.i.i.i, %NodeBlock
  %278 = call { i1, i64 } @llvm.amdgcn.else.i64.i64(i64 %277)
  %279 = extractvalue { i1, i64 } %278, 0
  %280 = extractvalue { i1, i64 } %278, 1
  br i1 %279, label %LeafBlock, label %Flow3862

LeafBlock:                                        ; preds = %Flow3861
  %SwitchLeaf = icmp eq i16 %241, 1
  %281 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %SwitchLeaf)
  %282 = extractvalue { i1, i64 } %281, 0
  %283 = extractvalue { i1, i64 } %281, 1
  br i1 %282, label %sw.bb1.i.i.i.i.i.i.i.i, label %Flow3860

sw.bb1.i.i.i.i.i.i.i.i:                           ; preds = %LeafBlock
  br label %Flow3860, !amdgpu.uniform !42

sw.bb2.i.i.i.i.i.i.i.i:                           ; preds = %NodeBlock
  br label %Flow3861, !amdgpu.uniform !42

Flow3858:                                         ; preds = %sw.bb4.i.i.i.i.i.i.i.i, %NodeBlock3352
  %284 = call { i1, i64 } @llvm.amdgcn.else.i64.i64(i64 %270)
  %285 = extractvalue { i1, i64 } %284, 0
  %286 = extractvalue { i1, i64 } %284, 1
  br i1 %285, label %sw.bb3.i.i.i.i.i.i.i.i, label %Flow3859

sw.bb3.i.i.i.i.i.i.i.i:                           ; preds = %Flow3858
  br label %Flow3859, !amdgpu.uniform !42

sw.bb4.i.i.i.i.i.i.i.i:                           ; preds = %NodeBlock3352
  br label %Flow3858, !amdgpu.uniform !42

Flow3854:                                         ; preds = %sw.bb6.i.i.i.i.i.i.i.i, %NodeBlock3356
  %287 = call { i1, i64 } @llvm.amdgcn.else.i64.i64(i64 %260)
  %288 = extractvalue { i1, i64 } %287, 0
  %289 = extractvalue { i1, i64 } %287, 1
  br i1 %288, label %sw.bb5.i.i.i.i.i.i.i.i, label %Flow3855

sw.bb5.i.i.i.i.i.i.i.i:                           ; preds = %Flow3854
  br label %Flow3855, !amdgpu.uniform !42

sw.bb6.i.i.i.i.i.i.i.i:                           ; preds = %NodeBlock3356
  br label %Flow3854, !amdgpu.uniform !42

Flow3852:                                         ; preds = %Flow3851, %NodeBlock3360
  %290 = phi i8 [ %297, %Flow3851 ], [ poison, %NodeBlock3360 ]
  %291 = call { i1, i64 } @llvm.amdgcn.else.i64.i64(i64 %250)
  %292 = extractvalue { i1, i64 } %291, 0
  %293 = extractvalue { i1, i64 } %291, 1
  br i1 %292, label %sw.bb7.i.i.i.i.i.i.i.i, label %Flow3853

sw.bb7.i.i.i.i.i.i.i.i:                           ; preds = %Flow3852
  br label %Flow3853, !amdgpu.uniform !42

Flow3850:                                         ; preds = %sw.bb9.i.i.i.i.i.i.i.i, %NodeBlock3358
  %294 = call { i1, i64 } @llvm.amdgcn.else.i64.i64(i64 %253)
  %295 = extractvalue { i1, i64 } %294, 0
  %296 = extractvalue { i1, i64 } %294, 1
  br i1 %295, label %sw.bb8.i.i.i.i.i.i.i.i, label %Flow3851

sw.bb8.i.i.i.i.i.i.i.i:                           ; preds = %Flow3850
  br label %Flow3851, !amdgpu.uniform !42

sw.bb9.i.i.i.i.i.i.i.i:                           ; preds = %NodeBlock3358
  br label %Flow3850, !amdgpu.uniform !42

Flow3851:                                         ; preds = %sw.bb8.i.i.i.i.i.i.i.i, %Flow3850
  %297 = phi i8 [ 56, %sw.bb8.i.i.i.i.i.i.i.i ], [ 57, %Flow3850 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %296)
  br label %Flow3852, !amdgpu.uniform !42

Flow3853:                                         ; preds = %sw.bb7.i.i.i.i.i.i.i.i, %Flow3852
  %298 = phi i8 [ 55, %sw.bb7.i.i.i.i.i.i.i.i ], [ %290, %Flow3852 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %293)
  br label %Flow3856, !amdgpu.uniform !42

Flow3855:                                         ; preds = %sw.bb5.i.i.i.i.i.i.i.i, %Flow3854
  %299 = phi i8 [ 53, %sw.bb5.i.i.i.i.i.i.i.i ], [ 54, %Flow3854 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %289)
  br label %Flow3857, !amdgpu.uniform !42

Flow3857:                                         ; preds = %Flow3855, %Flow3856
  %300 = phi i8 [ %299, %Flow3855 ], [ %254, %Flow3856 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %257)
  br label %Flow3865, !amdgpu.uniform !42

Flow3859:                                         ; preds = %sw.bb3.i.i.i.i.i.i.i.i, %Flow3858
  %301 = phi i8 [ 51, %sw.bb3.i.i.i.i.i.i.i.i ], [ 52, %Flow3858 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %286)
  br label %Flow3863, !amdgpu.uniform !42

Flow3860:                                         ; preds = %sw.bb1.i.i.i.i.i.i.i.i, %LeafBlock
  %302 = phi i8 [ 49, %sw.bb1.i.i.i.i.i.i.i.i ], [ 48, %LeafBlock ]
  call void @llvm.amdgcn.end.cf.i64(i64 %283)
  br label %Flow3862, !amdgpu.uniform !42

Flow3862:                                         ; preds = %Flow3860, %Flow3861
  %303 = phi i8 [ %302, %Flow3860 ], [ 50, %Flow3861 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %280)
  br label %Flow3864, !amdgpu.uniform !42

Flow3864:                                         ; preds = %Flow3862, %Flow3863
  %304 = phi i8 [ %303, %Flow3862 ], [ %271, %Flow3863 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %274)
  br label %if.then.i.i.i.i.i.i.i110, !amdgpu.uniform !42

if.then.i.i.i.i.i.i.i110:                         ; preds = %Flow3865, %Flow3864
  %retval.0.i.i.i.i.i.i.i.i = phi i8 [ %261, %Flow3865 ], [ %304, %Flow3864 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %264)
  %cmp.i.i.i.i.i.i.i.i = icmp eq i64 %writer.sroa.10.0.i.i.i.i, 11
  %cmp.i.i.i.i.i.i.i.i.inv = xor i1 %cmp.i.i.i.i.i.i.i.i, true
  br i1 %cmp.i.i.i.i.i.i.i.i.inv, label %if.then3.i.i.i.i.i.i.i, label %_ZN22__llvm_libc_22_0_0_git7details22StringBufferWriterImplILb0EE4pushEc.exit.i.i.i.i.i.i, !amdgpu.uniform !42

if.then3.i.i.i.i.i.i.i:                           ; preds = %if.then.i.i.i.i.i.i.i110
  %305 = trunc i64 %writer.sroa.10.0.i.i.i.i to i32
  %306 = xor i32 %305, -1
  %307 = getelementptr i8, ptr addrspace(5) %buffer.i.i.i, i32 %306
  %arrayidx.i.i.i.i.i.i.i.i1112700 = getelementptr i8, ptr addrspace(5) %307, i32 11
  %inc.i.i.i.i.i.i.i = add i64 %writer.sroa.10.0.i.i.i.i, 1
  store i8 %retval.0.i.i.i.i.i.i.i.i, ptr addrspace(5) %arrayidx.i.i.i.i.i.i.i.i1112700, align 1, !tbaa !57, !noalias !109
  br label %_ZN22__llvm_libc_22_0_0_git7details22StringBufferWriterImplILb0EE4pushEc.exit.i.i.i.i.i.i, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git7details22StringBufferWriterImplILb0EE4pushEc.exit.i.i.i.i.i.i: ; preds = %if.then3.i.i.i.i.i.i.i, %if.then.i.i.i.i.i.i.i110
  %writer.sroa.10.1.i.i.i.i = phi i64 [ %inc.i.i.i.i.i.i.i, %if.then3.i.i.i.i.i.i.i ], [ 11, %if.then.i.i.i.i.i.i.i110 ]
  %cmp.i.i.i.i.i.i112 = icmp samesign ult i32 %value.addr.09.i.i.i.i.i.i, 10
  %.not.i.i.i.i.i.i = or i1 %cmp.i.i.i.i.i.i.i.i, %cmp.i.i.i.i.i.i112
  %308 = call i64 @llvm.amdgcn.if.break.i64(i1 %.not.i.i.i.i.i.i, i64 %phi.broken3913)
  %309 = call i1 @llvm.amdgcn.loop.i64(i64 %308)
  br i1 %309, label %_ZN22__llvm_libc_22_0_0_git15IntegerToStringIiNS_7details3FmtILh10ELb0ELb0ELb0ELm1EEEE13IntegerWriter25write_unsigned_number_decEjRNS1_22StringBufferWriterImplILb0EEE.exit.i.i.i.i.i, label %while.body.i.i.i.i.i.i

_ZN22__llvm_libc_22_0_0_git15IntegerToStringIiNS_7details3FmtILh10ELb0ELb0ELb0ELm1EEEE13IntegerWriter25write_unsigned_number_decEjRNS1_22StringBufferWriterImplILb0EEE.exit.i.i.i.i.i: ; preds = %_ZN22__llvm_libc_22_0_0_git7details22StringBufferWriterImplILb0EE4pushEc.exit.i.i.i.i.i.i
  %writer.sroa.10.1.i.i.i.i.lcssa = phi i64 [ %writer.sroa.10.1.i.i.i.i, %_ZN22__llvm_libc_22_0_0_git7details22StringBufferWriterImplILb0EE4pushEc.exit.i.i.i.i.i.i ]
  %.lcssa4207 = phi i64 [ %308, %_ZN22__llvm_libc_22_0_0_git7details22StringBufferWriterImplILb0EE4pushEc.exit.i.i.i.i.i.i ]
  %cmp.i.i.i.i.i.i.i.i.lcssa = phi i1 [ %cmp.i.i.i.i.i.i.i.i, %_ZN22__llvm_libc_22_0_0_git7details22StringBufferWriterImplILb0EE4pushEc.exit.i.i.i.i.i.i ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa4207)
  %cmp.i19.i.i.i.i = icmp ne i64 %writer.sroa.10.1.i.i.i.i.lcssa, 0
  %or.cond.not.i.i.i.i = select i1 %cmp.i.i.i.i.i.i.i.i.lcssa, i1 true, i1 %cmp.i19.i.i.i.i
  %or.cond.not.i.i.i.i.inv = xor i1 %or.cond.not.i.i.i.i, true
  %310 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %or.cond.not.i.i.i.i.inv)
  %311 = extractvalue { i1, i64 } %310, 0
  %312 = extractvalue { i1, i64 } %310, 1
  br i1 %311, label %if.then3.i.i.lr.ph.i.i.i.i, label %while.end.i.i.i.i.i

if.then3.i.i.lr.ph.i.i.i.i:                       ; preds = %_ZN22__llvm_libc_22_0_0_git15IntegerToStringIiNS_7details3FmtILh10ELb0ELb0ELb0ELm1EEEE13IntegerWriter25write_unsigned_number_decEjRNS1_22StringBufferWriterImplILb0EEE.exit.i.i.i.i.i
  %arrayidx.i.i.i.i.i.i.i = getelementptr inbounds nuw i8, ptr addrspace(5) %buffer.i.i.i, i32 10
  store i8 48, ptr addrspace(5) %arrayidx.i.i.i.i.i.i.i, align 2, !tbaa !57, !noalias !109
  br label %while.end.i.i.i.i.i, !amdgpu.uniform !42

while.end.i.i.i.i.i:                              ; preds = %if.then3.i.i.lr.ph.i.i.i.i, %_ZN22__llvm_libc_22_0_0_git15IntegerToStringIiNS_7details3FmtILh10ELb0ELb0ELb0ELm1EEEE13IntegerWriter25write_unsigned_number_decEjRNS1_22StringBufferWriterImplILb0EEE.exit.i.i.i.i.i
  %writer.sroa.10.4.i.i.i.i = phi i64 [ 1, %if.then3.i.i.lr.ph.i.i.i.i ], [ %writer.sroa.10.1.i.i.i.i.lcssa, %_ZN22__llvm_libc_22_0_0_git15IntegerToStringIiNS_7details3FmtILh10ELb0ELb0ELb0ELm1EEEE13IntegerWriter25write_unsigned_number_decEjRNS1_22StringBufferWriterImplILb0EEE.exit.i.i.i.i.i ]
  call void @llvm.amdgcn.end.cf.i64(i64 %312)
  %written.i.i.i.i = getelementptr inbounds nuw i8, ptr addrspace(5) %buffer.i.i.i, i32 16
  store i64 %writer.sroa.10.4.i.i.i.i, ptr addrspace(5) %written.i.i.i.i, align 8, !tbaa !91, !noalias !109
  %cmp.i.not.i.i.i.i.i = icmp ne i64 %writer.sroa.10.4.i.i.i.i, 0
  %313 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.i.not.i.i.i.i.i)
  %314 = extractvalue { i1, i64 } %313, 0
  %315 = extractvalue { i1, i64 } %313, 1
  br i1 %314, label %if.then.i.i.i.i.i.i, label %Flow3849

Flow3848:                                         ; preds = %if.end4.i.i.i.i.i.i.i, %if.then.i.i.i.i.i.i
  %323 = phi ptr [ %cond.i.i.i, %if.end4.i.i.i.i.i.i.i ], [ addrspacecast (ptr addrspace(4) @_ZN22__llvm_libc_22_0_0_git3cpp6string14NULL_CHARACTERE to ptr), %if.then.i.i.i.i.i.i ]
  call void @llvm.amdgcn.end.cf.i64(i64 %319)
  br label %_ZN22__llvm_libc_22_0_0_git3cpp6string7reserveEm.exit.i.i.i.i.i.i, !amdgpu.uniform !42

Flow3849:                                         ; preds = %Flow3847, %while.end.i.i.i.i.i
  %325 = phi ptr [ %323, %Flow3847 ], [ addrspacecast (ptr addrspace(4) @_ZN22__llvm_libc_22_0_0_git3cpp6string14NULL_CHARACTERE to ptr), %while.end.i.i.i.i.i ]
  call void @llvm.amdgcn.end.cf.i64(i64 %315)
  br label %if.end.i.i.i.i.i.i, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git3cpp6string7reserveEm.exit.i.i.i.i.i.i: ; preds = %Flow3848
  br label %loadstoreloop, !amdgpu.uniform !42

loadstoreloop:                                    ; preds = %_ZN22__llvm_libc_22_0_0_git3cpp6string7reserveEm.exit.i.i.i.i.i.i, %loadstoreloop
  %phi.broken3914 = phi i64 [ 0, %_ZN22__llvm_libc_22_0_0_git3cpp6string7reserveEm.exit.i.i.i.i.i.i ], [ %330, %loadstoreloop ]
  %326 = phi i64 [ %328, %loadstoreloop ], [ 0, %_ZN22__llvm_libc_22_0_0_git3cpp6string7reserveEm.exit.i.i.i.i.i.i ]
  %327 = getelementptr i8, ptr %323, i64 %326
  store i8 0, ptr %327, align 1
  %328 = add i64 %326, 1
  %329 = icmp uge i64 %328, %writer.sroa.10.4.i.i.i.i
  %330 = call i64 @llvm.amdgcn.if.break.i64(i1 %329, i64 %phi.broken3914)
  %331 = call i1 @llvm.amdgcn.loop.i64(i64 %330)
  br i1 %331, label %Flow3847, label %loadstoreloop

Flow3847:                                         ; preds = %loadstoreloop
  %.lcssa4206 = phi i64 [ %330, %loadstoreloop ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa4206)
  br label %Flow3849, !amdgpu.uniform !42

if.end.i.i.i.i.i.i:                               ; preds = %Flow3849
  %cmp.not.i6.i.i.i.i.i.i = icmp ne ptr %325, addrspacecast (ptr addrspace(4) @_ZN22__llvm_libc_22_0_0_git3cpp6string14NULL_CHARACTERE to ptr)
  %332 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.not.i6.i.i.i.i.i.i)
  %333 = extractvalue { i1, i64 } %332, 0
  %334 = extractvalue { i1, i64 } %332, 1
  br i1 %333, label %if.then.i.i.i.i2.i.i.i, label %_ZN22__llvm_libc_22_0_0_git3cpp9to_stringEi.exit.i

if.then.i.i.i.i2.i.i.i:                           ; preds = %if.end.i.i.i.i.i.i
  %arrayidx.i.i.i.i3.i.i.i = getelementptr inbounds nuw i8, ptr %325, i64 %writer.sroa.10.4.i.i.i.i
  store i8 0, ptr %arrayidx.i.i.i.i3.i.i.i, align 1, !tbaa !57, !noalias !109
  br label %_ZN22__llvm_libc_22_0_0_git3cpp9to_stringEi.exit.i, !amdgpu.uniform !42

loop-memcpy-expansion2303.preheader:              ; preds = %_ZN22__llvm_libc_22_0_0_git3cpp9to_stringEi.exit.i
  %342 = sub i32 11, %335
  %scevgep3211 = getelementptr i8, ptr addrspace(5) %buffer.i.i.i, i32 %342
  br label %loop-memcpy-expansion2303, !amdgpu.uniform !42

Flow3846:                                         ; preds = %Flow3845, %_ZN22__llvm_libc_22_0_0_git3cpp9to_stringEi.exit.i
  call void @llvm.amdgcn.end.cf.i64(i64 %341)
  br label %loop-memcpy-residual-header2306, !amdgpu.uniform !42

loop-memcpy-expansion2303:                        ; preds = %loop-memcpy-expansion2303.preheader, %loop-memcpy-expansion2303
  %phi.broken3915 = phi i64 [ 0, %loop-memcpy-expansion2303.preheader ], [ %347, %loop-memcpy-expansion2303 ]
  %lsr.iv3212 = phi ptr addrspace(5) [ %scevgep3213, %loop-memcpy-expansion2303 ], [ %scevgep3211, %loop-memcpy-expansion2303.preheader ]
  %loop-index2304 = phi i64 [ %345, %loop-memcpy-expansion2303 ], [ 0, %loop-memcpy-expansion2303.preheader ]
  %343 = load <4 x i32>, ptr addrspace(5) %lsr.iv3212, align 1
  %344 = getelementptr i8, ptr %325, i64 %loop-index2304
  store <4 x i32> %343, ptr %344, align 1
  %345 = add i64 %loop-index2304, 16
  %scevgep3213 = getelementptr i8, ptr addrspace(5) %lsr.iv3212, i32 16
  %346 = icmp uge i64 %345, %337
  %347 = call i64 @llvm.amdgcn.if.break.i64(i1 %346, i64 %phi.broken3915)
  %348 = call i1 @llvm.amdgcn.loop.i64(i64 %347)
  br i1 %348, label %Flow3845, label %loop-memcpy-expansion2303

Flow3844:                                         ; preds = %Flow3843, %loop-memcpy-residual-header2306
  call void @llvm.amdgcn.end.cf.i64(i64 %2865)
  br label %post-loop-memcpy-expansion2302, !amdgpu.uniform !42

loop-memcpy-residual2305:                         ; preds = %loop-memcpy-residual2305.preheader, %loop-memcpy-residual2305
  %phi.broken3916 = phi i64 [ 0, %loop-memcpy-residual2305.preheader ], [ %352, %loop-memcpy-residual2305 ]
  %lsr.iv3207 = phi ptr addrspace(5) [ %scevgep3208, %loop-memcpy-residual2305 ], [ %scevgep3206, %loop-memcpy-residual2305.preheader ]
  %residual-loop-index2307 = phi i64 [ %350, %loop-memcpy-residual2305 ], [ 0, %loop-memcpy-residual2305.preheader ]
  %349 = load i8, ptr addrspace(5) %lsr.iv3207, align 1
  %scevgep3210 = getelementptr i8, ptr %scevgep3209, i64 %residual-loop-index2307
  store i8 %349, ptr %scevgep3210, align 1
  %350 = add i64 %residual-loop-index2307, 1
  %scevgep3208 = getelementptr i8, ptr addrspace(5) %lsr.iv3207, i32 1
  %351 = icmp uge i64 %350, %336
  %352 = call i64 @llvm.amdgcn.if.break.i64(i1 %351, i64 %phi.broken3916)
  %353 = call i1 @llvm.amdgcn.loop.i64(i64 %352)
  br i1 %353, label %Flow3843, label %loop-memcpy-residual2305

Flow3843:                                         ; preds = %loop-memcpy-residual2305
  %.lcssa4204 = phi i64 [ %352, %loop-memcpy-residual2305 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa4204)
  br label %Flow3844, !amdgpu.uniform !42

post-loop-memcpy-expansion2302:                   ; preds = %Flow3844
  call void @llvm.lifetime.end.p5(ptr addrspace(5) %buffer.i.i.i) #34, !noalias !109
  br label %for.cond.i.i.i118, !amdgpu.uniform !42

for.cond.i.i.i118:                                ; preds = %post-loop-memcpy-expansion2302, %Flow3842
  %354 = phi i32 [ %392, %Flow3842 ], [ poison, %post-loop-memcpy-expansion2302 ]
  %355 = phi i64 [ %393, %Flow3842 ], [ poison, %post-loop-memcpy-expansion2302 ]
  %356 = phi i32 [ %394, %Flow3842 ], [ poison, %post-loop-memcpy-expansion2302 ]
  %retval.sroa.6.0.i.i.i119 = phi i32 [ %397, %Flow3842 ], [ undef, %post-loop-memcpy-expansion2302 ]
  %retval.sroa.2.0.i.i.i121 = phi i64 [ %396, %Flow3842 ], [ undef, %post-loop-memcpy-expansion2302 ]
  %retval.sroa.8.0.i.i.i122 = phi i32 [ %395, %Flow3842 ], [ undef, %post-loop-memcpy-expansion2302 ]
  %index.0.i.i.i123 = phi i32 [ %399, %Flow3842 ], [ 0, %post-loop-memcpy-expansion2302 ]
  %357 = load i32, ptr addrspace(1) @__llvm_rpc_client, align 8, !tbaa !68
  %cmp.not.i.i.i124 = icmp ult i32 %index.0.i.i.i123, %357
  %spec.store.select.i.i.i125 = select i1 %cmp.not.i.i.i124, i32 %index.0.i.i.i123, i32 0
  %358 = tail call noundef i64 @llvm.amdgcn.ballot.i64(i1 true)
  %359 = and i32 %spec.store.select.i.i.i125, 31
  %360 = lshr i64 %358, %sh_prom.i.i.i.i
  %361 = trunc i64 %360 to i32
  %conv4.i.i.i.i.i129 = and i32 %361, 1
  %shl.i.i.i.i.i130 = shl nuw i32 %conv4.i.i.i.i.i129, %359
  %362 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2449, !amdgpu.uniform !42

363:                                              ; preds = %ComputeEnd2450
  %sunkaddr3291 = getelementptr inbounds i8, ptr addrspace(1) %3085, i64 40
  %364 = atomicrmw or ptr addrspace(1) %sunkaddr3291, i32 %.lcssa4198 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %365, !amdgpu.uniform !42

365:                                              ; preds = %363, %ComputeEnd2450
  %366 = phi i32 [ %364, %363 ], [ poison, %ComputeEnd2450 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %3094)
  %367 = call i32 @llvm.amdgcn.readfirstlane.i32(i32 %366)
  %368 = or i32 %367, %.lcssa4199
  %shl5.i.i.i.i.i131 = shl nuw i32 1, %359
  %and.i.i.i.i.i132 = and i32 %shl5.i.i.i.i.i131, %368
  %tobool3.i.i.i.i133 = icmp ne i32 %and.i.i.i.i.i132, 0
  %369 = tail call i64 @llvm.amdgcn.ballot.i64(i1 %tobool3.i.i.i.i133)
  %and.i.i.i.i.i.i134 = and i64 %369, %358
  %cmp.i.not.i.i.i135 = icmp ne i64 %358, %and.i.i.i.i.i.i134
  br i1 %cmp.i.not.i.i.i135, label %if.then.i.i.i.i136, label %Flow3841, !amdgpu.uniform !42

381:                                              ; preds = %ComputeEnd2455
  %sunkaddr3292 = getelementptr inbounds i8, ptr addrspace(1) %3085, i64 40
  %382 = atomicrmw and ptr addrspace(1) %sunkaddr3292, i32 %.lcssa4200 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %383, !amdgpu.uniform !42

383:                                              ; preds = %381, %ComputeEnd2455
  call void @llvm.amdgcn.end.cf.i64(i64 %3112)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %Flow3839, !amdgpu.uniform !42

if.end14.i.i.i206:                                ; preds = %Flow3839
  %384 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %358, i1 true)
  %iszero.i.i.i.i.i.i207 = icmp ne i64 %358, 0
  %cmp2.i.i.i.i.i208 = icmp eq i64 %384, %sh_prom.i.i.i.i
  %cmp.i.i.i.i.i209 = select i1 %iszero.i.i.i.i.i.i207, i1 %cmp2.i.i.i.i.i208, i1 false
  %385 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.i.i.i.i.i209)
  %386 = extractvalue { i1, i64 } %385, 0
  %387 = extractvalue { i1, i64 } %385, 1
  br i1 %386, label %if.then16.i.i.i211, label %if.end22.i.i.i210

Flow3840:                                         ; preds = %if.end22.i.i.i210, %Flow3839
  %389 = phi i32 [ %376, %if.end22.i.i.i210 ], [ %retval.sroa.8.0.i.i.i122, %Flow3839 ]
  %390 = phi i64 [ %358, %if.end22.i.i.i210 ], [ %retval.sroa.2.0.i.i.i121, %Flow3839 ]
  %391 = phi i32 [ %spec.store.select.i.i.i125, %if.end22.i.i.i210 ], [ %retval.sroa.6.0.i.i.i119, %Flow3839 ]
  br label %cleanup26.i.i.i147, !amdgpu.uniform !42

Flow3841:                                         ; preds = %cleanup26.i.i.i147, %365
  %392 = phi i32 [ %391, %cleanup26.i.i.i147 ], [ %354, %365 ]
  %393 = phi i64 [ %390, %cleanup26.i.i.i147 ], [ %355, %365 ]
  %394 = phi i32 [ %389, %cleanup26.i.i.i147 ], [ %356, %365 ]
  %395 = phi i32 [ %389, %cleanup26.i.i.i147 ], [ %retval.sroa.8.0.i.i.i122, %365 ]
  %396 = phi i64 [ %390, %cleanup26.i.i.i147 ], [ %retval.sroa.2.0.i.i.i121, %365 ]
  %397 = phi i32 [ %391, %cleanup26.i.i.i147 ], [ %retval.sroa.6.0.i.i.i119, %365 ]
  %398 = phi i1 [ %cmp.i46.not.i.i.i140, %cleanup26.i.i.i147 ], [ true, %365 ]
  br i1 %398, label %for.inc.i.i.i152, label %Flow3842, !amdgpu.uniform !42

cleanup26.i.i.i147:                               ; preds = %Flow3840
  br label %Flow3841, !amdgpu.uniform !42

for.inc.i.i.i152:                                 ; preds = %Flow3841
  %inc.i.i.i157 = add i32 %spec.store.select.i.i.i125, 1
  br label %Flow3842, !amdgpu.uniform !42

Flow3842:                                         ; preds = %for.inc.i.i.i152, %Flow3841
  %399 = phi i32 [ %inc.i.i.i157, %for.inc.i.i.i152 ], [ poison, %Flow3841 ]
  %400 = phi i1 [ false, %for.inc.i.i.i152 ], [ true, %Flow3841 ]
  br i1 %400, label %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i158, label %for.cond.i.i.i118, !amdgpu.uniform !42

_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i158: ; preds = %Flow3842
  %.lcssa4203 = phi i32 [ %392, %Flow3842 ]
  %.lcssa4202 = phi i64 [ %393, %Flow3842 ]
  %.lcssa4201 = phi i32 [ %394, %Flow3842 ]
  fence acquire
  %401 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %mul.i.i.i.i.i.i159 = shl i32 %.lcssa4203, 6
  %idxprom.i19.i.i.i.i.i160 = zext i32 %mul.i.i.i.i.i.i159 to i64
  %arrayidx.i20.i.i.i.i.i161 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %401, i64 %idxprom.i19.i.i.i.i.i160
  %arrayidx.i22.i.i.i.i.i162 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %arrayidx.i20.i.i.i.i.i161, i64 %sh_prom.i.i.i.i
  store i64 %writer.sroa.10.4.i.i.i.i, ptr %arrayidx.i22.i.i.i.i.i162, align 8, !tbaa !82
  %spec.select.i.i.i.i.i.i.i163 = tail call i64 @llvm.umin.i64(i64 %writer.sroa.10.4.i.i.i.i, i64 56)
  %402 = and i64 %spec.select.i.i.i.i.i.i.i163, 15
  %403 = sub i64 %spec.select.i.i.i.i.i.i.i163, %402
  %404 = icmp ne i64 %403, 0
  %405 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %404)
  %406 = extractvalue { i1, i64 } %405, 0
  %407 = extractvalue { i1, i64 } %405, 1
  br i1 %406, label %loop-memcpy-expansion2279.preheader, label %Flow3838

loop-memcpy-expansion2279.preheader:              ; preds = %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i158
  %408 = add nuw nsw i64 %75, 8
  %scevgep3203 = getelementptr i8, ptr %401, i64 %408
  %409 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i160, 6
  %scevgep3204 = getelementptr i8, ptr %scevgep3203, i64 %409
  br label %loop-memcpy-expansion2279, !amdgpu.uniform !42

Flow3838:                                         ; preds = %Flow3837, %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i158
  call void @llvm.amdgcn.end.cf.i64(i64 %407)
  br label %loop-memcpy-residual-header2282, !amdgpu.uniform !42

loop-memcpy-expansion2279:                        ; preds = %loop-memcpy-expansion2279.preheader, %loop-memcpy-expansion2279
  %phi.broken3917 = phi i64 [ 0, %loop-memcpy-expansion2279.preheader ], [ %414, %loop-memcpy-expansion2279 ]
  %loop-index2280 = phi i64 [ %412, %loop-memcpy-expansion2279 ], [ 0, %loop-memcpy-expansion2279.preheader ]
  %410 = getelementptr i8, ptr %325, i64 %loop-index2280
  %411 = load <4 x i32>, ptr %410, align 1
  %scevgep3205 = getelementptr i8, ptr %scevgep3204, i64 %loop-index2280
  store <4 x i32> %411, ptr %scevgep3205, align 1
  %412 = add i64 %loop-index2280, 16
  %413 = icmp uge i64 %412, %403
  %414 = call i64 @llvm.amdgcn.if.break.i64(i1 %413, i64 %phi.broken3917)
  %415 = call i1 @llvm.amdgcn.loop.i64(i64 %414)
  br i1 %415, label %Flow3837, label %loop-memcpy-expansion2279

Flow3836:                                         ; preds = %Flow3834, %Flow3835
  %416 = phi i64 [ %2858, %Flow3834 ], [ %2850, %Flow3835 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %2853)
  br label %post-loop-memcpy-expansion2278, !amdgpu.uniform !42

loop-memcpy-residual2281:                         ; preds = %loop-memcpy-residual2281.preheader, %loop-memcpy-residual2281
  %phi.broken3918 = phi i64 [ 0, %loop-memcpy-residual2281.preheader ], [ %420, %loop-memcpy-residual2281 ]
  %residual-loop-index2283 = phi i64 [ %418, %loop-memcpy-residual2281 ], [ 0, %loop-memcpy-residual2281.preheader ]
  %scevgep3202 = getelementptr i8, ptr %scevgep3201, i64 %residual-loop-index2283
  %417 = load i8, ptr %scevgep3202, align 1
  %scevgep3200 = getelementptr i8, ptr %scevgep3199, i64 %residual-loop-index2283
  store i8 %417, ptr %scevgep3200, align 1
  %418 = add i64 %residual-loop-index2283, 1
  %419 = icmp uge i64 %418, %402
  %420 = call i64 @llvm.amdgcn.if.break.i64(i1 %419, i64 %phi.broken3918)
  %421 = call i1 @llvm.amdgcn.loop.i64(i64 %420)
  br i1 %421, label %Flow3834, label %loop-memcpy-residual2281

Flow3834:                                         ; preds = %loop-memcpy-residual2281
  %.lcssa4196 = phi i64 [ %420, %loop-memcpy-residual2281 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa4196)
  br label %Flow3836, !amdgpu.uniform !42

post-loop-memcpy-expansion2278:                   ; preds = %Flow3836
  %tobool.not.i.i.i.i.i.i165 = icmp eq i32 %.lcssa4201, 0
  %conv.i.i.i.i.i.i166 = zext i1 %tobool.not.i.i.i.i.i.i165 to i32
  fence release
  %422 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %idxprom.i23.i.i.i.i.i167 = zext i32 %.lcssa4203 to i64
  %arrayidx.i24.i.i.i.i.i168 = getelementptr inbounds nuw i32, ptr %422, i64 %idxprom.i23.i.i.i.i.i167
  store atomic i32 %conv.i.i.i.i.i.i166, ptr %arrayidx.i24.i.i.i.i.i168 monotonic, align 4
  %423 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !84
  %arrayidx.i.i5.i.i169 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %423, i64 %idxprom.i23.i.i.i.i.i167, !amdgpu.uniform !42
  %424 = load i64, ptr %arrayidx.i.i5.i.i169, align 8, !tbaa !80
  %scevgep3185 = getelementptr i8, ptr %325, i64 56
  %425 = add i64 %writer.sroa.10.4.i.i.i.i, -56
  br label %while.cond.i.i.i.i170, !amdgpu.uniform !42

while.cond.i.i.i.i170:                            ; preds = %post-loop-memcpy-expansion2278, %Flow3833
  %phi.broken3921 = phi i64 [ 0, %post-loop-memcpy-expansion2278 ], [ %442, %Flow3833 ]
  %lsr.iv3191 = phi i64 [ %440, %Flow3833 ], [ %425, %post-loop-memcpy-expansion2278 ]
  %lsr.iv3186 = phi ptr [ %439, %Flow3833 ], [ %scevgep3185, %post-loop-memcpy-expansion2278 ]
  %port.sroa.37.0.i.i171 = phi i32 [ %438, %Flow3833 ], [ %conv.i.i.i.i.i.i166, %post-loop-memcpy-expansion2278 ]
  %idx.0.i.i.i.i172 = phi i64 [ %437, %Flow3833 ], [ 56, %post-loop-memcpy-expansion2278 ]
  %cmp.i.i.i.i173 = icmp ugt i64 %writer.sroa.10.4.i.i.i.i, %idx.0.i.i.i.i172
  %426 = tail call i64 @llvm.amdgcn.ballot.i64(i1 %cmp.i.i.i.i173)
  %and.i.i.i.i6.i.i174 = and i64 %426, %424
  %tobool.not.i.i.i.i175 = icmp ne i64 %and.i.i.i.i6.i.i174, 0
  %427 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i22.i.i.i176 = getelementptr inbounds nuw i32, ptr %427, i64 %idxprom.i23.i.i.i.i.i167, !amdgpu.uniform !42
  %428 = load atomic i32, ptr %arrayidx.i22.i.i.i176 monotonic, align 4
  %429 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %tobool.not.i.i.i.i175)
  %430 = extractvalue { i1, i64 } %429, 0
  %431 = extractvalue { i1, i64 } %429, 1
  br i1 %430, label %cond.false.i8.i.i.i.i177, label %Flow3833

while.body.i.i21.i.i.i.i178.preheader:            ; preds = %cond.false.i8.i.i.i.i177
  br label %while.body.i.i21.i.i.i.i178, !amdgpu.uniform !42

Flow3832:                                         ; preds = %Flow3831, %cond.false.i8.i.i.i.i177
  br label %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i181, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i178:                      ; preds = %while.body.i.i21.i.i.i.i178.preheader, %while.body.i.i21.i.i.i.i178
  tail call void @llvm.amdgcn.s.sleep(i32 2)
  %434 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i22.i.i.i.i179 = getelementptr inbounds nuw i32, ptr %434, i64 %idxprom.i23.i.i.i.i.i167, !amdgpu.uniform !42
  %435 = load atomic i32, ptr %arrayidx.i.i.i22.i.i.i.i179 monotonic, align 4
  %436 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %435)
  %cmp.i.not.i.i23.i.i.i.i180 = icmp eq i32 %436, %port.sroa.37.0.i.i171
  br i1 %cmp.i.not.i.i23.i.i.i.i180, label %Flow3831, label %while.body.i.i21.i.i.i.i178, !amdgpu.uniform !42

Flow3831:                                         ; preds = %while.body.i.i21.i.i.i.i178
  br label %Flow3832, !amdgpu.uniform !42

Flow3833:                                         ; preds = %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i182, %while.cond.i.i.i.i170
  %437 = phi i64 [ %add.i.i.i.i185, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i182 ], [ poison, %while.cond.i.i.i.i170 ]
  %438 = phi i32 [ %conv.i.i26.i.i.i.i183, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i182 ], [ poison, %while.cond.i.i.i.i170 ]
  %439 = phi ptr [ %scevgep3187, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i182 ], [ poison, %while.cond.i.i.i.i170 ]
  %440 = phi i64 [ %lsr.iv.next3192, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i182 ], [ poison, %while.cond.i.i.i.i170 ]
  %441 = phi i1 [ false, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i182 ], [ true, %while.cond.i.i.i.i170 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %431)
  %442 = call i64 @llvm.amdgcn.if.break.i64(i1 %441, i64 %phi.broken3921)
  %443 = call i1 @llvm.amdgcn.loop.i64(i64 %442)
  br i1 %443, label %cond.false.i.i.i191, label %while.cond.i.i.i.i170

_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i181: ; preds = %Flow3832
  fence acquire
  %444 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.i.i.i.i173)
  %445 = extractvalue { i1, i64 } %444, 0
  %446 = extractvalue { i1, i64 } %444, 1
  br i1 %445, label %if.then.i.i.i.i.i.i3.i, label %Flow3830

loop-memcpy-expansion2273.preheader:              ; preds = %if.then.i.i.i.i.i.i3.i
  %arrayidx.i20.i35.i.i.i.i189 = getelementptr %"struct.rpc::Buffer", ptr %447, i64 %idxprom.i19.i.i.i.i.i160
  %arrayidx.i22.i38.i.i.i.i190 = getelementptr %"struct.rpc::Buffer", ptr %arrayidx.i20.i35.i.i.i.i189, i64 %sh_prom.i.i.i.i
  br label %loop-memcpy-expansion2273, !amdgpu.uniform !42

Flow3829:                                         ; preds = %Flow3828, %if.then.i.i.i.i.i.i3.i
  call void @llvm.amdgcn.end.cf.i64(i64 %453)
  br label %loop-memcpy-residual-header2276, !amdgpu.uniform !42

loop-memcpy-expansion2273:                        ; preds = %loop-memcpy-expansion2273.preheader, %loop-memcpy-expansion2273
  %phi.broken3919 = phi i64 [ 0, %loop-memcpy-expansion2273.preheader ], [ %458, %loop-memcpy-expansion2273 ]
  %loop-index2274 = phi i64 [ %456, %loop-memcpy-expansion2273 ], [ 0, %loop-memcpy-expansion2273.preheader ]
  %scevgep3188 = getelementptr i8, ptr %lsr.iv3186, i64 %loop-index2274
  %454 = load <4 x i32>, ptr %scevgep3188, align 1
  %455 = getelementptr i8, ptr %arrayidx.i22.i38.i.i.i.i190, i64 %loop-index2274
  store <4 x i32> %454, ptr %455, align 1
  %456 = add i64 %loop-index2274, 16
  %457 = icmp uge i64 %456, %449
  %458 = call i64 @llvm.amdgcn.if.break.i64(i1 %457, i64 %phi.broken3919)
  %459 = call i1 @llvm.amdgcn.loop.i64(i64 %458)
  br i1 %459, label %Flow3828, label %loop-memcpy-expansion2273

Flow3827:                                         ; preds = %Flow3826, %loop-memcpy-residual-header2276
  call void @llvm.amdgcn.end.cf.i64(i64 %2843)
  br label %Flow3830, !amdgpu.uniform !42

loop-memcpy-residual2275:                         ; preds = %loop-memcpy-residual2275.preheader, %loop-memcpy-residual2275
  %phi.broken3920 = phi i64 [ 0, %loop-memcpy-residual2275.preheader ], [ %463, %loop-memcpy-residual2275 ]
  %residual-loop-index2277 = phi i64 [ %461, %loop-memcpy-residual2275 ], [ 0, %loop-memcpy-residual2275.preheader ]
  %scevgep3197 = getelementptr i8, ptr %scevgep3196, i64 %residual-loop-index2277
  %460 = load i8, ptr %scevgep3197, align 1
  %scevgep3195 = getelementptr i8, ptr %scevgep3194, i64 %residual-loop-index2277
  store i8 %460, ptr %scevgep3195, align 1
  %461 = add i64 %residual-loop-index2277, 1
  %462 = icmp uge i64 %461, %448
  %463 = call i64 @llvm.amdgcn.if.break.i64(i1 %462, i64 %phi.broken3920)
  %464 = call i1 @llvm.amdgcn.loop.i64(i64 %463)
  br i1 %464, label %Flow3826, label %loop-memcpy-residual2275

Flow3826:                                         ; preds = %loop-memcpy-residual2275
  %.lcssa4193 = phi i64 [ %463, %loop-memcpy-residual2275 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa4193)
  br label %Flow3827, !amdgpu.uniform !42

_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i182: ; preds = %Flow3830
  %conv.i.i26.i.i.i.i183 = xor i32 %port.sroa.37.0.i.i171, 1
  fence release
  %465 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %arrayidx.i24.i29.i.i.i.i184 = getelementptr inbounds nuw i32, ptr %465, i64 %idxprom.i23.i.i.i.i.i167
  store atomic i32 %conv.i.i26.i.i.i.i183, ptr %arrayidx.i24.i29.i.i.i.i184 monotonic, align 4
  %add.i.i.i.i185 = add i64 %idx.0.i.i.i.i172, 64
  %scevgep3187 = getelementptr i8, ptr %lsr.iv3186, i64 64
  %lsr.iv.next3192 = add i64 %lsr.iv3191, -64
  br label %Flow3833, !amdgpu.uniform !42

cond.false.i.i.i191:                              ; preds = %Flow3833
  %.lcssa4195 = phi i64 [ %442, %Flow3833 ]
  %port.sroa.37.0.i.i171.lcssa = phi i32 [ %port.sroa.37.0.i.i171, %Flow3833 ]
  %.lcssa4194 = phi i32 [ %428, %Flow3833 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa4195)
  %466 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %.lcssa4194)
  %467 = icmp ne i32 %466, %port.sroa.37.0.i.i171.lcssa
  %468 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %467)
  %469 = extractvalue { i1, i64 } %468, 0
  %470 = extractvalue { i1, i64 } %468, 1
  br i1 %469, label %while.body.i.i.i.i192.preheader, label %Flow3825

while.body.i.i.i.i192.preheader:                  ; preds = %cond.false.i.i.i191
  br label %while.body.i.i.i.i192, !amdgpu.uniform !42

Flow3825:                                         ; preds = %Flow3824, %cond.false.i.i.i191
  call void @llvm.amdgcn.end.cf.i64(i64 %470)
  br label %_ZN22__llvm_libc_22_0_0_git15write_to_stderrENS_3cpp11string_viewE.exit.i, !amdgpu.uniform !42

while.body.i.i.i.i192:                            ; preds = %while.body.i.i.i.i192.preheader, %while.body.i.i.i.i192
  %phi.broken3922 = phi i64 [ 0, %while.body.i.i.i.i192.preheader ], [ %474, %while.body.i.i.i.i192 ]
  tail call void @llvm.amdgcn.s.sleep(i32 2)
  %471 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i8.i.i193 = getelementptr inbounds nuw i32, ptr %471, i64 %idxprom.i23.i.i.i.i.i167, !amdgpu.uniform !42
  %472 = load atomic i32, ptr %arrayidx.i.i.i8.i.i193 monotonic, align 4
  %473 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %472)
  %cmp.i.not.i.i.i.i194 = icmp eq i32 %473, %port.sroa.37.0.i.i171.lcssa
  %474 = call i64 @llvm.amdgcn.if.break.i64(i1 %cmp.i.not.i.i.i.i194, i64 %phi.broken3922)
  %475 = call i1 @llvm.amdgcn.loop.i64(i64 %474)
  br i1 %475, label %Flow3824, label %while.body.i.i.i.i192

Flow3824:                                         ; preds = %while.body.i.i.i.i192
  %.lcssa4191 = phi i64 [ %474, %while.body.i.i.i.i192 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa4191)
  br label %Flow3825, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git15write_to_stderrENS_3cpp11string_viewE.exit.i: ; preds = %Flow3825
  fence acquire
  tail call void @llvm.amdgcn.wave.barrier()
  fence syncscope("agent") release
  %476 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %.lcssa4202, i1 true)
  %iszero.i.i.i.i.i13.i.i195 = icmp ne i64 %.lcssa4202, 0
  %cmp2.i.i.i.i14.i.i196 = icmp eq i64 %476, %sh_prom.i.i.i.i
  %cmp.i.i.i.i15.i.i197 = select i1 %iszero.i.i.i.i.i13.i.i195, i1 %cmp2.i.i.i.i14.i.i196, i1 false
  %477 = and i32 %.lcssa4203, 31
  %conv4.i.i.i19.i.i201 = zext i1 %cmp.i.i.i.i15.i.i197 to i32
  %shl.i.i.i20.i.i202 = shl nuw i32 %conv4.i.i.i19.i.i201, %477
  %xor.i.i.i21.i.i203 = xor i32 %shl.i.i.i20.i.i202, -1
  %478 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2458, !amdgpu.uniform !42

479:                                              ; preds = %ComputeEnd2459
  %div8.i.i.i16.i.i198 = lshr i32 %.lcssa4203, 5
  %idxprom.i.i.i17.i.i199 = zext nneg i32 %div8.i.i.i16.i.i198 to i64
  %480 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i17.i.i199
  %sunkaddr3293 = getelementptr inbounds i8, ptr addrspace(1) %480, i64 40
  %481 = atomicrmw and ptr addrspace(1) %sunkaddr3293, i32 %.lcssa4190 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %482, !amdgpu.uniform !42

482:                                              ; preds = %479, %ComputeEnd2459
  call void @llvm.amdgcn.end.cf.i64(i64 %3129)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %for.cond.i.i1.i222, !amdgpu.uniform !42

for.cond.i.i1.i222:                               ; preds = %482, %Flow3823
  %483 = phi i32 [ %521, %Flow3823 ], [ poison, %482 ]
  %484 = phi i64 [ %522, %Flow3823 ], [ poison, %482 ]
  %485 = phi i32 [ %523, %Flow3823 ], [ poison, %482 ]
  %retval.sroa.6.0.i.i.i223 = phi i32 [ %526, %Flow3823 ], [ undef, %482 ]
  %retval.sroa.2.0.i.i.i225 = phi i64 [ %525, %Flow3823 ], [ undef, %482 ]
  %retval.sroa.8.0.i.i.i226 = phi i32 [ %524, %Flow3823 ], [ undef, %482 ]
  %index.0.i.i.i227 = phi i32 [ %528, %Flow3823 ], [ 0, %482 ]
  %486 = load i32, ptr addrspace(1) @__llvm_rpc_client, align 8, !tbaa !68
  %cmp.not.i.i.i228 = icmp ult i32 %index.0.i.i.i227, %486
  %spec.store.select.i.i.i229 = select i1 %cmp.not.i.i.i228, i32 %index.0.i.i.i227, i32 0
  %487 = tail call noundef i64 @llvm.amdgcn.ballot.i64(i1 true)
  %488 = and i32 %spec.store.select.i.i.i229, 31
  %489 = lshr i64 %487, %sh_prom.i.i.i.i
  %490 = trunc i64 %489 to i32
  %conv4.i.i.i.i.i233 = and i32 %490, 1
  %shl.i.i.i.i.i234 = shl nuw i32 %conv4.i.i.i.i.i233, %488
  %491 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2462, !amdgpu.uniform !42

492:                                              ; preds = %ComputeEnd2463
  %sunkaddr3294 = getelementptr inbounds i8, ptr addrspace(1) %3139, i64 40
  %493 = atomicrmw or ptr addrspace(1) %sunkaddr3294, i32 %.lcssa4184 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %494, !amdgpu.uniform !42

494:                                              ; preds = %492, %ComputeEnd2463
  %495 = phi i32 [ %493, %492 ], [ poison, %ComputeEnd2463 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %3148)
  %496 = call i32 @llvm.amdgcn.readfirstlane.i32(i32 %495)
  %497 = or i32 %496, %.lcssa4185
  %shl5.i.i.i.i.i235 = shl nuw i32 1, %488
  %and.i.i.i.i.i236 = and i32 %shl5.i.i.i.i.i235, %497
  %tobool3.i.i.i.i237 = icmp ne i32 %and.i.i.i.i.i236, 0
  %498 = tail call i64 @llvm.amdgcn.ballot.i64(i1 %tobool3.i.i.i.i237)
  %and.i.i.i.i.i.i238 = and i64 %498, %487
  %cmp.i.not.i.i.i239 = icmp ne i64 %487, %and.i.i.i.i.i.i238
  br i1 %cmp.i.not.i.i.i239, label %if.then.i.i.i.i240, label %Flow3822, !amdgpu.uniform !42

510:                                              ; preds = %ComputeEnd2468
  %sunkaddr3295 = getelementptr inbounds i8, ptr addrspace(1) %3139, i64 40
  %511 = atomicrmw and ptr addrspace(1) %sunkaddr3295, i32 %.lcssa4186 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %512, !amdgpu.uniform !42

512:                                              ; preds = %510, %ComputeEnd2468
  call void @llvm.amdgcn.end.cf.i64(i64 %3166)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %Flow3820, !amdgpu.uniform !42

if.end14.i.i.i311:                                ; preds = %Flow3820
  %513 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %487, i1 true)
  %iszero.i.i.i.i.i.i312 = icmp ne i64 %487, 0
  %cmp2.i.i.i.i.i313 = icmp eq i64 %513, %sh_prom.i.i.i.i
  %cmp.i.i.i.i.i314 = select i1 %iszero.i.i.i.i.i.i312, i1 %cmp2.i.i.i.i.i313, i1 false
  %514 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.i.i.i.i.i314)
  %515 = extractvalue { i1, i64 } %514, 0
  %516 = extractvalue { i1, i64 } %514, 1
  br i1 %515, label %if.then16.i.i.i316, label %if.end22.i.i.i315

Flow3821:                                         ; preds = %if.end22.i.i.i315, %Flow3820
  %518 = phi i32 [ %505, %if.end22.i.i.i315 ], [ %retval.sroa.8.0.i.i.i226, %Flow3820 ]
  %519 = phi i64 [ %487, %if.end22.i.i.i315 ], [ %retval.sroa.2.0.i.i.i225, %Flow3820 ]
  %520 = phi i32 [ %spec.store.select.i.i.i229, %if.end22.i.i.i315 ], [ %retval.sroa.6.0.i.i.i223, %Flow3820 ]
  br label %cleanup26.i.i.i252, !amdgpu.uniform !42

Flow3822:                                         ; preds = %cleanup26.i.i.i252, %494
  %521 = phi i32 [ %520, %cleanup26.i.i.i252 ], [ %483, %494 ]
  %522 = phi i64 [ %519, %cleanup26.i.i.i252 ], [ %484, %494 ]
  %523 = phi i32 [ %518, %cleanup26.i.i.i252 ], [ %485, %494 ]
  %524 = phi i32 [ %518, %cleanup26.i.i.i252 ], [ %retval.sroa.8.0.i.i.i226, %494 ]
  %525 = phi i64 [ %519, %cleanup26.i.i.i252 ], [ %retval.sroa.2.0.i.i.i225, %494 ]
  %526 = phi i32 [ %520, %cleanup26.i.i.i252 ], [ %retval.sroa.6.0.i.i.i223, %494 ]
  %527 = phi i1 [ %cmp.i46.not.i.i.i244, %cleanup26.i.i.i252 ], [ true, %494 ]
  br i1 %527, label %for.inc.i.i.i257, label %Flow3823, !amdgpu.uniform !42

cleanup26.i.i.i252:                               ; preds = %Flow3821
  br label %Flow3822, !amdgpu.uniform !42

for.inc.i.i.i257:                                 ; preds = %Flow3822
  %inc.i.i.i262 = add i32 %spec.store.select.i.i.i229, 1
  br label %Flow3823, !amdgpu.uniform !42

Flow3823:                                         ; preds = %for.inc.i.i.i257, %Flow3822
  %528 = phi i32 [ %inc.i.i.i262, %for.inc.i.i.i257 ], [ poison, %Flow3822 ]
  %529 = phi i1 [ false, %for.inc.i.i.i257 ], [ true, %Flow3822 ]
  br i1 %529, label %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i263, label %for.cond.i.i1.i222, !amdgpu.uniform !42

_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i263: ; preds = %Flow3823
  %.lcssa4189 = phi i32 [ %521, %Flow3823 ]
  %.lcssa4188 = phi i64 [ %522, %Flow3823 ]
  %.lcssa4187 = phi i32 [ %523, %Flow3823 ]
  fence acquire
  %530 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %mul.i.i.i.i.i.i264 = shl i32 %.lcssa4189, 6
  %idxprom.i19.i.i.i.i.i265 = zext i32 %mul.i.i.i.i.i.i264 to i64
  %arrayidx.i20.i.i.i.i.i266 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %530, i64 %idxprom.i19.i.i.i.i.i265
  %arrayidx.i22.i.i.i.i.i267 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %arrayidx.i20.i.i.i.i.i266, i64 %sh_prom.i.i.i.i
  store i64 sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.7.11 to ptr), i64 10) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.7.11 to ptr) to i64)), ptr %arrayidx.i22.i.i.i.i.i267, align 8, !tbaa !82
  %spec.select.i.i.i.i.i.i.i268 = tail call i64 @llvm.umin.i64(i64 sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.7.11 to ptr), i64 10) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.7.11 to ptr) to i64)), i64 56)
  %531 = and i64 %spec.select.i.i.i.i.i.i.i268, 15
  %532 = sub i64 %spec.select.i.i.i.i.i.i.i268, %531
  %533 = icmp ne i64 %532, 0
  br i1 %533, label %loop-memcpy-expansion2345.preheader, label %Flow3819, !amdgpu.uniform !42

loop-memcpy-expansion2345.preheader:              ; preds = %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i263
  %534 = add nuw nsw i64 %75, 8
  %scevgep3182 = getelementptr i8, ptr %530, i64 %534
  %535 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i265, 6
  %scevgep3183 = getelementptr i8, ptr %scevgep3182, i64 %535
  br label %loop-memcpy-expansion2345, !amdgpu.uniform !42

Flow3819:                                         ; preds = %Flow3818, %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i263
  br label %loop-memcpy-residual-header2348, !amdgpu.uniform !42

loop-memcpy-expansion2345:                        ; preds = %loop-memcpy-expansion2345.preheader, %loop-memcpy-expansion2345
  %loop-index2346 = phi i64 [ %538, %loop-memcpy-expansion2345 ], [ 0, %loop-memcpy-expansion2345.preheader ]
  %536 = getelementptr i8, ptr addrspace(4) @.str.7.11, i64 %loop-index2346, !amdgpu.uniform !42
  %537 = load <4 x i32>, ptr addrspace(4) %536, align 1
  %scevgep3184 = getelementptr i8, ptr %scevgep3183, i64 %loop-index2346
  store <4 x i32> %537, ptr %scevgep3184, align 1
  %538 = add i64 %loop-index2346, 16
  %539 = icmp uge i64 %538, %532
  br i1 %539, label %Flow3818, label %loop-memcpy-expansion2345, !amdgpu.uniform !42

Flow3817:                                         ; preds = %Flow3815, %Flow3816
  %540 = phi i64 [ %2914, %Flow3815 ], [ %2911, %Flow3816 ]
  br label %post-loop-memcpy-expansion2344, !amdgpu.uniform !42

loop-memcpy-residual2347:                         ; preds = %loop-memcpy-residual2347.preheader, %loop-memcpy-residual2347
  %residual-loop-index2349 = phi i64 [ %542, %loop-memcpy-residual2347 ], [ 0, %loop-memcpy-residual2347.preheader ]
  %scevgep3181 = getelementptr i8, ptr addrspace(4) @.str.7.11, i64 %residual-loop-index2349, !amdgpu.uniform !42
  %541 = load i8, ptr addrspace(4) %scevgep3181, align 1
  %scevgep3180 = getelementptr i8, ptr %scevgep3179, i64 %residual-loop-index2349
  store i8 %541, ptr %scevgep3180, align 1
  %542 = add i64 %residual-loop-index2349, 1
  %543 = icmp uge i64 %542, %531
  br i1 %543, label %Flow3815, label %loop-memcpy-residual2347, !amdgpu.uniform !42

Flow3815:                                         ; preds = %loop-memcpy-residual2347
  br label %Flow3817, !amdgpu.uniform !42

post-loop-memcpy-expansion2344:                   ; preds = %Flow3817
  %tobool.not.i.i.i.i.i.i270 = icmp eq i32 %.lcssa4187, 0
  %conv.i.i.i.i.i.i271 = zext i1 %tobool.not.i.i.i.i.i.i270 to i32
  fence release
  %544 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %idxprom.i23.i.i.i.i.i272 = zext i32 %.lcssa4189 to i64
  %arrayidx.i24.i.i.i.i.i273 = getelementptr inbounds nuw i32, ptr %544, i64 %idxprom.i23.i.i.i.i.i272
  store atomic i32 %conv.i.i.i.i.i.i271, ptr %arrayidx.i24.i.i.i.i.i273 monotonic, align 4
  %545 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !84
  %arrayidx.i.i5.i.i274 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %545, i64 %idxprom.i23.i.i.i.i.i272, !amdgpu.uniform !42
  %546 = load i64, ptr %arrayidx.i.i5.i.i274, align 8, !tbaa !80
  br label %while.cond.i.i.i.i275, !amdgpu.uniform !42

while.cond.i.i.i.i275:                            ; preds = %post-loop-memcpy-expansion2344, %Flow3814
  %phi.broken3923 = phi i64 [ 0, %post-loop-memcpy-expansion2344 ], [ %563, %Flow3814 ]
  %lsr.iv3171 = phi i64 [ %561, %Flow3814 ], [ -46, %post-loop-memcpy-expansion2344 ]
  %lsr.iv3166 = phi ptr addrspace(4) [ %560, %Flow3814 ], [ getelementptr (i8, ptr addrspace(4) @.str.7.11, i64 56), %post-loop-memcpy-expansion2344 ]
  %port.sroa.37.0.i.i276 = phi i32 [ %559, %Flow3814 ], [ %conv.i.i.i.i.i.i271, %post-loop-memcpy-expansion2344 ]
  %idx.0.i.i.i.i277 = phi i64 [ %558, %Flow3814 ], [ 56, %post-loop-memcpy-expansion2344 ]
  %cmp.i.i.i.i278 = icmp ult i64 %idx.0.i.i.i.i277, sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.7.11 to ptr), i64 10) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.7.11 to ptr) to i64))
  %547 = tail call i64 @llvm.amdgcn.ballot.i64(i1 %cmp.i.i.i.i278)
  %and.i.i.i.i6.i.i279 = and i64 %547, %546
  %tobool.not.i.i.i.i280 = icmp ne i64 %and.i.i.i.i6.i.i279, 0
  %548 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i22.i.i.i281 = getelementptr inbounds nuw i32, ptr %548, i64 %idxprom.i23.i.i.i.i.i272, !amdgpu.uniform !42
  %549 = load atomic i32, ptr %arrayidx.i22.i.i.i281 monotonic, align 4
  %550 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %tobool.not.i.i.i.i280)
  %551 = extractvalue { i1, i64 } %550, 0
  %552 = extractvalue { i1, i64 } %550, 1
  br i1 %551, label %cond.false.i8.i.i.i.i282, label %Flow3814

while.body.i.i21.i.i.i.i283.preheader:            ; preds = %cond.false.i8.i.i.i.i282
  br label %while.body.i.i21.i.i.i.i283, !amdgpu.uniform !42

Flow3813:                                         ; preds = %Flow3812, %cond.false.i8.i.i.i.i282
  br label %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i286, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i283:                      ; preds = %while.body.i.i21.i.i.i.i283.preheader, %while.body.i.i21.i.i.i.i283
  tail call void @llvm.amdgcn.s.sleep(i32 2)
  %555 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i22.i.i.i.i284 = getelementptr inbounds nuw i32, ptr %555, i64 %idxprom.i23.i.i.i.i.i272, !amdgpu.uniform !42
  %556 = load atomic i32, ptr %arrayidx.i.i.i22.i.i.i.i284 monotonic, align 4
  %557 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %556)
  %cmp.i.not.i.i23.i.i.i.i285 = icmp eq i32 %557, %port.sroa.37.0.i.i276
  br i1 %cmp.i.not.i.i23.i.i.i.i285, label %Flow3812, label %while.body.i.i21.i.i.i.i283, !amdgpu.uniform !42

Flow3812:                                         ; preds = %while.body.i.i21.i.i.i.i283
  br label %Flow3813, !amdgpu.uniform !42

Flow3814:                                         ; preds = %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i287, %while.cond.i.i.i.i275
  %558 = phi i64 [ %add.i.i.i.i290, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i287 ], [ poison, %while.cond.i.i.i.i275 ]
  %559 = phi i32 [ %conv.i.i26.i.i.i.i288, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i287 ], [ poison, %while.cond.i.i.i.i275 ]
  %560 = phi ptr addrspace(4) [ %scevgep3167, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i287 ], [ poison, %while.cond.i.i.i.i275 ]
  %561 = phi i64 [ %lsr.iv.next3172, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i287 ], [ poison, %while.cond.i.i.i.i275 ]
  %562 = phi i1 [ false, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i287 ], [ true, %while.cond.i.i.i.i275 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %552)
  %563 = call i64 @llvm.amdgcn.if.break.i64(i1 %562, i64 %phi.broken3923)
  %564 = call i1 @llvm.amdgcn.loop.i64(i64 %563)
  br i1 %564, label %cond.false.i.i.i297, label %while.cond.i.i.i.i275

_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i286: ; preds = %Flow3813
  fence acquire
  br i1 %cmp.i.i.i.i278, label %if.then.i.i.i.i.i.i.i291, label %Flow3811, !amdgpu.uniform !42

loop-memcpy-expansion2351.preheader:              ; preds = %if.then.i.i.i.i.i.i.i291
  %arrayidx.i20.i35.i.i.i.i295 = getelementptr %"struct.rpc::Buffer", ptr %565, i64 %idxprom.i19.i.i.i.i.i265
  %arrayidx.i22.i38.i.i.i.i296 = getelementptr %"struct.rpc::Buffer", ptr %arrayidx.i20.i35.i.i.i.i295, i64 %sh_prom.i.i.i.i
  br label %loop-memcpy-expansion2351, !amdgpu.uniform !42

Flow3810:                                         ; preds = %Flow3809, %if.then.i.i.i.i.i.i.i291
  br label %loop-memcpy-residual-header2354, !amdgpu.uniform !42

loop-memcpy-expansion2351:                        ; preds = %loop-memcpy-expansion2351.preheader, %loop-memcpy-expansion2351
  %loop-index2352 = phi i64 [ %571, %loop-memcpy-expansion2351 ], [ 0, %loop-memcpy-expansion2351.preheader ]
  %scevgep3168 = getelementptr i8, ptr addrspace(4) %lsr.iv3166, i64 %loop-index2352, !amdgpu.uniform !42
  %569 = load <4 x i32>, ptr addrspace(4) %scevgep3168, align 1
  %570 = getelementptr i8, ptr %arrayidx.i22.i38.i.i.i.i296, i64 %loop-index2352
  store <4 x i32> %569, ptr %570, align 1
  %571 = add i64 %loop-index2352, 16
  %572 = icmp uge i64 %571, %567
  br i1 %572, label %Flow3809, label %loop-memcpy-expansion2351, !amdgpu.uniform !42

Flow3808:                                         ; preds = %Flow3807, %loop-memcpy-residual-header2354
  br label %Flow3811, !amdgpu.uniform !42

loop-memcpy-residual2353:                         ; preds = %loop-memcpy-residual2353.preheader, %loop-memcpy-residual2353
  %residual-loop-index2355 = phi i64 [ %574, %loop-memcpy-residual2353 ], [ 0, %loop-memcpy-residual2353.preheader ]
  %scevgep3177 = getelementptr i8, ptr addrspace(4) %scevgep3176, i64 %residual-loop-index2355, !amdgpu.uniform !42
  %573 = load i8, ptr addrspace(4) %scevgep3177, align 1
  %scevgep3175 = getelementptr i8, ptr %scevgep3174, i64 %residual-loop-index2355
  store i8 %573, ptr %scevgep3175, align 1
  %574 = add i64 %residual-loop-index2355, 1
  %575 = icmp uge i64 %574, %566
  br i1 %575, label %Flow3807, label %loop-memcpy-residual2353, !amdgpu.uniform !42

Flow3807:                                         ; preds = %loop-memcpy-residual2353
  br label %Flow3808, !amdgpu.uniform !42

_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i287: ; preds = %Flow3811
  %conv.i.i26.i.i.i.i288 = xor i32 %port.sroa.37.0.i.i276, 1
  fence release
  %576 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %arrayidx.i24.i29.i.i.i.i289 = getelementptr inbounds nuw i32, ptr %576, i64 %idxprom.i23.i.i.i.i.i272
  store atomic i32 %conv.i.i26.i.i.i.i288, ptr %arrayidx.i24.i29.i.i.i.i289 monotonic, align 4
  %add.i.i.i.i290 = add i64 %idx.0.i.i.i.i277, 64
  %scevgep3167 = getelementptr i8, ptr addrspace(4) %lsr.iv3166, i64 64
  %lsr.iv.next3172 = add i64 %lsr.iv3171, -64
  br label %Flow3814, !amdgpu.uniform !42

cond.false.i.i.i297:                              ; preds = %Flow3814
  %.lcssa4183 = phi i64 [ %563, %Flow3814 ]
  %port.sroa.37.0.i.i276.lcssa = phi i32 [ %port.sroa.37.0.i.i276, %Flow3814 ]
  %.lcssa4182 = phi i32 [ %549, %Flow3814 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa4183)
  %577 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %.lcssa4182)
  %578 = icmp ne i32 %577, %port.sroa.37.0.i.i276.lcssa
  %579 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %578)
  %580 = extractvalue { i1, i64 } %579, 0
  %581 = extractvalue { i1, i64 } %579, 1
  br i1 %580, label %while.body.i.i.i.i298.preheader, label %Flow3806

while.body.i.i.i.i298.preheader:                  ; preds = %cond.false.i.i.i297
  br label %while.body.i.i.i.i298, !amdgpu.uniform !42

Flow3806:                                         ; preds = %Flow3805, %cond.false.i.i.i297
  call void @llvm.amdgcn.end.cf.i64(i64 %581)
  br label %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit319, !amdgpu.uniform !42

while.body.i.i.i.i298:                            ; preds = %while.body.i.i.i.i298.preheader, %while.body.i.i.i.i298
  %phi.broken3924 = phi i64 [ 0, %while.body.i.i.i.i298.preheader ], [ %585, %while.body.i.i.i.i298 ]
  tail call void @llvm.amdgcn.s.sleep(i32 2)
  %582 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i8.i.i299 = getelementptr inbounds nuw i32, ptr %582, i64 %idxprom.i23.i.i.i.i.i272, !amdgpu.uniform !42
  %583 = load atomic i32, ptr %arrayidx.i.i.i8.i.i299 monotonic, align 4
  %584 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %583)
  %cmp.i.not.i.i.i.i300 = icmp eq i32 %584, %port.sroa.37.0.i.i276.lcssa
  %585 = call i64 @llvm.amdgcn.if.break.i64(i1 %cmp.i.not.i.i.i.i300, i64 %phi.broken3924)
  %586 = call i1 @llvm.amdgcn.loop.i64(i64 %585)
  br i1 %586, label %Flow3805, label %while.body.i.i.i.i298

Flow3805:                                         ; preds = %while.body.i.i.i.i298
  %.lcssa4181 = phi i64 [ %585, %while.body.i.i.i.i298 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa4181)
  br label %Flow3806, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit319: ; preds = %Flow3806
  fence acquire
  tail call void @llvm.amdgcn.wave.barrier()
  fence syncscope("agent") release
  %587 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %.lcssa4188, i1 true)
  %iszero.i.i.i.i.i13.i.i302 = icmp ne i64 %.lcssa4188, 0
  %cmp2.i.i.i.i14.i.i303 = icmp eq i64 %587, %sh_prom.i.i.i.i
  %cmp.i.i.i.i15.i.i304 = select i1 %iszero.i.i.i.i.i13.i.i302, i1 %cmp2.i.i.i.i14.i.i303, i1 false
  %588 = and i32 %.lcssa4189, 31
  %conv4.i.i.i19.i.i308 = zext i1 %cmp.i.i.i.i15.i.i304 to i32
  %shl.i.i.i20.i.i309 = shl nuw i32 %conv4.i.i.i19.i.i308, %588
  %xor.i.i.i21.i.i310 = xor i32 %shl.i.i.i20.i.i309, -1
  %589 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2471, !amdgpu.uniform !42

590:                                              ; preds = %ComputeEnd2472
  %div8.i.i.i16.i.i305 = lshr i32 %.lcssa4189, 5
  %idxprom.i.i.i17.i.i306 = zext nneg i32 %div8.i.i.i16.i.i305 to i64
  %591 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i17.i.i306
  %sunkaddr3296 = getelementptr inbounds i8, ptr addrspace(1) %591, i64 40
  %592 = atomicrmw and ptr addrspace(1) %sunkaddr3296, i32 %.lcssa4180 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %593, !amdgpu.uniform !42

593:                                              ; preds = %590, %ComputeEnd2472
  call void @llvm.amdgcn.end.cf.i64(i64 %3183)
  tail call void @llvm.amdgcn.wave.barrier()
  %inc.i.i.i32.i = add nuw i64 %spec.select.i, 1
  %cmp2.i.i.i35.i = icmp ult i64 %spec.select.i, 1676976733973595600
  br i1 %cmp2.i.i.i35.i, label %if.then3.i.i.i44.i, label %if.end4.i.i.i36.i, !amdgpu.uniform !42

loop-memcpy-expansion2297.preheader:              ; preds = %if.end4.i.i.i36.i
  br label %loop-memcpy-expansion2297, !amdgpu.uniform !42

Flow3804:                                         ; preds = %Flow3803, %if.end4.i.i.i36.i
  br label %loop-memcpy-residual-header2300, !amdgpu.uniform !42

loop-memcpy-expansion2297:                        ; preds = %loop-memcpy-expansion2297.preheader, %loop-memcpy-expansion2297
  %loop-index2298 = phi i64 [ %601, %loop-memcpy-expansion2297 ], [ 0, %loop-memcpy-expansion2297.preheader ]
  %598 = getelementptr i8, ptr %3, i64 %loop-index2298, !amdgpu.uniform !42
  %599 = load <4 x i32>, ptr %598, align 1
  %600 = getelementptr i8, ptr %594, i64 %loop-index2298
  store <4 x i32> %599, ptr %600, align 1
  %601 = add i64 %loop-index2298, 16
  %602 = icmp uge i64 %601, %596
  br i1 %602, label %Flow3803, label %loop-memcpy-expansion2297, !amdgpu.uniform !42

Flow3802:                                         ; preds = %Flow3801, %loop-memcpy-residual-header2300
  br label %post-loop-memcpy-expansion2296, !amdgpu.uniform !42

loop-memcpy-residual2299:                         ; preds = %loop-memcpy-residual2299.preheader, %loop-memcpy-residual2299
  %residual-loop-index2301 = phi i64 [ %604, %loop-memcpy-residual2299 ], [ 0, %loop-memcpy-residual2299.preheader ]
  %scevgep3165 = getelementptr i8, ptr %scevgep3164, i64 %residual-loop-index2301, !amdgpu.uniform !42
  %603 = load i8, ptr %scevgep3165, align 1
  %scevgep3163 = getelementptr i8, ptr %scevgep3162, i64 %residual-loop-index2301
  store i8 %603, ptr %scevgep3163, align 1
  %604 = add i64 %residual-loop-index2301, 1
  %605 = icmp uge i64 %604, %595
  br i1 %605, label %Flow3801, label %loop-memcpy-residual2299, !amdgpu.uniform !42

Flow3801:                                         ; preds = %loop-memcpy-residual2299
  br label %Flow3802, !amdgpu.uniform !42

post-loop-memcpy-expansion2296:                   ; preds = %Flow3802
  %sub.ptr.lhs.cast.i.i323 = ptrtoint ptr %add.ptr.i.i322 to i64
  %sub.ptr.sub.i.i324 = sub i64 %sub.ptr.lhs.cast.i.i323, ptrtoint (ptr addrspacecast (ptr addrspace(1) @_ZN12_GLOBAL__N_16memoryE to ptr) to i64)
  %cmp.i.i325 = icmp ugt i64 %sub.ptr.sub.i.i324, 65335
  %cond.i.i326 = select i1 %cmp.i.i325, ptr null, ptr %594
  %cmp.not.i10.i.i.i = icmp ne ptr %cond.i.i326, addrspacecast (ptr addrspace(4) @_ZN22__llvm_libc_22_0_0_git3cpp6string14NULL_CHARACTERE to ptr)
  br i1 %cmp.not.i10.i.i.i, label %if.then.i.i.i42.i, label %for.cond.i.i.i328.preheader, !amdgpu.uniform !42

if.then.i.i.i42.i:                                ; preds = %post-loop-memcpy-expansion2296
  %arrayidx.i.i.i43.i = getelementptr inbounds nuw i8, ptr %594, i64 %spec.select.i
  store i8 0, ptr %arrayidx.i.i.i43.i, align 1, !tbaa !57
  br label %for.cond.i.i.i328.preheader, !amdgpu.uniform !42

for.cond.i.i.i328.preheader:                      ; preds = %if.then.i.i.i42.i, %post-loop-memcpy-expansion2296
  br label %for.cond.i.i.i328, !amdgpu.uniform !42

for.cond.i.i.i328:                                ; preds = %for.cond.i.i.i328.preheader, %Flow3800
  %606 = phi i32 [ %644, %Flow3800 ], [ poison, %for.cond.i.i.i328.preheader ]
  %607 = phi i64 [ %645, %Flow3800 ], [ poison, %for.cond.i.i.i328.preheader ]
  %608 = phi i32 [ %646, %Flow3800 ], [ poison, %for.cond.i.i.i328.preheader ]
  %retval.sroa.6.0.i.i.i329 = phi i32 [ %649, %Flow3800 ], [ undef, %for.cond.i.i.i328.preheader ]
  %retval.sroa.2.0.i.i.i331 = phi i64 [ %648, %Flow3800 ], [ undef, %for.cond.i.i.i328.preheader ]
  %retval.sroa.8.0.i.i.i332 = phi i32 [ %647, %Flow3800 ], [ undef, %for.cond.i.i.i328.preheader ]
  %index.0.i.i.i333 = phi i32 [ %651, %Flow3800 ], [ 0, %for.cond.i.i.i328.preheader ]
  %609 = load i32, ptr addrspace(1) @__llvm_rpc_client, align 8, !tbaa !68
  %cmp.not.i.i.i334 = icmp ult i32 %index.0.i.i.i333, %609
  %spec.store.select.i.i.i335 = select i1 %cmp.not.i.i.i334, i32 %index.0.i.i.i333, i32 0
  %610 = tail call noundef i64 @llvm.amdgcn.ballot.i64(i1 true)
  %611 = and i32 %spec.store.select.i.i.i335, 31
  %612 = lshr i64 %610, %sh_prom.i.i.i.i
  %613 = trunc i64 %612 to i32
  %conv4.i.i.i.i.i339 = and i32 %613, 1
  %shl.i.i.i.i.i340 = shl nuw i32 %conv4.i.i.i.i.i339, %611
  %614 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2475, !amdgpu.uniform !42

615:                                              ; preds = %ComputeEnd2476
  %sunkaddr3297 = getelementptr inbounds i8, ptr addrspace(1) %3193, i64 40
  %616 = atomicrmw or ptr addrspace(1) %sunkaddr3297, i32 %.lcssa4174 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %617, !amdgpu.uniform !42

617:                                              ; preds = %615, %ComputeEnd2476
  %618 = phi i32 [ %616, %615 ], [ poison, %ComputeEnd2476 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %3202)
  %619 = call i32 @llvm.amdgcn.readfirstlane.i32(i32 %618)
  %620 = or i32 %619, %.lcssa4175
  %shl5.i.i.i.i.i341 = shl nuw i32 1, %611
  %and.i.i.i.i.i342 = and i32 %shl5.i.i.i.i.i341, %620
  %tobool3.i.i.i.i343 = icmp ne i32 %and.i.i.i.i.i342, 0
  %621 = tail call i64 @llvm.amdgcn.ballot.i64(i1 %tobool3.i.i.i.i343)
  %and.i.i.i.i.i.i344 = and i64 %621, %610
  %cmp.i.not.i.i.i345 = icmp ne i64 %610, %and.i.i.i.i.i.i344
  br i1 %cmp.i.not.i.i.i345, label %if.then.i.i.i.i346, label %Flow3799, !amdgpu.uniform !42

633:                                              ; preds = %ComputeEnd2481
  %sunkaddr3298 = getelementptr inbounds i8, ptr addrspace(1) %3193, i64 40
  %634 = atomicrmw and ptr addrspace(1) %sunkaddr3298, i32 %.lcssa4176 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %635, !amdgpu.uniform !42

635:                                              ; preds = %633, %ComputeEnd2481
  call void @llvm.amdgcn.end.cf.i64(i64 %3220)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %Flow3797, !amdgpu.uniform !42

if.end14.i.i.i417:                                ; preds = %Flow3797
  %636 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %610, i1 true)
  %iszero.i.i.i.i.i.i418 = icmp ne i64 %610, 0
  %cmp2.i.i.i.i.i419 = icmp eq i64 %636, %sh_prom.i.i.i.i
  %cmp.i.i.i.i.i420 = select i1 %iszero.i.i.i.i.i.i418, i1 %cmp2.i.i.i.i.i419, i1 false
  %637 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.i.i.i.i.i420)
  %638 = extractvalue { i1, i64 } %637, 0
  %639 = extractvalue { i1, i64 } %637, 1
  br i1 %638, label %if.then16.i.i.i422, label %if.end22.i.i.i421

Flow3798:                                         ; preds = %if.end22.i.i.i421, %Flow3797
  %641 = phi i32 [ %628, %if.end22.i.i.i421 ], [ %retval.sroa.8.0.i.i.i332, %Flow3797 ]
  %642 = phi i64 [ %610, %if.end22.i.i.i421 ], [ %retval.sroa.2.0.i.i.i331, %Flow3797 ]
  %643 = phi i32 [ %spec.store.select.i.i.i335, %if.end22.i.i.i421 ], [ %retval.sroa.6.0.i.i.i329, %Flow3797 ]
  br label %cleanup26.i.i.i358, !amdgpu.uniform !42

Flow3799:                                         ; preds = %cleanup26.i.i.i358, %617
  %644 = phi i32 [ %643, %cleanup26.i.i.i358 ], [ %606, %617 ]
  %645 = phi i64 [ %642, %cleanup26.i.i.i358 ], [ %607, %617 ]
  %646 = phi i32 [ %641, %cleanup26.i.i.i358 ], [ %608, %617 ]
  %647 = phi i32 [ %641, %cleanup26.i.i.i358 ], [ %retval.sroa.8.0.i.i.i332, %617 ]
  %648 = phi i64 [ %642, %cleanup26.i.i.i358 ], [ %retval.sroa.2.0.i.i.i331, %617 ]
  %649 = phi i32 [ %643, %cleanup26.i.i.i358 ], [ %retval.sroa.6.0.i.i.i329, %617 ]
  %650 = phi i1 [ %cmp.i46.not.i.i.i350, %cleanup26.i.i.i358 ], [ true, %617 ]
  br i1 %650, label %for.inc.i.i.i363, label %Flow3800, !amdgpu.uniform !42

cleanup26.i.i.i358:                               ; preds = %Flow3798
  br label %Flow3799, !amdgpu.uniform !42

for.inc.i.i.i363:                                 ; preds = %Flow3799
  %inc.i.i.i368 = add i32 %spec.store.select.i.i.i335, 1
  br label %Flow3800, !amdgpu.uniform !42

Flow3800:                                         ; preds = %for.inc.i.i.i363, %Flow3799
  %651 = phi i32 [ %inc.i.i.i368, %for.inc.i.i.i363 ], [ poison, %Flow3799 ]
  %652 = phi i1 [ false, %for.inc.i.i.i363 ], [ true, %Flow3799 ]
  br i1 %652, label %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i369, label %for.cond.i.i.i328, !amdgpu.uniform !42

_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i369: ; preds = %Flow3800
  %.lcssa4179 = phi i32 [ %644, %Flow3800 ]
  %.lcssa4178 = phi i64 [ %645, %Flow3800 ]
  %.lcssa4177 = phi i32 [ %646, %Flow3800 ]
  fence acquire
  %653 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %mul.i.i.i.i.i.i370 = shl i32 %.lcssa4179, 6
  %idxprom.i19.i.i.i.i.i371 = zext i32 %mul.i.i.i.i.i.i370 to i64
  %arrayidx.i20.i.i.i.i.i372 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %653, i64 %idxprom.i19.i.i.i.i.i371
  %arrayidx.i22.i.i.i.i.i373 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %arrayidx.i20.i.i.i.i.i372, i64 %sh_prom.i.i.i.i
  store i64 %spec.select.i, ptr %arrayidx.i22.i.i.i.i.i373, align 8, !tbaa !82
  %spec.select.i.i.i.i.i.i.i374 = tail call i64 @llvm.umin.i64(i64 %spec.select.i, i64 56)
  %654 = and i64 %spec.select.i.i.i.i.i.i.i374, 15
  %655 = sub i64 %spec.select.i.i.i.i.i.i.i374, %654
  %656 = icmp ne i64 %655, 0
  br i1 %656, label %loop-memcpy-expansion2267.preheader, label %Flow3796, !amdgpu.uniform !42

loop-memcpy-expansion2267.preheader:              ; preds = %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i369
  %657 = add nuw nsw i64 %75, 8
  %scevgep3159 = getelementptr i8, ptr %653, i64 %657
  %658 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i371, 6
  %scevgep3160 = getelementptr i8, ptr %scevgep3159, i64 %658
  br label %loop-memcpy-expansion2267, !amdgpu.uniform !42

Flow3796:                                         ; preds = %Flow3795, %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i369
  br label %loop-memcpy-residual-header2270, !amdgpu.uniform !42

loop-memcpy-expansion2267:                        ; preds = %loop-memcpy-expansion2267.preheader, %loop-memcpy-expansion2267
  %loop-index2268 = phi i64 [ %661, %loop-memcpy-expansion2267 ], [ 0, %loop-memcpy-expansion2267.preheader ]
  %659 = getelementptr i8, ptr %594, i64 %loop-index2268, !amdgpu.uniform !42
  %660 = load <4 x i32>, ptr %659, align 1
  %scevgep3161 = getelementptr i8, ptr %scevgep3160, i64 %loop-index2268
  store <4 x i32> %660, ptr %scevgep3161, align 1
  %661 = add i64 %loop-index2268, 16
  %662 = icmp uge i64 %661, %655
  br i1 %662, label %Flow3795, label %loop-memcpy-expansion2267, !amdgpu.uniform !42

Flow3794:                                         ; preds = %Flow3792, %Flow3793
  %663 = phi i64 [ %2839, %Flow3792 ], [ %2835, %Flow3793 ]
  br label %post-loop-memcpy-expansion2266, !amdgpu.uniform !42

loop-memcpy-residual2269:                         ; preds = %loop-memcpy-residual2269.preheader, %loop-memcpy-residual2269
  %residual-loop-index2271 = phi i64 [ %665, %loop-memcpy-residual2269 ], [ 0, %loop-memcpy-residual2269.preheader ]
  %scevgep3158 = getelementptr i8, ptr %scevgep3157, i64 %residual-loop-index2271, !amdgpu.uniform !42
  %664 = load i8, ptr %scevgep3158, align 1
  %scevgep3156 = getelementptr i8, ptr %scevgep3155, i64 %residual-loop-index2271
  store i8 %664, ptr %scevgep3156, align 1
  %665 = add i64 %residual-loop-index2271, 1
  %666 = icmp uge i64 %665, %654
  br i1 %666, label %Flow3792, label %loop-memcpy-residual2269, !amdgpu.uniform !42

Flow3792:                                         ; preds = %loop-memcpy-residual2269
  br label %Flow3794, !amdgpu.uniform !42

post-loop-memcpy-expansion2266:                   ; preds = %Flow3794
  %tobool.not.i.i.i.i.i.i376 = icmp eq i32 %.lcssa4177, 0
  %conv.i.i.i.i.i.i377 = zext i1 %tobool.not.i.i.i.i.i.i376 to i32
  fence release
  %667 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %idxprom.i23.i.i.i.i.i378 = zext i32 %.lcssa4179 to i64
  %arrayidx.i24.i.i.i.i.i379 = getelementptr inbounds nuw i32, ptr %667, i64 %idxprom.i23.i.i.i.i.i378
  store atomic i32 %conv.i.i.i.i.i.i377, ptr %arrayidx.i24.i.i.i.i.i379 monotonic, align 4
  %668 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !84
  %arrayidx.i.i5.i.i380 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %668, i64 %idxprom.i23.i.i.i.i.i378, !amdgpu.uniform !42
  %669 = load i64, ptr %arrayidx.i.i5.i.i380, align 8, !tbaa !80
  %scevgep3140 = getelementptr i8, ptr %594, i64 56
  %umin3146 = call i64 @llvm.umin.i64(i64 sub (i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) getelementptr inbounds nuw (i8, ptr addrspace(4) @.str.9, i64 8) to ptr) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.9 to ptr) to i64)), i64 2)
  %670 = sub i64 sub (i64 add (i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) getelementptr inbounds nuw (i8, ptr addrspace(4) @.str.9, i64 8) to ptr) to i64), i64 -56), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.9 to ptr) to i64)), %umin3146
  br label %while.cond.i.i.i.i381, !amdgpu.uniform !42

while.cond.i.i.i.i381:                            ; preds = %post-loop-memcpy-expansion2266, %Flow3791
  %phi.broken3925 = phi i64 [ 0, %post-loop-memcpy-expansion2266 ], [ %687, %Flow3791 ]
  %lsr.iv3147 = phi i64 [ %685, %Flow3791 ], [ %670, %post-loop-memcpy-expansion2266 ]
  %lsr.iv3141 = phi ptr [ %684, %Flow3791 ], [ %scevgep3140, %post-loop-memcpy-expansion2266 ]
  %port.sroa.37.0.i.i382 = phi i32 [ %683, %Flow3791 ], [ %conv.i.i.i.i.i.i377, %post-loop-memcpy-expansion2266 ]
  %idx.0.i.i.i.i383 = phi i64 [ %682, %Flow3791 ], [ 56, %post-loop-memcpy-expansion2266 ]
  %cmp.i.i.i.i384 = icmp ugt i64 %spec.select.i, %idx.0.i.i.i.i383
  %671 = tail call i64 @llvm.amdgcn.ballot.i64(i1 %cmp.i.i.i.i384)
  %and.i.i.i.i6.i.i385 = and i64 %671, %669
  %tobool.not.i.i.i.i386 = icmp ne i64 %and.i.i.i.i6.i.i385, 0
  %672 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i22.i.i.i387 = getelementptr inbounds nuw i32, ptr %672, i64 %idxprom.i23.i.i.i.i.i378, !amdgpu.uniform !42
  %673 = load atomic i32, ptr %arrayidx.i22.i.i.i387 monotonic, align 4
  %674 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %tobool.not.i.i.i.i386)
  %675 = extractvalue { i1, i64 } %674, 0
  %676 = extractvalue { i1, i64 } %674, 1
  br i1 %675, label %cond.false.i8.i.i.i.i388, label %Flow3791

while.body.i.i21.i.i.i.i389.preheader:            ; preds = %cond.false.i8.i.i.i.i388
  br label %while.body.i.i21.i.i.i.i389, !amdgpu.uniform !42

Flow3790:                                         ; preds = %Flow3789, %cond.false.i8.i.i.i.i388
  br label %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i392, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i389:                      ; preds = %while.body.i.i21.i.i.i.i389.preheader, %while.body.i.i21.i.i.i.i389
  tail call void @llvm.amdgcn.s.sleep(i32 2)
  %679 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i22.i.i.i.i390 = getelementptr inbounds nuw i32, ptr %679, i64 %idxprom.i23.i.i.i.i.i378, !amdgpu.uniform !42
  %680 = load atomic i32, ptr %arrayidx.i.i.i22.i.i.i.i390 monotonic, align 4
  %681 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %680)
  %cmp.i.not.i.i23.i.i.i.i391 = icmp eq i32 %681, %port.sroa.37.0.i.i382
  br i1 %cmp.i.not.i.i23.i.i.i.i391, label %Flow3789, label %while.body.i.i21.i.i.i.i389, !amdgpu.uniform !42

Flow3789:                                         ; preds = %while.body.i.i21.i.i.i.i389
  br label %Flow3790, !amdgpu.uniform !42

Flow3791:                                         ; preds = %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i393, %while.cond.i.i.i.i381
  %682 = phi i64 [ %add.i.i.i.i396, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i393 ], [ poison, %while.cond.i.i.i.i381 ]
  %683 = phi i32 [ %conv.i.i26.i.i.i.i394, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i393 ], [ poison, %while.cond.i.i.i.i381 ]
  %684 = phi ptr [ %scevgep3142, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i393 ], [ poison, %while.cond.i.i.i.i381 ]
  %685 = phi i64 [ %lsr.iv.next3148, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i393 ], [ poison, %while.cond.i.i.i.i381 ]
  %686 = phi i1 [ false, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i393 ], [ true, %while.cond.i.i.i.i381 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %676)
  %687 = call i64 @llvm.amdgcn.if.break.i64(i1 %686, i64 %phi.broken3925)
  %688 = call i1 @llvm.amdgcn.loop.i64(i64 %687)
  br i1 %688, label %cond.false.i.i.i403, label %while.cond.i.i.i.i381

_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i392: ; preds = %Flow3790
  fence acquire
  br i1 %cmp.i.i.i.i384, label %if.then.i.i.i.i.i.i.i397, label %Flow3788, !amdgpu.uniform !42

loop-memcpy-expansion2261.preheader:              ; preds = %if.then.i.i.i.i.i.i.i397
  %arrayidx.i20.i35.i.i.i.i401 = getelementptr %"struct.rpc::Buffer", ptr %689, i64 %idxprom.i19.i.i.i.i.i371
  %arrayidx.i22.i38.i.i.i.i402 = getelementptr %"struct.rpc::Buffer", ptr %arrayidx.i20.i35.i.i.i.i401, i64 %sh_prom.i.i.i.i
  br label %loop-memcpy-expansion2261, !amdgpu.uniform !42

Flow3787:                                         ; preds = %Flow3786, %if.then.i.i.i.i.i.i.i397
  br label %loop-memcpy-residual-header2264, !amdgpu.uniform !42

loop-memcpy-expansion2261:                        ; preds = %loop-memcpy-expansion2261.preheader, %loop-memcpy-expansion2261
  %loop-index2262 = phi i64 [ %695, %loop-memcpy-expansion2261 ], [ 0, %loop-memcpy-expansion2261.preheader ]
  %scevgep3143 = getelementptr i8, ptr %lsr.iv3141, i64 %loop-index2262, !amdgpu.uniform !42
  %693 = load <4 x i32>, ptr %scevgep3143, align 1
  %694 = getelementptr i8, ptr %arrayidx.i22.i38.i.i.i.i402, i64 %loop-index2262
  store <4 x i32> %693, ptr %694, align 1
  %695 = add i64 %loop-index2262, 16
  %696 = icmp uge i64 %695, %691
  br i1 %696, label %Flow3786, label %loop-memcpy-expansion2261, !amdgpu.uniform !42

Flow3785:                                         ; preds = %Flow3784, %loop-memcpy-residual-header2264
  br label %Flow3788, !amdgpu.uniform !42

loop-memcpy-residual2263:                         ; preds = %loop-memcpy-residual2263.preheader, %loop-memcpy-residual2263
  %residual-loop-index2265 = phi i64 [ %698, %loop-memcpy-residual2263 ], [ 0, %loop-memcpy-residual2263.preheader ]
  %scevgep3153 = getelementptr i8, ptr %scevgep3152, i64 %residual-loop-index2265, !amdgpu.uniform !42
  %697 = load i8, ptr %scevgep3153, align 1
  %scevgep3151 = getelementptr i8, ptr %scevgep3150, i64 %residual-loop-index2265
  store i8 %697, ptr %scevgep3151, align 1
  %698 = add i64 %residual-loop-index2265, 1
  %699 = icmp uge i64 %698, %690
  br i1 %699, label %Flow3784, label %loop-memcpy-residual2263, !amdgpu.uniform !42

Flow3784:                                         ; preds = %loop-memcpy-residual2263
  br label %Flow3785, !amdgpu.uniform !42

_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i393: ; preds = %Flow3788
  %conv.i.i26.i.i.i.i394 = xor i32 %port.sroa.37.0.i.i382, 1
  fence release
  %700 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %arrayidx.i24.i29.i.i.i.i395 = getelementptr inbounds nuw i32, ptr %700, i64 %idxprom.i23.i.i.i.i.i378
  store atomic i32 %conv.i.i26.i.i.i.i394, ptr %arrayidx.i24.i29.i.i.i.i395 monotonic, align 4
  %add.i.i.i.i396 = add i64 %idx.0.i.i.i.i383, 64
  %scevgep3142 = getelementptr i8, ptr %lsr.iv3141, i64 64
  %lsr.iv.next3148 = add i64 %lsr.iv3147, -64
  br label %Flow3791, !amdgpu.uniform !42

cond.false.i.i.i403:                              ; preds = %Flow3791
  %.lcssa4173 = phi i64 [ %687, %Flow3791 ]
  %port.sroa.37.0.i.i382.lcssa = phi i32 [ %port.sroa.37.0.i.i382, %Flow3791 ]
  %.lcssa4172 = phi i32 [ %673, %Flow3791 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa4173)
  %701 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %.lcssa4172)
  %702 = icmp ne i32 %701, %port.sroa.37.0.i.i382.lcssa
  %703 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %702)
  %704 = extractvalue { i1, i64 } %703, 0
  %705 = extractvalue { i1, i64 } %703, 1
  br i1 %704, label %while.body.i.i.i.i404.preheader, label %Flow3783

while.body.i.i.i.i404.preheader:                  ; preds = %cond.false.i.i.i403
  br label %while.body.i.i.i.i404, !amdgpu.uniform !42

Flow3783:                                         ; preds = %Flow3782, %cond.false.i.i.i403
  call void @llvm.amdgcn.end.cf.i64(i64 %705)
  br label %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsINS_3cpp6stringEEERS1_T_.exit, !amdgpu.uniform !42

while.body.i.i.i.i404:                            ; preds = %while.body.i.i.i.i404.preheader, %while.body.i.i.i.i404
  %phi.broken3926 = phi i64 [ 0, %while.body.i.i.i.i404.preheader ], [ %709, %while.body.i.i.i.i404 ]
  tail call void @llvm.amdgcn.s.sleep(i32 2)
  %706 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i8.i.i405 = getelementptr inbounds nuw i32, ptr %706, i64 %idxprom.i23.i.i.i.i.i378, !amdgpu.uniform !42
  %707 = load atomic i32, ptr %arrayidx.i.i.i8.i.i405 monotonic, align 4
  %708 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %707)
  %cmp.i.not.i.i.i.i406 = icmp eq i32 %708, %port.sroa.37.0.i.i382.lcssa
  %709 = call i64 @llvm.amdgcn.if.break.i64(i1 %cmp.i.not.i.i.i.i406, i64 %phi.broken3926)
  %710 = call i1 @llvm.amdgcn.loop.i64(i64 %709)
  br i1 %710, label %Flow3782, label %while.body.i.i.i.i404

Flow3782:                                         ; preds = %while.body.i.i.i.i404
  %.lcssa4171 = phi i64 [ %709, %while.body.i.i.i.i404 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa4171)
  br label %Flow3783, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsINS_3cpp6stringEEERS1_T_.exit: ; preds = %Flow3783
  fence acquire
  tail call void @llvm.amdgcn.wave.barrier()
  fence syncscope("agent") release
  %711 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %.lcssa4178, i1 true)
  %iszero.i.i.i.i.i13.i.i408 = icmp ne i64 %.lcssa4178, 0
  %cmp2.i.i.i.i14.i.i409 = icmp eq i64 %711, %sh_prom.i.i.i.i
  %cmp.i.i.i.i15.i.i410 = select i1 %iszero.i.i.i.i.i13.i.i408, i1 %cmp2.i.i.i.i14.i.i409, i1 false
  %712 = and i32 %.lcssa4179, 31
  %conv4.i.i.i19.i.i414 = zext i1 %cmp.i.i.i.i15.i.i410 to i32
  %shl.i.i.i20.i.i415 = shl nuw i32 %conv4.i.i.i19.i.i414, %712
  %xor.i.i.i21.i.i416 = xor i32 %shl.i.i.i20.i.i415, -1
  %713 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2484, !amdgpu.uniform !42

714:                                              ; preds = %ComputeEnd2485
  %div8.i.i.i16.i.i411 = lshr i32 %.lcssa4179, 5
  %idxprom.i.i.i17.i.i412 = zext nneg i32 %div8.i.i.i16.i.i411 to i64
  %715 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i17.i.i412
  %sunkaddr3299 = getelementptr inbounds i8, ptr addrspace(1) %715, i64 40
  %716 = atomicrmw and ptr addrspace(1) %sunkaddr3299, i32 %.lcssa4170 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %717, !amdgpu.uniform !42

717:                                              ; preds = %714, %ComputeEnd2485
  call void @llvm.amdgcn.end.cf.i64(i64 %3237)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %for.cond.i.i1.i433, !amdgpu.uniform !42

for.cond.i.i1.i433:                               ; preds = %717, %Flow3781
  %718 = phi i32 [ %756, %Flow3781 ], [ poison, %717 ]
  %719 = phi i64 [ %757, %Flow3781 ], [ poison, %717 ]
  %720 = phi i32 [ %758, %Flow3781 ], [ poison, %717 ]
  %retval.sroa.6.0.i.i.i434 = phi i32 [ %761, %Flow3781 ], [ undef, %717 ]
  %retval.sroa.2.0.i.i.i436 = phi i64 [ %760, %Flow3781 ], [ undef, %717 ]
  %retval.sroa.8.0.i.i.i437 = phi i32 [ %759, %Flow3781 ], [ undef, %717 ]
  %index.0.i.i.i438 = phi i32 [ %763, %Flow3781 ], [ 0, %717 ]
  %721 = load i32, ptr addrspace(1) @__llvm_rpc_client, align 8, !tbaa !68
  %cmp.not.i.i.i439 = icmp ult i32 %index.0.i.i.i438, %721
  %spec.store.select.i.i.i440 = select i1 %cmp.not.i.i.i439, i32 %index.0.i.i.i438, i32 0
  %722 = tail call noundef i64 @llvm.amdgcn.ballot.i64(i1 true)
  %723 = and i32 %spec.store.select.i.i.i440, 31
  %724 = lshr i64 %722, %sh_prom.i.i.i.i
  %725 = trunc i64 %724 to i32
  %conv4.i.i.i.i.i444 = and i32 %725, 1
  %shl.i.i.i.i.i445 = shl nuw i32 %conv4.i.i.i.i.i444, %723
  %726 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2488, !amdgpu.uniform !42

727:                                              ; preds = %ComputeEnd2489
  %sunkaddr3300 = getelementptr inbounds i8, ptr addrspace(1) %3247, i64 40
  %728 = atomicrmw or ptr addrspace(1) %sunkaddr3300, i32 %.lcssa4164 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %729, !amdgpu.uniform !42

729:                                              ; preds = %727, %ComputeEnd2489
  %730 = phi i32 [ %728, %727 ], [ poison, %ComputeEnd2489 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %3256)
  %731 = call i32 @llvm.amdgcn.readfirstlane.i32(i32 %730)
  %732 = or i32 %731, %.lcssa4165
  %shl5.i.i.i.i.i446 = shl nuw i32 1, %723
  %and.i.i.i.i.i447 = and i32 %shl5.i.i.i.i.i446, %732
  %tobool3.i.i.i.i448 = icmp ne i32 %and.i.i.i.i.i447, 0
  %733 = tail call i64 @llvm.amdgcn.ballot.i64(i1 %tobool3.i.i.i.i448)
  %and.i.i.i.i.i.i449 = and i64 %733, %722
  %cmp.i.not.i.i.i450 = icmp ne i64 %722, %and.i.i.i.i.i.i449
  br i1 %cmp.i.not.i.i.i450, label %if.then.i.i.i.i451, label %Flow3780, !amdgpu.uniform !42

745:                                              ; preds = %ComputeEnd2494
  %sunkaddr3301 = getelementptr inbounds i8, ptr addrspace(1) %3247, i64 40
  %746 = atomicrmw and ptr addrspace(1) %sunkaddr3301, i32 %.lcssa4166 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %747, !amdgpu.uniform !42

747:                                              ; preds = %745, %ComputeEnd2494
  call void @llvm.amdgcn.end.cf.i64(i64 %3274)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %Flow3778, !amdgpu.uniform !42

if.end14.i.i.i522:                                ; preds = %Flow3778
  %748 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %722, i1 true)
  %iszero.i.i.i.i.i.i523 = icmp ne i64 %722, 0
  %cmp2.i.i.i.i.i524 = icmp eq i64 %748, %sh_prom.i.i.i.i
  %cmp.i.i.i.i.i525 = select i1 %iszero.i.i.i.i.i.i523, i1 %cmp2.i.i.i.i.i524, i1 false
  %749 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.i.i.i.i.i525)
  %750 = extractvalue { i1, i64 } %749, 0
  %751 = extractvalue { i1, i64 } %749, 1
  br i1 %750, label %if.then16.i.i.i527, label %if.end22.i.i.i526

Flow3779:                                         ; preds = %if.end22.i.i.i526, %Flow3778
  %753 = phi i32 [ %740, %if.end22.i.i.i526 ], [ %retval.sroa.8.0.i.i.i437, %Flow3778 ]
  %754 = phi i64 [ %722, %if.end22.i.i.i526 ], [ %retval.sroa.2.0.i.i.i436, %Flow3778 ]
  %755 = phi i32 [ %spec.store.select.i.i.i440, %if.end22.i.i.i526 ], [ %retval.sroa.6.0.i.i.i434, %Flow3778 ]
  br label %cleanup26.i.i.i463, !amdgpu.uniform !42

Flow3780:                                         ; preds = %cleanup26.i.i.i463, %729
  %756 = phi i32 [ %755, %cleanup26.i.i.i463 ], [ %718, %729 ]
  %757 = phi i64 [ %754, %cleanup26.i.i.i463 ], [ %719, %729 ]
  %758 = phi i32 [ %753, %cleanup26.i.i.i463 ], [ %720, %729 ]
  %759 = phi i32 [ %753, %cleanup26.i.i.i463 ], [ %retval.sroa.8.0.i.i.i437, %729 ]
  %760 = phi i64 [ %754, %cleanup26.i.i.i463 ], [ %retval.sroa.2.0.i.i.i436, %729 ]
  %761 = phi i32 [ %755, %cleanup26.i.i.i463 ], [ %retval.sroa.6.0.i.i.i434, %729 ]
  %762 = phi i1 [ %cmp.i46.not.i.i.i455, %cleanup26.i.i.i463 ], [ true, %729 ]
  br i1 %762, label %for.inc.i.i.i468, label %Flow3781, !amdgpu.uniform !42

cleanup26.i.i.i463:                               ; preds = %Flow3779
  br label %Flow3780, !amdgpu.uniform !42

for.inc.i.i.i468:                                 ; preds = %Flow3780
  %inc.i.i.i473 = add i32 %spec.store.select.i.i.i440, 1
  br label %Flow3781, !amdgpu.uniform !42

Flow3781:                                         ; preds = %for.inc.i.i.i468, %Flow3780
  %763 = phi i32 [ %inc.i.i.i473, %for.inc.i.i.i468 ], [ poison, %Flow3780 ]
  %764 = phi i1 [ false, %for.inc.i.i.i468 ], [ true, %Flow3780 ]
  br i1 %764, label %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i474, label %for.cond.i.i1.i433, !amdgpu.uniform !42

_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i474: ; preds = %Flow3781
  %.lcssa4169 = phi i32 [ %756, %Flow3781 ]
  %.lcssa4168 = phi i64 [ %757, %Flow3781 ]
  %.lcssa4167 = phi i32 [ %758, %Flow3781 ]
  fence acquire
  %765 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %mul.i.i.i.i.i.i475 = shl i32 %.lcssa4169, 6
  %idxprom.i19.i.i.i.i.i476 = zext i32 %mul.i.i.i.i.i.i475 to i64
  %arrayidx.i20.i.i.i.i.i477 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %765, i64 %idxprom.i19.i.i.i.i.i476
  %arrayidx.i22.i.i.i.i.i478 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %arrayidx.i20.i.i.i.i.i477, i64 %sh_prom.i.i.i.i
  store i64 sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.34.12 to ptr), i64 10) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.34.12 to ptr) to i64)), ptr %arrayidx.i22.i.i.i.i.i478, align 8, !tbaa !82
  %spec.select.i.i.i.i.i.i.i479 = tail call i64 @llvm.umin.i64(i64 sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.34.12 to ptr), i64 10) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.34.12 to ptr) to i64)), i64 56)
  %766 = and i64 %spec.select.i.i.i.i.i.i.i479, 15
  %767 = sub i64 %spec.select.i.i.i.i.i.i.i479, %766
  %768 = icmp ne i64 %767, 0
  br i1 %768, label %loop-memcpy-expansion2357.preheader, label %Flow3777, !amdgpu.uniform !42

loop-memcpy-expansion2357.preheader:              ; preds = %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i474
  %769 = add nuw nsw i64 %75, 8
  %scevgep3137 = getelementptr i8, ptr %765, i64 %769
  %770 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i476, 6
  %scevgep3138 = getelementptr i8, ptr %scevgep3137, i64 %770
  br label %loop-memcpy-expansion2357, !amdgpu.uniform !42

Flow3777:                                         ; preds = %Flow3776, %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i474
  br label %loop-memcpy-residual-header2360, !amdgpu.uniform !42

loop-memcpy-expansion2357:                        ; preds = %loop-memcpy-expansion2357.preheader, %loop-memcpy-expansion2357
  %loop-index2358 = phi i64 [ %773, %loop-memcpy-expansion2357 ], [ 0, %loop-memcpy-expansion2357.preheader ]
  %771 = getelementptr i8, ptr addrspace(4) @.str.34.12, i64 %loop-index2358, !amdgpu.uniform !42
  %772 = load <4 x i32>, ptr addrspace(4) %771, align 1
  %scevgep3139 = getelementptr i8, ptr %scevgep3138, i64 %loop-index2358
  store <4 x i32> %772, ptr %scevgep3139, align 1
  %773 = add i64 %loop-index2358, 16
  %774 = icmp uge i64 %773, %767
  br i1 %774, label %Flow3776, label %loop-memcpy-expansion2357, !amdgpu.uniform !42

Flow3775:                                         ; preds = %Flow3773, %Flow3774
  %775 = phi i64 [ %2922, %Flow3773 ], [ %2919, %Flow3774 ]
  br label %post-loop-memcpy-expansion2356, !amdgpu.uniform !42

loop-memcpy-residual2359:                         ; preds = %loop-memcpy-residual2359.preheader, %loop-memcpy-residual2359
  %residual-loop-index2361 = phi i64 [ %777, %loop-memcpy-residual2359 ], [ 0, %loop-memcpy-residual2359.preheader ]
  %scevgep3136 = getelementptr i8, ptr addrspace(4) @.str.34.12, i64 %residual-loop-index2361, !amdgpu.uniform !42
  %776 = load i8, ptr addrspace(4) %scevgep3136, align 1
  %scevgep3135 = getelementptr i8, ptr %scevgep3134, i64 %residual-loop-index2361
  store i8 %776, ptr %scevgep3135, align 1
  %777 = add i64 %residual-loop-index2361, 1
  %778 = icmp uge i64 %777, %766
  br i1 %778, label %Flow3773, label %loop-memcpy-residual2359, !amdgpu.uniform !42

Flow3773:                                         ; preds = %loop-memcpy-residual2359
  br label %Flow3775, !amdgpu.uniform !42

post-loop-memcpy-expansion2356:                   ; preds = %Flow3775
  %tobool.not.i.i.i.i.i.i481 = icmp eq i32 %.lcssa4167, 0
  %conv.i.i.i.i.i.i482 = zext i1 %tobool.not.i.i.i.i.i.i481 to i32
  fence release
  %779 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %idxprom.i23.i.i.i.i.i483 = zext i32 %.lcssa4169 to i64
  %arrayidx.i24.i.i.i.i.i484 = getelementptr inbounds nuw i32, ptr %779, i64 %idxprom.i23.i.i.i.i.i483
  store atomic i32 %conv.i.i.i.i.i.i482, ptr %arrayidx.i24.i.i.i.i.i484 monotonic, align 4
  %780 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !84
  %arrayidx.i.i5.i.i485 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %780, i64 %idxprom.i23.i.i.i.i.i483, !amdgpu.uniform !42
  %781 = load i64, ptr %arrayidx.i.i5.i.i485, align 8, !tbaa !80
  br label %while.cond.i.i.i.i486, !amdgpu.uniform !42

while.cond.i.i.i.i486:                            ; preds = %post-loop-memcpy-expansion2356, %Flow3772
  %phi.broken3927 = phi i64 [ 0, %post-loop-memcpy-expansion2356 ], [ %798, %Flow3772 ]
  %lsr.iv3126 = phi i64 [ %796, %Flow3772 ], [ -46, %post-loop-memcpy-expansion2356 ]
  %lsr.iv3121 = phi ptr addrspace(4) [ %795, %Flow3772 ], [ getelementptr (i8, ptr addrspace(4) @.str.34.12, i64 56), %post-loop-memcpy-expansion2356 ]
  %port.sroa.37.0.i.i487 = phi i32 [ %794, %Flow3772 ], [ %conv.i.i.i.i.i.i482, %post-loop-memcpy-expansion2356 ]
  %idx.0.i.i.i.i488 = phi i64 [ %793, %Flow3772 ], [ 56, %post-loop-memcpy-expansion2356 ]
  %cmp.i.i.i.i489 = icmp ult i64 %idx.0.i.i.i.i488, sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.34.12 to ptr), i64 10) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.34.12 to ptr) to i64))
  %782 = tail call i64 @llvm.amdgcn.ballot.i64(i1 %cmp.i.i.i.i489)
  %and.i.i.i.i6.i.i490 = and i64 %782, %781
  %tobool.not.i.i.i.i491 = icmp ne i64 %and.i.i.i.i6.i.i490, 0
  %783 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i22.i.i.i492 = getelementptr inbounds nuw i32, ptr %783, i64 %idxprom.i23.i.i.i.i.i483, !amdgpu.uniform !42
  %784 = load atomic i32, ptr %arrayidx.i22.i.i.i492 monotonic, align 4
  %785 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %tobool.not.i.i.i.i491)
  %786 = extractvalue { i1, i64 } %785, 0
  %787 = extractvalue { i1, i64 } %785, 1
  br i1 %786, label %cond.false.i8.i.i.i.i493, label %Flow3772

while.body.i.i21.i.i.i.i494.preheader:            ; preds = %cond.false.i8.i.i.i.i493
  br label %while.body.i.i21.i.i.i.i494, !amdgpu.uniform !42

Flow3771:                                         ; preds = %Flow3770, %cond.false.i8.i.i.i.i493
  br label %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i497, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i494:                      ; preds = %while.body.i.i21.i.i.i.i494.preheader, %while.body.i.i21.i.i.i.i494
  tail call void @llvm.amdgcn.s.sleep(i32 2)
  %790 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i22.i.i.i.i495 = getelementptr inbounds nuw i32, ptr %790, i64 %idxprom.i23.i.i.i.i.i483, !amdgpu.uniform !42
  %791 = load atomic i32, ptr %arrayidx.i.i.i22.i.i.i.i495 monotonic, align 4
  %792 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %791)
  %cmp.i.not.i.i23.i.i.i.i496 = icmp eq i32 %792, %port.sroa.37.0.i.i487
  br i1 %cmp.i.not.i.i23.i.i.i.i496, label %Flow3770, label %while.body.i.i21.i.i.i.i494, !amdgpu.uniform !42

Flow3770:                                         ; preds = %while.body.i.i21.i.i.i.i494
  br label %Flow3771, !amdgpu.uniform !42

Flow3772:                                         ; preds = %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i498, %while.cond.i.i.i.i486
  %793 = phi i64 [ %add.i.i.i.i501, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i498 ], [ poison, %while.cond.i.i.i.i486 ]
  %794 = phi i32 [ %conv.i.i26.i.i.i.i499, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i498 ], [ poison, %while.cond.i.i.i.i486 ]
  %795 = phi ptr addrspace(4) [ %scevgep3122, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i498 ], [ poison, %while.cond.i.i.i.i486 ]
  %796 = phi i64 [ %lsr.iv.next3127, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i498 ], [ poison, %while.cond.i.i.i.i486 ]
  %797 = phi i1 [ false, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i498 ], [ true, %while.cond.i.i.i.i486 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %787)
  %798 = call i64 @llvm.amdgcn.if.break.i64(i1 %797, i64 %phi.broken3927)
  %799 = call i1 @llvm.amdgcn.loop.i64(i64 %798)
  br i1 %799, label %cond.false.i.i.i508, label %while.cond.i.i.i.i486

_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i497: ; preds = %Flow3771
  fence acquire
  br i1 %cmp.i.i.i.i489, label %if.then.i.i.i.i.i.i.i502, label %Flow3769, !amdgpu.uniform !42

loop-memcpy-expansion2363.preheader:              ; preds = %if.then.i.i.i.i.i.i.i502
  %arrayidx.i20.i35.i.i.i.i506 = getelementptr %"struct.rpc::Buffer", ptr %800, i64 %idxprom.i19.i.i.i.i.i476
  %arrayidx.i22.i38.i.i.i.i507 = getelementptr %"struct.rpc::Buffer", ptr %arrayidx.i20.i35.i.i.i.i506, i64 %sh_prom.i.i.i.i
  br label %loop-memcpy-expansion2363, !amdgpu.uniform !42

Flow3768:                                         ; preds = %Flow3767, %if.then.i.i.i.i.i.i.i502
  br label %loop-memcpy-residual-header2366, !amdgpu.uniform !42

loop-memcpy-expansion2363:                        ; preds = %loop-memcpy-expansion2363.preheader, %loop-memcpy-expansion2363
  %loop-index2364 = phi i64 [ %806, %loop-memcpy-expansion2363 ], [ 0, %loop-memcpy-expansion2363.preheader ]
  %scevgep3123 = getelementptr i8, ptr addrspace(4) %lsr.iv3121, i64 %loop-index2364, !amdgpu.uniform !42
  %804 = load <4 x i32>, ptr addrspace(4) %scevgep3123, align 1
  %805 = getelementptr i8, ptr %arrayidx.i22.i38.i.i.i.i507, i64 %loop-index2364
  store <4 x i32> %804, ptr %805, align 1
  %806 = add i64 %loop-index2364, 16
  %807 = icmp uge i64 %806, %802
  br i1 %807, label %Flow3767, label %loop-memcpy-expansion2363, !amdgpu.uniform !42

Flow3766:                                         ; preds = %Flow3765, %loop-memcpy-residual-header2366
  br label %Flow3769, !amdgpu.uniform !42

loop-memcpy-residual2365:                         ; preds = %loop-memcpy-residual2365.preheader, %loop-memcpy-residual2365
  %residual-loop-index2367 = phi i64 [ %809, %loop-memcpy-residual2365 ], [ 0, %loop-memcpy-residual2365.preheader ]
  %scevgep3132 = getelementptr i8, ptr addrspace(4) %scevgep3131, i64 %residual-loop-index2367, !amdgpu.uniform !42
  %808 = load i8, ptr addrspace(4) %scevgep3132, align 1
  %scevgep3130 = getelementptr i8, ptr %scevgep3129, i64 %residual-loop-index2367
  store i8 %808, ptr %scevgep3130, align 1
  %809 = add i64 %residual-loop-index2367, 1
  %810 = icmp uge i64 %809, %801
  br i1 %810, label %Flow3765, label %loop-memcpy-residual2365, !amdgpu.uniform !42

Flow3765:                                         ; preds = %loop-memcpy-residual2365
  br label %Flow3766, !amdgpu.uniform !42

_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i498: ; preds = %Flow3769
  %conv.i.i26.i.i.i.i499 = xor i32 %port.sroa.37.0.i.i487, 1
  fence release
  %811 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %arrayidx.i24.i29.i.i.i.i500 = getelementptr inbounds nuw i32, ptr %811, i64 %idxprom.i23.i.i.i.i.i483
  store atomic i32 %conv.i.i26.i.i.i.i499, ptr %arrayidx.i24.i29.i.i.i.i500 monotonic, align 4
  %add.i.i.i.i501 = add i64 %idx.0.i.i.i.i488, 64
  %scevgep3122 = getelementptr i8, ptr addrspace(4) %lsr.iv3121, i64 64
  %lsr.iv.next3127 = add i64 %lsr.iv3126, -64
  br label %Flow3772, !amdgpu.uniform !42

cond.false.i.i.i508:                              ; preds = %Flow3772
  %.lcssa4163 = phi i64 [ %798, %Flow3772 ]
  %port.sroa.37.0.i.i487.lcssa = phi i32 [ %port.sroa.37.0.i.i487, %Flow3772 ]
  %.lcssa4162 = phi i32 [ %784, %Flow3772 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa4163)
  %812 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %.lcssa4162)
  %813 = icmp ne i32 %812, %port.sroa.37.0.i.i487.lcssa
  %814 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %813)
  %815 = extractvalue { i1, i64 } %814, 0
  %816 = extractvalue { i1, i64 } %814, 1
  br i1 %815, label %while.body.i.i.i.i509.preheader, label %Flow3764

while.body.i.i.i.i509.preheader:                  ; preds = %cond.false.i.i.i508
  br label %while.body.i.i.i.i509, !amdgpu.uniform !42

Flow3764:                                         ; preds = %Flow3763, %cond.false.i.i.i508
  call void @llvm.amdgcn.end.cf.i64(i64 %816)
  br label %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit530, !amdgpu.uniform !42

while.body.i.i.i.i509:                            ; preds = %while.body.i.i.i.i509.preheader, %while.body.i.i.i.i509
  %phi.broken3928 = phi i64 [ 0, %while.body.i.i.i.i509.preheader ], [ %820, %while.body.i.i.i.i509 ]
  tail call void @llvm.amdgcn.s.sleep(i32 2)
  %817 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i8.i.i510 = getelementptr inbounds nuw i32, ptr %817, i64 %idxprom.i23.i.i.i.i.i483, !amdgpu.uniform !42
  %818 = load atomic i32, ptr %arrayidx.i.i.i8.i.i510 monotonic, align 4
  %819 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %818)
  %cmp.i.not.i.i.i.i511 = icmp eq i32 %819, %port.sroa.37.0.i.i487.lcssa
  %820 = call i64 @llvm.amdgcn.if.break.i64(i1 %cmp.i.not.i.i.i.i511, i64 %phi.broken3928)
  %821 = call i1 @llvm.amdgcn.loop.i64(i64 %820)
  br i1 %821, label %Flow3763, label %while.body.i.i.i.i509

Flow3763:                                         ; preds = %while.body.i.i.i.i509
  %.lcssa4161 = phi i64 [ %820, %while.body.i.i.i.i509 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa4161)
  br label %Flow3764, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit530: ; preds = %Flow3764
  fence acquire
  tail call void @llvm.amdgcn.wave.barrier()
  fence syncscope("agent") release
  %822 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %.lcssa4168, i1 true)
  %iszero.i.i.i.i.i13.i.i513 = icmp ne i64 %.lcssa4168, 0
  %cmp2.i.i.i.i14.i.i514 = icmp eq i64 %822, %sh_prom.i.i.i.i
  %cmp.i.i.i.i15.i.i515 = select i1 %iszero.i.i.i.i.i13.i.i513, i1 %cmp2.i.i.i.i14.i.i514, i1 false
  %823 = and i32 %.lcssa4169, 31
  %conv4.i.i.i19.i.i519 = zext i1 %cmp.i.i.i.i15.i.i515 to i32
  %shl.i.i.i20.i.i520 = shl nuw i32 %conv4.i.i.i19.i.i519, %823
  %xor.i.i.i21.i.i521 = xor i32 %shl.i.i.i20.i.i520, -1
  %824 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2497, !amdgpu.uniform !42

825:                                              ; preds = %ComputeEnd2498
  %div8.i.i.i16.i.i516 = lshr i32 %.lcssa4169, 5
  %idxprom.i.i.i17.i.i517 = zext nneg i32 %div8.i.i.i16.i.i516 to i64
  %826 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i17.i.i517
  %sunkaddr3302 = getelementptr inbounds i8, ptr addrspace(1) %826, i64 40
  %827 = atomicrmw and ptr addrspace(1) %sunkaddr3302, i32 %.lcssa4160 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %828, !amdgpu.uniform !42

828:                                              ; preds = %825, %ComputeEnd2498
  call void @llvm.amdgcn.end.cf.i64(i64 %3291)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %for.cond.i.i.i531, !amdgpu.uniform !42

for.cond.i.i.i531:                                ; preds = %828, %for.cond.i.i.i531
  %phi.broken3929 = phi i64 [ 0, %828 ], [ %831, %for.cond.i.i.i531 ]
  %lsr.iv3107 = phi i64 [ %lsr.iv.next3108, %for.cond.i.i.i531 ], [ -1, %828 ]
  %scevgep3119 = getelementptr i8, ptr %LHSStr, i64 %lsr.iv3107
  %scevgep3120 = getelementptr i8, ptr %scevgep3119, i64 1
  %829 = addrspacecast ptr %scevgep3120 to ptr addrspace(4)
  %830 = load i8, ptr addrspace(4) %829, align 1, !tbaa !57, !noalias.addrspace !101
  %cmp.i.i.i533 = icmp eq i8 %830, 0
  %lsr.iv.next3108 = add i64 %lsr.iv3107, 1
  %831 = call i64 @llvm.amdgcn.if.break.i64(i1 %cmp.i.i.i533, i64 %phi.broken3929)
  %832 = call i1 @llvm.amdgcn.loop.i64(i64 %831)
  br i1 %832, label %for.cond.i.i1.i539.preheader, label %for.cond.i.i.i531

for.cond.i.i1.i539.preheader:                     ; preds = %for.cond.i.i.i531
  %lsr.iv.next3108.lcssa = phi i64 [ %lsr.iv.next3108, %for.cond.i.i.i531 ]
  %.lcssa4159 = phi i64 [ %831, %for.cond.i.i.i531 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa4159)
  br label %for.cond.i.i1.i539, !amdgpu.uniform !42

for.cond.i.i1.i539:                               ; preds = %for.cond.i.i1.i539.preheader, %Flow3762
  %833 = phi i32 [ %871, %Flow3762 ], [ poison, %for.cond.i.i1.i539.preheader ]
  %834 = phi i64 [ %872, %Flow3762 ], [ poison, %for.cond.i.i1.i539.preheader ]
  %835 = phi i32 [ %873, %Flow3762 ], [ poison, %for.cond.i.i1.i539.preheader ]
  %retval.sroa.6.0.i.i.i540 = phi i32 [ %876, %Flow3762 ], [ undef, %for.cond.i.i1.i539.preheader ]
  %retval.sroa.2.0.i.i.i542 = phi i64 [ %875, %Flow3762 ], [ undef, %for.cond.i.i1.i539.preheader ]
  %retval.sroa.8.0.i.i.i543 = phi i32 [ %874, %Flow3762 ], [ undef, %for.cond.i.i1.i539.preheader ]
  %index.0.i.i.i544 = phi i32 [ %878, %Flow3762 ], [ 0, %for.cond.i.i1.i539.preheader ]
  %836 = load i32, ptr addrspace(1) @__llvm_rpc_client, align 8, !tbaa !68
  %cmp.not.i.i.i545 = icmp ult i32 %index.0.i.i.i544, %836
  %spec.store.select.i.i.i546 = select i1 %cmp.not.i.i.i545, i32 %index.0.i.i.i544, i32 0
  %837 = tail call noundef i64 @llvm.amdgcn.ballot.i64(i1 true)
  %838 = and i32 %spec.store.select.i.i.i546, 31
  %839 = lshr i64 %837, %sh_prom.i.i.i.i
  %840 = trunc i64 %839 to i32
  %conv4.i.i.i.i.i550 = and i32 %840, 1
  %shl.i.i.i.i.i551 = shl nuw i32 %conv4.i.i.i.i.i550, %838
  %841 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2501, !amdgpu.uniform !42

842:                                              ; preds = %ComputeEnd2502
  %sunkaddr3303 = getelementptr inbounds i8, ptr addrspace(1) %3301, i64 40
  %843 = atomicrmw or ptr addrspace(1) %sunkaddr3303, i32 %.lcssa4153 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %844, !amdgpu.uniform !42

844:                                              ; preds = %842, %ComputeEnd2502
  %845 = phi i32 [ %843, %842 ], [ poison, %ComputeEnd2502 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %3310)
  %846 = call i32 @llvm.amdgcn.readfirstlane.i32(i32 %845)
  %847 = or i32 %846, %.lcssa4154
  %shl5.i.i.i.i.i552 = shl nuw i32 1, %838
  %and.i.i.i.i.i553 = and i32 %shl5.i.i.i.i.i552, %847
  %tobool3.i.i.i.i554 = icmp ne i32 %and.i.i.i.i.i553, 0
  %848 = tail call i64 @llvm.amdgcn.ballot.i64(i1 %tobool3.i.i.i.i554)
  %and.i.i.i.i.i.i555 = and i64 %848, %837
  %cmp.i.not.i.i.i556 = icmp ne i64 %837, %and.i.i.i.i.i.i555
  br i1 %cmp.i.not.i.i.i556, label %if.then.i.i.i.i557, label %Flow3761, !amdgpu.uniform !42

if.then.i.i.i.i557:                               ; preds = %844
  fence syncscope("agent") acquire
  %849 = load <2 x i64>, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !75
  %850 = extractelement <2 x i64> %849, i32 0
  %851 = inttoptr i64 %850 to ptr
  %852 = extractelement <2 x i64> %849, i32 1
  %853 = inttoptr i64 %852 to ptr
  %idxprom.i.i.i.i558 = zext i32 %spec.store.select.i.i.i546 to i64
  %arrayidx.i.i.i.i559 = getelementptr inbounds nuw i32, ptr %851, i64 %idxprom.i.i.i.i558, !amdgpu.uniform !42
  %854 = load atomic i32, ptr %arrayidx.i.i.i.i559 monotonic, align 4
  %855 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %854)
  %arrayidx.i45.i.i.i560 = getelementptr inbounds nuw i32, ptr %853, i64 %idxprom.i.i.i.i558, !amdgpu.uniform !42
  %856 = load atomic i32, ptr %arrayidx.i45.i.i.i560 monotonic, align 4
  %857 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %856)
  %cmp.i46.not.i.i.i561 = icmp ne i32 %855, %857
  br i1 %cmp.i46.not.i.i.i561, label %if.then12.i.i.i562, label %Flow3759, !amdgpu.uniform !42

if.then12.i.i.i562:                               ; preds = %if.then.i.i.i.i557
  fence syncscope("agent") release
  %858 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %837, i1 true)
  %iszero.i.i.i.i.i.i.i563 = icmp ne i64 %837, 0
  %cmp2.i.i.i.i.i.i564 = icmp eq i64 %858, %sh_prom.i.i.i.i
  %cmp.i.i.i.i.i.i565 = select i1 %iszero.i.i.i.i.i.i.i563, i1 %cmp2.i.i.i.i.i.i564, i1 false
  %conv4.i.i51.i.i.i566 = zext i1 %cmp.i.i.i.i.i.i565 to i32
  %shl.i.i52.i.i.i567 = shl nuw i32 %conv4.i.i51.i.i.i566, %838
  %xor.i.i.i.i.i568 = xor i32 %shl.i.i52.i.i.i567, -1
  %859 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2506, !amdgpu.uniform !42

860:                                              ; preds = %ComputeEnd2507
  %sunkaddr3304 = getelementptr inbounds i8, ptr addrspace(1) %3301, i64 40
  %861 = atomicrmw and ptr addrspace(1) %sunkaddr3304, i32 %.lcssa4155 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %862, !amdgpu.uniform !42

862:                                              ; preds = %860, %ComputeEnd2507
  call void @llvm.amdgcn.end.cf.i64(i64 %3328)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %Flow3759, !amdgpu.uniform !42

if.end14.i.i.i628:                                ; preds = %Flow3759
  %863 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %837, i1 true)
  %iszero.i.i.i.i.i.i629 = icmp ne i64 %837, 0
  %cmp2.i.i.i.i.i630 = icmp eq i64 %863, %sh_prom.i.i.i.i
  %cmp.i.i.i.i.i631 = select i1 %iszero.i.i.i.i.i.i629, i1 %cmp2.i.i.i.i.i630, i1 false
  %864 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.i.i.i.i.i631)
  %865 = extractvalue { i1, i64 } %864, 0
  %866 = extractvalue { i1, i64 } %864, 1
  br i1 %865, label %if.then16.i.i.i633, label %if.end22.i.i.i632

if.then16.i.i.i633:                               ; preds = %if.end14.i.i.i628
  %867 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !76
  %opcode.i.i.i634 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %867, i64 %idxprom.i.i.i.i558, i32 1
  store i32 1660944387, ptr %opcode.i.i.i634, align 8, !tbaa !77
  %arrayidx21.i.i.i635 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %867, i64 %idxprom.i.i.i.i558
  store i64 %837, ptr %arrayidx21.i.i.i635, align 8, !tbaa !80
  br label %if.end22.i.i.i632, !amdgpu.uniform !42

Flow3760:                                         ; preds = %if.end22.i.i.i632, %Flow3759
  %868 = phi i32 [ %855, %if.end22.i.i.i632 ], [ %retval.sroa.8.0.i.i.i543, %Flow3759 ]
  %869 = phi i64 [ %837, %if.end22.i.i.i632 ], [ %retval.sroa.2.0.i.i.i542, %Flow3759 ]
  %870 = phi i32 [ %spec.store.select.i.i.i546, %if.end22.i.i.i632 ], [ %retval.sroa.6.0.i.i.i540, %Flow3759 ]
  br label %cleanup26.i.i.i569, !amdgpu.uniform !42

if.end22.i.i.i632:                                ; preds = %if.then16.i.i.i633, %if.end14.i.i.i628
  call void @llvm.amdgcn.end.cf.i64(i64 %866)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %Flow3760, !amdgpu.uniform !42

Flow3761:                                         ; preds = %cleanup26.i.i.i569, %844
  %871 = phi i32 [ %870, %cleanup26.i.i.i569 ], [ %833, %844 ]
  %872 = phi i64 [ %869, %cleanup26.i.i.i569 ], [ %834, %844 ]
  %873 = phi i32 [ %868, %cleanup26.i.i.i569 ], [ %835, %844 ]
  %874 = phi i32 [ %868, %cleanup26.i.i.i569 ], [ %retval.sroa.8.0.i.i.i543, %844 ]
  %875 = phi i64 [ %869, %cleanup26.i.i.i569 ], [ %retval.sroa.2.0.i.i.i542, %844 ]
  %876 = phi i32 [ %870, %cleanup26.i.i.i569 ], [ %retval.sroa.6.0.i.i.i540, %844 ]
  %877 = phi i1 [ %cmp.i46.not.i.i.i561, %cleanup26.i.i.i569 ], [ true, %844 ]
  br i1 %877, label %for.inc.i.i.i574, label %Flow3762, !amdgpu.uniform !42

cleanup26.i.i.i569:                               ; preds = %Flow3760
  br label %Flow3761, !amdgpu.uniform !42

for.inc.i.i.i574:                                 ; preds = %Flow3761
  %inc.i.i.i579 = add i32 %spec.store.select.i.i.i546, 1
  br label %Flow3762, !amdgpu.uniform !42

Flow3762:                                         ; preds = %for.inc.i.i.i574, %Flow3761
  %878 = phi i32 [ %inc.i.i.i579, %for.inc.i.i.i574 ], [ poison, %Flow3761 ]
  %879 = phi i1 [ false, %for.inc.i.i.i574 ], [ true, %Flow3761 ]
  br i1 %879, label %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i580, label %for.cond.i.i1.i539, !amdgpu.uniform !42

_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i580: ; preds = %Flow3762
  %.lcssa4158 = phi i32 [ %871, %Flow3762 ]
  %.lcssa4157 = phi i64 [ %872, %Flow3762 ]
  %.lcssa4156 = phi i32 [ %873, %Flow3762 ]
  %scevgep3118 = getelementptr i8, ptr %LHSStr, i64 %lsr.iv.next3108.lcssa
  %sub.ptr.lhs.cast.i.i.i536 = ptrtoint ptr %scevgep3118 to i64
  %sub.ptr.rhs.cast.i.i.i = ptrtoint ptr %LHSStr to i64
  %sub.ptr.sub.i.i.i537 = sub i64 %sub.ptr.lhs.cast.i.i.i536, %sub.ptr.rhs.cast.i.i.i
  fence acquire
  %880 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %mul.i.i.i.i.i.i581 = shl i32 %.lcssa4158, 6
  %idxprom.i19.i.i.i.i.i582 = zext i32 %mul.i.i.i.i.i.i581 to i64
  %arrayidx.i20.i.i.i.i.i583 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %880, i64 %idxprom.i19.i.i.i.i.i582
  %arrayidx.i22.i.i.i.i.i584 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %arrayidx.i20.i.i.i.i.i583, i64 %sh_prom.i.i.i.i
  store i64 %sub.ptr.sub.i.i.i537, ptr %arrayidx.i22.i.i.i.i.i584, align 8, !tbaa !82
  %spec.select.i.i.i.i.i.i.i585 = tail call i64 @llvm.umin.i64(i64 %sub.ptr.sub.i.i.i537, i64 56)
  %881 = and i64 %spec.select.i.i.i.i.i.i.i585, 15
  %882 = sub i64 %spec.select.i.i.i.i.i.i.i585, %881
  %883 = icmp ne i64 %882, 0
  %884 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %883)
  %885 = extractvalue { i1, i64 } %884, 0
  %886 = extractvalue { i1, i64 } %884, 1
  br i1 %885, label %loop-memcpy-expansion2255.preheader, label %Flow3758

loop-memcpy-expansion2255.preheader:              ; preds = %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i580
  %887 = add nuw nsw i64 %75, 8
  %scevgep3115 = getelementptr i8, ptr %880, i64 %887
  %888 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i582, 6
  %scevgep3116 = getelementptr i8, ptr %scevgep3115, i64 %888
  br label %loop-memcpy-expansion2255, !amdgpu.uniform !42

Flow3758:                                         ; preds = %Flow3757, %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i580
  call void @llvm.amdgcn.end.cf.i64(i64 %886)
  br label %loop-memcpy-residual-header2258, !amdgpu.uniform !42

loop-memcpy-expansion2255:                        ; preds = %loop-memcpy-expansion2255.preheader, %loop-memcpy-expansion2255
  %phi.broken3930 = phi i64 [ 0, %loop-memcpy-expansion2255.preheader ], [ %893, %loop-memcpy-expansion2255 ]
  %loop-index2256 = phi i64 [ %891, %loop-memcpy-expansion2255 ], [ 0, %loop-memcpy-expansion2255.preheader ]
  %889 = getelementptr i8, ptr %LHSStr, i64 %loop-index2256
  %890 = load <4 x i32>, ptr %889, align 1
  %scevgep3117 = getelementptr i8, ptr %scevgep3116, i64 %loop-index2256
  store <4 x i32> %890, ptr %scevgep3117, align 1
  %891 = add i64 %loop-index2256, 16
  %892 = icmp uge i64 %891, %882
  %893 = call i64 @llvm.amdgcn.if.break.i64(i1 %892, i64 %phi.broken3930)
  %894 = call i1 @llvm.amdgcn.loop.i64(i64 %893)
  br i1 %894, label %Flow3757, label %loop-memcpy-expansion2255

Flow3756:                                         ; preds = %Flow3754, %Flow3755
  %895 = phi i64 [ %2830, %Flow3754 ], [ %2823, %Flow3755 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %2826)
  br label %post-loop-memcpy-expansion2254, !amdgpu.uniform !42

loop-memcpy-residual2257:                         ; preds = %loop-memcpy-residual2257.preheader, %loop-memcpy-residual2257
  %phi.broken3931 = phi i64 [ 0, %loop-memcpy-residual2257.preheader ], [ %899, %loop-memcpy-residual2257 ]
  %residual-loop-index2259 = phi i64 [ %897, %loop-memcpy-residual2257 ], [ 0, %loop-memcpy-residual2257.preheader ]
  %scevgep3114 = getelementptr i8, ptr %scevgep3113, i64 %residual-loop-index2259
  %896 = load i8, ptr %scevgep3114, align 1
  %scevgep3112 = getelementptr i8, ptr %scevgep3111, i64 %residual-loop-index2259
  store i8 %896, ptr %scevgep3112, align 1
  %897 = add i64 %residual-loop-index2259, 1
  %898 = icmp uge i64 %897, %881
  %899 = call i64 @llvm.amdgcn.if.break.i64(i1 %898, i64 %phi.broken3931)
  %900 = call i1 @llvm.amdgcn.loop.i64(i64 %899)
  br i1 %900, label %Flow3754, label %loop-memcpy-residual2257

Flow3754:                                         ; preds = %loop-memcpy-residual2257
  %.lcssa4151 = phi i64 [ %899, %loop-memcpy-residual2257 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa4151)
  br label %Flow3756, !amdgpu.uniform !42

post-loop-memcpy-expansion2254:                   ; preds = %Flow3756
  %901 = add i64 %lsr.iv.next3108.lcssa, -56
  %tobool.not.i.i.i.i.i.i587 = icmp eq i32 %.lcssa4156, 0
  %conv.i.i.i.i.i.i588 = zext i1 %tobool.not.i.i.i.i.i.i587 to i32
  fence release
  %902 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %idxprom.i23.i.i.i.i.i589 = zext i32 %.lcssa4158 to i64
  %arrayidx.i24.i.i.i.i.i590 = getelementptr inbounds nuw i32, ptr %902, i64 %idxprom.i23.i.i.i.i.i589
  store atomic i32 %conv.i.i.i.i.i.i588, ptr %arrayidx.i24.i.i.i.i.i590 monotonic, align 4
  %903 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !84
  %arrayidx.i.i5.i.i591 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %903, i64 %idxprom.i23.i.i.i.i.i589, !amdgpu.uniform !42
  %904 = load i64, ptr %arrayidx.i.i5.i.i591, align 8, !tbaa !80
  %scevgep3090 = getelementptr i8, ptr %LHSStr, i64 56
  br label %while.cond.i.i.i.i592, !amdgpu.uniform !42

while.cond.i.i.i.i592:                            ; preds = %post-loop-memcpy-expansion2254, %Flow3753
  %phi.broken3934 = phi i64 [ 0, %post-loop-memcpy-expansion2254 ], [ %921, %Flow3753 ]
  %lsr.iv3098 = phi i64 [ %919, %Flow3753 ], [ %901, %post-loop-memcpy-expansion2254 ]
  %lsr.iv3091 = phi ptr [ %918, %Flow3753 ], [ %scevgep3090, %post-loop-memcpy-expansion2254 ]
  %port.sroa.37.0.i.i593 = phi i32 [ %917, %Flow3753 ], [ %conv.i.i.i.i.i.i588, %post-loop-memcpy-expansion2254 ]
  %idx.0.i.i.i.i594 = phi i64 [ %916, %Flow3753 ], [ 56, %post-loop-memcpy-expansion2254 ]
  %cmp.i.i.i.i595 = icmp ugt i64 %sub.ptr.sub.i.i.i537, %idx.0.i.i.i.i594
  %905 = tail call i64 @llvm.amdgcn.ballot.i64(i1 %cmp.i.i.i.i595)
  %and.i.i.i.i6.i.i596 = and i64 %905, %904
  %tobool.not.i.i.i.i597 = icmp ne i64 %and.i.i.i.i6.i.i596, 0
  %906 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i22.i.i.i598 = getelementptr inbounds nuw i32, ptr %906, i64 %idxprom.i23.i.i.i.i.i589, !amdgpu.uniform !42
  %907 = load atomic i32, ptr %arrayidx.i22.i.i.i598 monotonic, align 4
  %908 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %tobool.not.i.i.i.i597)
  %909 = extractvalue { i1, i64 } %908, 0
  %910 = extractvalue { i1, i64 } %908, 1
  br i1 %909, label %cond.false.i8.i.i.i.i599, label %Flow3753

cond.false.i8.i.i.i.i599:                         ; preds = %while.cond.i.i.i.i592
  %911 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %907)
  %912 = icmp ne i32 %911, %port.sroa.37.0.i.i593
  br i1 %912, label %while.body.i.i21.i.i.i.i600.preheader, label %Flow3752, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i600.preheader:            ; preds = %cond.false.i8.i.i.i.i599
  br label %while.body.i.i21.i.i.i.i600, !amdgpu.uniform !42

Flow3752:                                         ; preds = %Flow3751, %cond.false.i8.i.i.i.i599
  br label %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i603, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i600:                      ; preds = %while.body.i.i21.i.i.i.i600.preheader, %while.body.i.i21.i.i.i.i600
  tail call void @llvm.amdgcn.s.sleep(i32 2)
  %913 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i22.i.i.i.i601 = getelementptr inbounds nuw i32, ptr %913, i64 %idxprom.i23.i.i.i.i.i589, !amdgpu.uniform !42
  %914 = load atomic i32, ptr %arrayidx.i.i.i22.i.i.i.i601 monotonic, align 4
  %915 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %914)
  %cmp.i.not.i.i23.i.i.i.i602 = icmp eq i32 %915, %port.sroa.37.0.i.i593
  br i1 %cmp.i.not.i.i23.i.i.i.i602, label %Flow3751, label %while.body.i.i21.i.i.i.i600, !amdgpu.uniform !42

Flow3751:                                         ; preds = %while.body.i.i21.i.i.i.i600
  br label %Flow3752, !amdgpu.uniform !42

Flow3753:                                         ; preds = %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i604, %while.cond.i.i.i.i592
  %916 = phi i64 [ %add.i.i.i.i607, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i604 ], [ poison, %while.cond.i.i.i.i592 ]
  %917 = phi i32 [ %conv.i.i26.i.i.i.i605, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i604 ], [ poison, %while.cond.i.i.i.i592 ]
  %918 = phi ptr [ %scevgep3092, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i604 ], [ poison, %while.cond.i.i.i.i592 ]
  %919 = phi i64 [ %lsr.iv.next3099, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i604 ], [ poison, %while.cond.i.i.i.i592 ]
  %920 = phi i1 [ false, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i604 ], [ true, %while.cond.i.i.i.i592 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %910)
  %921 = call i64 @llvm.amdgcn.if.break.i64(i1 %920, i64 %phi.broken3934)
  %922 = call i1 @llvm.amdgcn.loop.i64(i64 %921)
  br i1 %922, label %cond.false.i.i.i614, label %while.cond.i.i.i.i592

_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i603: ; preds = %Flow3752
  fence acquire
  %923 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.i.i.i.i595)
  %924 = extractvalue { i1, i64 } %923, 0
  %925 = extractvalue { i1, i64 } %923, 1
  br i1 %924, label %if.then.i.i.i.i.i.i.i608, label %Flow3750

if.then.i.i.i.i.i.i.i608:                         ; preds = %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i603
  %sub.i.i.i.i.i.i.i610 = sub nuw i64 %sub.ptr.sub.i.i.i537, %idx.0.i.i.i.i594
  %spec.select.i.i.i36.i.i.i.i611 = tail call i64 @llvm.umin.i64(i64 %sub.i.i.i.i.i.i.i610, i64 64)
  %926 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %927 = and i64 %spec.select.i.i.i36.i.i.i.i611, 15
  %928 = sub i64 %spec.select.i.i.i36.i.i.i.i611, %927
  %929 = icmp ne i64 %928, 0
  %930 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %929)
  %931 = extractvalue { i1, i64 } %930, 0
  %932 = extractvalue { i1, i64 } %930, 1
  br i1 %931, label %loop-memcpy-expansion2249.preheader, label %Flow3749

loop-memcpy-expansion2249.preheader:              ; preds = %if.then.i.i.i.i.i.i.i608
  %arrayidx.i20.i35.i.i.i.i612 = getelementptr %"struct.rpc::Buffer", ptr %926, i64 %idxprom.i19.i.i.i.i.i582
  %arrayidx.i22.i38.i.i.i.i613 = getelementptr %"struct.rpc::Buffer", ptr %arrayidx.i20.i35.i.i.i.i612, i64 %sh_prom.i.i.i.i
  br label %loop-memcpy-expansion2249, !amdgpu.uniform !42

Flow3749:                                         ; preds = %Flow3748, %if.then.i.i.i.i.i.i.i608
  call void @llvm.amdgcn.end.cf.i64(i64 %932)
  br label %loop-memcpy-residual-header2252, !amdgpu.uniform !42

loop-memcpy-expansion2249:                        ; preds = %loop-memcpy-expansion2249.preheader, %loop-memcpy-expansion2249
  %phi.broken3932 = phi i64 [ 0, %loop-memcpy-expansion2249.preheader ], [ %937, %loop-memcpy-expansion2249 ]
  %loop-index2250 = phi i64 [ %935, %loop-memcpy-expansion2249 ], [ 0, %loop-memcpy-expansion2249.preheader ]
  %scevgep3093 = getelementptr i8, ptr %lsr.iv3091, i64 %loop-index2250
  %933 = load <4 x i32>, ptr %scevgep3093, align 1
  %934 = getelementptr i8, ptr %arrayidx.i22.i38.i.i.i.i613, i64 %loop-index2250
  store <4 x i32> %933, ptr %934, align 1
  %935 = add i64 %loop-index2250, 16
  %936 = icmp uge i64 %935, %928
  %937 = call i64 @llvm.amdgcn.if.break.i64(i1 %936, i64 %phi.broken3932)
  %938 = call i1 @llvm.amdgcn.loop.i64(i64 %937)
  br i1 %938, label %Flow3748, label %loop-memcpy-expansion2249

Flow3747:                                         ; preds = %Flow3746, %loop-memcpy-residual-header2252
  call void @llvm.amdgcn.end.cf.i64(i64 %2816)
  br label %Flow3750, !amdgpu.uniform !42

loop-memcpy-residual2251:                         ; preds = %loop-memcpy-residual2251.preheader, %loop-memcpy-residual2251
  %phi.broken3933 = phi i64 [ 0, %loop-memcpy-residual2251.preheader ], [ %942, %loop-memcpy-residual2251 ]
  %residual-loop-index2253 = phi i64 [ %940, %loop-memcpy-residual2251 ], [ 0, %loop-memcpy-residual2251.preheader ]
  %scevgep3105 = getelementptr i8, ptr %scevgep3104, i64 %residual-loop-index2253
  %939 = load i8, ptr %scevgep3105, align 1
  %scevgep3102 = getelementptr i8, ptr %scevgep3101, i64 %residual-loop-index2253
  store i8 %939, ptr %scevgep3102, align 1
  %940 = add i64 %residual-loop-index2253, 1
  %941 = icmp uge i64 %940, %927
  %942 = call i64 @llvm.amdgcn.if.break.i64(i1 %941, i64 %phi.broken3933)
  %943 = call i1 @llvm.amdgcn.loop.i64(i64 %942)
  br i1 %943, label %Flow3746, label %loop-memcpy-residual2251

Flow3746:                                         ; preds = %loop-memcpy-residual2251
  %.lcssa4148 = phi i64 [ %942, %loop-memcpy-residual2251 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa4148)
  br label %Flow3747, !amdgpu.uniform !42

_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i604: ; preds = %Flow3750
  %conv.i.i26.i.i.i.i605 = xor i32 %port.sroa.37.0.i.i593, 1
  fence release
  %944 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %arrayidx.i24.i29.i.i.i.i606 = getelementptr inbounds nuw i32, ptr %944, i64 %idxprom.i23.i.i.i.i.i589
  store atomic i32 %conv.i.i26.i.i.i.i605, ptr %arrayidx.i24.i29.i.i.i.i606 monotonic, align 4
  %add.i.i.i.i607 = add i64 %idx.0.i.i.i.i594, 64
  %scevgep3092 = getelementptr i8, ptr %lsr.iv3091, i64 64
  %lsr.iv.next3099 = add i64 %lsr.iv3098, -64
  br label %Flow3753, !amdgpu.uniform !42

cond.false.i.i.i614:                              ; preds = %Flow3753
  %.lcssa4150 = phi i64 [ %921, %Flow3753 ]
  %port.sroa.37.0.i.i593.lcssa = phi i32 [ %port.sroa.37.0.i.i593, %Flow3753 ]
  %.lcssa4149 = phi i32 [ %907, %Flow3753 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa4150)
  %945 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %.lcssa4149)
  %946 = icmp ne i32 %945, %port.sroa.37.0.i.i593.lcssa
  %947 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %946)
  %948 = extractvalue { i1, i64 } %947, 0
  %949 = extractvalue { i1, i64 } %947, 1
  br i1 %948, label %while.body.i.i.i.i615.preheader, label %Flow3745

while.body.i.i.i.i615.preheader:                  ; preds = %cond.false.i.i.i614
  br label %while.body.i.i.i.i615, !amdgpu.uniform !42

Flow3745:                                         ; preds = %Flow3744, %cond.false.i.i.i614
  call void @llvm.amdgcn.end.cf.i64(i64 %949)
  br label %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit636, !amdgpu.uniform !42

while.body.i.i.i.i615:                            ; preds = %while.body.i.i.i.i615.preheader, %while.body.i.i.i.i615
  %phi.broken3935 = phi i64 [ 0, %while.body.i.i.i.i615.preheader ], [ %953, %while.body.i.i.i.i615 ]
  tail call void @llvm.amdgcn.s.sleep(i32 2)
  %950 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i8.i.i616 = getelementptr inbounds nuw i32, ptr %950, i64 %idxprom.i23.i.i.i.i.i589, !amdgpu.uniform !42
  %951 = load atomic i32, ptr %arrayidx.i.i.i8.i.i616 monotonic, align 4
  %952 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %951)
  %cmp.i.not.i.i.i.i617 = icmp eq i32 %952, %port.sroa.37.0.i.i593.lcssa
  %953 = call i64 @llvm.amdgcn.if.break.i64(i1 %cmp.i.not.i.i.i.i617, i64 %phi.broken3935)
  %954 = call i1 @llvm.amdgcn.loop.i64(i64 %953)
  br i1 %954, label %Flow3744, label %while.body.i.i.i.i615

Flow3744:                                         ; preds = %while.body.i.i.i.i615
  %.lcssa4146 = phi i64 [ %953, %while.body.i.i.i.i615 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa4146)
  br label %Flow3745, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit636: ; preds = %Flow3745
  fence acquire
  tail call void @llvm.amdgcn.wave.barrier()
  fence syncscope("agent") release
  %955 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %.lcssa4157, i1 true)
  %iszero.i.i.i.i.i13.i.i619 = icmp ne i64 %.lcssa4157, 0
  %cmp2.i.i.i.i14.i.i620 = icmp eq i64 %955, %sh_prom.i.i.i.i
  %cmp.i.i.i.i15.i.i621 = select i1 %iszero.i.i.i.i.i13.i.i619, i1 %cmp2.i.i.i.i14.i.i620, i1 false
  %956 = and i32 %.lcssa4158, 31
  %conv4.i.i.i19.i.i625 = zext i1 %cmp.i.i.i.i15.i.i621 to i32
  %shl.i.i.i20.i.i626 = shl nuw i32 %conv4.i.i.i19.i.i625, %956
  %xor.i.i.i21.i.i627 = xor i32 %shl.i.i.i20.i.i626, -1
  %957 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2510, !amdgpu.uniform !42

958:                                              ; preds = %ComputeEnd2511
  %div8.i.i.i16.i.i622 = lshr i32 %.lcssa4158, 5
  %idxprom.i.i.i17.i.i623 = zext nneg i32 %div8.i.i.i16.i.i622 to i64
  %959 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i17.i.i623
  %sunkaddr3305 = getelementptr inbounds i8, ptr addrspace(1) %959, i64 40
  %960 = atomicrmw and ptr addrspace(1) %sunkaddr3305, i32 %.lcssa4145 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %961, !amdgpu.uniform !42

961:                                              ; preds = %958, %ComputeEnd2511
  call void @llvm.amdgcn.end.cf.i64(i64 %3345)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %for.cond.i.i.i638, !amdgpu.uniform !42

for.cond.i.i.i638:                                ; preds = %961, %Flow3743
  %962 = phi i32 [ %1000, %Flow3743 ], [ poison, %961 ]
  %963 = phi i64 [ %1001, %Flow3743 ], [ poison, %961 ]
  %964 = phi i32 [ %1002, %Flow3743 ], [ poison, %961 ]
  %retval.sroa.6.0.i.i.i639 = phi i32 [ %1005, %Flow3743 ], [ undef, %961 ]
  %retval.sroa.2.0.i.i.i641 = phi i64 [ %1004, %Flow3743 ], [ undef, %961 ]
  %retval.sroa.8.0.i.i.i642 = phi i32 [ %1003, %Flow3743 ], [ undef, %961 ]
  %index.0.i.i.i643 = phi i32 [ %1007, %Flow3743 ], [ 0, %961 ]
  %965 = load i32, ptr addrspace(1) @__llvm_rpc_client, align 8, !tbaa !68
  %cmp.not.i.i.i644 = icmp ult i32 %index.0.i.i.i643, %965
  %spec.store.select.i.i.i645 = select i1 %cmp.not.i.i.i644, i32 %index.0.i.i.i643, i32 0
  %966 = tail call noundef i64 @llvm.amdgcn.ballot.i64(i1 true)
  %967 = and i32 %spec.store.select.i.i.i645, 31
  %968 = lshr i64 %966, %sh_prom.i.i.i.i
  %969 = trunc i64 %968 to i32
  %conv4.i.i.i.i.i649 = and i32 %969, 1
  %shl.i.i.i.i.i650 = shl nuw i32 %conv4.i.i.i.i.i649, %967
  %970 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2514, !amdgpu.uniform !42

971:                                              ; preds = %ComputeEnd2515
  %sunkaddr3306 = getelementptr inbounds i8, ptr addrspace(1) %3355, i64 40
  %972 = atomicrmw or ptr addrspace(1) %sunkaddr3306, i32 %.lcssa4139 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %973, !amdgpu.uniform !42

973:                                              ; preds = %971, %ComputeEnd2515
  %974 = phi i32 [ %972, %971 ], [ poison, %ComputeEnd2515 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %3364)
  %975 = call i32 @llvm.amdgcn.readfirstlane.i32(i32 %974)
  %976 = or i32 %975, %.lcssa4140
  %shl5.i.i.i.i.i651 = shl nuw i32 1, %967
  %and.i.i.i.i.i652 = and i32 %shl5.i.i.i.i.i651, %976
  %tobool3.i.i.i.i653 = icmp ne i32 %and.i.i.i.i.i652, 0
  %977 = tail call i64 @llvm.amdgcn.ballot.i64(i1 %tobool3.i.i.i.i653)
  %and.i.i.i.i.i.i654 = and i64 %977, %966
  %cmp.i.not.i.i.i655 = icmp ne i64 %966, %and.i.i.i.i.i.i654
  br i1 %cmp.i.not.i.i.i655, label %if.then.i.i.i.i656, label %Flow3742, !amdgpu.uniform !42

if.then.i.i.i.i656:                               ; preds = %973
  fence syncscope("agent") acquire
  %978 = load <2 x i64>, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !75
  %979 = extractelement <2 x i64> %978, i32 0
  %980 = inttoptr i64 %979 to ptr
  %981 = extractelement <2 x i64> %978, i32 1
  %982 = inttoptr i64 %981 to ptr
  %idxprom.i.i.i.i657 = zext i32 %spec.store.select.i.i.i645 to i64
  %arrayidx.i.i.i.i658 = getelementptr inbounds nuw i32, ptr %980, i64 %idxprom.i.i.i.i657, !amdgpu.uniform !42
  %983 = load atomic i32, ptr %arrayidx.i.i.i.i658 monotonic, align 4
  %984 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %983)
  %arrayidx.i45.i.i.i659 = getelementptr inbounds nuw i32, ptr %982, i64 %idxprom.i.i.i.i657, !amdgpu.uniform !42
  %985 = load atomic i32, ptr %arrayidx.i45.i.i.i659 monotonic, align 4
  %986 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %985)
  %cmp.i46.not.i.i.i660 = icmp ne i32 %984, %986
  br i1 %cmp.i46.not.i.i.i660, label %if.then12.i.i.i661, label %Flow3740, !amdgpu.uniform !42

if.then12.i.i.i661:                               ; preds = %if.then.i.i.i.i656
  fence syncscope("agent") release
  %987 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %966, i1 true)
  %iszero.i.i.i.i.i.i.i662 = icmp ne i64 %966, 0
  %cmp2.i.i.i.i.i.i663 = icmp eq i64 %987, %sh_prom.i.i.i.i
  %cmp.i.i.i.i.i.i664 = select i1 %iszero.i.i.i.i.i.i.i662, i1 %cmp2.i.i.i.i.i.i663, i1 false
  %conv4.i.i51.i.i.i665 = zext i1 %cmp.i.i.i.i.i.i664 to i32
  %shl.i.i52.i.i.i666 = shl nuw i32 %conv4.i.i51.i.i.i665, %967
  %xor.i.i.i.i.i667 = xor i32 %shl.i.i52.i.i.i666, -1
  %988 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2519, !amdgpu.uniform !42

989:                                              ; preds = %ComputeEnd2520
  %sunkaddr3307 = getelementptr inbounds i8, ptr addrspace(1) %3355, i64 40
  %990 = atomicrmw and ptr addrspace(1) %sunkaddr3307, i32 %.lcssa4141 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %991, !amdgpu.uniform !42

991:                                              ; preds = %989, %ComputeEnd2520
  call void @llvm.amdgcn.end.cf.i64(i64 %3382)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %Flow3740, !amdgpu.uniform !42

if.end14.i.i.i725:                                ; preds = %Flow3740
  %992 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %966, i1 true)
  %iszero.i.i.i.i.i.i726 = icmp ne i64 %966, 0
  %cmp2.i.i.i.i.i727 = icmp eq i64 %992, %sh_prom.i.i.i.i
  %cmp.i.i.i.i.i728 = select i1 %iszero.i.i.i.i.i.i726, i1 %cmp2.i.i.i.i.i727, i1 false
  %993 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.i.i.i.i.i728)
  %994 = extractvalue { i1, i64 } %993, 0
  %995 = extractvalue { i1, i64 } %993, 1
  br i1 %994, label %if.then16.i.i.i730, label %if.end22.i.i.i729

if.then16.i.i.i730:                               ; preds = %if.end14.i.i.i725
  %996 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !76
  %opcode.i.i.i731 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %996, i64 %idxprom.i.i.i.i657, i32 1
  store i32 1660944387, ptr %opcode.i.i.i731, align 8, !tbaa !77
  %arrayidx21.i.i.i732 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %996, i64 %idxprom.i.i.i.i657
  store i64 %966, ptr %arrayidx21.i.i.i732, align 8, !tbaa !80
  br label %if.end22.i.i.i729, !amdgpu.uniform !42

Flow3741:                                         ; preds = %if.end22.i.i.i729, %Flow3740
  %997 = phi i32 [ %984, %if.end22.i.i.i729 ], [ %retval.sroa.8.0.i.i.i642, %Flow3740 ]
  %998 = phi i64 [ %966, %if.end22.i.i.i729 ], [ %retval.sroa.2.0.i.i.i641, %Flow3740 ]
  %999 = phi i32 [ %spec.store.select.i.i.i645, %if.end22.i.i.i729 ], [ %retval.sroa.6.0.i.i.i639, %Flow3740 ]
  br label %cleanup26.i.i.i668, !amdgpu.uniform !42

if.end22.i.i.i729:                                ; preds = %if.then16.i.i.i730, %if.end14.i.i.i725
  call void @llvm.amdgcn.end.cf.i64(i64 %995)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %Flow3741, !amdgpu.uniform !42

Flow3742:                                         ; preds = %cleanup26.i.i.i668, %973
  %1000 = phi i32 [ %999, %cleanup26.i.i.i668 ], [ %962, %973 ]
  %1001 = phi i64 [ %998, %cleanup26.i.i.i668 ], [ %963, %973 ]
  %1002 = phi i32 [ %997, %cleanup26.i.i.i668 ], [ %964, %973 ]
  %1003 = phi i32 [ %997, %cleanup26.i.i.i668 ], [ %retval.sroa.8.0.i.i.i642, %973 ]
  %1004 = phi i64 [ %998, %cleanup26.i.i.i668 ], [ %retval.sroa.2.0.i.i.i641, %973 ]
  %1005 = phi i32 [ %999, %cleanup26.i.i.i668 ], [ %retval.sroa.6.0.i.i.i639, %973 ]
  %1006 = phi i1 [ %cmp.i46.not.i.i.i660, %cleanup26.i.i.i668 ], [ true, %973 ]
  br i1 %1006, label %for.inc.i.i.i673, label %Flow3743, !amdgpu.uniform !42

cleanup26.i.i.i668:                               ; preds = %Flow3741
  br label %Flow3742, !amdgpu.uniform !42

for.inc.i.i.i673:                                 ; preds = %Flow3742
  %inc.i.i.i678 = add i32 %spec.store.select.i.i.i645, 1
  br label %Flow3743, !amdgpu.uniform !42

Flow3743:                                         ; preds = %for.inc.i.i.i673, %Flow3742
  %1007 = phi i32 [ %inc.i.i.i678, %for.inc.i.i.i673 ], [ poison, %Flow3742 ]
  %1008 = phi i1 [ false, %for.inc.i.i.i673 ], [ true, %Flow3742 ]
  br i1 %1008, label %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i679, label %for.cond.i.i.i638, !amdgpu.uniform !42

_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i679: ; preds = %Flow3743
  %.lcssa4144 = phi i32 [ %1000, %Flow3743 ]
  %.lcssa4143 = phi i64 [ %1001, %Flow3743 ]
  %.lcssa4142 = phi i32 [ %1002, %Flow3743 ]
  fence acquire
  %1009 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %mul.i.i.i.i.i.i680 = shl i32 %.lcssa4144, 6
  %idxprom.i19.i.i.i.i.i681 = zext i32 %mul.i.i.i.i.i.i680 to i64
  %arrayidx.i20.i.i.i.i.i682 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %1009, i64 %idxprom.i19.i.i.i.i.i681
  %arrayidx.i22.i.i.i.i.i683 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %arrayidx.i20.i.i.i.i.i682, i64 %sh_prom.i.i.i.i
  store i64 1, ptr %arrayidx.i22.i.i.i.i.i683, align 8, !tbaa !82
  %arrayidx6.i.i.i.i.i.i.i684 = getelementptr inbounds nuw i8, ptr %arrayidx.i22.i.i.i.i.i683, i64 8
  store i8 10, ptr %arrayidx6.i.i.i.i.i.i.i684, align 8
  %tobool.not.i.i.i.i.i.i685 = icmp eq i32 %.lcssa4142, 0
  %conv.i.i.i.i.i.i686 = zext i1 %tobool.not.i.i.i.i.i.i685 to i32
  fence release
  %1010 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %idxprom.i23.i.i.i.i.i687 = zext i32 %.lcssa4144 to i64
  %arrayidx.i24.i.i.i.i.i688 = getelementptr inbounds nuw i32, ptr %1010, i64 %idxprom.i23.i.i.i.i.i687
  store atomic i32 %conv.i.i.i.i.i.i686, ptr %arrayidx.i24.i.i.i.i.i688 monotonic, align 4
  %1011 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !84
  %arrayidx.i.i5.i.i689 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %1011, i64 %idxprom.i23.i.i.i.i.i687, !amdgpu.uniform !42
  %1012 = load i64, ptr %arrayidx.i.i5.i.i689, align 8, !tbaa !80
  br label %while.cond.i.i.i.i690, !amdgpu.uniform !42

while.cond.i.i.i.i690:                            ; preds = %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i679, %Flow3739
  %phi.broken3936 = phi i64 [ 0, %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i679 ], [ %1027, %Flow3739 ]
  %port.sroa.37.0.i.i691 = phi i32 [ %1025, %Flow3739 ], [ %conv.i.i.i.i.i.i686, %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i679 ]
  %idx.0.i.i.i.i692 = phi i64 [ %1024, %Flow3739 ], [ 56, %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i679 ]
  %cmp.i.i.i.i693 = icmp eq i64 %idx.0.i.i.i.i692, 0
  %1013 = tail call i64 @llvm.amdgcn.ballot.i64(i1 %cmp.i.i.i.i693)
  %and.i.i.i.i6.i.i694 = and i64 %1013, %1012
  %tobool.not.i.i.i.i695 = icmp ne i64 %and.i.i.i.i6.i.i694, 0
  %1014 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i22.i.i.i696 = getelementptr inbounds nuw i32, ptr %1014, i64 %idxprom.i23.i.i.i.i.i687, !amdgpu.uniform !42
  %1015 = load atomic i32, ptr %arrayidx.i22.i.i.i696 monotonic, align 4
  %1016 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %tobool.not.i.i.i.i695)
  %1017 = extractvalue { i1, i64 } %1016, 0
  %1018 = extractvalue { i1, i64 } %1016, 1
  br i1 %1017, label %cond.false.i8.i.i.i.i697, label %Flow3739

cond.false.i8.i.i.i.i697:                         ; preds = %while.cond.i.i.i.i690
  %1019 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %1015)
  %1020 = icmp ne i32 %1019, %port.sroa.37.0.i.i691
  br i1 %1020, label %while.body.i.i21.i.i.i.i698.preheader, label %Flow3738, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i698.preheader:            ; preds = %cond.false.i8.i.i.i.i697
  br label %while.body.i.i21.i.i.i.i698, !amdgpu.uniform !42

Flow3738:                                         ; preds = %Flow3737, %cond.false.i8.i.i.i.i697
  br label %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i701, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i698:                      ; preds = %while.body.i.i21.i.i.i.i698.preheader, %while.body.i.i21.i.i.i.i698
  tail call void @llvm.amdgcn.s.sleep(i32 2)
  %1021 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i22.i.i.i.i699 = getelementptr inbounds nuw i32, ptr %1021, i64 %idxprom.i23.i.i.i.i.i687, !amdgpu.uniform !42
  %1022 = load atomic i32, ptr %arrayidx.i.i.i22.i.i.i.i699 monotonic, align 4
  %1023 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %1022)
  %cmp.i.not.i.i23.i.i.i.i700 = icmp eq i32 %1023, %port.sroa.37.0.i.i691
  br i1 %cmp.i.not.i.i23.i.i.i.i700, label %Flow3737, label %while.body.i.i21.i.i.i.i698, !amdgpu.uniform !42

Flow3737:                                         ; preds = %while.body.i.i21.i.i.i.i698
  br label %Flow3738, !amdgpu.uniform !42

Flow3739:                                         ; preds = %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i702, %while.cond.i.i.i.i690
  %1024 = phi i64 [ %add.i.i.i.i705, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i702 ], [ poison, %while.cond.i.i.i.i690 ]
  %1025 = phi i32 [ %conv.i.i26.i.i.i.i703, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i702 ], [ poison, %while.cond.i.i.i.i690 ]
  %1026 = phi i1 [ false, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i702 ], [ true, %while.cond.i.i.i.i690 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %1018)
  %1027 = call i64 @llvm.amdgcn.if.break.i64(i1 %1026, i64 %phi.broken3936)
  %1028 = call i1 @llvm.amdgcn.loop.i64(i64 %1027)
  br i1 %1028, label %cond.false.i.i.i711, label %while.cond.i.i.i.i690

_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i701: ; preds = %Flow3738
  fence acquire
  br i1 %cmp.i.i.i.i693, label %if.then.i.i.i.i.i.i.i706, label %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i702, !amdgpu.uniform !42

if.then.i.i.i.i.i.i.i706:                         ; preds = %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i701
  %1029 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %arrayidx.i20.i35.i.i.i.i709 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %1029, i64 %idxprom.i19.i.i.i.i.i681
  %arrayidx.i22.i38.i.i.i.i710 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %arrayidx.i20.i35.i.i.i.i709, i64 %sh_prom.i.i.i.i
  store i8 10, ptr %arrayidx.i22.i38.i.i.i.i710, align 1
  br label %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i702, !amdgpu.uniform !42

_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i702: ; preds = %if.then.i.i.i.i.i.i.i706, %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i701
  %conv.i.i26.i.i.i.i703 = xor i32 %port.sroa.37.0.i.i691, 1
  fence release
  %1030 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %arrayidx.i24.i29.i.i.i.i704 = getelementptr inbounds nuw i32, ptr %1030, i64 %idxprom.i23.i.i.i.i.i687
  store atomic i32 %conv.i.i26.i.i.i.i703, ptr %arrayidx.i24.i29.i.i.i.i704 monotonic, align 4
  %add.i.i.i.i705 = add i64 %idx.0.i.i.i.i692, 64
  br label %Flow3739, !amdgpu.uniform !42

cond.false.i.i.i711:                              ; preds = %Flow3739
  %.lcssa4138 = phi i64 [ %1027, %Flow3739 ]
  %port.sroa.37.0.i.i691.lcssa = phi i32 [ %port.sroa.37.0.i.i691, %Flow3739 ]
  %.lcssa4137 = phi i32 [ %1015, %Flow3739 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa4138)
  %1031 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %.lcssa4137)
  %1032 = icmp ne i32 %1031, %port.sroa.37.0.i.i691.lcssa
  %1033 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %1032)
  %1034 = extractvalue { i1, i64 } %1033, 0
  %1035 = extractvalue { i1, i64 } %1033, 1
  br i1 %1034, label %while.body.i.i.i.i712.preheader, label %Flow3736

while.body.i.i.i.i712.preheader:                  ; preds = %cond.false.i.i.i711
  br label %while.body.i.i.i.i712, !amdgpu.uniform !42

Flow3736:                                         ; preds = %Flow3735, %cond.false.i.i.i711
  call void @llvm.amdgcn.end.cf.i64(i64 %1035)
  br label %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIcEERS1_T_.exit, !amdgpu.uniform !42

while.body.i.i.i.i712:                            ; preds = %while.body.i.i.i.i712.preheader, %while.body.i.i.i.i712
  %phi.broken3937 = phi i64 [ 0, %while.body.i.i.i.i712.preheader ], [ %1039, %while.body.i.i.i.i712 ]
  tail call void @llvm.amdgcn.s.sleep(i32 2)
  %1036 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i8.i.i713 = getelementptr inbounds nuw i32, ptr %1036, i64 %idxprom.i23.i.i.i.i.i687, !amdgpu.uniform !42
  %1037 = load atomic i32, ptr %arrayidx.i.i.i8.i.i713 monotonic, align 4
  %1038 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %1037)
  %cmp.i.not.i.i.i.i714 = icmp eq i32 %1038, %port.sroa.37.0.i.i691.lcssa
  %1039 = call i64 @llvm.amdgcn.if.break.i64(i1 %cmp.i.not.i.i.i.i714, i64 %phi.broken3937)
  %1040 = call i1 @llvm.amdgcn.loop.i64(i64 %1039)
  br i1 %1040, label %Flow3735, label %while.body.i.i.i.i712

Flow3735:                                         ; preds = %while.body.i.i.i.i712
  %.lcssa4136 = phi i64 [ %1039, %while.body.i.i.i.i712 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa4136)
  br label %Flow3736, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIcEERS1_T_.exit: ; preds = %Flow3736
  fence acquire
  tail call void @llvm.amdgcn.wave.barrier()
  fence syncscope("agent") release
  %1041 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %.lcssa4143, i1 true)
  %iszero.i.i.i.i.i13.i.i716 = icmp ne i64 %.lcssa4143, 0
  %cmp2.i.i.i.i14.i.i717 = icmp eq i64 %1041, %sh_prom.i.i.i.i
  %cmp.i.i.i.i15.i.i718 = select i1 %iszero.i.i.i.i.i13.i.i716, i1 %cmp2.i.i.i.i14.i.i717, i1 false
  %1042 = and i32 %.lcssa4144, 31
  %conv4.i.i.i19.i.i722 = zext i1 %cmp.i.i.i.i15.i.i718 to i32
  %shl.i.i.i20.i.i723 = shl nuw i32 %conv4.i.i.i19.i.i722, %1042
  %xor.i.i.i21.i.i724 = xor i32 %shl.i.i.i20.i.i723, -1
  %1043 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2523, !amdgpu.uniform !42

1044:                                             ; preds = %ComputeEnd2524
  %div8.i.i.i16.i.i719 = lshr i32 %.lcssa4144, 5
  %idxprom.i.i.i17.i.i720 = zext nneg i32 %div8.i.i.i16.i.i719 to i64
  %1045 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i17.i.i720
  %sunkaddr3308 = getelementptr inbounds i8, ptr addrspace(1) %1045, i64 40
  %1046 = atomicrmw and ptr addrspace(1) %sunkaddr3308, i32 %.lcssa4135 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %1047, !amdgpu.uniform !42

1047:                                             ; preds = %1044, %ComputeEnd2524
  call void @llvm.amdgcn.end.cf.i64(i64 %3399)
  tail call void @llvm.amdgcn.wave.barrier()
  br i1 %cmp2.i.i.i35.i, label %if.then3.i.i.i65.i, label %if.end4.i.i.i54.i, !amdgpu.uniform !42

if.then3.i.i.i65.i:                               ; preds = %1047
  %mul.i.i.i66.i = mul nuw i64 %inc.i.i.i32.i, 11
  %div18.i.i.i67.i = lshr i64 %mul.i.i.i66.i, 3
  br label %if.end4.i.i.i54.i, !amdgpu.uniform !42

if.end4.i.i.i54.i:                                ; preds = %if.then3.i.i.i65.i, %1047
  %new_capacity.addr.0.i.i.i55.i = phi i64 [ %div18.i.i.i67.i, %if.then3.i.i.i65.i ], [ %inc.i.i.i32.i, %1047 ]
  %sub.i.i733 = add i64 %new_capacity.addr.0.i.i.i55.i, 7
  %div2.i.i734 = and i64 %sub.i.i733, -8
  %1048 = load ptr, ptr addrspace(1) @_ZN12_GLOBAL__N_13ptrE, align 8, !tbaa !55
  %add.ptr.i.i735 = getelementptr inbounds nuw i8, ptr %1048, i64 %div2.i.i734
  store ptr %add.ptr.i.i735, ptr addrspace(1) @_ZN12_GLOBAL__N_13ptrE, align 8, !tbaa !55
  br i1 %597, label %loop-memcpy-expansion2291.preheader, label %Flow3734, !amdgpu.uniform !42

loop-memcpy-expansion2291.preheader:              ; preds = %if.end4.i.i.i54.i
  br label %loop-memcpy-expansion2291, !amdgpu.uniform !42

Flow3734:                                         ; preds = %Flow3733, %if.end4.i.i.i54.i
  br label %loop-memcpy-residual-header2294, !amdgpu.uniform !42

loop-memcpy-expansion2291:                        ; preds = %loop-memcpy-expansion2291.preheader, %loop-memcpy-expansion2291
  %loop-index2292 = phi i64 [ %1052, %loop-memcpy-expansion2291 ], [ 0, %loop-memcpy-expansion2291.preheader ]
  %1049 = getelementptr i8, ptr %3, i64 %loop-index2292, !amdgpu.uniform !42
  %1050 = load <4 x i32>, ptr %1049, align 1
  %1051 = getelementptr i8, ptr %1048, i64 %loop-index2292
  store <4 x i32> %1050, ptr %1051, align 1
  %1052 = add i64 %loop-index2292, 16
  %1053 = icmp uge i64 %1052, %596
  br i1 %1053, label %Flow3733, label %loop-memcpy-expansion2291, !amdgpu.uniform !42

Flow3732:                                         ; preds = %Flow3731, %loop-memcpy-residual-header2294
  br label %post-loop-memcpy-expansion2290, !amdgpu.uniform !42

loop-memcpy-residual2293:                         ; preds = %loop-memcpy-residual2293.preheader, %loop-memcpy-residual2293
  %residual-loop-index2295 = phi i64 [ %1055, %loop-memcpy-residual2293 ], [ 0, %loop-memcpy-residual2293.preheader ]
  %scevgep3089 = getelementptr i8, ptr %scevgep3088, i64 %residual-loop-index2295, !amdgpu.uniform !42
  %1054 = load i8, ptr %scevgep3089, align 1
  %scevgep3087 = getelementptr i8, ptr %scevgep3086, i64 %residual-loop-index2295
  store i8 %1054, ptr %scevgep3087, align 1
  %1055 = add i64 %residual-loop-index2295, 1
  %1056 = icmp uge i64 %1055, %595
  br i1 %1056, label %Flow3731, label %loop-memcpy-residual2293, !amdgpu.uniform !42

Flow3731:                                         ; preds = %loop-memcpy-residual2293
  br label %Flow3732, !amdgpu.uniform !42

post-loop-memcpy-expansion2290:                   ; preds = %Flow3732
  %sub.ptr.lhs.cast.i.i736 = ptrtoint ptr %add.ptr.i.i735 to i64
  %sub.ptr.sub.i.i737 = sub i64 %sub.ptr.lhs.cast.i.i736, ptrtoint (ptr addrspacecast (ptr addrspace(1) @_ZN12_GLOBAL__N_16memoryE to ptr) to i64)
  %cmp.i.i738 = icmp ugt i64 %sub.ptr.sub.i.i737, 65335
  %cond.i.i739 = select i1 %cmp.i.i738, ptr null, ptr %1048
  %cmp.not.i10.i.i62.i = icmp ne ptr %cond.i.i739, addrspacecast (ptr addrspace(4) @_ZN22__llvm_libc_22_0_0_git3cpp6string14NULL_CHARACTERE to ptr)
  br i1 %cmp.not.i10.i.i62.i, label %if.then.i.i.i63.i, label %for.cond.i.i.i741.preheader, !amdgpu.uniform !42

if.then.i.i.i63.i:                                ; preds = %post-loop-memcpy-expansion2290
  %arrayidx.i.i.i64.i = getelementptr inbounds nuw i8, ptr %1048, i64 %spec.select.i
  store i8 0, ptr %arrayidx.i.i.i64.i, align 1, !tbaa !57
  br label %for.cond.i.i.i741.preheader, !amdgpu.uniform !42

for.cond.i.i.i741.preheader:                      ; preds = %if.then.i.i.i63.i, %post-loop-memcpy-expansion2290
  br label %for.cond.i.i.i741, !amdgpu.uniform !42

for.cond.i.i.i741:                                ; preds = %for.cond.i.i.i741.preheader, %Flow3730
  %1057 = phi i32 [ %1095, %Flow3730 ], [ poison, %for.cond.i.i.i741.preheader ]
  %1058 = phi i64 [ %1096, %Flow3730 ], [ poison, %for.cond.i.i.i741.preheader ]
  %1059 = phi i32 [ %1097, %Flow3730 ], [ poison, %for.cond.i.i.i741.preheader ]
  %retval.sroa.6.0.i.i.i742 = phi i32 [ %1100, %Flow3730 ], [ undef, %for.cond.i.i.i741.preheader ]
  %retval.sroa.2.0.i.i.i744 = phi i64 [ %1099, %Flow3730 ], [ undef, %for.cond.i.i.i741.preheader ]
  %retval.sroa.8.0.i.i.i745 = phi i32 [ %1098, %Flow3730 ], [ undef, %for.cond.i.i.i741.preheader ]
  %index.0.i.i.i746 = phi i32 [ %1102, %Flow3730 ], [ 0, %for.cond.i.i.i741.preheader ]
  %1060 = load i32, ptr addrspace(1) @__llvm_rpc_client, align 8, !tbaa !68
  %cmp.not.i.i.i747 = icmp ult i32 %index.0.i.i.i746, %1060
  %spec.store.select.i.i.i748 = select i1 %cmp.not.i.i.i747, i32 %index.0.i.i.i746, i32 0
  %1061 = tail call noundef i64 @llvm.amdgcn.ballot.i64(i1 true)
  %1062 = and i32 %spec.store.select.i.i.i748, 31
  %1063 = lshr i64 %1061, %sh_prom.i.i.i.i
  %1064 = trunc i64 %1063 to i32
  %conv4.i.i.i.i.i752 = and i32 %1064, 1
  %shl.i.i.i.i.i753 = shl nuw i32 %conv4.i.i.i.i.i752, %1062
  %1065 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2527, !amdgpu.uniform !42

1066:                                             ; preds = %ComputeEnd2528
  %sunkaddr3309 = getelementptr inbounds i8, ptr addrspace(1) %3409, i64 40
  %1067 = atomicrmw or ptr addrspace(1) %sunkaddr3309, i32 %.lcssa4129 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %1068, !amdgpu.uniform !42

1068:                                             ; preds = %1066, %ComputeEnd2528
  %1069 = phi i32 [ %1067, %1066 ], [ poison, %ComputeEnd2528 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %3418)
  %1070 = call i32 @llvm.amdgcn.readfirstlane.i32(i32 %1069)
  %1071 = or i32 %1070, %.lcssa4130
  %shl5.i.i.i.i.i754 = shl nuw i32 1, %1062
  %and.i.i.i.i.i755 = and i32 %shl5.i.i.i.i.i754, %1071
  %tobool3.i.i.i.i756 = icmp ne i32 %and.i.i.i.i.i755, 0
  %1072 = tail call i64 @llvm.amdgcn.ballot.i64(i1 %tobool3.i.i.i.i756)
  %and.i.i.i.i.i.i757 = and i64 %1072, %1061
  %cmp.i.not.i.i.i758 = icmp ne i64 %1061, %and.i.i.i.i.i.i757
  br i1 %cmp.i.not.i.i.i758, label %if.then.i.i.i.i759, label %Flow3729, !amdgpu.uniform !42

if.then.i.i.i.i759:                               ; preds = %1068
  fence syncscope("agent") acquire
  %1073 = load <2 x i64>, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !75
  %1074 = extractelement <2 x i64> %1073, i32 0
  %1075 = inttoptr i64 %1074 to ptr
  %1076 = extractelement <2 x i64> %1073, i32 1
  %1077 = inttoptr i64 %1076 to ptr
  %idxprom.i.i.i.i760 = zext i32 %spec.store.select.i.i.i748 to i64
  %arrayidx.i.i.i.i761 = getelementptr inbounds nuw i32, ptr %1075, i64 %idxprom.i.i.i.i760, !amdgpu.uniform !42
  %1078 = load atomic i32, ptr %arrayidx.i.i.i.i761 monotonic, align 4
  %1079 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %1078)
  %arrayidx.i45.i.i.i762 = getelementptr inbounds nuw i32, ptr %1077, i64 %idxprom.i.i.i.i760, !amdgpu.uniform !42
  %1080 = load atomic i32, ptr %arrayidx.i45.i.i.i762 monotonic, align 4
  %1081 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %1080)
  %cmp.i46.not.i.i.i763 = icmp ne i32 %1079, %1081
  br i1 %cmp.i46.not.i.i.i763, label %if.then12.i.i.i764, label %Flow3727, !amdgpu.uniform !42

if.then12.i.i.i764:                               ; preds = %if.then.i.i.i.i759
  fence syncscope("agent") release
  %1082 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %1061, i1 true)
  %iszero.i.i.i.i.i.i.i765 = icmp ne i64 %1061, 0
  %cmp2.i.i.i.i.i.i766 = icmp eq i64 %1082, %sh_prom.i.i.i.i
  %cmp.i.i.i.i.i.i767 = select i1 %iszero.i.i.i.i.i.i.i765, i1 %cmp2.i.i.i.i.i.i766, i1 false
  %conv4.i.i51.i.i.i768 = zext i1 %cmp.i.i.i.i.i.i767 to i32
  %shl.i.i52.i.i.i769 = shl nuw i32 %conv4.i.i51.i.i.i768, %1062
  %xor.i.i.i.i.i770 = xor i32 %shl.i.i52.i.i.i769, -1
  %1083 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2532, !amdgpu.uniform !42

1084:                                             ; preds = %ComputeEnd2533
  %sunkaddr3310 = getelementptr inbounds i8, ptr addrspace(1) %3409, i64 40
  %1085 = atomicrmw and ptr addrspace(1) %sunkaddr3310, i32 %.lcssa4131 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %1086, !amdgpu.uniform !42

1086:                                             ; preds = %1084, %ComputeEnd2533
  call void @llvm.amdgcn.end.cf.i64(i64 %3436)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %Flow3727, !amdgpu.uniform !42

if.end14.i.i.i830:                                ; preds = %Flow3727
  %1087 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %1061, i1 true)
  %iszero.i.i.i.i.i.i831 = icmp ne i64 %1061, 0
  %cmp2.i.i.i.i.i832 = icmp eq i64 %1087, %sh_prom.i.i.i.i
  %cmp.i.i.i.i.i833 = select i1 %iszero.i.i.i.i.i.i831, i1 %cmp2.i.i.i.i.i832, i1 false
  %1088 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.i.i.i.i.i833)
  %1089 = extractvalue { i1, i64 } %1088, 0
  %1090 = extractvalue { i1, i64 } %1088, 1
  br i1 %1089, label %if.then16.i.i.i835, label %if.end22.i.i.i834

if.then16.i.i.i835:                               ; preds = %if.end14.i.i.i830
  %1091 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !76
  %opcode.i.i.i836 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %1091, i64 %idxprom.i.i.i.i760, i32 1
  store i32 1660944387, ptr %opcode.i.i.i836, align 8, !tbaa !77
  %arrayidx21.i.i.i837 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %1091, i64 %idxprom.i.i.i.i760
  store i64 %1061, ptr %arrayidx21.i.i.i837, align 8, !tbaa !80
  br label %if.end22.i.i.i834, !amdgpu.uniform !42

Flow3728:                                         ; preds = %if.end22.i.i.i834, %Flow3727
  %1092 = phi i32 [ %1079, %if.end22.i.i.i834 ], [ %retval.sroa.8.0.i.i.i745, %Flow3727 ]
  %1093 = phi i64 [ %1061, %if.end22.i.i.i834 ], [ %retval.sroa.2.0.i.i.i744, %Flow3727 ]
  %1094 = phi i32 [ %spec.store.select.i.i.i748, %if.end22.i.i.i834 ], [ %retval.sroa.6.0.i.i.i742, %Flow3727 ]
  br label %cleanup26.i.i.i771, !amdgpu.uniform !42

if.end22.i.i.i834:                                ; preds = %if.then16.i.i.i835, %if.end14.i.i.i830
  call void @llvm.amdgcn.end.cf.i64(i64 %1090)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %Flow3728, !amdgpu.uniform !42

Flow3729:                                         ; preds = %cleanup26.i.i.i771, %1068
  %1095 = phi i32 [ %1094, %cleanup26.i.i.i771 ], [ %1057, %1068 ]
  %1096 = phi i64 [ %1093, %cleanup26.i.i.i771 ], [ %1058, %1068 ]
  %1097 = phi i32 [ %1092, %cleanup26.i.i.i771 ], [ %1059, %1068 ]
  %1098 = phi i32 [ %1092, %cleanup26.i.i.i771 ], [ %retval.sroa.8.0.i.i.i745, %1068 ]
  %1099 = phi i64 [ %1093, %cleanup26.i.i.i771 ], [ %retval.sroa.2.0.i.i.i744, %1068 ]
  %1100 = phi i32 [ %1094, %cleanup26.i.i.i771 ], [ %retval.sroa.6.0.i.i.i742, %1068 ]
  %1101 = phi i1 [ %cmp.i46.not.i.i.i763, %cleanup26.i.i.i771 ], [ true, %1068 ]
  br i1 %1101, label %for.inc.i.i.i776, label %Flow3730, !amdgpu.uniform !42

cleanup26.i.i.i771:                               ; preds = %Flow3728
  br label %Flow3729, !amdgpu.uniform !42

for.inc.i.i.i776:                                 ; preds = %Flow3729
  %inc.i.i.i781 = add i32 %spec.store.select.i.i.i748, 1
  br label %Flow3730, !amdgpu.uniform !42

Flow3730:                                         ; preds = %for.inc.i.i.i776, %Flow3729
  %1102 = phi i32 [ %inc.i.i.i781, %for.inc.i.i.i776 ], [ poison, %Flow3729 ]
  %1103 = phi i1 [ false, %for.inc.i.i.i776 ], [ true, %Flow3729 ]
  br i1 %1103, label %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i782, label %for.cond.i.i.i741, !amdgpu.uniform !42

_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i782: ; preds = %Flow3730
  %.lcssa4134 = phi i32 [ %1095, %Flow3730 ]
  %.lcssa4133 = phi i64 [ %1096, %Flow3730 ]
  %.lcssa4132 = phi i32 [ %1097, %Flow3730 ]
  fence acquire
  %1104 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %mul.i.i.i.i.i.i783 = shl i32 %.lcssa4134, 6
  %idxprom.i19.i.i.i.i.i784 = zext i32 %mul.i.i.i.i.i.i783 to i64
  %arrayidx.i20.i.i.i.i.i785 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %1104, i64 %idxprom.i19.i.i.i.i.i784
  %arrayidx.i22.i.i.i.i.i786 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %arrayidx.i20.i.i.i.i.i785, i64 %sh_prom.i.i.i.i
  store i64 %spec.select.i, ptr %arrayidx.i22.i.i.i.i.i786, align 8, !tbaa !82
  br i1 %656, label %loop-memcpy-expansion2243.preheader, label %Flow3726, !amdgpu.uniform !42

loop-memcpy-expansion2243.preheader:              ; preds = %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i782
  %1105 = add nuw nsw i64 %75, 8
  %scevgep3083 = getelementptr i8, ptr %1104, i64 %1105
  %1106 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i784, 6
  %scevgep3084 = getelementptr i8, ptr %scevgep3083, i64 %1106
  br label %loop-memcpy-expansion2243, !amdgpu.uniform !42

Flow3726:                                         ; preds = %Flow3725, %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i782
  br label %loop-memcpy-residual-header2246, !amdgpu.uniform !42

loop-memcpy-expansion2243:                        ; preds = %loop-memcpy-expansion2243.preheader, %loop-memcpy-expansion2243
  %loop-index2244 = phi i64 [ %1109, %loop-memcpy-expansion2243 ], [ 0, %loop-memcpy-expansion2243.preheader ]
  %1107 = getelementptr i8, ptr %1048, i64 %loop-index2244, !amdgpu.uniform !42
  %1108 = load <4 x i32>, ptr %1107, align 1
  %scevgep3085 = getelementptr i8, ptr %scevgep3084, i64 %loop-index2244
  store <4 x i32> %1108, ptr %scevgep3085, align 1
  %1109 = add i64 %loop-index2244, 16
  %1110 = icmp uge i64 %1109, %655
  br i1 %1110, label %Flow3725, label %loop-memcpy-expansion2243, !amdgpu.uniform !42

Flow3724:                                         ; preds = %Flow3722, %Flow3723
  %1111 = phi i64 [ %2812, %Flow3722 ], [ %2808, %Flow3723 ]
  br label %post-loop-memcpy-expansion2242, !amdgpu.uniform !42

loop-memcpy-residual2245:                         ; preds = %loop-memcpy-residual2245.preheader, %loop-memcpy-residual2245
  %residual-loop-index2247 = phi i64 [ %1113, %loop-memcpy-residual2245 ], [ 0, %loop-memcpy-residual2245.preheader ]
  %scevgep3082 = getelementptr i8, ptr %scevgep3081, i64 %residual-loop-index2247, !amdgpu.uniform !42
  %1112 = load i8, ptr %scevgep3082, align 1
  %scevgep3080 = getelementptr i8, ptr %scevgep3079, i64 %residual-loop-index2247
  store i8 %1112, ptr %scevgep3080, align 1
  %1113 = add i64 %residual-loop-index2247, 1
  %1114 = icmp uge i64 %1113, %654
  br i1 %1114, label %Flow3722, label %loop-memcpy-residual2245, !amdgpu.uniform !42

Flow3722:                                         ; preds = %loop-memcpy-residual2245
  br label %Flow3724, !amdgpu.uniform !42

post-loop-memcpy-expansion2242:                   ; preds = %Flow3724
  %tobool.not.i.i.i.i.i.i789 = icmp eq i32 %.lcssa4132, 0
  %conv.i.i.i.i.i.i790 = zext i1 %tobool.not.i.i.i.i.i.i789 to i32
  fence release
  %1115 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %idxprom.i23.i.i.i.i.i791 = zext i32 %.lcssa4134 to i64
  %arrayidx.i24.i.i.i.i.i792 = getelementptr inbounds nuw i32, ptr %1115, i64 %idxprom.i23.i.i.i.i.i791
  store atomic i32 %conv.i.i.i.i.i.i790, ptr %arrayidx.i24.i.i.i.i.i792 monotonic, align 4
  %1116 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !84
  %arrayidx.i.i5.i.i793 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %1116, i64 %idxprom.i23.i.i.i.i.i791, !amdgpu.uniform !42
  %1117 = load i64, ptr %arrayidx.i.i5.i.i793, align 8, !tbaa !80
  %scevgep3064 = getelementptr i8, ptr %1048, i64 56
  br label %while.cond.i.i.i.i794, !amdgpu.uniform !42

while.cond.i.i.i.i794:                            ; preds = %post-loop-memcpy-expansion2242, %Flow3721
  %phi.broken3938 = phi i64 [ 0, %post-loop-memcpy-expansion2242 ], [ %1134, %Flow3721 ]
  %lsr.iv3071 = phi i64 [ %1132, %Flow3721 ], [ %670, %post-loop-memcpy-expansion2242 ]
  %lsr.iv3065 = phi ptr [ %1131, %Flow3721 ], [ %scevgep3064, %post-loop-memcpy-expansion2242 ]
  %port.sroa.37.0.i.i795 = phi i32 [ %1130, %Flow3721 ], [ %conv.i.i.i.i.i.i790, %post-loop-memcpy-expansion2242 ]
  %idx.0.i.i.i.i796 = phi i64 [ %1129, %Flow3721 ], [ 56, %post-loop-memcpy-expansion2242 ]
  %cmp.i.i.i.i797 = icmp ugt i64 %spec.select.i, %idx.0.i.i.i.i796
  %1118 = tail call i64 @llvm.amdgcn.ballot.i64(i1 %cmp.i.i.i.i797)
  %and.i.i.i.i6.i.i798 = and i64 %1118, %1117
  %tobool.not.i.i.i.i799 = icmp ne i64 %and.i.i.i.i6.i.i798, 0
  %1119 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i22.i.i.i800 = getelementptr inbounds nuw i32, ptr %1119, i64 %idxprom.i23.i.i.i.i.i791, !amdgpu.uniform !42
  %1120 = load atomic i32, ptr %arrayidx.i22.i.i.i800 monotonic, align 4
  %1121 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %tobool.not.i.i.i.i799)
  %1122 = extractvalue { i1, i64 } %1121, 0
  %1123 = extractvalue { i1, i64 } %1121, 1
  br i1 %1122, label %cond.false.i8.i.i.i.i801, label %Flow3721

cond.false.i8.i.i.i.i801:                         ; preds = %while.cond.i.i.i.i794
  %1124 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %1120)
  %1125 = icmp ne i32 %1124, %port.sroa.37.0.i.i795
  br i1 %1125, label %while.body.i.i21.i.i.i.i802.preheader, label %Flow3720, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i802.preheader:            ; preds = %cond.false.i8.i.i.i.i801
  br label %while.body.i.i21.i.i.i.i802, !amdgpu.uniform !42

Flow3720:                                         ; preds = %Flow3719, %cond.false.i8.i.i.i.i801
  br label %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i805, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i802:                      ; preds = %while.body.i.i21.i.i.i.i802.preheader, %while.body.i.i21.i.i.i.i802
  tail call void @llvm.amdgcn.s.sleep(i32 2)
  %1126 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i22.i.i.i.i803 = getelementptr inbounds nuw i32, ptr %1126, i64 %idxprom.i23.i.i.i.i.i791, !amdgpu.uniform !42
  %1127 = load atomic i32, ptr %arrayidx.i.i.i22.i.i.i.i803 monotonic, align 4
  %1128 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %1127)
  %cmp.i.not.i.i23.i.i.i.i804 = icmp eq i32 %1128, %port.sroa.37.0.i.i795
  br i1 %cmp.i.not.i.i23.i.i.i.i804, label %Flow3719, label %while.body.i.i21.i.i.i.i802, !amdgpu.uniform !42

Flow3719:                                         ; preds = %while.body.i.i21.i.i.i.i802
  br label %Flow3720, !amdgpu.uniform !42

Flow3721:                                         ; preds = %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i806, %while.cond.i.i.i.i794
  %1129 = phi i64 [ %add.i.i.i.i809, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i806 ], [ poison, %while.cond.i.i.i.i794 ]
  %1130 = phi i32 [ %conv.i.i26.i.i.i.i807, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i806 ], [ poison, %while.cond.i.i.i.i794 ]
  %1131 = phi ptr [ %scevgep3066, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i806 ], [ poison, %while.cond.i.i.i.i794 ]
  %1132 = phi i64 [ %lsr.iv.next3072, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i806 ], [ poison, %while.cond.i.i.i.i794 ]
  %1133 = phi i1 [ false, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i806 ], [ true, %while.cond.i.i.i.i794 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %1123)
  %1134 = call i64 @llvm.amdgcn.if.break.i64(i1 %1133, i64 %phi.broken3938)
  %1135 = call i1 @llvm.amdgcn.loop.i64(i64 %1134)
  br i1 %1135, label %cond.false.i.i.i816, label %while.cond.i.i.i.i794

_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i805: ; preds = %Flow3720
  fence acquire
  br i1 %cmp.i.i.i.i797, label %if.then.i.i.i.i.i.i.i810, label %Flow3718, !amdgpu.uniform !42

if.then.i.i.i.i.i.i.i810:                         ; preds = %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i805
  %sub.i.i.i.i.i.i.i812 = sub nuw i64 %spec.select.i, %idx.0.i.i.i.i796
  %spec.select.i.i.i36.i.i.i.i813 = tail call i64 @llvm.umin.i64(i64 %sub.i.i.i.i.i.i.i812, i64 64)
  %1136 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %1137 = and i64 %spec.select.i.i.i36.i.i.i.i813, 15
  %1138 = sub i64 %spec.select.i.i.i36.i.i.i.i813, %1137
  %1139 = icmp ne i64 %1138, 0
  br i1 %1139, label %loop-memcpy-expansion2237.preheader, label %Flow3717, !amdgpu.uniform !42

loop-memcpy-expansion2237.preheader:              ; preds = %if.then.i.i.i.i.i.i.i810
  %arrayidx.i20.i35.i.i.i.i814 = getelementptr %"struct.rpc::Buffer", ptr %1136, i64 %idxprom.i19.i.i.i.i.i784
  %arrayidx.i22.i38.i.i.i.i815 = getelementptr %"struct.rpc::Buffer", ptr %arrayidx.i20.i35.i.i.i.i814, i64 %sh_prom.i.i.i.i
  br label %loop-memcpy-expansion2237, !amdgpu.uniform !42

Flow3717:                                         ; preds = %Flow3716, %if.then.i.i.i.i.i.i.i810
  br label %loop-memcpy-residual-header2240, !amdgpu.uniform !42

loop-memcpy-expansion2237:                        ; preds = %loop-memcpy-expansion2237.preheader, %loop-memcpy-expansion2237
  %loop-index2238 = phi i64 [ %1142, %loop-memcpy-expansion2237 ], [ 0, %loop-memcpy-expansion2237.preheader ]
  %scevgep3067 = getelementptr i8, ptr %lsr.iv3065, i64 %loop-index2238, !amdgpu.uniform !42
  %1140 = load <4 x i32>, ptr %scevgep3067, align 1
  %1141 = getelementptr i8, ptr %arrayidx.i22.i38.i.i.i.i815, i64 %loop-index2238
  store <4 x i32> %1140, ptr %1141, align 1
  %1142 = add i64 %loop-index2238, 16
  %1143 = icmp uge i64 %1142, %1138
  br i1 %1143, label %Flow3716, label %loop-memcpy-expansion2237, !amdgpu.uniform !42

Flow3715:                                         ; preds = %Flow3714, %loop-memcpy-residual-header2240
  br label %Flow3718, !amdgpu.uniform !42

loop-memcpy-residual2239:                         ; preds = %loop-memcpy-residual2239.preheader, %loop-memcpy-residual2239
  %residual-loop-index2241 = phi i64 [ %1145, %loop-memcpy-residual2239 ], [ 0, %loop-memcpy-residual2239.preheader ]
  %scevgep3077 = getelementptr i8, ptr %scevgep3076, i64 %residual-loop-index2241, !amdgpu.uniform !42
  %1144 = load i8, ptr %scevgep3077, align 1
  %scevgep3075 = getelementptr i8, ptr %scevgep3074, i64 %residual-loop-index2241
  store i8 %1144, ptr %scevgep3075, align 1
  %1145 = add i64 %residual-loop-index2241, 1
  %1146 = icmp uge i64 %1145, %1137
  br i1 %1146, label %Flow3714, label %loop-memcpy-residual2239, !amdgpu.uniform !42

Flow3714:                                         ; preds = %loop-memcpy-residual2239
  br label %Flow3715, !amdgpu.uniform !42

_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i806: ; preds = %Flow3718
  %conv.i.i26.i.i.i.i807 = xor i32 %port.sroa.37.0.i.i795, 1
  fence release
  %1147 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %arrayidx.i24.i29.i.i.i.i808 = getelementptr inbounds nuw i32, ptr %1147, i64 %idxprom.i23.i.i.i.i.i791
  store atomic i32 %conv.i.i26.i.i.i.i807, ptr %arrayidx.i24.i29.i.i.i.i808 monotonic, align 4
  %add.i.i.i.i809 = add i64 %idx.0.i.i.i.i796, 64
  %scevgep3066 = getelementptr i8, ptr %lsr.iv3065, i64 64
  %lsr.iv.next3072 = add i64 %lsr.iv3071, -64
  br label %Flow3721, !amdgpu.uniform !42

cond.false.i.i.i816:                              ; preds = %Flow3721
  %.lcssa4128 = phi i64 [ %1134, %Flow3721 ]
  %port.sroa.37.0.i.i795.lcssa = phi i32 [ %port.sroa.37.0.i.i795, %Flow3721 ]
  %.lcssa4127 = phi i32 [ %1120, %Flow3721 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa4128)
  %1148 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %.lcssa4127)
  %1149 = icmp ne i32 %1148, %port.sroa.37.0.i.i795.lcssa
  %1150 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %1149)
  %1151 = extractvalue { i1, i64 } %1150, 0
  %1152 = extractvalue { i1, i64 } %1150, 1
  br i1 %1151, label %while.body.i.i.i.i817.preheader, label %Flow3713

while.body.i.i.i.i817.preheader:                  ; preds = %cond.false.i.i.i816
  br label %while.body.i.i.i.i817, !amdgpu.uniform !42

Flow3713:                                         ; preds = %Flow3712, %cond.false.i.i.i816
  call void @llvm.amdgcn.end.cf.i64(i64 %1152)
  br label %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsINS_3cpp6stringEEERS1_T_.exit838, !amdgpu.uniform !42

while.body.i.i.i.i817:                            ; preds = %while.body.i.i.i.i817.preheader, %while.body.i.i.i.i817
  %phi.broken3939 = phi i64 [ 0, %while.body.i.i.i.i817.preheader ], [ %1156, %while.body.i.i.i.i817 ]
  tail call void @llvm.amdgcn.s.sleep(i32 2)
  %1153 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i8.i.i818 = getelementptr inbounds nuw i32, ptr %1153, i64 %idxprom.i23.i.i.i.i.i791, !amdgpu.uniform !42
  %1154 = load atomic i32, ptr %arrayidx.i.i.i8.i.i818 monotonic, align 4
  %1155 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %1154)
  %cmp.i.not.i.i.i.i819 = icmp eq i32 %1155, %port.sroa.37.0.i.i795.lcssa
  %1156 = call i64 @llvm.amdgcn.if.break.i64(i1 %cmp.i.not.i.i.i.i819, i64 %phi.broken3939)
  %1157 = call i1 @llvm.amdgcn.loop.i64(i64 %1156)
  br i1 %1157, label %Flow3712, label %while.body.i.i.i.i817

Flow3712:                                         ; preds = %while.body.i.i.i.i817
  %.lcssa4126 = phi i64 [ %1156, %while.body.i.i.i.i817 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa4126)
  br label %Flow3713, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsINS_3cpp6stringEEERS1_T_.exit838: ; preds = %Flow3713
  fence acquire
  tail call void @llvm.amdgcn.wave.barrier()
  fence syncscope("agent") release
  %1158 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %.lcssa4133, i1 true)
  %iszero.i.i.i.i.i13.i.i821 = icmp ne i64 %.lcssa4133, 0
  %cmp2.i.i.i.i14.i.i822 = icmp eq i64 %1158, %sh_prom.i.i.i.i
  %cmp.i.i.i.i15.i.i823 = select i1 %iszero.i.i.i.i.i13.i.i821, i1 %cmp2.i.i.i.i14.i.i822, i1 false
  %1159 = and i32 %.lcssa4134, 31
  %conv4.i.i.i19.i.i827 = zext i1 %cmp.i.i.i.i15.i.i823 to i32
  %shl.i.i.i20.i.i828 = shl nuw i32 %conv4.i.i.i19.i.i827, %1159
  %xor.i.i.i21.i.i829 = xor i32 %shl.i.i.i20.i.i828, -1
  %1160 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2536, !amdgpu.uniform !42

1161:                                             ; preds = %ComputeEnd2537
  %div8.i.i.i16.i.i824 = lshr i32 %.lcssa4134, 5
  %idxprom.i.i.i17.i.i825 = zext nneg i32 %div8.i.i.i16.i.i824 to i64
  %1162 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i17.i.i825
  %sunkaddr3311 = getelementptr inbounds i8, ptr addrspace(1) %1162, i64 40
  %1163 = atomicrmw and ptr addrspace(1) %sunkaddr3311, i32 %.lcssa4125 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %1164, !amdgpu.uniform !42

1164:                                             ; preds = %1161, %ComputeEnd2537
  call void @llvm.amdgcn.end.cf.i64(i64 %3453)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %for.cond.i.i1.i847, !amdgpu.uniform !42

for.cond.i.i1.i847:                               ; preds = %1164, %Flow3711
  %1165 = phi i32 [ %1203, %Flow3711 ], [ poison, %1164 ]
  %1166 = phi i64 [ %1204, %Flow3711 ], [ poison, %1164 ]
  %1167 = phi i32 [ %1205, %Flow3711 ], [ poison, %1164 ]
  %retval.sroa.6.0.i.i.i848 = phi i32 [ %1208, %Flow3711 ], [ undef, %1164 ]
  %retval.sroa.2.0.i.i.i850 = phi i64 [ %1207, %Flow3711 ], [ undef, %1164 ]
  %retval.sroa.8.0.i.i.i851 = phi i32 [ %1206, %Flow3711 ], [ undef, %1164 ]
  %index.0.i.i.i852 = phi i32 [ %1210, %Flow3711 ], [ 0, %1164 ]
  %1168 = load i32, ptr addrspace(1) @__llvm_rpc_client, align 8, !tbaa !68
  %cmp.not.i.i.i853 = icmp ult i32 %index.0.i.i.i852, %1168
  %spec.store.select.i.i.i854 = select i1 %cmp.not.i.i.i853, i32 %index.0.i.i.i852, i32 0
  %1169 = tail call noundef i64 @llvm.amdgcn.ballot.i64(i1 true)
  %1170 = and i32 %spec.store.select.i.i.i854, 31
  %1171 = lshr i64 %1169, %sh_prom.i.i.i.i
  %1172 = trunc i64 %1171 to i32
  %conv4.i.i.i.i.i858 = and i32 %1172, 1
  %shl.i.i.i.i.i859 = shl nuw i32 %conv4.i.i.i.i.i858, %1170
  %1173 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2540, !amdgpu.uniform !42

1174:                                             ; preds = %ComputeEnd2541
  %sunkaddr3312 = getelementptr inbounds i8, ptr addrspace(1) %3463, i64 40
  %1175 = atomicrmw or ptr addrspace(1) %sunkaddr3312, i32 %.lcssa4119 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %1176, !amdgpu.uniform !42

1176:                                             ; preds = %1174, %ComputeEnd2541
  %1177 = phi i32 [ %1175, %1174 ], [ poison, %ComputeEnd2541 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %3472)
  %1178 = call i32 @llvm.amdgcn.readfirstlane.i32(i32 %1177)
  %1179 = or i32 %1178, %.lcssa4120
  %shl5.i.i.i.i.i860 = shl nuw i32 1, %1170
  %and.i.i.i.i.i861 = and i32 %shl5.i.i.i.i.i860, %1179
  %tobool3.i.i.i.i862 = icmp ne i32 %and.i.i.i.i.i861, 0
  %1180 = tail call i64 @llvm.amdgcn.ballot.i64(i1 %tobool3.i.i.i.i862)
  %and.i.i.i.i.i.i863 = and i64 %1180, %1169
  %cmp.i.not.i.i.i864 = icmp ne i64 %1169, %and.i.i.i.i.i.i863
  br i1 %cmp.i.not.i.i.i864, label %if.then.i.i.i.i865, label %Flow3710, !amdgpu.uniform !42

if.then.i.i.i.i865:                               ; preds = %1176
  fence syncscope("agent") acquire
  %1181 = load <2 x i64>, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !75
  %1182 = extractelement <2 x i64> %1181, i32 0
  %1183 = inttoptr i64 %1182 to ptr
  %1184 = extractelement <2 x i64> %1181, i32 1
  %1185 = inttoptr i64 %1184 to ptr
  %idxprom.i.i.i.i866 = zext i32 %spec.store.select.i.i.i854 to i64
  %arrayidx.i.i.i.i867 = getelementptr inbounds nuw i32, ptr %1183, i64 %idxprom.i.i.i.i866, !amdgpu.uniform !42
  %1186 = load atomic i32, ptr %arrayidx.i.i.i.i867 monotonic, align 4
  %1187 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %1186)
  %arrayidx.i45.i.i.i868 = getelementptr inbounds nuw i32, ptr %1185, i64 %idxprom.i.i.i.i866, !amdgpu.uniform !42
  %1188 = load atomic i32, ptr %arrayidx.i45.i.i.i868 monotonic, align 4
  %1189 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %1188)
  %cmp.i46.not.i.i.i869 = icmp ne i32 %1187, %1189
  br i1 %cmp.i46.not.i.i.i869, label %if.then12.i.i.i870, label %Flow3708, !amdgpu.uniform !42

if.then12.i.i.i870:                               ; preds = %if.then.i.i.i.i865
  fence syncscope("agent") release
  %1190 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %1169, i1 true)
  %iszero.i.i.i.i.i.i.i871 = icmp ne i64 %1169, 0
  %cmp2.i.i.i.i.i.i872 = icmp eq i64 %1190, %sh_prom.i.i.i.i
  %cmp.i.i.i.i.i.i873 = select i1 %iszero.i.i.i.i.i.i.i871, i1 %cmp2.i.i.i.i.i.i872, i1 false
  %conv4.i.i51.i.i.i874 = zext i1 %cmp.i.i.i.i.i.i873 to i32
  %shl.i.i52.i.i.i875 = shl nuw i32 %conv4.i.i51.i.i.i874, %1170
  %xor.i.i.i.i.i876 = xor i32 %shl.i.i52.i.i.i875, -1
  %1191 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2545, !amdgpu.uniform !42

1192:                                             ; preds = %ComputeEnd2546
  %sunkaddr3313 = getelementptr inbounds i8, ptr addrspace(1) %3463, i64 40
  %1193 = atomicrmw and ptr addrspace(1) %sunkaddr3313, i32 %.lcssa4121 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %1194, !amdgpu.uniform !42

1194:                                             ; preds = %1192, %ComputeEnd2546
  call void @llvm.amdgcn.end.cf.i64(i64 %3490)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %Flow3708, !amdgpu.uniform !42

if.end14.i.i.i936:                                ; preds = %Flow3708
  %1195 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %1169, i1 true)
  %iszero.i.i.i.i.i.i937 = icmp ne i64 %1169, 0
  %cmp2.i.i.i.i.i938 = icmp eq i64 %1195, %sh_prom.i.i.i.i
  %cmp.i.i.i.i.i939 = select i1 %iszero.i.i.i.i.i.i937, i1 %cmp2.i.i.i.i.i938, i1 false
  %1196 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.i.i.i.i.i939)
  %1197 = extractvalue { i1, i64 } %1196, 0
  %1198 = extractvalue { i1, i64 } %1196, 1
  br i1 %1197, label %if.then16.i.i.i941, label %if.end22.i.i.i940

if.then16.i.i.i941:                               ; preds = %if.end14.i.i.i936
  %1199 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !76
  %opcode.i.i.i942 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %1199, i64 %idxprom.i.i.i.i866, i32 1
  store i32 1660944387, ptr %opcode.i.i.i942, align 8, !tbaa !77
  %arrayidx21.i.i.i943 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %1199, i64 %idxprom.i.i.i.i866
  store i64 %1169, ptr %arrayidx21.i.i.i943, align 8, !tbaa !80
  br label %if.end22.i.i.i940, !amdgpu.uniform !42

Flow3709:                                         ; preds = %if.end22.i.i.i940, %Flow3708
  %1200 = phi i32 [ %1187, %if.end22.i.i.i940 ], [ %retval.sroa.8.0.i.i.i851, %Flow3708 ]
  %1201 = phi i64 [ %1169, %if.end22.i.i.i940 ], [ %retval.sroa.2.0.i.i.i850, %Flow3708 ]
  %1202 = phi i32 [ %spec.store.select.i.i.i854, %if.end22.i.i.i940 ], [ %retval.sroa.6.0.i.i.i848, %Flow3708 ]
  br label %cleanup26.i.i.i877, !amdgpu.uniform !42

if.end22.i.i.i940:                                ; preds = %if.then16.i.i.i941, %if.end14.i.i.i936
  call void @llvm.amdgcn.end.cf.i64(i64 %1198)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %Flow3709, !amdgpu.uniform !42

Flow3710:                                         ; preds = %cleanup26.i.i.i877, %1176
  %1203 = phi i32 [ %1202, %cleanup26.i.i.i877 ], [ %1165, %1176 ]
  %1204 = phi i64 [ %1201, %cleanup26.i.i.i877 ], [ %1166, %1176 ]
  %1205 = phi i32 [ %1200, %cleanup26.i.i.i877 ], [ %1167, %1176 ]
  %1206 = phi i32 [ %1200, %cleanup26.i.i.i877 ], [ %retval.sroa.8.0.i.i.i851, %1176 ]
  %1207 = phi i64 [ %1201, %cleanup26.i.i.i877 ], [ %retval.sroa.2.0.i.i.i850, %1176 ]
  %1208 = phi i32 [ %1202, %cleanup26.i.i.i877 ], [ %retval.sroa.6.0.i.i.i848, %1176 ]
  %1209 = phi i1 [ %cmp.i46.not.i.i.i869, %cleanup26.i.i.i877 ], [ true, %1176 ]
  br i1 %1209, label %for.inc.i.i.i882, label %Flow3711, !amdgpu.uniform !42

cleanup26.i.i.i877:                               ; preds = %Flow3709
  br label %Flow3710, !amdgpu.uniform !42

for.inc.i.i.i882:                                 ; preds = %Flow3710
  %inc.i.i.i887 = add i32 %spec.store.select.i.i.i854, 1
  br label %Flow3711, !amdgpu.uniform !42

Flow3711:                                         ; preds = %for.inc.i.i.i882, %Flow3710
  %1210 = phi i32 [ %inc.i.i.i887, %for.inc.i.i.i882 ], [ poison, %Flow3710 ]
  %1211 = phi i1 [ false, %for.inc.i.i.i882 ], [ true, %Flow3710 ]
  br i1 %1211, label %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i888, label %for.cond.i.i1.i847, !amdgpu.uniform !42

_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i888: ; preds = %Flow3711
  %.lcssa4124 = phi i32 [ %1203, %Flow3711 ]
  %.lcssa4123 = phi i64 [ %1204, %Flow3711 ]
  %.lcssa4122 = phi i32 [ %1205, %Flow3711 ]
  fence acquire
  %1212 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %mul.i.i.i.i.i.i889 = shl i32 %.lcssa4124, 6
  %idxprom.i19.i.i.i.i.i890 = zext i32 %mul.i.i.i.i.i.i889 to i64
  %arrayidx.i20.i.i.i.i.i891 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %1212, i64 %idxprom.i19.i.i.i.i.i890
  %arrayidx.i22.i.i.i.i.i892 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %arrayidx.i20.i.i.i.i.i891, i64 %sh_prom.i.i.i.i
  store i64 sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.35 to ptr), i64 10) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.35 to ptr) to i64)), ptr %arrayidx.i22.i.i.i.i.i892, align 8, !tbaa !82
  %spec.select.i.i.i.i.i.i.i893 = tail call i64 @llvm.umin.i64(i64 sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.35 to ptr), i64 10) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.35 to ptr) to i64)), i64 56)
  %1213 = and i64 %spec.select.i.i.i.i.i.i.i893, 15
  %1214 = sub i64 %spec.select.i.i.i.i.i.i.i893, %1213
  %1215 = icmp ne i64 %1214, 0
  br i1 %1215, label %loop-memcpy-expansion2369.preheader, label %Flow3707, !amdgpu.uniform !42

loop-memcpy-expansion2369.preheader:              ; preds = %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i888
  %1216 = add nuw nsw i64 %75, 8
  %scevgep3061 = getelementptr i8, ptr %1212, i64 %1216
  %1217 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i890, 6
  %scevgep3062 = getelementptr i8, ptr %scevgep3061, i64 %1217
  br label %loop-memcpy-expansion2369, !amdgpu.uniform !42

Flow3707:                                         ; preds = %Flow3706, %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i888
  br label %loop-memcpy-residual-header2372, !amdgpu.uniform !42

loop-memcpy-expansion2369:                        ; preds = %loop-memcpy-expansion2369.preheader, %loop-memcpy-expansion2369
  %loop-index2370 = phi i64 [ %1220, %loop-memcpy-expansion2369 ], [ 0, %loop-memcpy-expansion2369.preheader ]
  %1218 = getelementptr i8, ptr addrspace(4) @.str.35, i64 %loop-index2370, !amdgpu.uniform !42
  %1219 = load <4 x i32>, ptr addrspace(4) %1218, align 1
  %scevgep3063 = getelementptr i8, ptr %scevgep3062, i64 %loop-index2370
  store <4 x i32> %1219, ptr %scevgep3063, align 1
  %1220 = add i64 %loop-index2370, 16
  %1221 = icmp uge i64 %1220, %1214
  br i1 %1221, label %Flow3706, label %loop-memcpy-expansion2369, !amdgpu.uniform !42

Flow3705:                                         ; preds = %Flow3703, %Flow3704
  %1222 = phi i64 [ %2930, %Flow3703 ], [ %2927, %Flow3704 ]
  br label %post-loop-memcpy-expansion2368, !amdgpu.uniform !42

loop-memcpy-residual2371:                         ; preds = %loop-memcpy-residual2371.preheader, %loop-memcpy-residual2371
  %residual-loop-index2373 = phi i64 [ %1224, %loop-memcpy-residual2371 ], [ 0, %loop-memcpy-residual2371.preheader ]
  %scevgep3060 = getelementptr i8, ptr addrspace(4) @.str.35, i64 %residual-loop-index2373, !amdgpu.uniform !42
  %1223 = load i8, ptr addrspace(4) %scevgep3060, align 1
  %scevgep3059 = getelementptr i8, ptr %scevgep3058, i64 %residual-loop-index2373
  store i8 %1223, ptr %scevgep3059, align 1
  %1224 = add i64 %residual-loop-index2373, 1
  %1225 = icmp uge i64 %1224, %1213
  br i1 %1225, label %Flow3703, label %loop-memcpy-residual2371, !amdgpu.uniform !42

Flow3703:                                         ; preds = %loop-memcpy-residual2371
  br label %Flow3705, !amdgpu.uniform !42

post-loop-memcpy-expansion2368:                   ; preds = %Flow3705
  %tobool.not.i.i.i.i.i.i895 = icmp eq i32 %.lcssa4122, 0
  %conv.i.i.i.i.i.i896 = zext i1 %tobool.not.i.i.i.i.i.i895 to i32
  fence release
  %1226 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %idxprom.i23.i.i.i.i.i897 = zext i32 %.lcssa4124 to i64
  %arrayidx.i24.i.i.i.i.i898 = getelementptr inbounds nuw i32, ptr %1226, i64 %idxprom.i23.i.i.i.i.i897
  store atomic i32 %conv.i.i.i.i.i.i896, ptr %arrayidx.i24.i.i.i.i.i898 monotonic, align 4
  %1227 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !84
  %arrayidx.i.i5.i.i899 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %1227, i64 %idxprom.i23.i.i.i.i.i897, !amdgpu.uniform !42
  %1228 = load i64, ptr %arrayidx.i.i5.i.i899, align 8, !tbaa !80
  br label %while.cond.i.i.i.i900, !amdgpu.uniform !42

while.cond.i.i.i.i900:                            ; preds = %post-loop-memcpy-expansion2368, %Flow3701
  %phi.broken3940 = phi i64 [ 0, %post-loop-memcpy-expansion2368 ], [ %1245, %Flow3701 ]
  %lsr.iv3050 = phi i64 [ %1243, %Flow3701 ], [ -46, %post-loop-memcpy-expansion2368 ]
  %lsr.iv3045 = phi ptr addrspace(4) [ %1242, %Flow3701 ], [ getelementptr (i8, ptr addrspace(4) @.str.35, i64 56), %post-loop-memcpy-expansion2368 ]
  %port.sroa.37.0.i.i901 = phi i32 [ %1241, %Flow3701 ], [ %conv.i.i.i.i.i.i896, %post-loop-memcpy-expansion2368 ]
  %idx.0.i.i.i.i902 = phi i64 [ %1240, %Flow3701 ], [ 56, %post-loop-memcpy-expansion2368 ]
  %cmp.i.i.i.i903 = icmp ult i64 %idx.0.i.i.i.i902, sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.35 to ptr), i64 10) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.35 to ptr) to i64))
  %1229 = tail call i64 @llvm.amdgcn.ballot.i64(i1 %cmp.i.i.i.i903)
  %and.i.i.i.i6.i.i904 = and i64 %1229, %1228
  %tobool.not.i.i.i.i905 = icmp ne i64 %and.i.i.i.i6.i.i904, 0
  %1230 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i22.i.i.i906 = getelementptr inbounds nuw i32, ptr %1230, i64 %idxprom.i23.i.i.i.i.i897, !amdgpu.uniform !42
  %1231 = load atomic i32, ptr %arrayidx.i22.i.i.i906 monotonic, align 4
  %1232 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %tobool.not.i.i.i.i905)
  %1233 = extractvalue { i1, i64 } %1232, 0
  %1234 = extractvalue { i1, i64 } %1232, 1
  br i1 %1233, label %cond.false.i8.i.i.i.i907, label %Flow3701

cond.false.i8.i.i.i.i907:                         ; preds = %while.cond.i.i.i.i900
  %1235 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %1231)
  %1236 = icmp ne i32 %1235, %port.sroa.37.0.i.i901
  br i1 %1236, label %while.body.i.i21.i.i.i.i908.preheader, label %Flow3700, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i908.preheader:            ; preds = %cond.false.i8.i.i.i.i907
  br label %while.body.i.i21.i.i.i.i908, !amdgpu.uniform !42

Flow3700:                                         ; preds = %Flow3698, %cond.false.i8.i.i.i.i907
  br label %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i911, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i908:                      ; preds = %while.body.i.i21.i.i.i.i908.preheader, %while.body.i.i21.i.i.i.i908
  tail call void @llvm.amdgcn.s.sleep(i32 2)
  %1237 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i22.i.i.i.i909 = getelementptr inbounds nuw i32, ptr %1237, i64 %idxprom.i23.i.i.i.i.i897, !amdgpu.uniform !42
  %1238 = load atomic i32, ptr %arrayidx.i.i.i22.i.i.i.i909 monotonic, align 4
  %1239 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %1238)
  %cmp.i.not.i.i23.i.i.i.i910 = icmp eq i32 %1239, %port.sroa.37.0.i.i901
  br i1 %cmp.i.not.i.i23.i.i.i.i910, label %Flow3698, label %while.body.i.i21.i.i.i.i908, !amdgpu.uniform !42

Flow3698:                                         ; preds = %while.body.i.i21.i.i.i.i908
  br label %Flow3700, !amdgpu.uniform !42

Flow3701:                                         ; preds = %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i912, %while.cond.i.i.i.i900
  %1240 = phi i64 [ %add.i.i.i.i915, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i912 ], [ poison, %while.cond.i.i.i.i900 ]
  %1241 = phi i32 [ %conv.i.i26.i.i.i.i913, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i912 ], [ poison, %while.cond.i.i.i.i900 ]
  %1242 = phi ptr addrspace(4) [ %scevgep3046, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i912 ], [ poison, %while.cond.i.i.i.i900 ]
  %1243 = phi i64 [ %lsr.iv.next3051, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i912 ], [ poison, %while.cond.i.i.i.i900 ]
  %1244 = phi i1 [ false, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i912 ], [ true, %while.cond.i.i.i.i900 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %1234)
  %1245 = call i64 @llvm.amdgcn.if.break.i64(i1 %1244, i64 %phi.broken3940)
  %1246 = call i1 @llvm.amdgcn.loop.i64(i64 %1245)
  br i1 %1246, label %cond.false.i.i.i922, label %while.cond.i.i.i.i900

_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i911: ; preds = %Flow3700
  fence acquire
  br i1 %cmp.i.i.i.i903, label %if.then.i.i.i.i.i.i.i916, label %Flow3697, !amdgpu.uniform !42

if.then.i.i.i.i.i.i.i916:                         ; preds = %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i911
  %sub.i.i.i.i.i.i.i918 = sub nuw i64 sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.35 to ptr), i64 10) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.35 to ptr) to i64)), %idx.0.i.i.i.i902
  %spec.select.i.i.i36.i.i.i.i919 = tail call i64 @llvm.umin.i64(i64 %sub.i.i.i.i.i.i.i918, i64 64)
  %1247 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %1248 = and i64 %spec.select.i.i.i36.i.i.i.i919, 15
  %1249 = sub i64 %spec.select.i.i.i36.i.i.i.i919, %1248
  %1250 = icmp ne i64 %1249, 0
  br i1 %1250, label %loop-memcpy-expansion2381.preheader, label %Flow3696, !amdgpu.uniform !42

loop-memcpy-expansion2381.preheader:              ; preds = %if.then.i.i.i.i.i.i.i916
  %arrayidx.i20.i35.i.i.i.i920 = getelementptr %"struct.rpc::Buffer", ptr %1247, i64 %idxprom.i19.i.i.i.i.i890
  %arrayidx.i22.i38.i.i.i.i921 = getelementptr %"struct.rpc::Buffer", ptr %arrayidx.i20.i35.i.i.i.i920, i64 %sh_prom.i.i.i.i
  br label %loop-memcpy-expansion2381, !amdgpu.uniform !42

Flow3696:                                         ; preds = %Flow3695, %if.then.i.i.i.i.i.i.i916
  br label %loop-memcpy-residual-header2384, !amdgpu.uniform !42

loop-memcpy-expansion2381:                        ; preds = %loop-memcpy-expansion2381.preheader, %loop-memcpy-expansion2381
  %loop-index2382 = phi i64 [ %1253, %loop-memcpy-expansion2381 ], [ 0, %loop-memcpy-expansion2381.preheader ]
  %scevgep3047 = getelementptr i8, ptr addrspace(4) %lsr.iv3045, i64 %loop-index2382, !amdgpu.uniform !42
  %1251 = load <4 x i32>, ptr addrspace(4) %scevgep3047, align 1
  %1252 = getelementptr i8, ptr %arrayidx.i22.i38.i.i.i.i921, i64 %loop-index2382
  store <4 x i32> %1251, ptr %1252, align 1
  %1253 = add i64 %loop-index2382, 16
  %1254 = icmp uge i64 %1253, %1249
  br i1 %1254, label %Flow3695, label %loop-memcpy-expansion2381, !amdgpu.uniform !42

Flow3693:                                         ; preds = %Flow3692, %loop-memcpy-residual-header2384
  br label %Flow3697, !amdgpu.uniform !42

loop-memcpy-residual2383:                         ; preds = %loop-memcpy-residual2383.preheader, %loop-memcpy-residual2383
  %residual-loop-index2385 = phi i64 [ %1256, %loop-memcpy-residual2383 ], [ 0, %loop-memcpy-residual2383.preheader ]
  %scevgep3056 = getelementptr i8, ptr addrspace(4) %scevgep3055, i64 %residual-loop-index2385, !amdgpu.uniform !42
  %1255 = load i8, ptr addrspace(4) %scevgep3056, align 1
  %scevgep3054 = getelementptr i8, ptr %scevgep3053, i64 %residual-loop-index2385
  store i8 %1255, ptr %scevgep3054, align 1
  %1256 = add i64 %residual-loop-index2385, 1
  %1257 = icmp uge i64 %1256, %1248
  br i1 %1257, label %Flow3692, label %loop-memcpy-residual2383, !amdgpu.uniform !42

Flow3692:                                         ; preds = %loop-memcpy-residual2383
  br label %Flow3693, !amdgpu.uniform !42

_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i912: ; preds = %Flow3697
  %conv.i.i26.i.i.i.i913 = xor i32 %port.sroa.37.0.i.i901, 1
  fence release
  %1258 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %arrayidx.i24.i29.i.i.i.i914 = getelementptr inbounds nuw i32, ptr %1258, i64 %idxprom.i23.i.i.i.i.i897
  store atomic i32 %conv.i.i26.i.i.i.i913, ptr %arrayidx.i24.i29.i.i.i.i914 monotonic, align 4
  %add.i.i.i.i915 = add i64 %idx.0.i.i.i.i902, 64
  %scevgep3046 = getelementptr i8, ptr addrspace(4) %lsr.iv3045, i64 64
  %lsr.iv.next3051 = add i64 %lsr.iv3050, -64
  br label %Flow3701, !amdgpu.uniform !42

cond.false.i.i.i922:                              ; preds = %Flow3701
  %.lcssa4118 = phi i64 [ %1245, %Flow3701 ]
  %port.sroa.37.0.i.i901.lcssa = phi i32 [ %port.sroa.37.0.i.i901, %Flow3701 ]
  %.lcssa4117 = phi i32 [ %1231, %Flow3701 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa4118)
  %1259 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %.lcssa4117)
  %1260 = icmp ne i32 %1259, %port.sroa.37.0.i.i901.lcssa
  %1261 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %1260)
  %1262 = extractvalue { i1, i64 } %1261, 0
  %1263 = extractvalue { i1, i64 } %1261, 1
  br i1 %1262, label %while.body.i.i.i.i923.preheader, label %Flow3690

while.body.i.i.i.i923.preheader:                  ; preds = %cond.false.i.i.i922
  br label %while.body.i.i.i.i923, !amdgpu.uniform !42

Flow3690:                                         ; preds = %Flow3688, %cond.false.i.i.i922
  call void @llvm.amdgcn.end.cf.i64(i64 %1263)
  br label %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit944, !amdgpu.uniform !42

while.body.i.i.i.i923:                            ; preds = %while.body.i.i.i.i923.preheader, %while.body.i.i.i.i923
  %phi.broken3941 = phi i64 [ 0, %while.body.i.i.i.i923.preheader ], [ %1267, %while.body.i.i.i.i923 ]
  tail call void @llvm.amdgcn.s.sleep(i32 2)
  %1264 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i8.i.i924 = getelementptr inbounds nuw i32, ptr %1264, i64 %idxprom.i23.i.i.i.i.i897, !amdgpu.uniform !42
  %1265 = load atomic i32, ptr %arrayidx.i.i.i8.i.i924 monotonic, align 4
  %1266 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %1265)
  %cmp.i.not.i.i.i.i925 = icmp eq i32 %1266, %port.sroa.37.0.i.i901.lcssa
  %1267 = call i64 @llvm.amdgcn.if.break.i64(i1 %cmp.i.not.i.i.i.i925, i64 %phi.broken3941)
  %1268 = call i1 @llvm.amdgcn.loop.i64(i64 %1267)
  br i1 %1268, label %Flow3688, label %while.body.i.i.i.i923

Flow3688:                                         ; preds = %while.body.i.i.i.i923
  %.lcssa4116 = phi i64 [ %1267, %while.body.i.i.i.i923 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa4116)
  br label %Flow3690, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit944: ; preds = %Flow3690
  fence acquire
  tail call void @llvm.amdgcn.wave.barrier()
  fence syncscope("agent") release
  %1269 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %.lcssa4123, i1 true)
  %iszero.i.i.i.i.i13.i.i927 = icmp ne i64 %.lcssa4123, 0
  %cmp2.i.i.i.i14.i.i928 = icmp eq i64 %1269, %sh_prom.i.i.i.i
  %cmp.i.i.i.i15.i.i929 = select i1 %iszero.i.i.i.i.i13.i.i927, i1 %cmp2.i.i.i.i14.i.i928, i1 false
  %1270 = and i32 %.lcssa4124, 31
  %conv4.i.i.i19.i.i933 = zext i1 %cmp.i.i.i.i15.i.i929 to i32
  %shl.i.i.i20.i.i934 = shl nuw i32 %conv4.i.i.i19.i.i933, %1270
  %xor.i.i.i21.i.i935 = xor i32 %shl.i.i.i20.i.i934, -1
  %1271 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2549, !amdgpu.uniform !42

1272:                                             ; preds = %ComputeEnd2550
  %div8.i.i.i16.i.i930 = lshr i32 %.lcssa4124, 5
  %idxprom.i.i.i17.i.i931 = zext nneg i32 %div8.i.i.i16.i.i930 to i64
  %1273 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i17.i.i931
  %sunkaddr3314 = getelementptr inbounds i8, ptr addrspace(1) %1273, i64 40
  %1274 = atomicrmw and ptr addrspace(1) %sunkaddr3314, i32 %.lcssa4115 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %1275, !amdgpu.uniform !42

1275:                                             ; preds = %1272, %ComputeEnd2550
  call void @llvm.amdgcn.end.cf.i64(i64 %3507)
  tail call void @llvm.amdgcn.wave.barrier()
  call void @llvm.lifetime.start.p5(ptr addrspace(5) %buffer.i.i.i) #34, !noalias !114
  %cmp7.i.i.i.i.i.i.i = icmp ne i32 %LHS, 0
  %1276 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp7.i.i.i.i.i.i.i)
  %1277 = extractvalue { i1, i64 } %1276, 0
  %1278 = extractvalue { i1, i64 } %1276, 1
  br i1 %1277, label %while.body.lr.ph.i.i.i.i.i.i.i, label %Flow3687

while.body.lr.ph.i.i.i.i.i.i.i:                   ; preds = %1275
  %retval.0.i.i.i.i.i.i.i = tail call noundef i32 @llvm.abs.i32(i32 %LHS, i1 false)
  br label %while.body.i.i.i.i.i.i.i, !amdgpu.uniform !42

Flow3687:                                         ; preds = %Flow3686, %1275
  %1279 = phi i1 [ %cmp.i.i.i.i.i.i.i.i.i.lcssa, %Flow3686 ], [ false, %1275 ]
  %1280 = phi i64 [ %writer.sroa.10.1.i.i.i.i.i.lcssa, %Flow3686 ], [ 0, %1275 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %1278)
  br label %_ZN22__llvm_libc_22_0_0_git15IntegerToStringIiNS_7details3FmtILh10ELb0ELb0ELb0ELm1EEEE13IntegerWriter25write_unsigned_number_decEjRNS1_22StringBufferWriterImplILb0EEE.exit.i.i.i.i.i.i, !amdgpu.uniform !42

while.body.i.i.i.i.i.i.i:                         ; preds = %while.body.lr.ph.i.i.i.i.i.i.i, %_ZN22__llvm_libc_22_0_0_git7details22StringBufferWriterImplILb0EE4pushEc.exit.i.i.i.i.i.i.i
  %phi.broken3942 = phi i64 [ 0, %while.body.lr.ph.i.i.i.i.i.i.i ], [ %1348, %_ZN22__llvm_libc_22_0_0_git7details22StringBufferWriterImplILb0EE4pushEc.exit.i.i.i.i.i.i.i ]
  %writer.sroa.10.0.i.i.i.i.i = phi i64 [ %writer.sroa.10.1.i.i.i.i.i, %_ZN22__llvm_libc_22_0_0_git7details22StringBufferWriterImplILb0EE4pushEc.exit.i.i.i.i.i.i.i ], [ 0, %while.body.lr.ph.i.i.i.i.i.i.i ]
  %value.addr.09.i.i.i.i.i.i.i = phi i32 [ %div.i.i.i.i.i.i.i.i, %_ZN22__llvm_libc_22_0_0_git7details22StringBufferWriterImplILb0EE4pushEc.exit.i.i.i.i.i.i.i ], [ %retval.0.i.i.i.i.i.i.i, %while.body.lr.ph.i.i.i.i.i.i.i ]
  br label %NodeBlock3382, !amdgpu.uniform !42

NodeBlock3382:                                    ; preds = %while.body.i.i.i.i.i.i.i
  %div.i.i.i.i.i.i.i.i = udiv i32 %value.addr.09.i.i.i.i.i.i.i, 10
  %.neg = mul nuw nsw i32 %div.i.i.i.i.i.i.i.i, 246
  %rem.i.i.i.i.i.i.i.i.decomposed = add nuw nsw i32 %.neg, %value.addr.09.i.i.i.i.i.i.i
  %conv.i.i.i.i.i.i.i.i = trunc i32 %rem.i.i.i.i.i.i.i.i.decomposed to i8
  %1281 = zext i8 %conv.i.i.i.i.i.i.i.i to i16
  %Pivot3383 = icmp sge i16 %1281, 5
  %1282 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %Pivot3383)
  %1283 = extractvalue { i1, i64 } %1282, 0
  %1284 = extractvalue { i1, i64 } %1282, 1
  br i1 %1283, label %NodeBlock3380, label %Flow3685

NodeBlock3380:                                    ; preds = %NodeBlock3382
  %Pivot3381 = icmp sge i16 %1281, 7
  %1285 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %Pivot3381)
  %1286 = extractvalue { i1, i64 } %1285, 0
  %1287 = extractvalue { i1, i64 } %1285, 1
  br i1 %1286, label %NodeBlock3378, label %Flow3676

NodeBlock3378:                                    ; preds = %NodeBlock3380
  %Pivot3379 = icmp sge i16 %1281, 8
  %1288 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %Pivot3379)
  %1289 = extractvalue { i1, i64 } %1288, 0
  %1290 = extractvalue { i1, i64 } %1288, 1
  br i1 %1289, label %NodeBlock3376, label %Flow3672

NodeBlock3376:                                    ; preds = %NodeBlock3378
  %Pivot3377 = icmp sge i16 %1281, 9
  %1291 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %Pivot3377)
  %1292 = extractvalue { i1, i64 } %1291, 0
  %1293 = extractvalue { i1, i64 } %1291, 1
  br i1 %1292, label %sw.bb9.i.i.i.i.i.i.i.i.i, label %Flow3670

Flow3676:                                         ; preds = %Flow3673, %NodeBlock3380
  %1294 = phi i8 [ %1338, %Flow3673 ], [ poison, %NodeBlock3380 ]
  %1295 = call { i1, i64 } @llvm.amdgcn.else.i64.i64(i64 %1287)
  %1296 = extractvalue { i1, i64 } %1295, 0
  %1297 = extractvalue { i1, i64 } %1295, 1
  br i1 %1296, label %NodeBlock3374, label %Flow3677

NodeBlock3374:                                    ; preds = %Flow3676
  %Pivot3375 = icmp sge i16 %1281, 6
  %1298 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %Pivot3375)
  %1299 = extractvalue { i1, i64 } %1298, 0
  %1300 = extractvalue { i1, i64 } %1298, 1
  br i1 %1299, label %sw.bb6.i.i.i.i.i.i.i.i.i, label %Flow3674

Flow3685:                                         ; preds = %Flow3677, %NodeBlock3382
  %1301 = phi i8 [ %1340, %Flow3677 ], [ poison, %NodeBlock3382 ]
  %1302 = call { i1, i64 } @llvm.amdgcn.else.i64.i64(i64 %1284)
  %1303 = extractvalue { i1, i64 } %1302, 0
  %1304 = extractvalue { i1, i64 } %1302, 1
  br i1 %1303, label %NodeBlock3372, label %if.then.i.i.i.i.i.i.i.i

NodeBlock3372:                                    ; preds = %Flow3685
  %Pivot3373 = icmp sge i16 %1281, 3
  %1305 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %Pivot3373)
  %1306 = extractvalue { i1, i64 } %1305, 0
  %1307 = extractvalue { i1, i64 } %1305, 1
  br i1 %1306, label %NodeBlock3370, label %Flow3683

NodeBlock3370:                                    ; preds = %NodeBlock3372
  %Pivot3371 = icmp sge i16 %1281, 4
  %1308 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %Pivot3371)
  %1309 = extractvalue { i1, i64 } %1308, 0
  %1310 = extractvalue { i1, i64 } %1308, 1
  br i1 %1309, label %sw.bb4.i.i.i.i.i.i.i.i.i, label %Flow3678

Flow3683:                                         ; preds = %Flow3679, %NodeBlock3372
  %1311 = phi i8 [ %1341, %Flow3679 ], [ poison, %NodeBlock3372 ]
  %1312 = call { i1, i64 } @llvm.amdgcn.else.i64.i64(i64 %1307)
  %1313 = extractvalue { i1, i64 } %1312, 0
  %1314 = extractvalue { i1, i64 } %1312, 1
  br i1 %1313, label %NodeBlock3368, label %Flow3684

NodeBlock3368:                                    ; preds = %Flow3683
  %Pivot3369 = icmp sge i16 %1281, 2
  %1315 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %Pivot3369)
  %1316 = extractvalue { i1, i64 } %1315, 0
  %1317 = extractvalue { i1, i64 } %1315, 1
  br i1 %1316, label %sw.bb2.i.i.i.i.i.i.i.i.i, label %Flow3681

Flow3681:                                         ; preds = %sw.bb2.i.i.i.i.i.i.i.i.i, %NodeBlock3368
  %1318 = call { i1, i64 } @llvm.amdgcn.else.i64.i64(i64 %1317)
  %1319 = extractvalue { i1, i64 } %1318, 0
  %1320 = extractvalue { i1, i64 } %1318, 1
  br i1 %1319, label %LeafBlock3366, label %Flow3682

LeafBlock3366:                                    ; preds = %Flow3681
  %SwitchLeaf3367 = icmp eq i16 %1281, 1
  %1321 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %SwitchLeaf3367)
  %1322 = extractvalue { i1, i64 } %1321, 0
  %1323 = extractvalue { i1, i64 } %1321, 1
  br i1 %1322, label %sw.bb1.i.i.i.i.i.i.i.i.i, label %Flow3680

sw.bb1.i.i.i.i.i.i.i.i.i:                         ; preds = %LeafBlock3366
  br label %Flow3680, !amdgpu.uniform !42

sw.bb2.i.i.i.i.i.i.i.i.i:                         ; preds = %NodeBlock3368
  br label %Flow3681, !amdgpu.uniform !42

Flow3678:                                         ; preds = %sw.bb4.i.i.i.i.i.i.i.i.i, %NodeBlock3370
  %1324 = call { i1, i64 } @llvm.amdgcn.else.i64.i64(i64 %1310)
  %1325 = extractvalue { i1, i64 } %1324, 0
  %1326 = extractvalue { i1, i64 } %1324, 1
  br i1 %1325, label %sw.bb3.i.i.i.i.i.i.i.i.i, label %Flow3679

sw.bb3.i.i.i.i.i.i.i.i.i:                         ; preds = %Flow3678
  br label %Flow3679, !amdgpu.uniform !42

sw.bb4.i.i.i.i.i.i.i.i.i:                         ; preds = %NodeBlock3370
  br label %Flow3678, !amdgpu.uniform !42

Flow3674:                                         ; preds = %sw.bb6.i.i.i.i.i.i.i.i.i, %NodeBlock3374
  %1327 = call { i1, i64 } @llvm.amdgcn.else.i64.i64(i64 %1300)
  %1328 = extractvalue { i1, i64 } %1327, 0
  %1329 = extractvalue { i1, i64 } %1327, 1
  br i1 %1328, label %sw.bb5.i.i.i.i.i.i.i.i.i, label %Flow3675

sw.bb5.i.i.i.i.i.i.i.i.i:                         ; preds = %Flow3674
  br label %Flow3675, !amdgpu.uniform !42

sw.bb6.i.i.i.i.i.i.i.i.i:                         ; preds = %NodeBlock3374
  br label %Flow3674, !amdgpu.uniform !42

Flow3672:                                         ; preds = %Flow3671, %NodeBlock3378
  %1330 = phi i8 [ %1337, %Flow3671 ], [ poison, %NodeBlock3378 ]
  %1331 = call { i1, i64 } @llvm.amdgcn.else.i64.i64(i64 %1290)
  %1332 = extractvalue { i1, i64 } %1331, 0
  %1333 = extractvalue { i1, i64 } %1331, 1
  br i1 %1332, label %sw.bb7.i.i.i.i.i.i.i.i.i, label %Flow3673

sw.bb7.i.i.i.i.i.i.i.i.i:                         ; preds = %Flow3672
  br label %Flow3673, !amdgpu.uniform !42

Flow3670:                                         ; preds = %sw.bb9.i.i.i.i.i.i.i.i.i, %NodeBlock3376
  %1334 = call { i1, i64 } @llvm.amdgcn.else.i64.i64(i64 %1293)
  %1335 = extractvalue { i1, i64 } %1334, 0
  %1336 = extractvalue { i1, i64 } %1334, 1
  br i1 %1335, label %sw.bb8.i.i.i.i.i.i.i.i.i, label %Flow3671

sw.bb8.i.i.i.i.i.i.i.i.i:                         ; preds = %Flow3670
  br label %Flow3671, !amdgpu.uniform !42

sw.bb9.i.i.i.i.i.i.i.i.i:                         ; preds = %NodeBlock3376
  br label %Flow3670, !amdgpu.uniform !42

Flow3671:                                         ; preds = %sw.bb8.i.i.i.i.i.i.i.i.i, %Flow3670
  %1337 = phi i8 [ 56, %sw.bb8.i.i.i.i.i.i.i.i.i ], [ 57, %Flow3670 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %1336)
  br label %Flow3672, !amdgpu.uniform !42

Flow3673:                                         ; preds = %sw.bb7.i.i.i.i.i.i.i.i.i, %Flow3672
  %1338 = phi i8 [ 55, %sw.bb7.i.i.i.i.i.i.i.i.i ], [ %1330, %Flow3672 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %1333)
  br label %Flow3676, !amdgpu.uniform !42

Flow3675:                                         ; preds = %sw.bb5.i.i.i.i.i.i.i.i.i, %Flow3674
  %1339 = phi i8 [ 53, %sw.bb5.i.i.i.i.i.i.i.i.i ], [ 54, %Flow3674 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %1329)
  br label %Flow3677, !amdgpu.uniform !42

Flow3677:                                         ; preds = %Flow3675, %Flow3676
  %1340 = phi i8 [ %1339, %Flow3675 ], [ %1294, %Flow3676 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %1297)
  br label %Flow3685, !amdgpu.uniform !42

Flow3679:                                         ; preds = %sw.bb3.i.i.i.i.i.i.i.i.i, %Flow3678
  %1341 = phi i8 [ 51, %sw.bb3.i.i.i.i.i.i.i.i.i ], [ 52, %Flow3678 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %1326)
  br label %Flow3683, !amdgpu.uniform !42

Flow3680:                                         ; preds = %sw.bb1.i.i.i.i.i.i.i.i.i, %LeafBlock3366
  %1342 = phi i8 [ 49, %sw.bb1.i.i.i.i.i.i.i.i.i ], [ 48, %LeafBlock3366 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %1323)
  br label %Flow3682, !amdgpu.uniform !42

Flow3682:                                         ; preds = %Flow3680, %Flow3681
  %1343 = phi i8 [ %1342, %Flow3680 ], [ 50, %Flow3681 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %1320)
  br label %Flow3684, !amdgpu.uniform !42

Flow3684:                                         ; preds = %Flow3682, %Flow3683
  %1344 = phi i8 [ %1343, %Flow3682 ], [ %1311, %Flow3683 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %1314)
  br label %if.then.i.i.i.i.i.i.i.i, !amdgpu.uniform !42

if.then.i.i.i.i.i.i.i.i:                          ; preds = %Flow3685, %Flow3684
  %retval.0.i.i.i.i.i.i.i.i.i = phi i8 [ %1301, %Flow3685 ], [ %1344, %Flow3684 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %1304)
  %cmp.i.i.i.i.i.i.i.i.i = icmp eq i64 %writer.sroa.10.0.i.i.i.i.i, 11
  %cmp.i.i.i.i.i.i.i.i.i.inv = xor i1 %cmp.i.i.i.i.i.i.i.i.i, true
  br i1 %cmp.i.i.i.i.i.i.i.i.i.inv, label %if.then3.i.i.i.i.i.i.i.i, label %_ZN22__llvm_libc_22_0_0_git7details22StringBufferWriterImplILb0EE4pushEc.exit.i.i.i.i.i.i.i, !amdgpu.uniform !42

if.then3.i.i.i.i.i.i.i.i:                         ; preds = %if.then.i.i.i.i.i.i.i.i
  %1345 = trunc i64 %writer.sroa.10.0.i.i.i.i.i to i32
  %1346 = xor i32 %1345, -1
  %1347 = getelementptr i8, ptr addrspace(5) %buffer.i.i.i, i32 %1346
  %arrayidx.i.i.i.i.i.i.i.i.i2719 = getelementptr i8, ptr addrspace(5) %1347, i32 11
  %inc.i.i.i.i.i.i.i.i = add i64 %writer.sroa.10.0.i.i.i.i.i, 1
  store i8 %retval.0.i.i.i.i.i.i.i.i.i, ptr addrspace(5) %arrayidx.i.i.i.i.i.i.i.i.i2719, align 1, !tbaa !57, !noalias !114
  br label %_ZN22__llvm_libc_22_0_0_git7details22StringBufferWriterImplILb0EE4pushEc.exit.i.i.i.i.i.i.i, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git7details22StringBufferWriterImplILb0EE4pushEc.exit.i.i.i.i.i.i.i: ; preds = %if.then3.i.i.i.i.i.i.i.i, %if.then.i.i.i.i.i.i.i.i
  %writer.sroa.10.1.i.i.i.i.i = phi i64 [ %inc.i.i.i.i.i.i.i.i, %if.then3.i.i.i.i.i.i.i.i ], [ 11, %if.then.i.i.i.i.i.i.i.i ]
  %cmp.i.i.i.i.i.i.i = icmp ult i32 %value.addr.09.i.i.i.i.i.i.i, 10
  %.not.i.i.i.i.i.i.i = or i1 %cmp.i.i.i.i.i.i.i.i.i, %cmp.i.i.i.i.i.i.i
  %1348 = call i64 @llvm.amdgcn.if.break.i64(i1 %.not.i.i.i.i.i.i.i, i64 %phi.broken3942)
  %1349 = call i1 @llvm.amdgcn.loop.i64(i64 %1348)
  br i1 %1349, label %Flow3686, label %while.body.i.i.i.i.i.i.i

Flow3686:                                         ; preds = %_ZN22__llvm_libc_22_0_0_git7details22StringBufferWriterImplILb0EE4pushEc.exit.i.i.i.i.i.i.i
  %writer.sroa.10.1.i.i.i.i.i.lcssa = phi i64 [ %writer.sroa.10.1.i.i.i.i.i, %_ZN22__llvm_libc_22_0_0_git7details22StringBufferWriterImplILb0EE4pushEc.exit.i.i.i.i.i.i.i ]
  %.lcssa4114 = phi i64 [ %1348, %_ZN22__llvm_libc_22_0_0_git7details22StringBufferWriterImplILb0EE4pushEc.exit.i.i.i.i.i.i.i ]
  %cmp.i.i.i.i.i.i.i.i.i.lcssa = phi i1 [ %cmp.i.i.i.i.i.i.i.i.i, %_ZN22__llvm_libc_22_0_0_git7details22StringBufferWriterImplILb0EE4pushEc.exit.i.i.i.i.i.i.i ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa4114)
  br label %Flow3687, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git15IntegerToStringIiNS_7details3FmtILh10ELb0ELb0ELb0ELm1EEEE13IntegerWriter25write_unsigned_number_decEjRNS1_22StringBufferWriterImplILb0EEE.exit.i.i.i.i.i.i: ; preds = %Flow3687
  %cmp.i19.i.i.i.i.i = icmp ne i64 %1280, 0
  %or.cond.not.i.i.i.i.i = select i1 %1279, i1 true, i1 %cmp.i19.i.i.i.i.i
  %or.cond.not.i.i.i.i.i.inv = xor i1 %or.cond.not.i.i.i.i.i, true
  %1350 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %or.cond.not.i.i.i.i.i.inv)
  %1351 = extractvalue { i1, i64 } %1350, 0
  %1352 = extractvalue { i1, i64 } %1350, 1
  br i1 %1351, label %if.then3.i.i.lr.ph.i.i.i.i.i, label %while.end.i.i.i.i.i.i

if.then3.i.i.lr.ph.i.i.i.i.i:                     ; preds = %_ZN22__llvm_libc_22_0_0_git15IntegerToStringIiNS_7details3FmtILh10ELb0ELb0ELb0ELm1EEEE13IntegerWriter25write_unsigned_number_decEjRNS1_22StringBufferWriterImplILb0EEE.exit.i.i.i.i.i.i
  %arrayidx.i.i.i.i.i.i.i.i = getelementptr inbounds nuw i8, ptr addrspace(5) %buffer.i.i.i, i32 10
  store i8 48, ptr addrspace(5) %arrayidx.i.i.i.i.i.i.i.i, align 2, !tbaa !57, !noalias !114
  br label %while.end.i.i.i.i.i.i, !amdgpu.uniform !42

while.end.i.i.i.i.i.i:                            ; preds = %if.then3.i.i.lr.ph.i.i.i.i.i, %_ZN22__llvm_libc_22_0_0_git15IntegerToStringIiNS_7details3FmtILh10ELb0ELb0ELb0ELm1EEEE13IntegerWriter25write_unsigned_number_decEjRNS1_22StringBufferWriterImplILb0EEE.exit.i.i.i.i.i.i
  %writer.sroa.10.4.i.i.i.i.i = phi i64 [ 1, %if.then3.i.i.lr.ph.i.i.i.i.i ], [ %1280, %_ZN22__llvm_libc_22_0_0_git15IntegerToStringIiNS_7details3FmtILh10ELb0ELb0ELb0ELm1EEEE13IntegerWriter25write_unsigned_number_decEjRNS1_22StringBufferWriterImplILb0EEE.exit.i.i.i.i.i.i ]
  call void @llvm.amdgcn.end.cf.i64(i64 %1352)
  %cmp4.i.i.i.i.i.i = icmp sgt i32 %LHS, -1
  %brmerge.i.i.i.i.i.i = or i1 %cmp4.i.i.i.i.i.i, %1279
  %brmerge.i.i.i.i.i.i.inv = xor i1 %brmerge.i.i.i.i.i.i, true
  %1353 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %brmerge.i.i.i.i.i.i.inv)
  %1354 = extractvalue { i1, i64 } %1353, 0
  %1355 = extractvalue { i1, i64 } %1353, 1
  br i1 %1354, label %if.then.i14.i.i.i.i.i.i, label %_ZN22__llvm_libc_22_0_0_git15IntegerToStringIiNS_7details3FmtILh10ELb0ELb0ELb0ELm1EEEEC2Ei.exit.i.i.i.i

if.then.i14.i.i.i.i.i.i:                          ; preds = %while.end.i.i.i.i.i.i
  %cmp.i.i17.i.i.i.i.i.i = icmp ne i64 %writer.sroa.10.4.i.i.i.i.i, 11
  %1356 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.i.i17.i.i.i.i.i.i)
  %1357 = extractvalue { i1, i64 } %1356, 0
  %1358 = extractvalue { i1, i64 } %1356, 1
  br i1 %1357, label %if.then3.i18.i.i.i.i.i.i, label %Flow3669

if.then3.i18.i.i.i.i.i.i:                         ; preds = %if.then.i14.i.i.i.i.i.i
  %1359 = trunc i64 %writer.sroa.10.4.i.i.i.i.i to i32
  %1360 = xor i32 %1359, -1
  %1361 = getelementptr i8, ptr addrspace(5) %buffer.i.i.i, i32 %1360
  %arrayidx.i.i19.i.i.i.i.i.i2720 = getelementptr i8, ptr addrspace(5) %1361, i32 11
  %inc.i20.i.i.i.i.i.i = add i64 %writer.sroa.10.4.i.i.i.i.i, 1
  store i8 45, ptr addrspace(5) %arrayidx.i.i19.i.i.i.i.i.i2720, align 1, !tbaa !57, !noalias !114
  br label %Flow3669, !amdgpu.uniform !42

Flow3669:                                         ; preds = %if.then3.i18.i.i.i.i.i.i, %if.then.i14.i.i.i.i.i.i
  %1362 = phi i64 [ %inc.i20.i.i.i.i.i.i, %if.then3.i18.i.i.i.i.i.i ], [ 11, %if.then.i14.i.i.i.i.i.i ]
  call void @llvm.amdgcn.end.cf.i64(i64 %1358)
  br label %_ZN22__llvm_libc_22_0_0_git15IntegerToStringIiNS_7details3FmtILh10ELb0ELb0ELb0ELm1EEEEC2Ei.exit.i.i.i.i, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git15IntegerToStringIiNS_7details3FmtILh10ELb0ELb0ELb0ELm1EEEEC2Ei.exit.i.i.i.i: ; preds = %while.end.i.i.i.i.i.i, %Flow3669
  %writer.sroa.10.5.i.i.i.i.i = phi i64 [ %writer.sroa.10.4.i.i.i.i.i, %while.end.i.i.i.i.i.i ], [ %1362, %Flow3669 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %1355)
  %written.i.i.i.i.i = getelementptr inbounds nuw i8, ptr addrspace(5) %buffer.i.i.i, i32 16
  store i64 %writer.sroa.10.5.i.i.i.i.i, ptr addrspace(5) %written.i.i.i.i.i, align 8, !tbaa !91, !noalias !114
  %cmp.i.not.i.i.i.i.i.i = icmp ne i64 %writer.sroa.10.5.i.i.i.i.i, 0
  %1363 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.i.not.i.i.i.i.i.i)
  %1364 = extractvalue { i1, i64 } %1363, 0
  %1365 = extractvalue { i1, i64 } %1363, 1
  br i1 %1364, label %if.then.i.i.i.i.i.i.i, label %Flow3668

if.then.i.i.i.i.i.i.i:                            ; preds = %_ZN22__llvm_libc_22_0_0_git15IntegerToStringIiNS_7details3FmtILh10ELb0ELb0ELb0ELm1EEEEC2Ei.exit.i.i.i.i
  %1366 = call { i64, i1 } @llvm.uadd.with.overflow.i64(i64 %writer.sroa.10.5.i.i.i.i.i, i64 1)
  %ov3316 = extractvalue { i64, i1 } %1366, 1
  %ov3316.inv = xor i1 %ov3316, true
  %1367 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %ov3316.inv)
  %1368 = extractvalue { i1, i64 } %1367, 0
  %1369 = extractvalue { i1, i64 } %1367, 1
  br i1 %1368, label %if.end.i.i.i.i.i.i.i.i, label %Flow3667

if.end.i.i.i.i.i.i.i.i:                           ; preds = %if.then.i.i.i.i.i.i.i
  %math3315 = extractvalue { i64, i1 } %1366, 0
  %cmp2.i.i.i.i.i.i.i.i = icmp ult i64 %math3315, 1676976733973595601
  %1370 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp2.i.i.i.i.i.i.i.i)
  %1371 = extractvalue { i1, i64 } %1370, 0
  %1372 = extractvalue { i1, i64 } %1370, 1
  br i1 %1371, label %if.then3.i.i.i.i4.i.i.i.i, label %if.end4.i.i.i.i.i.i.i.i

if.then3.i.i.i.i4.i.i.i.i:                        ; preds = %if.end.i.i.i.i.i.i.i.i
  %mul.i.i.i.i.i.i.i.i = mul nuw i64 %math3315, 11
  %div18.i.i.i.i.i.i.i.i = lshr i64 %mul.i.i.i.i.i.i.i.i, 3
  br label %if.end4.i.i.i.i.i.i.i.i, !amdgpu.uniform !42

Flow3667:                                         ; preds = %if.end4.i.i.i.i.i.i.i.i, %if.then.i.i.i.i.i.i.i
  %1373 = phi ptr [ %cond.i.i951, %if.end4.i.i.i.i.i.i.i.i ], [ addrspacecast (ptr addrspace(4) @_ZN22__llvm_libc_22_0_0_git3cpp6string14NULL_CHARACTERE to ptr), %if.then.i.i.i.i.i.i.i ]
  call void @llvm.amdgcn.end.cf.i64(i64 %1369)
  br label %_ZN22__llvm_libc_22_0_0_git3cpp6string7reserveEm.exit.i.i.i.i.i.i.i, !amdgpu.uniform !42

if.end4.i.i.i.i.i.i.i.i:                          ; preds = %if.then3.i.i.i.i4.i.i.i.i, %if.end.i.i.i.i.i.i.i.i
  %new_capacity.addr.0.i.i.i.i.i.i.i.i = phi i64 [ %div18.i.i.i.i.i.i.i.i, %if.then3.i.i.i.i4.i.i.i.i ], [ %math3315, %if.end.i.i.i.i.i.i.i.i ]
  call void @llvm.amdgcn.end.cf.i64(i64 %1372)
  %sub.i.i945 = add i64 %new_capacity.addr.0.i.i.i.i.i.i.i.i, 7
  %div2.i.i946 = and i64 %sub.i.i945, -8
  %1374 = load ptr, ptr addrspace(1) @_ZN12_GLOBAL__N_13ptrE, align 8, !tbaa !55
  %add.ptr.i.i947 = getelementptr inbounds nuw i8, ptr %1374, i64 %div2.i.i946
  store ptr %add.ptr.i.i947, ptr addrspace(1) @_ZN12_GLOBAL__N_13ptrE, align 8, !tbaa !55
  %sub.ptr.lhs.cast.i.i948 = ptrtoint ptr %add.ptr.i.i947 to i64
  %sub.ptr.sub.i.i949 = sub i64 %sub.ptr.lhs.cast.i.i948, ptrtoint (ptr addrspacecast (ptr addrspace(1) @_ZN12_GLOBAL__N_16memoryE to ptr) to i64)
  %cmp.i.i950 = icmp ugt i64 %sub.ptr.sub.i.i949, 65335
  %cond.i.i951 = select i1 %cmp.i.i950, ptr null, ptr %1374
  br label %Flow3667, !amdgpu.uniform !42

Flow3668:                                         ; preds = %Flow3666, %_ZN22__llvm_libc_22_0_0_git15IntegerToStringIiNS_7details3FmtILh10ELb0ELb0ELb0ELm1EEEEC2Ei.exit.i.i.i.i
  %1375 = phi ptr [ %1373, %Flow3666 ], [ addrspacecast (ptr addrspace(4) @_ZN22__llvm_libc_22_0_0_git3cpp6string14NULL_CHARACTERE to ptr), %_ZN22__llvm_libc_22_0_0_git15IntegerToStringIiNS_7details3FmtILh10ELb0ELb0ELb0ELm1EEEEC2Ei.exit.i.i.i.i ]
  call void @llvm.amdgcn.end.cf.i64(i64 %1365)
  br label %if.end.i.i.i.i.i.i.i, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git3cpp6string7reserveEm.exit.i.i.i.i.i.i.i: ; preds = %Flow3667
  br label %loadstoreloop2189, !amdgpu.uniform !42

loadstoreloop2189:                                ; preds = %_ZN22__llvm_libc_22_0_0_git3cpp6string7reserveEm.exit.i.i.i.i.i.i.i, %loadstoreloop2189
  %phi.broken3943 = phi i64 [ 0, %_ZN22__llvm_libc_22_0_0_git3cpp6string7reserveEm.exit.i.i.i.i.i.i.i ], [ %1380, %loadstoreloop2189 ]
  %1376 = phi i64 [ %1378, %loadstoreloop2189 ], [ 0, %_ZN22__llvm_libc_22_0_0_git3cpp6string7reserveEm.exit.i.i.i.i.i.i.i ]
  %1377 = getelementptr i8, ptr %1373, i64 %1376
  store i8 0, ptr %1377, align 1
  %1378 = add i64 %1376, 1
  %1379 = icmp uge i64 %1378, %writer.sroa.10.5.i.i.i.i.i
  %1380 = call i64 @llvm.amdgcn.if.break.i64(i1 %1379, i64 %phi.broken3943)
  %1381 = call i1 @llvm.amdgcn.loop.i64(i64 %1380)
  br i1 %1381, label %Flow3666, label %loadstoreloop2189

Flow3666:                                         ; preds = %loadstoreloop2189
  %.lcssa4113 = phi i64 [ %1380, %loadstoreloop2189 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa4113)
  br label %Flow3668, !amdgpu.uniform !42

if.end.i.i.i.i.i.i.i:                             ; preds = %Flow3668
  %cmp.not.i6.i.i.i.i.i.i.i = icmp ne ptr %1375, addrspacecast (ptr addrspace(4) @_ZN22__llvm_libc_22_0_0_git3cpp6string14NULL_CHARACTERE to ptr)
  %1382 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.not.i6.i.i.i.i.i.i.i)
  %1383 = extractvalue { i1, i64 } %1382, 0
  %1384 = extractvalue { i1, i64 } %1382, 1
  br i1 %1383, label %if.then.i.i.i.i2.i.i.i.i, label %_ZN22__llvm_libc_22_0_0_git7testing8internal13describeValueIiEENS_3cpp9enable_ifIXaasr3cppE13is_integral_vIT_ElestS5_Lm8EENS3_6stringEE4typeES5_.exit.i

if.then.i.i.i.i2.i.i.i.i:                         ; preds = %if.end.i.i.i.i.i.i.i
  %arrayidx.i.i.i.i3.i.i.i.i = getelementptr inbounds nuw i8, ptr %1375, i64 %writer.sroa.10.5.i.i.i.i.i
  store i8 0, ptr %arrayidx.i.i.i.i3.i.i.i.i, align 1, !tbaa !57, !noalias !114
  br label %_ZN22__llvm_libc_22_0_0_git7testing8internal13describeValueIiEENS_3cpp9enable_ifIXaasr3cppE13is_integral_vIT_ElestS5_Lm8EENS3_6stringEE4typeES5_.exit.i, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git7testing8internal13describeValueIiEENS_3cpp9enable_ifIXaasr3cppE13is_integral_vIT_ElestS5_Lm8EENS3_6stringEE4typeES5_.exit.i: ; preds = %if.then.i.i.i.i2.i.i.i.i, %if.end.i.i.i.i.i.i.i
  call void @llvm.amdgcn.end.cf.i64(i64 %1384)
  %1385 = trunc i64 %writer.sroa.10.5.i.i.i.i.i to i32
  %1386 = and i64 %writer.sroa.10.5.i.i.i.i.i, 15
  %1387 = sub i64 %writer.sroa.10.5.i.i.i.i.i, %1386
  %1388 = icmp ne i64 %1387, 0
  %1389 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %1388)
  %1390 = extractvalue { i1, i64 } %1389, 0
  %1391 = extractvalue { i1, i64 } %1389, 1
  br i1 %1390, label %loop-memcpy-expansion2315.preheader, label %Flow3664

loop-memcpy-expansion2315.preheader:              ; preds = %_ZN22__llvm_libc_22_0_0_git7testing8internal13describeValueIiEENS_3cpp9enable_ifIXaasr3cppE13is_integral_vIT_ElestS5_Lm8EENS3_6stringEE4typeES5_.exit.i
  %1392 = sub i32 11, %1385
  %scevgep3042 = getelementptr i8, ptr addrspace(5) %buffer.i.i.i, i32 %1392
  br label %loop-memcpy-expansion2315, !amdgpu.uniform !42

Flow3664:                                         ; preds = %Flow3663, %_ZN22__llvm_libc_22_0_0_git7testing8internal13describeValueIiEENS_3cpp9enable_ifIXaasr3cppE13is_integral_vIT_ElestS5_Lm8EENS3_6stringEE4typeES5_.exit.i
  call void @llvm.amdgcn.end.cf.i64(i64 %1391)
  br label %loop-memcpy-residual-header2318, !amdgpu.uniform !42

loop-memcpy-expansion2315:                        ; preds = %loop-memcpy-expansion2315.preheader, %loop-memcpy-expansion2315
  %phi.broken3944 = phi i64 [ 0, %loop-memcpy-expansion2315.preheader ], [ %1397, %loop-memcpy-expansion2315 ]
  %lsr.iv3043 = phi ptr addrspace(5) [ %scevgep3044, %loop-memcpy-expansion2315 ], [ %scevgep3042, %loop-memcpy-expansion2315.preheader ]
  %loop-index2316 = phi i64 [ %1395, %loop-memcpy-expansion2315 ], [ 0, %loop-memcpy-expansion2315.preheader ]
  %1393 = load <4 x i32>, ptr addrspace(5) %lsr.iv3043, align 1
  %1394 = getelementptr i8, ptr %1375, i64 %loop-index2316
  store <4 x i32> %1393, ptr %1394, align 1
  %1395 = add i64 %loop-index2316, 16
  %scevgep3044 = getelementptr i8, ptr addrspace(5) %lsr.iv3043, i32 16
  %1396 = icmp uge i64 %1395, %1387
  %1397 = call i64 @llvm.amdgcn.if.break.i64(i1 %1396, i64 %phi.broken3944)
  %1398 = call i1 @llvm.amdgcn.loop.i64(i64 %1397)
  br i1 %1398, label %Flow3663, label %loop-memcpy-expansion2315

Flow3661:                                         ; preds = %Flow3660, %loop-memcpy-residual-header2318
  call void @llvm.amdgcn.end.cf.i64(i64 %2885)
  br label %post-loop-memcpy-expansion2314, !amdgpu.uniform !42

loop-memcpy-residual2317:                         ; preds = %loop-memcpy-residual2317.preheader, %loop-memcpy-residual2317
  %phi.broken3945 = phi i64 [ 0, %loop-memcpy-residual2317.preheader ], [ %1402, %loop-memcpy-residual2317 ]
  %lsr.iv3038 = phi ptr addrspace(5) [ %scevgep3039, %loop-memcpy-residual2317 ], [ %scevgep3037, %loop-memcpy-residual2317.preheader ]
  %residual-loop-index2319 = phi i64 [ %1400, %loop-memcpy-residual2317 ], [ 0, %loop-memcpy-residual2317.preheader ]
  %1399 = load i8, ptr addrspace(5) %lsr.iv3038, align 1
  %scevgep3041 = getelementptr i8, ptr %scevgep3040, i64 %residual-loop-index2319
  store i8 %1399, ptr %scevgep3041, align 1
  %1400 = add i64 %residual-loop-index2319, 1
  %scevgep3039 = getelementptr i8, ptr addrspace(5) %lsr.iv3038, i32 1
  %1401 = icmp uge i64 %1400, %1386
  %1402 = call i64 @llvm.amdgcn.if.break.i64(i1 %1401, i64 %phi.broken3945)
  %1403 = call i1 @llvm.amdgcn.loop.i64(i64 %1402)
  br i1 %1403, label %Flow3660, label %loop-memcpy-residual2317

Flow3660:                                         ; preds = %loop-memcpy-residual2317
  %.lcssa4111 = phi i64 [ %1402, %loop-memcpy-residual2317 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa4111)
  br label %Flow3661, !amdgpu.uniform !42

post-loop-memcpy-expansion2314:                   ; preds = %Flow3661
  call void @llvm.lifetime.end.p5(ptr addrspace(5) %buffer.i.i.i) #34, !noalias !114
  br label %for.cond.i.i.i953, !amdgpu.uniform !42

for.cond.i.i.i953:                                ; preds = %post-loop-memcpy-expansion2314, %Flow3658
  %1404 = phi i32 [ %1442, %Flow3658 ], [ poison, %post-loop-memcpy-expansion2314 ]
  %1405 = phi i64 [ %1443, %Flow3658 ], [ poison, %post-loop-memcpy-expansion2314 ]
  %1406 = phi i32 [ %1444, %Flow3658 ], [ poison, %post-loop-memcpy-expansion2314 ]
  %retval.sroa.6.0.i.i.i954 = phi i32 [ %1447, %Flow3658 ], [ undef, %post-loop-memcpy-expansion2314 ]
  %retval.sroa.2.0.i.i.i956 = phi i64 [ %1446, %Flow3658 ], [ undef, %post-loop-memcpy-expansion2314 ]
  %retval.sroa.8.0.i.i.i957 = phi i32 [ %1445, %Flow3658 ], [ undef, %post-loop-memcpy-expansion2314 ]
  %index.0.i.i.i958 = phi i32 [ %1449, %Flow3658 ], [ 0, %post-loop-memcpy-expansion2314 ]
  %1407 = load i32, ptr addrspace(1) @__llvm_rpc_client, align 8, !tbaa !68
  %cmp.not.i.i.i959 = icmp ult i32 %index.0.i.i.i958, %1407
  %spec.store.select.i.i.i960 = select i1 %cmp.not.i.i.i959, i32 %index.0.i.i.i958, i32 0
  %1408 = tail call noundef i64 @llvm.amdgcn.ballot.i64(i1 true)
  %1409 = and i32 %spec.store.select.i.i.i960, 31
  %1410 = lshr i64 %1408, %sh_prom.i.i.i.i
  %1411 = trunc i64 %1410 to i32
  %conv4.i.i.i.i.i964 = and i32 %1411, 1
  %shl.i.i.i.i.i965 = shl nuw i32 %conv4.i.i.i.i.i964, %1409
  %1412 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2553, !amdgpu.uniform !42

1413:                                             ; preds = %ComputeEnd2554
  %sunkaddr3317 = getelementptr inbounds i8, ptr addrspace(1) %3517, i64 40
  %1414 = atomicrmw or ptr addrspace(1) %sunkaddr3317, i32 %.lcssa4105 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %1415, !amdgpu.uniform !42

1415:                                             ; preds = %1413, %ComputeEnd2554
  %1416 = phi i32 [ %1414, %1413 ], [ poison, %ComputeEnd2554 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %3526)
  %1417 = call i32 @llvm.amdgcn.readfirstlane.i32(i32 %1416)
  %1418 = or i32 %1417, %.lcssa4106
  %shl5.i.i.i.i.i966 = shl nuw i32 1, %1409
  %and.i.i.i.i.i967 = and i32 %shl5.i.i.i.i.i966, %1418
  %tobool3.i.i.i.i968 = icmp ne i32 %and.i.i.i.i.i967, 0
  %1419 = tail call i64 @llvm.amdgcn.ballot.i64(i1 %tobool3.i.i.i.i968)
  %and.i.i.i.i.i.i969 = and i64 %1419, %1408
  %cmp.i.not.i.i.i970 = icmp ne i64 %1408, %and.i.i.i.i.i.i969
  br i1 %cmp.i.not.i.i.i970, label %if.then.i.i.i.i971, label %Flow3657, !amdgpu.uniform !42

if.then.i.i.i.i971:                               ; preds = %1415
  fence syncscope("agent") acquire
  %1420 = load <2 x i64>, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !75
  %1421 = extractelement <2 x i64> %1420, i32 0
  %1422 = inttoptr i64 %1421 to ptr
  %1423 = extractelement <2 x i64> %1420, i32 1
  %1424 = inttoptr i64 %1423 to ptr
  %idxprom.i.i.i.i972 = zext i32 %spec.store.select.i.i.i960 to i64
  %arrayidx.i.i.i.i973 = getelementptr inbounds nuw i32, ptr %1422, i64 %idxprom.i.i.i.i972, !amdgpu.uniform !42
  %1425 = load atomic i32, ptr %arrayidx.i.i.i.i973 monotonic, align 4
  %1426 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %1425)
  %arrayidx.i45.i.i.i974 = getelementptr inbounds nuw i32, ptr %1424, i64 %idxprom.i.i.i.i972, !amdgpu.uniform !42
  %1427 = load atomic i32, ptr %arrayidx.i45.i.i.i974 monotonic, align 4
  %1428 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %1427)
  %cmp.i46.not.i.i.i975 = icmp ne i32 %1426, %1428
  br i1 %cmp.i46.not.i.i.i975, label %if.then12.i.i.i976, label %Flow3655, !amdgpu.uniform !42

if.then12.i.i.i976:                               ; preds = %if.then.i.i.i.i971
  fence syncscope("agent") release
  %1429 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %1408, i1 true)
  %iszero.i.i.i.i.i.i.i977 = icmp ne i64 %1408, 0
  %cmp2.i.i.i.i.i.i978 = icmp eq i64 %1429, %sh_prom.i.i.i.i
  %cmp.i.i.i.i.i.i979 = select i1 %iszero.i.i.i.i.i.i.i977, i1 %cmp2.i.i.i.i.i.i978, i1 false
  %conv4.i.i51.i.i.i980 = zext i1 %cmp.i.i.i.i.i.i979 to i32
  %shl.i.i52.i.i.i981 = shl nuw i32 %conv4.i.i51.i.i.i980, %1409
  %xor.i.i.i.i.i982 = xor i32 %shl.i.i52.i.i.i981, -1
  %1430 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2558, !amdgpu.uniform !42

1431:                                             ; preds = %ComputeEnd2559
  %sunkaddr3318 = getelementptr inbounds i8, ptr addrspace(1) %3517, i64 40
  %1432 = atomicrmw and ptr addrspace(1) %sunkaddr3318, i32 %.lcssa4107 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %1433, !amdgpu.uniform !42

1433:                                             ; preds = %1431, %ComputeEnd2559
  call void @llvm.amdgcn.end.cf.i64(i64 %3544)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %Flow3655, !amdgpu.uniform !42

if.end14.i.i.i1042:                               ; preds = %Flow3655
  %1434 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %1408, i1 true)
  %iszero.i.i.i.i.i.i1043 = icmp ne i64 %1408, 0
  %cmp2.i.i.i.i.i1044 = icmp eq i64 %1434, %sh_prom.i.i.i.i
  %cmp.i.i.i.i.i1045 = select i1 %iszero.i.i.i.i.i.i1043, i1 %cmp2.i.i.i.i.i1044, i1 false
  %1435 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.i.i.i.i.i1045)
  %1436 = extractvalue { i1, i64 } %1435, 0
  %1437 = extractvalue { i1, i64 } %1435, 1
  br i1 %1436, label %if.then16.i.i.i1047, label %if.end22.i.i.i1046

if.then16.i.i.i1047:                              ; preds = %if.end14.i.i.i1042
  %1438 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !76
  %opcode.i.i.i1048 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %1438, i64 %idxprom.i.i.i.i972, i32 1
  store i32 1660944387, ptr %opcode.i.i.i1048, align 8, !tbaa !77
  %arrayidx21.i.i.i1049 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %1438, i64 %idxprom.i.i.i.i972
  store i64 %1408, ptr %arrayidx21.i.i.i1049, align 8, !tbaa !80
  br label %if.end22.i.i.i1046, !amdgpu.uniform !42

Flow3656:                                         ; preds = %if.end22.i.i.i1046, %Flow3655
  %1439 = phi i32 [ %1426, %if.end22.i.i.i1046 ], [ %retval.sroa.8.0.i.i.i957, %Flow3655 ]
  %1440 = phi i64 [ %1408, %if.end22.i.i.i1046 ], [ %retval.sroa.2.0.i.i.i956, %Flow3655 ]
  %1441 = phi i32 [ %spec.store.select.i.i.i960, %if.end22.i.i.i1046 ], [ %retval.sroa.6.0.i.i.i954, %Flow3655 ]
  br label %cleanup26.i.i.i983, !amdgpu.uniform !42

if.end22.i.i.i1046:                               ; preds = %if.then16.i.i.i1047, %if.end14.i.i.i1042
  call void @llvm.amdgcn.end.cf.i64(i64 %1437)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %Flow3656, !amdgpu.uniform !42

Flow3657:                                         ; preds = %cleanup26.i.i.i983, %1415
  %1442 = phi i32 [ %1441, %cleanup26.i.i.i983 ], [ %1404, %1415 ]
  %1443 = phi i64 [ %1440, %cleanup26.i.i.i983 ], [ %1405, %1415 ]
  %1444 = phi i32 [ %1439, %cleanup26.i.i.i983 ], [ %1406, %1415 ]
  %1445 = phi i32 [ %1439, %cleanup26.i.i.i983 ], [ %retval.sroa.8.0.i.i.i957, %1415 ]
  %1446 = phi i64 [ %1440, %cleanup26.i.i.i983 ], [ %retval.sroa.2.0.i.i.i956, %1415 ]
  %1447 = phi i32 [ %1441, %cleanup26.i.i.i983 ], [ %retval.sroa.6.0.i.i.i954, %1415 ]
  %1448 = phi i1 [ %cmp.i46.not.i.i.i975, %cleanup26.i.i.i983 ], [ true, %1415 ]
  br i1 %1448, label %for.inc.i.i.i988, label %Flow3658, !amdgpu.uniform !42

cleanup26.i.i.i983:                               ; preds = %Flow3656
  br label %Flow3657, !amdgpu.uniform !42

for.inc.i.i.i988:                                 ; preds = %Flow3657
  %inc.i.i.i993 = add i32 %spec.store.select.i.i.i960, 1
  br label %Flow3658, !amdgpu.uniform !42

Flow3658:                                         ; preds = %for.inc.i.i.i988, %Flow3657
  %1449 = phi i32 [ %inc.i.i.i993, %for.inc.i.i.i988 ], [ poison, %Flow3657 ]
  %1450 = phi i1 [ false, %for.inc.i.i.i988 ], [ true, %Flow3657 ]
  br i1 %1450, label %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i994, label %for.cond.i.i.i953, !amdgpu.uniform !42

_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i994: ; preds = %Flow3658
  %.lcssa4110 = phi i32 [ %1442, %Flow3658 ]
  %.lcssa4109 = phi i64 [ %1443, %Flow3658 ]
  %.lcssa4108 = phi i32 [ %1444, %Flow3658 ]
  fence acquire
  %1451 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %mul.i.i.i.i.i.i995 = shl i32 %.lcssa4110, 6
  %idxprom.i19.i.i.i.i.i996 = zext i32 %mul.i.i.i.i.i.i995 to i64
  %arrayidx.i20.i.i.i.i.i997 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %1451, i64 %idxprom.i19.i.i.i.i.i996
  %arrayidx.i22.i.i.i.i.i998 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %arrayidx.i20.i.i.i.i.i997, i64 %sh_prom.i.i.i.i
  store i64 %writer.sroa.10.5.i.i.i.i.i, ptr %arrayidx.i22.i.i.i.i.i998, align 8, !tbaa !82
  %spec.select.i.i.i.i.i.i.i999 = tail call i64 @llvm.umin.i64(i64 %writer.sroa.10.5.i.i.i.i.i, i64 56)
  %1452 = and i64 %spec.select.i.i.i.i.i.i.i999, 15
  %1453 = sub i64 %spec.select.i.i.i.i.i.i.i999, %1452
  %1454 = icmp ne i64 %1453, 0
  %1455 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %1454)
  %1456 = extractvalue { i1, i64 } %1455, 0
  %1457 = extractvalue { i1, i64 } %1455, 1
  br i1 %1456, label %loop-memcpy-expansion2231.preheader, label %Flow3654

loop-memcpy-expansion2231.preheader:              ; preds = %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i994
  %1458 = add nuw nsw i64 %75, 8
  %scevgep3034 = getelementptr i8, ptr %1451, i64 %1458
  %1459 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i996, 6
  %scevgep3035 = getelementptr i8, ptr %scevgep3034, i64 %1459
  br label %loop-memcpy-expansion2231, !amdgpu.uniform !42

Flow3654:                                         ; preds = %Flow3653, %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i994
  call void @llvm.amdgcn.end.cf.i64(i64 %1457)
  br label %loop-memcpy-residual-header2234, !amdgpu.uniform !42

loop-memcpy-expansion2231:                        ; preds = %loop-memcpy-expansion2231.preheader, %loop-memcpy-expansion2231
  %phi.broken3946 = phi i64 [ 0, %loop-memcpy-expansion2231.preheader ], [ %1464, %loop-memcpy-expansion2231 ]
  %loop-index2232 = phi i64 [ %1462, %loop-memcpy-expansion2231 ], [ 0, %loop-memcpy-expansion2231.preheader ]
  %1460 = getelementptr i8, ptr %1375, i64 %loop-index2232
  %1461 = load <4 x i32>, ptr %1460, align 1
  %scevgep3036 = getelementptr i8, ptr %scevgep3035, i64 %loop-index2232
  store <4 x i32> %1461, ptr %scevgep3036, align 1
  %1462 = add i64 %loop-index2232, 16
  %1463 = icmp uge i64 %1462, %1453
  %1464 = call i64 @llvm.amdgcn.if.break.i64(i1 %1463, i64 %phi.broken3946)
  %1465 = call i1 @llvm.amdgcn.loop.i64(i64 %1464)
  br i1 %1465, label %Flow3653, label %loop-memcpy-expansion2231

Flow3651:                                         ; preds = %Flow3648, %Flow3650
  %1466 = phi i64 [ %2804, %Flow3648 ], [ %2796, %Flow3650 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %2799)
  br label %post-loop-memcpy-expansion2230, !amdgpu.uniform !42

loop-memcpy-residual2233:                         ; preds = %loop-memcpy-residual2233.preheader, %loop-memcpy-residual2233
  %phi.broken3947 = phi i64 [ 0, %loop-memcpy-residual2233.preheader ], [ %1470, %loop-memcpy-residual2233 ]
  %residual-loop-index2235 = phi i64 [ %1468, %loop-memcpy-residual2233 ], [ 0, %loop-memcpy-residual2233.preheader ]
  %scevgep3033 = getelementptr i8, ptr %scevgep3032, i64 %residual-loop-index2235
  %1467 = load i8, ptr %scevgep3033, align 1
  %scevgep3031 = getelementptr i8, ptr %scevgep3030, i64 %residual-loop-index2235
  store i8 %1467, ptr %scevgep3031, align 1
  %1468 = add i64 %residual-loop-index2235, 1
  %1469 = icmp uge i64 %1468, %1452
  %1470 = call i64 @llvm.amdgcn.if.break.i64(i1 %1469, i64 %phi.broken3947)
  %1471 = call i1 @llvm.amdgcn.loop.i64(i64 %1470)
  br i1 %1471, label %Flow3648, label %loop-memcpy-residual2233

Flow3648:                                         ; preds = %loop-memcpy-residual2233
  %.lcssa4103 = phi i64 [ %1470, %loop-memcpy-residual2233 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa4103)
  br label %Flow3651, !amdgpu.uniform !42

post-loop-memcpy-expansion2230:                   ; preds = %Flow3651
  %tobool.not.i.i.i.i.i.i1001 = icmp eq i32 %.lcssa4108, 0
  %conv.i.i.i.i.i.i1002 = zext i1 %tobool.not.i.i.i.i.i.i1001 to i32
  fence release
  %1472 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %idxprom.i23.i.i.i.i.i1003 = zext i32 %.lcssa4110 to i64
  %arrayidx.i24.i.i.i.i.i1004 = getelementptr inbounds nuw i32, ptr %1472, i64 %idxprom.i23.i.i.i.i.i1003
  store atomic i32 %conv.i.i.i.i.i.i1002, ptr %arrayidx.i24.i.i.i.i.i1004 monotonic, align 4
  %1473 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !84
  %arrayidx.i.i5.i.i1005 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %1473, i64 %idxprom.i23.i.i.i.i.i1003, !amdgpu.uniform !42
  %1474 = load i64, ptr %arrayidx.i.i5.i.i1005, align 8, !tbaa !80
  %scevgep3016 = getelementptr i8, ptr %1375, i64 56
  %1475 = add i64 %writer.sroa.10.5.i.i.i.i.i, -56
  br label %while.cond.i.i.i.i1006, !amdgpu.uniform !42

while.cond.i.i.i.i1006:                           ; preds = %post-loop-memcpy-expansion2230, %Flow3646
  %phi.broken3950 = phi i64 [ 0, %post-loop-memcpy-expansion2230 ], [ %1492, %Flow3646 ]
  %lsr.iv3022 = phi i64 [ %1490, %Flow3646 ], [ %1475, %post-loop-memcpy-expansion2230 ]
  %lsr.iv3017 = phi ptr [ %1489, %Flow3646 ], [ %scevgep3016, %post-loop-memcpy-expansion2230 ]
  %port.sroa.37.0.i.i1007 = phi i32 [ %1488, %Flow3646 ], [ %conv.i.i.i.i.i.i1002, %post-loop-memcpy-expansion2230 ]
  %idx.0.i.i.i.i1008 = phi i64 [ %1487, %Flow3646 ], [ 56, %post-loop-memcpy-expansion2230 ]
  %cmp.i.i.i.i1009 = icmp ugt i64 %writer.sroa.10.5.i.i.i.i.i, %idx.0.i.i.i.i1008
  %1476 = tail call i64 @llvm.amdgcn.ballot.i64(i1 %cmp.i.i.i.i1009)
  %and.i.i.i.i6.i.i1010 = and i64 %1476, %1474
  %tobool.not.i.i.i.i1011 = icmp ne i64 %and.i.i.i.i6.i.i1010, 0
  %1477 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i22.i.i.i1012 = getelementptr inbounds nuw i32, ptr %1477, i64 %idxprom.i23.i.i.i.i.i1003, !amdgpu.uniform !42
  %1478 = load atomic i32, ptr %arrayidx.i22.i.i.i1012 monotonic, align 4
  %1479 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %tobool.not.i.i.i.i1011)
  %1480 = extractvalue { i1, i64 } %1479, 0
  %1481 = extractvalue { i1, i64 } %1479, 1
  br i1 %1480, label %cond.false.i8.i.i.i.i1013, label %Flow3646

cond.false.i8.i.i.i.i1013:                        ; preds = %while.cond.i.i.i.i1006
  %1482 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %1478)
  %1483 = icmp ne i32 %1482, %port.sroa.37.0.i.i1007
  br i1 %1483, label %while.body.i.i21.i.i.i.i1014.preheader, label %Flow3645, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i1014.preheader:           ; preds = %cond.false.i8.i.i.i.i1013
  br label %while.body.i.i21.i.i.i.i1014, !amdgpu.uniform !42

Flow3645:                                         ; preds = %Flow3643, %cond.false.i8.i.i.i.i1013
  br label %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i1017, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i1014:                     ; preds = %while.body.i.i21.i.i.i.i1014.preheader, %while.body.i.i21.i.i.i.i1014
  tail call void @llvm.amdgcn.s.sleep(i32 2)
  %1484 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i22.i.i.i.i1015 = getelementptr inbounds nuw i32, ptr %1484, i64 %idxprom.i23.i.i.i.i.i1003, !amdgpu.uniform !42
  %1485 = load atomic i32, ptr %arrayidx.i.i.i22.i.i.i.i1015 monotonic, align 4
  %1486 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %1485)
  %cmp.i.not.i.i23.i.i.i.i1016 = icmp eq i32 %1486, %port.sroa.37.0.i.i1007
  br i1 %cmp.i.not.i.i23.i.i.i.i1016, label %Flow3643, label %while.body.i.i21.i.i.i.i1014, !amdgpu.uniform !42

Flow3643:                                         ; preds = %while.body.i.i21.i.i.i.i1014
  br label %Flow3645, !amdgpu.uniform !42

Flow3646:                                         ; preds = %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1018, %while.cond.i.i.i.i1006
  %1487 = phi i64 [ %add.i.i.i.i1021, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1018 ], [ poison, %while.cond.i.i.i.i1006 ]
  %1488 = phi i32 [ %conv.i.i26.i.i.i.i1019, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1018 ], [ poison, %while.cond.i.i.i.i1006 ]
  %1489 = phi ptr [ %scevgep3018, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1018 ], [ poison, %while.cond.i.i.i.i1006 ]
  %1490 = phi i64 [ %lsr.iv.next3023, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1018 ], [ poison, %while.cond.i.i.i.i1006 ]
  %1491 = phi i1 [ false, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1018 ], [ true, %while.cond.i.i.i.i1006 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %1481)
  %1492 = call i64 @llvm.amdgcn.if.break.i64(i1 %1491, i64 %phi.broken3950)
  %1493 = call i1 @llvm.amdgcn.loop.i64(i64 %1492)
  br i1 %1493, label %cond.false.i.i.i1028, label %while.cond.i.i.i.i1006

_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i1017: ; preds = %Flow3645
  fence acquire
  %1494 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.i.i.i.i1009)
  %1495 = extractvalue { i1, i64 } %1494, 0
  %1496 = extractvalue { i1, i64 } %1494, 1
  br i1 %1495, label %if.then.i.i.i.i.i.i.i1022, label %Flow3642

if.then.i.i.i.i.i.i.i1022:                        ; preds = %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i1017
  %sub.i.i.i.i.i.i.i1024 = sub nuw i64 %writer.sroa.10.5.i.i.i.i.i, %idx.0.i.i.i.i1008
  %spec.select.i.i.i36.i.i.i.i1025 = tail call i64 @llvm.umin.i64(i64 %sub.i.i.i.i.i.i.i1024, i64 64)
  %1497 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %1498 = and i64 %spec.select.i.i.i36.i.i.i.i1025, 15
  %1499 = sub i64 %spec.select.i.i.i36.i.i.i.i1025, %1498
  %1500 = icmp ne i64 %1499, 0
  %1501 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %1500)
  %1502 = extractvalue { i1, i64 } %1501, 0
  %1503 = extractvalue { i1, i64 } %1501, 1
  br i1 %1502, label %loop-memcpy-expansion2225.preheader, label %Flow3641

loop-memcpy-expansion2225.preheader:              ; preds = %if.then.i.i.i.i.i.i.i1022
  %arrayidx.i20.i35.i.i.i.i1026 = getelementptr %"struct.rpc::Buffer", ptr %1497, i64 %idxprom.i19.i.i.i.i.i996
  %arrayidx.i22.i38.i.i.i.i1027 = getelementptr %"struct.rpc::Buffer", ptr %arrayidx.i20.i35.i.i.i.i1026, i64 %sh_prom.i.i.i.i
  br label %loop-memcpy-expansion2225, !amdgpu.uniform !42

Flow3641:                                         ; preds = %Flow3640, %if.then.i.i.i.i.i.i.i1022
  call void @llvm.amdgcn.end.cf.i64(i64 %1503)
  br label %loop-memcpy-residual-header2228, !amdgpu.uniform !42

loop-memcpy-expansion2225:                        ; preds = %loop-memcpy-expansion2225.preheader, %loop-memcpy-expansion2225
  %phi.broken3948 = phi i64 [ 0, %loop-memcpy-expansion2225.preheader ], [ %1508, %loop-memcpy-expansion2225 ]
  %loop-index2226 = phi i64 [ %1506, %loop-memcpy-expansion2225 ], [ 0, %loop-memcpy-expansion2225.preheader ]
  %scevgep3019 = getelementptr i8, ptr %lsr.iv3017, i64 %loop-index2226
  %1504 = load <4 x i32>, ptr %scevgep3019, align 1
  %1505 = getelementptr i8, ptr %arrayidx.i22.i38.i.i.i.i1027, i64 %loop-index2226
  store <4 x i32> %1504, ptr %1505, align 1
  %1506 = add i64 %loop-index2226, 16
  %1507 = icmp uge i64 %1506, %1499
  %1508 = call i64 @llvm.amdgcn.if.break.i64(i1 %1507, i64 %phi.broken3948)
  %1509 = call i1 @llvm.amdgcn.loop.i64(i64 %1508)
  br i1 %1509, label %Flow3640, label %loop-memcpy-expansion2225

Flow3638:                                         ; preds = %Flow3637, %loop-memcpy-residual-header2228
  call void @llvm.amdgcn.end.cf.i64(i64 %2789)
  br label %Flow3642, !amdgpu.uniform !42

loop-memcpy-residual2227:                         ; preds = %loop-memcpy-residual2227.preheader, %loop-memcpy-residual2227
  %phi.broken3949 = phi i64 [ 0, %loop-memcpy-residual2227.preheader ], [ %1513, %loop-memcpy-residual2227 ]
  %residual-loop-index2229 = phi i64 [ %1511, %loop-memcpy-residual2227 ], [ 0, %loop-memcpy-residual2227.preheader ]
  %scevgep3028 = getelementptr i8, ptr %scevgep3027, i64 %residual-loop-index2229
  %1510 = load i8, ptr %scevgep3028, align 1
  %scevgep3026 = getelementptr i8, ptr %scevgep3025, i64 %residual-loop-index2229
  store i8 %1510, ptr %scevgep3026, align 1
  %1511 = add i64 %residual-loop-index2229, 1
  %1512 = icmp uge i64 %1511, %1498
  %1513 = call i64 @llvm.amdgcn.if.break.i64(i1 %1512, i64 %phi.broken3949)
  %1514 = call i1 @llvm.amdgcn.loop.i64(i64 %1513)
  br i1 %1514, label %Flow3637, label %loop-memcpy-residual2227

Flow3637:                                         ; preds = %loop-memcpy-residual2227
  %.lcssa4100 = phi i64 [ %1513, %loop-memcpy-residual2227 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa4100)
  br label %Flow3638, !amdgpu.uniform !42

_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1018: ; preds = %Flow3642
  %conv.i.i26.i.i.i.i1019 = xor i32 %port.sroa.37.0.i.i1007, 1
  fence release
  %1515 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %arrayidx.i24.i29.i.i.i.i1020 = getelementptr inbounds nuw i32, ptr %1515, i64 %idxprom.i23.i.i.i.i.i1003
  store atomic i32 %conv.i.i26.i.i.i.i1019, ptr %arrayidx.i24.i29.i.i.i.i1020 monotonic, align 4
  %add.i.i.i.i1021 = add i64 %idx.0.i.i.i.i1008, 64
  %scevgep3018 = getelementptr i8, ptr %lsr.iv3017, i64 64
  %lsr.iv.next3023 = add i64 %lsr.iv3022, -64
  br label %Flow3646, !amdgpu.uniform !42

cond.false.i.i.i1028:                             ; preds = %Flow3646
  %.lcssa4102 = phi i64 [ %1492, %Flow3646 ]
  %port.sroa.37.0.i.i1007.lcssa = phi i32 [ %port.sroa.37.0.i.i1007, %Flow3646 ]
  %.lcssa4101 = phi i32 [ %1478, %Flow3646 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa4102)
  %1516 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %.lcssa4101)
  %1517 = icmp ne i32 %1516, %port.sroa.37.0.i.i1007.lcssa
  %1518 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %1517)
  %1519 = extractvalue { i1, i64 } %1518, 0
  %1520 = extractvalue { i1, i64 } %1518, 1
  br i1 %1519, label %while.body.i.i.i.i1029.preheader, label %Flow3635

while.body.i.i.i.i1029.preheader:                 ; preds = %cond.false.i.i.i1028
  br label %while.body.i.i.i.i1029, !amdgpu.uniform !42

Flow3635:                                         ; preds = %Flow3633, %cond.false.i.i.i1028
  call void @llvm.amdgcn.end.cf.i64(i64 %1520)
  br label %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsINS_3cpp6stringEEERS1_T_.exit1050, !amdgpu.uniform !42

while.body.i.i.i.i1029:                           ; preds = %while.body.i.i.i.i1029.preheader, %while.body.i.i.i.i1029
  %phi.broken3951 = phi i64 [ 0, %while.body.i.i.i.i1029.preheader ], [ %1524, %while.body.i.i.i.i1029 ]
  tail call void @llvm.amdgcn.s.sleep(i32 2)
  %1521 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i8.i.i1030 = getelementptr inbounds nuw i32, ptr %1521, i64 %idxprom.i23.i.i.i.i.i1003, !amdgpu.uniform !42
  %1522 = load atomic i32, ptr %arrayidx.i.i.i8.i.i1030 monotonic, align 4
  %1523 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %1522)
  %cmp.i.not.i.i.i.i1031 = icmp eq i32 %1523, %port.sroa.37.0.i.i1007.lcssa
  %1524 = call i64 @llvm.amdgcn.if.break.i64(i1 %cmp.i.not.i.i.i.i1031, i64 %phi.broken3951)
  %1525 = call i1 @llvm.amdgcn.loop.i64(i64 %1524)
  br i1 %1525, label %Flow3633, label %while.body.i.i.i.i1029

Flow3633:                                         ; preds = %while.body.i.i.i.i1029
  %.lcssa4098 = phi i64 [ %1524, %while.body.i.i.i.i1029 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa4098)
  br label %Flow3635, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsINS_3cpp6stringEEERS1_T_.exit1050: ; preds = %Flow3635
  fence acquire
  tail call void @llvm.amdgcn.wave.barrier()
  fence syncscope("agent") release
  %1526 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %.lcssa4109, i1 true)
  %iszero.i.i.i.i.i13.i.i1033 = icmp ne i64 %.lcssa4109, 0
  %cmp2.i.i.i.i14.i.i1034 = icmp eq i64 %1526, %sh_prom.i.i.i.i
  %cmp.i.i.i.i15.i.i1035 = select i1 %iszero.i.i.i.i.i13.i.i1033, i1 %cmp2.i.i.i.i14.i.i1034, i1 false
  %1527 = and i32 %.lcssa4110, 31
  %conv4.i.i.i19.i.i1039 = zext i1 %cmp.i.i.i.i15.i.i1035 to i32
  %shl.i.i.i20.i.i1040 = shl nuw i32 %conv4.i.i.i19.i.i1039, %1527
  %xor.i.i.i21.i.i1041 = xor i32 %shl.i.i.i20.i.i1040, -1
  %1528 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2562, !amdgpu.uniform !42

1529:                                             ; preds = %ComputeEnd2563
  %div8.i.i.i16.i.i1036 = lshr i32 %.lcssa4110, 5
  %idxprom.i.i.i17.i.i1037 = zext nneg i32 %div8.i.i.i16.i.i1036 to i64
  %1530 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i17.i.i1037
  %sunkaddr3319 = getelementptr inbounds i8, ptr addrspace(1) %1530, i64 40
  %1531 = atomicrmw and ptr addrspace(1) %sunkaddr3319, i32 %.lcssa4097 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %1532, !amdgpu.uniform !42

1532:                                             ; preds = %1529, %ComputeEnd2563
  call void @llvm.amdgcn.end.cf.i64(i64 %3561)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %for.cond.i.i.i1054, !amdgpu.uniform !42

for.cond.i.i.i1054:                               ; preds = %1532, %Flow3632
  %1533 = phi i32 [ %1571, %Flow3632 ], [ poison, %1532 ]
  %1534 = phi i64 [ %1572, %Flow3632 ], [ poison, %1532 ]
  %1535 = phi i32 [ %1573, %Flow3632 ], [ poison, %1532 ]
  %retval.sroa.6.0.i.i.i1055 = phi i32 [ %1576, %Flow3632 ], [ undef, %1532 ]
  %retval.sroa.2.0.i.i.i1057 = phi i64 [ %1575, %Flow3632 ], [ undef, %1532 ]
  %retval.sroa.8.0.i.i.i1058 = phi i32 [ %1574, %Flow3632 ], [ undef, %1532 ]
  %index.0.i.i.i1059 = phi i32 [ %1578, %Flow3632 ], [ 0, %1532 ]
  %1536 = load i32, ptr addrspace(1) @__llvm_rpc_client, align 8, !tbaa !68
  %cmp.not.i.i.i1060 = icmp ult i32 %index.0.i.i.i1059, %1536
  %spec.store.select.i.i.i1061 = select i1 %cmp.not.i.i.i1060, i32 %index.0.i.i.i1059, i32 0
  %1537 = tail call noundef i64 @llvm.amdgcn.ballot.i64(i1 true)
  %1538 = and i32 %spec.store.select.i.i.i1061, 31
  %1539 = lshr i64 %1537, %sh_prom.i.i.i.i
  %1540 = trunc i64 %1539 to i32
  %conv4.i.i.i.i.i1065 = and i32 %1540, 1
  %shl.i.i.i.i.i1066 = shl nuw i32 %conv4.i.i.i.i.i1065, %1538
  %1541 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2566, !amdgpu.uniform !42

1542:                                             ; preds = %ComputeEnd2567
  %sunkaddr3320 = getelementptr inbounds i8, ptr addrspace(1) %3571, i64 40
  %1543 = atomicrmw or ptr addrspace(1) %sunkaddr3320, i32 %.lcssa4091 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %1544, !amdgpu.uniform !42

1544:                                             ; preds = %1542, %ComputeEnd2567
  %1545 = phi i32 [ %1543, %1542 ], [ poison, %ComputeEnd2567 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %3580)
  %1546 = call i32 @llvm.amdgcn.readfirstlane.i32(i32 %1545)
  %1547 = or i32 %1546, %.lcssa4092
  %shl5.i.i.i.i.i1067 = shl nuw i32 1, %1538
  %and.i.i.i.i.i1068 = and i32 %shl5.i.i.i.i.i1067, %1547
  %tobool3.i.i.i.i1069 = icmp ne i32 %and.i.i.i.i.i1068, 0
  %1548 = tail call i64 @llvm.amdgcn.ballot.i64(i1 %tobool3.i.i.i.i1069)
  %and.i.i.i.i.i.i1070 = and i64 %1548, %1537
  %cmp.i.not.i.i.i1071 = icmp ne i64 %1537, %and.i.i.i.i.i.i1070
  br i1 %cmp.i.not.i.i.i1071, label %if.then.i.i.i.i1072, label %Flow3631, !amdgpu.uniform !42

if.then.i.i.i.i1072:                              ; preds = %1544
  fence syncscope("agent") acquire
  %1549 = load <2 x i64>, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !75
  %1550 = extractelement <2 x i64> %1549, i32 0
  %1551 = inttoptr i64 %1550 to ptr
  %1552 = extractelement <2 x i64> %1549, i32 1
  %1553 = inttoptr i64 %1552 to ptr
  %idxprom.i.i.i.i1073 = zext i32 %spec.store.select.i.i.i1061 to i64
  %arrayidx.i.i.i.i1074 = getelementptr inbounds nuw i32, ptr %1551, i64 %idxprom.i.i.i.i1073, !amdgpu.uniform !42
  %1554 = load atomic i32, ptr %arrayidx.i.i.i.i1074 monotonic, align 4
  %1555 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %1554)
  %arrayidx.i45.i.i.i1075 = getelementptr inbounds nuw i32, ptr %1553, i64 %idxprom.i.i.i.i1073, !amdgpu.uniform !42
  %1556 = load atomic i32, ptr %arrayidx.i45.i.i.i1075 monotonic, align 4
  %1557 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %1556)
  %cmp.i46.not.i.i.i1076 = icmp ne i32 %1555, %1557
  br i1 %cmp.i46.not.i.i.i1076, label %if.then12.i.i.i1077, label %Flow3629, !amdgpu.uniform !42

if.then12.i.i.i1077:                              ; preds = %if.then.i.i.i.i1072
  fence syncscope("agent") release
  %1558 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %1537, i1 true)
  %iszero.i.i.i.i.i.i.i1078 = icmp ne i64 %1537, 0
  %cmp2.i.i.i.i.i.i1079 = icmp eq i64 %1558, %sh_prom.i.i.i.i
  %cmp.i.i.i.i.i.i1080 = select i1 %iszero.i.i.i.i.i.i.i1078, i1 %cmp2.i.i.i.i.i.i1079, i1 false
  %conv4.i.i51.i.i.i1081 = zext i1 %cmp.i.i.i.i.i.i1080 to i32
  %shl.i.i52.i.i.i1082 = shl nuw i32 %conv4.i.i51.i.i.i1081, %1538
  %xor.i.i.i.i.i1083 = xor i32 %shl.i.i52.i.i.i1082, -1
  %1559 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2571, !amdgpu.uniform !42

1560:                                             ; preds = %ComputeEnd2572
  %sunkaddr3321 = getelementptr inbounds i8, ptr addrspace(1) %3571, i64 40
  %1561 = atomicrmw and ptr addrspace(1) %sunkaddr3321, i32 %.lcssa4093 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %1562, !amdgpu.uniform !42

1562:                                             ; preds = %1560, %ComputeEnd2572
  call void @llvm.amdgcn.end.cf.i64(i64 %3598)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %Flow3629, !amdgpu.uniform !42

if.end14.i.i.i1141:                               ; preds = %Flow3629
  %1563 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %1537, i1 true)
  %iszero.i.i.i.i.i.i1142 = icmp ne i64 %1537, 0
  %cmp2.i.i.i.i.i1143 = icmp eq i64 %1563, %sh_prom.i.i.i.i
  %cmp.i.i.i.i.i1144 = select i1 %iszero.i.i.i.i.i.i1142, i1 %cmp2.i.i.i.i.i1143, i1 false
  %1564 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.i.i.i.i.i1144)
  %1565 = extractvalue { i1, i64 } %1564, 0
  %1566 = extractvalue { i1, i64 } %1564, 1
  br i1 %1565, label %if.then16.i.i.i1146, label %if.end22.i.i.i1145

if.then16.i.i.i1146:                              ; preds = %if.end14.i.i.i1141
  %1567 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !76
  %opcode.i.i.i1147 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %1567, i64 %idxprom.i.i.i.i1073, i32 1
  store i32 1660944387, ptr %opcode.i.i.i1147, align 8, !tbaa !77
  %arrayidx21.i.i.i1148 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %1567, i64 %idxprom.i.i.i.i1073
  store i64 %1537, ptr %arrayidx21.i.i.i1148, align 8, !tbaa !80
  br label %if.end22.i.i.i1145, !amdgpu.uniform !42

Flow3630:                                         ; preds = %if.end22.i.i.i1145, %Flow3629
  %1568 = phi i32 [ %1555, %if.end22.i.i.i1145 ], [ %retval.sroa.8.0.i.i.i1058, %Flow3629 ]
  %1569 = phi i64 [ %1537, %if.end22.i.i.i1145 ], [ %retval.sroa.2.0.i.i.i1057, %Flow3629 ]
  %1570 = phi i32 [ %spec.store.select.i.i.i1061, %if.end22.i.i.i1145 ], [ %retval.sroa.6.0.i.i.i1055, %Flow3629 ]
  br label %cleanup26.i.i.i1084, !amdgpu.uniform !42

if.end22.i.i.i1145:                               ; preds = %if.then16.i.i.i1146, %if.end14.i.i.i1141
  call void @llvm.amdgcn.end.cf.i64(i64 %1566)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %Flow3630, !amdgpu.uniform !42

Flow3631:                                         ; preds = %cleanup26.i.i.i1084, %1544
  %1571 = phi i32 [ %1570, %cleanup26.i.i.i1084 ], [ %1533, %1544 ]
  %1572 = phi i64 [ %1569, %cleanup26.i.i.i1084 ], [ %1534, %1544 ]
  %1573 = phi i32 [ %1568, %cleanup26.i.i.i1084 ], [ %1535, %1544 ]
  %1574 = phi i32 [ %1568, %cleanup26.i.i.i1084 ], [ %retval.sroa.8.0.i.i.i1058, %1544 ]
  %1575 = phi i64 [ %1569, %cleanup26.i.i.i1084 ], [ %retval.sroa.2.0.i.i.i1057, %1544 ]
  %1576 = phi i32 [ %1570, %cleanup26.i.i.i1084 ], [ %retval.sroa.6.0.i.i.i1055, %1544 ]
  %1577 = phi i1 [ %cmp.i46.not.i.i.i1076, %cleanup26.i.i.i1084 ], [ true, %1544 ]
  br i1 %1577, label %for.inc.i.i.i1089, label %Flow3632, !amdgpu.uniform !42

cleanup26.i.i.i1084:                              ; preds = %Flow3630
  br label %Flow3631, !amdgpu.uniform !42

for.inc.i.i.i1089:                                ; preds = %Flow3631
  %inc.i.i.i1094 = add i32 %spec.store.select.i.i.i1061, 1
  br label %Flow3632, !amdgpu.uniform !42

Flow3632:                                         ; preds = %for.inc.i.i.i1089, %Flow3631
  %1578 = phi i32 [ %inc.i.i.i1094, %for.inc.i.i.i1089 ], [ poison, %Flow3631 ]
  %1579 = phi i1 [ false, %for.inc.i.i.i1089 ], [ true, %Flow3631 ]
  br i1 %1579, label %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1095, label %for.cond.i.i.i1054, !amdgpu.uniform !42

_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1095: ; preds = %Flow3632
  %.lcssa4096 = phi i32 [ %1571, %Flow3632 ]
  %.lcssa4095 = phi i64 [ %1572, %Flow3632 ]
  %.lcssa4094 = phi i32 [ %1573, %Flow3632 ]
  fence acquire
  %1580 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %mul.i.i.i.i.i.i1096 = shl i32 %.lcssa4096, 6
  %idxprom.i19.i.i.i.i.i1097 = zext i32 %mul.i.i.i.i.i.i1096 to i64
  %arrayidx.i20.i.i.i.i.i1098 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %1580, i64 %idxprom.i19.i.i.i.i.i1097
  %arrayidx.i22.i.i.i.i.i1099 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %arrayidx.i20.i.i.i.i.i1098, i64 %sh_prom.i.i.i.i
  store i64 1, ptr %arrayidx.i22.i.i.i.i.i1099, align 8, !tbaa !82
  %arrayidx6.i.i.i.i.i.i.i1100 = getelementptr inbounds nuw i8, ptr %arrayidx.i22.i.i.i.i.i1099, i64 8
  store i8 10, ptr %arrayidx6.i.i.i.i.i.i.i1100, align 8
  %tobool.not.i.i.i.i.i.i1101 = icmp eq i32 %.lcssa4094, 0
  %conv.i.i.i.i.i.i1102 = zext i1 %tobool.not.i.i.i.i.i.i1101 to i32
  fence release
  %1581 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %idxprom.i23.i.i.i.i.i1103 = zext i32 %.lcssa4096 to i64
  %arrayidx.i24.i.i.i.i.i1104 = getelementptr inbounds nuw i32, ptr %1581, i64 %idxprom.i23.i.i.i.i.i1103
  store atomic i32 %conv.i.i.i.i.i.i1102, ptr %arrayidx.i24.i.i.i.i.i1104 monotonic, align 4
  %1582 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !84
  %arrayidx.i.i5.i.i1105 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %1582, i64 %idxprom.i23.i.i.i.i.i1103, !amdgpu.uniform !42
  %1583 = load i64, ptr %arrayidx.i.i5.i.i1105, align 8, !tbaa !80
  br label %while.cond.i.i.i.i1106, !amdgpu.uniform !42

while.cond.i.i.i.i1106:                           ; preds = %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1095, %Flow3628
  %phi.broken3952 = phi i64 [ 0, %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1095 ], [ %1598, %Flow3628 ]
  %port.sroa.37.0.i.i1107 = phi i32 [ %1596, %Flow3628 ], [ %conv.i.i.i.i.i.i1102, %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1095 ]
  %idx.0.i.i.i.i1108 = phi i64 [ %1595, %Flow3628 ], [ 56, %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1095 ]
  %cmp.i.i.i.i1109 = icmp eq i64 %idx.0.i.i.i.i1108, 0
  %1584 = tail call i64 @llvm.amdgcn.ballot.i64(i1 %cmp.i.i.i.i1109)
  %and.i.i.i.i6.i.i1110 = and i64 %1584, %1583
  %tobool.not.i.i.i.i1111 = icmp ne i64 %and.i.i.i.i6.i.i1110, 0
  %1585 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i22.i.i.i1112 = getelementptr inbounds nuw i32, ptr %1585, i64 %idxprom.i23.i.i.i.i.i1103, !amdgpu.uniform !42
  %1586 = load atomic i32, ptr %arrayidx.i22.i.i.i1112 monotonic, align 4
  %1587 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %tobool.not.i.i.i.i1111)
  %1588 = extractvalue { i1, i64 } %1587, 0
  %1589 = extractvalue { i1, i64 } %1587, 1
  br i1 %1588, label %cond.false.i8.i.i.i.i1113, label %Flow3628

cond.false.i8.i.i.i.i1113:                        ; preds = %while.cond.i.i.i.i1106
  %1590 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %1586)
  %1591 = icmp ne i32 %1590, %port.sroa.37.0.i.i1107
  br i1 %1591, label %while.body.i.i21.i.i.i.i1114.preheader, label %Flow3627, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i1114.preheader:           ; preds = %cond.false.i8.i.i.i.i1113
  br label %while.body.i.i21.i.i.i.i1114, !amdgpu.uniform !42

Flow3627:                                         ; preds = %Flow3625, %cond.false.i8.i.i.i.i1113
  br label %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i1117, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i1114:                     ; preds = %while.body.i.i21.i.i.i.i1114.preheader, %while.body.i.i21.i.i.i.i1114
  tail call void @llvm.amdgcn.s.sleep(i32 2)
  %1592 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i22.i.i.i.i1115 = getelementptr inbounds nuw i32, ptr %1592, i64 %idxprom.i23.i.i.i.i.i1103, !amdgpu.uniform !42
  %1593 = load atomic i32, ptr %arrayidx.i.i.i22.i.i.i.i1115 monotonic, align 4
  %1594 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %1593)
  %cmp.i.not.i.i23.i.i.i.i1116 = icmp eq i32 %1594, %port.sroa.37.0.i.i1107
  br i1 %cmp.i.not.i.i23.i.i.i.i1116, label %Flow3625, label %while.body.i.i21.i.i.i.i1114, !amdgpu.uniform !42

Flow3625:                                         ; preds = %while.body.i.i21.i.i.i.i1114
  br label %Flow3627, !amdgpu.uniform !42

Flow3628:                                         ; preds = %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1118, %while.cond.i.i.i.i1106
  %1595 = phi i64 [ %add.i.i.i.i1121, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1118 ], [ poison, %while.cond.i.i.i.i1106 ]
  %1596 = phi i32 [ %conv.i.i26.i.i.i.i1119, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1118 ], [ poison, %while.cond.i.i.i.i1106 ]
  %1597 = phi i1 [ false, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1118 ], [ true, %while.cond.i.i.i.i1106 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %1589)
  %1598 = call i64 @llvm.amdgcn.if.break.i64(i1 %1597, i64 %phi.broken3952)
  %1599 = call i1 @llvm.amdgcn.loop.i64(i64 %1598)
  br i1 %1599, label %cond.false.i.i.i1127, label %while.cond.i.i.i.i1106

_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i1117: ; preds = %Flow3627
  fence acquire
  br i1 %cmp.i.i.i.i1109, label %if.then.i.i.i.i.i.i.i1122, label %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1118, !amdgpu.uniform !42

if.then.i.i.i.i.i.i.i1122:                        ; preds = %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i1117
  %1600 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %arrayidx.i20.i35.i.i.i.i1125 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %1600, i64 %idxprom.i19.i.i.i.i.i1097
  %arrayidx.i22.i38.i.i.i.i1126 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %arrayidx.i20.i35.i.i.i.i1125, i64 %sh_prom.i.i.i.i
  store i8 10, ptr %arrayidx.i22.i38.i.i.i.i1126, align 1
  br label %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1118, !amdgpu.uniform !42

_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1118: ; preds = %if.then.i.i.i.i.i.i.i1122, %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i1117
  %conv.i.i26.i.i.i.i1119 = xor i32 %port.sroa.37.0.i.i1107, 1
  fence release
  %1601 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %arrayidx.i24.i29.i.i.i.i1120 = getelementptr inbounds nuw i32, ptr %1601, i64 %idxprom.i23.i.i.i.i.i1103
  store atomic i32 %conv.i.i26.i.i.i.i1119, ptr %arrayidx.i24.i29.i.i.i.i1120 monotonic, align 4
  %add.i.i.i.i1121 = add i64 %idx.0.i.i.i.i1108, 64
  br label %Flow3628, !amdgpu.uniform !42

cond.false.i.i.i1127:                             ; preds = %Flow3628
  %.lcssa4090 = phi i64 [ %1598, %Flow3628 ]
  %port.sroa.37.0.i.i1107.lcssa = phi i32 [ %port.sroa.37.0.i.i1107, %Flow3628 ]
  %.lcssa4089 = phi i32 [ %1586, %Flow3628 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa4090)
  %1602 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %.lcssa4089)
  %1603 = icmp ne i32 %1602, %port.sroa.37.0.i.i1107.lcssa
  %1604 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %1603)
  %1605 = extractvalue { i1, i64 } %1604, 0
  %1606 = extractvalue { i1, i64 } %1604, 1
  br i1 %1605, label %while.body.i.i.i.i1128.preheader, label %Flow3624

while.body.i.i.i.i1128.preheader:                 ; preds = %cond.false.i.i.i1127
  br label %while.body.i.i.i.i1128, !amdgpu.uniform !42

Flow3624:                                         ; preds = %Flow3622, %cond.false.i.i.i1127
  call void @llvm.amdgcn.end.cf.i64(i64 %1606)
  br label %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIcEERS1_T_.exit1149, !amdgpu.uniform !42

while.body.i.i.i.i1128:                           ; preds = %while.body.i.i.i.i1128.preheader, %while.body.i.i.i.i1128
  %phi.broken3953 = phi i64 [ 0, %while.body.i.i.i.i1128.preheader ], [ %1610, %while.body.i.i.i.i1128 ]
  tail call void @llvm.amdgcn.s.sleep(i32 2)
  %1607 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i8.i.i1129 = getelementptr inbounds nuw i32, ptr %1607, i64 %idxprom.i23.i.i.i.i.i1103, !amdgpu.uniform !42
  %1608 = load atomic i32, ptr %arrayidx.i.i.i8.i.i1129 monotonic, align 4
  %1609 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %1608)
  %cmp.i.not.i.i.i.i1130 = icmp eq i32 %1609, %port.sroa.37.0.i.i1107.lcssa
  %1610 = call i64 @llvm.amdgcn.if.break.i64(i1 %cmp.i.not.i.i.i.i1130, i64 %phi.broken3953)
  %1611 = call i1 @llvm.amdgcn.loop.i64(i64 %1610)
  br i1 %1611, label %Flow3622, label %while.body.i.i.i.i1128

Flow3622:                                         ; preds = %while.body.i.i.i.i1128
  %.lcssa4088 = phi i64 [ %1610, %while.body.i.i.i.i1128 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa4088)
  br label %Flow3624, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIcEERS1_T_.exit1149: ; preds = %Flow3624
  fence acquire
  tail call void @llvm.amdgcn.wave.barrier()
  fence syncscope("agent") release
  %1612 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %.lcssa4095, i1 true)
  %iszero.i.i.i.i.i13.i.i1132 = icmp ne i64 %.lcssa4095, 0
  %cmp2.i.i.i.i14.i.i1133 = icmp eq i64 %1612, %sh_prom.i.i.i.i
  %cmp.i.i.i.i15.i.i1134 = select i1 %iszero.i.i.i.i.i13.i.i1132, i1 %cmp2.i.i.i.i14.i.i1133, i1 false
  %1613 = and i32 %.lcssa4096, 31
  %conv4.i.i.i19.i.i1138 = zext i1 %cmp.i.i.i.i15.i.i1134 to i32
  %shl.i.i.i20.i.i1139 = shl nuw i32 %conv4.i.i.i19.i.i1138, %1613
  %xor.i.i.i21.i.i1140 = xor i32 %shl.i.i.i20.i.i1139, -1
  %1614 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2575, !amdgpu.uniform !42

1615:                                             ; preds = %ComputeEnd2576
  %div8.i.i.i16.i.i1135 = lshr i32 %.lcssa4096, 5
  %idxprom.i.i.i17.i.i1136 = zext nneg i32 %div8.i.i.i16.i.i1135 to i64
  %1616 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i17.i.i1136
  %sunkaddr3322 = getelementptr inbounds i8, ptr addrspace(1) %1616, i64 40
  %1617 = atomicrmw and ptr addrspace(1) %sunkaddr3322, i32 %.lcssa4087 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %1618, !amdgpu.uniform !42

1618:                                             ; preds = %1615, %ComputeEnd2576
  call void @llvm.amdgcn.end.cf.i64(i64 %3615)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %for.cond.i.i1.i1158, !amdgpu.uniform !42

for.cond.i.i1.i1158:                              ; preds = %1618, %Flow3621
  %1619 = phi i32 [ %1657, %Flow3621 ], [ poison, %1618 ]
  %1620 = phi i64 [ %1658, %Flow3621 ], [ poison, %1618 ]
  %1621 = phi i32 [ %1659, %Flow3621 ], [ poison, %1618 ]
  %retval.sroa.6.0.i.i.i1159 = phi i32 [ %1662, %Flow3621 ], [ undef, %1618 ]
  %retval.sroa.2.0.i.i.i1161 = phi i64 [ %1661, %Flow3621 ], [ undef, %1618 ]
  %retval.sroa.8.0.i.i.i1162 = phi i32 [ %1660, %Flow3621 ], [ undef, %1618 ]
  %index.0.i.i.i1163 = phi i32 [ %1664, %Flow3621 ], [ 0, %1618 ]
  %1622 = load i32, ptr addrspace(1) @__llvm_rpc_client, align 8, !tbaa !68
  %cmp.not.i.i.i1164 = icmp ult i32 %index.0.i.i.i1163, %1622
  %spec.store.select.i.i.i1165 = select i1 %cmp.not.i.i.i1164, i32 %index.0.i.i.i1163, i32 0
  %1623 = tail call noundef i64 @llvm.amdgcn.ballot.i64(i1 true)
  %1624 = and i32 %spec.store.select.i.i.i1165, 31
  %1625 = lshr i64 %1623, %sh_prom.i.i.i.i
  %1626 = trunc i64 %1625 to i32
  %conv4.i.i.i.i.i1169 = and i32 %1626, 1
  %shl.i.i.i.i.i1170 = shl nuw i32 %conv4.i.i.i.i.i1169, %1624
  %1627 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2579, !amdgpu.uniform !42

1628:                                             ; preds = %ComputeEnd2580
  %sunkaddr3323 = getelementptr inbounds i8, ptr addrspace(1) %3625, i64 40
  %1629 = atomicrmw or ptr addrspace(1) %sunkaddr3323, i32 %.lcssa4081 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %1630, !amdgpu.uniform !42

1630:                                             ; preds = %1628, %ComputeEnd2580
  %1631 = phi i32 [ %1629, %1628 ], [ poison, %ComputeEnd2580 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %3634)
  %1632 = call i32 @llvm.amdgcn.readfirstlane.i32(i32 %1631)
  %1633 = or i32 %1632, %.lcssa4082
  %shl5.i.i.i.i.i1171 = shl nuw i32 1, %1624
  %and.i.i.i.i.i1172 = and i32 %shl5.i.i.i.i.i1171, %1633
  %tobool3.i.i.i.i1173 = icmp ne i32 %and.i.i.i.i.i1172, 0
  %1634 = tail call i64 @llvm.amdgcn.ballot.i64(i1 %tobool3.i.i.i.i1173)
  %and.i.i.i.i.i.i1174 = and i64 %1634, %1623
  %cmp.i.not.i.i.i1175 = icmp ne i64 %1623, %and.i.i.i.i.i.i1174
  br i1 %cmp.i.not.i.i.i1175, label %if.then.i.i.i.i1176, label %Flow3620, !amdgpu.uniform !42

if.then.i.i.i.i1176:                              ; preds = %1630
  fence syncscope("agent") acquire
  %1635 = load <2 x i64>, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !75
  %1636 = extractelement <2 x i64> %1635, i32 0
  %1637 = inttoptr i64 %1636 to ptr
  %1638 = extractelement <2 x i64> %1635, i32 1
  %1639 = inttoptr i64 %1638 to ptr
  %idxprom.i.i.i.i1177 = zext i32 %spec.store.select.i.i.i1165 to i64
  %arrayidx.i.i.i.i1178 = getelementptr inbounds nuw i32, ptr %1637, i64 %idxprom.i.i.i.i1177, !amdgpu.uniform !42
  %1640 = load atomic i32, ptr %arrayidx.i.i.i.i1178 monotonic, align 4
  %1641 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %1640)
  %arrayidx.i45.i.i.i1179 = getelementptr inbounds nuw i32, ptr %1639, i64 %idxprom.i.i.i.i1177, !amdgpu.uniform !42
  %1642 = load atomic i32, ptr %arrayidx.i45.i.i.i1179 monotonic, align 4
  %1643 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %1642)
  %cmp.i46.not.i.i.i1180 = icmp ne i32 %1641, %1643
  br i1 %cmp.i46.not.i.i.i1180, label %if.then12.i.i.i1181, label %Flow3618, !amdgpu.uniform !42

if.then12.i.i.i1181:                              ; preds = %if.then.i.i.i.i1176
  fence syncscope("agent") release
  %1644 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %1623, i1 true)
  %iszero.i.i.i.i.i.i.i1182 = icmp ne i64 %1623, 0
  %cmp2.i.i.i.i.i.i1183 = icmp eq i64 %1644, %sh_prom.i.i.i.i
  %cmp.i.i.i.i.i.i1184 = select i1 %iszero.i.i.i.i.i.i.i1182, i1 %cmp2.i.i.i.i.i.i1183, i1 false
  %conv4.i.i51.i.i.i1185 = zext i1 %cmp.i.i.i.i.i.i1184 to i32
  %shl.i.i52.i.i.i1186 = shl nuw i32 %conv4.i.i51.i.i.i1185, %1624
  %xor.i.i.i.i.i1187 = xor i32 %shl.i.i52.i.i.i1186, -1
  %1645 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2584, !amdgpu.uniform !42

1646:                                             ; preds = %ComputeEnd2585
  %sunkaddr3324 = getelementptr inbounds i8, ptr addrspace(1) %3625, i64 40
  %1647 = atomicrmw and ptr addrspace(1) %sunkaddr3324, i32 %.lcssa4083 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %1648, !amdgpu.uniform !42

1648:                                             ; preds = %1646, %ComputeEnd2585
  call void @llvm.amdgcn.end.cf.i64(i64 %3652)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %Flow3618, !amdgpu.uniform !42

if.end14.i.i.i1247:                               ; preds = %Flow3618
  %1649 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %1623, i1 true)
  %iszero.i.i.i.i.i.i1248 = icmp ne i64 %1623, 0
  %cmp2.i.i.i.i.i1249 = icmp eq i64 %1649, %sh_prom.i.i.i.i
  %cmp.i.i.i.i.i1250 = select i1 %iszero.i.i.i.i.i.i1248, i1 %cmp2.i.i.i.i.i1249, i1 false
  %1650 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.i.i.i.i.i1250)
  %1651 = extractvalue { i1, i64 } %1650, 0
  %1652 = extractvalue { i1, i64 } %1650, 1
  br i1 %1651, label %if.then16.i.i.i1252, label %if.end22.i.i.i1251

if.then16.i.i.i1252:                              ; preds = %if.end14.i.i.i1247
  %1653 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !76
  %opcode.i.i.i1253 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %1653, i64 %idxprom.i.i.i.i1177, i32 1
  store i32 1660944387, ptr %opcode.i.i.i1253, align 8, !tbaa !77
  %arrayidx21.i.i.i1254 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %1653, i64 %idxprom.i.i.i.i1177
  store i64 %1623, ptr %arrayidx21.i.i.i1254, align 8, !tbaa !80
  br label %if.end22.i.i.i1251, !amdgpu.uniform !42

Flow3619:                                         ; preds = %if.end22.i.i.i1251, %Flow3618
  %1654 = phi i32 [ %1641, %if.end22.i.i.i1251 ], [ %retval.sroa.8.0.i.i.i1162, %Flow3618 ]
  %1655 = phi i64 [ %1623, %if.end22.i.i.i1251 ], [ %retval.sroa.2.0.i.i.i1161, %Flow3618 ]
  %1656 = phi i32 [ %spec.store.select.i.i.i1165, %if.end22.i.i.i1251 ], [ %retval.sroa.6.0.i.i.i1159, %Flow3618 ]
  br label %cleanup26.i.i.i1188, !amdgpu.uniform !42

if.end22.i.i.i1251:                               ; preds = %if.then16.i.i.i1252, %if.end14.i.i.i1247
  call void @llvm.amdgcn.end.cf.i64(i64 %1652)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %Flow3619, !amdgpu.uniform !42

Flow3620:                                         ; preds = %cleanup26.i.i.i1188, %1630
  %1657 = phi i32 [ %1656, %cleanup26.i.i.i1188 ], [ %1619, %1630 ]
  %1658 = phi i64 [ %1655, %cleanup26.i.i.i1188 ], [ %1620, %1630 ]
  %1659 = phi i32 [ %1654, %cleanup26.i.i.i1188 ], [ %1621, %1630 ]
  %1660 = phi i32 [ %1654, %cleanup26.i.i.i1188 ], [ %retval.sroa.8.0.i.i.i1162, %1630 ]
  %1661 = phi i64 [ %1655, %cleanup26.i.i.i1188 ], [ %retval.sroa.2.0.i.i.i1161, %1630 ]
  %1662 = phi i32 [ %1656, %cleanup26.i.i.i1188 ], [ %retval.sroa.6.0.i.i.i1159, %1630 ]
  %1663 = phi i1 [ %cmp.i46.not.i.i.i1180, %cleanup26.i.i.i1188 ], [ true, %1630 ]
  br i1 %1663, label %for.inc.i.i.i1193, label %Flow3621, !amdgpu.uniform !42

cleanup26.i.i.i1188:                              ; preds = %Flow3619
  br label %Flow3620, !amdgpu.uniform !42

for.inc.i.i.i1193:                                ; preds = %Flow3620
  %inc.i.i.i1198 = add i32 %spec.store.select.i.i.i1165, 1
  br label %Flow3621, !amdgpu.uniform !42

Flow3621:                                         ; preds = %for.inc.i.i.i1193, %Flow3620
  %1664 = phi i32 [ %inc.i.i.i1198, %for.inc.i.i.i1193 ], [ poison, %Flow3620 ]
  %1665 = phi i1 [ false, %for.inc.i.i.i1193 ], [ true, %Flow3620 ]
  br i1 %1665, label %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1199, label %for.cond.i.i1.i1158, !amdgpu.uniform !42

_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1199: ; preds = %Flow3621
  %.lcssa4086 = phi i32 [ %1657, %Flow3621 ]
  %.lcssa4085 = phi i64 [ %1658, %Flow3621 ]
  %.lcssa4084 = phi i32 [ %1659, %Flow3621 ]
  fence acquire
  %1666 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %mul.i.i.i.i.i.i1200 = shl i32 %.lcssa4086, 6
  %idxprom.i19.i.i.i.i.i1201 = zext i32 %mul.i.i.i.i.i.i1200 to i64
  %arrayidx.i20.i.i.i.i.i1202 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %1666, i64 %idxprom.i19.i.i.i.i.i1201
  %arrayidx.i22.i.i.i.i.i1203 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %arrayidx.i20.i.i.i.i.i1202, i64 %sh_prom.i.i.i.i
  store i64 sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.36.13 to ptr), i64 6) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.36.13 to ptr) to i64)), ptr %arrayidx.i22.i.i.i.i.i1203, align 8, !tbaa !82
  %spec.select.i.i.i.i.i.i.i1204 = tail call i64 @llvm.umin.i64(i64 sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.36.13 to ptr), i64 6) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.36.13 to ptr) to i64)), i64 56)
  %1667 = and i64 %spec.select.i.i.i.i.i.i.i1204, 15
  %1668 = sub i64 %spec.select.i.i.i.i.i.i.i1204, %1667
  %1669 = icmp ne i64 %1668, 0
  br i1 %1669, label %loop-memcpy-expansion2387.preheader, label %Flow3617, !amdgpu.uniform !42

loop-memcpy-expansion2387.preheader:              ; preds = %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1199
  %1670 = add nuw nsw i64 %75, 8
  %scevgep3013 = getelementptr i8, ptr %1666, i64 %1670
  %1671 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i1201, 6
  %scevgep3014 = getelementptr i8, ptr %scevgep3013, i64 %1671
  br label %loop-memcpy-expansion2387, !amdgpu.uniform !42

Flow3617:                                         ; preds = %Flow3616, %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1199
  br label %loop-memcpy-residual-header2390, !amdgpu.uniform !42

loop-memcpy-expansion2387:                        ; preds = %loop-memcpy-expansion2387.preheader, %loop-memcpy-expansion2387
  %loop-index2388 = phi i64 [ %1674, %loop-memcpy-expansion2387 ], [ 0, %loop-memcpy-expansion2387.preheader ]
  %1672 = getelementptr i8, ptr addrspace(4) @.str.36.13, i64 %loop-index2388, !amdgpu.uniform !42
  %1673 = load <4 x i32>, ptr addrspace(4) %1672, align 1
  %scevgep3015 = getelementptr i8, ptr %scevgep3014, i64 %loop-index2388
  store <4 x i32> %1673, ptr %scevgep3015, align 1
  %1674 = add i64 %loop-index2388, 16
  %1675 = icmp uge i64 %1674, %1668
  br i1 %1675, label %Flow3616, label %loop-memcpy-expansion2387, !amdgpu.uniform !42

Flow3614:                                         ; preds = %Flow3611, %Flow3613
  %1676 = phi i64 [ %2942, %Flow3611 ], [ %2939, %Flow3613 ]
  br label %post-loop-memcpy-expansion2386, !amdgpu.uniform !42

loop-memcpy-residual2389:                         ; preds = %loop-memcpy-residual2389.preheader, %loop-memcpy-residual2389
  %residual-loop-index2391 = phi i64 [ %1678, %loop-memcpy-residual2389 ], [ 0, %loop-memcpy-residual2389.preheader ]
  %scevgep3012 = getelementptr i8, ptr addrspace(4) @.str.36.13, i64 %residual-loop-index2391, !amdgpu.uniform !42
  %1677 = load i8, ptr addrspace(4) %scevgep3012, align 1
  %scevgep3011 = getelementptr i8, ptr %scevgep3010, i64 %residual-loop-index2391
  store i8 %1677, ptr %scevgep3011, align 1
  %1678 = add i64 %residual-loop-index2391, 1
  %1679 = icmp uge i64 %1678, %1667
  br i1 %1679, label %Flow3611, label %loop-memcpy-residual2389, !amdgpu.uniform !42

Flow3611:                                         ; preds = %loop-memcpy-residual2389
  br label %Flow3614, !amdgpu.uniform !42

post-loop-memcpy-expansion2386:                   ; preds = %Flow3614
  %tobool.not.i.i.i.i.i.i1206 = icmp eq i32 %.lcssa4084, 0
  %conv.i.i.i.i.i.i1207 = zext i1 %tobool.not.i.i.i.i.i.i1206 to i32
  fence release
  %1680 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %idxprom.i23.i.i.i.i.i1208 = zext i32 %.lcssa4086 to i64
  %arrayidx.i24.i.i.i.i.i1209 = getelementptr inbounds nuw i32, ptr %1680, i64 %idxprom.i23.i.i.i.i.i1208
  store atomic i32 %conv.i.i.i.i.i.i1207, ptr %arrayidx.i24.i.i.i.i.i1209 monotonic, align 4
  %1681 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !84
  %arrayidx.i.i5.i.i1210 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %1681, i64 %idxprom.i23.i.i.i.i.i1208, !amdgpu.uniform !42
  %1682 = load i64, ptr %arrayidx.i.i5.i.i1210, align 8, !tbaa !80
  br label %while.cond.i.i.i.i1211, !amdgpu.uniform !42

while.cond.i.i.i.i1211:                           ; preds = %post-loop-memcpy-expansion2386, %Flow3609
  %phi.broken3954 = phi i64 [ 0, %post-loop-memcpy-expansion2386 ], [ %1699, %Flow3609 ]
  %lsr.iv3002 = phi i64 [ %1697, %Flow3609 ], [ -50, %post-loop-memcpy-expansion2386 ]
  %lsr.iv2997 = phi ptr addrspace(4) [ %1696, %Flow3609 ], [ getelementptr (i8, ptr addrspace(4) @.str.36.13, i64 56), %post-loop-memcpy-expansion2386 ]
  %port.sroa.37.0.i.i1212 = phi i32 [ %1695, %Flow3609 ], [ %conv.i.i.i.i.i.i1207, %post-loop-memcpy-expansion2386 ]
  %idx.0.i.i.i.i1213 = phi i64 [ %1694, %Flow3609 ], [ 56, %post-loop-memcpy-expansion2386 ]
  %cmp.i.i.i.i1214 = icmp ult i64 %idx.0.i.i.i.i1213, sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.36.13 to ptr), i64 6) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.36.13 to ptr) to i64))
  %1683 = tail call i64 @llvm.amdgcn.ballot.i64(i1 %cmp.i.i.i.i1214)
  %and.i.i.i.i6.i.i1215 = and i64 %1683, %1682
  %tobool.not.i.i.i.i1216 = icmp ne i64 %and.i.i.i.i6.i.i1215, 0
  %1684 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i22.i.i.i1217 = getelementptr inbounds nuw i32, ptr %1684, i64 %idxprom.i23.i.i.i.i.i1208, !amdgpu.uniform !42
  %1685 = load atomic i32, ptr %arrayidx.i22.i.i.i1217 monotonic, align 4
  %1686 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %tobool.not.i.i.i.i1216)
  %1687 = extractvalue { i1, i64 } %1686, 0
  %1688 = extractvalue { i1, i64 } %1686, 1
  br i1 %1687, label %cond.false.i8.i.i.i.i1218, label %Flow3609

cond.false.i8.i.i.i.i1218:                        ; preds = %while.cond.i.i.i.i1211
  %1689 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %1685)
  %1690 = icmp ne i32 %1689, %port.sroa.37.0.i.i1212
  br i1 %1690, label %while.body.i.i21.i.i.i.i1219.preheader, label %Flow3608, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i1219.preheader:           ; preds = %cond.false.i8.i.i.i.i1218
  br label %while.body.i.i21.i.i.i.i1219, !amdgpu.uniform !42

Flow3608:                                         ; preds = %Flow3606, %cond.false.i8.i.i.i.i1218
  br label %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i1222, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i1219:                     ; preds = %while.body.i.i21.i.i.i.i1219.preheader, %while.body.i.i21.i.i.i.i1219
  tail call void @llvm.amdgcn.s.sleep(i32 2)
  %1691 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i22.i.i.i.i1220 = getelementptr inbounds nuw i32, ptr %1691, i64 %idxprom.i23.i.i.i.i.i1208, !amdgpu.uniform !42
  %1692 = load atomic i32, ptr %arrayidx.i.i.i22.i.i.i.i1220 monotonic, align 4
  %1693 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %1692)
  %cmp.i.not.i.i23.i.i.i.i1221 = icmp eq i32 %1693, %port.sroa.37.0.i.i1212
  br i1 %cmp.i.not.i.i23.i.i.i.i1221, label %Flow3606, label %while.body.i.i21.i.i.i.i1219, !amdgpu.uniform !42

Flow3606:                                         ; preds = %while.body.i.i21.i.i.i.i1219
  br label %Flow3608, !amdgpu.uniform !42

Flow3609:                                         ; preds = %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1223, %while.cond.i.i.i.i1211
  %1694 = phi i64 [ %add.i.i.i.i1226, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1223 ], [ poison, %while.cond.i.i.i.i1211 ]
  %1695 = phi i32 [ %conv.i.i26.i.i.i.i1224, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1223 ], [ poison, %while.cond.i.i.i.i1211 ]
  %1696 = phi ptr addrspace(4) [ %scevgep2998, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1223 ], [ poison, %while.cond.i.i.i.i1211 ]
  %1697 = phi i64 [ %lsr.iv.next3003, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1223 ], [ poison, %while.cond.i.i.i.i1211 ]
  %1698 = phi i1 [ false, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1223 ], [ true, %while.cond.i.i.i.i1211 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %1688)
  %1699 = call i64 @llvm.amdgcn.if.break.i64(i1 %1698, i64 %phi.broken3954)
  %1700 = call i1 @llvm.amdgcn.loop.i64(i64 %1699)
  br i1 %1700, label %cond.false.i.i.i1233, label %while.cond.i.i.i.i1211

_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i1222: ; preds = %Flow3608
  fence acquire
  br i1 %cmp.i.i.i.i1214, label %if.then.i.i.i.i.i.i.i1227, label %Flow3605, !amdgpu.uniform !42

if.then.i.i.i.i.i.i.i1227:                        ; preds = %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i1222
  %sub.i.i.i.i.i.i.i1229 = sub nuw i64 sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.36.13 to ptr), i64 6) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.36.13 to ptr) to i64)), %idx.0.i.i.i.i1213
  %spec.select.i.i.i36.i.i.i.i1230 = tail call i64 @llvm.umin.i64(i64 %sub.i.i.i.i.i.i.i1229, i64 64)
  %1701 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %1702 = and i64 %spec.select.i.i.i36.i.i.i.i1230, 15
  %1703 = sub i64 %spec.select.i.i.i36.i.i.i.i1230, %1702
  %1704 = icmp ne i64 %1703, 0
  br i1 %1704, label %loop-memcpy-expansion2393.preheader, label %Flow3604, !amdgpu.uniform !42

loop-memcpy-expansion2393.preheader:              ; preds = %if.then.i.i.i.i.i.i.i1227
  %arrayidx.i20.i35.i.i.i.i1231 = getelementptr %"struct.rpc::Buffer", ptr %1701, i64 %idxprom.i19.i.i.i.i.i1201
  %arrayidx.i22.i38.i.i.i.i1232 = getelementptr %"struct.rpc::Buffer", ptr %arrayidx.i20.i35.i.i.i.i1231, i64 %sh_prom.i.i.i.i
  br label %loop-memcpy-expansion2393, !amdgpu.uniform !42

Flow3604:                                         ; preds = %Flow3603, %if.then.i.i.i.i.i.i.i1227
  br label %loop-memcpy-residual-header2396, !amdgpu.uniform !42

loop-memcpy-expansion2393:                        ; preds = %loop-memcpy-expansion2393.preheader, %loop-memcpy-expansion2393
  %loop-index2394 = phi i64 [ %1707, %loop-memcpy-expansion2393 ], [ 0, %loop-memcpy-expansion2393.preheader ]
  %scevgep2999 = getelementptr i8, ptr addrspace(4) %lsr.iv2997, i64 %loop-index2394, !amdgpu.uniform !42
  %1705 = load <4 x i32>, ptr addrspace(4) %scevgep2999, align 1
  %1706 = getelementptr i8, ptr %arrayidx.i22.i38.i.i.i.i1232, i64 %loop-index2394
  store <4 x i32> %1705, ptr %1706, align 1
  %1707 = add i64 %loop-index2394, 16
  %1708 = icmp uge i64 %1707, %1703
  br i1 %1708, label %Flow3603, label %loop-memcpy-expansion2393, !amdgpu.uniform !42

Flow3601:                                         ; preds = %Flow3600, %loop-memcpy-residual-header2396
  br label %Flow3605, !amdgpu.uniform !42

loop-memcpy-residual2395:                         ; preds = %loop-memcpy-residual2395.preheader, %loop-memcpy-residual2395
  %residual-loop-index2397 = phi i64 [ %1710, %loop-memcpy-residual2395 ], [ 0, %loop-memcpy-residual2395.preheader ]
  %scevgep3008 = getelementptr i8, ptr addrspace(4) %scevgep3007, i64 %residual-loop-index2397, !amdgpu.uniform !42
  %1709 = load i8, ptr addrspace(4) %scevgep3008, align 1
  %scevgep3006 = getelementptr i8, ptr %scevgep3005, i64 %residual-loop-index2397
  store i8 %1709, ptr %scevgep3006, align 1
  %1710 = add i64 %residual-loop-index2397, 1
  %1711 = icmp uge i64 %1710, %1702
  br i1 %1711, label %Flow3600, label %loop-memcpy-residual2395, !amdgpu.uniform !42

Flow3600:                                         ; preds = %loop-memcpy-residual2395
  br label %Flow3601, !amdgpu.uniform !42

_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1223: ; preds = %Flow3605
  %conv.i.i26.i.i.i.i1224 = xor i32 %port.sroa.37.0.i.i1212, 1
  fence release
  %1712 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %arrayidx.i24.i29.i.i.i.i1225 = getelementptr inbounds nuw i32, ptr %1712, i64 %idxprom.i23.i.i.i.i.i1208
  store atomic i32 %conv.i.i26.i.i.i.i1224, ptr %arrayidx.i24.i29.i.i.i.i1225 monotonic, align 4
  %add.i.i.i.i1226 = add i64 %idx.0.i.i.i.i1213, 64
  %scevgep2998 = getelementptr i8, ptr addrspace(4) %lsr.iv2997, i64 64
  %lsr.iv.next3003 = add i64 %lsr.iv3002, -64
  br label %Flow3609, !amdgpu.uniform !42

cond.false.i.i.i1233:                             ; preds = %Flow3609
  %.lcssa4080 = phi i64 [ %1699, %Flow3609 ]
  %port.sroa.37.0.i.i1212.lcssa = phi i32 [ %port.sroa.37.0.i.i1212, %Flow3609 ]
  %.lcssa4079 = phi i32 [ %1685, %Flow3609 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa4080)
  %1713 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %.lcssa4079)
  %1714 = icmp ne i32 %1713, %port.sroa.37.0.i.i1212.lcssa
  %1715 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %1714)
  %1716 = extractvalue { i1, i64 } %1715, 0
  %1717 = extractvalue { i1, i64 } %1715, 1
  br i1 %1716, label %while.body.i.i.i.i1234.preheader, label %Flow3598

while.body.i.i.i.i1234.preheader:                 ; preds = %cond.false.i.i.i1233
  br label %while.body.i.i.i.i1234, !amdgpu.uniform !42

Flow3598:                                         ; preds = %Flow3596, %cond.false.i.i.i1233
  call void @llvm.amdgcn.end.cf.i64(i64 %1717)
  br label %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit1255, !amdgpu.uniform !42

while.body.i.i.i.i1234:                           ; preds = %while.body.i.i.i.i1234.preheader, %while.body.i.i.i.i1234
  %phi.broken3955 = phi i64 [ 0, %while.body.i.i.i.i1234.preheader ], [ %1721, %while.body.i.i.i.i1234 ]
  tail call void @llvm.amdgcn.s.sleep(i32 2)
  %1718 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i8.i.i1235 = getelementptr inbounds nuw i32, ptr %1718, i64 %idxprom.i23.i.i.i.i.i1208, !amdgpu.uniform !42
  %1719 = load atomic i32, ptr %arrayidx.i.i.i8.i.i1235 monotonic, align 4
  %1720 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %1719)
  %cmp.i.not.i.i.i.i1236 = icmp eq i32 %1720, %port.sroa.37.0.i.i1212.lcssa
  %1721 = call i64 @llvm.amdgcn.if.break.i64(i1 %cmp.i.not.i.i.i.i1236, i64 %phi.broken3955)
  %1722 = call i1 @llvm.amdgcn.loop.i64(i64 %1721)
  br i1 %1722, label %Flow3596, label %while.body.i.i.i.i1234

Flow3596:                                         ; preds = %while.body.i.i.i.i1234
  %.lcssa4078 = phi i64 [ %1721, %while.body.i.i.i.i1234 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa4078)
  br label %Flow3598, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit1255: ; preds = %Flow3598
  fence acquire
  tail call void @llvm.amdgcn.wave.barrier()
  fence syncscope("agent") release
  %1723 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %.lcssa4085, i1 true)
  %iszero.i.i.i.i.i13.i.i1238 = icmp ne i64 %.lcssa4085, 0
  %cmp2.i.i.i.i14.i.i1239 = icmp eq i64 %1723, %sh_prom.i.i.i.i
  %cmp.i.i.i.i15.i.i1240 = select i1 %iszero.i.i.i.i.i13.i.i1238, i1 %cmp2.i.i.i.i14.i.i1239, i1 false
  %1724 = and i32 %.lcssa4086, 31
  %conv4.i.i.i19.i.i1244 = zext i1 %cmp.i.i.i.i15.i.i1240 to i32
  %shl.i.i.i20.i.i1245 = shl nuw i32 %conv4.i.i.i19.i.i1244, %1724
  %xor.i.i.i21.i.i1246 = xor i32 %shl.i.i.i20.i.i1245, -1
  %1725 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2588, !amdgpu.uniform !42

1726:                                             ; preds = %ComputeEnd2589
  %div8.i.i.i16.i.i1241 = lshr i32 %.lcssa4086, 5
  %idxprom.i.i.i17.i.i1242 = zext nneg i32 %div8.i.i.i16.i.i1241 to i64
  %1727 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i17.i.i1242
  %sunkaddr3325 = getelementptr inbounds i8, ptr addrspace(1) %1727, i64 40
  %1728 = atomicrmw and ptr addrspace(1) %sunkaddr3325, i32 %.lcssa4077 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %1729, !amdgpu.uniform !42

1729:                                             ; preds = %1726, %ComputeEnd2589
  call void @llvm.amdgcn.end.cf.i64(i64 %3669)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %for.cond.i.i.i1257, !amdgpu.uniform !42

for.cond.i.i.i1257:                               ; preds = %1729, %Flow3595
  %1730 = phi i32 [ %1768, %Flow3595 ], [ poison, %1729 ]
  %1731 = phi i64 [ %1769, %Flow3595 ], [ poison, %1729 ]
  %1732 = phi i32 [ %1770, %Flow3595 ], [ poison, %1729 ]
  %retval.sroa.6.0.i.i.i1258 = phi i32 [ %1773, %Flow3595 ], [ undef, %1729 ]
  %retval.sroa.2.0.i.i.i1260 = phi i64 [ %1772, %Flow3595 ], [ undef, %1729 ]
  %retval.sroa.8.0.i.i.i1261 = phi i32 [ %1771, %Flow3595 ], [ undef, %1729 ]
  %index.0.i.i.i1262 = phi i32 [ %1775, %Flow3595 ], [ 0, %1729 ]
  %1733 = load i32, ptr addrspace(1) @__llvm_rpc_client, align 8, !tbaa !68
  %cmp.not.i.i.i1263 = icmp ult i32 %index.0.i.i.i1262, %1733
  %spec.store.select.i.i.i1264 = select i1 %cmp.not.i.i.i1263, i32 %index.0.i.i.i1262, i32 0
  %1734 = tail call noundef i64 @llvm.amdgcn.ballot.i64(i1 true)
  %1735 = and i32 %spec.store.select.i.i.i1264, 31
  %1736 = lshr i64 %1734, %sh_prom.i.i.i.i
  %1737 = trunc i64 %1736 to i32
  %conv4.i.i.i.i.i1268 = and i32 %1737, 1
  %shl.i.i.i.i.i1269 = shl nuw i32 %conv4.i.i.i.i.i1268, %1735
  %1738 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2592, !amdgpu.uniform !42

1739:                                             ; preds = %ComputeEnd2593
  %sunkaddr3326 = getelementptr inbounds i8, ptr addrspace(1) %3679, i64 40
  %1740 = atomicrmw or ptr addrspace(1) %sunkaddr3326, i32 %.lcssa4071 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %1741, !amdgpu.uniform !42

1741:                                             ; preds = %1739, %ComputeEnd2593
  %1742 = phi i32 [ %1740, %1739 ], [ poison, %ComputeEnd2593 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %3688)
  %1743 = call i32 @llvm.amdgcn.readfirstlane.i32(i32 %1742)
  %1744 = or i32 %1743, %.lcssa4072
  %shl5.i.i.i.i.i1270 = shl nuw i32 1, %1735
  %and.i.i.i.i.i1271 = and i32 %shl5.i.i.i.i.i1270, %1744
  %tobool3.i.i.i.i1272 = icmp ne i32 %and.i.i.i.i.i1271, 0
  %1745 = tail call i64 @llvm.amdgcn.ballot.i64(i1 %tobool3.i.i.i.i1272)
  %and.i.i.i.i.i.i1273 = and i64 %1745, %1734
  %cmp.i.not.i.i.i1274 = icmp ne i64 %1734, %and.i.i.i.i.i.i1273
  br i1 %cmp.i.not.i.i.i1274, label %if.then.i.i.i.i1275, label %Flow3594, !amdgpu.uniform !42

if.then.i.i.i.i1275:                              ; preds = %1741
  fence syncscope("agent") acquire
  %1746 = load <2 x i64>, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !75
  %1747 = extractelement <2 x i64> %1746, i32 0
  %1748 = inttoptr i64 %1747 to ptr
  %1749 = extractelement <2 x i64> %1746, i32 1
  %1750 = inttoptr i64 %1749 to ptr
  %idxprom.i.i.i.i1276 = zext i32 %spec.store.select.i.i.i1264 to i64
  %arrayidx.i.i.i.i1277 = getelementptr inbounds nuw i32, ptr %1748, i64 %idxprom.i.i.i.i1276, !amdgpu.uniform !42
  %1751 = load atomic i32, ptr %arrayidx.i.i.i.i1277 monotonic, align 4
  %1752 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %1751)
  %arrayidx.i45.i.i.i1278 = getelementptr inbounds nuw i32, ptr %1750, i64 %idxprom.i.i.i.i1276, !amdgpu.uniform !42
  %1753 = load atomic i32, ptr %arrayidx.i45.i.i.i1278 monotonic, align 4
  %1754 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %1753)
  %cmp.i46.not.i.i.i1279 = icmp ne i32 %1752, %1754
  br i1 %cmp.i46.not.i.i.i1279, label %if.then12.i.i.i1280, label %Flow3592, !amdgpu.uniform !42

if.then12.i.i.i1280:                              ; preds = %if.then.i.i.i.i1275
  fence syncscope("agent") release
  %1755 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %1734, i1 true)
  %iszero.i.i.i.i.i.i.i1281 = icmp ne i64 %1734, 0
  %cmp2.i.i.i.i.i.i1282 = icmp eq i64 %1755, %sh_prom.i.i.i.i
  %cmp.i.i.i.i.i.i1283 = select i1 %iszero.i.i.i.i.i.i.i1281, i1 %cmp2.i.i.i.i.i.i1282, i1 false
  %conv4.i.i51.i.i.i1284 = zext i1 %cmp.i.i.i.i.i.i1283 to i32
  %shl.i.i52.i.i.i1285 = shl nuw i32 %conv4.i.i51.i.i.i1284, %1735
  %xor.i.i.i.i.i1286 = xor i32 %shl.i.i52.i.i.i1285, -1
  %1756 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2597, !amdgpu.uniform !42

1757:                                             ; preds = %ComputeEnd2598
  %sunkaddr3327 = getelementptr inbounds i8, ptr addrspace(1) %3679, i64 40
  %1758 = atomicrmw and ptr addrspace(1) %sunkaddr3327, i32 %.lcssa4073 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %1759, !amdgpu.uniform !42

1759:                                             ; preds = %1757, %ComputeEnd2598
  call void @llvm.amdgcn.end.cf.i64(i64 %3706)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %Flow3592, !amdgpu.uniform !42

if.end14.i.i.i1346:                               ; preds = %Flow3592
  %1760 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %1734, i1 true)
  %iszero.i.i.i.i.i.i1347 = icmp ne i64 %1734, 0
  %cmp2.i.i.i.i.i1348 = icmp eq i64 %1760, %sh_prom.i.i.i.i
  %cmp.i.i.i.i.i1349 = select i1 %iszero.i.i.i.i.i.i1347, i1 %cmp2.i.i.i.i.i1348, i1 false
  %1761 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.i.i.i.i.i1349)
  %1762 = extractvalue { i1, i64 } %1761, 0
  %1763 = extractvalue { i1, i64 } %1761, 1
  br i1 %1762, label %if.then16.i.i.i1351, label %if.end22.i.i.i1350

if.then16.i.i.i1351:                              ; preds = %if.end14.i.i.i1346
  %1764 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !76
  %opcode.i.i.i1352 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %1764, i64 %idxprom.i.i.i.i1276, i32 1
  store i32 1660944387, ptr %opcode.i.i.i1352, align 8, !tbaa !77
  %arrayidx21.i.i.i1353 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %1764, i64 %idxprom.i.i.i.i1276
  store i64 %1734, ptr %arrayidx21.i.i.i1353, align 8, !tbaa !80
  br label %if.end22.i.i.i1350, !amdgpu.uniform !42

Flow3593:                                         ; preds = %if.end22.i.i.i1350, %Flow3592
  %1765 = phi i32 [ %1752, %if.end22.i.i.i1350 ], [ %retval.sroa.8.0.i.i.i1261, %Flow3592 ]
  %1766 = phi i64 [ %1734, %if.end22.i.i.i1350 ], [ %retval.sroa.2.0.i.i.i1260, %Flow3592 ]
  %1767 = phi i32 [ %spec.store.select.i.i.i1264, %if.end22.i.i.i1350 ], [ %retval.sroa.6.0.i.i.i1258, %Flow3592 ]
  br label %cleanup26.i.i.i1287, !amdgpu.uniform !42

if.end22.i.i.i1350:                               ; preds = %if.then16.i.i.i1351, %if.end14.i.i.i1346
  call void @llvm.amdgcn.end.cf.i64(i64 %1763)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %Flow3593, !amdgpu.uniform !42

Flow3594:                                         ; preds = %cleanup26.i.i.i1287, %1741
  %1768 = phi i32 [ %1767, %cleanup26.i.i.i1287 ], [ %1730, %1741 ]
  %1769 = phi i64 [ %1766, %cleanup26.i.i.i1287 ], [ %1731, %1741 ]
  %1770 = phi i32 [ %1765, %cleanup26.i.i.i1287 ], [ %1732, %1741 ]
  %1771 = phi i32 [ %1765, %cleanup26.i.i.i1287 ], [ %retval.sroa.8.0.i.i.i1261, %1741 ]
  %1772 = phi i64 [ %1766, %cleanup26.i.i.i1287 ], [ %retval.sroa.2.0.i.i.i1260, %1741 ]
  %1773 = phi i32 [ %1767, %cleanup26.i.i.i1287 ], [ %retval.sroa.6.0.i.i.i1258, %1741 ]
  %1774 = phi i1 [ %cmp.i46.not.i.i.i1279, %cleanup26.i.i.i1287 ], [ true, %1741 ]
  br i1 %1774, label %for.inc.i.i.i1292, label %Flow3595, !amdgpu.uniform !42

cleanup26.i.i.i1287:                              ; preds = %Flow3593
  br label %Flow3594, !amdgpu.uniform !42

for.inc.i.i.i1292:                                ; preds = %Flow3594
  %inc.i.i.i1297 = add i32 %spec.store.select.i.i.i1264, 1
  br label %Flow3595, !amdgpu.uniform !42

Flow3595:                                         ; preds = %for.inc.i.i.i1292, %Flow3594
  %1775 = phi i32 [ %inc.i.i.i1297, %for.inc.i.i.i1292 ], [ poison, %Flow3594 ]
  %1776 = phi i1 [ false, %for.inc.i.i.i1292 ], [ true, %Flow3594 ]
  br i1 %1776, label %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1298, label %for.cond.i.i.i1257, !amdgpu.uniform !42

_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1298: ; preds = %Flow3595
  %.lcssa4076 = phi i32 [ %1768, %Flow3595 ]
  %.lcssa4075 = phi i64 [ %1769, %Flow3595 ]
  %.lcssa4074 = phi i32 [ %1770, %Flow3595 ]
  fence acquire
  %1777 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %mul.i.i.i.i.i.i1299 = shl i32 %.lcssa4076, 6
  %idxprom.i19.i.i.i.i.i1300 = zext i32 %mul.i.i.i.i.i.i1299 to i64
  %arrayidx.i20.i.i.i.i.i1301 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %1777, i64 %idxprom.i19.i.i.i.i.i1300
  %arrayidx.i22.i.i.i.i.i1302 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %arrayidx.i20.i.i.i.i.i1301, i64 %sh_prom.i.i.i.i
  store i64 sub (i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) getelementptr inbounds nuw (i8, ptr addrspace(4) @.str.9, i64 8) to ptr) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.9 to ptr) to i64)), ptr %arrayidx.i22.i.i.i.i.i1302, align 8, !tbaa !82
  %spec.select.i.i.i.i.i.i.i1303 = tail call i64 @llvm.umin.i64(i64 sub (i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) getelementptr inbounds nuw (i8, ptr addrspace(4) @.str.9, i64 8) to ptr) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.9 to ptr) to i64)), i64 56)
  %1778 = and i64 %spec.select.i.i.i.i.i.i.i1303, 15
  %1779 = sub i64 %spec.select.i.i.i.i.i.i.i1303, %1778
  %1780 = icmp ne i64 %1779, 0
  br i1 %1780, label %loop-memcpy-expansion2399.preheader, label %Flow3591, !amdgpu.uniform !42

loop-memcpy-expansion2399.preheader:              ; preds = %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1298
  %1781 = add nuw nsw i64 %75, 8
  %scevgep2994 = getelementptr i8, ptr %1777, i64 %1781
  %1782 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i1300, 6
  %scevgep2995 = getelementptr i8, ptr %scevgep2994, i64 %1782
  br label %loop-memcpy-expansion2399, !amdgpu.uniform !42

Flow3591:                                         ; preds = %Flow3590, %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1298
  br label %loop-memcpy-residual-header2402, !amdgpu.uniform !42

loop-memcpy-expansion2399:                        ; preds = %loop-memcpy-expansion2399.preheader, %loop-memcpy-expansion2399
  %loop-index2400 = phi i64 [ %1785, %loop-memcpy-expansion2399 ], [ 0, %loop-memcpy-expansion2399.preheader ]
  %1783 = getelementptr i8, ptr addrspace(4) @.str.9, i64 %loop-index2400, !amdgpu.uniform !42
  %1784 = load <4 x i32>, ptr addrspace(4) %1783, align 1
  %scevgep2996 = getelementptr i8, ptr %scevgep2995, i64 %loop-index2400
  store <4 x i32> %1784, ptr %scevgep2996, align 1
  %1785 = add i64 %loop-index2400, 16
  %1786 = icmp uge i64 %1785, %1779
  br i1 %1786, label %Flow3590, label %loop-memcpy-expansion2399, !amdgpu.uniform !42

Flow3588:                                         ; preds = %Flow3585, %Flow3587
  %1787 = phi i64 [ %2953, %Flow3585 ], [ %2947, %Flow3587 ]
  br label %post-loop-memcpy-expansion2398, !amdgpu.uniform !42

loop-memcpy-residual2401:                         ; preds = %loop-memcpy-residual2401.preheader, %loop-memcpy-residual2401
  %residual-loop-index2403 = phi i64 [ %1789, %loop-memcpy-residual2401 ], [ 0, %loop-memcpy-residual2401.preheader ]
  %scevgep2993 = getelementptr i8, ptr addrspace(4) %scevgep2992, i64 %residual-loop-index2403, !amdgpu.uniform !42
  %1788 = load i8, ptr addrspace(4) %scevgep2993, align 1
  %scevgep2991 = getelementptr i8, ptr %scevgep2990, i64 %residual-loop-index2403
  store i8 %1788, ptr %scevgep2991, align 1
  %1789 = add i64 %residual-loop-index2403, 1
  %1790 = icmp uge i64 %1789, %1778
  br i1 %1790, label %Flow3585, label %loop-memcpy-residual2401, !amdgpu.uniform !42

Flow3585:                                         ; preds = %loop-memcpy-residual2401
  br label %Flow3588, !amdgpu.uniform !42

post-loop-memcpy-expansion2398:                   ; preds = %Flow3588
  %tobool.not.i.i.i.i.i.i1305 = icmp eq i32 %.lcssa4074, 0
  %conv.i.i.i.i.i.i1306 = zext i1 %tobool.not.i.i.i.i.i.i1305 to i32
  fence release
  %1791 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %idxprom.i23.i.i.i.i.i1307 = zext i32 %.lcssa4076 to i64
  %arrayidx.i24.i.i.i.i.i1308 = getelementptr inbounds nuw i32, ptr %1791, i64 %idxprom.i23.i.i.i.i.i1307
  store atomic i32 %conv.i.i.i.i.i.i1306, ptr %arrayidx.i24.i.i.i.i.i1308 monotonic, align 4
  %1792 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !84
  %arrayidx.i.i5.i.i1309 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %1792, i64 %idxprom.i23.i.i.i.i.i1307, !amdgpu.uniform !42
  %1793 = load i64, ptr %arrayidx.i.i5.i.i1309, align 8, !tbaa !80
  br label %while.cond.i.i.i.i1310, !amdgpu.uniform !42

while.cond.i.i.i.i1310:                           ; preds = %post-loop-memcpy-expansion2398, %Flow3583
  %phi.broken3956 = phi i64 [ 0, %post-loop-memcpy-expansion2398 ], [ %1810, %Flow3583 ]
  %lsr.iv2982 = phi i64 [ %1808, %Flow3583 ], [ sub (i64 add (i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) getelementptr inbounds nuw (i8, ptr addrspace(4) @.str.9, i64 8) to ptr) to i64), i64 -56), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.9 to ptr) to i64)), %post-loop-memcpy-expansion2398 ]
  %lsr.iv2977 = phi ptr addrspace(4) [ %1807, %Flow3583 ], [ getelementptr (i8, ptr addrspace(4) @.str.9, i64 56), %post-loop-memcpy-expansion2398 ]
  %port.sroa.37.0.i.i1311 = phi i32 [ %1806, %Flow3583 ], [ %conv.i.i.i.i.i.i1306, %post-loop-memcpy-expansion2398 ]
  %idx.0.i.i.i.i1312 = phi i64 [ %1805, %Flow3583 ], [ 56, %post-loop-memcpy-expansion2398 ]
  %cmp.i.i.i.i1313 = icmp ult i64 %idx.0.i.i.i.i1312, sub (i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) getelementptr inbounds nuw (i8, ptr addrspace(4) @.str.9, i64 8) to ptr) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.9 to ptr) to i64))
  %1794 = tail call i64 @llvm.amdgcn.ballot.i64(i1 %cmp.i.i.i.i1313)
  %and.i.i.i.i6.i.i1314 = and i64 %1794, %1793
  %tobool.not.i.i.i.i1315 = icmp ne i64 %and.i.i.i.i6.i.i1314, 0
  %1795 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i22.i.i.i1316 = getelementptr inbounds nuw i32, ptr %1795, i64 %idxprom.i23.i.i.i.i.i1307, !amdgpu.uniform !42
  %1796 = load atomic i32, ptr %arrayidx.i22.i.i.i1316 monotonic, align 4
  %1797 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %tobool.not.i.i.i.i1315)
  %1798 = extractvalue { i1, i64 } %1797, 0
  %1799 = extractvalue { i1, i64 } %1797, 1
  br i1 %1798, label %cond.false.i8.i.i.i.i1317, label %Flow3583

cond.false.i8.i.i.i.i1317:                        ; preds = %while.cond.i.i.i.i1310
  %1800 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %1796)
  %1801 = icmp ne i32 %1800, %port.sroa.37.0.i.i1311
  br i1 %1801, label %while.body.i.i21.i.i.i.i1318.preheader, label %Flow3582, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i1318.preheader:           ; preds = %cond.false.i8.i.i.i.i1317
  br label %while.body.i.i21.i.i.i.i1318, !amdgpu.uniform !42

Flow3582:                                         ; preds = %Flow3580, %cond.false.i8.i.i.i.i1317
  br label %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i1321, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i1318:                     ; preds = %while.body.i.i21.i.i.i.i1318.preheader, %while.body.i.i21.i.i.i.i1318
  tail call void @llvm.amdgcn.s.sleep(i32 2)
  %1802 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i22.i.i.i.i1319 = getelementptr inbounds nuw i32, ptr %1802, i64 %idxprom.i23.i.i.i.i.i1307, !amdgpu.uniform !42
  %1803 = load atomic i32, ptr %arrayidx.i.i.i22.i.i.i.i1319 monotonic, align 4
  %1804 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %1803)
  %cmp.i.not.i.i23.i.i.i.i1320 = icmp eq i32 %1804, %port.sroa.37.0.i.i1311
  br i1 %cmp.i.not.i.i23.i.i.i.i1320, label %Flow3580, label %while.body.i.i21.i.i.i.i1318, !amdgpu.uniform !42

Flow3580:                                         ; preds = %while.body.i.i21.i.i.i.i1318
  br label %Flow3582, !amdgpu.uniform !42

Flow3583:                                         ; preds = %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1322, %while.cond.i.i.i.i1310
  %1805 = phi i64 [ %add.i.i.i.i1325, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1322 ], [ poison, %while.cond.i.i.i.i1310 ]
  %1806 = phi i32 [ %conv.i.i26.i.i.i.i1323, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1322 ], [ poison, %while.cond.i.i.i.i1310 ]
  %1807 = phi ptr addrspace(4) [ %scevgep2978, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1322 ], [ poison, %while.cond.i.i.i.i1310 ]
  %1808 = phi i64 [ %lsr.iv.next2983, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1322 ], [ poison, %while.cond.i.i.i.i1310 ]
  %1809 = phi i1 [ false, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1322 ], [ true, %while.cond.i.i.i.i1310 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %1799)
  %1810 = call i64 @llvm.amdgcn.if.break.i64(i1 %1809, i64 %phi.broken3956)
  %1811 = call i1 @llvm.amdgcn.loop.i64(i64 %1810)
  br i1 %1811, label %cond.false.i.i.i1332, label %while.cond.i.i.i.i1310

_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i1321: ; preds = %Flow3582
  fence acquire
  br i1 %cmp.i.i.i.i1313, label %if.then.i.i.i.i.i.i.i1326, label %Flow3579, !amdgpu.uniform !42

if.then.i.i.i.i.i.i.i1326:                        ; preds = %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i1321
  %sub.i.i.i.i.i.i.i1328 = sub nuw i64 sub (i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) getelementptr inbounds nuw (i8, ptr addrspace(4) @.str.9, i64 8) to ptr) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.9 to ptr) to i64)), %idx.0.i.i.i.i1312
  %spec.select.i.i.i36.i.i.i.i1329 = tail call i64 @llvm.umin.i64(i64 %sub.i.i.i.i.i.i.i1328, i64 64)
  %1812 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %1813 = and i64 %spec.select.i.i.i36.i.i.i.i1329, 15
  %1814 = sub i64 %spec.select.i.i.i36.i.i.i.i1329, %1813
  %1815 = icmp ne i64 %1814, 0
  br i1 %1815, label %loop-memcpy-expansion2405.preheader, label %Flow3578, !amdgpu.uniform !42

loop-memcpy-expansion2405.preheader:              ; preds = %if.then.i.i.i.i.i.i.i1326
  %arrayidx.i20.i35.i.i.i.i1330 = getelementptr %"struct.rpc::Buffer", ptr %1812, i64 %idxprom.i19.i.i.i.i.i1300
  %arrayidx.i22.i38.i.i.i.i1331 = getelementptr %"struct.rpc::Buffer", ptr %arrayidx.i20.i35.i.i.i.i1330, i64 %sh_prom.i.i.i.i
  br label %loop-memcpy-expansion2405, !amdgpu.uniform !42

Flow3578:                                         ; preds = %Flow3577, %if.then.i.i.i.i.i.i.i1326
  br label %loop-memcpy-residual-header2408, !amdgpu.uniform !42

loop-memcpy-expansion2405:                        ; preds = %loop-memcpy-expansion2405.preheader, %loop-memcpy-expansion2405
  %loop-index2406 = phi i64 [ %1818, %loop-memcpy-expansion2405 ], [ 0, %loop-memcpy-expansion2405.preheader ]
  %scevgep2979 = getelementptr i8, ptr addrspace(4) %lsr.iv2977, i64 %loop-index2406, !amdgpu.uniform !42
  %1816 = load <4 x i32>, ptr addrspace(4) %scevgep2979, align 1
  %1817 = getelementptr i8, ptr %arrayidx.i22.i38.i.i.i.i1331, i64 %loop-index2406
  store <4 x i32> %1816, ptr %1817, align 1
  %1818 = add i64 %loop-index2406, 16
  %1819 = icmp uge i64 %1818, %1814
  br i1 %1819, label %Flow3577, label %loop-memcpy-expansion2405, !amdgpu.uniform !42

Flow3575:                                         ; preds = %Flow3574, %loop-memcpy-residual-header2408
  br label %Flow3579, !amdgpu.uniform !42

loop-memcpy-residual2407:                         ; preds = %loop-memcpy-residual2407.preheader, %loop-memcpy-residual2407
  %residual-loop-index2409 = phi i64 [ %1821, %loop-memcpy-residual2407 ], [ 0, %loop-memcpy-residual2407.preheader ]
  %scevgep2988 = getelementptr i8, ptr addrspace(4) %scevgep2987, i64 %residual-loop-index2409, !amdgpu.uniform !42
  %1820 = load i8, ptr addrspace(4) %scevgep2988, align 1
  %scevgep2986 = getelementptr i8, ptr %scevgep2985, i64 %residual-loop-index2409
  store i8 %1820, ptr %scevgep2986, align 1
  %1821 = add i64 %residual-loop-index2409, 1
  %1822 = icmp uge i64 %1821, %1813
  br i1 %1822, label %Flow3574, label %loop-memcpy-residual2407, !amdgpu.uniform !42

Flow3574:                                         ; preds = %loop-memcpy-residual2407
  br label %Flow3575, !amdgpu.uniform !42

_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1322: ; preds = %Flow3579
  %conv.i.i26.i.i.i.i1323 = xor i32 %port.sroa.37.0.i.i1311, 1
  fence release
  %1823 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %arrayidx.i24.i29.i.i.i.i1324 = getelementptr inbounds nuw i32, ptr %1823, i64 %idxprom.i23.i.i.i.i.i1307
  store atomic i32 %conv.i.i26.i.i.i.i1323, ptr %arrayidx.i24.i29.i.i.i.i1324 monotonic, align 4
  %add.i.i.i.i1325 = add i64 %idx.0.i.i.i.i1312, 64
  %scevgep2978 = getelementptr i8, ptr addrspace(4) %lsr.iv2977, i64 64
  %lsr.iv.next2983 = add i64 %lsr.iv2982, -64
  br label %Flow3583, !amdgpu.uniform !42

cond.false.i.i.i1332:                             ; preds = %Flow3583
  %.lcssa4070 = phi i64 [ %1810, %Flow3583 ]
  %port.sroa.37.0.i.i1311.lcssa = phi i32 [ %port.sroa.37.0.i.i1311, %Flow3583 ]
  %.lcssa4069 = phi i32 [ %1796, %Flow3583 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa4070)
  %1824 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %.lcssa4069)
  %1825 = icmp ne i32 %1824, %port.sroa.37.0.i.i1311.lcssa
  %1826 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %1825)
  %1827 = extractvalue { i1, i64 } %1826, 0
  %1828 = extractvalue { i1, i64 } %1826, 1
  br i1 %1827, label %while.body.i.i.i.i1333.preheader, label %Flow3572

while.body.i.i.i.i1333.preheader:                 ; preds = %cond.false.i.i.i1332
  br label %while.body.i.i.i.i1333, !amdgpu.uniform !42

Flow3572:                                         ; preds = %Flow3570, %cond.false.i.i.i1332
  call void @llvm.amdgcn.end.cf.i64(i64 %1828)
  br label %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsINS_3cpp11string_viewEEERS1_T_.exit, !amdgpu.uniform !42

while.body.i.i.i.i1333:                           ; preds = %while.body.i.i.i.i1333.preheader, %while.body.i.i.i.i1333
  %phi.broken3957 = phi i64 [ 0, %while.body.i.i.i.i1333.preheader ], [ %1832, %while.body.i.i.i.i1333 ]
  tail call void @llvm.amdgcn.s.sleep(i32 2)
  %1829 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i8.i.i1334 = getelementptr inbounds nuw i32, ptr %1829, i64 %idxprom.i23.i.i.i.i.i1307, !amdgpu.uniform !42
  %1830 = load atomic i32, ptr %arrayidx.i.i.i8.i.i1334 monotonic, align 4
  %1831 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %1830)
  %cmp.i.not.i.i.i.i1335 = icmp eq i32 %1831, %port.sroa.37.0.i.i1311.lcssa
  %1832 = call i64 @llvm.amdgcn.if.break.i64(i1 %cmp.i.not.i.i.i.i1335, i64 %phi.broken3957)
  %1833 = call i1 @llvm.amdgcn.loop.i64(i64 %1832)
  br i1 %1833, label %Flow3570, label %while.body.i.i.i.i1333

Flow3570:                                         ; preds = %while.body.i.i.i.i1333
  %.lcssa4068 = phi i64 [ %1832, %while.body.i.i.i.i1333 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa4068)
  br label %Flow3572, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsINS_3cpp11string_viewEEERS1_T_.exit: ; preds = %Flow3572
  fence acquire
  tail call void @llvm.amdgcn.wave.barrier()
  fence syncscope("agent") release
  %1834 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %.lcssa4075, i1 true)
  %iszero.i.i.i.i.i13.i.i1337 = icmp ne i64 %.lcssa4075, 0
  %cmp2.i.i.i.i14.i.i1338 = icmp eq i64 %1834, %sh_prom.i.i.i.i
  %cmp.i.i.i.i15.i.i1339 = select i1 %iszero.i.i.i.i.i13.i.i1337, i1 %cmp2.i.i.i.i14.i.i1338, i1 false
  %1835 = and i32 %.lcssa4076, 31
  %conv4.i.i.i19.i.i1343 = zext i1 %cmp.i.i.i.i15.i.i1339 to i32
  %shl.i.i.i20.i.i1344 = shl nuw i32 %conv4.i.i.i19.i.i1343, %1835
  %xor.i.i.i21.i.i1345 = xor i32 %shl.i.i.i20.i.i1344, -1
  %1836 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2601, !amdgpu.uniform !42

1837:                                             ; preds = %ComputeEnd2602
  %div8.i.i.i16.i.i1340 = lshr i32 %.lcssa4076, 5
  %idxprom.i.i.i17.i.i1341 = zext nneg i32 %div8.i.i.i16.i.i1340 to i64
  %1838 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i17.i.i1341
  %sunkaddr3328 = getelementptr inbounds i8, ptr addrspace(1) %1838, i64 40
  %1839 = atomicrmw and ptr addrspace(1) %sunkaddr3328, i32 %.lcssa4067 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %1840, !amdgpu.uniform !42

1840:                                             ; preds = %1837, %ComputeEnd2602
  call void @llvm.amdgcn.end.cf.i64(i64 %3723)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %for.cond.i.i1.i1362, !amdgpu.uniform !42

for.cond.i.i1.i1362:                              ; preds = %1840, %Flow3569
  %1841 = phi i32 [ %1879, %Flow3569 ], [ poison, %1840 ]
  %1842 = phi i64 [ %1880, %Flow3569 ], [ poison, %1840 ]
  %1843 = phi i32 [ %1881, %Flow3569 ], [ poison, %1840 ]
  %retval.sroa.6.0.i.i.i1363 = phi i32 [ %1884, %Flow3569 ], [ undef, %1840 ]
  %retval.sroa.2.0.i.i.i1365 = phi i64 [ %1883, %Flow3569 ], [ undef, %1840 ]
  %retval.sroa.8.0.i.i.i1366 = phi i32 [ %1882, %Flow3569 ], [ undef, %1840 ]
  %index.0.i.i.i1367 = phi i32 [ %1886, %Flow3569 ], [ 0, %1840 ]
  %1844 = load i32, ptr addrspace(1) @__llvm_rpc_client, align 8, !tbaa !68
  %cmp.not.i.i.i1368 = icmp ult i32 %index.0.i.i.i1367, %1844
  %spec.store.select.i.i.i1369 = select i1 %cmp.not.i.i.i1368, i32 %index.0.i.i.i1367, i32 0
  %1845 = tail call noundef i64 @llvm.amdgcn.ballot.i64(i1 true)
  %1846 = and i32 %spec.store.select.i.i.i1369, 31
  %1847 = lshr i64 %1845, %sh_prom.i.i.i.i
  %1848 = trunc i64 %1847 to i32
  %conv4.i.i.i.i.i1373 = and i32 %1848, 1
  %shl.i.i.i.i.i1374 = shl nuw i32 %conv4.i.i.i.i.i1373, %1846
  %1849 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2605, !amdgpu.uniform !42

1850:                                             ; preds = %ComputeEnd2606
  %sunkaddr3329 = getelementptr inbounds i8, ptr addrspace(1) %3733, i64 40
  %1851 = atomicrmw or ptr addrspace(1) %sunkaddr3329, i32 %.lcssa4061 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %1852, !amdgpu.uniform !42

1852:                                             ; preds = %1850, %ComputeEnd2606
  %1853 = phi i32 [ %1851, %1850 ], [ poison, %ComputeEnd2606 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %3742)
  %1854 = call i32 @llvm.amdgcn.readfirstlane.i32(i32 %1853)
  %1855 = or i32 %1854, %.lcssa4062
  %shl5.i.i.i.i.i1375 = shl nuw i32 1, %1846
  %and.i.i.i.i.i1376 = and i32 %shl5.i.i.i.i.i1375, %1855
  %tobool3.i.i.i.i1377 = icmp ne i32 %and.i.i.i.i.i1376, 0
  %1856 = tail call i64 @llvm.amdgcn.ballot.i64(i1 %tobool3.i.i.i.i1377)
  %and.i.i.i.i.i.i1378 = and i64 %1856, %1845
  %cmp.i.not.i.i.i1379 = icmp ne i64 %1845, %and.i.i.i.i.i.i1378
  br i1 %cmp.i.not.i.i.i1379, label %if.then.i.i.i.i1380, label %Flow3568, !amdgpu.uniform !42

if.then.i.i.i.i1380:                              ; preds = %1852
  fence syncscope("agent") acquire
  %1857 = load <2 x i64>, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !75
  %1858 = extractelement <2 x i64> %1857, i32 0
  %1859 = inttoptr i64 %1858 to ptr
  %1860 = extractelement <2 x i64> %1857, i32 1
  %1861 = inttoptr i64 %1860 to ptr
  %idxprom.i.i.i.i1381 = zext i32 %spec.store.select.i.i.i1369 to i64
  %arrayidx.i.i.i.i1382 = getelementptr inbounds nuw i32, ptr %1859, i64 %idxprom.i.i.i.i1381, !amdgpu.uniform !42
  %1862 = load atomic i32, ptr %arrayidx.i.i.i.i1382 monotonic, align 4
  %1863 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %1862)
  %arrayidx.i45.i.i.i1383 = getelementptr inbounds nuw i32, ptr %1861, i64 %idxprom.i.i.i.i1381, !amdgpu.uniform !42
  %1864 = load atomic i32, ptr %arrayidx.i45.i.i.i1383 monotonic, align 4
  %1865 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %1864)
  %cmp.i46.not.i.i.i1384 = icmp ne i32 %1863, %1865
  br i1 %cmp.i46.not.i.i.i1384, label %if.then12.i.i.i1385, label %Flow3566, !amdgpu.uniform !42

if.then12.i.i.i1385:                              ; preds = %if.then.i.i.i.i1380
  fence syncscope("agent") release
  %1866 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %1845, i1 true)
  %iszero.i.i.i.i.i.i.i1386 = icmp ne i64 %1845, 0
  %cmp2.i.i.i.i.i.i1387 = icmp eq i64 %1866, %sh_prom.i.i.i.i
  %cmp.i.i.i.i.i.i1388 = select i1 %iszero.i.i.i.i.i.i.i1386, i1 %cmp2.i.i.i.i.i.i1387, i1 false
  %conv4.i.i51.i.i.i1389 = zext i1 %cmp.i.i.i.i.i.i1388 to i32
  %shl.i.i52.i.i.i1390 = shl nuw i32 %conv4.i.i51.i.i.i1389, %1846
  %xor.i.i.i.i.i1391 = xor i32 %shl.i.i52.i.i.i1390, -1
  %1867 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2610, !amdgpu.uniform !42

1868:                                             ; preds = %ComputeEnd2611
  %sunkaddr3330 = getelementptr inbounds i8, ptr addrspace(1) %3733, i64 40
  %1869 = atomicrmw and ptr addrspace(1) %sunkaddr3330, i32 %.lcssa4063 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %1870, !amdgpu.uniform !42

1870:                                             ; preds = %1868, %ComputeEnd2611
  call void @llvm.amdgcn.end.cf.i64(i64 %3760)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %Flow3566, !amdgpu.uniform !42

if.end14.i.i.i1451:                               ; preds = %Flow3566
  %1871 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %1845, i1 true)
  %iszero.i.i.i.i.i.i1452 = icmp ne i64 %1845, 0
  %cmp2.i.i.i.i.i1453 = icmp eq i64 %1871, %sh_prom.i.i.i.i
  %cmp.i.i.i.i.i1454 = select i1 %iszero.i.i.i.i.i.i1452, i1 %cmp2.i.i.i.i.i1453, i1 false
  %1872 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.i.i.i.i.i1454)
  %1873 = extractvalue { i1, i64 } %1872, 0
  %1874 = extractvalue { i1, i64 } %1872, 1
  br i1 %1873, label %if.then16.i.i.i1456, label %if.end22.i.i.i1455

if.then16.i.i.i1456:                              ; preds = %if.end14.i.i.i1451
  %1875 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !76
  %opcode.i.i.i1457 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %1875, i64 %idxprom.i.i.i.i1381, i32 1
  store i32 1660944387, ptr %opcode.i.i.i1457, align 8, !tbaa !77
  %arrayidx21.i.i.i1458 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %1875, i64 %idxprom.i.i.i.i1381
  store i64 %1845, ptr %arrayidx21.i.i.i1458, align 8, !tbaa !80
  br label %if.end22.i.i.i1455, !amdgpu.uniform !42

Flow3567:                                         ; preds = %if.end22.i.i.i1455, %Flow3566
  %1876 = phi i32 [ %1863, %if.end22.i.i.i1455 ], [ %retval.sroa.8.0.i.i.i1366, %Flow3566 ]
  %1877 = phi i64 [ %1845, %if.end22.i.i.i1455 ], [ %retval.sroa.2.0.i.i.i1365, %Flow3566 ]
  %1878 = phi i32 [ %spec.store.select.i.i.i1369, %if.end22.i.i.i1455 ], [ %retval.sroa.6.0.i.i.i1363, %Flow3566 ]
  br label %cleanup26.i.i.i1392, !amdgpu.uniform !42

if.end22.i.i.i1455:                               ; preds = %if.then16.i.i.i1456, %if.end14.i.i.i1451
  call void @llvm.amdgcn.end.cf.i64(i64 %1874)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %Flow3567, !amdgpu.uniform !42

Flow3568:                                         ; preds = %cleanup26.i.i.i1392, %1852
  %1879 = phi i32 [ %1878, %cleanup26.i.i.i1392 ], [ %1841, %1852 ]
  %1880 = phi i64 [ %1877, %cleanup26.i.i.i1392 ], [ %1842, %1852 ]
  %1881 = phi i32 [ %1876, %cleanup26.i.i.i1392 ], [ %1843, %1852 ]
  %1882 = phi i32 [ %1876, %cleanup26.i.i.i1392 ], [ %retval.sroa.8.0.i.i.i1366, %1852 ]
  %1883 = phi i64 [ %1877, %cleanup26.i.i.i1392 ], [ %retval.sroa.2.0.i.i.i1365, %1852 ]
  %1884 = phi i32 [ %1878, %cleanup26.i.i.i1392 ], [ %retval.sroa.6.0.i.i.i1363, %1852 ]
  %1885 = phi i1 [ %cmp.i46.not.i.i.i1384, %cleanup26.i.i.i1392 ], [ true, %1852 ]
  br i1 %1885, label %for.inc.i.i.i1397, label %Flow3569, !amdgpu.uniform !42

cleanup26.i.i.i1392:                              ; preds = %Flow3567
  br label %Flow3568, !amdgpu.uniform !42

for.inc.i.i.i1397:                                ; preds = %Flow3568
  %inc.i.i.i1402 = add i32 %spec.store.select.i.i.i1369, 1
  br label %Flow3569, !amdgpu.uniform !42

Flow3569:                                         ; preds = %for.inc.i.i.i1397, %Flow3568
  %1886 = phi i32 [ %inc.i.i.i1402, %for.inc.i.i.i1397 ], [ poison, %Flow3568 ]
  %1887 = phi i1 [ false, %for.inc.i.i.i1397 ], [ true, %Flow3568 ]
  br i1 %1887, label %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1403, label %for.cond.i.i1.i1362, !amdgpu.uniform !42

_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1403: ; preds = %Flow3569
  %.lcssa4066 = phi i32 [ %1879, %Flow3569 ]
  %.lcssa4065 = phi i64 [ %1880, %Flow3569 ]
  %.lcssa4064 = phi i32 [ %1881, %Flow3569 ]
  fence acquire
  %1888 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %mul.i.i.i.i.i.i1404 = shl i32 %.lcssa4066, 6
  %idxprom.i19.i.i.i.i.i1405 = zext i32 %mul.i.i.i.i.i.i1404 to i64
  %arrayidx.i20.i.i.i.i.i1406 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %1888, i64 %idxprom.i19.i.i.i.i.i1405
  %arrayidx.i22.i.i.i.i.i1407 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %arrayidx.i20.i.i.i.i.i1406, i64 %sh_prom.i.i.i.i
  store i64 sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.37.14 to ptr), i64 2) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.37.14 to ptr) to i64)), ptr %arrayidx.i22.i.i.i.i.i1407, align 8, !tbaa !82
  %spec.select.i.i.i.i.i.i.i1408 = tail call i64 @llvm.umin.i64(i64 sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.37.14 to ptr), i64 2) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.37.14 to ptr) to i64)), i64 56)
  %1889 = and i64 %spec.select.i.i.i.i.i.i.i1408, 15
  %1890 = sub i64 %spec.select.i.i.i.i.i.i.i1408, %1889
  %1891 = icmp ne i64 %1890, 0
  br i1 %1891, label %loop-memcpy-expansion2411.preheader, label %Flow3565, !amdgpu.uniform !42

loop-memcpy-expansion2411.preheader:              ; preds = %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1403
  %1892 = add nuw nsw i64 %75, 8
  %scevgep2974 = getelementptr i8, ptr %1888, i64 %1892
  %1893 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i1405, 6
  %scevgep2975 = getelementptr i8, ptr %scevgep2974, i64 %1893
  br label %loop-memcpy-expansion2411, !amdgpu.uniform !42

Flow3565:                                         ; preds = %Flow3564, %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1403
  br label %loop-memcpy-residual-header2414, !amdgpu.uniform !42

loop-memcpy-expansion2411:                        ; preds = %loop-memcpy-expansion2411.preheader, %loop-memcpy-expansion2411
  %loop-index2412 = phi i64 [ %1896, %loop-memcpy-expansion2411 ], [ 0, %loop-memcpy-expansion2411.preheader ]
  %1894 = getelementptr i8, ptr addrspace(4) @.str.37.14, i64 %loop-index2412, !amdgpu.uniform !42
  %1895 = load <4 x i32>, ptr addrspace(4) %1894, align 1
  %scevgep2976 = getelementptr i8, ptr %scevgep2975, i64 %loop-index2412
  store <4 x i32> %1895, ptr %scevgep2976, align 1
  %1896 = add i64 %loop-index2412, 16
  %1897 = icmp uge i64 %1896, %1890
  br i1 %1897, label %Flow3564, label %loop-memcpy-expansion2411, !amdgpu.uniform !42

Flow3562:                                         ; preds = %Flow3559, %Flow3561
  %1898 = phi i64 [ %2961, %Flow3559 ], [ %2958, %Flow3561 ]
  br label %post-loop-memcpy-expansion2410, !amdgpu.uniform !42

loop-memcpy-residual2413:                         ; preds = %loop-memcpy-residual2413.preheader, %loop-memcpy-residual2413
  %residual-loop-index2415 = phi i64 [ %1900, %loop-memcpy-residual2413 ], [ 0, %loop-memcpy-residual2413.preheader ]
  %scevgep2973 = getelementptr i8, ptr addrspace(4) @.str.37.14, i64 %residual-loop-index2415, !amdgpu.uniform !42
  %1899 = load i8, ptr addrspace(4) %scevgep2973, align 1
  %scevgep2972 = getelementptr i8, ptr %scevgep2971, i64 %residual-loop-index2415
  store i8 %1899, ptr %scevgep2972, align 1
  %1900 = add i64 %residual-loop-index2415, 1
  %1901 = icmp uge i64 %1900, %1889
  br i1 %1901, label %Flow3559, label %loop-memcpy-residual2413, !amdgpu.uniform !42

Flow3559:                                         ; preds = %loop-memcpy-residual2413
  br label %Flow3562, !amdgpu.uniform !42

post-loop-memcpy-expansion2410:                   ; preds = %Flow3562
  %tobool.not.i.i.i.i.i.i1410 = icmp eq i32 %.lcssa4064, 0
  %conv.i.i.i.i.i.i1411 = zext i1 %tobool.not.i.i.i.i.i.i1410 to i32
  fence release
  %1902 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %idxprom.i23.i.i.i.i.i1412 = zext i32 %.lcssa4066 to i64
  %arrayidx.i24.i.i.i.i.i1413 = getelementptr inbounds nuw i32, ptr %1902, i64 %idxprom.i23.i.i.i.i.i1412
  store atomic i32 %conv.i.i.i.i.i.i1411, ptr %arrayidx.i24.i.i.i.i.i1413 monotonic, align 4
  %1903 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !84
  %arrayidx.i.i5.i.i1414 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %1903, i64 %idxprom.i23.i.i.i.i.i1412, !amdgpu.uniform !42
  %1904 = load i64, ptr %arrayidx.i.i5.i.i1414, align 8, !tbaa !80
  br label %while.cond.i.i.i.i1415, !amdgpu.uniform !42

while.cond.i.i.i.i1415:                           ; preds = %post-loop-memcpy-expansion2410, %Flow3557
  %phi.broken3958 = phi i64 [ 0, %post-loop-memcpy-expansion2410 ], [ %1921, %Flow3557 ]
  %lsr.iv2963 = phi i64 [ %1919, %Flow3557 ], [ -54, %post-loop-memcpy-expansion2410 ]
  %lsr.iv2958 = phi ptr addrspace(4) [ %1918, %Flow3557 ], [ getelementptr (i8, ptr addrspace(4) @.str.37.14, i64 56), %post-loop-memcpy-expansion2410 ]
  %port.sroa.37.0.i.i1416 = phi i32 [ %1917, %Flow3557 ], [ %conv.i.i.i.i.i.i1411, %post-loop-memcpy-expansion2410 ]
  %idx.0.i.i.i.i1417 = phi i64 [ %1916, %Flow3557 ], [ 56, %post-loop-memcpy-expansion2410 ]
  %cmp.i.i.i.i1418 = icmp ult i64 %idx.0.i.i.i.i1417, sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.37.14 to ptr), i64 2) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.37.14 to ptr) to i64))
  %1905 = tail call i64 @llvm.amdgcn.ballot.i64(i1 %cmp.i.i.i.i1418)
  %and.i.i.i.i6.i.i1419 = and i64 %1905, %1904
  %tobool.not.i.i.i.i1420 = icmp ne i64 %and.i.i.i.i6.i.i1419, 0
  %1906 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i22.i.i.i1421 = getelementptr inbounds nuw i32, ptr %1906, i64 %idxprom.i23.i.i.i.i.i1412, !amdgpu.uniform !42
  %1907 = load atomic i32, ptr %arrayidx.i22.i.i.i1421 monotonic, align 4
  %1908 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %tobool.not.i.i.i.i1420)
  %1909 = extractvalue { i1, i64 } %1908, 0
  %1910 = extractvalue { i1, i64 } %1908, 1
  br i1 %1909, label %cond.false.i8.i.i.i.i1422, label %Flow3557

cond.false.i8.i.i.i.i1422:                        ; preds = %while.cond.i.i.i.i1415
  %1911 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %1907)
  %1912 = icmp ne i32 %1911, %port.sroa.37.0.i.i1416
  br i1 %1912, label %while.body.i.i21.i.i.i.i1423.preheader, label %Flow3556, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i1423.preheader:           ; preds = %cond.false.i8.i.i.i.i1422
  br label %while.body.i.i21.i.i.i.i1423, !amdgpu.uniform !42

Flow3556:                                         ; preds = %Flow3554, %cond.false.i8.i.i.i.i1422
  br label %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i1426, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i1423:                     ; preds = %while.body.i.i21.i.i.i.i1423.preheader, %while.body.i.i21.i.i.i.i1423
  tail call void @llvm.amdgcn.s.sleep(i32 2)
  %1913 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i22.i.i.i.i1424 = getelementptr inbounds nuw i32, ptr %1913, i64 %idxprom.i23.i.i.i.i.i1412, !amdgpu.uniform !42
  %1914 = load atomic i32, ptr %arrayidx.i.i.i22.i.i.i.i1424 monotonic, align 4
  %1915 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %1914)
  %cmp.i.not.i.i23.i.i.i.i1425 = icmp eq i32 %1915, %port.sroa.37.0.i.i1416
  br i1 %cmp.i.not.i.i23.i.i.i.i1425, label %Flow3554, label %while.body.i.i21.i.i.i.i1423, !amdgpu.uniform !42

Flow3554:                                         ; preds = %while.body.i.i21.i.i.i.i1423
  br label %Flow3556, !amdgpu.uniform !42

Flow3557:                                         ; preds = %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1427, %while.cond.i.i.i.i1415
  %1916 = phi i64 [ %add.i.i.i.i1430, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1427 ], [ poison, %while.cond.i.i.i.i1415 ]
  %1917 = phi i32 [ %conv.i.i26.i.i.i.i1428, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1427 ], [ poison, %while.cond.i.i.i.i1415 ]
  %1918 = phi ptr addrspace(4) [ %scevgep2959, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1427 ], [ poison, %while.cond.i.i.i.i1415 ]
  %1919 = phi i64 [ %lsr.iv.next2964, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1427 ], [ poison, %while.cond.i.i.i.i1415 ]
  %1920 = phi i1 [ false, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1427 ], [ true, %while.cond.i.i.i.i1415 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %1910)
  %1921 = call i64 @llvm.amdgcn.if.break.i64(i1 %1920, i64 %phi.broken3958)
  %1922 = call i1 @llvm.amdgcn.loop.i64(i64 %1921)
  br i1 %1922, label %cond.false.i.i.i1437, label %while.cond.i.i.i.i1415

_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i1426: ; preds = %Flow3556
  fence acquire
  br i1 %cmp.i.i.i.i1418, label %if.then.i.i.i.i.i.i.i1431, label %Flow3553, !amdgpu.uniform !42

if.then.i.i.i.i.i.i.i1431:                        ; preds = %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i1426
  %sub.i.i.i.i.i.i.i1433 = sub nuw i64 sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.37.14 to ptr), i64 2) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.37.14 to ptr) to i64)), %idx.0.i.i.i.i1417
  %spec.select.i.i.i36.i.i.i.i1434 = tail call i64 @llvm.umin.i64(i64 %sub.i.i.i.i.i.i.i1433, i64 64)
  %1923 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %1924 = and i64 %spec.select.i.i.i36.i.i.i.i1434, 15
  %1925 = sub i64 %spec.select.i.i.i36.i.i.i.i1434, %1924
  %1926 = icmp ne i64 %1925, 0
  br i1 %1926, label %loop-memcpy-expansion2417.preheader, label %Flow3552, !amdgpu.uniform !42

loop-memcpy-expansion2417.preheader:              ; preds = %if.then.i.i.i.i.i.i.i1431
  %arrayidx.i20.i35.i.i.i.i1435 = getelementptr %"struct.rpc::Buffer", ptr %1923, i64 %idxprom.i19.i.i.i.i.i1405
  %arrayidx.i22.i38.i.i.i.i1436 = getelementptr %"struct.rpc::Buffer", ptr %arrayidx.i20.i35.i.i.i.i1435, i64 %sh_prom.i.i.i.i
  br label %loop-memcpy-expansion2417, !amdgpu.uniform !42

Flow3552:                                         ; preds = %Flow3551, %if.then.i.i.i.i.i.i.i1431
  br label %loop-memcpy-residual-header2420, !amdgpu.uniform !42

loop-memcpy-expansion2417:                        ; preds = %loop-memcpy-expansion2417.preheader, %loop-memcpy-expansion2417
  %loop-index2418 = phi i64 [ %1929, %loop-memcpy-expansion2417 ], [ 0, %loop-memcpy-expansion2417.preheader ]
  %scevgep2960 = getelementptr i8, ptr addrspace(4) %lsr.iv2958, i64 %loop-index2418, !amdgpu.uniform !42
  %1927 = load <4 x i32>, ptr addrspace(4) %scevgep2960, align 1
  %1928 = getelementptr i8, ptr %arrayidx.i22.i38.i.i.i.i1436, i64 %loop-index2418
  store <4 x i32> %1927, ptr %1928, align 1
  %1929 = add i64 %loop-index2418, 16
  %1930 = icmp uge i64 %1929, %1925
  br i1 %1930, label %Flow3551, label %loop-memcpy-expansion2417, !amdgpu.uniform !42

Flow3549:                                         ; preds = %Flow3548, %loop-memcpy-residual-header2420
  br label %Flow3553, !amdgpu.uniform !42

loop-memcpy-residual2419:                         ; preds = %loop-memcpy-residual2419.preheader, %loop-memcpy-residual2419
  %residual-loop-index2421 = phi i64 [ %1932, %loop-memcpy-residual2419 ], [ 0, %loop-memcpy-residual2419.preheader ]
  %scevgep2969 = getelementptr i8, ptr addrspace(4) %scevgep2968, i64 %residual-loop-index2421, !amdgpu.uniform !42
  %1931 = load i8, ptr addrspace(4) %scevgep2969, align 1
  %scevgep2967 = getelementptr i8, ptr %scevgep2966, i64 %residual-loop-index2421
  store i8 %1931, ptr %scevgep2967, align 1
  %1932 = add i64 %residual-loop-index2421, 1
  %1933 = icmp uge i64 %1932, %1924
  br i1 %1933, label %Flow3548, label %loop-memcpy-residual2419, !amdgpu.uniform !42

Flow3548:                                         ; preds = %loop-memcpy-residual2419
  br label %Flow3549, !amdgpu.uniform !42

_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1427: ; preds = %Flow3553
  %conv.i.i26.i.i.i.i1428 = xor i32 %port.sroa.37.0.i.i1416, 1
  fence release
  %1934 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %arrayidx.i24.i29.i.i.i.i1429 = getelementptr inbounds nuw i32, ptr %1934, i64 %idxprom.i23.i.i.i.i.i1412
  store atomic i32 %conv.i.i26.i.i.i.i1428, ptr %arrayidx.i24.i29.i.i.i.i1429 monotonic, align 4
  %add.i.i.i.i1430 = add i64 %idx.0.i.i.i.i1417, 64
  %scevgep2959 = getelementptr i8, ptr addrspace(4) %lsr.iv2958, i64 64
  %lsr.iv.next2964 = add i64 %lsr.iv2963, -64
  br label %Flow3557, !amdgpu.uniform !42

cond.false.i.i.i1437:                             ; preds = %Flow3557
  %.lcssa4060 = phi i64 [ %1921, %Flow3557 ]
  %port.sroa.37.0.i.i1416.lcssa = phi i32 [ %port.sroa.37.0.i.i1416, %Flow3557 ]
  %.lcssa4059 = phi i32 [ %1907, %Flow3557 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa4060)
  %1935 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %.lcssa4059)
  %1936 = icmp ne i32 %1935, %port.sroa.37.0.i.i1416.lcssa
  %1937 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %1936)
  %1938 = extractvalue { i1, i64 } %1937, 0
  %1939 = extractvalue { i1, i64 } %1937, 1
  br i1 %1938, label %while.body.i.i.i.i1438.preheader, label %Flow3546

while.body.i.i.i.i1438.preheader:                 ; preds = %cond.false.i.i.i1437
  br label %while.body.i.i.i.i1438, !amdgpu.uniform !42

Flow3546:                                         ; preds = %Flow3544, %cond.false.i.i.i1437
  call void @llvm.amdgcn.end.cf.i64(i64 %1939)
  br label %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit1459, !amdgpu.uniform !42

while.body.i.i.i.i1438:                           ; preds = %while.body.i.i.i.i1438.preheader, %while.body.i.i.i.i1438
  %phi.broken3959 = phi i64 [ 0, %while.body.i.i.i.i1438.preheader ], [ %1943, %while.body.i.i.i.i1438 ]
  tail call void @llvm.amdgcn.s.sleep(i32 2)
  %1940 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i8.i.i1439 = getelementptr inbounds nuw i32, ptr %1940, i64 %idxprom.i23.i.i.i.i.i1412, !amdgpu.uniform !42
  %1941 = load atomic i32, ptr %arrayidx.i.i.i8.i.i1439 monotonic, align 4
  %1942 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %1941)
  %cmp.i.not.i.i.i.i1440 = icmp eq i32 %1942, %port.sroa.37.0.i.i1416.lcssa
  %1943 = call i64 @llvm.amdgcn.if.break.i64(i1 %cmp.i.not.i.i.i.i1440, i64 %phi.broken3959)
  %1944 = call i1 @llvm.amdgcn.loop.i64(i64 %1943)
  br i1 %1944, label %Flow3544, label %while.body.i.i.i.i1438

Flow3544:                                         ; preds = %while.body.i.i.i.i1438
  %.lcssa4058 = phi i64 [ %1943, %while.body.i.i.i.i1438 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa4058)
  br label %Flow3546, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit1459: ; preds = %Flow3546
  fence acquire
  tail call void @llvm.amdgcn.wave.barrier()
  fence syncscope("agent") release
  %1945 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %.lcssa4065, i1 true)
  %iszero.i.i.i.i.i13.i.i1442 = icmp ne i64 %.lcssa4065, 0
  %cmp2.i.i.i.i14.i.i1443 = icmp eq i64 %1945, %sh_prom.i.i.i.i
  %cmp.i.i.i.i15.i.i1444 = select i1 %iszero.i.i.i.i.i13.i.i1442, i1 %cmp2.i.i.i.i14.i.i1443, i1 false
  %1946 = and i32 %.lcssa4066, 31
  %conv4.i.i.i19.i.i1448 = zext i1 %cmp.i.i.i.i15.i.i1444 to i32
  %shl.i.i.i20.i.i1449 = shl nuw i32 %conv4.i.i.i19.i.i1448, %1946
  %xor.i.i.i21.i.i1450 = xor i32 %shl.i.i.i20.i.i1449, -1
  %1947 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2614, !amdgpu.uniform !42

1948:                                             ; preds = %ComputeEnd2615
  %div8.i.i.i16.i.i1445 = lshr i32 %.lcssa4066, 5
  %idxprom.i.i.i17.i.i1446 = zext nneg i32 %div8.i.i.i16.i.i1445 to i64
  %1949 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i17.i.i1446
  %sunkaddr3331 = getelementptr inbounds i8, ptr addrspace(1) %1949, i64 40
  %1950 = atomicrmw and ptr addrspace(1) %sunkaddr3331, i32 %.lcssa4057 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %1951, !amdgpu.uniform !42

1951:                                             ; preds = %1948, %ComputeEnd2615
  call void @llvm.amdgcn.end.cf.i64(i64 %3777)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %for.cond.i.i.i1460, !amdgpu.uniform !42

for.cond.i.i.i1460:                               ; preds = %1951, %for.cond.i.i.i1460
  %phi.broken3960 = phi i64 [ 0, %1951 ], [ %1954, %for.cond.i.i.i1460 ]
  %lsr.iv2944 = phi i64 [ %lsr.iv.next2945, %for.cond.i.i.i1460 ], [ -1, %1951 ]
  %scevgep2956 = getelementptr i8, ptr %RHSStr, i64 %lsr.iv2944
  %scevgep2957 = getelementptr i8, ptr %scevgep2956, i64 1
  %1952 = addrspacecast ptr %scevgep2957 to ptr addrspace(4)
  %1953 = load i8, ptr addrspace(4) %1952, align 1, !tbaa !57, !noalias.addrspace !101
  %cmp.i.i.i1462 = icmp eq i8 %1953, 0
  %lsr.iv.next2945 = add i64 %lsr.iv2944, 1
  %1954 = call i64 @llvm.amdgcn.if.break.i64(i1 %cmp.i.i.i1462, i64 %phi.broken3960)
  %1955 = call i1 @llvm.amdgcn.loop.i64(i64 %1954)
  br i1 %1955, label %for.cond.i.i1.i1469.preheader, label %for.cond.i.i.i1460

for.cond.i.i1.i1469.preheader:                    ; preds = %for.cond.i.i.i1460
  %lsr.iv.next2945.lcssa = phi i64 [ %lsr.iv.next2945, %for.cond.i.i.i1460 ]
  %.lcssa4056 = phi i64 [ %1954, %for.cond.i.i.i1460 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa4056)
  br label %for.cond.i.i1.i1469, !amdgpu.uniform !42

for.cond.i.i1.i1469:                              ; preds = %for.cond.i.i1.i1469.preheader, %Flow3543
  %1956 = phi i32 [ %1994, %Flow3543 ], [ poison, %for.cond.i.i1.i1469.preheader ]
  %1957 = phi i64 [ %1995, %Flow3543 ], [ poison, %for.cond.i.i1.i1469.preheader ]
  %1958 = phi i32 [ %1996, %Flow3543 ], [ poison, %for.cond.i.i1.i1469.preheader ]
  %retval.sroa.6.0.i.i.i1470 = phi i32 [ %1999, %Flow3543 ], [ undef, %for.cond.i.i1.i1469.preheader ]
  %retval.sroa.2.0.i.i.i1472 = phi i64 [ %1998, %Flow3543 ], [ undef, %for.cond.i.i1.i1469.preheader ]
  %retval.sroa.8.0.i.i.i1473 = phi i32 [ %1997, %Flow3543 ], [ undef, %for.cond.i.i1.i1469.preheader ]
  %index.0.i.i.i1474 = phi i32 [ %2001, %Flow3543 ], [ 0, %for.cond.i.i1.i1469.preheader ]
  %1959 = load i32, ptr addrspace(1) @__llvm_rpc_client, align 8, !tbaa !68
  %cmp.not.i.i.i1475 = icmp ult i32 %index.0.i.i.i1474, %1959
  %spec.store.select.i.i.i1476 = select i1 %cmp.not.i.i.i1475, i32 %index.0.i.i.i1474, i32 0
  %1960 = tail call noundef i64 @llvm.amdgcn.ballot.i64(i1 true)
  %1961 = and i32 %spec.store.select.i.i.i1476, 31
  %1962 = lshr i64 %1960, %sh_prom.i.i.i.i
  %1963 = trunc i64 %1962 to i32
  %conv4.i.i.i.i.i1480 = and i32 %1963, 1
  %shl.i.i.i.i.i1481 = shl nuw i32 %conv4.i.i.i.i.i1480, %1961
  %1964 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2618, !amdgpu.uniform !42

1965:                                             ; preds = %ComputeEnd2619
  %sunkaddr3332 = getelementptr inbounds i8, ptr addrspace(1) %3787, i64 40
  %1966 = atomicrmw or ptr addrspace(1) %sunkaddr3332, i32 %.lcssa4050 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %1967, !amdgpu.uniform !42

1967:                                             ; preds = %1965, %ComputeEnd2619
  %1968 = phi i32 [ %1966, %1965 ], [ poison, %ComputeEnd2619 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %3796)
  %1969 = call i32 @llvm.amdgcn.readfirstlane.i32(i32 %1968)
  %1970 = or i32 %1969, %.lcssa4051
  %shl5.i.i.i.i.i1482 = shl nuw i32 1, %1961
  %and.i.i.i.i.i1483 = and i32 %shl5.i.i.i.i.i1482, %1970
  %tobool3.i.i.i.i1484 = icmp ne i32 %and.i.i.i.i.i1483, 0
  %1971 = tail call i64 @llvm.amdgcn.ballot.i64(i1 %tobool3.i.i.i.i1484)
  %and.i.i.i.i.i.i1485 = and i64 %1971, %1960
  %cmp.i.not.i.i.i1486 = icmp ne i64 %1960, %and.i.i.i.i.i.i1485
  br i1 %cmp.i.not.i.i.i1486, label %if.then.i.i.i.i1487, label %Flow3542, !amdgpu.uniform !42

if.then.i.i.i.i1487:                              ; preds = %1967
  fence syncscope("agent") acquire
  %1972 = load <2 x i64>, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !75
  %1973 = extractelement <2 x i64> %1972, i32 0
  %1974 = inttoptr i64 %1973 to ptr
  %1975 = extractelement <2 x i64> %1972, i32 1
  %1976 = inttoptr i64 %1975 to ptr
  %idxprom.i.i.i.i1488 = zext i32 %spec.store.select.i.i.i1476 to i64
  %arrayidx.i.i.i.i1489 = getelementptr inbounds nuw i32, ptr %1974, i64 %idxprom.i.i.i.i1488, !amdgpu.uniform !42
  %1977 = load atomic i32, ptr %arrayidx.i.i.i.i1489 monotonic, align 4
  %1978 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %1977)
  %arrayidx.i45.i.i.i1490 = getelementptr inbounds nuw i32, ptr %1976, i64 %idxprom.i.i.i.i1488, !amdgpu.uniform !42
  %1979 = load atomic i32, ptr %arrayidx.i45.i.i.i1490 monotonic, align 4
  %1980 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %1979)
  %cmp.i46.not.i.i.i1491 = icmp ne i32 %1978, %1980
  br i1 %cmp.i46.not.i.i.i1491, label %if.then12.i.i.i1492, label %Flow3540, !amdgpu.uniform !42

if.then12.i.i.i1492:                              ; preds = %if.then.i.i.i.i1487
  fence syncscope("agent") release
  %1981 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %1960, i1 true)
  %iszero.i.i.i.i.i.i.i1493 = icmp ne i64 %1960, 0
  %cmp2.i.i.i.i.i.i1494 = icmp eq i64 %1981, %sh_prom.i.i.i.i
  %cmp.i.i.i.i.i.i1495 = select i1 %iszero.i.i.i.i.i.i.i1493, i1 %cmp2.i.i.i.i.i.i1494, i1 false
  %conv4.i.i51.i.i.i1496 = zext i1 %cmp.i.i.i.i.i.i1495 to i32
  %shl.i.i52.i.i.i1497 = shl nuw i32 %conv4.i.i51.i.i.i1496, %1961
  %xor.i.i.i.i.i1498 = xor i32 %shl.i.i52.i.i.i1497, -1
  %1982 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2623, !amdgpu.uniform !42

1983:                                             ; preds = %ComputeEnd2624
  %sunkaddr3333 = getelementptr inbounds i8, ptr addrspace(1) %3787, i64 40
  %1984 = atomicrmw and ptr addrspace(1) %sunkaddr3333, i32 %.lcssa4052 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %1985, !amdgpu.uniform !42

1985:                                             ; preds = %1983, %ComputeEnd2624
  call void @llvm.amdgcn.end.cf.i64(i64 %3814)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %Flow3540, !amdgpu.uniform !42

if.end14.i.i.i1558:                               ; preds = %Flow3540
  %1986 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %1960, i1 true)
  %iszero.i.i.i.i.i.i1559 = icmp ne i64 %1960, 0
  %cmp2.i.i.i.i.i1560 = icmp eq i64 %1986, %sh_prom.i.i.i.i
  %cmp.i.i.i.i.i1561 = select i1 %iszero.i.i.i.i.i.i1559, i1 %cmp2.i.i.i.i.i1560, i1 false
  %1987 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.i.i.i.i.i1561)
  %1988 = extractvalue { i1, i64 } %1987, 0
  %1989 = extractvalue { i1, i64 } %1987, 1
  br i1 %1988, label %if.then16.i.i.i1563, label %if.end22.i.i.i1562

if.then16.i.i.i1563:                              ; preds = %if.end14.i.i.i1558
  %1990 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !76
  %opcode.i.i.i1564 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %1990, i64 %idxprom.i.i.i.i1488, i32 1
  store i32 1660944387, ptr %opcode.i.i.i1564, align 8, !tbaa !77
  %arrayidx21.i.i.i1565 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %1990, i64 %idxprom.i.i.i.i1488
  store i64 %1960, ptr %arrayidx21.i.i.i1565, align 8, !tbaa !80
  br label %if.end22.i.i.i1562, !amdgpu.uniform !42

Flow3541:                                         ; preds = %if.end22.i.i.i1562, %Flow3540
  %1991 = phi i32 [ %1978, %if.end22.i.i.i1562 ], [ %retval.sroa.8.0.i.i.i1473, %Flow3540 ]
  %1992 = phi i64 [ %1960, %if.end22.i.i.i1562 ], [ %retval.sroa.2.0.i.i.i1472, %Flow3540 ]
  %1993 = phi i32 [ %spec.store.select.i.i.i1476, %if.end22.i.i.i1562 ], [ %retval.sroa.6.0.i.i.i1470, %Flow3540 ]
  br label %cleanup26.i.i.i1499, !amdgpu.uniform !42

if.end22.i.i.i1562:                               ; preds = %if.then16.i.i.i1563, %if.end14.i.i.i1558
  call void @llvm.amdgcn.end.cf.i64(i64 %1989)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %Flow3541, !amdgpu.uniform !42

Flow3542:                                         ; preds = %cleanup26.i.i.i1499, %1967
  %1994 = phi i32 [ %1993, %cleanup26.i.i.i1499 ], [ %1956, %1967 ]
  %1995 = phi i64 [ %1992, %cleanup26.i.i.i1499 ], [ %1957, %1967 ]
  %1996 = phi i32 [ %1991, %cleanup26.i.i.i1499 ], [ %1958, %1967 ]
  %1997 = phi i32 [ %1991, %cleanup26.i.i.i1499 ], [ %retval.sroa.8.0.i.i.i1473, %1967 ]
  %1998 = phi i64 [ %1992, %cleanup26.i.i.i1499 ], [ %retval.sroa.2.0.i.i.i1472, %1967 ]
  %1999 = phi i32 [ %1993, %cleanup26.i.i.i1499 ], [ %retval.sroa.6.0.i.i.i1470, %1967 ]
  %2000 = phi i1 [ %cmp.i46.not.i.i.i1491, %cleanup26.i.i.i1499 ], [ true, %1967 ]
  br i1 %2000, label %for.inc.i.i.i1504, label %Flow3543, !amdgpu.uniform !42

cleanup26.i.i.i1499:                              ; preds = %Flow3541
  br label %Flow3542, !amdgpu.uniform !42

for.inc.i.i.i1504:                                ; preds = %Flow3542
  %inc.i.i.i1509 = add i32 %spec.store.select.i.i.i1476, 1
  br label %Flow3543, !amdgpu.uniform !42

Flow3543:                                         ; preds = %for.inc.i.i.i1504, %Flow3542
  %2001 = phi i32 [ %inc.i.i.i1509, %for.inc.i.i.i1504 ], [ poison, %Flow3542 ]
  %2002 = phi i1 [ false, %for.inc.i.i.i1504 ], [ true, %Flow3542 ]
  br i1 %2002, label %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1510, label %for.cond.i.i1.i1469, !amdgpu.uniform !42

_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1510: ; preds = %Flow3543
  %.lcssa4055 = phi i32 [ %1994, %Flow3543 ]
  %.lcssa4054 = phi i64 [ %1995, %Flow3543 ]
  %.lcssa4053 = phi i32 [ %1996, %Flow3543 ]
  %scevgep2955 = getelementptr i8, ptr %RHSStr, i64 %lsr.iv.next2945.lcssa
  %sub.ptr.lhs.cast.i.i.i1465 = ptrtoint ptr %scevgep2955 to i64
  %sub.ptr.rhs.cast.i.i.i1466 = ptrtoint ptr %RHSStr to i64
  %sub.ptr.sub.i.i.i1467 = sub i64 %sub.ptr.lhs.cast.i.i.i1465, %sub.ptr.rhs.cast.i.i.i1466
  fence acquire
  %2003 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %mul.i.i.i.i.i.i1511 = shl i32 %.lcssa4055, 6
  %idxprom.i19.i.i.i.i.i1512 = zext i32 %mul.i.i.i.i.i.i1511 to i64
  %arrayidx.i20.i.i.i.i.i1513 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %2003, i64 %idxprom.i19.i.i.i.i.i1512
  %arrayidx.i22.i.i.i.i.i1514 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %arrayidx.i20.i.i.i.i.i1513, i64 %sh_prom.i.i.i.i
  store i64 %sub.ptr.sub.i.i.i1467, ptr %arrayidx.i22.i.i.i.i.i1514, align 8, !tbaa !82
  %spec.select.i.i.i.i.i.i.i1515 = tail call i64 @llvm.umin.i64(i64 %sub.ptr.sub.i.i.i1467, i64 56)
  %2004 = and i64 %spec.select.i.i.i.i.i.i.i1515, 15
  %2005 = sub i64 %spec.select.i.i.i.i.i.i.i1515, %2004
  %2006 = icmp ne i64 %2005, 0
  %2007 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %2006)
  %2008 = extractvalue { i1, i64 } %2007, 0
  %2009 = extractvalue { i1, i64 } %2007, 1
  br i1 %2008, label %loop-memcpy-expansion2219.preheader, label %Flow3539

loop-memcpy-expansion2219.preheader:              ; preds = %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1510
  %2010 = add nuw nsw i64 %75, 8
  %scevgep2952 = getelementptr i8, ptr %2003, i64 %2010
  %2011 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i1512, 6
  %scevgep2953 = getelementptr i8, ptr %scevgep2952, i64 %2011
  br label %loop-memcpy-expansion2219, !amdgpu.uniform !42

Flow3539:                                         ; preds = %Flow3538, %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1510
  call void @llvm.amdgcn.end.cf.i64(i64 %2009)
  br label %loop-memcpy-residual-header2222, !amdgpu.uniform !42

loop-memcpy-expansion2219:                        ; preds = %loop-memcpy-expansion2219.preheader, %loop-memcpy-expansion2219
  %phi.broken3961 = phi i64 [ 0, %loop-memcpy-expansion2219.preheader ], [ %2016, %loop-memcpy-expansion2219 ]
  %loop-index2220 = phi i64 [ %2014, %loop-memcpy-expansion2219 ], [ 0, %loop-memcpy-expansion2219.preheader ]
  %2012 = getelementptr i8, ptr %RHSStr, i64 %loop-index2220
  %2013 = load <4 x i32>, ptr %2012, align 1
  %scevgep2954 = getelementptr i8, ptr %scevgep2953, i64 %loop-index2220
  store <4 x i32> %2013, ptr %scevgep2954, align 1
  %2014 = add i64 %loop-index2220, 16
  %2015 = icmp uge i64 %2014, %2005
  %2016 = call i64 @llvm.amdgcn.if.break.i64(i1 %2015, i64 %phi.broken3961)
  %2017 = call i1 @llvm.amdgcn.loop.i64(i64 %2016)
  br i1 %2017, label %Flow3538, label %loop-memcpy-expansion2219

Flow3536:                                         ; preds = %Flow3533, %Flow3535
  %2018 = phi i64 [ %2785, %Flow3533 ], [ %2778, %Flow3535 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %2781)
  br label %post-loop-memcpy-expansion2218, !amdgpu.uniform !42

loop-memcpy-residual2221:                         ; preds = %loop-memcpy-residual2221.preheader, %loop-memcpy-residual2221
  %phi.broken3962 = phi i64 [ 0, %loop-memcpy-residual2221.preheader ], [ %2022, %loop-memcpy-residual2221 ]
  %residual-loop-index2223 = phi i64 [ %2020, %loop-memcpy-residual2221 ], [ 0, %loop-memcpy-residual2221.preheader ]
  %scevgep2951 = getelementptr i8, ptr %scevgep2950, i64 %residual-loop-index2223
  %2019 = load i8, ptr %scevgep2951, align 1
  %scevgep2949 = getelementptr i8, ptr %scevgep2948, i64 %residual-loop-index2223
  store i8 %2019, ptr %scevgep2949, align 1
  %2020 = add i64 %residual-loop-index2223, 1
  %2021 = icmp uge i64 %2020, %2004
  %2022 = call i64 @llvm.amdgcn.if.break.i64(i1 %2021, i64 %phi.broken3962)
  %2023 = call i1 @llvm.amdgcn.loop.i64(i64 %2022)
  br i1 %2023, label %Flow3533, label %loop-memcpy-residual2221

Flow3533:                                         ; preds = %loop-memcpy-residual2221
  %.lcssa4048 = phi i64 [ %2022, %loop-memcpy-residual2221 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa4048)
  br label %Flow3536, !amdgpu.uniform !42

post-loop-memcpy-expansion2218:                   ; preds = %Flow3536
  %2024 = add i64 %lsr.iv.next2945.lcssa, -56
  %tobool.not.i.i.i.i.i.i1517 = icmp eq i32 %.lcssa4053, 0
  %conv.i.i.i.i.i.i1518 = zext i1 %tobool.not.i.i.i.i.i.i1517 to i32
  fence release
  %2025 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %idxprom.i23.i.i.i.i.i1519 = zext i32 %.lcssa4055 to i64
  %arrayidx.i24.i.i.i.i.i1520 = getelementptr inbounds nuw i32, ptr %2025, i64 %idxprom.i23.i.i.i.i.i1519
  store atomic i32 %conv.i.i.i.i.i.i1518, ptr %arrayidx.i24.i.i.i.i.i1520 monotonic, align 4
  %2026 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !84
  %arrayidx.i.i5.i.i1521 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %2026, i64 %idxprom.i23.i.i.i.i.i1519, !amdgpu.uniform !42
  %2027 = load i64, ptr %arrayidx.i.i5.i.i1521, align 8, !tbaa !80
  %scevgep2927 = getelementptr i8, ptr %RHSStr, i64 56
  br label %while.cond.i.i.i.i1522, !amdgpu.uniform !42

while.cond.i.i.i.i1522:                           ; preds = %post-loop-memcpy-expansion2218, %Flow3531
  %phi.broken3965 = phi i64 [ 0, %post-loop-memcpy-expansion2218 ], [ %2044, %Flow3531 ]
  %lsr.iv2935 = phi i64 [ %2042, %Flow3531 ], [ %2024, %post-loop-memcpy-expansion2218 ]
  %lsr.iv2928 = phi ptr [ %2041, %Flow3531 ], [ %scevgep2927, %post-loop-memcpy-expansion2218 ]
  %port.sroa.37.0.i.i1523 = phi i32 [ %2040, %Flow3531 ], [ %conv.i.i.i.i.i.i1518, %post-loop-memcpy-expansion2218 ]
  %idx.0.i.i.i.i1524 = phi i64 [ %2039, %Flow3531 ], [ 56, %post-loop-memcpy-expansion2218 ]
  %cmp.i.i.i.i1525 = icmp ugt i64 %sub.ptr.sub.i.i.i1467, %idx.0.i.i.i.i1524
  %2028 = tail call i64 @llvm.amdgcn.ballot.i64(i1 %cmp.i.i.i.i1525)
  %and.i.i.i.i6.i.i1526 = and i64 %2028, %2027
  %tobool.not.i.i.i.i1527 = icmp ne i64 %and.i.i.i.i6.i.i1526, 0
  %2029 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i22.i.i.i1528 = getelementptr inbounds nuw i32, ptr %2029, i64 %idxprom.i23.i.i.i.i.i1519, !amdgpu.uniform !42
  %2030 = load atomic i32, ptr %arrayidx.i22.i.i.i1528 monotonic, align 4
  %2031 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %tobool.not.i.i.i.i1527)
  %2032 = extractvalue { i1, i64 } %2031, 0
  %2033 = extractvalue { i1, i64 } %2031, 1
  br i1 %2032, label %cond.false.i8.i.i.i.i1529, label %Flow3531

cond.false.i8.i.i.i.i1529:                        ; preds = %while.cond.i.i.i.i1522
  %2034 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %2030)
  %2035 = icmp ne i32 %2034, %port.sroa.37.0.i.i1523
  br i1 %2035, label %while.body.i.i21.i.i.i.i1530.preheader, label %Flow3530, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i1530.preheader:           ; preds = %cond.false.i8.i.i.i.i1529
  br label %while.body.i.i21.i.i.i.i1530, !amdgpu.uniform !42

Flow3530:                                         ; preds = %Flow3528, %cond.false.i8.i.i.i.i1529
  br label %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i1533, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i1530:                     ; preds = %while.body.i.i21.i.i.i.i1530.preheader, %while.body.i.i21.i.i.i.i1530
  tail call void @llvm.amdgcn.s.sleep(i32 2)
  %2036 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i22.i.i.i.i1531 = getelementptr inbounds nuw i32, ptr %2036, i64 %idxprom.i23.i.i.i.i.i1519, !amdgpu.uniform !42
  %2037 = load atomic i32, ptr %arrayidx.i.i.i22.i.i.i.i1531 monotonic, align 4
  %2038 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %2037)
  %cmp.i.not.i.i23.i.i.i.i1532 = icmp eq i32 %2038, %port.sroa.37.0.i.i1523
  br i1 %cmp.i.not.i.i23.i.i.i.i1532, label %Flow3528, label %while.body.i.i21.i.i.i.i1530, !amdgpu.uniform !42

Flow3528:                                         ; preds = %while.body.i.i21.i.i.i.i1530
  br label %Flow3530, !amdgpu.uniform !42

Flow3531:                                         ; preds = %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1534, %while.cond.i.i.i.i1522
  %2039 = phi i64 [ %add.i.i.i.i1537, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1534 ], [ poison, %while.cond.i.i.i.i1522 ]
  %2040 = phi i32 [ %conv.i.i26.i.i.i.i1535, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1534 ], [ poison, %while.cond.i.i.i.i1522 ]
  %2041 = phi ptr [ %scevgep2929, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1534 ], [ poison, %while.cond.i.i.i.i1522 ]
  %2042 = phi i64 [ %lsr.iv.next2936, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1534 ], [ poison, %while.cond.i.i.i.i1522 ]
  %2043 = phi i1 [ false, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1534 ], [ true, %while.cond.i.i.i.i1522 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %2033)
  %2044 = call i64 @llvm.amdgcn.if.break.i64(i1 %2043, i64 %phi.broken3965)
  %2045 = call i1 @llvm.amdgcn.loop.i64(i64 %2044)
  br i1 %2045, label %cond.false.i.i.i1544, label %while.cond.i.i.i.i1522

_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i1533: ; preds = %Flow3530
  fence acquire
  %2046 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.i.i.i.i1525)
  %2047 = extractvalue { i1, i64 } %2046, 0
  %2048 = extractvalue { i1, i64 } %2046, 1
  br i1 %2047, label %if.then.i.i.i.i.i.i.i1538, label %Flow3527

if.then.i.i.i.i.i.i.i1538:                        ; preds = %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i1533
  %sub.i.i.i.i.i.i.i1540 = sub nuw i64 %sub.ptr.sub.i.i.i1467, %idx.0.i.i.i.i1524
  %spec.select.i.i.i36.i.i.i.i1541 = tail call i64 @llvm.umin.i64(i64 %sub.i.i.i.i.i.i.i1540, i64 64)
  %2049 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %2050 = and i64 %spec.select.i.i.i36.i.i.i.i1541, 15
  %2051 = sub i64 %spec.select.i.i.i36.i.i.i.i1541, %2050
  %2052 = icmp ne i64 %2051, 0
  %2053 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %2052)
  %2054 = extractvalue { i1, i64 } %2053, 0
  %2055 = extractvalue { i1, i64 } %2053, 1
  br i1 %2054, label %loop-memcpy-expansion2213.preheader, label %Flow3526

loop-memcpy-expansion2213.preheader:              ; preds = %if.then.i.i.i.i.i.i.i1538
  %arrayidx.i20.i35.i.i.i.i1542 = getelementptr %"struct.rpc::Buffer", ptr %2049, i64 %idxprom.i19.i.i.i.i.i1512
  %arrayidx.i22.i38.i.i.i.i1543 = getelementptr %"struct.rpc::Buffer", ptr %arrayidx.i20.i35.i.i.i.i1542, i64 %sh_prom.i.i.i.i
  br label %loop-memcpy-expansion2213, !amdgpu.uniform !42

Flow3526:                                         ; preds = %Flow3525, %if.then.i.i.i.i.i.i.i1538
  call void @llvm.amdgcn.end.cf.i64(i64 %2055)
  br label %loop-memcpy-residual-header2216, !amdgpu.uniform !42

loop-memcpy-expansion2213:                        ; preds = %loop-memcpy-expansion2213.preheader, %loop-memcpy-expansion2213
  %phi.broken3963 = phi i64 [ 0, %loop-memcpy-expansion2213.preheader ], [ %2060, %loop-memcpy-expansion2213 ]
  %loop-index2214 = phi i64 [ %2058, %loop-memcpy-expansion2213 ], [ 0, %loop-memcpy-expansion2213.preheader ]
  %scevgep2930 = getelementptr i8, ptr %lsr.iv2928, i64 %loop-index2214
  %2056 = load <4 x i32>, ptr %scevgep2930, align 1
  %2057 = getelementptr i8, ptr %arrayidx.i22.i38.i.i.i.i1543, i64 %loop-index2214
  store <4 x i32> %2056, ptr %2057, align 1
  %2058 = add i64 %loop-index2214, 16
  %2059 = icmp uge i64 %2058, %2051
  %2060 = call i64 @llvm.amdgcn.if.break.i64(i1 %2059, i64 %phi.broken3963)
  %2061 = call i1 @llvm.amdgcn.loop.i64(i64 %2060)
  br i1 %2061, label %Flow3525, label %loop-memcpy-expansion2213

Flow3523:                                         ; preds = %Flow3522, %loop-memcpy-residual-header2216
  call void @llvm.amdgcn.end.cf.i64(i64 %2771)
  br label %Flow3527, !amdgpu.uniform !42

loop-memcpy-residual2215:                         ; preds = %loop-memcpy-residual2215.preheader, %loop-memcpy-residual2215
  %phi.broken3964 = phi i64 [ 0, %loop-memcpy-residual2215.preheader ], [ %2065, %loop-memcpy-residual2215 ]
  %residual-loop-index2217 = phi i64 [ %2063, %loop-memcpy-residual2215 ], [ 0, %loop-memcpy-residual2215.preheader ]
  %scevgep2942 = getelementptr i8, ptr %scevgep2941, i64 %residual-loop-index2217
  %2062 = load i8, ptr %scevgep2942, align 1
  %scevgep2939 = getelementptr i8, ptr %scevgep2938, i64 %residual-loop-index2217
  store i8 %2062, ptr %scevgep2939, align 1
  %2063 = add i64 %residual-loop-index2217, 1
  %2064 = icmp uge i64 %2063, %2050
  %2065 = call i64 @llvm.amdgcn.if.break.i64(i1 %2064, i64 %phi.broken3964)
  %2066 = call i1 @llvm.amdgcn.loop.i64(i64 %2065)
  br i1 %2066, label %Flow3522, label %loop-memcpy-residual2215

Flow3522:                                         ; preds = %loop-memcpy-residual2215
  %.lcssa4045 = phi i64 [ %2065, %loop-memcpy-residual2215 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa4045)
  br label %Flow3523, !amdgpu.uniform !42

_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1534: ; preds = %Flow3527
  %conv.i.i26.i.i.i.i1535 = xor i32 %port.sroa.37.0.i.i1523, 1
  fence release
  %2067 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %arrayidx.i24.i29.i.i.i.i1536 = getelementptr inbounds nuw i32, ptr %2067, i64 %idxprom.i23.i.i.i.i.i1519
  store atomic i32 %conv.i.i26.i.i.i.i1535, ptr %arrayidx.i24.i29.i.i.i.i1536 monotonic, align 4
  %add.i.i.i.i1537 = add i64 %idx.0.i.i.i.i1524, 64
  %scevgep2929 = getelementptr i8, ptr %lsr.iv2928, i64 64
  %lsr.iv.next2936 = add i64 %lsr.iv2935, -64
  br label %Flow3531, !amdgpu.uniform !42

cond.false.i.i.i1544:                             ; preds = %Flow3531
  %.lcssa4047 = phi i64 [ %2044, %Flow3531 ]
  %port.sroa.37.0.i.i1523.lcssa = phi i32 [ %port.sroa.37.0.i.i1523, %Flow3531 ]
  %.lcssa4046 = phi i32 [ %2030, %Flow3531 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa4047)
  %2068 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %.lcssa4046)
  %2069 = icmp ne i32 %2068, %port.sroa.37.0.i.i1523.lcssa
  %2070 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %2069)
  %2071 = extractvalue { i1, i64 } %2070, 0
  %2072 = extractvalue { i1, i64 } %2070, 1
  br i1 %2071, label %while.body.i.i.i.i1545.preheader, label %Flow3520

while.body.i.i.i.i1545.preheader:                 ; preds = %cond.false.i.i.i1544
  br label %while.body.i.i.i.i1545, !amdgpu.uniform !42

Flow3520:                                         ; preds = %Flow3518, %cond.false.i.i.i1544
  call void @llvm.amdgcn.end.cf.i64(i64 %2072)
  br label %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit1566, !amdgpu.uniform !42

while.body.i.i.i.i1545:                           ; preds = %while.body.i.i.i.i1545.preheader, %while.body.i.i.i.i1545
  %phi.broken3966 = phi i64 [ 0, %while.body.i.i.i.i1545.preheader ], [ %2076, %while.body.i.i.i.i1545 ]
  tail call void @llvm.amdgcn.s.sleep(i32 2)
  %2073 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i8.i.i1546 = getelementptr inbounds nuw i32, ptr %2073, i64 %idxprom.i23.i.i.i.i.i1519, !amdgpu.uniform !42
  %2074 = load atomic i32, ptr %arrayidx.i.i.i8.i.i1546 monotonic, align 4
  %2075 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %2074)
  %cmp.i.not.i.i.i.i1547 = icmp eq i32 %2075, %port.sroa.37.0.i.i1523.lcssa
  %2076 = call i64 @llvm.amdgcn.if.break.i64(i1 %cmp.i.not.i.i.i.i1547, i64 %phi.broken3966)
  %2077 = call i1 @llvm.amdgcn.loop.i64(i64 %2076)
  br i1 %2077, label %Flow3518, label %while.body.i.i.i.i1545

Flow3518:                                         ; preds = %while.body.i.i.i.i1545
  %.lcssa4043 = phi i64 [ %2076, %while.body.i.i.i.i1545 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa4043)
  br label %Flow3520, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit1566: ; preds = %Flow3520
  fence acquire
  tail call void @llvm.amdgcn.wave.barrier()
  fence syncscope("agent") release
  %2078 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %.lcssa4054, i1 true)
  %iszero.i.i.i.i.i13.i.i1549 = icmp ne i64 %.lcssa4054, 0
  %cmp2.i.i.i.i14.i.i1550 = icmp eq i64 %2078, %sh_prom.i.i.i.i
  %cmp.i.i.i.i15.i.i1551 = select i1 %iszero.i.i.i.i.i13.i.i1549, i1 %cmp2.i.i.i.i14.i.i1550, i1 false
  %2079 = and i32 %.lcssa4055, 31
  %conv4.i.i.i19.i.i1555 = zext i1 %cmp.i.i.i.i15.i.i1551 to i32
  %shl.i.i.i20.i.i1556 = shl nuw i32 %conv4.i.i.i19.i.i1555, %2079
  %xor.i.i.i21.i.i1557 = xor i32 %shl.i.i.i20.i.i1556, -1
  %2080 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2627, !amdgpu.uniform !42

2081:                                             ; preds = %ComputeEnd2628
  %div8.i.i.i16.i.i1552 = lshr i32 %.lcssa4055, 5
  %idxprom.i.i.i17.i.i1553 = zext nneg i32 %div8.i.i.i16.i.i1552 to i64
  %2082 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i17.i.i1553
  %sunkaddr3334 = getelementptr inbounds i8, ptr addrspace(1) %2082, i64 40
  %2083 = atomicrmw and ptr addrspace(1) %sunkaddr3334, i32 %.lcssa4042 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %2084, !amdgpu.uniform !42

2084:                                             ; preds = %2081, %ComputeEnd2628
  call void @llvm.amdgcn.end.cf.i64(i64 %3831)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %for.cond.i.i.i1570, !amdgpu.uniform !42

for.cond.i.i.i1570:                               ; preds = %2084, %Flow3517
  %2085 = phi i32 [ %2123, %Flow3517 ], [ poison, %2084 ]
  %2086 = phi i64 [ %2124, %Flow3517 ], [ poison, %2084 ]
  %2087 = phi i32 [ %2125, %Flow3517 ], [ poison, %2084 ]
  %retval.sroa.6.0.i.i.i1571 = phi i32 [ %2128, %Flow3517 ], [ undef, %2084 ]
  %retval.sroa.2.0.i.i.i1573 = phi i64 [ %2127, %Flow3517 ], [ undef, %2084 ]
  %retval.sroa.8.0.i.i.i1574 = phi i32 [ %2126, %Flow3517 ], [ undef, %2084 ]
  %index.0.i.i.i1575 = phi i32 [ %2130, %Flow3517 ], [ 0, %2084 ]
  %2088 = load i32, ptr addrspace(1) @__llvm_rpc_client, align 8, !tbaa !68
  %cmp.not.i.i.i1576 = icmp ult i32 %index.0.i.i.i1575, %2088
  %spec.store.select.i.i.i1577 = select i1 %cmp.not.i.i.i1576, i32 %index.0.i.i.i1575, i32 0
  %2089 = tail call noundef i64 @llvm.amdgcn.ballot.i64(i1 true)
  %2090 = and i32 %spec.store.select.i.i.i1577, 31
  %2091 = lshr i64 %2089, %sh_prom.i.i.i.i
  %2092 = trunc i64 %2091 to i32
  %conv4.i.i.i.i.i1581 = and i32 %2092, 1
  %shl.i.i.i.i.i1582 = shl nuw i32 %conv4.i.i.i.i.i1581, %2090
  %2093 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2631, !amdgpu.uniform !42

2094:                                             ; preds = %ComputeEnd2632
  %sunkaddr3335 = getelementptr inbounds i8, ptr addrspace(1) %3841, i64 40
  %2095 = atomicrmw or ptr addrspace(1) %sunkaddr3335, i32 %.lcssa4036 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %2096, !amdgpu.uniform !42

2096:                                             ; preds = %2094, %ComputeEnd2632
  %2097 = phi i32 [ %2095, %2094 ], [ poison, %ComputeEnd2632 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %3850)
  %2098 = call i32 @llvm.amdgcn.readfirstlane.i32(i32 %2097)
  %2099 = or i32 %2098, %.lcssa4037
  %shl5.i.i.i.i.i1583 = shl nuw i32 1, %2090
  %and.i.i.i.i.i1584 = and i32 %shl5.i.i.i.i.i1583, %2099
  %tobool3.i.i.i.i1585 = icmp ne i32 %and.i.i.i.i.i1584, 0
  %2100 = tail call i64 @llvm.amdgcn.ballot.i64(i1 %tobool3.i.i.i.i1585)
  %and.i.i.i.i.i.i1586 = and i64 %2100, %2089
  %cmp.i.not.i.i.i1587 = icmp ne i64 %2089, %and.i.i.i.i.i.i1586
  br i1 %cmp.i.not.i.i.i1587, label %if.then.i.i.i.i1588, label %Flow3516, !amdgpu.uniform !42

if.then.i.i.i.i1588:                              ; preds = %2096
  fence syncscope("agent") acquire
  %2101 = load <2 x i64>, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !75
  %2102 = extractelement <2 x i64> %2101, i32 0
  %2103 = inttoptr i64 %2102 to ptr
  %2104 = extractelement <2 x i64> %2101, i32 1
  %2105 = inttoptr i64 %2104 to ptr
  %idxprom.i.i.i.i1589 = zext i32 %spec.store.select.i.i.i1577 to i64
  %arrayidx.i.i.i.i1590 = getelementptr inbounds nuw i32, ptr %2103, i64 %idxprom.i.i.i.i1589, !amdgpu.uniform !42
  %2106 = load atomic i32, ptr %arrayidx.i.i.i.i1590 monotonic, align 4
  %2107 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %2106)
  %arrayidx.i45.i.i.i1591 = getelementptr inbounds nuw i32, ptr %2105, i64 %idxprom.i.i.i.i1589, !amdgpu.uniform !42
  %2108 = load atomic i32, ptr %arrayidx.i45.i.i.i1591 monotonic, align 4
  %2109 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %2108)
  %cmp.i46.not.i.i.i1592 = icmp ne i32 %2107, %2109
  br i1 %cmp.i46.not.i.i.i1592, label %if.then12.i.i.i1593, label %Flow3514, !amdgpu.uniform !42

if.then12.i.i.i1593:                              ; preds = %if.then.i.i.i.i1588
  fence syncscope("agent") release
  %2110 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %2089, i1 true)
  %iszero.i.i.i.i.i.i.i1594 = icmp ne i64 %2089, 0
  %cmp2.i.i.i.i.i.i1595 = icmp eq i64 %2110, %sh_prom.i.i.i.i
  %cmp.i.i.i.i.i.i1596 = select i1 %iszero.i.i.i.i.i.i.i1594, i1 %cmp2.i.i.i.i.i.i1595, i1 false
  %conv4.i.i51.i.i.i1597 = zext i1 %cmp.i.i.i.i.i.i1596 to i32
  %shl.i.i52.i.i.i1598 = shl nuw i32 %conv4.i.i51.i.i.i1597, %2090
  %xor.i.i.i.i.i1599 = xor i32 %shl.i.i52.i.i.i1598, -1
  %2111 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2636, !amdgpu.uniform !42

2112:                                             ; preds = %ComputeEnd2637
  %sunkaddr3336 = getelementptr inbounds i8, ptr addrspace(1) %3841, i64 40
  %2113 = atomicrmw and ptr addrspace(1) %sunkaddr3336, i32 %.lcssa4038 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %2114, !amdgpu.uniform !42

2114:                                             ; preds = %2112, %ComputeEnd2637
  call void @llvm.amdgcn.end.cf.i64(i64 %3868)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %Flow3514, !amdgpu.uniform !42

if.end14.i.i.i1657:                               ; preds = %Flow3514
  %2115 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %2089, i1 true)
  %iszero.i.i.i.i.i.i1658 = icmp ne i64 %2089, 0
  %cmp2.i.i.i.i.i1659 = icmp eq i64 %2115, %sh_prom.i.i.i.i
  %cmp.i.i.i.i.i1660 = select i1 %iszero.i.i.i.i.i.i1658, i1 %cmp2.i.i.i.i.i1659, i1 false
  %2116 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.i.i.i.i.i1660)
  %2117 = extractvalue { i1, i64 } %2116, 0
  %2118 = extractvalue { i1, i64 } %2116, 1
  br i1 %2117, label %if.then16.i.i.i1662, label %if.end22.i.i.i1661

if.then16.i.i.i1662:                              ; preds = %if.end14.i.i.i1657
  %2119 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !76
  %opcode.i.i.i1663 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %2119, i64 %idxprom.i.i.i.i1589, i32 1
  store i32 1660944387, ptr %opcode.i.i.i1663, align 8, !tbaa !77
  %arrayidx21.i.i.i1664 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %2119, i64 %idxprom.i.i.i.i1589
  store i64 %2089, ptr %arrayidx21.i.i.i1664, align 8, !tbaa !80
  br label %if.end22.i.i.i1661, !amdgpu.uniform !42

Flow3515:                                         ; preds = %if.end22.i.i.i1661, %Flow3514
  %2120 = phi i32 [ %2107, %if.end22.i.i.i1661 ], [ %retval.sroa.8.0.i.i.i1574, %Flow3514 ]
  %2121 = phi i64 [ %2089, %if.end22.i.i.i1661 ], [ %retval.sroa.2.0.i.i.i1573, %Flow3514 ]
  %2122 = phi i32 [ %spec.store.select.i.i.i1577, %if.end22.i.i.i1661 ], [ %retval.sroa.6.0.i.i.i1571, %Flow3514 ]
  br label %cleanup26.i.i.i1600, !amdgpu.uniform !42

if.end22.i.i.i1661:                               ; preds = %if.then16.i.i.i1662, %if.end14.i.i.i1657
  call void @llvm.amdgcn.end.cf.i64(i64 %2118)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %Flow3515, !amdgpu.uniform !42

Flow3516:                                         ; preds = %cleanup26.i.i.i1600, %2096
  %2123 = phi i32 [ %2122, %cleanup26.i.i.i1600 ], [ %2085, %2096 ]
  %2124 = phi i64 [ %2121, %cleanup26.i.i.i1600 ], [ %2086, %2096 ]
  %2125 = phi i32 [ %2120, %cleanup26.i.i.i1600 ], [ %2087, %2096 ]
  %2126 = phi i32 [ %2120, %cleanup26.i.i.i1600 ], [ %retval.sroa.8.0.i.i.i1574, %2096 ]
  %2127 = phi i64 [ %2121, %cleanup26.i.i.i1600 ], [ %retval.sroa.2.0.i.i.i1573, %2096 ]
  %2128 = phi i32 [ %2122, %cleanup26.i.i.i1600 ], [ %retval.sroa.6.0.i.i.i1571, %2096 ]
  %2129 = phi i1 [ %cmp.i46.not.i.i.i1592, %cleanup26.i.i.i1600 ], [ true, %2096 ]
  br i1 %2129, label %for.inc.i.i.i1605, label %Flow3517, !amdgpu.uniform !42

cleanup26.i.i.i1600:                              ; preds = %Flow3515
  br label %Flow3516, !amdgpu.uniform !42

for.inc.i.i.i1605:                                ; preds = %Flow3516
  %inc.i.i.i1610 = add i32 %spec.store.select.i.i.i1577, 1
  br label %Flow3517, !amdgpu.uniform !42

Flow3517:                                         ; preds = %for.inc.i.i.i1605, %Flow3516
  %2130 = phi i32 [ %inc.i.i.i1610, %for.inc.i.i.i1605 ], [ poison, %Flow3516 ]
  %2131 = phi i1 [ false, %for.inc.i.i.i1605 ], [ true, %Flow3516 ]
  br i1 %2131, label %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1611, label %for.cond.i.i.i1570, !amdgpu.uniform !42

_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1611: ; preds = %Flow3517
  %.lcssa4041 = phi i32 [ %2123, %Flow3517 ]
  %.lcssa4040 = phi i64 [ %2124, %Flow3517 ]
  %.lcssa4039 = phi i32 [ %2125, %Flow3517 ]
  fence acquire
  %2132 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %mul.i.i.i.i.i.i1612 = shl i32 %.lcssa4041, 6
  %idxprom.i19.i.i.i.i.i1613 = zext i32 %mul.i.i.i.i.i.i1612 to i64
  %arrayidx.i20.i.i.i.i.i1614 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %2132, i64 %idxprom.i19.i.i.i.i.i1613
  %arrayidx.i22.i.i.i.i.i1615 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %arrayidx.i20.i.i.i.i.i1614, i64 %sh_prom.i.i.i.i
  store i64 1, ptr %arrayidx.i22.i.i.i.i.i1615, align 8, !tbaa !82
  %arrayidx6.i.i.i.i.i.i.i1616 = getelementptr inbounds nuw i8, ptr %arrayidx.i22.i.i.i.i.i1615, i64 8
  store i8 10, ptr %arrayidx6.i.i.i.i.i.i.i1616, align 8
  %tobool.not.i.i.i.i.i.i1617 = icmp eq i32 %.lcssa4039, 0
  %conv.i.i.i.i.i.i1618 = zext i1 %tobool.not.i.i.i.i.i.i1617 to i32
  fence release
  %2133 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %idxprom.i23.i.i.i.i.i1619 = zext i32 %.lcssa4041 to i64
  %arrayidx.i24.i.i.i.i.i1620 = getelementptr inbounds nuw i32, ptr %2133, i64 %idxprom.i23.i.i.i.i.i1619
  store atomic i32 %conv.i.i.i.i.i.i1618, ptr %arrayidx.i24.i.i.i.i.i1620 monotonic, align 4
  %2134 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !84
  %arrayidx.i.i5.i.i1621 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %2134, i64 %idxprom.i23.i.i.i.i.i1619, !amdgpu.uniform !42
  %2135 = load i64, ptr %arrayidx.i.i5.i.i1621, align 8, !tbaa !80
  br label %while.cond.i.i.i.i1622, !amdgpu.uniform !42

while.cond.i.i.i.i1622:                           ; preds = %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1611, %Flow3513
  %phi.broken3967 = phi i64 [ 0, %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1611 ], [ %2150, %Flow3513 ]
  %port.sroa.37.0.i.i1623 = phi i32 [ %2148, %Flow3513 ], [ %conv.i.i.i.i.i.i1618, %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1611 ]
  %idx.0.i.i.i.i1624 = phi i64 [ %2147, %Flow3513 ], [ 56, %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1611 ]
  %cmp.i.i.i.i1625 = icmp eq i64 %idx.0.i.i.i.i1624, 0
  %2136 = tail call i64 @llvm.amdgcn.ballot.i64(i1 %cmp.i.i.i.i1625)
  %and.i.i.i.i6.i.i1626 = and i64 %2136, %2135
  %tobool.not.i.i.i.i1627 = icmp ne i64 %and.i.i.i.i6.i.i1626, 0
  %2137 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i22.i.i.i1628 = getelementptr inbounds nuw i32, ptr %2137, i64 %idxprom.i23.i.i.i.i.i1619, !amdgpu.uniform !42
  %2138 = load atomic i32, ptr %arrayidx.i22.i.i.i1628 monotonic, align 4
  %2139 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %tobool.not.i.i.i.i1627)
  %2140 = extractvalue { i1, i64 } %2139, 0
  %2141 = extractvalue { i1, i64 } %2139, 1
  br i1 %2140, label %cond.false.i8.i.i.i.i1629, label %Flow3513

cond.false.i8.i.i.i.i1629:                        ; preds = %while.cond.i.i.i.i1622
  %2142 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %2138)
  %2143 = icmp ne i32 %2142, %port.sroa.37.0.i.i1623
  br i1 %2143, label %while.body.i.i21.i.i.i.i1630.preheader, label %Flow3512, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i1630.preheader:           ; preds = %cond.false.i8.i.i.i.i1629
  br label %while.body.i.i21.i.i.i.i1630, !amdgpu.uniform !42

Flow3512:                                         ; preds = %Flow3510, %cond.false.i8.i.i.i.i1629
  br label %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i1633, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i1630:                     ; preds = %while.body.i.i21.i.i.i.i1630.preheader, %while.body.i.i21.i.i.i.i1630
  tail call void @llvm.amdgcn.s.sleep(i32 2)
  %2144 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i22.i.i.i.i1631 = getelementptr inbounds nuw i32, ptr %2144, i64 %idxprom.i23.i.i.i.i.i1619, !amdgpu.uniform !42
  %2145 = load atomic i32, ptr %arrayidx.i.i.i22.i.i.i.i1631 monotonic, align 4
  %2146 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %2145)
  %cmp.i.not.i.i23.i.i.i.i1632 = icmp eq i32 %2146, %port.sroa.37.0.i.i1623
  br i1 %cmp.i.not.i.i23.i.i.i.i1632, label %Flow3510, label %while.body.i.i21.i.i.i.i1630, !amdgpu.uniform !42

Flow3510:                                         ; preds = %while.body.i.i21.i.i.i.i1630
  br label %Flow3512, !amdgpu.uniform !42

Flow3513:                                         ; preds = %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1634, %while.cond.i.i.i.i1622
  %2147 = phi i64 [ %add.i.i.i.i1637, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1634 ], [ poison, %while.cond.i.i.i.i1622 ]
  %2148 = phi i32 [ %conv.i.i26.i.i.i.i1635, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1634 ], [ poison, %while.cond.i.i.i.i1622 ]
  %2149 = phi i1 [ false, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1634 ], [ true, %while.cond.i.i.i.i1622 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %2141)
  %2150 = call i64 @llvm.amdgcn.if.break.i64(i1 %2149, i64 %phi.broken3967)
  %2151 = call i1 @llvm.amdgcn.loop.i64(i64 %2150)
  br i1 %2151, label %cond.false.i.i.i1643, label %while.cond.i.i.i.i1622

_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i1633: ; preds = %Flow3512
  fence acquire
  br i1 %cmp.i.i.i.i1625, label %if.then.i.i.i.i.i.i.i1638, label %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1634, !amdgpu.uniform !42

if.then.i.i.i.i.i.i.i1638:                        ; preds = %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i1633
  %2152 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %arrayidx.i20.i35.i.i.i.i1641 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %2152, i64 %idxprom.i19.i.i.i.i.i1613
  %arrayidx.i22.i38.i.i.i.i1642 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %arrayidx.i20.i35.i.i.i.i1641, i64 %sh_prom.i.i.i.i
  store i8 10, ptr %arrayidx.i22.i38.i.i.i.i1642, align 1
  br label %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1634, !amdgpu.uniform !42

_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1634: ; preds = %if.then.i.i.i.i.i.i.i1638, %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i1633
  %conv.i.i26.i.i.i.i1635 = xor i32 %port.sroa.37.0.i.i1623, 1
  fence release
  %2153 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %arrayidx.i24.i29.i.i.i.i1636 = getelementptr inbounds nuw i32, ptr %2153, i64 %idxprom.i23.i.i.i.i.i1619
  store atomic i32 %conv.i.i26.i.i.i.i1635, ptr %arrayidx.i24.i29.i.i.i.i1636 monotonic, align 4
  %add.i.i.i.i1637 = add i64 %idx.0.i.i.i.i1624, 64
  br label %Flow3513, !amdgpu.uniform !42

cond.false.i.i.i1643:                             ; preds = %Flow3513
  %.lcssa4035 = phi i64 [ %2150, %Flow3513 ]
  %port.sroa.37.0.i.i1623.lcssa = phi i32 [ %port.sroa.37.0.i.i1623, %Flow3513 ]
  %.lcssa4034 = phi i32 [ %2138, %Flow3513 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa4035)
  %2154 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %.lcssa4034)
  %2155 = icmp ne i32 %2154, %port.sroa.37.0.i.i1623.lcssa
  %2156 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %2155)
  %2157 = extractvalue { i1, i64 } %2156, 0
  %2158 = extractvalue { i1, i64 } %2156, 1
  br i1 %2157, label %while.body.i.i.i.i1644.preheader, label %Flow3509

while.body.i.i.i.i1644.preheader:                 ; preds = %cond.false.i.i.i1643
  br label %while.body.i.i.i.i1644, !amdgpu.uniform !42

Flow3509:                                         ; preds = %Flow3507, %cond.false.i.i.i1643
  call void @llvm.amdgcn.end.cf.i64(i64 %2158)
  br label %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIcEERS1_T_.exit1665, !amdgpu.uniform !42

while.body.i.i.i.i1644:                           ; preds = %while.body.i.i.i.i1644.preheader, %while.body.i.i.i.i1644
  %phi.broken3968 = phi i64 [ 0, %while.body.i.i.i.i1644.preheader ], [ %2162, %while.body.i.i.i.i1644 ]
  tail call void @llvm.amdgcn.s.sleep(i32 2)
  %2159 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i8.i.i1645 = getelementptr inbounds nuw i32, ptr %2159, i64 %idxprom.i23.i.i.i.i.i1619, !amdgpu.uniform !42
  %2160 = load atomic i32, ptr %arrayidx.i.i.i8.i.i1645 monotonic, align 4
  %2161 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %2160)
  %cmp.i.not.i.i.i.i1646 = icmp eq i32 %2161, %port.sroa.37.0.i.i1623.lcssa
  %2162 = call i64 @llvm.amdgcn.if.break.i64(i1 %cmp.i.not.i.i.i.i1646, i64 %phi.broken3968)
  %2163 = call i1 @llvm.amdgcn.loop.i64(i64 %2162)
  br i1 %2163, label %Flow3507, label %while.body.i.i.i.i1644

Flow3507:                                         ; preds = %while.body.i.i.i.i1644
  %.lcssa4033 = phi i64 [ %2162, %while.body.i.i.i.i1644 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa4033)
  br label %Flow3509, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIcEERS1_T_.exit1665: ; preds = %Flow3509
  fence acquire
  tail call void @llvm.amdgcn.wave.barrier()
  fence syncscope("agent") release
  %2164 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %.lcssa4040, i1 true)
  %iszero.i.i.i.i.i13.i.i1648 = icmp ne i64 %.lcssa4040, 0
  %cmp2.i.i.i.i14.i.i1649 = icmp eq i64 %2164, %sh_prom.i.i.i.i
  %cmp.i.i.i.i15.i.i1650 = select i1 %iszero.i.i.i.i.i13.i.i1648, i1 %cmp2.i.i.i.i14.i.i1649, i1 false
  %2165 = and i32 %.lcssa4041, 31
  %conv4.i.i.i19.i.i1654 = zext i1 %cmp.i.i.i.i15.i.i1650 to i32
  %shl.i.i.i20.i.i1655 = shl nuw i32 %conv4.i.i.i19.i.i1654, %2165
  %xor.i.i.i21.i.i1656 = xor i32 %shl.i.i.i20.i.i1655, -1
  %2166 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2640, !amdgpu.uniform !42

2167:                                             ; preds = %ComputeEnd2641
  %div8.i.i.i16.i.i1651 = lshr i32 %.lcssa4041, 5
  %idxprom.i.i.i17.i.i1652 = zext nneg i32 %div8.i.i.i16.i.i1651 to i64
  %2168 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i17.i.i1652
  %sunkaddr3337 = getelementptr inbounds i8, ptr addrspace(1) %2168, i64 40
  %2169 = atomicrmw and ptr addrspace(1) %sunkaddr3337, i32 %.lcssa4032 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %2170, !amdgpu.uniform !42

2170:                                             ; preds = %2167, %ComputeEnd2641
  call void @llvm.amdgcn.end.cf.i64(i64 %3885)
  tail call void @llvm.amdgcn.wave.barrier()
  br i1 %cmp2.i.i.i35.i, label %if.then3.i.i.i87.i, label %if.end4.i.i.i76.i, !amdgpu.uniform !42

if.then3.i.i.i87.i:                               ; preds = %2170
  %mul.i.i.i88.i = mul nuw i64 %inc.i.i.i32.i, 11
  %div18.i.i.i89.i = lshr i64 %mul.i.i.i88.i, 3
  br label %if.end4.i.i.i76.i, !amdgpu.uniform !42

if.end4.i.i.i76.i:                                ; preds = %if.then3.i.i.i87.i, %2170
  %new_capacity.addr.0.i.i.i77.i = phi i64 [ %div18.i.i.i89.i, %if.then3.i.i.i87.i ], [ %inc.i.i.i32.i, %2170 ]
  %sub.i.i1666 = add i64 %new_capacity.addr.0.i.i.i77.i, 7
  %div2.i.i1667 = and i64 %sub.i.i1666, -8
  %2171 = load ptr, ptr addrspace(1) @_ZN12_GLOBAL__N_13ptrE, align 8, !tbaa !55
  %add.ptr.i.i1668 = getelementptr inbounds nuw i8, ptr %2171, i64 %div2.i.i1667
  store ptr %add.ptr.i.i1668, ptr addrspace(1) @_ZN12_GLOBAL__N_13ptrE, align 8, !tbaa !55
  br i1 %597, label %loop-memcpy-expansion2285.preheader, label %Flow3506, !amdgpu.uniform !42

loop-memcpy-expansion2285.preheader:              ; preds = %if.end4.i.i.i76.i
  br label %loop-memcpy-expansion2285, !amdgpu.uniform !42

Flow3506:                                         ; preds = %Flow3505, %if.end4.i.i.i76.i
  br label %loop-memcpy-residual-header2288, !amdgpu.uniform !42

loop-memcpy-expansion2285:                        ; preds = %loop-memcpy-expansion2285.preheader, %loop-memcpy-expansion2285
  %loop-index2286 = phi i64 [ %2175, %loop-memcpy-expansion2285 ], [ 0, %loop-memcpy-expansion2285.preheader ]
  %2172 = getelementptr i8, ptr %3, i64 %loop-index2286, !amdgpu.uniform !42
  %2173 = load <4 x i32>, ptr %2172, align 1
  %2174 = getelementptr i8, ptr %2171, i64 %loop-index2286
  store <4 x i32> %2173, ptr %2174, align 1
  %2175 = add i64 %loop-index2286, 16
  %2176 = icmp uge i64 %2175, %596
  br i1 %2176, label %Flow3505, label %loop-memcpy-expansion2285, !amdgpu.uniform !42

Flow3503:                                         ; preds = %Flow3502, %loop-memcpy-residual-header2288
  br label %post-loop-memcpy-expansion2284, !amdgpu.uniform !42

loop-memcpy-residual2287:                         ; preds = %loop-memcpy-residual2287.preheader, %loop-memcpy-residual2287
  %residual-loop-index2289 = phi i64 [ %2178, %loop-memcpy-residual2287 ], [ 0, %loop-memcpy-residual2287.preheader ]
  %scevgep2926 = getelementptr i8, ptr %scevgep2925, i64 %residual-loop-index2289, !amdgpu.uniform !42
  %2177 = load i8, ptr %scevgep2926, align 1
  %scevgep2924 = getelementptr i8, ptr %scevgep2923, i64 %residual-loop-index2289
  store i8 %2177, ptr %scevgep2924, align 1
  %2178 = add i64 %residual-loop-index2289, 1
  %2179 = icmp uge i64 %2178, %595
  br i1 %2179, label %Flow3502, label %loop-memcpy-residual2287, !amdgpu.uniform !42

Flow3502:                                         ; preds = %loop-memcpy-residual2287
  br label %Flow3503, !amdgpu.uniform !42

post-loop-memcpy-expansion2284:                   ; preds = %Flow3503
  %sub.ptr.lhs.cast.i.i1669 = ptrtoint ptr %add.ptr.i.i1668 to i64
  %sub.ptr.sub.i.i1670 = sub i64 %sub.ptr.lhs.cast.i.i1669, ptrtoint (ptr addrspacecast (ptr addrspace(1) @_ZN12_GLOBAL__N_16memoryE to ptr) to i64)
  %cmp.i.i1671 = icmp ugt i64 %sub.ptr.sub.i.i1670, 65335
  %cond.i.i1672 = select i1 %cmp.i.i1671, ptr null, ptr %2171
  %cmp.not.i10.i.i84.i = icmp ne ptr %cond.i.i1672, addrspacecast (ptr addrspace(4) @_ZN22__llvm_libc_22_0_0_git3cpp6string14NULL_CHARACTERE to ptr)
  br i1 %cmp.not.i10.i.i84.i, label %if.then.i.i.i85.i, label %for.cond.i.i.i1674.preheader, !amdgpu.uniform !42

if.then.i.i.i85.i:                                ; preds = %post-loop-memcpy-expansion2284
  %arrayidx.i.i.i86.i = getelementptr inbounds nuw i8, ptr %2171, i64 %spec.select.i
  store i8 0, ptr %arrayidx.i.i.i86.i, align 1, !tbaa !57
  br label %for.cond.i.i.i1674.preheader, !amdgpu.uniform !42

for.cond.i.i.i1674.preheader:                     ; preds = %if.then.i.i.i85.i, %post-loop-memcpy-expansion2284
  br label %for.cond.i.i.i1674, !amdgpu.uniform !42

for.cond.i.i.i1674:                               ; preds = %for.cond.i.i.i1674.preheader, %Flow3500
  %2180 = phi i32 [ %2218, %Flow3500 ], [ poison, %for.cond.i.i.i1674.preheader ]
  %2181 = phi i64 [ %2219, %Flow3500 ], [ poison, %for.cond.i.i.i1674.preheader ]
  %2182 = phi i32 [ %2220, %Flow3500 ], [ poison, %for.cond.i.i.i1674.preheader ]
  %retval.sroa.6.0.i.i.i1675 = phi i32 [ %2223, %Flow3500 ], [ undef, %for.cond.i.i.i1674.preheader ]
  %retval.sroa.2.0.i.i.i1677 = phi i64 [ %2222, %Flow3500 ], [ undef, %for.cond.i.i.i1674.preheader ]
  %retval.sroa.8.0.i.i.i1678 = phi i32 [ %2221, %Flow3500 ], [ undef, %for.cond.i.i.i1674.preheader ]
  %index.0.i.i.i1679 = phi i32 [ %2225, %Flow3500 ], [ 0, %for.cond.i.i.i1674.preheader ]
  %2183 = load i32, ptr addrspace(1) @__llvm_rpc_client, align 8, !tbaa !68
  %cmp.not.i.i.i1680 = icmp ult i32 %index.0.i.i.i1679, %2183
  %spec.store.select.i.i.i1681 = select i1 %cmp.not.i.i.i1680, i32 %index.0.i.i.i1679, i32 0
  %2184 = tail call noundef i64 @llvm.amdgcn.ballot.i64(i1 true)
  %2185 = and i32 %spec.store.select.i.i.i1681, 31
  %2186 = lshr i64 %2184, %sh_prom.i.i.i.i
  %2187 = trunc i64 %2186 to i32
  %conv4.i.i.i.i.i1685 = and i32 %2187, 1
  %shl.i.i.i.i.i1686 = shl nuw i32 %conv4.i.i.i.i.i1685, %2185
  %2188 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2644, !amdgpu.uniform !42

2189:                                             ; preds = %ComputeEnd2645
  %sunkaddr3338 = getelementptr inbounds i8, ptr addrspace(1) %3895, i64 40
  %2190 = atomicrmw or ptr addrspace(1) %sunkaddr3338, i32 %.lcssa4026 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %2191, !amdgpu.uniform !42

2191:                                             ; preds = %2189, %ComputeEnd2645
  %2192 = phi i32 [ %2190, %2189 ], [ poison, %ComputeEnd2645 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %3904)
  %2193 = call i32 @llvm.amdgcn.readfirstlane.i32(i32 %2192)
  %2194 = or i32 %2193, %.lcssa4027
  %shl5.i.i.i.i.i1687 = shl nuw i32 1, %2185
  %and.i.i.i.i.i1688 = and i32 %shl5.i.i.i.i.i1687, %2194
  %tobool3.i.i.i.i1689 = icmp ne i32 %and.i.i.i.i.i1688, 0
  %2195 = tail call i64 @llvm.amdgcn.ballot.i64(i1 %tobool3.i.i.i.i1689)
  %and.i.i.i.i.i.i1690 = and i64 %2195, %2184
  %cmp.i.not.i.i.i1691 = icmp ne i64 %2184, %and.i.i.i.i.i.i1690
  br i1 %cmp.i.not.i.i.i1691, label %if.then.i.i.i.i1692, label %Flow3499, !amdgpu.uniform !42

if.then.i.i.i.i1692:                              ; preds = %2191
  fence syncscope("agent") acquire
  %2196 = load <2 x i64>, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !75
  %2197 = extractelement <2 x i64> %2196, i32 0
  %2198 = inttoptr i64 %2197 to ptr
  %2199 = extractelement <2 x i64> %2196, i32 1
  %2200 = inttoptr i64 %2199 to ptr
  %idxprom.i.i.i.i1693 = zext i32 %spec.store.select.i.i.i1681 to i64
  %arrayidx.i.i.i.i1694 = getelementptr inbounds nuw i32, ptr %2198, i64 %idxprom.i.i.i.i1693, !amdgpu.uniform !42
  %2201 = load atomic i32, ptr %arrayidx.i.i.i.i1694 monotonic, align 4
  %2202 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %2201)
  %arrayidx.i45.i.i.i1695 = getelementptr inbounds nuw i32, ptr %2200, i64 %idxprom.i.i.i.i1693, !amdgpu.uniform !42
  %2203 = load atomic i32, ptr %arrayidx.i45.i.i.i1695 monotonic, align 4
  %2204 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %2203)
  %cmp.i46.not.i.i.i1696 = icmp ne i32 %2202, %2204
  br i1 %cmp.i46.not.i.i.i1696, label %if.then12.i.i.i1697, label %Flow3497, !amdgpu.uniform !42

if.then12.i.i.i1697:                              ; preds = %if.then.i.i.i.i1692
  fence syncscope("agent") release
  %2205 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %2184, i1 true)
  %iszero.i.i.i.i.i.i.i1698 = icmp ne i64 %2184, 0
  %cmp2.i.i.i.i.i.i1699 = icmp eq i64 %2205, %sh_prom.i.i.i.i
  %cmp.i.i.i.i.i.i1700 = select i1 %iszero.i.i.i.i.i.i.i1698, i1 %cmp2.i.i.i.i.i.i1699, i1 false
  %conv4.i.i51.i.i.i1701 = zext i1 %cmp.i.i.i.i.i.i1700 to i32
  %shl.i.i52.i.i.i1702 = shl nuw i32 %conv4.i.i51.i.i.i1701, %2185
  %xor.i.i.i.i.i1703 = xor i32 %shl.i.i52.i.i.i1702, -1
  %2206 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2649, !amdgpu.uniform !42

2207:                                             ; preds = %ComputeEnd2650
  %sunkaddr3339 = getelementptr inbounds i8, ptr addrspace(1) %3895, i64 40
  %2208 = atomicrmw and ptr addrspace(1) %sunkaddr3339, i32 %.lcssa4028 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %2209, !amdgpu.uniform !42

2209:                                             ; preds = %2207, %ComputeEnd2650
  call void @llvm.amdgcn.end.cf.i64(i64 %3922)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %Flow3497, !amdgpu.uniform !42

if.end14.i.i.i1763:                               ; preds = %Flow3497
  %2210 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %2184, i1 true)
  %iszero.i.i.i.i.i.i1764 = icmp ne i64 %2184, 0
  %cmp2.i.i.i.i.i1765 = icmp eq i64 %2210, %sh_prom.i.i.i.i
  %cmp.i.i.i.i.i1766 = select i1 %iszero.i.i.i.i.i.i1764, i1 %cmp2.i.i.i.i.i1765, i1 false
  %2211 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.i.i.i.i.i1766)
  %2212 = extractvalue { i1, i64 } %2211, 0
  %2213 = extractvalue { i1, i64 } %2211, 1
  br i1 %2212, label %if.then16.i.i.i1768, label %if.end22.i.i.i1767

if.then16.i.i.i1768:                              ; preds = %if.end14.i.i.i1763
  %2214 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !76
  %opcode.i.i.i1769 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %2214, i64 %idxprom.i.i.i.i1693, i32 1
  store i32 1660944387, ptr %opcode.i.i.i1769, align 8, !tbaa !77
  %arrayidx21.i.i.i1770 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %2214, i64 %idxprom.i.i.i.i1693
  store i64 %2184, ptr %arrayidx21.i.i.i1770, align 8, !tbaa !80
  br label %if.end22.i.i.i1767, !amdgpu.uniform !42

Flow3498:                                         ; preds = %if.end22.i.i.i1767, %Flow3497
  %2215 = phi i32 [ %2202, %if.end22.i.i.i1767 ], [ %retval.sroa.8.0.i.i.i1678, %Flow3497 ]
  %2216 = phi i64 [ %2184, %if.end22.i.i.i1767 ], [ %retval.sroa.2.0.i.i.i1677, %Flow3497 ]
  %2217 = phi i32 [ %spec.store.select.i.i.i1681, %if.end22.i.i.i1767 ], [ %retval.sroa.6.0.i.i.i1675, %Flow3497 ]
  br label %cleanup26.i.i.i1704, !amdgpu.uniform !42

if.end22.i.i.i1767:                               ; preds = %if.then16.i.i.i1768, %if.end14.i.i.i1763
  call void @llvm.amdgcn.end.cf.i64(i64 %2213)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %Flow3498, !amdgpu.uniform !42

Flow3499:                                         ; preds = %cleanup26.i.i.i1704, %2191
  %2218 = phi i32 [ %2217, %cleanup26.i.i.i1704 ], [ %2180, %2191 ]
  %2219 = phi i64 [ %2216, %cleanup26.i.i.i1704 ], [ %2181, %2191 ]
  %2220 = phi i32 [ %2215, %cleanup26.i.i.i1704 ], [ %2182, %2191 ]
  %2221 = phi i32 [ %2215, %cleanup26.i.i.i1704 ], [ %retval.sroa.8.0.i.i.i1678, %2191 ]
  %2222 = phi i64 [ %2216, %cleanup26.i.i.i1704 ], [ %retval.sroa.2.0.i.i.i1677, %2191 ]
  %2223 = phi i32 [ %2217, %cleanup26.i.i.i1704 ], [ %retval.sroa.6.0.i.i.i1675, %2191 ]
  %2224 = phi i1 [ %cmp.i46.not.i.i.i1696, %cleanup26.i.i.i1704 ], [ true, %2191 ]
  br i1 %2224, label %for.inc.i.i.i1709, label %Flow3500, !amdgpu.uniform !42

cleanup26.i.i.i1704:                              ; preds = %Flow3498
  br label %Flow3499, !amdgpu.uniform !42

for.inc.i.i.i1709:                                ; preds = %Flow3499
  %inc.i.i.i1714 = add i32 %spec.store.select.i.i.i1681, 1
  br label %Flow3500, !amdgpu.uniform !42

Flow3500:                                         ; preds = %for.inc.i.i.i1709, %Flow3499
  %2225 = phi i32 [ %inc.i.i.i1714, %for.inc.i.i.i1709 ], [ poison, %Flow3499 ]
  %2226 = phi i1 [ false, %for.inc.i.i.i1709 ], [ true, %Flow3499 ]
  br i1 %2226, label %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1715, label %for.cond.i.i.i1674, !amdgpu.uniform !42

_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1715: ; preds = %Flow3500
  %.lcssa4031 = phi i32 [ %2218, %Flow3500 ]
  %.lcssa4030 = phi i64 [ %2219, %Flow3500 ]
  %.lcssa4029 = phi i32 [ %2220, %Flow3500 ]
  fence acquire
  %2227 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %mul.i.i.i.i.i.i1716 = shl i32 %.lcssa4031, 6
  %idxprom.i19.i.i.i.i.i1717 = zext i32 %mul.i.i.i.i.i.i1716 to i64
  %arrayidx.i20.i.i.i.i.i1718 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %2227, i64 %idxprom.i19.i.i.i.i.i1717
  %arrayidx.i22.i.i.i.i.i1719 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %arrayidx.i20.i.i.i.i.i1718, i64 %sh_prom.i.i.i.i
  store i64 %spec.select.i, ptr %arrayidx.i22.i.i.i.i.i1719, align 8, !tbaa !82
  br i1 %656, label %loop-memcpy-expansion2207.preheader, label %Flow3496, !amdgpu.uniform !42

loop-memcpy-expansion2207.preheader:              ; preds = %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1715
  %2228 = add nuw nsw i64 %75, 8
  %scevgep2919 = getelementptr i8, ptr %2227, i64 %2228
  %2229 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i1717, 6
  %scevgep2920 = getelementptr i8, ptr %scevgep2919, i64 %2229
  br label %loop-memcpy-expansion2207, !amdgpu.uniform !42

Flow3496:                                         ; preds = %Flow3495, %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1715
  br label %loop-memcpy-residual-header2210, !amdgpu.uniform !42

loop-memcpy-expansion2207:                        ; preds = %loop-memcpy-expansion2207.preheader, %loop-memcpy-expansion2207
  %loop-index2208 = phi i64 [ %2232, %loop-memcpy-expansion2207 ], [ 0, %loop-memcpy-expansion2207.preheader ]
  %2230 = getelementptr i8, ptr %2171, i64 %loop-index2208, !amdgpu.uniform !42
  %2231 = load <4 x i32>, ptr %2230, align 1
  %scevgep2921 = getelementptr i8, ptr %scevgep2920, i64 %loop-index2208
  store <4 x i32> %2231, ptr %scevgep2921, align 1
  %2232 = add i64 %loop-index2208, 16
  %2233 = icmp uge i64 %2232, %655
  br i1 %2233, label %Flow3495, label %loop-memcpy-expansion2207, !amdgpu.uniform !42

Flow3493:                                         ; preds = %Flow3490, %Flow3492
  %2234 = phi i64 [ %2767, %Flow3490 ], [ %2761, %Flow3492 ]
  br label %post-loop-memcpy-expansion2206, !amdgpu.uniform !42

loop-memcpy-residual2209:                         ; preds = %loop-memcpy-residual2209.preheader, %loop-memcpy-residual2209
  %residual-loop-index2211 = phi i64 [ %2236, %loop-memcpy-residual2209 ], [ 0, %loop-memcpy-residual2209.preheader ]
  %scevgep2918 = getelementptr i8, ptr %scevgep2917, i64 %residual-loop-index2211, !amdgpu.uniform !42
  %2235 = load i8, ptr %scevgep2918, align 1
  %scevgep2916 = getelementptr i8, ptr %scevgep2915, i64 %residual-loop-index2211
  store i8 %2235, ptr %scevgep2916, align 1
  %2236 = add i64 %residual-loop-index2211, 1
  %2237 = icmp uge i64 %2236, %654
  br i1 %2237, label %Flow3490, label %loop-memcpy-residual2209, !amdgpu.uniform !42

Flow3490:                                         ; preds = %loop-memcpy-residual2209
  br label %Flow3493, !amdgpu.uniform !42

post-loop-memcpy-expansion2206:                   ; preds = %Flow3493
  %tobool.not.i.i.i.i.i.i1722 = icmp eq i32 %.lcssa4029, 0
  %conv.i.i.i.i.i.i1723 = zext i1 %tobool.not.i.i.i.i.i.i1722 to i32
  fence release
  %2238 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %idxprom.i23.i.i.i.i.i1724 = zext i32 %.lcssa4031 to i64
  %arrayidx.i24.i.i.i.i.i1725 = getelementptr inbounds nuw i32, ptr %2238, i64 %idxprom.i23.i.i.i.i.i1724
  store atomic i32 %conv.i.i.i.i.i.i1723, ptr %arrayidx.i24.i.i.i.i.i1725 monotonic, align 4
  %2239 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !84
  %arrayidx.i.i5.i.i1726 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %2239, i64 %idxprom.i23.i.i.i.i.i1724, !amdgpu.uniform !42
  %2240 = load i64, ptr %arrayidx.i.i5.i.i1726, align 8, !tbaa !80
  %scevgep2900 = getelementptr i8, ptr %2171, i64 56
  br label %while.cond.i.i.i.i1727, !amdgpu.uniform !42

while.cond.i.i.i.i1727:                           ; preds = %post-loop-memcpy-expansion2206, %Flow3488
  %phi.broken3969 = phi i64 [ 0, %post-loop-memcpy-expansion2206 ], [ %2257, %Flow3488 ]
  %lsr.iv2907 = phi i64 [ %2255, %Flow3488 ], [ %670, %post-loop-memcpy-expansion2206 ]
  %lsr.iv2901 = phi ptr [ %2254, %Flow3488 ], [ %scevgep2900, %post-loop-memcpy-expansion2206 ]
  %port.sroa.37.0.i.i1728 = phi i32 [ %2253, %Flow3488 ], [ %conv.i.i.i.i.i.i1723, %post-loop-memcpy-expansion2206 ]
  %idx.0.i.i.i.i1729 = phi i64 [ %2252, %Flow3488 ], [ 56, %post-loop-memcpy-expansion2206 ]
  %cmp.i.i.i.i1730 = icmp ugt i64 %spec.select.i, %idx.0.i.i.i.i1729
  %2241 = tail call i64 @llvm.amdgcn.ballot.i64(i1 %cmp.i.i.i.i1730)
  %and.i.i.i.i6.i.i1731 = and i64 %2241, %2240
  %tobool.not.i.i.i.i1732 = icmp ne i64 %and.i.i.i.i6.i.i1731, 0
  %2242 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i22.i.i.i1733 = getelementptr inbounds nuw i32, ptr %2242, i64 %idxprom.i23.i.i.i.i.i1724, !amdgpu.uniform !42
  %2243 = load atomic i32, ptr %arrayidx.i22.i.i.i1733 monotonic, align 4
  %2244 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %tobool.not.i.i.i.i1732)
  %2245 = extractvalue { i1, i64 } %2244, 0
  %2246 = extractvalue { i1, i64 } %2244, 1
  br i1 %2245, label %cond.false.i8.i.i.i.i1734, label %Flow3488

cond.false.i8.i.i.i.i1734:                        ; preds = %while.cond.i.i.i.i1727
  %2247 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %2243)
  %2248 = icmp ne i32 %2247, %port.sroa.37.0.i.i1728
  br i1 %2248, label %while.body.i.i21.i.i.i.i1735.preheader, label %Flow3487, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i1735.preheader:           ; preds = %cond.false.i8.i.i.i.i1734
  br label %while.body.i.i21.i.i.i.i1735, !amdgpu.uniform !42

Flow3487:                                         ; preds = %Flow3485, %cond.false.i8.i.i.i.i1734
  br label %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i1738, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i1735:                     ; preds = %while.body.i.i21.i.i.i.i1735.preheader, %while.body.i.i21.i.i.i.i1735
  tail call void @llvm.amdgcn.s.sleep(i32 2)
  %2249 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i22.i.i.i.i1736 = getelementptr inbounds nuw i32, ptr %2249, i64 %idxprom.i23.i.i.i.i.i1724, !amdgpu.uniform !42
  %2250 = load atomic i32, ptr %arrayidx.i.i.i22.i.i.i.i1736 monotonic, align 4
  %2251 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %2250)
  %cmp.i.not.i.i23.i.i.i.i1737 = icmp eq i32 %2251, %port.sroa.37.0.i.i1728
  br i1 %cmp.i.not.i.i23.i.i.i.i1737, label %Flow3485, label %while.body.i.i21.i.i.i.i1735, !amdgpu.uniform !42

Flow3485:                                         ; preds = %while.body.i.i21.i.i.i.i1735
  br label %Flow3487, !amdgpu.uniform !42

Flow3488:                                         ; preds = %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1739, %while.cond.i.i.i.i1727
  %2252 = phi i64 [ %add.i.i.i.i1742, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1739 ], [ poison, %while.cond.i.i.i.i1727 ]
  %2253 = phi i32 [ %conv.i.i26.i.i.i.i1740, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1739 ], [ poison, %while.cond.i.i.i.i1727 ]
  %2254 = phi ptr [ %scevgep2902, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1739 ], [ poison, %while.cond.i.i.i.i1727 ]
  %2255 = phi i64 [ %lsr.iv.next2908, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1739 ], [ poison, %while.cond.i.i.i.i1727 ]
  %2256 = phi i1 [ false, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1739 ], [ true, %while.cond.i.i.i.i1727 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %2246)
  %2257 = call i64 @llvm.amdgcn.if.break.i64(i1 %2256, i64 %phi.broken3969)
  %2258 = call i1 @llvm.amdgcn.loop.i64(i64 %2257)
  br i1 %2258, label %cond.false.i.i.i1749, label %while.cond.i.i.i.i1727

_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i1738: ; preds = %Flow3487
  fence acquire
  br i1 %cmp.i.i.i.i1730, label %if.then.i.i.i.i.i.i.i1743, label %Flow3484, !amdgpu.uniform !42

if.then.i.i.i.i.i.i.i1743:                        ; preds = %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i1738
  %sub.i.i.i.i.i.i.i1745 = sub nuw i64 %spec.select.i, %idx.0.i.i.i.i1729
  %spec.select.i.i.i36.i.i.i.i1746 = tail call i64 @llvm.umin.i64(i64 %sub.i.i.i.i.i.i.i1745, i64 64)
  %2259 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %2260 = and i64 %spec.select.i.i.i36.i.i.i.i1746, 15
  %2261 = sub i64 %spec.select.i.i.i36.i.i.i.i1746, %2260
  %2262 = icmp ne i64 %2261, 0
  br i1 %2262, label %loop-memcpy-expansion2201.preheader, label %Flow3483, !amdgpu.uniform !42

loop-memcpy-expansion2201.preheader:              ; preds = %if.then.i.i.i.i.i.i.i1743
  %arrayidx.i20.i35.i.i.i.i1747 = getelementptr %"struct.rpc::Buffer", ptr %2259, i64 %idxprom.i19.i.i.i.i.i1717
  %arrayidx.i22.i38.i.i.i.i1748 = getelementptr %"struct.rpc::Buffer", ptr %arrayidx.i20.i35.i.i.i.i1747, i64 %sh_prom.i.i.i.i
  br label %loop-memcpy-expansion2201, !amdgpu.uniform !42

Flow3483:                                         ; preds = %Flow3482, %if.then.i.i.i.i.i.i.i1743
  br label %loop-memcpy-residual-header2204, !amdgpu.uniform !42

loop-memcpy-expansion2201:                        ; preds = %loop-memcpy-expansion2201.preheader, %loop-memcpy-expansion2201
  %loop-index2202 = phi i64 [ %2265, %loop-memcpy-expansion2201 ], [ 0, %loop-memcpy-expansion2201.preheader ]
  %scevgep2903 = getelementptr i8, ptr %lsr.iv2901, i64 %loop-index2202, !amdgpu.uniform !42
  %2263 = load <4 x i32>, ptr %scevgep2903, align 1
  %2264 = getelementptr i8, ptr %arrayidx.i22.i38.i.i.i.i1748, i64 %loop-index2202
  store <4 x i32> %2263, ptr %2264, align 1
  %2265 = add i64 %loop-index2202, 16
  %2266 = icmp uge i64 %2265, %2261
  br i1 %2266, label %Flow3482, label %loop-memcpy-expansion2201, !amdgpu.uniform !42

Flow3480:                                         ; preds = %Flow3479, %loop-memcpy-residual-header2204
  br label %Flow3484, !amdgpu.uniform !42

loop-memcpy-residual2203:                         ; preds = %loop-memcpy-residual2203.preheader, %loop-memcpy-residual2203
  %residual-loop-index2205 = phi i64 [ %2268, %loop-memcpy-residual2203 ], [ 0, %loop-memcpy-residual2203.preheader ]
  %scevgep2913 = getelementptr i8, ptr %scevgep2912, i64 %residual-loop-index2205, !amdgpu.uniform !42
  %2267 = load i8, ptr %scevgep2913, align 1
  %scevgep2911 = getelementptr i8, ptr %scevgep2910, i64 %residual-loop-index2205
  store i8 %2267, ptr %scevgep2911, align 1
  %2268 = add i64 %residual-loop-index2205, 1
  %2269 = icmp uge i64 %2268, %2260
  br i1 %2269, label %Flow3479, label %loop-memcpy-residual2203, !amdgpu.uniform !42

Flow3479:                                         ; preds = %loop-memcpy-residual2203
  br label %Flow3480, !amdgpu.uniform !42

_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1739: ; preds = %Flow3484
  %conv.i.i26.i.i.i.i1740 = xor i32 %port.sroa.37.0.i.i1728, 1
  fence release
  %2270 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %arrayidx.i24.i29.i.i.i.i1741 = getelementptr inbounds nuw i32, ptr %2270, i64 %idxprom.i23.i.i.i.i.i1724
  store atomic i32 %conv.i.i26.i.i.i.i1740, ptr %arrayidx.i24.i29.i.i.i.i1741 monotonic, align 4
  %add.i.i.i.i1742 = add i64 %idx.0.i.i.i.i1729, 64
  %scevgep2902 = getelementptr i8, ptr %lsr.iv2901, i64 64
  %lsr.iv.next2908 = add i64 %lsr.iv2907, -64
  br label %Flow3488, !amdgpu.uniform !42

cond.false.i.i.i1749:                             ; preds = %Flow3488
  %.lcssa4025 = phi i64 [ %2257, %Flow3488 ]
  %port.sroa.37.0.i.i1728.lcssa = phi i32 [ %port.sroa.37.0.i.i1728, %Flow3488 ]
  %.lcssa4024 = phi i32 [ %2243, %Flow3488 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa4025)
  %2271 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %.lcssa4024)
  %2272 = icmp ne i32 %2271, %port.sroa.37.0.i.i1728.lcssa
  %2273 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %2272)
  %2274 = extractvalue { i1, i64 } %2273, 0
  %2275 = extractvalue { i1, i64 } %2273, 1
  br i1 %2274, label %while.body.i.i.i.i1750.preheader, label %Flow3477

while.body.i.i.i.i1750.preheader:                 ; preds = %cond.false.i.i.i1749
  br label %while.body.i.i.i.i1750, !amdgpu.uniform !42

Flow3477:                                         ; preds = %Flow3475, %cond.false.i.i.i1749
  call void @llvm.amdgcn.end.cf.i64(i64 %2275)
  br label %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsINS_3cpp6stringEEERS1_T_.exit1771, !amdgpu.uniform !42

while.body.i.i.i.i1750:                           ; preds = %while.body.i.i.i.i1750.preheader, %while.body.i.i.i.i1750
  %phi.broken3970 = phi i64 [ 0, %while.body.i.i.i.i1750.preheader ], [ %2279, %while.body.i.i.i.i1750 ]
  tail call void @llvm.amdgcn.s.sleep(i32 2)
  %2276 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i8.i.i1751 = getelementptr inbounds nuw i32, ptr %2276, i64 %idxprom.i23.i.i.i.i.i1724, !amdgpu.uniform !42
  %2277 = load atomic i32, ptr %arrayidx.i.i.i8.i.i1751 monotonic, align 4
  %2278 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %2277)
  %cmp.i.not.i.i.i.i1752 = icmp eq i32 %2278, %port.sroa.37.0.i.i1728.lcssa
  %2279 = call i64 @llvm.amdgcn.if.break.i64(i1 %cmp.i.not.i.i.i.i1752, i64 %phi.broken3970)
  %2280 = call i1 @llvm.amdgcn.loop.i64(i64 %2279)
  br i1 %2280, label %Flow3475, label %while.body.i.i.i.i1750

Flow3475:                                         ; preds = %while.body.i.i.i.i1750
  %.lcssa4023 = phi i64 [ %2279, %while.body.i.i.i.i1750 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa4023)
  br label %Flow3477, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsINS_3cpp6stringEEERS1_T_.exit1771: ; preds = %Flow3477
  fence acquire
  tail call void @llvm.amdgcn.wave.barrier()
  fence syncscope("agent") release
  %2281 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %.lcssa4030, i1 true)
  %iszero.i.i.i.i.i13.i.i1754 = icmp ne i64 %.lcssa4030, 0
  %cmp2.i.i.i.i14.i.i1755 = icmp eq i64 %2281, %sh_prom.i.i.i.i
  %cmp.i.i.i.i15.i.i1756 = select i1 %iszero.i.i.i.i.i13.i.i1754, i1 %cmp2.i.i.i.i14.i.i1755, i1 false
  %2282 = and i32 %.lcssa4031, 31
  %conv4.i.i.i19.i.i1760 = zext i1 %cmp.i.i.i.i15.i.i1756 to i32
  %shl.i.i.i20.i.i1761 = shl nuw i32 %conv4.i.i.i19.i.i1760, %2282
  %xor.i.i.i21.i.i1762 = xor i32 %shl.i.i.i20.i.i1761, -1
  %2283 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2653, !amdgpu.uniform !42

2284:                                             ; preds = %ComputeEnd2654
  %div8.i.i.i16.i.i1757 = lshr i32 %.lcssa4031, 5
  %idxprom.i.i.i17.i.i1758 = zext nneg i32 %div8.i.i.i16.i.i1757 to i64
  %2285 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i17.i.i1758
  %sunkaddr3340 = getelementptr inbounds i8, ptr addrspace(1) %2285, i64 40
  %2286 = atomicrmw and ptr addrspace(1) %sunkaddr3340, i32 %.lcssa4022 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %2287, !amdgpu.uniform !42

2287:                                             ; preds = %2284, %ComputeEnd2654
  call void @llvm.amdgcn.end.cf.i64(i64 %3939)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %for.cond.i.i1.i1780, !amdgpu.uniform !42

for.cond.i.i1.i1780:                              ; preds = %2287, %Flow3474
  %2288 = phi i32 [ %2326, %Flow3474 ], [ poison, %2287 ]
  %2289 = phi i64 [ %2327, %Flow3474 ], [ poison, %2287 ]
  %2290 = phi i32 [ %2328, %Flow3474 ], [ poison, %2287 ]
  %retval.sroa.6.0.i.i.i1781 = phi i32 [ %2331, %Flow3474 ], [ undef, %2287 ]
  %retval.sroa.2.0.i.i.i1783 = phi i64 [ %2330, %Flow3474 ], [ undef, %2287 ]
  %retval.sroa.8.0.i.i.i1784 = phi i32 [ %2329, %Flow3474 ], [ undef, %2287 ]
  %index.0.i.i.i1785 = phi i32 [ %2333, %Flow3474 ], [ 0, %2287 ]
  %2291 = load i32, ptr addrspace(1) @__llvm_rpc_client, align 8, !tbaa !68
  %cmp.not.i.i.i1786 = icmp ult i32 %index.0.i.i.i1785, %2291
  %spec.store.select.i.i.i1787 = select i1 %cmp.not.i.i.i1786, i32 %index.0.i.i.i1785, i32 0
  %2292 = tail call noundef i64 @llvm.amdgcn.ballot.i64(i1 true)
  %2293 = and i32 %spec.store.select.i.i.i1787, 31
  %2294 = lshr i64 %2292, %sh_prom.i.i.i.i
  %2295 = trunc i64 %2294 to i32
  %conv4.i.i.i.i.i1791 = and i32 %2295, 1
  %shl.i.i.i.i.i1792 = shl nuw i32 %conv4.i.i.i.i.i1791, %2293
  %2296 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2657, !amdgpu.uniform !42

2297:                                             ; preds = %ComputeEnd2658
  %sunkaddr3341 = getelementptr inbounds i8, ptr addrspace(1) %3949, i64 40
  %2298 = atomicrmw or ptr addrspace(1) %sunkaddr3341, i32 %.lcssa4016 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %2299, !amdgpu.uniform !42

2299:                                             ; preds = %2297, %ComputeEnd2658
  %2300 = phi i32 [ %2298, %2297 ], [ poison, %ComputeEnd2658 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %3958)
  %2301 = call i32 @llvm.amdgcn.readfirstlane.i32(i32 %2300)
  %2302 = or i32 %2301, %.lcssa4017
  %shl5.i.i.i.i.i1793 = shl nuw i32 1, %2293
  %and.i.i.i.i.i1794 = and i32 %shl5.i.i.i.i.i1793, %2302
  %tobool3.i.i.i.i1795 = icmp ne i32 %and.i.i.i.i.i1794, 0
  %2303 = tail call i64 @llvm.amdgcn.ballot.i64(i1 %tobool3.i.i.i.i1795)
  %and.i.i.i.i.i.i1796 = and i64 %2303, %2292
  %cmp.i.not.i.i.i1797 = icmp ne i64 %2292, %and.i.i.i.i.i.i1796
  br i1 %cmp.i.not.i.i.i1797, label %if.then.i.i.i.i1798, label %Flow3473, !amdgpu.uniform !42

if.then.i.i.i.i1798:                              ; preds = %2299
  fence syncscope("agent") acquire
  %2304 = load <2 x i64>, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !75
  %2305 = extractelement <2 x i64> %2304, i32 0
  %2306 = inttoptr i64 %2305 to ptr
  %2307 = extractelement <2 x i64> %2304, i32 1
  %2308 = inttoptr i64 %2307 to ptr
  %idxprom.i.i.i.i1799 = zext i32 %spec.store.select.i.i.i1787 to i64
  %arrayidx.i.i.i.i1800 = getelementptr inbounds nuw i32, ptr %2306, i64 %idxprom.i.i.i.i1799, !amdgpu.uniform !42
  %2309 = load atomic i32, ptr %arrayidx.i.i.i.i1800 monotonic, align 4
  %2310 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %2309)
  %arrayidx.i45.i.i.i1801 = getelementptr inbounds nuw i32, ptr %2308, i64 %idxprom.i.i.i.i1799, !amdgpu.uniform !42
  %2311 = load atomic i32, ptr %arrayidx.i45.i.i.i1801 monotonic, align 4
  %2312 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %2311)
  %cmp.i46.not.i.i.i1802 = icmp ne i32 %2310, %2312
  br i1 %cmp.i46.not.i.i.i1802, label %if.then12.i.i.i1803, label %Flow3471, !amdgpu.uniform !42

if.then12.i.i.i1803:                              ; preds = %if.then.i.i.i.i1798
  fence syncscope("agent") release
  %2313 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %2292, i1 true)
  %iszero.i.i.i.i.i.i.i1804 = icmp ne i64 %2292, 0
  %cmp2.i.i.i.i.i.i1805 = icmp eq i64 %2313, %sh_prom.i.i.i.i
  %cmp.i.i.i.i.i.i1806 = select i1 %iszero.i.i.i.i.i.i.i1804, i1 %cmp2.i.i.i.i.i.i1805, i1 false
  %conv4.i.i51.i.i.i1807 = zext i1 %cmp.i.i.i.i.i.i1806 to i32
  %shl.i.i52.i.i.i1808 = shl nuw i32 %conv4.i.i51.i.i.i1807, %2293
  %xor.i.i.i.i.i1809 = xor i32 %shl.i.i52.i.i.i1808, -1
  %2314 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2662, !amdgpu.uniform !42

2315:                                             ; preds = %ComputeEnd2663
  %sunkaddr3342 = getelementptr inbounds i8, ptr addrspace(1) %3949, i64 40
  %2316 = atomicrmw and ptr addrspace(1) %sunkaddr3342, i32 %.lcssa4018 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %2317, !amdgpu.uniform !42

2317:                                             ; preds = %2315, %ComputeEnd2663
  call void @llvm.amdgcn.end.cf.i64(i64 %3976)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %Flow3471, !amdgpu.uniform !42

if.end14.i.i.i1869:                               ; preds = %Flow3471
  %2318 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %2292, i1 true)
  %iszero.i.i.i.i.i.i1870 = icmp ne i64 %2292, 0
  %cmp2.i.i.i.i.i1871 = icmp eq i64 %2318, %sh_prom.i.i.i.i
  %cmp.i.i.i.i.i1872 = select i1 %iszero.i.i.i.i.i.i1870, i1 %cmp2.i.i.i.i.i1871, i1 false
  %2319 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.i.i.i.i.i1872)
  %2320 = extractvalue { i1, i64 } %2319, 0
  %2321 = extractvalue { i1, i64 } %2319, 1
  br i1 %2320, label %if.then16.i.i.i1874, label %if.end22.i.i.i1873

if.then16.i.i.i1874:                              ; preds = %if.end14.i.i.i1869
  %2322 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !76
  %opcode.i.i.i1875 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %2322, i64 %idxprom.i.i.i.i1799, i32 1
  store i32 1660944387, ptr %opcode.i.i.i1875, align 8, !tbaa !77
  %arrayidx21.i.i.i1876 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %2322, i64 %idxprom.i.i.i.i1799
  store i64 %2292, ptr %arrayidx21.i.i.i1876, align 8, !tbaa !80
  br label %if.end22.i.i.i1873, !amdgpu.uniform !42

Flow3472:                                         ; preds = %if.end22.i.i.i1873, %Flow3471
  %2323 = phi i32 [ %2310, %if.end22.i.i.i1873 ], [ %retval.sroa.8.0.i.i.i1784, %Flow3471 ]
  %2324 = phi i64 [ %2292, %if.end22.i.i.i1873 ], [ %retval.sroa.2.0.i.i.i1783, %Flow3471 ]
  %2325 = phi i32 [ %spec.store.select.i.i.i1787, %if.end22.i.i.i1873 ], [ %retval.sroa.6.0.i.i.i1781, %Flow3471 ]
  br label %cleanup26.i.i.i1810, !amdgpu.uniform !42

if.end22.i.i.i1873:                               ; preds = %if.then16.i.i.i1874, %if.end14.i.i.i1869
  call void @llvm.amdgcn.end.cf.i64(i64 %2321)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %Flow3472, !amdgpu.uniform !42

Flow3473:                                         ; preds = %cleanup26.i.i.i1810, %2299
  %2326 = phi i32 [ %2325, %cleanup26.i.i.i1810 ], [ %2288, %2299 ]
  %2327 = phi i64 [ %2324, %cleanup26.i.i.i1810 ], [ %2289, %2299 ]
  %2328 = phi i32 [ %2323, %cleanup26.i.i.i1810 ], [ %2290, %2299 ]
  %2329 = phi i32 [ %2323, %cleanup26.i.i.i1810 ], [ %retval.sroa.8.0.i.i.i1784, %2299 ]
  %2330 = phi i64 [ %2324, %cleanup26.i.i.i1810 ], [ %retval.sroa.2.0.i.i.i1783, %2299 ]
  %2331 = phi i32 [ %2325, %cleanup26.i.i.i1810 ], [ %retval.sroa.6.0.i.i.i1781, %2299 ]
  %2332 = phi i1 [ %cmp.i46.not.i.i.i1802, %cleanup26.i.i.i1810 ], [ true, %2299 ]
  br i1 %2332, label %for.inc.i.i.i1815, label %Flow3474, !amdgpu.uniform !42

cleanup26.i.i.i1810:                              ; preds = %Flow3472
  br label %Flow3473, !amdgpu.uniform !42

for.inc.i.i.i1815:                                ; preds = %Flow3473
  %inc.i.i.i1820 = add i32 %spec.store.select.i.i.i1787, 1
  br label %Flow3474, !amdgpu.uniform !42

Flow3474:                                         ; preds = %for.inc.i.i.i1815, %Flow3473
  %2333 = phi i32 [ %inc.i.i.i1820, %for.inc.i.i.i1815 ], [ poison, %Flow3473 ]
  %2334 = phi i1 [ false, %for.inc.i.i.i1815 ], [ true, %Flow3473 ]
  br i1 %2334, label %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1821, label %for.cond.i.i1.i1780, !amdgpu.uniform !42

_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1821: ; preds = %Flow3474
  %.lcssa4021 = phi i32 [ %2326, %Flow3474 ]
  %.lcssa4020 = phi i64 [ %2327, %Flow3474 ]
  %.lcssa4019 = phi i32 [ %2328, %Flow3474 ]
  fence acquire
  %2335 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %mul.i.i.i.i.i.i1822 = shl i32 %.lcssa4021, 6
  %idxprom.i19.i.i.i.i.i1823 = zext i32 %mul.i.i.i.i.i.i1822 to i64
  %arrayidx.i20.i.i.i.i.i1824 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %2335, i64 %idxprom.i19.i.i.i.i.i1823
  %arrayidx.i22.i.i.i.i.i1825 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %arrayidx.i20.i.i.i.i.i1824, i64 %sh_prom.i.i.i.i
  store i64 sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.35 to ptr), i64 10) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.35 to ptr) to i64)), ptr %arrayidx.i22.i.i.i.i.i1825, align 8, !tbaa !82
  br i1 %1215, label %loop-memcpy-expansion2375.preheader, label %Flow3470, !amdgpu.uniform !42

loop-memcpy-expansion2375.preheader:              ; preds = %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1821
  %2336 = add nuw nsw i64 %75, 8
  %scevgep2897 = getelementptr i8, ptr %2335, i64 %2336
  %2337 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i1823, 6
  %scevgep2898 = getelementptr i8, ptr %scevgep2897, i64 %2337
  br label %loop-memcpy-expansion2375, !amdgpu.uniform !42

Flow3470:                                         ; preds = %Flow3469, %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1821
  br label %loop-memcpy-residual-header2378, !amdgpu.uniform !42

loop-memcpy-expansion2375:                        ; preds = %loop-memcpy-expansion2375.preheader, %loop-memcpy-expansion2375
  %loop-index2376 = phi i64 [ %2340, %loop-memcpy-expansion2375 ], [ 0, %loop-memcpy-expansion2375.preheader ]
  %2338 = getelementptr i8, ptr addrspace(4) @.str.35, i64 %loop-index2376, !amdgpu.uniform !42
  %2339 = load <4 x i32>, ptr addrspace(4) %2338, align 1
  %scevgep2899 = getelementptr i8, ptr %scevgep2898, i64 %loop-index2376
  store <4 x i32> %2339, ptr %scevgep2899, align 1
  %2340 = add i64 %loop-index2376, 16
  %2341 = icmp uge i64 %2340, %1214
  br i1 %2341, label %Flow3469, label %loop-memcpy-expansion2375, !amdgpu.uniform !42

Flow3467:                                         ; preds = %Flow3465, %Flow3466
  %2342 = phi i64 [ %2934, %Flow3465 ], [ %2931, %Flow3466 ]
  br label %post-loop-memcpy-expansion2374, !amdgpu.uniform !42

loop-memcpy-residual2377:                         ; preds = %loop-memcpy-residual2377.preheader, %loop-memcpy-residual2377
  %residual-loop-index2379 = phi i64 [ %2344, %loop-memcpy-residual2377 ], [ 0, %loop-memcpy-residual2377.preheader ]
  %scevgep2896 = getelementptr i8, ptr addrspace(4) @.str.35, i64 %residual-loop-index2379, !amdgpu.uniform !42
  %2343 = load i8, ptr addrspace(4) %scevgep2896, align 1
  %scevgep2895 = getelementptr i8, ptr %scevgep2894, i64 %residual-loop-index2379
  store i8 %2343, ptr %scevgep2895, align 1
  %2344 = add i64 %residual-loop-index2379, 1
  %2345 = icmp uge i64 %2344, %1213
  br i1 %2345, label %Flow3465, label %loop-memcpy-residual2377, !amdgpu.uniform !42

Flow3465:                                         ; preds = %loop-memcpy-residual2377
  br label %Flow3467, !amdgpu.uniform !42

post-loop-memcpy-expansion2374:                   ; preds = %Flow3467
  %tobool.not.i.i.i.i.i.i1828 = icmp eq i32 %.lcssa4019, 0
  %conv.i.i.i.i.i.i1829 = zext i1 %tobool.not.i.i.i.i.i.i1828 to i32
  fence release
  %2346 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %idxprom.i23.i.i.i.i.i1830 = zext i32 %.lcssa4021 to i64
  %arrayidx.i24.i.i.i.i.i1831 = getelementptr inbounds nuw i32, ptr %2346, i64 %idxprom.i23.i.i.i.i.i1830
  store atomic i32 %conv.i.i.i.i.i.i1829, ptr %arrayidx.i24.i.i.i.i.i1831 monotonic, align 4
  %2347 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !84
  %arrayidx.i.i5.i.i1832 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %2347, i64 %idxprom.i23.i.i.i.i.i1830, !amdgpu.uniform !42
  %2348 = load i64, ptr %arrayidx.i.i5.i.i1832, align 8, !tbaa !80
  br label %while.cond.i.i.i.i1833, !amdgpu.uniform !42

while.cond.i.i.i.i1833:                           ; preds = %post-loop-memcpy-expansion2374, %Flow3464
  %phi.broken3971 = phi i64 [ 0, %post-loop-memcpy-expansion2374 ], [ %2365, %Flow3464 ]
  %lsr.iv2886 = phi i64 [ %2363, %Flow3464 ], [ -46, %post-loop-memcpy-expansion2374 ]
  %lsr.iv2881 = phi ptr addrspace(4) [ %2362, %Flow3464 ], [ getelementptr (i8, ptr addrspace(4) @.str.35, i64 56), %post-loop-memcpy-expansion2374 ]
  %port.sroa.37.0.i.i1834 = phi i32 [ %2361, %Flow3464 ], [ %conv.i.i.i.i.i.i1829, %post-loop-memcpy-expansion2374 ]
  %idx.0.i.i.i.i1835 = phi i64 [ %2360, %Flow3464 ], [ 56, %post-loop-memcpy-expansion2374 ]
  %cmp.i.i.i.i1836 = icmp ult i64 %idx.0.i.i.i.i1835, sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.35 to ptr), i64 10) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.35 to ptr) to i64))
  %2349 = tail call i64 @llvm.amdgcn.ballot.i64(i1 %cmp.i.i.i.i1836)
  %and.i.i.i.i6.i.i1837 = and i64 %2349, %2348
  %tobool.not.i.i.i.i1838 = icmp ne i64 %and.i.i.i.i6.i.i1837, 0
  %2350 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i22.i.i.i1839 = getelementptr inbounds nuw i32, ptr %2350, i64 %idxprom.i23.i.i.i.i.i1830, !amdgpu.uniform !42
  %2351 = load atomic i32, ptr %arrayidx.i22.i.i.i1839 monotonic, align 4
  %2352 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %tobool.not.i.i.i.i1838)
  %2353 = extractvalue { i1, i64 } %2352, 0
  %2354 = extractvalue { i1, i64 } %2352, 1
  br i1 %2353, label %cond.false.i8.i.i.i.i1840, label %Flow3464

cond.false.i8.i.i.i.i1840:                        ; preds = %while.cond.i.i.i.i1833
  %2355 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %2351)
  %2356 = icmp ne i32 %2355, %port.sroa.37.0.i.i1834
  br i1 %2356, label %while.body.i.i21.i.i.i.i1841.preheader, label %Flow3463, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i1841.preheader:           ; preds = %cond.false.i8.i.i.i.i1840
  br label %while.body.i.i21.i.i.i.i1841, !amdgpu.uniform !42

Flow3463:                                         ; preds = %Flow3462, %cond.false.i8.i.i.i.i1840
  br label %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i1844, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i1841:                     ; preds = %while.body.i.i21.i.i.i.i1841.preheader, %while.body.i.i21.i.i.i.i1841
  tail call void @llvm.amdgcn.s.sleep(i32 2)
  %2357 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i22.i.i.i.i1842 = getelementptr inbounds nuw i32, ptr %2357, i64 %idxprom.i23.i.i.i.i.i1830, !amdgpu.uniform !42
  %2358 = load atomic i32, ptr %arrayidx.i.i.i22.i.i.i.i1842 monotonic, align 4
  %2359 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %2358)
  %cmp.i.not.i.i23.i.i.i.i1843 = icmp eq i32 %2359, %port.sroa.37.0.i.i1834
  br i1 %cmp.i.not.i.i23.i.i.i.i1843, label %Flow3462, label %while.body.i.i21.i.i.i.i1841, !amdgpu.uniform !42

Flow3462:                                         ; preds = %while.body.i.i21.i.i.i.i1841
  br label %Flow3463, !amdgpu.uniform !42

Flow3464:                                         ; preds = %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1845, %while.cond.i.i.i.i1833
  %2360 = phi i64 [ %add.i.i.i.i1848, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1845 ], [ poison, %while.cond.i.i.i.i1833 ]
  %2361 = phi i32 [ %conv.i.i26.i.i.i.i1846, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1845 ], [ poison, %while.cond.i.i.i.i1833 ]
  %2362 = phi ptr addrspace(4) [ %scevgep2882, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1845 ], [ poison, %while.cond.i.i.i.i1833 ]
  %2363 = phi i64 [ %lsr.iv.next2887, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1845 ], [ poison, %while.cond.i.i.i.i1833 ]
  %2364 = phi i1 [ false, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1845 ], [ true, %while.cond.i.i.i.i1833 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %2354)
  %2365 = call i64 @llvm.amdgcn.if.break.i64(i1 %2364, i64 %phi.broken3971)
  %2366 = call i1 @llvm.amdgcn.loop.i64(i64 %2365)
  br i1 %2366, label %cond.false.i.i.i1855, label %while.cond.i.i.i.i1833

_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i1844: ; preds = %Flow3463
  fence acquire
  br i1 %cmp.i.i.i.i1836, label %if.then.i.i.i.i.i.i.i1849, label %Flow3461, !amdgpu.uniform !42

if.then.i.i.i.i.i.i.i1849:                        ; preds = %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i1844
  %sub.i.i.i.i.i.i.i1851 = sub nuw i64 sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.35 to ptr), i64 10) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.35 to ptr) to i64)), %idx.0.i.i.i.i1835
  %spec.select.i.i.i36.i.i.i.i1852 = tail call i64 @llvm.umin.i64(i64 %sub.i.i.i.i.i.i.i1851, i64 64)
  %2367 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %2368 = and i64 %spec.select.i.i.i36.i.i.i.i1852, 15
  %2369 = sub i64 %spec.select.i.i.i36.i.i.i.i1852, %2368
  %2370 = icmp ne i64 %2369, 0
  br i1 %2370, label %loop-memcpy-expansion2423.preheader, label %Flow3460, !amdgpu.uniform !42

loop-memcpy-expansion2423.preheader:              ; preds = %if.then.i.i.i.i.i.i.i1849
  %arrayidx.i20.i35.i.i.i.i1853 = getelementptr %"struct.rpc::Buffer", ptr %2367, i64 %idxprom.i19.i.i.i.i.i1823
  %arrayidx.i22.i38.i.i.i.i1854 = getelementptr %"struct.rpc::Buffer", ptr %arrayidx.i20.i35.i.i.i.i1853, i64 %sh_prom.i.i.i.i
  br label %loop-memcpy-expansion2423, !amdgpu.uniform !42

Flow3460:                                         ; preds = %Flow3459, %if.then.i.i.i.i.i.i.i1849
  br label %loop-memcpy-residual-header2426, !amdgpu.uniform !42

loop-memcpy-expansion2423:                        ; preds = %loop-memcpy-expansion2423.preheader, %loop-memcpy-expansion2423
  %loop-index2424 = phi i64 [ %2373, %loop-memcpy-expansion2423 ], [ 0, %loop-memcpy-expansion2423.preheader ]
  %scevgep2883 = getelementptr i8, ptr addrspace(4) %lsr.iv2881, i64 %loop-index2424, !amdgpu.uniform !42
  %2371 = load <4 x i32>, ptr addrspace(4) %scevgep2883, align 1
  %2372 = getelementptr i8, ptr %arrayidx.i22.i38.i.i.i.i1854, i64 %loop-index2424
  store <4 x i32> %2371, ptr %2372, align 1
  %2373 = add i64 %loop-index2424, 16
  %2374 = icmp uge i64 %2373, %2369
  br i1 %2374, label %Flow3459, label %loop-memcpy-expansion2423, !amdgpu.uniform !42

Flow3458:                                         ; preds = %Flow3457, %loop-memcpy-residual-header2426
  br label %Flow3461, !amdgpu.uniform !42

loop-memcpy-residual2425:                         ; preds = %loop-memcpy-residual2425.preheader, %loop-memcpy-residual2425
  %residual-loop-index2427 = phi i64 [ %2376, %loop-memcpy-residual2425 ], [ 0, %loop-memcpy-residual2425.preheader ]
  %scevgep2892 = getelementptr i8, ptr addrspace(4) %scevgep2891, i64 %residual-loop-index2427, !amdgpu.uniform !42
  %2375 = load i8, ptr addrspace(4) %scevgep2892, align 1
  %scevgep2890 = getelementptr i8, ptr %scevgep2889, i64 %residual-loop-index2427
  store i8 %2375, ptr %scevgep2890, align 1
  %2376 = add i64 %residual-loop-index2427, 1
  %2377 = icmp uge i64 %2376, %2368
  br i1 %2377, label %Flow3457, label %loop-memcpy-residual2425, !amdgpu.uniform !42

Flow3457:                                         ; preds = %loop-memcpy-residual2425
  br label %Flow3458, !amdgpu.uniform !42

_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1845: ; preds = %Flow3461
  %conv.i.i26.i.i.i.i1846 = xor i32 %port.sroa.37.0.i.i1834, 1
  fence release
  %2378 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %arrayidx.i24.i29.i.i.i.i1847 = getelementptr inbounds nuw i32, ptr %2378, i64 %idxprom.i23.i.i.i.i.i1830
  store atomic i32 %conv.i.i26.i.i.i.i1846, ptr %arrayidx.i24.i29.i.i.i.i1847 monotonic, align 4
  %add.i.i.i.i1848 = add i64 %idx.0.i.i.i.i1835, 64
  %scevgep2882 = getelementptr i8, ptr addrspace(4) %lsr.iv2881, i64 64
  %lsr.iv.next2887 = add i64 %lsr.iv2886, -64
  br label %Flow3464, !amdgpu.uniform !42

cond.false.i.i.i1855:                             ; preds = %Flow3464
  %.lcssa4015 = phi i64 [ %2365, %Flow3464 ]
  %port.sroa.37.0.i.i1834.lcssa = phi i32 [ %port.sroa.37.0.i.i1834, %Flow3464 ]
  %.lcssa4014 = phi i32 [ %2351, %Flow3464 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa4015)
  %2379 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %.lcssa4014)
  %2380 = icmp ne i32 %2379, %port.sroa.37.0.i.i1834.lcssa
  %2381 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %2380)
  %2382 = extractvalue { i1, i64 } %2381, 0
  %2383 = extractvalue { i1, i64 } %2381, 1
  br i1 %2382, label %while.body.i.i.i.i1856.preheader, label %Flow3456

while.body.i.i.i.i1856.preheader:                 ; preds = %cond.false.i.i.i1855
  br label %while.body.i.i.i.i1856, !amdgpu.uniform !42

Flow3456:                                         ; preds = %Flow3455, %cond.false.i.i.i1855
  call void @llvm.amdgcn.end.cf.i64(i64 %2383)
  br label %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit1877, !amdgpu.uniform !42

while.body.i.i.i.i1856:                           ; preds = %while.body.i.i.i.i1856.preheader, %while.body.i.i.i.i1856
  %phi.broken3972 = phi i64 [ 0, %while.body.i.i.i.i1856.preheader ], [ %2387, %while.body.i.i.i.i1856 ]
  tail call void @llvm.amdgcn.s.sleep(i32 2)
  %2384 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i8.i.i1857 = getelementptr inbounds nuw i32, ptr %2384, i64 %idxprom.i23.i.i.i.i.i1830, !amdgpu.uniform !42
  %2385 = load atomic i32, ptr %arrayidx.i.i.i8.i.i1857 monotonic, align 4
  %2386 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %2385)
  %cmp.i.not.i.i.i.i1858 = icmp eq i32 %2386, %port.sroa.37.0.i.i1834.lcssa
  %2387 = call i64 @llvm.amdgcn.if.break.i64(i1 %cmp.i.not.i.i.i.i1858, i64 %phi.broken3972)
  %2388 = call i1 @llvm.amdgcn.loop.i64(i64 %2387)
  br i1 %2388, label %Flow3455, label %while.body.i.i.i.i1856

Flow3455:                                         ; preds = %while.body.i.i.i.i1856
  %.lcssa4013 = phi i64 [ %2387, %while.body.i.i.i.i1856 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa4013)
  br label %Flow3456, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit1877: ; preds = %Flow3456
  fence acquire
  tail call void @llvm.amdgcn.wave.barrier()
  fence syncscope("agent") release
  %2389 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %.lcssa4020, i1 true)
  %iszero.i.i.i.i.i13.i.i1860 = icmp ne i64 %.lcssa4020, 0
  %cmp2.i.i.i.i14.i.i1861 = icmp eq i64 %2389, %sh_prom.i.i.i.i
  %cmp.i.i.i.i15.i.i1862 = select i1 %iszero.i.i.i.i.i13.i.i1860, i1 %cmp2.i.i.i.i14.i.i1861, i1 false
  %2390 = and i32 %.lcssa4021, 31
  %conv4.i.i.i19.i.i1866 = zext i1 %cmp.i.i.i.i15.i.i1862 to i32
  %shl.i.i.i20.i.i1867 = shl nuw i32 %conv4.i.i.i19.i.i1866, %2390
  %xor.i.i.i21.i.i1868 = xor i32 %shl.i.i.i20.i.i1867, -1
  %2391 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2666, !amdgpu.uniform !42

2392:                                             ; preds = %ComputeEnd2667
  %div8.i.i.i16.i.i1863 = lshr i32 %.lcssa4021, 5
  %idxprom.i.i.i17.i.i1864 = zext nneg i32 %div8.i.i.i16.i.i1863 to i64
  %2393 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i17.i.i1864
  %sunkaddr3343 = getelementptr inbounds i8, ptr addrspace(1) %2393, i64 40
  %2394 = atomicrmw and ptr addrspace(1) %sunkaddr3343, i32 %.lcssa4012 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %2395, !amdgpu.uniform !42

2395:                                             ; preds = %2392, %ComputeEnd2667
  call void @llvm.amdgcn.end.cf.i64(i64 %3993)
  tail call void @llvm.amdgcn.wave.barrier()
  call void @llvm.lifetime.start.p5(ptr addrspace(5) %buffer.i.i.i) #34, !noalias !121
  %cmp7.i.i.i.i.i.i92.i = icmp ne i32 %RHS, 0
  %2396 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp7.i.i.i.i.i.i92.i)
  %2397 = extractvalue { i1, i64 } %2396, 0
  %2398 = extractvalue { i1, i64 } %2396, 1
  br i1 %2397, label %while.body.lr.ph.i.i.i.i.i.i93.i, label %Flow3454

while.body.lr.ph.i.i.i.i.i.i93.i:                 ; preds = %2395
  %retval.0.i.i.i.i.i.i94.i = tail call noundef i32 @llvm.abs.i32(i32 %RHS, i1 false)
  br label %while.body.i.i.i.i.i.i95.i, !amdgpu.uniform !42

Flow3454:                                         ; preds = %Flow3453, %2395
  %2399 = phi i1 [ %cmp.i.i.i.i.i.i.i.i104.i.lcssa, %Flow3453 ], [ false, %2395 ]
  %2400 = phi i64 [ %writer.sroa.10.1.i.i.i.i109.i.lcssa, %Flow3453 ], [ 0, %2395 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %2398)
  br label %_ZN22__llvm_libc_22_0_0_git15IntegerToStringIiNS_7details3FmtILh10ELb0ELb0ELb0ELm1EEEE13IntegerWriter25write_unsigned_number_decEjRNS1_22StringBufferWriterImplILb0EEE.exit.i.i.i.i.i112.i, !amdgpu.uniform !42

while.body.i.i.i.i.i.i95.i:                       ; preds = %while.body.lr.ph.i.i.i.i.i.i93.i, %_ZN22__llvm_libc_22_0_0_git7details22StringBufferWriterImplILb0EE4pushEc.exit.i.i.i.i.i.i108.i
  %phi.broken3973 = phi i64 [ 0, %while.body.lr.ph.i.i.i.i.i.i93.i ], [ %2468, %_ZN22__llvm_libc_22_0_0_git7details22StringBufferWriterImplILb0EE4pushEc.exit.i.i.i.i.i.i108.i ]
  %writer.sroa.10.0.i.i.i.i96.i = phi i64 [ %writer.sroa.10.1.i.i.i.i109.i, %_ZN22__llvm_libc_22_0_0_git7details22StringBufferWriterImplILb0EE4pushEc.exit.i.i.i.i.i.i108.i ], [ 0, %while.body.lr.ph.i.i.i.i.i.i93.i ]
  %value.addr.09.i.i.i.i.i.i97.i = phi i32 [ %div.i.i.i.i.i.i.i100.i, %_ZN22__llvm_libc_22_0_0_git7details22StringBufferWriterImplILb0EE4pushEc.exit.i.i.i.i.i.i108.i ], [ %retval.0.i.i.i.i.i.i94.i, %while.body.lr.ph.i.i.i.i.i.i93.i ]
  br label %NodeBlock3400, !amdgpu.uniform !42

NodeBlock3400:                                    ; preds = %while.body.i.i.i.i.i.i95.i
  %div.i.i.i.i.i.i.i100.i = udiv i32 %value.addr.09.i.i.i.i.i.i97.i, 10
  %.neg1 = mul nuw nsw i32 %div.i.i.i.i.i.i.i100.i, 246
  %rem.i.i.i.i.i.i.i98.i.decomposed = add nuw nsw i32 %.neg1, %value.addr.09.i.i.i.i.i.i97.i
  %conv.i.i.i.i.i.i.i99.i = trunc i32 %rem.i.i.i.i.i.i.i98.i.decomposed to i8
  %2401 = zext i8 %conv.i.i.i.i.i.i.i99.i to i16
  %Pivot3401 = icmp sge i16 %2401, 5
  %2402 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %Pivot3401)
  %2403 = extractvalue { i1, i64 } %2402, 0
  %2404 = extractvalue { i1, i64 } %2402, 1
  br i1 %2403, label %NodeBlock3398, label %Flow3452

NodeBlock3398:                                    ; preds = %NodeBlock3400
  %Pivot3399 = icmp sge i16 %2401, 7
  %2405 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %Pivot3399)
  %2406 = extractvalue { i1, i64 } %2405, 0
  %2407 = extractvalue { i1, i64 } %2405, 1
  br i1 %2406, label %NodeBlock3396, label %Flow3443

NodeBlock3396:                                    ; preds = %NodeBlock3398
  %Pivot3397 = icmp sge i16 %2401, 8
  %2408 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %Pivot3397)
  %2409 = extractvalue { i1, i64 } %2408, 0
  %2410 = extractvalue { i1, i64 } %2408, 1
  br i1 %2409, label %NodeBlock3394, label %Flow3439

NodeBlock3394:                                    ; preds = %NodeBlock3396
  %Pivot3395 = icmp sge i16 %2401, 9
  %2411 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %Pivot3395)
  %2412 = extractvalue { i1, i64 } %2411, 0
  %2413 = extractvalue { i1, i64 } %2411, 1
  br i1 %2412, label %sw.bb9.i.i.i.i.i.i.i.i101.i, label %Flow3437

Flow3443:                                         ; preds = %Flow3440, %NodeBlock3398
  %2414 = phi i8 [ %2458, %Flow3440 ], [ poison, %NodeBlock3398 ]
  %2415 = call { i1, i64 } @llvm.amdgcn.else.i64.i64(i64 %2407)
  %2416 = extractvalue { i1, i64 } %2415, 0
  %2417 = extractvalue { i1, i64 } %2415, 1
  br i1 %2416, label %NodeBlock3392, label %Flow3444

NodeBlock3392:                                    ; preds = %Flow3443
  %Pivot3393 = icmp sge i16 %2401, 6
  %2418 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %Pivot3393)
  %2419 = extractvalue { i1, i64 } %2418, 0
  %2420 = extractvalue { i1, i64 } %2418, 1
  br i1 %2419, label %sw.bb6.i.i.i.i.i.i.i.i157.i, label %Flow3441

Flow3452:                                         ; preds = %Flow3444, %NodeBlock3400
  %2421 = phi i8 [ %2460, %Flow3444 ], [ poison, %NodeBlock3400 ]
  %2422 = call { i1, i64 } @llvm.amdgcn.else.i64.i64(i64 %2404)
  %2423 = extractvalue { i1, i64 } %2422, 0
  %2424 = extractvalue { i1, i64 } %2422, 1
  br i1 %2423, label %NodeBlock3390, label %if.then.i.i.i.i.i.i.i102.i

NodeBlock3390:                                    ; preds = %Flow3452
  %Pivot3391 = icmp sge i16 %2401, 3
  %2425 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %Pivot3391)
  %2426 = extractvalue { i1, i64 } %2425, 0
  %2427 = extractvalue { i1, i64 } %2425, 1
  br i1 %2426, label %NodeBlock3388, label %Flow3450

NodeBlock3388:                                    ; preds = %NodeBlock3390
  %Pivot3389 = icmp sge i16 %2401, 4
  %2428 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %Pivot3389)
  %2429 = extractvalue { i1, i64 } %2428, 0
  %2430 = extractvalue { i1, i64 } %2428, 1
  br i1 %2429, label %sw.bb4.i.i.i.i.i.i.i.i159.i, label %Flow3445

Flow3450:                                         ; preds = %Flow3446, %NodeBlock3390
  %2431 = phi i8 [ %2461, %Flow3446 ], [ poison, %NodeBlock3390 ]
  %2432 = call { i1, i64 } @llvm.amdgcn.else.i64.i64(i64 %2427)
  %2433 = extractvalue { i1, i64 } %2432, 0
  %2434 = extractvalue { i1, i64 } %2432, 1
  br i1 %2433, label %NodeBlock3386, label %Flow3451

NodeBlock3386:                                    ; preds = %Flow3450
  %Pivot3387 = icmp sge i16 %2401, 2
  %2435 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %Pivot3387)
  %2436 = extractvalue { i1, i64 } %2435, 0
  %2437 = extractvalue { i1, i64 } %2435, 1
  br i1 %2436, label %sw.bb2.i.i.i.i.i.i.i.i161.i, label %Flow3448

Flow3448:                                         ; preds = %sw.bb2.i.i.i.i.i.i.i.i161.i, %NodeBlock3386
  %2438 = call { i1, i64 } @llvm.amdgcn.else.i64.i64(i64 %2437)
  %2439 = extractvalue { i1, i64 } %2438, 0
  %2440 = extractvalue { i1, i64 } %2438, 1
  br i1 %2439, label %LeafBlock3384, label %Flow3449

LeafBlock3384:                                    ; preds = %Flow3448
  %SwitchLeaf3385 = icmp eq i16 %2401, 1
  %2441 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %SwitchLeaf3385)
  %2442 = extractvalue { i1, i64 } %2441, 0
  %2443 = extractvalue { i1, i64 } %2441, 1
  br i1 %2442, label %sw.bb1.i.i.i.i.i.i.i.i162.i, label %Flow3447

sw.bb1.i.i.i.i.i.i.i.i162.i:                      ; preds = %LeafBlock3384
  br label %Flow3447, !amdgpu.uniform !42

sw.bb2.i.i.i.i.i.i.i.i161.i:                      ; preds = %NodeBlock3386
  br label %Flow3448, !amdgpu.uniform !42

Flow3445:                                         ; preds = %sw.bb4.i.i.i.i.i.i.i.i159.i, %NodeBlock3388
  %2444 = call { i1, i64 } @llvm.amdgcn.else.i64.i64(i64 %2430)
  %2445 = extractvalue { i1, i64 } %2444, 0
  %2446 = extractvalue { i1, i64 } %2444, 1
  br i1 %2445, label %sw.bb3.i.i.i.i.i.i.i.i160.i, label %Flow3446

sw.bb3.i.i.i.i.i.i.i.i160.i:                      ; preds = %Flow3445
  br label %Flow3446, !amdgpu.uniform !42

sw.bb4.i.i.i.i.i.i.i.i159.i:                      ; preds = %NodeBlock3388
  br label %Flow3445, !amdgpu.uniform !42

Flow3441:                                         ; preds = %sw.bb6.i.i.i.i.i.i.i.i157.i, %NodeBlock3392
  %2447 = call { i1, i64 } @llvm.amdgcn.else.i64.i64(i64 %2420)
  %2448 = extractvalue { i1, i64 } %2447, 0
  %2449 = extractvalue { i1, i64 } %2447, 1
  br i1 %2448, label %sw.bb5.i.i.i.i.i.i.i.i158.i, label %Flow3442

sw.bb5.i.i.i.i.i.i.i.i158.i:                      ; preds = %Flow3441
  br label %Flow3442, !amdgpu.uniform !42

sw.bb6.i.i.i.i.i.i.i.i157.i:                      ; preds = %NodeBlock3392
  br label %Flow3441, !amdgpu.uniform !42

Flow3439:                                         ; preds = %Flow3438, %NodeBlock3396
  %2450 = phi i8 [ %2457, %Flow3438 ], [ poison, %NodeBlock3396 ]
  %2451 = call { i1, i64 } @llvm.amdgcn.else.i64.i64(i64 %2410)
  %2452 = extractvalue { i1, i64 } %2451, 0
  %2453 = extractvalue { i1, i64 } %2451, 1
  br i1 %2452, label %sw.bb7.i.i.i.i.i.i.i.i156.i, label %Flow3440

sw.bb7.i.i.i.i.i.i.i.i156.i:                      ; preds = %Flow3439
  br label %Flow3440, !amdgpu.uniform !42

Flow3437:                                         ; preds = %sw.bb9.i.i.i.i.i.i.i.i101.i, %NodeBlock3394
  %2454 = call { i1, i64 } @llvm.amdgcn.else.i64.i64(i64 %2413)
  %2455 = extractvalue { i1, i64 } %2454, 0
  %2456 = extractvalue { i1, i64 } %2454, 1
  br i1 %2455, label %sw.bb8.i.i.i.i.i.i.i.i155.i, label %Flow3438

sw.bb8.i.i.i.i.i.i.i.i155.i:                      ; preds = %Flow3437
  br label %Flow3438, !amdgpu.uniform !42

sw.bb9.i.i.i.i.i.i.i.i101.i:                      ; preds = %NodeBlock3394
  br label %Flow3437, !amdgpu.uniform !42

Flow3438:                                         ; preds = %sw.bb8.i.i.i.i.i.i.i.i155.i, %Flow3437
  %2457 = phi i8 [ 56, %sw.bb8.i.i.i.i.i.i.i.i155.i ], [ 57, %Flow3437 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %2456)
  br label %Flow3439, !amdgpu.uniform !42

Flow3440:                                         ; preds = %sw.bb7.i.i.i.i.i.i.i.i156.i, %Flow3439
  %2458 = phi i8 [ 55, %sw.bb7.i.i.i.i.i.i.i.i156.i ], [ %2450, %Flow3439 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %2453)
  br label %Flow3443, !amdgpu.uniform !42

Flow3442:                                         ; preds = %sw.bb5.i.i.i.i.i.i.i.i158.i, %Flow3441
  %2459 = phi i8 [ 53, %sw.bb5.i.i.i.i.i.i.i.i158.i ], [ 54, %Flow3441 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %2449)
  br label %Flow3444, !amdgpu.uniform !42

Flow3444:                                         ; preds = %Flow3442, %Flow3443
  %2460 = phi i8 [ %2459, %Flow3442 ], [ %2414, %Flow3443 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %2417)
  br label %Flow3452, !amdgpu.uniform !42

Flow3446:                                         ; preds = %sw.bb3.i.i.i.i.i.i.i.i160.i, %Flow3445
  %2461 = phi i8 [ 51, %sw.bb3.i.i.i.i.i.i.i.i160.i ], [ 52, %Flow3445 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %2446)
  br label %Flow3450, !amdgpu.uniform !42

Flow3447:                                         ; preds = %sw.bb1.i.i.i.i.i.i.i.i162.i, %LeafBlock3384
  %2462 = phi i8 [ 49, %sw.bb1.i.i.i.i.i.i.i.i162.i ], [ 48, %LeafBlock3384 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %2443)
  br label %Flow3449, !amdgpu.uniform !42

Flow3449:                                         ; preds = %Flow3447, %Flow3448
  %2463 = phi i8 [ %2462, %Flow3447 ], [ 50, %Flow3448 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %2440)
  br label %Flow3451, !amdgpu.uniform !42

Flow3451:                                         ; preds = %Flow3449, %Flow3450
  %2464 = phi i8 [ %2463, %Flow3449 ], [ %2431, %Flow3450 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %2434)
  br label %if.then.i.i.i.i.i.i.i102.i, !amdgpu.uniform !42

if.then.i.i.i.i.i.i.i102.i:                       ; preds = %Flow3452, %Flow3451
  %retval.0.i.i.i.i.i.i.i.i103.i = phi i8 [ %2421, %Flow3452 ], [ %2464, %Flow3451 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %2424)
  %cmp.i.i.i.i.i.i.i.i104.i = icmp eq i64 %writer.sroa.10.0.i.i.i.i96.i, 11
  %cmp.i.i.i.i.i.i.i.i104.i.inv = xor i1 %cmp.i.i.i.i.i.i.i.i104.i, true
  br i1 %cmp.i.i.i.i.i.i.i.i104.i.inv, label %if.then3.i.i.i.i.i.i.i105.i, label %_ZN22__llvm_libc_22_0_0_git7details22StringBufferWriterImplILb0EE4pushEc.exit.i.i.i.i.i.i108.i, !amdgpu.uniform !42

if.then3.i.i.i.i.i.i.i105.i:                      ; preds = %if.then.i.i.i.i.i.i.i102.i
  %2465 = trunc i64 %writer.sroa.10.0.i.i.i.i96.i to i32
  %2466 = xor i32 %2465, -1
  %2467 = getelementptr i8, ptr addrspace(5) %buffer.i.i.i, i32 %2466
  %arrayidx.i.i.i.i.i.i.i.i106.i2742 = getelementptr i8, ptr addrspace(5) %2467, i32 11
  %inc.i.i.i.i.i.i.i107.i = add i64 %writer.sroa.10.0.i.i.i.i96.i, 1
  store i8 %retval.0.i.i.i.i.i.i.i.i103.i, ptr addrspace(5) %arrayidx.i.i.i.i.i.i.i.i106.i2742, align 1, !tbaa !57, !noalias !121
  br label %_ZN22__llvm_libc_22_0_0_git7details22StringBufferWriterImplILb0EE4pushEc.exit.i.i.i.i.i.i108.i, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git7details22StringBufferWriterImplILb0EE4pushEc.exit.i.i.i.i.i.i108.i: ; preds = %if.then3.i.i.i.i.i.i.i105.i, %if.then.i.i.i.i.i.i.i102.i
  %writer.sroa.10.1.i.i.i.i109.i = phi i64 [ %inc.i.i.i.i.i.i.i107.i, %if.then3.i.i.i.i.i.i.i105.i ], [ 11, %if.then.i.i.i.i.i.i.i102.i ]
  %cmp.i.i.i.i.i.i110.i = icmp ult i32 %value.addr.09.i.i.i.i.i.i97.i, 10
  %.not.i.i.i.i.i.i111.i = or i1 %cmp.i.i.i.i.i.i.i.i104.i, %cmp.i.i.i.i.i.i110.i
  %2468 = call i64 @llvm.amdgcn.if.break.i64(i1 %.not.i.i.i.i.i.i111.i, i64 %phi.broken3973)
  %2469 = call i1 @llvm.amdgcn.loop.i64(i64 %2468)
  br i1 %2469, label %Flow3453, label %while.body.i.i.i.i.i.i95.i

Flow3453:                                         ; preds = %_ZN22__llvm_libc_22_0_0_git7details22StringBufferWriterImplILb0EE4pushEc.exit.i.i.i.i.i.i108.i
  %writer.sroa.10.1.i.i.i.i109.i.lcssa = phi i64 [ %writer.sroa.10.1.i.i.i.i109.i, %_ZN22__llvm_libc_22_0_0_git7details22StringBufferWriterImplILb0EE4pushEc.exit.i.i.i.i.i.i108.i ]
  %.lcssa4011 = phi i64 [ %2468, %_ZN22__llvm_libc_22_0_0_git7details22StringBufferWriterImplILb0EE4pushEc.exit.i.i.i.i.i.i108.i ]
  %cmp.i.i.i.i.i.i.i.i104.i.lcssa = phi i1 [ %cmp.i.i.i.i.i.i.i.i104.i, %_ZN22__llvm_libc_22_0_0_git7details22StringBufferWriterImplILb0EE4pushEc.exit.i.i.i.i.i.i108.i ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa4011)
  br label %Flow3454, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git15IntegerToStringIiNS_7details3FmtILh10ELb0ELb0ELb0ELm1EEEE13IntegerWriter25write_unsigned_number_decEjRNS1_22StringBufferWriterImplILb0EEE.exit.i.i.i.i.i112.i: ; preds = %Flow3454
  %cmp.i19.i.i.i.i115.i = icmp ne i64 %2400, 0
  %or.cond.not.i.i.i.i116.i = select i1 %2399, i1 true, i1 %cmp.i19.i.i.i.i115.i
  %or.cond.not.i.i.i.i116.i.inv = xor i1 %or.cond.not.i.i.i.i116.i, true
  %2470 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %or.cond.not.i.i.i.i116.i.inv)
  %2471 = extractvalue { i1, i64 } %2470, 0
  %2472 = extractvalue { i1, i64 } %2470, 1
  br i1 %2471, label %if.then3.i.i.lr.ph.i.i.i.i117.i, label %while.end.i.i.i.i.i119.i

if.then3.i.i.lr.ph.i.i.i.i117.i:                  ; preds = %_ZN22__llvm_libc_22_0_0_git15IntegerToStringIiNS_7details3FmtILh10ELb0ELb0ELb0ELm1EEEE13IntegerWriter25write_unsigned_number_decEjRNS1_22StringBufferWriterImplILb0EEE.exit.i.i.i.i.i112.i
  %arrayidx.i.i.i.i.i.i.i118.i = getelementptr inbounds nuw i8, ptr addrspace(5) %buffer.i.i.i, i32 10
  store i8 48, ptr addrspace(5) %arrayidx.i.i.i.i.i.i.i118.i, align 2, !tbaa !57, !noalias !121
  br label %while.end.i.i.i.i.i119.i, !amdgpu.uniform !42

while.end.i.i.i.i.i119.i:                         ; preds = %if.then3.i.i.lr.ph.i.i.i.i117.i, %_ZN22__llvm_libc_22_0_0_git15IntegerToStringIiNS_7details3FmtILh10ELb0ELb0ELb0ELm1EEEE13IntegerWriter25write_unsigned_number_decEjRNS1_22StringBufferWriterImplILb0EEE.exit.i.i.i.i.i112.i
  %writer.sroa.10.4.i.i.i.i120.i = phi i64 [ 1, %if.then3.i.i.lr.ph.i.i.i.i117.i ], [ %2400, %_ZN22__llvm_libc_22_0_0_git15IntegerToStringIiNS_7details3FmtILh10ELb0ELb0ELb0ELm1EEEE13IntegerWriter25write_unsigned_number_decEjRNS1_22StringBufferWriterImplILb0EEE.exit.i.i.i.i.i112.i ]
  call void @llvm.amdgcn.end.cf.i64(i64 %2472)
  %cmp4.i.i.i.i.i121.i = icmp sgt i32 %RHS, -1
  %brmerge.i.i.i.i.i122.i = or i1 %cmp4.i.i.i.i.i121.i, %2399
  %brmerge.i.i.i.i.i122.i.inv = xor i1 %brmerge.i.i.i.i.i122.i, true
  %2473 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %brmerge.i.i.i.i.i122.i.inv)
  %2474 = extractvalue { i1, i64 } %2473, 0
  %2475 = extractvalue { i1, i64 } %2473, 1
  br i1 %2474, label %if.then.i14.i.i.i.i.i123.i, label %_ZN22__llvm_libc_22_0_0_git15IntegerToStringIiNS_7details3FmtILh10ELb0ELb0ELb0ELm1EEEEC2Ei.exit.i.i.i128.i

if.then.i14.i.i.i.i.i123.i:                       ; preds = %while.end.i.i.i.i.i119.i
  %cmp.i.i17.i.i.i.i.i124.i = icmp ne i64 %writer.sroa.10.4.i.i.i.i120.i, 11
  %2476 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.i.i17.i.i.i.i.i124.i)
  %2477 = extractvalue { i1, i64 } %2476, 0
  %2478 = extractvalue { i1, i64 } %2476, 1
  br i1 %2477, label %if.then3.i18.i.i.i.i.i125.i, label %Flow3436

if.then3.i18.i.i.i.i.i125.i:                      ; preds = %if.then.i14.i.i.i.i.i123.i
  %2479 = trunc i64 %writer.sroa.10.4.i.i.i.i120.i to i32
  %2480 = xor i32 %2479, -1
  %2481 = getelementptr i8, ptr addrspace(5) %buffer.i.i.i, i32 %2480
  %arrayidx.i.i19.i.i.i.i.i126.i2743 = getelementptr i8, ptr addrspace(5) %2481, i32 11
  %inc.i20.i.i.i.i.i127.i = add i64 %writer.sroa.10.4.i.i.i.i120.i, 1
  store i8 45, ptr addrspace(5) %arrayidx.i.i19.i.i.i.i.i126.i2743, align 1, !tbaa !57, !noalias !121
  br label %Flow3436, !amdgpu.uniform !42

Flow3436:                                         ; preds = %if.then3.i18.i.i.i.i.i125.i, %if.then.i14.i.i.i.i.i123.i
  %2482 = phi i64 [ %inc.i20.i.i.i.i.i127.i, %if.then3.i18.i.i.i.i.i125.i ], [ 11, %if.then.i14.i.i.i.i.i123.i ]
  call void @llvm.amdgcn.end.cf.i64(i64 %2478)
  br label %_ZN22__llvm_libc_22_0_0_git15IntegerToStringIiNS_7details3FmtILh10ELb0ELb0ELb0ELm1EEEEC2Ei.exit.i.i.i128.i, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git15IntegerToStringIiNS_7details3FmtILh10ELb0ELb0ELb0ELm1EEEEC2Ei.exit.i.i.i128.i: ; preds = %while.end.i.i.i.i.i119.i, %Flow3436
  %writer.sroa.10.5.i.i.i.i129.i = phi i64 [ %writer.sroa.10.4.i.i.i.i120.i, %while.end.i.i.i.i.i119.i ], [ %2482, %Flow3436 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %2475)
  %written.i.i.i.i130.i = getelementptr inbounds nuw i8, ptr addrspace(5) %buffer.i.i.i, i32 16
  store i64 %writer.sroa.10.5.i.i.i.i129.i, ptr addrspace(5) %written.i.i.i.i130.i, align 8, !tbaa !91, !noalias !121
  %cmp.i.not.i.i.i.i.i133.i = icmp ne i64 %writer.sroa.10.5.i.i.i.i129.i, 0
  %2483 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.i.not.i.i.i.i.i133.i)
  %2484 = extractvalue { i1, i64 } %2483, 0
  %2485 = extractvalue { i1, i64 } %2483, 1
  br i1 %2484, label %if.then.i.i.i.i.i.i134.i, label %Flow3435

if.then.i.i.i.i.i.i134.i:                         ; preds = %_ZN22__llvm_libc_22_0_0_git15IntegerToStringIiNS_7details3FmtILh10ELb0ELb0ELb0ELm1EEEEC2Ei.exit.i.i.i128.i
  %2486 = call { i64, i1 } @llvm.uadd.with.overflow.i64(i64 %writer.sroa.10.5.i.i.i.i129.i, i64 1)
  %ov3345 = extractvalue { i64, i1 } %2486, 1
  %ov3345.inv = xor i1 %ov3345, true
  %2487 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %ov3345.inv)
  %2488 = extractvalue { i1, i64 } %2487, 0
  %2489 = extractvalue { i1, i64 } %2487, 1
  br i1 %2488, label %if.end.i.i.i.i.i.i.i137.i, label %Flow3434

if.end.i.i.i.i.i.i.i137.i:                        ; preds = %if.then.i.i.i.i.i.i134.i
  %math3344 = extractvalue { i64, i1 } %2486, 0
  %cmp2.i.i.i.i.i.i.i138.i = icmp ult i64 %math3344, 1676976733973595601
  %2490 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp2.i.i.i.i.i.i.i138.i)
  %2491 = extractvalue { i1, i64 } %2490, 0
  %2492 = extractvalue { i1, i64 } %2490, 1
  br i1 %2491, label %if.then3.i.i.i.i4.i.i.i152.i, label %if.end4.i.i.i.i.i.i.i139.i

if.then3.i.i.i.i4.i.i.i152.i:                     ; preds = %if.end.i.i.i.i.i.i.i137.i
  %mul.i.i.i.i.i.i.i153.i = mul nuw i64 %math3344, 11
  %div18.i.i.i.i.i.i.i154.i = lshr i64 %mul.i.i.i.i.i.i.i153.i, 3
  br label %if.end4.i.i.i.i.i.i.i139.i, !amdgpu.uniform !42

Flow3434:                                         ; preds = %if.end4.i.i.i.i.i.i.i139.i, %if.then.i.i.i.i.i.i134.i
  %2493 = phi ptr [ %cond.i.i1884, %if.end4.i.i.i.i.i.i.i139.i ], [ addrspacecast (ptr addrspace(4) @_ZN22__llvm_libc_22_0_0_git3cpp6string14NULL_CHARACTERE to ptr), %if.then.i.i.i.i.i.i134.i ]
  call void @llvm.amdgcn.end.cf.i64(i64 %2489)
  br label %_ZN22__llvm_libc_22_0_0_git3cpp6string7reserveEm.exit.i.i.i.i.i.i143.i, !amdgpu.uniform !42

if.end4.i.i.i.i.i.i.i139.i:                       ; preds = %if.then3.i.i.i.i4.i.i.i152.i, %if.end.i.i.i.i.i.i.i137.i
  %new_capacity.addr.0.i.i.i.i.i.i.i140.i = phi i64 [ %div18.i.i.i.i.i.i.i154.i, %if.then3.i.i.i.i4.i.i.i152.i ], [ %math3344, %if.end.i.i.i.i.i.i.i137.i ]
  call void @llvm.amdgcn.end.cf.i64(i64 %2492)
  %sub.i.i1878 = add i64 %new_capacity.addr.0.i.i.i.i.i.i.i140.i, 7
  %div2.i.i1879 = and i64 %sub.i.i1878, -8
  %2494 = load ptr, ptr addrspace(1) @_ZN12_GLOBAL__N_13ptrE, align 8, !tbaa !55
  %add.ptr.i.i1880 = getelementptr inbounds nuw i8, ptr %2494, i64 %div2.i.i1879
  store ptr %add.ptr.i.i1880, ptr addrspace(1) @_ZN12_GLOBAL__N_13ptrE, align 8, !tbaa !55
  %sub.ptr.lhs.cast.i.i1881 = ptrtoint ptr %add.ptr.i.i1880 to i64
  %sub.ptr.sub.i.i1882 = sub i64 %sub.ptr.lhs.cast.i.i1881, ptrtoint (ptr addrspacecast (ptr addrspace(1) @_ZN12_GLOBAL__N_16memoryE to ptr) to i64)
  %cmp.i.i1883 = icmp ugt i64 %sub.ptr.sub.i.i1882, 65335
  %cond.i.i1884 = select i1 %cmp.i.i1883, ptr null, ptr %2494
  br label %Flow3434, !amdgpu.uniform !42

Flow3435:                                         ; preds = %Flow3433, %_ZN22__llvm_libc_22_0_0_git15IntegerToStringIiNS_7details3FmtILh10ELb0ELb0ELb0ELm1EEEEC2Ei.exit.i.i.i128.i
  %2495 = phi ptr [ %2493, %Flow3433 ], [ addrspacecast (ptr addrspace(4) @_ZN22__llvm_libc_22_0_0_git3cpp6string14NULL_CHARACTERE to ptr), %_ZN22__llvm_libc_22_0_0_git15IntegerToStringIiNS_7details3FmtILh10ELb0ELb0ELb0ELm1EEEEC2Ei.exit.i.i.i128.i ]
  call void @llvm.amdgcn.end.cf.i64(i64 %2485)
  br label %if.end.i.i.i.i.i.i145.i, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git3cpp6string7reserveEm.exit.i.i.i.i.i.i143.i: ; preds = %Flow3434
  br label %loadstoreloop2187, !amdgpu.uniform !42

loadstoreloop2187:                                ; preds = %_ZN22__llvm_libc_22_0_0_git3cpp6string7reserveEm.exit.i.i.i.i.i.i143.i, %loadstoreloop2187
  %phi.broken3974 = phi i64 [ 0, %_ZN22__llvm_libc_22_0_0_git3cpp6string7reserveEm.exit.i.i.i.i.i.i143.i ], [ %2500, %loadstoreloop2187 ]
  %2496 = phi i64 [ %2498, %loadstoreloop2187 ], [ 0, %_ZN22__llvm_libc_22_0_0_git3cpp6string7reserveEm.exit.i.i.i.i.i.i143.i ]
  %2497 = getelementptr i8, ptr %2493, i64 %2496
  store i8 0, ptr %2497, align 1
  %2498 = add i64 %2496, 1
  %2499 = icmp uge i64 %2498, %writer.sroa.10.5.i.i.i.i129.i
  %2500 = call i64 @llvm.amdgcn.if.break.i64(i1 %2499, i64 %phi.broken3974)
  %2501 = call i1 @llvm.amdgcn.loop.i64(i64 %2500)
  br i1 %2501, label %Flow3433, label %loadstoreloop2187

Flow3433:                                         ; preds = %loadstoreloop2187
  %.lcssa4010 = phi i64 [ %2500, %loadstoreloop2187 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa4010)
  br label %Flow3435, !amdgpu.uniform !42

if.end.i.i.i.i.i.i145.i:                          ; preds = %Flow3435
  %cmp.not.i6.i.i.i.i.i.i147.i = icmp ne ptr %2495, addrspacecast (ptr addrspace(4) @_ZN22__llvm_libc_22_0_0_git3cpp6string14NULL_CHARACTERE to ptr)
  %2502 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.not.i6.i.i.i.i.i.i147.i)
  %2503 = extractvalue { i1, i64 } %2502, 0
  %2504 = extractvalue { i1, i64 } %2502, 1
  br i1 %2503, label %if.then.i.i.i.i2.i.i.i148.i, label %_ZN22__llvm_libc_22_0_0_git7testing8internal13describeValueIiEENS_3cpp9enable_ifIXaasr3cppE13is_integral_vIT_ElestS5_Lm8EENS3_6stringEE4typeES5_.exit164.i

if.then.i.i.i.i2.i.i.i148.i:                      ; preds = %if.end.i.i.i.i.i.i145.i
  %arrayidx.i.i.i.i3.i.i.i149.i = getelementptr inbounds nuw i8, ptr %2495, i64 %writer.sroa.10.5.i.i.i.i129.i
  store i8 0, ptr %arrayidx.i.i.i.i3.i.i.i149.i, align 1, !tbaa !57, !noalias !121
  br label %_ZN22__llvm_libc_22_0_0_git7testing8internal13describeValueIiEENS_3cpp9enable_ifIXaasr3cppE13is_integral_vIT_ElestS5_Lm8EENS3_6stringEE4typeES5_.exit164.i, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git7testing8internal13describeValueIiEENS_3cpp9enable_ifIXaasr3cppE13is_integral_vIT_ElestS5_Lm8EENS3_6stringEE4typeES5_.exit164.i: ; preds = %if.then.i.i.i.i2.i.i.i148.i, %if.end.i.i.i.i.i.i145.i
  call void @llvm.amdgcn.end.cf.i64(i64 %2504)
  %2505 = trunc i64 %writer.sroa.10.5.i.i.i.i129.i to i32
  %2506 = and i64 %writer.sroa.10.5.i.i.i.i129.i, 15
  %2507 = sub i64 %writer.sroa.10.5.i.i.i.i129.i, %2506
  %2508 = icmp ne i64 %2507, 0
  %2509 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %2508)
  %2510 = extractvalue { i1, i64 } %2509, 0
  %2511 = extractvalue { i1, i64 } %2509, 1
  br i1 %2510, label %loop-memcpy-expansion2309.preheader, label %Flow3432

loop-memcpy-expansion2309.preheader:              ; preds = %_ZN22__llvm_libc_22_0_0_git7testing8internal13describeValueIiEENS_3cpp9enable_ifIXaasr3cppE13is_integral_vIT_ElestS5_Lm8EENS3_6stringEE4typeES5_.exit164.i
  %2512 = sub i32 11, %2505
  %scevgep2878 = getelementptr i8, ptr addrspace(5) %buffer.i.i.i, i32 %2512
  br label %loop-memcpy-expansion2309, !amdgpu.uniform !42

Flow3432:                                         ; preds = %Flow3431, %_ZN22__llvm_libc_22_0_0_git7testing8internal13describeValueIiEENS_3cpp9enable_ifIXaasr3cppE13is_integral_vIT_ElestS5_Lm8EENS3_6stringEE4typeES5_.exit164.i
  call void @llvm.amdgcn.end.cf.i64(i64 %2511)
  br label %loop-memcpy-residual-header2312, !amdgpu.uniform !42

loop-memcpy-expansion2309:                        ; preds = %loop-memcpy-expansion2309.preheader, %loop-memcpy-expansion2309
  %phi.broken3975 = phi i64 [ 0, %loop-memcpy-expansion2309.preheader ], [ %2517, %loop-memcpy-expansion2309 ]
  %lsr.iv2879 = phi ptr addrspace(5) [ %scevgep2880, %loop-memcpy-expansion2309 ], [ %scevgep2878, %loop-memcpy-expansion2309.preheader ]
  %loop-index2310 = phi i64 [ %2515, %loop-memcpy-expansion2309 ], [ 0, %loop-memcpy-expansion2309.preheader ]
  %2513 = load <4 x i32>, ptr addrspace(5) %lsr.iv2879, align 1
  %2514 = getelementptr i8, ptr %2495, i64 %loop-index2310
  store <4 x i32> %2513, ptr %2514, align 1
  %2515 = add i64 %loop-index2310, 16
  %scevgep2880 = getelementptr i8, ptr addrspace(5) %lsr.iv2879, i32 16
  %2516 = icmp uge i64 %2515, %2507
  %2517 = call i64 @llvm.amdgcn.if.break.i64(i1 %2516, i64 %phi.broken3975)
  %2518 = call i1 @llvm.amdgcn.loop.i64(i64 %2517)
  br i1 %2518, label %Flow3431, label %loop-memcpy-expansion2309

Flow3430:                                         ; preds = %Flow3429, %loop-memcpy-residual-header2312
  call void @llvm.amdgcn.end.cf.i64(i64 %2875)
  br label %post-loop-memcpy-expansion2308, !amdgpu.uniform !42

loop-memcpy-residual2311:                         ; preds = %loop-memcpy-residual2311.preheader, %loop-memcpy-residual2311
  %phi.broken3976 = phi i64 [ 0, %loop-memcpy-residual2311.preheader ], [ %2522, %loop-memcpy-residual2311 ]
  %lsr.iv2874 = phi ptr addrspace(5) [ %scevgep2875, %loop-memcpy-residual2311 ], [ %scevgep2873, %loop-memcpy-residual2311.preheader ]
  %residual-loop-index2313 = phi i64 [ %2520, %loop-memcpy-residual2311 ], [ 0, %loop-memcpy-residual2311.preheader ]
  %2519 = load i8, ptr addrspace(5) %lsr.iv2874, align 1
  %scevgep2877 = getelementptr i8, ptr %scevgep2876, i64 %residual-loop-index2313
  store i8 %2519, ptr %scevgep2877, align 1
  %2520 = add i64 %residual-loop-index2313, 1
  %scevgep2875 = getelementptr i8, ptr addrspace(5) %lsr.iv2874, i32 1
  %2521 = icmp uge i64 %2520, %2506
  %2522 = call i64 @llvm.amdgcn.if.break.i64(i1 %2521, i64 %phi.broken3976)
  %2523 = call i1 @llvm.amdgcn.loop.i64(i64 %2522)
  br i1 %2523, label %Flow3429, label %loop-memcpy-residual2311

Flow3429:                                         ; preds = %loop-memcpy-residual2311
  %.lcssa4008 = phi i64 [ %2522, %loop-memcpy-residual2311 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa4008)
  br label %Flow3430, !amdgpu.uniform !42

post-loop-memcpy-expansion2308:                   ; preds = %Flow3430
  call void @llvm.lifetime.end.p5(ptr addrspace(5) %buffer.i.i.i) #34, !noalias !121
  br label %for.cond.i.i.i1886, !amdgpu.uniform !42

for.cond.i.i.i1886:                               ; preds = %post-loop-memcpy-expansion2308, %Flow3428
  %2524 = phi i32 [ %2562, %Flow3428 ], [ poison, %post-loop-memcpy-expansion2308 ]
  %2525 = phi i64 [ %2563, %Flow3428 ], [ poison, %post-loop-memcpy-expansion2308 ]
  %2526 = phi i32 [ %2564, %Flow3428 ], [ poison, %post-loop-memcpy-expansion2308 ]
  %retval.sroa.6.0.i.i.i1887 = phi i32 [ %2567, %Flow3428 ], [ undef, %post-loop-memcpy-expansion2308 ]
  %retval.sroa.2.0.i.i.i1889 = phi i64 [ %2566, %Flow3428 ], [ undef, %post-loop-memcpy-expansion2308 ]
  %retval.sroa.8.0.i.i.i1890 = phi i32 [ %2565, %Flow3428 ], [ undef, %post-loop-memcpy-expansion2308 ]
  %index.0.i.i.i1891 = phi i32 [ %2569, %Flow3428 ], [ 0, %post-loop-memcpy-expansion2308 ]
  %2527 = load i32, ptr addrspace(1) @__llvm_rpc_client, align 8, !tbaa !68
  %cmp.not.i.i.i1892 = icmp ult i32 %index.0.i.i.i1891, %2527
  %spec.store.select.i.i.i1893 = select i1 %cmp.not.i.i.i1892, i32 %index.0.i.i.i1891, i32 0
  %2528 = tail call noundef i64 @llvm.amdgcn.ballot.i64(i1 true)
  %2529 = and i32 %spec.store.select.i.i.i1893, 31
  %2530 = lshr i64 %2528, %sh_prom.i.i.i.i
  %2531 = trunc i64 %2530 to i32
  %conv4.i.i.i.i.i1897 = and i32 %2531, 1
  %shl.i.i.i.i.i1898 = shl nuw i32 %conv4.i.i.i.i.i1897, %2529
  %2532 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2670, !amdgpu.uniform !42

2533:                                             ; preds = %ComputeEnd2671
  %sunkaddr3346 = getelementptr inbounds i8, ptr addrspace(1) %4003, i64 40
  %2534 = atomicrmw or ptr addrspace(1) %sunkaddr3346, i32 %.lcssa4002 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %2535, !amdgpu.uniform !42

2535:                                             ; preds = %2533, %ComputeEnd2671
  %2536 = phi i32 [ %2534, %2533 ], [ poison, %ComputeEnd2671 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %4012)
  %2537 = call i32 @llvm.amdgcn.readfirstlane.i32(i32 %2536)
  %2538 = or i32 %2537, %.lcssa4003
  %shl5.i.i.i.i.i1899 = shl nuw i32 1, %2529
  %and.i.i.i.i.i1900 = and i32 %shl5.i.i.i.i.i1899, %2538
  %tobool3.i.i.i.i1901 = icmp ne i32 %and.i.i.i.i.i1900, 0
  %2539 = tail call i64 @llvm.amdgcn.ballot.i64(i1 %tobool3.i.i.i.i1901)
  %and.i.i.i.i.i.i1902 = and i64 %2539, %2528
  %cmp.i.not.i.i.i1903 = icmp ne i64 %2528, %and.i.i.i.i.i.i1902
  br i1 %cmp.i.not.i.i.i1903, label %if.then.i.i.i.i1904, label %Flow3427, !amdgpu.uniform !42

if.then.i.i.i.i1904:                              ; preds = %2535
  fence syncscope("agent") acquire
  %2540 = load <2 x i64>, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !75
  %2541 = extractelement <2 x i64> %2540, i32 0
  %2542 = inttoptr i64 %2541 to ptr
  %2543 = extractelement <2 x i64> %2540, i32 1
  %2544 = inttoptr i64 %2543 to ptr
  %idxprom.i.i.i.i1905 = zext i32 %spec.store.select.i.i.i1893 to i64
  %arrayidx.i.i.i.i1906 = getelementptr inbounds nuw i32, ptr %2542, i64 %idxprom.i.i.i.i1905, !amdgpu.uniform !42
  %2545 = load atomic i32, ptr %arrayidx.i.i.i.i1906 monotonic, align 4
  %2546 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %2545)
  %arrayidx.i45.i.i.i1907 = getelementptr inbounds nuw i32, ptr %2544, i64 %idxprom.i.i.i.i1905, !amdgpu.uniform !42
  %2547 = load atomic i32, ptr %arrayidx.i45.i.i.i1907 monotonic, align 4
  %2548 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %2547)
  %cmp.i46.not.i.i.i1908 = icmp ne i32 %2546, %2548
  br i1 %cmp.i46.not.i.i.i1908, label %if.then12.i.i.i1909, label %Flow3425, !amdgpu.uniform !42

if.then12.i.i.i1909:                              ; preds = %if.then.i.i.i.i1904
  fence syncscope("agent") release
  %2549 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %2528, i1 true)
  %iszero.i.i.i.i.i.i.i1910 = icmp ne i64 %2528, 0
  %cmp2.i.i.i.i.i.i1911 = icmp eq i64 %2549, %sh_prom.i.i.i.i
  %cmp.i.i.i.i.i.i1912 = select i1 %iszero.i.i.i.i.i.i.i1910, i1 %cmp2.i.i.i.i.i.i1911, i1 false
  %conv4.i.i51.i.i.i1913 = zext i1 %cmp.i.i.i.i.i.i1912 to i32
  %shl.i.i52.i.i.i1914 = shl nuw i32 %conv4.i.i51.i.i.i1913, %2529
  %xor.i.i.i.i.i1915 = xor i32 %shl.i.i52.i.i.i1914, -1
  %2550 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2675, !amdgpu.uniform !42

2551:                                             ; preds = %ComputeEnd2676
  %sunkaddr3347 = getelementptr inbounds i8, ptr addrspace(1) %4003, i64 40
  %2552 = atomicrmw and ptr addrspace(1) %sunkaddr3347, i32 %.lcssa4004 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %2553, !amdgpu.uniform !42

2553:                                             ; preds = %2551, %ComputeEnd2676
  call void @llvm.amdgcn.end.cf.i64(i64 %4030)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %Flow3425, !amdgpu.uniform !42

if.end14.i.i.i1975:                               ; preds = %Flow3425
  %2554 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %2528, i1 true)
  %iszero.i.i.i.i.i.i1976 = icmp ne i64 %2528, 0
  %cmp2.i.i.i.i.i1977 = icmp eq i64 %2554, %sh_prom.i.i.i.i
  %cmp.i.i.i.i.i1978 = select i1 %iszero.i.i.i.i.i.i1976, i1 %cmp2.i.i.i.i.i1977, i1 false
  %2555 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.i.i.i.i.i1978)
  %2556 = extractvalue { i1, i64 } %2555, 0
  %2557 = extractvalue { i1, i64 } %2555, 1
  br i1 %2556, label %if.then16.i.i.i1980, label %if.end22.i.i.i1979

if.then16.i.i.i1980:                              ; preds = %if.end14.i.i.i1975
  %2558 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !76
  %opcode.i.i.i1981 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %2558, i64 %idxprom.i.i.i.i1905, i32 1
  store i32 1660944387, ptr %opcode.i.i.i1981, align 8, !tbaa !77
  %arrayidx21.i.i.i1982 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %2558, i64 %idxprom.i.i.i.i1905
  store i64 %2528, ptr %arrayidx21.i.i.i1982, align 8, !tbaa !80
  br label %if.end22.i.i.i1979, !amdgpu.uniform !42

Flow3426:                                         ; preds = %if.end22.i.i.i1979, %Flow3425
  %2559 = phi i32 [ %2546, %if.end22.i.i.i1979 ], [ %retval.sroa.8.0.i.i.i1890, %Flow3425 ]
  %2560 = phi i64 [ %2528, %if.end22.i.i.i1979 ], [ %retval.sroa.2.0.i.i.i1889, %Flow3425 ]
  %2561 = phi i32 [ %spec.store.select.i.i.i1893, %if.end22.i.i.i1979 ], [ %retval.sroa.6.0.i.i.i1887, %Flow3425 ]
  br label %cleanup26.i.i.i1916, !amdgpu.uniform !42

if.end22.i.i.i1979:                               ; preds = %if.then16.i.i.i1980, %if.end14.i.i.i1975
  call void @llvm.amdgcn.end.cf.i64(i64 %2557)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %Flow3426, !amdgpu.uniform !42

Flow3427:                                         ; preds = %cleanup26.i.i.i1916, %2535
  %2562 = phi i32 [ %2561, %cleanup26.i.i.i1916 ], [ %2524, %2535 ]
  %2563 = phi i64 [ %2560, %cleanup26.i.i.i1916 ], [ %2525, %2535 ]
  %2564 = phi i32 [ %2559, %cleanup26.i.i.i1916 ], [ %2526, %2535 ]
  %2565 = phi i32 [ %2559, %cleanup26.i.i.i1916 ], [ %retval.sroa.8.0.i.i.i1890, %2535 ]
  %2566 = phi i64 [ %2560, %cleanup26.i.i.i1916 ], [ %retval.sroa.2.0.i.i.i1889, %2535 ]
  %2567 = phi i32 [ %2561, %cleanup26.i.i.i1916 ], [ %retval.sroa.6.0.i.i.i1887, %2535 ]
  %2568 = phi i1 [ %cmp.i46.not.i.i.i1908, %cleanup26.i.i.i1916 ], [ true, %2535 ]
  br i1 %2568, label %for.inc.i.i.i1921, label %Flow3428, !amdgpu.uniform !42

cleanup26.i.i.i1916:                              ; preds = %Flow3426
  br label %Flow3427, !amdgpu.uniform !42

for.inc.i.i.i1921:                                ; preds = %Flow3427
  %inc.i.i.i1926 = add i32 %spec.store.select.i.i.i1893, 1
  br label %Flow3428, !amdgpu.uniform !42

Flow3428:                                         ; preds = %for.inc.i.i.i1921, %Flow3427
  %2569 = phi i32 [ %inc.i.i.i1926, %for.inc.i.i.i1921 ], [ poison, %Flow3427 ]
  %2570 = phi i1 [ false, %for.inc.i.i.i1921 ], [ true, %Flow3427 ]
  br i1 %2570, label %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1927, label %for.cond.i.i.i1886, !amdgpu.uniform !42

_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1927: ; preds = %Flow3428
  %.lcssa4007 = phi i32 [ %2562, %Flow3428 ]
  %.lcssa4006 = phi i64 [ %2563, %Flow3428 ]
  %.lcssa4005 = phi i32 [ %2564, %Flow3428 ]
  fence acquire
  %2571 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %mul.i.i.i.i.i.i1928 = shl i32 %.lcssa4007, 6
  %idxprom.i19.i.i.i.i.i1929 = zext i32 %mul.i.i.i.i.i.i1928 to i64
  %arrayidx.i20.i.i.i.i.i1930 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %2571, i64 %idxprom.i19.i.i.i.i.i1929
  %arrayidx.i22.i.i.i.i.i1931 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %arrayidx.i20.i.i.i.i.i1930, i64 %sh_prom.i.i.i.i
  store i64 %writer.sroa.10.5.i.i.i.i129.i, ptr %arrayidx.i22.i.i.i.i.i1931, align 8, !tbaa !82
  %spec.select.i.i.i.i.i.i.i1932 = tail call i64 @llvm.umin.i64(i64 %writer.sroa.10.5.i.i.i.i129.i, i64 56)
  %2572 = and i64 %spec.select.i.i.i.i.i.i.i1932, 15
  %2573 = sub i64 %spec.select.i.i.i.i.i.i.i1932, %2572
  %2574 = icmp ne i64 %2573, 0
  %2575 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %2574)
  %2576 = extractvalue { i1, i64 } %2575, 0
  %2577 = extractvalue { i1, i64 } %2575, 1
  br i1 %2576, label %loop-memcpy-expansion2195.preheader, label %Flow3424

loop-memcpy-expansion2195.preheader:              ; preds = %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1927
  %2578 = add nuw nsw i64 %75, 8
  %scevgep2870 = getelementptr i8, ptr %2571, i64 %2578
  %2579 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i1929, 6
  %scevgep2871 = getelementptr i8, ptr %scevgep2870, i64 %2579
  br label %loop-memcpy-expansion2195, !amdgpu.uniform !42

Flow3424:                                         ; preds = %Flow3423, %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1927
  call void @llvm.amdgcn.end.cf.i64(i64 %2577)
  br label %loop-memcpy-residual-header2198, !amdgpu.uniform !42

loop-memcpy-expansion2195:                        ; preds = %loop-memcpy-expansion2195.preheader, %loop-memcpy-expansion2195
  %phi.broken3977 = phi i64 [ 0, %loop-memcpy-expansion2195.preheader ], [ %2584, %loop-memcpy-expansion2195 ]
  %loop-index2196 = phi i64 [ %2582, %loop-memcpy-expansion2195 ], [ 0, %loop-memcpy-expansion2195.preheader ]
  %2580 = getelementptr i8, ptr %2495, i64 %loop-index2196
  %2581 = load <4 x i32>, ptr %2580, align 1
  %scevgep2872 = getelementptr i8, ptr %scevgep2871, i64 %loop-index2196
  store <4 x i32> %2581, ptr %scevgep2872, align 1
  %2582 = add i64 %loop-index2196, 16
  %2583 = icmp uge i64 %2582, %2573
  %2584 = call i64 @llvm.amdgcn.if.break.i64(i1 %2583, i64 %phi.broken3977)
  %2585 = call i1 @llvm.amdgcn.loop.i64(i64 %2584)
  br i1 %2585, label %Flow3423, label %loop-memcpy-expansion2195

Flow3422:                                         ; preds = %Flow3420, %Flow3421
  %2586 = phi i64 [ %2757, %Flow3420 ], [ %2749, %Flow3421 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %2752)
  br label %post-loop-memcpy-expansion2194, !amdgpu.uniform !42

loop-memcpy-residual2197:                         ; preds = %loop-memcpy-residual2197.preheader, %loop-memcpy-residual2197
  %phi.broken3978 = phi i64 [ 0, %loop-memcpy-residual2197.preheader ], [ %2590, %loop-memcpy-residual2197 ]
  %residual-loop-index2199 = phi i64 [ %2588, %loop-memcpy-residual2197 ], [ 0, %loop-memcpy-residual2197.preheader ]
  %scevgep2869 = getelementptr i8, ptr %scevgep2868, i64 %residual-loop-index2199
  %2587 = load i8, ptr %scevgep2869, align 1
  %scevgep2867 = getelementptr i8, ptr %scevgep2866, i64 %residual-loop-index2199
  store i8 %2587, ptr %scevgep2867, align 1
  %2588 = add i64 %residual-loop-index2199, 1
  %2589 = icmp uge i64 %2588, %2572
  %2590 = call i64 @llvm.amdgcn.if.break.i64(i1 %2589, i64 %phi.broken3978)
  %2591 = call i1 @llvm.amdgcn.loop.i64(i64 %2590)
  br i1 %2591, label %Flow3420, label %loop-memcpy-residual2197

Flow3420:                                         ; preds = %loop-memcpy-residual2197
  %.lcssa4000 = phi i64 [ %2590, %loop-memcpy-residual2197 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa4000)
  br label %Flow3422, !amdgpu.uniform !42

post-loop-memcpy-expansion2194:                   ; preds = %Flow3422
  %tobool.not.i.i.i.i.i.i1934 = icmp eq i32 %.lcssa4005, 0
  %conv.i.i.i.i.i.i1935 = zext i1 %tobool.not.i.i.i.i.i.i1934 to i32
  fence release
  %2592 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %idxprom.i23.i.i.i.i.i1936 = zext i32 %.lcssa4007 to i64
  %arrayidx.i24.i.i.i.i.i1937 = getelementptr inbounds nuw i32, ptr %2592, i64 %idxprom.i23.i.i.i.i.i1936
  store atomic i32 %conv.i.i.i.i.i.i1935, ptr %arrayidx.i24.i.i.i.i.i1937 monotonic, align 4
  %2593 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !84
  %arrayidx.i.i5.i.i1938 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %2593, i64 %idxprom.i23.i.i.i.i.i1936, !amdgpu.uniform !42
  %2594 = load i64, ptr %arrayidx.i.i5.i.i1938, align 8, !tbaa !80
  %scevgep = getelementptr i8, ptr %2495, i64 56
  %2595 = add i64 %writer.sroa.10.5.i.i.i.i129.i, -56
  br label %while.cond.i.i.i.i1939, !amdgpu.uniform !42

while.cond.i.i.i.i1939:                           ; preds = %post-loop-memcpy-expansion2194, %Flow3419
  %phi.broken3981 = phi i64 [ 0, %post-loop-memcpy-expansion2194 ], [ %2612, %Flow3419 ]
  %lsr.iv2860 = phi i64 [ %2610, %Flow3419 ], [ %2595, %post-loop-memcpy-expansion2194 ]
  %lsr.iv = phi ptr [ %2609, %Flow3419 ], [ %scevgep, %post-loop-memcpy-expansion2194 ]
  %port.sroa.37.0.i.i1940 = phi i32 [ %2608, %Flow3419 ], [ %conv.i.i.i.i.i.i1935, %post-loop-memcpy-expansion2194 ]
  %idx.0.i.i.i.i1941 = phi i64 [ %2607, %Flow3419 ], [ 56, %post-loop-memcpy-expansion2194 ]
  %cmp.i.i.i.i1942 = icmp ugt i64 %writer.sroa.10.5.i.i.i.i129.i, %idx.0.i.i.i.i1941
  %2596 = tail call i64 @llvm.amdgcn.ballot.i64(i1 %cmp.i.i.i.i1942)
  %and.i.i.i.i6.i.i1943 = and i64 %2596, %2594
  %tobool.not.i.i.i.i1944 = icmp ne i64 %and.i.i.i.i6.i.i1943, 0
  %2597 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i22.i.i.i1945 = getelementptr inbounds nuw i32, ptr %2597, i64 %idxprom.i23.i.i.i.i.i1936, !amdgpu.uniform !42
  %2598 = load atomic i32, ptr %arrayidx.i22.i.i.i1945 monotonic, align 4
  %2599 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %tobool.not.i.i.i.i1944)
  %2600 = extractvalue { i1, i64 } %2599, 0
  %2601 = extractvalue { i1, i64 } %2599, 1
  br i1 %2600, label %cond.false.i8.i.i.i.i1946, label %Flow3419

cond.false.i8.i.i.i.i1946:                        ; preds = %while.cond.i.i.i.i1939
  %2602 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %2598)
  %2603 = icmp ne i32 %2602, %port.sroa.37.0.i.i1940
  br i1 %2603, label %while.body.i.i21.i.i.i.i1947.preheader, label %Flow3418, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i1947.preheader:           ; preds = %cond.false.i8.i.i.i.i1946
  br label %while.body.i.i21.i.i.i.i1947, !amdgpu.uniform !42

Flow3418:                                         ; preds = %Flow3417, %cond.false.i8.i.i.i.i1946
  br label %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i1950, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i1947:                     ; preds = %while.body.i.i21.i.i.i.i1947.preheader, %while.body.i.i21.i.i.i.i1947
  tail call void @llvm.amdgcn.s.sleep(i32 2)
  %2604 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i22.i.i.i.i1948 = getelementptr inbounds nuw i32, ptr %2604, i64 %idxprom.i23.i.i.i.i.i1936, !amdgpu.uniform !42
  %2605 = load atomic i32, ptr %arrayidx.i.i.i22.i.i.i.i1948 monotonic, align 4
  %2606 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %2605)
  %cmp.i.not.i.i23.i.i.i.i1949 = icmp eq i32 %2606, %port.sroa.37.0.i.i1940
  br i1 %cmp.i.not.i.i23.i.i.i.i1949, label %Flow3417, label %while.body.i.i21.i.i.i.i1947, !amdgpu.uniform !42

Flow3417:                                         ; preds = %while.body.i.i21.i.i.i.i1947
  br label %Flow3418, !amdgpu.uniform !42

Flow3419:                                         ; preds = %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1951, %while.cond.i.i.i.i1939
  %2607 = phi i64 [ %add.i.i.i.i1954, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1951 ], [ poison, %while.cond.i.i.i.i1939 ]
  %2608 = phi i32 [ %conv.i.i26.i.i.i.i1952, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1951 ], [ poison, %while.cond.i.i.i.i1939 ]
  %2609 = phi ptr [ %scevgep2856, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1951 ], [ poison, %while.cond.i.i.i.i1939 ]
  %2610 = phi i64 [ %lsr.iv.next, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1951 ], [ poison, %while.cond.i.i.i.i1939 ]
  %2611 = phi i1 [ false, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1951 ], [ true, %while.cond.i.i.i.i1939 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %2601)
  %2612 = call i64 @llvm.amdgcn.if.break.i64(i1 %2611, i64 %phi.broken3981)
  %2613 = call i1 @llvm.amdgcn.loop.i64(i64 %2612)
  br i1 %2613, label %cond.false.i.i.i1961, label %while.cond.i.i.i.i1939

_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i1950: ; preds = %Flow3418
  fence acquire
  %2614 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.i.i.i.i1942)
  %2615 = extractvalue { i1, i64 } %2614, 0
  %2616 = extractvalue { i1, i64 } %2614, 1
  br i1 %2615, label %if.then.i.i.i.i.i.i.i1955, label %Flow3416

if.then.i.i.i.i.i.i.i1955:                        ; preds = %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i1950
  %sub.i.i.i.i.i.i.i1957 = sub nuw i64 %writer.sroa.10.5.i.i.i.i129.i, %idx.0.i.i.i.i1941
  %spec.select.i.i.i36.i.i.i.i1958 = tail call i64 @llvm.umin.i64(i64 %sub.i.i.i.i.i.i.i1957, i64 64)
  %2617 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %2618 = and i64 %spec.select.i.i.i36.i.i.i.i1958, 15
  %2619 = sub i64 %spec.select.i.i.i36.i.i.i.i1958, %2618
  %2620 = icmp ne i64 %2619, 0
  %2621 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %2620)
  %2622 = extractvalue { i1, i64 } %2621, 0
  %2623 = extractvalue { i1, i64 } %2621, 1
  br i1 %2622, label %loop-memcpy-expansion.preheader, label %Flow3415

loop-memcpy-expansion.preheader:                  ; preds = %if.then.i.i.i.i.i.i.i1955
  %arrayidx.i20.i35.i.i.i.i1959 = getelementptr %"struct.rpc::Buffer", ptr %2617, i64 %idxprom.i19.i.i.i.i.i1929
  %arrayidx.i22.i38.i.i.i.i1960 = getelementptr %"struct.rpc::Buffer", ptr %arrayidx.i20.i35.i.i.i.i1959, i64 %sh_prom.i.i.i.i
  br label %loop-memcpy-expansion, !amdgpu.uniform !42

Flow3415:                                         ; preds = %Flow3414, %if.then.i.i.i.i.i.i.i1955
  call void @llvm.amdgcn.end.cf.i64(i64 %2623)
  br label %loop-memcpy-residual-header, !amdgpu.uniform !42

loop-memcpy-expansion:                            ; preds = %loop-memcpy-expansion.preheader, %loop-memcpy-expansion
  %phi.broken3979 = phi i64 [ 0, %loop-memcpy-expansion.preheader ], [ %2628, %loop-memcpy-expansion ]
  %loop-index = phi i64 [ %2626, %loop-memcpy-expansion ], [ 0, %loop-memcpy-expansion.preheader ]
  %scevgep2857 = getelementptr i8, ptr %lsr.iv, i64 %loop-index
  %2624 = load <4 x i32>, ptr %scevgep2857, align 1
  %2625 = getelementptr i8, ptr %arrayidx.i22.i38.i.i.i.i1960, i64 %loop-index
  store <4 x i32> %2624, ptr %2625, align 1
  %2626 = add i64 %loop-index, 16
  %2627 = icmp uge i64 %2626, %2619
  %2628 = call i64 @llvm.amdgcn.if.break.i64(i1 %2627, i64 %phi.broken3979)
  %2629 = call i1 @llvm.amdgcn.loop.i64(i64 %2628)
  br i1 %2629, label %Flow3414, label %loop-memcpy-expansion

Flow3413:                                         ; preds = %Flow3412, %loop-memcpy-residual-header
  call void @llvm.amdgcn.end.cf.i64(i64 %2742)
  br label %Flow3416, !amdgpu.uniform !42

loop-memcpy-residual:                             ; preds = %loop-memcpy-residual.preheader, %loop-memcpy-residual
  %phi.broken3980 = phi i64 [ 0, %loop-memcpy-residual.preheader ], [ %2633, %loop-memcpy-residual ]
  %residual-loop-index = phi i64 [ %2631, %loop-memcpy-residual ], [ 0, %loop-memcpy-residual.preheader ]
  %scevgep2864 = getelementptr i8, ptr %scevgep2863, i64 %residual-loop-index
  %2630 = load i8, ptr %scevgep2864, align 1
  %scevgep2862 = getelementptr i8, ptr %scevgep2861, i64 %residual-loop-index
  store i8 %2630, ptr %scevgep2862, align 1
  %2631 = add i64 %residual-loop-index, 1
  %2632 = icmp uge i64 %2631, %2618
  %2633 = call i64 @llvm.amdgcn.if.break.i64(i1 %2632, i64 %phi.broken3980)
  %2634 = call i1 @llvm.amdgcn.loop.i64(i64 %2633)
  br i1 %2634, label %Flow3412, label %loop-memcpy-residual

Flow3412:                                         ; preds = %loop-memcpy-residual
  %.lcssa3997 = phi i64 [ %2633, %loop-memcpy-residual ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa3997)
  br label %Flow3413, !amdgpu.uniform !42

_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1951: ; preds = %Flow3416
  %conv.i.i26.i.i.i.i1952 = xor i32 %port.sroa.37.0.i.i1940, 1
  fence release
  %2635 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %arrayidx.i24.i29.i.i.i.i1953 = getelementptr inbounds nuw i32, ptr %2635, i64 %idxprom.i23.i.i.i.i.i1936
  store atomic i32 %conv.i.i26.i.i.i.i1952, ptr %arrayidx.i24.i29.i.i.i.i1953 monotonic, align 4
  %add.i.i.i.i1954 = add i64 %idx.0.i.i.i.i1941, 64
  %scevgep2856 = getelementptr i8, ptr %lsr.iv, i64 64
  %lsr.iv.next = add i64 %lsr.iv2860, -64
  br label %Flow3419, !amdgpu.uniform !42

cond.false.i.i.i1961:                             ; preds = %Flow3419
  %.lcssa3999 = phi i64 [ %2612, %Flow3419 ]
  %port.sroa.37.0.i.i1940.lcssa = phi i32 [ %port.sroa.37.0.i.i1940, %Flow3419 ]
  %.lcssa3998 = phi i32 [ %2598, %Flow3419 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa3999)
  %2636 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %.lcssa3998)
  %2637 = icmp ne i32 %2636, %port.sroa.37.0.i.i1940.lcssa
  %2638 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %2637)
  %2639 = extractvalue { i1, i64 } %2638, 0
  %2640 = extractvalue { i1, i64 } %2638, 1
  br i1 %2639, label %while.body.i.i.i.i1962.preheader, label %Flow3411

while.body.i.i.i.i1962.preheader:                 ; preds = %cond.false.i.i.i1961
  br label %while.body.i.i.i.i1962, !amdgpu.uniform !42

Flow3411:                                         ; preds = %Flow3410, %cond.false.i.i.i1961
  call void @llvm.amdgcn.end.cf.i64(i64 %2640)
  br label %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsINS_3cpp6stringEEERS1_T_.exit1983, !amdgpu.uniform !42

while.body.i.i.i.i1962:                           ; preds = %while.body.i.i.i.i1962.preheader, %while.body.i.i.i.i1962
  %phi.broken3982 = phi i64 [ 0, %while.body.i.i.i.i1962.preheader ], [ %2644, %while.body.i.i.i.i1962 ]
  tail call void @llvm.amdgcn.s.sleep(i32 2)
  %2641 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i8.i.i1963 = getelementptr inbounds nuw i32, ptr %2641, i64 %idxprom.i23.i.i.i.i.i1936, !amdgpu.uniform !42
  %2642 = load atomic i32, ptr %arrayidx.i.i.i8.i.i1963 monotonic, align 4
  %2643 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %2642)
  %cmp.i.not.i.i.i.i1964 = icmp eq i32 %2643, %port.sroa.37.0.i.i1940.lcssa
  %2644 = call i64 @llvm.amdgcn.if.break.i64(i1 %cmp.i.not.i.i.i.i1964, i64 %phi.broken3982)
  %2645 = call i1 @llvm.amdgcn.loop.i64(i64 %2644)
  br i1 %2645, label %Flow3410, label %while.body.i.i.i.i1962

Flow3410:                                         ; preds = %while.body.i.i.i.i1962
  %.lcssa3995 = phi i64 [ %2644, %while.body.i.i.i.i1962 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa3995)
  br label %Flow3411, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsINS_3cpp6stringEEERS1_T_.exit1983: ; preds = %Flow3411
  fence acquire
  tail call void @llvm.amdgcn.wave.barrier()
  fence syncscope("agent") release
  %2646 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %.lcssa4006, i1 true)
  %iszero.i.i.i.i.i13.i.i1966 = icmp ne i64 %.lcssa4006, 0
  %cmp2.i.i.i.i14.i.i1967 = icmp eq i64 %2646, %sh_prom.i.i.i.i
  %cmp.i.i.i.i15.i.i1968 = select i1 %iszero.i.i.i.i.i13.i.i1966, i1 %cmp2.i.i.i.i14.i.i1967, i1 false
  %2647 = and i32 %.lcssa4007, 31
  %conv4.i.i.i19.i.i1972 = zext i1 %cmp.i.i.i.i15.i.i1968 to i32
  %shl.i.i.i20.i.i1973 = shl nuw i32 %conv4.i.i.i19.i.i1972, %2647
  %xor.i.i.i21.i.i1974 = xor i32 %shl.i.i.i20.i.i1973, -1
  %2648 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2679, !amdgpu.uniform !42

2649:                                             ; preds = %ComputeEnd2680
  %div8.i.i.i16.i.i1969 = lshr i32 %.lcssa4007, 5
  %idxprom.i.i.i17.i.i1970 = zext nneg i32 %div8.i.i.i16.i.i1969 to i64
  %2650 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i17.i.i1970
  %sunkaddr3348 = getelementptr inbounds i8, ptr addrspace(1) %2650, i64 40
  %2651 = atomicrmw and ptr addrspace(1) %sunkaddr3348, i32 %.lcssa3994 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %2652, !amdgpu.uniform !42

2652:                                             ; preds = %2649, %ComputeEnd2680
  call void @llvm.amdgcn.end.cf.i64(i64 %4047)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %for.cond.i.i.i1987, !amdgpu.uniform !42

for.cond.i.i.i1987:                               ; preds = %2652, %Flow3409
  %2653 = phi i32 [ %2691, %Flow3409 ], [ poison, %2652 ]
  %2654 = phi i64 [ %2692, %Flow3409 ], [ poison, %2652 ]
  %2655 = phi i32 [ %2693, %Flow3409 ], [ poison, %2652 ]
  %retval.sroa.6.0.i.i.i1988 = phi i32 [ %2696, %Flow3409 ], [ undef, %2652 ]
  %retval.sroa.2.0.i.i.i1990 = phi i64 [ %2695, %Flow3409 ], [ undef, %2652 ]
  %retval.sroa.8.0.i.i.i1991 = phi i32 [ %2694, %Flow3409 ], [ undef, %2652 ]
  %index.0.i.i.i1992 = phi i32 [ %2698, %Flow3409 ], [ 0, %2652 ]
  %2656 = load i32, ptr addrspace(1) @__llvm_rpc_client, align 8, !tbaa !68
  %cmp.not.i.i.i1993 = icmp ult i32 %index.0.i.i.i1992, %2656
  %spec.store.select.i.i.i1994 = select i1 %cmp.not.i.i.i1993, i32 %index.0.i.i.i1992, i32 0
  %2657 = tail call noundef i64 @llvm.amdgcn.ballot.i64(i1 true)
  %2658 = and i32 %spec.store.select.i.i.i1994, 31
  %2659 = lshr i64 %2657, %sh_prom.i.i.i.i
  %2660 = trunc i64 %2659 to i32
  %conv4.i.i.i.i.i1998 = and i32 %2660, 1
  %shl.i.i.i.i.i1999 = shl nuw i32 %conv4.i.i.i.i.i1998, %2658
  %2661 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2683, !amdgpu.uniform !42

2662:                                             ; preds = %ComputeEnd2684
  %sunkaddr3349 = getelementptr inbounds i8, ptr addrspace(1) %4057, i64 40
  %2663 = atomicrmw or ptr addrspace(1) %sunkaddr3349, i32 %.lcssa3988 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %2664, !amdgpu.uniform !42

2664:                                             ; preds = %2662, %ComputeEnd2684
  %2665 = phi i32 [ %2663, %2662 ], [ poison, %ComputeEnd2684 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %4066)
  %2666 = call i32 @llvm.amdgcn.readfirstlane.i32(i32 %2665)
  %2667 = or i32 %2666, %.lcssa3989
  %shl5.i.i.i.i.i2000 = shl nuw i32 1, %2658
  %and.i.i.i.i.i2001 = and i32 %shl5.i.i.i.i.i2000, %2667
  %tobool3.i.i.i.i2002 = icmp ne i32 %and.i.i.i.i.i2001, 0
  %2668 = tail call i64 @llvm.amdgcn.ballot.i64(i1 %tobool3.i.i.i.i2002)
  %and.i.i.i.i.i.i2003 = and i64 %2668, %2657
  %cmp.i.not.i.i.i2004 = icmp ne i64 %2657, %and.i.i.i.i.i.i2003
  br i1 %cmp.i.not.i.i.i2004, label %if.then.i.i.i.i2005, label %Flow3408, !amdgpu.uniform !42

if.then.i.i.i.i2005:                              ; preds = %2664
  fence syncscope("agent") acquire
  %2669 = load <2 x i64>, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !75
  %2670 = extractelement <2 x i64> %2669, i32 0
  %2671 = inttoptr i64 %2670 to ptr
  %2672 = extractelement <2 x i64> %2669, i32 1
  %2673 = inttoptr i64 %2672 to ptr
  %idxprom.i.i.i.i2006 = zext i32 %spec.store.select.i.i.i1994 to i64
  %arrayidx.i.i.i.i2007 = getelementptr inbounds nuw i32, ptr %2671, i64 %idxprom.i.i.i.i2006, !amdgpu.uniform !42
  %2674 = load atomic i32, ptr %arrayidx.i.i.i.i2007 monotonic, align 4
  %2675 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %2674)
  %arrayidx.i45.i.i.i2008 = getelementptr inbounds nuw i32, ptr %2673, i64 %idxprom.i.i.i.i2006, !amdgpu.uniform !42
  %2676 = load atomic i32, ptr %arrayidx.i45.i.i.i2008 monotonic, align 4
  %2677 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %2676)
  %cmp.i46.not.i.i.i2009 = icmp ne i32 %2675, %2677
  br i1 %cmp.i46.not.i.i.i2009, label %if.then12.i.i.i2010, label %Flow3406, !amdgpu.uniform !42

if.then12.i.i.i2010:                              ; preds = %if.then.i.i.i.i2005
  fence syncscope("agent") release
  %2678 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %2657, i1 true)
  %iszero.i.i.i.i.i.i.i2011 = icmp ne i64 %2657, 0
  %cmp2.i.i.i.i.i.i2012 = icmp eq i64 %2678, %sh_prom.i.i.i.i
  %cmp.i.i.i.i.i.i2013 = select i1 %iszero.i.i.i.i.i.i.i2011, i1 %cmp2.i.i.i.i.i.i2012, i1 false
  %conv4.i.i51.i.i.i2014 = zext i1 %cmp.i.i.i.i.i.i2013 to i32
  %shl.i.i52.i.i.i2015 = shl nuw i32 %conv4.i.i51.i.i.i2014, %2658
  %xor.i.i.i.i.i2016 = xor i32 %shl.i.i52.i.i.i2015, -1
  %2679 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2688, !amdgpu.uniform !42

2680:                                             ; preds = %ComputeEnd2689
  %sunkaddr3350 = getelementptr inbounds i8, ptr addrspace(1) %4057, i64 40
  %2681 = atomicrmw and ptr addrspace(1) %sunkaddr3350, i32 %.lcssa3990 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %2682, !amdgpu.uniform !42

2682:                                             ; preds = %2680, %ComputeEnd2689
  call void @llvm.amdgcn.end.cf.i64(i64 %4084)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %Flow3406, !amdgpu.uniform !42

if.end14.i.i.i2074:                               ; preds = %Flow3406
  %2683 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %2657, i1 true)
  %iszero.i.i.i.i.i.i2075 = icmp ne i64 %2657, 0
  %cmp2.i.i.i.i.i2076 = icmp eq i64 %2683, %sh_prom.i.i.i.i
  %cmp.i.i.i.i.i2077 = select i1 %iszero.i.i.i.i.i.i2075, i1 %cmp2.i.i.i.i.i2076, i1 false
  %2684 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.i.i.i.i.i2077)
  %2685 = extractvalue { i1, i64 } %2684, 0
  %2686 = extractvalue { i1, i64 } %2684, 1
  br i1 %2685, label %if.then16.i.i.i2079, label %if.end22.i.i.i2078

if.then16.i.i.i2079:                              ; preds = %if.end14.i.i.i2074
  %2687 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !76
  %opcode.i.i.i2080 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %2687, i64 %idxprom.i.i.i.i2006, i32 1
  store i32 1660944387, ptr %opcode.i.i.i2080, align 8, !tbaa !77
  %arrayidx21.i.i.i2081 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %2687, i64 %idxprom.i.i.i.i2006
  store i64 %2657, ptr %arrayidx21.i.i.i2081, align 8, !tbaa !80
  br label %if.end22.i.i.i2078, !amdgpu.uniform !42

Flow3407:                                         ; preds = %if.end22.i.i.i2078, %Flow3406
  %2688 = phi i32 [ %2675, %if.end22.i.i.i2078 ], [ %retval.sroa.8.0.i.i.i1991, %Flow3406 ]
  %2689 = phi i64 [ %2657, %if.end22.i.i.i2078 ], [ %retval.sroa.2.0.i.i.i1990, %Flow3406 ]
  %2690 = phi i32 [ %spec.store.select.i.i.i1994, %if.end22.i.i.i2078 ], [ %retval.sroa.6.0.i.i.i1988, %Flow3406 ]
  br label %cleanup26.i.i.i2017, !amdgpu.uniform !42

if.end22.i.i.i2078:                               ; preds = %if.then16.i.i.i2079, %if.end14.i.i.i2074
  call void @llvm.amdgcn.end.cf.i64(i64 %2686)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %Flow3407, !amdgpu.uniform !42

Flow3408:                                         ; preds = %cleanup26.i.i.i2017, %2664
  %2691 = phi i32 [ %2690, %cleanup26.i.i.i2017 ], [ %2653, %2664 ]
  %2692 = phi i64 [ %2689, %cleanup26.i.i.i2017 ], [ %2654, %2664 ]
  %2693 = phi i32 [ %2688, %cleanup26.i.i.i2017 ], [ %2655, %2664 ]
  %2694 = phi i32 [ %2688, %cleanup26.i.i.i2017 ], [ %retval.sroa.8.0.i.i.i1991, %2664 ]
  %2695 = phi i64 [ %2689, %cleanup26.i.i.i2017 ], [ %retval.sroa.2.0.i.i.i1990, %2664 ]
  %2696 = phi i32 [ %2690, %cleanup26.i.i.i2017 ], [ %retval.sroa.6.0.i.i.i1988, %2664 ]
  %2697 = phi i1 [ %cmp.i46.not.i.i.i2009, %cleanup26.i.i.i2017 ], [ true, %2664 ]
  br i1 %2697, label %for.inc.i.i.i2022, label %Flow3409, !amdgpu.uniform !42

cleanup26.i.i.i2017:                              ; preds = %Flow3407
  br label %Flow3408, !amdgpu.uniform !42

for.inc.i.i.i2022:                                ; preds = %Flow3408
  %inc.i.i.i2027 = add i32 %spec.store.select.i.i.i1994, 1
  br label %Flow3409, !amdgpu.uniform !42

Flow3409:                                         ; preds = %for.inc.i.i.i2022, %Flow3408
  %2698 = phi i32 [ %inc.i.i.i2027, %for.inc.i.i.i2022 ], [ poison, %Flow3408 ]
  %2699 = phi i1 [ false, %for.inc.i.i.i2022 ], [ true, %Flow3408 ]
  br i1 %2699, label %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i2028, label %for.cond.i.i.i1987, !amdgpu.uniform !42

_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i2028: ; preds = %Flow3409
  %.lcssa3993 = phi i32 [ %2691, %Flow3409 ]
  %.lcssa3992 = phi i64 [ %2692, %Flow3409 ]
  %.lcssa3991 = phi i32 [ %2693, %Flow3409 ]
  fence acquire
  %2700 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %mul.i.i.i.i.i.i2029 = shl i32 %.lcssa3993, 6
  %idxprom.i19.i.i.i.i.i2030 = zext i32 %mul.i.i.i.i.i.i2029 to i64
  %arrayidx.i20.i.i.i.i.i2031 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %2700, i64 %idxprom.i19.i.i.i.i.i2030
  %arrayidx.i22.i.i.i.i.i2032 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %arrayidx.i20.i.i.i.i.i2031, i64 %sh_prom.i.i.i.i
  store i64 1, ptr %arrayidx.i22.i.i.i.i.i2032, align 8, !tbaa !82
  %arrayidx6.i.i.i.i.i.i.i2033 = getelementptr inbounds nuw i8, ptr %arrayidx.i22.i.i.i.i.i2032, i64 8
  store i8 10, ptr %arrayidx6.i.i.i.i.i.i.i2033, align 8
  %tobool.not.i.i.i.i.i.i2034 = icmp eq i32 %.lcssa3991, 0
  %conv.i.i.i.i.i.i2035 = zext i1 %tobool.not.i.i.i.i.i.i2034 to i32
  fence release
  %2701 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %idxprom.i23.i.i.i.i.i2036 = zext i32 %.lcssa3993 to i64
  %arrayidx.i24.i.i.i.i.i2037 = getelementptr inbounds nuw i32, ptr %2701, i64 %idxprom.i23.i.i.i.i.i2036
  store atomic i32 %conv.i.i.i.i.i.i2035, ptr %arrayidx.i24.i.i.i.i.i2037 monotonic, align 4
  %2702 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !84
  %arrayidx.i.i5.i.i2038 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %2702, i64 %idxprom.i23.i.i.i.i.i2036, !amdgpu.uniform !42
  %2703 = load i64, ptr %arrayidx.i.i5.i.i2038, align 8, !tbaa !80
  br label %while.cond.i.i.i.i2039, !amdgpu.uniform !42

while.cond.i.i.i.i2039:                           ; preds = %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i2028, %Flow3405
  %phi.broken3983 = phi i64 [ 0, %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i2028 ], [ %2718, %Flow3405 ]
  %port.sroa.37.0.i.i2040 = phi i32 [ %2716, %Flow3405 ], [ %conv.i.i.i.i.i.i2035, %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i2028 ]
  %idx.0.i.i.i.i2041 = phi i64 [ %2715, %Flow3405 ], [ 56, %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i2028 ]
  %cmp.i.i.i.i2042 = icmp eq i64 %idx.0.i.i.i.i2041, 0
  %2704 = tail call i64 @llvm.amdgcn.ballot.i64(i1 %cmp.i.i.i.i2042)
  %and.i.i.i.i6.i.i2043 = and i64 %2704, %2703
  %tobool.not.i.i.i.i2044 = icmp ne i64 %and.i.i.i.i6.i.i2043, 0
  %2705 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i22.i.i.i2045 = getelementptr inbounds nuw i32, ptr %2705, i64 %idxprom.i23.i.i.i.i.i2036, !amdgpu.uniform !42
  %2706 = load atomic i32, ptr %arrayidx.i22.i.i.i2045 monotonic, align 4
  %2707 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %tobool.not.i.i.i.i2044)
  %2708 = extractvalue { i1, i64 } %2707, 0
  %2709 = extractvalue { i1, i64 } %2707, 1
  br i1 %2708, label %cond.false.i8.i.i.i.i2046, label %Flow3405

cond.false.i8.i.i.i.i2046:                        ; preds = %while.cond.i.i.i.i2039
  %2710 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %2706)
  %2711 = icmp ne i32 %2710, %port.sroa.37.0.i.i2040
  br i1 %2711, label %while.body.i.i21.i.i.i.i2047.preheader, label %Flow3404, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i2047.preheader:           ; preds = %cond.false.i8.i.i.i.i2046
  br label %while.body.i.i21.i.i.i.i2047, !amdgpu.uniform !42

Flow3404:                                         ; preds = %Flow3403, %cond.false.i8.i.i.i.i2046
  br label %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i2050, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i2047:                     ; preds = %while.body.i.i21.i.i.i.i2047.preheader, %while.body.i.i21.i.i.i.i2047
  tail call void @llvm.amdgcn.s.sleep(i32 2)
  %2712 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i22.i.i.i.i2048 = getelementptr inbounds nuw i32, ptr %2712, i64 %idxprom.i23.i.i.i.i.i2036, !amdgpu.uniform !42
  %2713 = load atomic i32, ptr %arrayidx.i.i.i22.i.i.i.i2048 monotonic, align 4
  %2714 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %2713)
  %cmp.i.not.i.i23.i.i.i.i2049 = icmp eq i32 %2714, %port.sroa.37.0.i.i2040
  br i1 %cmp.i.not.i.i23.i.i.i.i2049, label %Flow3403, label %while.body.i.i21.i.i.i.i2047, !amdgpu.uniform !42

Flow3403:                                         ; preds = %while.body.i.i21.i.i.i.i2047
  br label %Flow3404, !amdgpu.uniform !42

Flow3405:                                         ; preds = %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i2051, %while.cond.i.i.i.i2039
  %2715 = phi i64 [ %add.i.i.i.i2054, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i2051 ], [ poison, %while.cond.i.i.i.i2039 ]
  %2716 = phi i32 [ %conv.i.i26.i.i.i.i2052, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i2051 ], [ poison, %while.cond.i.i.i.i2039 ]
  %2717 = phi i1 [ false, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i2051 ], [ true, %while.cond.i.i.i.i2039 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %2709)
  %2718 = call i64 @llvm.amdgcn.if.break.i64(i1 %2717, i64 %phi.broken3983)
  %2719 = call i1 @llvm.amdgcn.loop.i64(i64 %2718)
  br i1 %2719, label %cond.false.i.i.i2060, label %while.cond.i.i.i.i2039

_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i2050: ; preds = %Flow3404
  fence acquire
  br i1 %cmp.i.i.i.i2042, label %if.then.i.i.i.i.i.i.i2055, label %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i2051, !amdgpu.uniform !42

if.then.i.i.i.i.i.i.i2055:                        ; preds = %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i2050
  %2720 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %arrayidx.i20.i35.i.i.i.i2058 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %2720, i64 %idxprom.i19.i.i.i.i.i2030
  %arrayidx.i22.i38.i.i.i.i2059 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %arrayidx.i20.i35.i.i.i.i2058, i64 %sh_prom.i.i.i.i
  store i8 10, ptr %arrayidx.i22.i38.i.i.i.i2059, align 1
  br label %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i2051, !amdgpu.uniform !42

_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i2051: ; preds = %if.then.i.i.i.i.i.i.i2055, %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i2050
  %conv.i.i26.i.i.i.i2052 = xor i32 %port.sroa.37.0.i.i2040, 1
  fence release
  %2721 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %arrayidx.i24.i29.i.i.i.i2053 = getelementptr inbounds nuw i32, ptr %2721, i64 %idxprom.i23.i.i.i.i.i2036
  store atomic i32 %conv.i.i26.i.i.i.i2052, ptr %arrayidx.i24.i29.i.i.i.i2053 monotonic, align 4
  %add.i.i.i.i2054 = add i64 %idx.0.i.i.i.i2041, 64
  br label %Flow3405, !amdgpu.uniform !42

cond.false.i.i.i2060:                             ; preds = %Flow3405
  %.lcssa3987 = phi i64 [ %2718, %Flow3405 ]
  %port.sroa.37.0.i.i2040.lcssa = phi i32 [ %port.sroa.37.0.i.i2040, %Flow3405 ]
  %.lcssa3986 = phi i32 [ %2706, %Flow3405 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa3987)
  %2722 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %.lcssa3986)
  %2723 = icmp ne i32 %2722, %port.sroa.37.0.i.i2040.lcssa
  %2724 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %2723)
  %2725 = extractvalue { i1, i64 } %2724, 0
  %2726 = extractvalue { i1, i64 } %2724, 1
  br i1 %2725, label %while.body.i.i.i.i2061.preheader, label %Flow3402

while.body.i.i.i.i2061.preheader:                 ; preds = %cond.false.i.i.i2060
  br label %while.body.i.i.i.i2061, !amdgpu.uniform !42

Flow3402:                                         ; preds = %Flow, %cond.false.i.i.i2060
  call void @llvm.amdgcn.end.cf.i64(i64 %2726)
  br label %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIcEERS1_T_.exit2082, !amdgpu.uniform !42

while.body.i.i.i.i2061:                           ; preds = %while.body.i.i.i.i2061.preheader, %while.body.i.i.i.i2061
  %phi.broken3984 = phi i64 [ 0, %while.body.i.i.i.i2061.preheader ], [ %2730, %while.body.i.i.i.i2061 ]
  tail call void @llvm.amdgcn.s.sleep(i32 2)
  %2727 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i8.i.i2062 = getelementptr inbounds nuw i32, ptr %2727, i64 %idxprom.i23.i.i.i.i.i2036, !amdgpu.uniform !42
  %2728 = load atomic i32, ptr %arrayidx.i.i.i8.i.i2062 monotonic, align 4
  %2729 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %2728)
  %cmp.i.not.i.i.i.i2063 = icmp eq i32 %2729, %port.sroa.37.0.i.i2040.lcssa
  %2730 = call i64 @llvm.amdgcn.if.break.i64(i1 %cmp.i.not.i.i.i.i2063, i64 %phi.broken3984)
  %2731 = call i1 @llvm.amdgcn.loop.i64(i64 %2730)
  br i1 %2731, label %Flow, label %while.body.i.i.i.i2061

Flow:                                             ; preds = %while.body.i.i.i.i2061
  %.lcssa3985 = phi i64 [ %2730, %while.body.i.i.i.i2061 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa3985)
  br label %Flow3402, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIcEERS1_T_.exit2082: ; preds = %Flow3402
  fence acquire
  tail call void @llvm.amdgcn.wave.barrier()
  fence syncscope("agent") release
  %2732 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %.lcssa3992, i1 true)
  %iszero.i.i.i.i.i13.i.i2065 = icmp ne i64 %.lcssa3992, 0
  %cmp2.i.i.i.i14.i.i2066 = icmp eq i64 %2732, %sh_prom.i.i.i.i
  %cmp.i.i.i.i15.i.i2067 = select i1 %iszero.i.i.i.i.i13.i.i2065, i1 %cmp2.i.i.i.i14.i.i2066, i1 false
  %2733 = and i32 %.lcssa3993, 31
  %conv4.i.i.i19.i.i2071 = zext i1 %cmp.i.i.i.i15.i.i2067 to i32
  %shl.i.i.i20.i.i2072 = shl nuw i32 %conv4.i.i.i19.i.i2071, %2733
  %xor.i.i.i21.i.i2073 = xor i32 %shl.i.i.i20.i.i2072, -1
  %2734 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2692, !amdgpu.uniform !42

2735:                                             ; preds = %ComputeEnd2693
  %div8.i.i.i16.i.i2068 = lshr i32 %.lcssa3993, 5
  %idxprom.i.i.i17.i.i2069 = zext nneg i32 %div8.i.i.i16.i.i2068 to i64
  %2736 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i17.i.i2069
  %sunkaddr3351 = getelementptr inbounds i8, ptr addrspace(1) %2736, i64 40
  %2737 = atomicrmw and ptr addrspace(1) %sunkaddr3351, i32 %.lcssa syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %2738, !amdgpu.uniform !42

2738:                                             ; preds = %2735, %ComputeEnd2693
  call void @llvm.amdgcn.end.cf.i64(i64 %4101)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %Flow3909, !amdgpu.uniform !42

_ZZN22__llvm_libc_22_0_0_git7testing8internal4testIiEEbPNS1_10RunContextENS0_8TestCondET_S6_PKcS8_NS1_8LocationEENKUlbNS_3cpp11string_viewEE_clEbSB_.exit: ; preds = %Flow3909
  ret i1 %cmp

Flow3414:                                         ; preds = %loop-memcpy-expansion
  %.lcssa3996 = phi i64 [ %2628, %loop-memcpy-expansion ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa3996)
  br label %Flow3415, !amdgpu.uniform !42

Flow3416:                                         ; preds = %Flow3413, %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i1950
  call void @llvm.amdgcn.end.cf.i64(i64 %2616)
  br label %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1951, !amdgpu.uniform !42

loop-memcpy-residual-header:                      ; preds = %Flow3415
  %2739 = icmp ne i64 %2618, 0
  %2740 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %2739)
  %2741 = extractvalue { i1, i64 } %2740, 0
  %2742 = extractvalue { i1, i64 } %2740, 1
  br i1 %2741, label %loop-memcpy-residual.preheader, label %Flow3413

loop-memcpy-residual.preheader:                   ; preds = %loop-memcpy-residual-header
  %umin = call i64 @llvm.umin.i64(i64 %lsr.iv2860, i64 64)
  %2743 = lshr i64 %umin, 4
  %2744 = shl nuw nsw i64 %2743, 4
  %scevgep2863 = getelementptr i8, ptr %lsr.iv, i64 %2744
  %scevgep2858 = getelementptr i8, ptr %2617, i64 %75
  %scevgep2859 = getelementptr i8, ptr %scevgep2858, i64 %2586
  %scevgep2861 = getelementptr i8, ptr %scevgep2859, i64 %2744
  br label %loop-memcpy-residual, !amdgpu.uniform !42

Flow3423:                                         ; preds = %loop-memcpy-expansion2195
  %.lcssa4001 = phi i64 [ %2584, %loop-memcpy-expansion2195 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa4001)
  br label %Flow3424, !amdgpu.uniform !42

loop-memcpy-residual-header2198:                  ; preds = %Flow3424
  %2745 = icmp eq i64 %2572, 0
  %2746 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %2745)
  %2747 = extractvalue { i1, i64 } %2746, 0
  %2748 = extractvalue { i1, i64 } %2746, 1
  br i1 %2747, label %loop-memcpy-residual-header2198.post-loop-memcpy-expansion2194_crit_edge, label %Flow3421

loop-memcpy-residual-header2198.post-loop-memcpy-expansion2194_crit_edge: ; preds = %loop-memcpy-residual-header2198
  %.pre3284 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i1929, 6
  br label %Flow3421, !amdgpu.uniform !42

Flow3421:                                         ; preds = %loop-memcpy-residual-header2198.post-loop-memcpy-expansion2194_crit_edge, %loop-memcpy-residual-header2198
  %2749 = phi i64 [ %.pre3284, %loop-memcpy-residual-header2198.post-loop-memcpy-expansion2194_crit_edge ], [ poison, %loop-memcpy-residual-header2198 ]
  %2750 = call { i1, i64 } @llvm.amdgcn.else.i64.i64(i64 %2748)
  %2751 = extractvalue { i1, i64 } %2750, 0
  %2752 = extractvalue { i1, i64 } %2750, 1
  br i1 %2751, label %loop-memcpy-residual2197.preheader, label %Flow3422

loop-memcpy-residual2197.preheader:               ; preds = %Flow3421
  %2753 = lshr i64 %spec.select.i.i.i.i.i.i.i1932, 4
  %2754 = shl nuw nsw i64 %2753, 4
  %2755 = add i64 %75, %2754
  %2756 = add i64 %2755, 8
  %scevgep2865 = getelementptr i8, ptr %2571, i64 %2756
  %2757 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i1929, 6
  %scevgep2866 = getelementptr i8, ptr %scevgep2865, i64 %2757
  %scevgep2868 = getelementptr i8, ptr %2495, i64 %2754
  br label %loop-memcpy-residual2197, !amdgpu.uniform !42

Flow3482:                                         ; preds = %loop-memcpy-expansion2201
  br label %Flow3483, !amdgpu.uniform !42

Flow3484:                                         ; preds = %Flow3480, %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i1738
  br label %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1739, !amdgpu.uniform !42

loop-memcpy-residual-header2204:                  ; preds = %Flow3483
  %2758 = icmp ne i64 %2260, 0
  br i1 %2758, label %loop-memcpy-residual2203.preheader, label %Flow3480, !amdgpu.uniform !42

loop-memcpy-residual2203.preheader:               ; preds = %loop-memcpy-residual-header2204
  %umin2909 = call i64 @llvm.umin.i64(i64 %lsr.iv2907, i64 64)
  %2759 = lshr i64 %umin2909, 4
  %2760 = shl nuw nsw i64 %2759, 4
  %scevgep2912 = getelementptr i8, ptr %lsr.iv2901, i64 %2760
  %scevgep2904 = getelementptr i8, ptr %2259, i64 %75
  %scevgep2905 = getelementptr i8, ptr %scevgep2904, i64 %2234
  %scevgep2910 = getelementptr i8, ptr %scevgep2905, i64 %2760
  br label %loop-memcpy-residual2203, !amdgpu.uniform !42

Flow3495:                                         ; preds = %loop-memcpy-expansion2207
  br label %Flow3496, !amdgpu.uniform !42

loop-memcpy-residual-header2210:                  ; preds = %Flow3496
  br i1 %2834, label %loop-memcpy-residual-header2210.post-loop-memcpy-expansion2206_crit_edge, label %Flow3492, !amdgpu.uniform !42

loop-memcpy-residual-header2210.post-loop-memcpy-expansion2206_crit_edge: ; preds = %loop-memcpy-residual-header2210
  %.pre3280 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i1717, 6
  br label %Flow3492, !amdgpu.uniform !42

Flow3492:                                         ; preds = %loop-memcpy-residual-header2210.post-loop-memcpy-expansion2206_crit_edge, %loop-memcpy-residual-header2210
  %2761 = phi i64 [ %.pre3280, %loop-memcpy-residual-header2210.post-loop-memcpy-expansion2206_crit_edge ], [ poison, %loop-memcpy-residual-header2210 ]
  %2762 = phi i1 [ false, %loop-memcpy-residual-header2210.post-loop-memcpy-expansion2206_crit_edge ], [ true, %loop-memcpy-residual-header2210 ]
  br i1 %2762, label %loop-memcpy-residual2209.preheader, label %Flow3493, !amdgpu.uniform !42

loop-memcpy-residual2209.preheader:               ; preds = %Flow3492
  %2763 = lshr i64 %spec.select.i.i.i.i.i.i.i374, 4
  %2764 = shl nuw nsw i64 %2763, 4
  %2765 = add i64 %75, %2764
  %2766 = add i64 %2765, 8
  %scevgep2914 = getelementptr i8, ptr %2227, i64 %2766
  %2767 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i1717, 6
  %scevgep2915 = getelementptr i8, ptr %scevgep2914, i64 %2767
  %scevgep2917 = getelementptr i8, ptr %2171, i64 %2764
  br label %loop-memcpy-residual2209, !amdgpu.uniform !42

Flow3525:                                         ; preds = %loop-memcpy-expansion2213
  %.lcssa4044 = phi i64 [ %2060, %loop-memcpy-expansion2213 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa4044)
  br label %Flow3526, !amdgpu.uniform !42

Flow3527:                                         ; preds = %Flow3523, %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i1533
  call void @llvm.amdgcn.end.cf.i64(i64 %2048)
  br label %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1534, !amdgpu.uniform !42

loop-memcpy-residual-header2216:                  ; preds = %Flow3526
  %2768 = icmp ne i64 %2050, 0
  %2769 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %2768)
  %2770 = extractvalue { i1, i64 } %2769, 0
  %2771 = extractvalue { i1, i64 } %2769, 1
  br i1 %2770, label %loop-memcpy-residual2215.preheader, label %Flow3523

loop-memcpy-residual2215.preheader:               ; preds = %loop-memcpy-residual-header2216
  %umin2940 = call i64 @llvm.umin.i64(i64 %lsr.iv2935, i64 64)
  %2772 = lshr i64 %umin2940, 4
  %2773 = shl nuw nsw i64 %2772, 4
  %scevgep2941 = getelementptr i8, ptr %lsr.iv2928, i64 %2773
  %scevgep2931 = getelementptr i8, ptr %2049, i64 %75
  %scevgep2932 = getelementptr i8, ptr %scevgep2931, i64 %2018
  %scevgep2938 = getelementptr i8, ptr %scevgep2932, i64 %2773
  br label %loop-memcpy-residual2215, !amdgpu.uniform !42

Flow3538:                                         ; preds = %loop-memcpy-expansion2219
  %.lcssa4049 = phi i64 [ %2016, %loop-memcpy-expansion2219 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa4049)
  br label %Flow3539, !amdgpu.uniform !42

loop-memcpy-residual-header2222:                  ; preds = %Flow3539
  %2774 = icmp eq i64 %2004, 0
  %2775 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %2774)
  %2776 = extractvalue { i1, i64 } %2775, 0
  %2777 = extractvalue { i1, i64 } %2775, 1
  br i1 %2776, label %loop-memcpy-residual-header2222.post-loop-memcpy-expansion2218_crit_edge, label %Flow3535

loop-memcpy-residual-header2222.post-loop-memcpy-expansion2218_crit_edge: ; preds = %loop-memcpy-residual-header2222
  %.pre3278 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i1512, 6
  br label %Flow3535, !amdgpu.uniform !42

Flow3535:                                         ; preds = %loop-memcpy-residual-header2222.post-loop-memcpy-expansion2218_crit_edge, %loop-memcpy-residual-header2222
  %2778 = phi i64 [ %.pre3278, %loop-memcpy-residual-header2222.post-loop-memcpy-expansion2218_crit_edge ], [ poison, %loop-memcpy-residual-header2222 ]
  %2779 = call { i1, i64 } @llvm.amdgcn.else.i64.i64(i64 %2777)
  %2780 = extractvalue { i1, i64 } %2779, 0
  %2781 = extractvalue { i1, i64 } %2779, 1
  br i1 %2780, label %loop-memcpy-residual2221.preheader, label %Flow3536

loop-memcpy-residual2221.preheader:               ; preds = %Flow3535
  %2782 = add nuw nsw i64 %75, 8
  %scevgep2943 = getelementptr i8, ptr %2003, i64 %2782
  %umin2946 = call i64 @llvm.umin.i64(i64 %lsr.iv.next2945.lcssa, i64 56)
  %2783 = lshr i64 %umin2946, 4
  %2784 = shl nuw nsw i64 %2783, 4
  %scevgep2947 = getelementptr i8, ptr %scevgep2943, i64 %2784
  %2785 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i1512, 6
  %scevgep2948 = getelementptr i8, ptr %scevgep2947, i64 %2785
  %scevgep2950 = getelementptr i8, ptr %RHSStr, i64 %2784
  br label %loop-memcpy-residual2221, !amdgpu.uniform !42

Flow3640:                                         ; preds = %loop-memcpy-expansion2225
  %.lcssa4099 = phi i64 [ %1508, %loop-memcpy-expansion2225 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa4099)
  br label %Flow3641, !amdgpu.uniform !42

Flow3642:                                         ; preds = %Flow3638, %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i1017
  call void @llvm.amdgcn.end.cf.i64(i64 %1496)
  br label %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1018, !amdgpu.uniform !42

loop-memcpy-residual-header2228:                  ; preds = %Flow3641
  %2786 = icmp ne i64 %1498, 0
  %2787 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %2786)
  %2788 = extractvalue { i1, i64 } %2787, 0
  %2789 = extractvalue { i1, i64 } %2787, 1
  br i1 %2788, label %loop-memcpy-residual2227.preheader, label %Flow3638

loop-memcpy-residual2227.preheader:               ; preds = %loop-memcpy-residual-header2228
  %umin3024 = call i64 @llvm.umin.i64(i64 %lsr.iv3022, i64 64)
  %2790 = lshr i64 %umin3024, 4
  %2791 = shl nuw nsw i64 %2790, 4
  %scevgep3027 = getelementptr i8, ptr %lsr.iv3017, i64 %2791
  %scevgep3020 = getelementptr i8, ptr %1497, i64 %75
  %scevgep3021 = getelementptr i8, ptr %scevgep3020, i64 %1466
  %scevgep3025 = getelementptr i8, ptr %scevgep3021, i64 %2791
  br label %loop-memcpy-residual2227, !amdgpu.uniform !42

Flow3653:                                         ; preds = %loop-memcpy-expansion2231
  %.lcssa4104 = phi i64 [ %1464, %loop-memcpy-expansion2231 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa4104)
  br label %Flow3654, !amdgpu.uniform !42

loop-memcpy-residual-header2234:                  ; preds = %Flow3654
  %2792 = icmp eq i64 %1452, 0
  %2793 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %2792)
  %2794 = extractvalue { i1, i64 } %2793, 0
  %2795 = extractvalue { i1, i64 } %2793, 1
  br i1 %2794, label %loop-memcpy-residual-header2234.post-loop-memcpy-expansion2230_crit_edge, label %Flow3650

loop-memcpy-residual-header2234.post-loop-memcpy-expansion2230_crit_edge: ; preds = %loop-memcpy-residual-header2234
  %.pre3270 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i996, 6
  br label %Flow3650, !amdgpu.uniform !42

Flow3650:                                         ; preds = %loop-memcpy-residual-header2234.post-loop-memcpy-expansion2230_crit_edge, %loop-memcpy-residual-header2234
  %2796 = phi i64 [ %.pre3270, %loop-memcpy-residual-header2234.post-loop-memcpy-expansion2230_crit_edge ], [ poison, %loop-memcpy-residual-header2234 ]
  %2797 = call { i1, i64 } @llvm.amdgcn.else.i64.i64(i64 %2795)
  %2798 = extractvalue { i1, i64 } %2797, 0
  %2799 = extractvalue { i1, i64 } %2797, 1
  br i1 %2798, label %loop-memcpy-residual2233.preheader, label %Flow3651

loop-memcpy-residual2233.preheader:               ; preds = %Flow3650
  %2800 = lshr i64 %spec.select.i.i.i.i.i.i.i999, 4
  %2801 = shl nuw nsw i64 %2800, 4
  %2802 = add i64 %75, %2801
  %2803 = add i64 %2802, 8
  %scevgep3029 = getelementptr i8, ptr %1451, i64 %2803
  %2804 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i996, 6
  %scevgep3030 = getelementptr i8, ptr %scevgep3029, i64 %2804
  %scevgep3032 = getelementptr i8, ptr %1375, i64 %2801
  br label %loop-memcpy-residual2233, !amdgpu.uniform !42

Flow3716:                                         ; preds = %loop-memcpy-expansion2237
  br label %Flow3717, !amdgpu.uniform !42

Flow3718:                                         ; preds = %Flow3715, %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i805
  br label %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i806, !amdgpu.uniform !42

loop-memcpy-residual-header2240:                  ; preds = %Flow3717
  %2805 = icmp ne i64 %1137, 0
  br i1 %2805, label %loop-memcpy-residual2239.preheader, label %Flow3715, !amdgpu.uniform !42

loop-memcpy-residual2239.preheader:               ; preds = %loop-memcpy-residual-header2240
  %umin3073 = call i64 @llvm.umin.i64(i64 %lsr.iv3071, i64 64)
  %2806 = lshr i64 %umin3073, 4
  %2807 = shl nuw nsw i64 %2806, 4
  %scevgep3076 = getelementptr i8, ptr %lsr.iv3065, i64 %2807
  %scevgep3068 = getelementptr i8, ptr %1136, i64 %75
  %scevgep3069 = getelementptr i8, ptr %scevgep3068, i64 %1111
  %scevgep3074 = getelementptr i8, ptr %scevgep3069, i64 %2807
  br label %loop-memcpy-residual2239, !amdgpu.uniform !42

Flow3725:                                         ; preds = %loop-memcpy-expansion2243
  br label %Flow3726, !amdgpu.uniform !42

loop-memcpy-residual-header2246:                  ; preds = %Flow3726
  br i1 %2834, label %loop-memcpy-residual-header2246.post-loop-memcpy-expansion2242_crit_edge, label %Flow3723, !amdgpu.uniform !42

loop-memcpy-residual-header2246.post-loop-memcpy-expansion2242_crit_edge: ; preds = %loop-memcpy-residual-header2246
  %.pre3266 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i784, 6
  br label %Flow3723, !amdgpu.uniform !42

Flow3723:                                         ; preds = %loop-memcpy-residual-header2246.post-loop-memcpy-expansion2242_crit_edge, %loop-memcpy-residual-header2246
  %2808 = phi i64 [ %.pre3266, %loop-memcpy-residual-header2246.post-loop-memcpy-expansion2242_crit_edge ], [ poison, %loop-memcpy-residual-header2246 ]
  %2809 = phi i1 [ false, %loop-memcpy-residual-header2246.post-loop-memcpy-expansion2242_crit_edge ], [ true, %loop-memcpy-residual-header2246 ]
  br i1 %2809, label %loop-memcpy-residual2245.preheader, label %Flow3724, !amdgpu.uniform !42

loop-memcpy-residual2245.preheader:               ; preds = %Flow3723
  %2810 = add i64 %75, %655
  %2811 = add i64 %2810, 8
  %scevgep3078 = getelementptr i8, ptr %1104, i64 %2811
  %2812 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i784, 6
  %scevgep3079 = getelementptr i8, ptr %scevgep3078, i64 %2812
  %scevgep3081 = getelementptr i8, ptr %1048, i64 %655
  br label %loop-memcpy-residual2245, !amdgpu.uniform !42

Flow3748:                                         ; preds = %loop-memcpy-expansion2249
  %.lcssa4147 = phi i64 [ %937, %loop-memcpy-expansion2249 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa4147)
  br label %Flow3749, !amdgpu.uniform !42

Flow3750:                                         ; preds = %Flow3747, %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i603
  call void @llvm.amdgcn.end.cf.i64(i64 %925)
  br label %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i604, !amdgpu.uniform !42

loop-memcpy-residual-header2252:                  ; preds = %Flow3749
  %2813 = icmp ne i64 %927, 0
  %2814 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %2813)
  %2815 = extractvalue { i1, i64 } %2814, 0
  %2816 = extractvalue { i1, i64 } %2814, 1
  br i1 %2815, label %loop-memcpy-residual2251.preheader, label %Flow3747

loop-memcpy-residual2251.preheader:               ; preds = %loop-memcpy-residual-header2252
  %umin3103 = call i64 @llvm.umin.i64(i64 %lsr.iv3098, i64 64)
  %2817 = lshr i64 %umin3103, 4
  %2818 = shl nuw nsw i64 %2817, 4
  %scevgep3104 = getelementptr i8, ptr %lsr.iv3091, i64 %2818
  %scevgep3094 = getelementptr i8, ptr %926, i64 %75
  %scevgep3095 = getelementptr i8, ptr %scevgep3094, i64 %895
  %scevgep3101 = getelementptr i8, ptr %scevgep3095, i64 %2818
  br label %loop-memcpy-residual2251, !amdgpu.uniform !42

Flow3757:                                         ; preds = %loop-memcpy-expansion2255
  %.lcssa4152 = phi i64 [ %893, %loop-memcpy-expansion2255 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa4152)
  br label %Flow3758, !amdgpu.uniform !42

loop-memcpy-residual-header2258:                  ; preds = %Flow3758
  %2819 = icmp eq i64 %881, 0
  %2820 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %2819)
  %2821 = extractvalue { i1, i64 } %2820, 0
  %2822 = extractvalue { i1, i64 } %2820, 1
  br i1 %2821, label %loop-memcpy-residual-header2258.post-loop-memcpy-expansion2254_crit_edge, label %Flow3755

loop-memcpy-residual-header2258.post-loop-memcpy-expansion2254_crit_edge: ; preds = %loop-memcpy-residual-header2258
  %.pre3264 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i582, 6
  br label %Flow3755, !amdgpu.uniform !42

Flow3755:                                         ; preds = %loop-memcpy-residual-header2258.post-loop-memcpy-expansion2254_crit_edge, %loop-memcpy-residual-header2258
  %2823 = phi i64 [ %.pre3264, %loop-memcpy-residual-header2258.post-loop-memcpy-expansion2254_crit_edge ], [ poison, %loop-memcpy-residual-header2258 ]
  %2824 = call { i1, i64 } @llvm.amdgcn.else.i64.i64(i64 %2822)
  %2825 = extractvalue { i1, i64 } %2824, 0
  %2826 = extractvalue { i1, i64 } %2824, 1
  br i1 %2825, label %loop-memcpy-residual2257.preheader, label %Flow3756

loop-memcpy-residual2257.preheader:               ; preds = %Flow3755
  %2827 = add nuw nsw i64 %75, 8
  %scevgep3106 = getelementptr i8, ptr %880, i64 %2827
  %umin3109 = call i64 @llvm.umin.i64(i64 %lsr.iv.next3108.lcssa, i64 56)
  %2828 = lshr i64 %umin3109, 4
  %2829 = shl nuw nsw i64 %2828, 4
  %scevgep3110 = getelementptr i8, ptr %scevgep3106, i64 %2829
  %2830 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i582, 6
  %scevgep3111 = getelementptr i8, ptr %scevgep3110, i64 %2830
  %scevgep3113 = getelementptr i8, ptr %LHSStr, i64 %2829
  br label %loop-memcpy-residual2257, !amdgpu.uniform !42

Flow3786:                                         ; preds = %loop-memcpy-expansion2261
  br label %Flow3787, !amdgpu.uniform !42

Flow3788:                                         ; preds = %Flow3785, %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i392
  br label %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i393, !amdgpu.uniform !42

loop-memcpy-residual-header2264:                  ; preds = %Flow3787
  %2831 = icmp ne i64 %690, 0
  br i1 %2831, label %loop-memcpy-residual2263.preheader, label %Flow3785, !amdgpu.uniform !42

loop-memcpy-residual2263.preheader:               ; preds = %loop-memcpy-residual-header2264
  %umin3149 = call i64 @llvm.umin.i64(i64 %lsr.iv3147, i64 64)
  %2832 = lshr i64 %umin3149, 4
  %2833 = shl nuw nsw i64 %2832, 4
  %scevgep3152 = getelementptr i8, ptr %lsr.iv3141, i64 %2833
  %scevgep3144 = getelementptr i8, ptr %689, i64 %75
  %scevgep3145 = getelementptr i8, ptr %scevgep3144, i64 %663
  %scevgep3150 = getelementptr i8, ptr %scevgep3145, i64 %2833
  br label %loop-memcpy-residual2263, !amdgpu.uniform !42

Flow3795:                                         ; preds = %loop-memcpy-expansion2267
  br label %Flow3796, !amdgpu.uniform !42

loop-memcpy-residual-header2270:                  ; preds = %Flow3796
  %2834 = icmp eq i64 %654, 0
  br i1 %2834, label %loop-memcpy-residual-header2270.post-loop-memcpy-expansion2266_crit_edge, label %Flow3793, !amdgpu.uniform !42

loop-memcpy-residual-header2270.post-loop-memcpy-expansion2266_crit_edge: ; preds = %loop-memcpy-residual-header2270
  %.pre3260 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i371, 6
  br label %Flow3793, !amdgpu.uniform !42

Flow3793:                                         ; preds = %loop-memcpy-residual-header2270.post-loop-memcpy-expansion2266_crit_edge, %loop-memcpy-residual-header2270
  %2835 = phi i64 [ %.pre3260, %loop-memcpy-residual-header2270.post-loop-memcpy-expansion2266_crit_edge ], [ poison, %loop-memcpy-residual-header2270 ]
  %2836 = phi i1 [ false, %loop-memcpy-residual-header2270.post-loop-memcpy-expansion2266_crit_edge ], [ true, %loop-memcpy-residual-header2270 ]
  br i1 %2836, label %loop-memcpy-residual2269.preheader, label %Flow3794, !amdgpu.uniform !42

loop-memcpy-residual2269.preheader:               ; preds = %Flow3793
  %2837 = add i64 %75, %655
  %2838 = add i64 %2837, 8
  %scevgep3154 = getelementptr i8, ptr %653, i64 %2838
  %2839 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i371, 6
  %scevgep3155 = getelementptr i8, ptr %scevgep3154, i64 %2839
  %scevgep3157 = getelementptr i8, ptr %594, i64 %655
  br label %loop-memcpy-residual2269, !amdgpu.uniform !42

Flow3828:                                         ; preds = %loop-memcpy-expansion2273
  %.lcssa4192 = phi i64 [ %458, %loop-memcpy-expansion2273 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa4192)
  br label %Flow3829, !amdgpu.uniform !42

Flow3830:                                         ; preds = %Flow3827, %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i181
  call void @llvm.amdgcn.end.cf.i64(i64 %446)
  br label %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i182, !amdgpu.uniform !42

loop-memcpy-residual-header2276:                  ; preds = %Flow3829
  %2840 = icmp ne i64 %448, 0
  %2841 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %2840)
  %2842 = extractvalue { i1, i64 } %2841, 0
  %2843 = extractvalue { i1, i64 } %2841, 1
  br i1 %2842, label %loop-memcpy-residual2275.preheader, label %Flow3827

loop-memcpy-residual2275.preheader:               ; preds = %loop-memcpy-residual-header2276
  %umin3193 = call i64 @llvm.umin.i64(i64 %lsr.iv3191, i64 64)
  %2844 = lshr i64 %umin3193, 4
  %2845 = shl nuw nsw i64 %2844, 4
  %scevgep3196 = getelementptr i8, ptr %lsr.iv3186, i64 %2845
  %scevgep3189 = getelementptr i8, ptr %447, i64 %75
  %scevgep3190 = getelementptr i8, ptr %scevgep3189, i64 %416
  %scevgep3194 = getelementptr i8, ptr %scevgep3190, i64 %2845
  br label %loop-memcpy-residual2275, !amdgpu.uniform !42

Flow3837:                                         ; preds = %loop-memcpy-expansion2279
  %.lcssa4197 = phi i64 [ %414, %loop-memcpy-expansion2279 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa4197)
  br label %Flow3838, !amdgpu.uniform !42

loop-memcpy-residual-header2282:                  ; preds = %Flow3838
  %2846 = icmp eq i64 %402, 0
  %2847 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %2846)
  %2848 = extractvalue { i1, i64 } %2847, 0
  %2849 = extractvalue { i1, i64 } %2847, 1
  br i1 %2848, label %loop-memcpy-residual-header2282.post-loop-memcpy-expansion2278_crit_edge, label %Flow3835

loop-memcpy-residual-header2282.post-loop-memcpy-expansion2278_crit_edge: ; preds = %loop-memcpy-residual-header2282
  %.pre3256 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i160, 6
  br label %Flow3835, !amdgpu.uniform !42

Flow3835:                                         ; preds = %loop-memcpy-residual-header2282.post-loop-memcpy-expansion2278_crit_edge, %loop-memcpy-residual-header2282
  %2850 = phi i64 [ %.pre3256, %loop-memcpy-residual-header2282.post-loop-memcpy-expansion2278_crit_edge ], [ poison, %loop-memcpy-residual-header2282 ]
  %2851 = call { i1, i64 } @llvm.amdgcn.else.i64.i64(i64 %2849)
  %2852 = extractvalue { i1, i64 } %2851, 0
  %2853 = extractvalue { i1, i64 } %2851, 1
  br i1 %2852, label %loop-memcpy-residual2281.preheader, label %Flow3836

loop-memcpy-residual2281.preheader:               ; preds = %Flow3835
  %2854 = lshr i64 %spec.select.i.i.i.i.i.i.i163, 4
  %2855 = shl nuw nsw i64 %2854, 4
  %2856 = add i64 %75, %2855
  %2857 = add i64 %2856, 8
  %scevgep3198 = getelementptr i8, ptr %401, i64 %2857
  %2858 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i160, 6
  %scevgep3199 = getelementptr i8, ptr %scevgep3198, i64 %2858
  %scevgep3201 = getelementptr i8, ptr %325, i64 %2855
  br label %loop-memcpy-residual2281, !amdgpu.uniform !42

Flow3505:                                         ; preds = %loop-memcpy-expansion2285
  br label %Flow3506, !amdgpu.uniform !42

loop-memcpy-residual-header2288:                  ; preds = %Flow3506
  br i1 %2861, label %loop-memcpy-residual2287.preheader, label %Flow3503, !amdgpu.uniform !42

loop-memcpy-residual2287.preheader:               ; preds = %loop-memcpy-residual-header2288
  %2859 = sub i64 sub (i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) getelementptr inbounds nuw (i8, ptr addrspace(4) @.str.9, i64 8) to ptr) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.9 to ptr) to i64)), %umin3146
  %2860 = sub i64 %2859, %595
  %scevgep2923 = getelementptr i8, ptr %2171, i64 %2860
  %scevgep2925 = getelementptr i8, ptr %3, i64 %2860
  br label %loop-memcpy-residual2287, !amdgpu.uniform !42

Flow3733:                                         ; preds = %loop-memcpy-expansion2291
  br label %Flow3734, !amdgpu.uniform !42

loop-memcpy-residual-header2294:                  ; preds = %Flow3734
  br i1 %2861, label %loop-memcpy-residual2293.preheader, label %Flow3732, !amdgpu.uniform !42

loop-memcpy-residual2293.preheader:               ; preds = %loop-memcpy-residual-header2294
  %scevgep3086 = getelementptr i8, ptr %1048, i64 %596
  %scevgep3088 = getelementptr i8, ptr %3, i64 %596
  br label %loop-memcpy-residual2293, !amdgpu.uniform !42

Flow3803:                                         ; preds = %loop-memcpy-expansion2297
  br label %Flow3804, !amdgpu.uniform !42

loop-memcpy-residual-header2300:                  ; preds = %Flow3804
  %2861 = icmp ne i64 %595, 0
  br i1 %2861, label %loop-memcpy-residual2299.preheader, label %Flow3802, !amdgpu.uniform !42

loop-memcpy-residual2299.preheader:               ; preds = %loop-memcpy-residual-header2300
  %scevgep3162 = getelementptr i8, ptr %594, i64 %596
  %scevgep3164 = getelementptr i8, ptr %3, i64 %596
  br label %loop-memcpy-residual2299, !amdgpu.uniform !42

Flow3845:                                         ; preds = %loop-memcpy-expansion2303
  %.lcssa4205 = phi i64 [ %347, %loop-memcpy-expansion2303 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa4205)
  br label %Flow3846, !amdgpu.uniform !42

loop-memcpy-residual-header2306:                  ; preds = %Flow3846
  %2862 = icmp ne i64 %336, 0
  %2863 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %2862)
  %2864 = extractvalue { i1, i64 } %2863, 0
  %2865 = extractvalue { i1, i64 } %2863, 1
  br i1 %2864, label %loop-memcpy-residual2305.preheader, label %Flow3844

loop-memcpy-residual2305.preheader:               ; preds = %loop-memcpy-residual-header2306
  %2866 = lshr i64 %writer.sroa.10.4.i.i.i.i, 4
  %2867 = trunc i64 %2866 to i32
  %2868 = shl i32 %2867, 4
  %2869 = add i32 %2868, 11
  %2870 = sub i32 %2869, %335
  %scevgep3206 = getelementptr i8, ptr addrspace(5) %buffer.i.i.i, i32 %2870
  %2871 = shl nuw i64 %2866, 4
  %scevgep3209 = getelementptr i8, ptr %325, i64 %2871
  br label %loop-memcpy-residual2305, !amdgpu.uniform !42

Flow3431:                                         ; preds = %loop-memcpy-expansion2309
  %.lcssa4009 = phi i64 [ %2517, %loop-memcpy-expansion2309 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa4009)
  br label %Flow3432, !amdgpu.uniform !42

loop-memcpy-residual-header2312:                  ; preds = %Flow3432
  %2872 = icmp ne i64 %2506, 0
  %2873 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %2872)
  %2874 = extractvalue { i1, i64 } %2873, 0
  %2875 = extractvalue { i1, i64 } %2873, 1
  br i1 %2874, label %loop-memcpy-residual2311.preheader, label %Flow3430

loop-memcpy-residual2311.preheader:               ; preds = %loop-memcpy-residual-header2312
  %2876 = lshr i64 %writer.sroa.10.5.i.i.i.i129.i, 4
  %2877 = trunc i64 %2876 to i32
  %2878 = shl i32 %2877, 4
  %2879 = add i32 %2878, 11
  %2880 = sub i32 %2879, %2505
  %scevgep2873 = getelementptr i8, ptr addrspace(5) %buffer.i.i.i, i32 %2880
  %2881 = shl nuw i64 %2876, 4
  %scevgep2876 = getelementptr i8, ptr %2495, i64 %2881
  br label %loop-memcpy-residual2311, !amdgpu.uniform !42

Flow3663:                                         ; preds = %loop-memcpy-expansion2315
  %.lcssa4112 = phi i64 [ %1397, %loop-memcpy-expansion2315 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa4112)
  br label %Flow3664, !amdgpu.uniform !42

loop-memcpy-residual-header2318:                  ; preds = %Flow3664
  %2882 = icmp ne i64 %1386, 0
  %2883 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %2882)
  %2884 = extractvalue { i1, i64 } %2883, 0
  %2885 = extractvalue { i1, i64 } %2883, 1
  br i1 %2884, label %loop-memcpy-residual2317.preheader, label %Flow3661

loop-memcpy-residual2317.preheader:               ; preds = %loop-memcpy-residual-header2318
  %2886 = lshr i64 %writer.sroa.10.5.i.i.i.i.i, 4
  %2887 = trunc i64 %2886 to i32
  %2888 = shl i32 %2887, 4
  %2889 = add i32 %2888, 11
  %2890 = sub i32 %2889, %1385
  %scevgep3037 = getelementptr i8, ptr addrspace(5) %buffer.i.i.i, i32 %2890
  %2891 = shl nuw i64 %2886, 4
  %scevgep3040 = getelementptr i8, ptr %1375, i64 %2891
  br label %loop-memcpy-residual2317, !amdgpu.uniform !42

Flow3898:                                         ; preds = %loop-memcpy-expansion2321
  br label %Flow3899, !amdgpu.uniform !42

loop-memcpy-residual-header2324:                  ; preds = %Flow3899
  %2892 = icmp eq i64 %65, 0
  br i1 %2892, label %loop-memcpy-residual-header2324.post-loop-memcpy-expansion2320_crit_edge, label %Flow3896, !amdgpu.uniform !42

loop-memcpy-residual-header2324.post-loop-memcpy-expansion2320_crit_edge: ; preds = %loop-memcpy-residual-header2324
  %.pre = shl nuw nsw i64 %sh_prom.i.i.i.i, 6
  %.pre3252 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i, 6
  br label %Flow3896, !amdgpu.uniform !42

Flow3896:                                         ; preds = %loop-memcpy-residual-header2324.post-loop-memcpy-expansion2320_crit_edge, %loop-memcpy-residual-header2324
  %2893 = phi i64 [ %.pre, %loop-memcpy-residual-header2324.post-loop-memcpy-expansion2320_crit_edge ], [ poison, %loop-memcpy-residual-header2324 ]
  %2894 = phi i64 [ %.pre3252, %loop-memcpy-residual-header2324.post-loop-memcpy-expansion2320_crit_edge ], [ poison, %loop-memcpy-residual-header2324 ]
  %2895 = phi i1 [ false, %loop-memcpy-residual-header2324.post-loop-memcpy-expansion2320_crit_edge ], [ true, %loop-memcpy-residual-header2324 ]
  br i1 %2895, label %loop-memcpy-residual2323.preheader, label %Flow3897, !amdgpu.uniform !42

loop-memcpy-residual2323.preheader:               ; preds = %Flow3896
  %2896 = shl nuw nsw i64 %sh_prom.i.i.i.i, 6
  %2897 = add nuw nsw i64 %2896, 56
  %scevgep3245 = getelementptr i8, ptr %64, i64 %2897
  %2898 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i, 6
  %scevgep3246 = getelementptr i8, ptr %scevgep3245, i64 %2898
  br label %loop-memcpy-residual2323, !amdgpu.uniform !42

Flow3889:                                         ; preds = %loop-memcpy-expansion2327
  br label %Flow3890, !amdgpu.uniform !42

Flow3891:                                         ; preds = %Flow3888, %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i
  br label %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i, !amdgpu.uniform !42

loop-memcpy-residual-header2330:                  ; preds = %Flow3890
  %2899 = icmp ne i64 %102, 0
  br i1 %2899, label %loop-memcpy-residual2329.preheader, label %Flow3888, !amdgpu.uniform !42

loop-memcpy-residual2329.preheader:               ; preds = %loop-memcpy-residual-header2330
  %umin3240 = call i64 @llvm.umin.i64(i64 %lsr.iv3238, i64 64)
  %2900 = lshr i64 %umin3240, 4
  %2901 = shl nuw nsw i64 %2900, 4
  %scevgep3243 = getelementptr i8, ptr addrspace(4) %lsr.iv3233, i64 %2901
  %scevgep3236 = getelementptr i8, ptr %101, i64 %75
  %scevgep3237 = getelementptr i8, ptr %scevgep3236, i64 %76
  %scevgep3241 = getelementptr i8, ptr %scevgep3237, i64 %2901
  br label %loop-memcpy-residual2329, !amdgpu.uniform !42

Flow3879:                                         ; preds = %loop-memcpy-expansion2333
  br label %Flow3880, !amdgpu.uniform !42

loop-memcpy-residual-header2336:                  ; preds = %Flow3880
  %2902 = icmp eq i64 %178, 0
  br i1 %2902, label %loop-memcpy-residual-header2336.post-loop-memcpy-expansion2332_crit_edge, label %Flow3877, !amdgpu.uniform !42

loop-memcpy-residual-header2336.post-loop-memcpy-expansion2332_crit_edge: ; preds = %loop-memcpy-residual-header2336
  %.pre3254 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i56, 6
  br label %Flow3877, !amdgpu.uniform !42

Flow3877:                                         ; preds = %loop-memcpy-residual-header2336.post-loop-memcpy-expansion2332_crit_edge, %loop-memcpy-residual-header2336
  %2903 = phi i64 [ %.pre3254, %loop-memcpy-residual-header2336.post-loop-memcpy-expansion2332_crit_edge ], [ poison, %loop-memcpy-residual-header2336 ]
  %2904 = phi i1 [ false, %loop-memcpy-residual-header2336.post-loop-memcpy-expansion2332_crit_edge ], [ true, %loop-memcpy-residual-header2336 ]
  br i1 %2904, label %loop-memcpy-residual2335.preheader, label %Flow3878, !amdgpu.uniform !42

loop-memcpy-residual2335.preheader:               ; preds = %Flow3877
  %2905 = add nuw nsw i64 %75, 8
  %scevgep3226 = getelementptr i8, ptr %177, i64 %2905
  %2906 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i56, 6
  %scevgep3227 = getelementptr i8, ptr %scevgep3226, i64 %2906
  br label %loop-memcpy-residual2335, !amdgpu.uniform !42

Flow3870:                                         ; preds = %loop-memcpy-expansion2339
  br label %Flow3871, !amdgpu.uniform !42

Flow3872:                                         ; preds = %Flow3869, %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i77
  br label %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i78, !amdgpu.uniform !42

loop-memcpy-residual-header2342:                  ; preds = %Flow3871
  %2907 = icmp ne i64 %213, 0
  br i1 %2907, label %loop-memcpy-residual2341.preheader, label %Flow3869, !amdgpu.uniform !42

loop-memcpy-residual2341.preheader:               ; preds = %loop-memcpy-residual-header2342
  %umin3221 = call i64 @llvm.umin.i64(i64 %lsr.iv3219, i64 64)
  %2908 = lshr i64 %umin3221, 4
  %2909 = shl nuw nsw i64 %2908, 4
  %scevgep3224 = getelementptr i8, ptr addrspace(4) %lsr.iv3214, i64 %2909
  %scevgep3217 = getelementptr i8, ptr %212, i64 %75
  %scevgep3218 = getelementptr i8, ptr %scevgep3217, i64 %187
  %scevgep3222 = getelementptr i8, ptr %scevgep3218, i64 %2909
  br label %loop-memcpy-residual2341, !amdgpu.uniform !42

Flow3818:                                         ; preds = %loop-memcpy-expansion2345
  br label %Flow3819, !amdgpu.uniform !42

loop-memcpy-residual-header2348:                  ; preds = %Flow3819
  %2910 = icmp eq i64 %531, 0
  br i1 %2910, label %loop-memcpy-residual-header2348.post-loop-memcpy-expansion2344_crit_edge, label %Flow3816, !amdgpu.uniform !42

loop-memcpy-residual-header2348.post-loop-memcpy-expansion2344_crit_edge: ; preds = %loop-memcpy-residual-header2348
  %.pre3258 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i265, 6
  br label %Flow3816, !amdgpu.uniform !42

Flow3816:                                         ; preds = %loop-memcpy-residual-header2348.post-loop-memcpy-expansion2344_crit_edge, %loop-memcpy-residual-header2348
  %2911 = phi i64 [ %.pre3258, %loop-memcpy-residual-header2348.post-loop-memcpy-expansion2344_crit_edge ], [ poison, %loop-memcpy-residual-header2348 ]
  %2912 = phi i1 [ false, %loop-memcpy-residual-header2348.post-loop-memcpy-expansion2344_crit_edge ], [ true, %loop-memcpy-residual-header2348 ]
  br i1 %2912, label %loop-memcpy-residual2347.preheader, label %Flow3817, !amdgpu.uniform !42

loop-memcpy-residual2347.preheader:               ; preds = %Flow3816
  %2913 = add nuw nsw i64 %75, 8
  %scevgep3178 = getelementptr i8, ptr %530, i64 %2913
  %2914 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i265, 6
  %scevgep3179 = getelementptr i8, ptr %scevgep3178, i64 %2914
  br label %loop-memcpy-residual2347, !amdgpu.uniform !42

Flow3809:                                         ; preds = %loop-memcpy-expansion2351
  br label %Flow3810, !amdgpu.uniform !42

Flow3811:                                         ; preds = %Flow3808, %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i286
  br label %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i287, !amdgpu.uniform !42

loop-memcpy-residual-header2354:                  ; preds = %Flow3810
  %2915 = icmp ne i64 %566, 0
  br i1 %2915, label %loop-memcpy-residual2353.preheader, label %Flow3808, !amdgpu.uniform !42

loop-memcpy-residual2353.preheader:               ; preds = %loop-memcpy-residual-header2354
  %umin3173 = call i64 @llvm.umin.i64(i64 %lsr.iv3171, i64 64)
  %2916 = lshr i64 %umin3173, 4
  %2917 = shl nuw nsw i64 %2916, 4
  %scevgep3176 = getelementptr i8, ptr addrspace(4) %lsr.iv3166, i64 %2917
  %scevgep3169 = getelementptr i8, ptr %565, i64 %75
  %scevgep3170 = getelementptr i8, ptr %scevgep3169, i64 %540
  %scevgep3174 = getelementptr i8, ptr %scevgep3170, i64 %2917
  br label %loop-memcpy-residual2353, !amdgpu.uniform !42

Flow3776:                                         ; preds = %loop-memcpy-expansion2357
  br label %Flow3777, !amdgpu.uniform !42

loop-memcpy-residual-header2360:                  ; preds = %Flow3777
  %2918 = icmp eq i64 %766, 0
  br i1 %2918, label %loop-memcpy-residual-header2360.post-loop-memcpy-expansion2356_crit_edge, label %Flow3774, !amdgpu.uniform !42

loop-memcpy-residual-header2360.post-loop-memcpy-expansion2356_crit_edge: ; preds = %loop-memcpy-residual-header2360
  %.pre3262 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i476, 6
  br label %Flow3774, !amdgpu.uniform !42

Flow3774:                                         ; preds = %loop-memcpy-residual-header2360.post-loop-memcpy-expansion2356_crit_edge, %loop-memcpy-residual-header2360
  %2919 = phi i64 [ %.pre3262, %loop-memcpy-residual-header2360.post-loop-memcpy-expansion2356_crit_edge ], [ poison, %loop-memcpy-residual-header2360 ]
  %2920 = phi i1 [ false, %loop-memcpy-residual-header2360.post-loop-memcpy-expansion2356_crit_edge ], [ true, %loop-memcpy-residual-header2360 ]
  br i1 %2920, label %loop-memcpy-residual2359.preheader, label %Flow3775, !amdgpu.uniform !42

loop-memcpy-residual2359.preheader:               ; preds = %Flow3774
  %2921 = add nuw nsw i64 %75, 8
  %scevgep3133 = getelementptr i8, ptr %765, i64 %2921
  %2922 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i476, 6
  %scevgep3134 = getelementptr i8, ptr %scevgep3133, i64 %2922
  br label %loop-memcpy-residual2359, !amdgpu.uniform !42

Flow3767:                                         ; preds = %loop-memcpy-expansion2363
  br label %Flow3768, !amdgpu.uniform !42

Flow3769:                                         ; preds = %Flow3766, %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i497
  br label %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i498, !amdgpu.uniform !42

loop-memcpy-residual-header2366:                  ; preds = %Flow3768
  %2923 = icmp ne i64 %801, 0
  br i1 %2923, label %loop-memcpy-residual2365.preheader, label %Flow3766, !amdgpu.uniform !42

loop-memcpy-residual2365.preheader:               ; preds = %loop-memcpy-residual-header2366
  %umin3128 = call i64 @llvm.umin.i64(i64 %lsr.iv3126, i64 64)
  %2924 = lshr i64 %umin3128, 4
  %2925 = shl nuw nsw i64 %2924, 4
  %scevgep3131 = getelementptr i8, ptr addrspace(4) %lsr.iv3121, i64 %2925
  %scevgep3124 = getelementptr i8, ptr %800, i64 %75
  %scevgep3125 = getelementptr i8, ptr %scevgep3124, i64 %775
  %scevgep3129 = getelementptr i8, ptr %scevgep3125, i64 %2925
  br label %loop-memcpy-residual2365, !amdgpu.uniform !42

Flow3706:                                         ; preds = %loop-memcpy-expansion2369
  br label %Flow3707, !amdgpu.uniform !42

loop-memcpy-residual-header2372:                  ; preds = %Flow3707
  %2926 = icmp eq i64 %1213, 0
  br i1 %2926, label %loop-memcpy-residual-header2372.post-loop-memcpy-expansion2368_crit_edge, label %Flow3704, !amdgpu.uniform !42

loop-memcpy-residual-header2372.post-loop-memcpy-expansion2368_crit_edge: ; preds = %loop-memcpy-residual-header2372
  %.pre3268 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i890, 6
  br label %Flow3704, !amdgpu.uniform !42

Flow3704:                                         ; preds = %loop-memcpy-residual-header2372.post-loop-memcpy-expansion2368_crit_edge, %loop-memcpy-residual-header2372
  %2927 = phi i64 [ %.pre3268, %loop-memcpy-residual-header2372.post-loop-memcpy-expansion2368_crit_edge ], [ poison, %loop-memcpy-residual-header2372 ]
  %2928 = phi i1 [ false, %loop-memcpy-residual-header2372.post-loop-memcpy-expansion2368_crit_edge ], [ true, %loop-memcpy-residual-header2372 ]
  br i1 %2928, label %loop-memcpy-residual2371.preheader, label %Flow3705, !amdgpu.uniform !42

loop-memcpy-residual2371.preheader:               ; preds = %Flow3704
  %2929 = add nuw nsw i64 %75, 8
  %scevgep3057 = getelementptr i8, ptr %1212, i64 %2929
  %2930 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i890, 6
  %scevgep3058 = getelementptr i8, ptr %scevgep3057, i64 %2930
  br label %loop-memcpy-residual2371, !amdgpu.uniform !42

Flow3469:                                         ; preds = %loop-memcpy-expansion2375
  br label %Flow3470, !amdgpu.uniform !42

loop-memcpy-residual-header2378:                  ; preds = %Flow3470
  br i1 %2926, label %loop-memcpy-residual-header2378.post-loop-memcpy-expansion2374_crit_edge, label %Flow3466, !amdgpu.uniform !42

loop-memcpy-residual-header2378.post-loop-memcpy-expansion2374_crit_edge: ; preds = %loop-memcpy-residual-header2378
  %.pre3282 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i1823, 6
  br label %Flow3466, !amdgpu.uniform !42

Flow3466:                                         ; preds = %loop-memcpy-residual-header2378.post-loop-memcpy-expansion2374_crit_edge, %loop-memcpy-residual-header2378
  %2931 = phi i64 [ %.pre3282, %loop-memcpy-residual-header2378.post-loop-memcpy-expansion2374_crit_edge ], [ poison, %loop-memcpy-residual-header2378 ]
  %2932 = phi i1 [ false, %loop-memcpy-residual-header2378.post-loop-memcpy-expansion2374_crit_edge ], [ true, %loop-memcpy-residual-header2378 ]
  br i1 %2932, label %loop-memcpy-residual2377.preheader, label %Flow3467, !amdgpu.uniform !42

loop-memcpy-residual2377.preheader:               ; preds = %Flow3466
  %2933 = add nuw nsw i64 %75, 8
  %scevgep2893 = getelementptr i8, ptr %2335, i64 %2933
  %2934 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i1823, 6
  %scevgep2894 = getelementptr i8, ptr %scevgep2893, i64 %2934
  br label %loop-memcpy-residual2377, !amdgpu.uniform !42

Flow3695:                                         ; preds = %loop-memcpy-expansion2381
  br label %Flow3696, !amdgpu.uniform !42

Flow3697:                                         ; preds = %Flow3693, %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i911
  br label %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i912, !amdgpu.uniform !42

loop-memcpy-residual-header2384:                  ; preds = %Flow3696
  %2935 = icmp ne i64 %1248, 0
  br i1 %2935, label %loop-memcpy-residual2383.preheader, label %Flow3693, !amdgpu.uniform !42

loop-memcpy-residual2383.preheader:               ; preds = %loop-memcpy-residual-header2384
  %umin3052 = call i64 @llvm.umin.i64(i64 %lsr.iv3050, i64 64)
  %2936 = lshr i64 %umin3052, 4
  %2937 = shl nuw nsw i64 %2936, 4
  %scevgep3055 = getelementptr i8, ptr addrspace(4) %lsr.iv3045, i64 %2937
  %scevgep3048 = getelementptr i8, ptr %1247, i64 %75
  %scevgep3049 = getelementptr i8, ptr %scevgep3048, i64 %1222
  %scevgep3053 = getelementptr i8, ptr %scevgep3049, i64 %2937
  br label %loop-memcpy-residual2383, !amdgpu.uniform !42

Flow3616:                                         ; preds = %loop-memcpy-expansion2387
  br label %Flow3617, !amdgpu.uniform !42

loop-memcpy-residual-header2390:                  ; preds = %Flow3617
  %2938 = icmp eq i64 %1667, 0
  br i1 %2938, label %loop-memcpy-residual-header2390.post-loop-memcpy-expansion2386_crit_edge, label %Flow3613, !amdgpu.uniform !42

loop-memcpy-residual-header2390.post-loop-memcpy-expansion2386_crit_edge: ; preds = %loop-memcpy-residual-header2390
  %.pre3272 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i1201, 6
  br label %Flow3613, !amdgpu.uniform !42

Flow3613:                                         ; preds = %loop-memcpy-residual-header2390.post-loop-memcpy-expansion2386_crit_edge, %loop-memcpy-residual-header2390
  %2939 = phi i64 [ %.pre3272, %loop-memcpy-residual-header2390.post-loop-memcpy-expansion2386_crit_edge ], [ poison, %loop-memcpy-residual-header2390 ]
  %2940 = phi i1 [ false, %loop-memcpy-residual-header2390.post-loop-memcpy-expansion2386_crit_edge ], [ true, %loop-memcpy-residual-header2390 ]
  br i1 %2940, label %loop-memcpy-residual2389.preheader, label %Flow3614, !amdgpu.uniform !42

loop-memcpy-residual2389.preheader:               ; preds = %Flow3613
  %2941 = add nuw nsw i64 %75, 8
  %scevgep3009 = getelementptr i8, ptr %1666, i64 %2941
  %2942 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i1201, 6
  %scevgep3010 = getelementptr i8, ptr %scevgep3009, i64 %2942
  br label %loop-memcpy-residual2389, !amdgpu.uniform !42

Flow3603:                                         ; preds = %loop-memcpy-expansion2393
  br label %Flow3604, !amdgpu.uniform !42

Flow3605:                                         ; preds = %Flow3601, %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i1222
  br label %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1223, !amdgpu.uniform !42

loop-memcpy-residual-header2396:                  ; preds = %Flow3604
  %2943 = icmp ne i64 %1702, 0
  br i1 %2943, label %loop-memcpy-residual2395.preheader, label %Flow3601, !amdgpu.uniform !42

loop-memcpy-residual2395.preheader:               ; preds = %loop-memcpy-residual-header2396
  %umin3004 = call i64 @llvm.umin.i64(i64 %lsr.iv3002, i64 64)
  %2944 = lshr i64 %umin3004, 4
  %2945 = shl nuw nsw i64 %2944, 4
  %scevgep3007 = getelementptr i8, ptr addrspace(4) %lsr.iv2997, i64 %2945
  %scevgep3000 = getelementptr i8, ptr %1701, i64 %75
  %scevgep3001 = getelementptr i8, ptr %scevgep3000, i64 %1676
  %scevgep3005 = getelementptr i8, ptr %scevgep3001, i64 %2945
  br label %loop-memcpy-residual2395, !amdgpu.uniform !42

Flow3590:                                         ; preds = %loop-memcpy-expansion2399
  br label %Flow3591, !amdgpu.uniform !42

loop-memcpy-residual-header2402:                  ; preds = %Flow3591
  %2946 = icmp eq i64 %1778, 0
  br i1 %2946, label %loop-memcpy-residual-header2402.post-loop-memcpy-expansion2398_crit_edge, label %Flow3587, !amdgpu.uniform !42

loop-memcpy-residual-header2402.post-loop-memcpy-expansion2398_crit_edge: ; preds = %loop-memcpy-residual-header2402
  %.pre3274 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i1300, 6
  br label %Flow3587, !amdgpu.uniform !42

Flow3587:                                         ; preds = %loop-memcpy-residual-header2402.post-loop-memcpy-expansion2398_crit_edge, %loop-memcpy-residual-header2402
  %2947 = phi i64 [ %.pre3274, %loop-memcpy-residual-header2402.post-loop-memcpy-expansion2398_crit_edge ], [ poison, %loop-memcpy-residual-header2402 ]
  %2948 = phi i1 [ false, %loop-memcpy-residual-header2402.post-loop-memcpy-expansion2398_crit_edge ], [ true, %loop-memcpy-residual-header2402 ]
  br i1 %2948, label %loop-memcpy-residual2401.preheader, label %Flow3588, !amdgpu.uniform !42

loop-memcpy-residual2401.preheader:               ; preds = %Flow3587
  %2949 = lshr i64 %spec.select.i.i.i.i.i.i.i1303, 4
  %2950 = shl nuw nsw i64 %2949, 4
  %2951 = add i64 %75, %2950
  %2952 = add i64 %2951, 8
  %scevgep2989 = getelementptr i8, ptr %1777, i64 %2952
  %2953 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i1300, 6
  %scevgep2990 = getelementptr i8, ptr %scevgep2989, i64 %2953
  %scevgep2992 = getelementptr i8, ptr addrspace(4) @.str.9, i64 %2950
  br label %loop-memcpy-residual2401, !amdgpu.uniform !42

Flow3577:                                         ; preds = %loop-memcpy-expansion2405
  br label %Flow3578, !amdgpu.uniform !42

Flow3579:                                         ; preds = %Flow3575, %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i1321
  br label %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1322, !amdgpu.uniform !42

loop-memcpy-residual-header2408:                  ; preds = %Flow3578
  %2954 = icmp ne i64 %1813, 0
  br i1 %2954, label %loop-memcpy-residual2407.preheader, label %Flow3575, !amdgpu.uniform !42

loop-memcpy-residual2407.preheader:               ; preds = %loop-memcpy-residual-header2408
  %umin2984 = call i64 @llvm.umin.i64(i64 %lsr.iv2982, i64 64)
  %2955 = lshr i64 %umin2984, 4
  %2956 = shl nuw nsw i64 %2955, 4
  %scevgep2987 = getelementptr i8, ptr addrspace(4) %lsr.iv2977, i64 %2956
  %scevgep2980 = getelementptr i8, ptr %1812, i64 %75
  %scevgep2981 = getelementptr i8, ptr %scevgep2980, i64 %1787
  %scevgep2985 = getelementptr i8, ptr %scevgep2981, i64 %2956
  br label %loop-memcpy-residual2407, !amdgpu.uniform !42

Flow3564:                                         ; preds = %loop-memcpy-expansion2411
  br label %Flow3565, !amdgpu.uniform !42

loop-memcpy-residual-header2414:                  ; preds = %Flow3565
  %2957 = icmp eq i64 %1889, 0
  br i1 %2957, label %loop-memcpy-residual-header2414.post-loop-memcpy-expansion2410_crit_edge, label %Flow3561, !amdgpu.uniform !42

loop-memcpy-residual-header2414.post-loop-memcpy-expansion2410_crit_edge: ; preds = %loop-memcpy-residual-header2414
  %.pre3276 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i1405, 6
  br label %Flow3561, !amdgpu.uniform !42

Flow3561:                                         ; preds = %loop-memcpy-residual-header2414.post-loop-memcpy-expansion2410_crit_edge, %loop-memcpy-residual-header2414
  %2958 = phi i64 [ %.pre3276, %loop-memcpy-residual-header2414.post-loop-memcpy-expansion2410_crit_edge ], [ poison, %loop-memcpy-residual-header2414 ]
  %2959 = phi i1 [ false, %loop-memcpy-residual-header2414.post-loop-memcpy-expansion2410_crit_edge ], [ true, %loop-memcpy-residual-header2414 ]
  br i1 %2959, label %loop-memcpy-residual2413.preheader, label %Flow3562, !amdgpu.uniform !42

loop-memcpy-residual2413.preheader:               ; preds = %Flow3561
  %2960 = add nuw nsw i64 %75, 8
  %scevgep2970 = getelementptr i8, ptr %1888, i64 %2960
  %2961 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i1405, 6
  %scevgep2971 = getelementptr i8, ptr %scevgep2970, i64 %2961
  br label %loop-memcpy-residual2413, !amdgpu.uniform !42

Flow3551:                                         ; preds = %loop-memcpy-expansion2417
  br label %Flow3552, !amdgpu.uniform !42

Flow3553:                                         ; preds = %Flow3549, %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i1426
  br label %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1427, !amdgpu.uniform !42

loop-memcpy-residual-header2420:                  ; preds = %Flow3552
  %2962 = icmp ne i64 %1924, 0
  br i1 %2962, label %loop-memcpy-residual2419.preheader, label %Flow3549, !amdgpu.uniform !42

loop-memcpy-residual2419.preheader:               ; preds = %loop-memcpy-residual-header2420
  %umin2965 = call i64 @llvm.umin.i64(i64 %lsr.iv2963, i64 64)
  %2963 = lshr i64 %umin2965, 4
  %2964 = shl nuw nsw i64 %2963, 4
  %scevgep2968 = getelementptr i8, ptr addrspace(4) %lsr.iv2958, i64 %2964
  %scevgep2961 = getelementptr i8, ptr %1923, i64 %75
  %scevgep2962 = getelementptr i8, ptr %scevgep2961, i64 %1898
  %scevgep2966 = getelementptr i8, ptr %scevgep2962, i64 %2964
  br label %loop-memcpy-residual2419, !amdgpu.uniform !42

Flow3459:                                         ; preds = %loop-memcpy-expansion2423
  br label %Flow3460, !amdgpu.uniform !42

Flow3461:                                         ; preds = %Flow3458, %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i1844
  br label %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1845, !amdgpu.uniform !42

loop-memcpy-residual-header2426:                  ; preds = %Flow3460
  %2965 = icmp ne i64 %2368, 0
  br i1 %2965, label %loop-memcpy-residual2425.preheader, label %Flow3458, !amdgpu.uniform !42

loop-memcpy-residual2425.preheader:               ; preds = %loop-memcpy-residual-header2426
  %umin2888 = call i64 @llvm.umin.i64(i64 %lsr.iv2886, i64 64)
  %2966 = lshr i64 %umin2888, 4
  %2967 = shl nuw nsw i64 %2966, 4
  %scevgep2891 = getelementptr i8, ptr addrspace(4) %lsr.iv2881, i64 %2967
  %scevgep2884 = getelementptr i8, ptr %2367, i64 %75
  %scevgep2885 = getelementptr i8, ptr %scevgep2884, i64 %2342
  %scevgep2889 = getelementptr i8, ptr %scevgep2885, i64 %2967
  br label %loop-memcpy-residual2425, !amdgpu.uniform !42

ComputeLoop:                                      ; preds = %for.cond.i.i1.i, %ComputeLoop
  %Accumulator = phi i32 [ %2972, %ComputeLoop ], [ 0, %for.cond.i.i1.i ]
  %OldValuePhi = phi i32 [ %2971, %ComputeLoop ], [ poison, %for.cond.i.i1.i ]
  %ActiveBits = phi i64 [ %2975, %ComputeLoop ], [ %25, %for.cond.i.i1.i ]
  %2968 = call i64 @llvm.cttz.i64(i64 %ActiveBits, i1 true)
  %2969 = trunc i64 %2968 to i32
  %2970 = call i32 @llvm.amdgcn.readlane.i32(i32 %shl.i.i.i.i.i, i32 %2969)
  %2971 = call i32 @llvm.amdgcn.writelane.i32(i32 %Accumulator, i32 %2969, i32 %OldValuePhi)
  %2972 = or i32 %Accumulator, %2970
  %2973 = shl i64 1, %2968
  %2974 = xor i64 %2973, -1
  %2975 = and i64 %ActiveBits, %2974
  %2976 = icmp eq i64 %2975, 0
  br i1 %2976, label %ComputeEnd, label %ComputeLoop, !amdgpu.uniform !42

ComputeEnd:                                       ; preds = %ComputeLoop
  %.lcssa4223 = phi i32 [ %2971, %ComputeLoop ]
  %.lcssa4222 = phi i32 [ %2972, %ComputeLoop ]
  %div8.i.i.i.i.i = lshr i32 %spec.store.select.i.i.i, 5
  %idxprom.i.i.i.i.i = zext nneg i32 %div8.i.i.i.i.i to i64
  %2977 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i.i.i
  %2978 = trunc i64 %25 to i32
  %2979 = lshr i64 %25, 32
  %2980 = trunc i64 %2979 to i32
  %2981 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %2978, i32 0)
  %2982 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %2980, i32 %2981)
  %2983 = icmp eq i32 %2982, 0
  %2984 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %2983)
  %2985 = extractvalue { i1, i64 } %2984, 0
  %2986 = extractvalue { i1, i64 } %2984, 1
  br i1 %2985, label %26, label %28

Flow3900:                                         ; preds = %46, %if.then.i.i.i.i2
  %2987 = phi i1 [ false, %46 ], [ true, %if.then.i.i.i.i2 ]
  br i1 %2987, label %if.end14.i.i.i, label %Flow3901, !amdgpu.uniform !42

ComputeLoop2428:                                  ; preds = %if.then12.i.i.i, %ComputeLoop2428
  %Accumulator2430 = phi i32 [ %2991, %ComputeLoop2428 ], [ -1, %if.then12.i.i.i ]
  %ActiveBits2431 = phi i64 [ %2994, %ComputeLoop2428 ], [ %43, %if.then12.i.i.i ]
  %2988 = call i64 @llvm.cttz.i64(i64 %ActiveBits2431, i1 true)
  %2989 = trunc i64 %2988 to i32
  %2990 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i.i.i, i32 %2989)
  %2991 = and i32 %Accumulator2430, %2990
  %2992 = shl i64 1, %2988
  %2993 = xor i64 %2992, -1
  %2994 = and i64 %ActiveBits2431, %2993
  %2995 = icmp eq i64 %2994, 0
  br i1 %2995, label %ComputeEnd2429, label %ComputeLoop2428, !amdgpu.uniform !42

ComputeEnd2429:                                   ; preds = %ComputeLoop2428
  %.lcssa4224 = phi i32 [ %2991, %ComputeLoop2428 ]
  %2996 = trunc i64 %43 to i32
  %2997 = lshr i64 %43, 32
  %2998 = trunc i64 %2997 to i32
  %2999 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %2996, i32 0)
  %3000 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %2998, i32 %2999)
  %3001 = icmp eq i32 %3000, 0
  %3002 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3001)
  %3003 = extractvalue { i1, i64 } %3002, 0
  %3004 = extractvalue { i1, i64 } %3002, 1
  br i1 %3003, label %44, label %46

ComputeLoop2432:                                  ; preds = %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit, %ComputeLoop2432
  %Accumulator2434 = phi i32 [ %3008, %ComputeLoop2432 ], [ -1, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit ]
  %ActiveBits2435 = phi i64 [ %3011, %ComputeLoop2432 ], [ %125, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit ]
  %3005 = call i64 @llvm.cttz.i64(i64 %ActiveBits2435, i1 true)
  %3006 = trunc i64 %3005 to i32
  %3007 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i21.i.i, i32 %3006)
  %3008 = and i32 %Accumulator2434, %3007
  %3009 = shl i64 1, %3005
  %3010 = xor i64 %3009, -1
  %3011 = and i64 %ActiveBits2435, %3010
  %3012 = icmp eq i64 %3011, 0
  br i1 %3012, label %ComputeEnd2433, label %ComputeLoop2432, !amdgpu.uniform !42

ComputeEnd2433:                                   ; preds = %ComputeLoop2432
  %.lcssa4218 = phi i32 [ %3008, %ComputeLoop2432 ]
  %3013 = trunc i64 %125 to i32
  %3014 = lshr i64 %125, 32
  %3015 = trunc i64 %3014 to i32
  %3016 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %3013, i32 0)
  %3017 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %3015, i32 %3016)
  %3018 = icmp eq i32 %3017, 0
  %3019 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3018)
  %3020 = extractvalue { i1, i64 } %3019, 0
  %3021 = extractvalue { i1, i64 } %3019, 1
  br i1 %3020, label %126, label %129

ComputeLoop2436:                                  ; preds = %for.cond.i.i1.i13, %ComputeLoop2436
  %Accumulator2438 = phi i32 [ %3026, %ComputeLoop2436 ], [ 0, %for.cond.i.i1.i13 ]
  %OldValuePhi2439 = phi i32 [ %3025, %ComputeLoop2436 ], [ poison, %for.cond.i.i1.i13 ]
  %ActiveBits2440 = phi i64 [ %3029, %ComputeLoop2436 ], [ %138, %for.cond.i.i1.i13 ]
  %3022 = call i64 @llvm.cttz.i64(i64 %ActiveBits2440, i1 true)
  %3023 = trunc i64 %3022 to i32
  %3024 = call i32 @llvm.amdgcn.readlane.i32(i32 %shl.i.i.i.i.i25, i32 %3023)
  %3025 = call i32 @llvm.amdgcn.writelane.i32(i32 %Accumulator2438, i32 %3023, i32 %OldValuePhi2439)
  %3026 = or i32 %Accumulator2438, %3024
  %3027 = shl i64 1, %3022
  %3028 = xor i64 %3027, -1
  %3029 = and i64 %ActiveBits2440, %3028
  %3030 = icmp eq i64 %3029, 0
  br i1 %3030, label %ComputeEnd2437, label %ComputeLoop2436, !amdgpu.uniform !42

ComputeEnd2437:                                   ; preds = %ComputeLoop2436
  %.lcssa4213 = phi i32 [ %3025, %ComputeLoop2436 ]
  %.lcssa4212 = phi i32 [ %3026, %ComputeLoop2436 ]
  %div8.i.i.i.i.i21 = lshr i32 %spec.store.select.i.i.i20, 5
  %idxprom.i.i.i.i.i22 = zext nneg i32 %div8.i.i.i.i.i21 to i64
  %3031 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i.i.i22
  %3032 = trunc i64 %138 to i32
  %3033 = lshr i64 %138, 32
  %3034 = trunc i64 %3033 to i32
  %3035 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %3032, i32 0)
  %3036 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %3034, i32 %3035)
  %3037 = icmp eq i32 %3036, 0
  %3038 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3037)
  %3039 = extractvalue { i1, i64 } %3038, 0
  %3040 = extractvalue { i1, i64 } %3038, 1
  br i1 %3039, label %139, label %141

Flow3881:                                         ; preds = %159, %if.then.i.i.i.i31
  %3041 = phi i1 [ false, %159 ], [ true, %if.then.i.i.i.i31 ]
  br i1 %3041, label %if.end14.i.i.i101, label %Flow3882, !amdgpu.uniform !42

ComputeLoop2441:                                  ; preds = %if.then12.i.i.i36, %ComputeLoop2441
  %Accumulator2443 = phi i32 [ %3045, %ComputeLoop2441 ], [ -1, %if.then12.i.i.i36 ]
  %ActiveBits2444 = phi i64 [ %3048, %ComputeLoop2441 ], [ %156, %if.then12.i.i.i36 ]
  %3042 = call i64 @llvm.cttz.i64(i64 %ActiveBits2444, i1 true)
  %3043 = trunc i64 %3042 to i32
  %3044 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i.i.i42, i32 %3043)
  %3045 = and i32 %Accumulator2443, %3044
  %3046 = shl i64 1, %3042
  %3047 = xor i64 %3046, -1
  %3048 = and i64 %ActiveBits2444, %3047
  %3049 = icmp eq i64 %3048, 0
  br i1 %3049, label %ComputeEnd2442, label %ComputeLoop2441, !amdgpu.uniform !42

ComputeEnd2442:                                   ; preds = %ComputeLoop2441
  %.lcssa4214 = phi i32 [ %3045, %ComputeLoop2441 ]
  %3050 = trunc i64 %156 to i32
  %3051 = lshr i64 %156, 32
  %3052 = trunc i64 %3051 to i32
  %3053 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %3050, i32 0)
  %3054 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %3052, i32 %3053)
  %3055 = icmp eq i32 %3054, 0
  %3056 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3055)
  %3057 = extractvalue { i1, i64 } %3056, 0
  %3058 = extractvalue { i1, i64 } %3056, 1
  br i1 %3057, label %157, label %159

ComputeLoop2445:                                  ; preds = %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit109, %ComputeLoop2445
  %Accumulator2447 = phi i32 [ %3062, %ComputeLoop2445 ], [ -1, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit109 ]
  %ActiveBits2448 = phi i64 [ %3065, %ComputeLoop2445 ], [ %236, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit109 ]
  %3059 = call i64 @llvm.cttz.i64(i64 %ActiveBits2448, i1 true)
  %3060 = trunc i64 %3059 to i32
  %3061 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i21.i.i100, i32 %3060)
  %3062 = and i32 %Accumulator2447, %3061
  %3063 = shl i64 1, %3059
  %3064 = xor i64 %3063, -1
  %3065 = and i64 %ActiveBits2448, %3064
  %3066 = icmp eq i64 %3065, 0
  br i1 %3066, label %ComputeEnd2446, label %ComputeLoop2445, !amdgpu.uniform !42

ComputeEnd2446:                                   ; preds = %ComputeLoop2445
  %.lcssa4208 = phi i32 [ %3062, %ComputeLoop2445 ]
  %3067 = trunc i64 %236 to i32
  %3068 = lshr i64 %236, 32
  %3069 = trunc i64 %3068 to i32
  %3070 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %3067, i32 0)
  %3071 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %3069, i32 %3070)
  %3072 = icmp eq i32 %3071, 0
  %3073 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3072)
  %3074 = extractvalue { i1, i64 } %3073, 0
  %3075 = extractvalue { i1, i64 } %3073, 1
  br i1 %3074, label %237, label %240

ComputeLoop2449:                                  ; preds = %for.cond.i.i.i118, %ComputeLoop2449
  %Accumulator2451 = phi i32 [ %3080, %ComputeLoop2449 ], [ 0, %for.cond.i.i.i118 ]
  %OldValuePhi2452 = phi i32 [ %3079, %ComputeLoop2449 ], [ poison, %for.cond.i.i.i118 ]
  %ActiveBits2453 = phi i64 [ %3083, %ComputeLoop2449 ], [ %362, %for.cond.i.i.i118 ]
  %3076 = call i64 @llvm.cttz.i64(i64 %ActiveBits2453, i1 true)
  %3077 = trunc i64 %3076 to i32
  %3078 = call i32 @llvm.amdgcn.readlane.i32(i32 %shl.i.i.i.i.i130, i32 %3077)
  %3079 = call i32 @llvm.amdgcn.writelane.i32(i32 %Accumulator2451, i32 %3077, i32 %OldValuePhi2452)
  %3080 = or i32 %Accumulator2451, %3078
  %3081 = shl i64 1, %3076
  %3082 = xor i64 %3081, -1
  %3083 = and i64 %ActiveBits2453, %3082
  %3084 = icmp eq i64 %3083, 0
  br i1 %3084, label %ComputeEnd2450, label %ComputeLoop2449, !amdgpu.uniform !42

ComputeEnd2450:                                   ; preds = %ComputeLoop2449
  %.lcssa4199 = phi i32 [ %3079, %ComputeLoop2449 ]
  %.lcssa4198 = phi i32 [ %3080, %ComputeLoop2449 ]
  %div8.i.i.i.i.i126 = lshr i32 %spec.store.select.i.i.i125, 5
  %idxprom.i.i.i.i.i127 = zext nneg i32 %div8.i.i.i.i.i126 to i64
  %3085 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i.i.i127
  %3086 = trunc i64 %362 to i32
  %3087 = lshr i64 %362, 32
  %3088 = trunc i64 %3087 to i32
  %3089 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %3086, i32 0)
  %3090 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %3088, i32 %3089)
  %3091 = icmp eq i32 %3090, 0
  %3092 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3091)
  %3093 = extractvalue { i1, i64 } %3092, 0
  %3094 = extractvalue { i1, i64 } %3092, 1
  br i1 %3093, label %363, label %365

Flow3839:                                         ; preds = %383, %if.then.i.i.i.i136
  %3095 = phi i1 [ false, %383 ], [ true, %if.then.i.i.i.i136 ]
  br i1 %3095, label %if.end14.i.i.i206, label %Flow3840, !amdgpu.uniform !42

ComputeLoop2454:                                  ; preds = %if.then12.i.i.i141, %ComputeLoop2454
  %Accumulator2456 = phi i32 [ %3099, %ComputeLoop2454 ], [ -1, %if.then12.i.i.i141 ]
  %ActiveBits2457 = phi i64 [ %3102, %ComputeLoop2454 ], [ %380, %if.then12.i.i.i141 ]
  %3096 = call i64 @llvm.cttz.i64(i64 %ActiveBits2457, i1 true)
  %3097 = trunc i64 %3096 to i32
  %3098 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i.i.i146, i32 %3097)
  %3099 = and i32 %Accumulator2456, %3098
  %3100 = shl i64 1, %3096
  %3101 = xor i64 %3100, -1
  %3102 = and i64 %ActiveBits2457, %3101
  %3103 = icmp eq i64 %3102, 0
  br i1 %3103, label %ComputeEnd2455, label %ComputeLoop2454, !amdgpu.uniform !42

ComputeEnd2455:                                   ; preds = %ComputeLoop2454
  %.lcssa4200 = phi i32 [ %3099, %ComputeLoop2454 ]
  %3104 = trunc i64 %380 to i32
  %3105 = lshr i64 %380, 32
  %3106 = trunc i64 %3105 to i32
  %3107 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %3104, i32 0)
  %3108 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %3106, i32 %3107)
  %3109 = icmp eq i32 %3108, 0
  %3110 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3109)
  %3111 = extractvalue { i1, i64 } %3110, 0
  %3112 = extractvalue { i1, i64 } %3110, 1
  br i1 %3111, label %381, label %383

ComputeLoop2458:                                  ; preds = %_ZN22__llvm_libc_22_0_0_git15write_to_stderrENS_3cpp11string_viewE.exit.i, %ComputeLoop2458
  %Accumulator2460 = phi i32 [ %3116, %ComputeLoop2458 ], [ -1, %_ZN22__llvm_libc_22_0_0_git15write_to_stderrENS_3cpp11string_viewE.exit.i ]
  %ActiveBits2461 = phi i64 [ %3119, %ComputeLoop2458 ], [ %478, %_ZN22__llvm_libc_22_0_0_git15write_to_stderrENS_3cpp11string_viewE.exit.i ]
  %3113 = call i64 @llvm.cttz.i64(i64 %ActiveBits2461, i1 true)
  %3114 = trunc i64 %3113 to i32
  %3115 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i21.i.i203, i32 %3114)
  %3116 = and i32 %Accumulator2460, %3115
  %3117 = shl i64 1, %3113
  %3118 = xor i64 %3117, -1
  %3119 = and i64 %ActiveBits2461, %3118
  %3120 = icmp eq i64 %3119, 0
  br i1 %3120, label %ComputeEnd2459, label %ComputeLoop2458, !amdgpu.uniform !42

ComputeEnd2459:                                   ; preds = %ComputeLoop2458
  %.lcssa4190 = phi i32 [ %3116, %ComputeLoop2458 ]
  %3121 = trunc i64 %478 to i32
  %3122 = lshr i64 %478, 32
  %3123 = trunc i64 %3122 to i32
  %3124 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %3121, i32 0)
  %3125 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %3123, i32 %3124)
  %3126 = icmp eq i32 %3125, 0
  %3127 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3126)
  %3128 = extractvalue { i1, i64 } %3127, 0
  %3129 = extractvalue { i1, i64 } %3127, 1
  br i1 %3128, label %479, label %482

ComputeLoop2462:                                  ; preds = %for.cond.i.i1.i222, %ComputeLoop2462
  %Accumulator2464 = phi i32 [ %3134, %ComputeLoop2462 ], [ 0, %for.cond.i.i1.i222 ]
  %OldValuePhi2465 = phi i32 [ %3133, %ComputeLoop2462 ], [ poison, %for.cond.i.i1.i222 ]
  %ActiveBits2466 = phi i64 [ %3137, %ComputeLoop2462 ], [ %491, %for.cond.i.i1.i222 ]
  %3130 = call i64 @llvm.cttz.i64(i64 %ActiveBits2466, i1 true)
  %3131 = trunc i64 %3130 to i32
  %3132 = call i32 @llvm.amdgcn.readlane.i32(i32 %shl.i.i.i.i.i234, i32 %3131)
  %3133 = call i32 @llvm.amdgcn.writelane.i32(i32 %Accumulator2464, i32 %3131, i32 %OldValuePhi2465)
  %3134 = or i32 %Accumulator2464, %3132
  %3135 = shl i64 1, %3130
  %3136 = xor i64 %3135, -1
  %3137 = and i64 %ActiveBits2466, %3136
  %3138 = icmp eq i64 %3137, 0
  br i1 %3138, label %ComputeEnd2463, label %ComputeLoop2462, !amdgpu.uniform !42

ComputeEnd2463:                                   ; preds = %ComputeLoop2462
  %.lcssa4185 = phi i32 [ %3133, %ComputeLoop2462 ]
  %.lcssa4184 = phi i32 [ %3134, %ComputeLoop2462 ]
  %div8.i.i.i.i.i230 = lshr i32 %spec.store.select.i.i.i229, 5
  %idxprom.i.i.i.i.i231 = zext nneg i32 %div8.i.i.i.i.i230 to i64
  %3139 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i.i.i231
  %3140 = trunc i64 %491 to i32
  %3141 = lshr i64 %491, 32
  %3142 = trunc i64 %3141 to i32
  %3143 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %3140, i32 0)
  %3144 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %3142, i32 %3143)
  %3145 = icmp eq i32 %3144, 0
  %3146 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3145)
  %3147 = extractvalue { i1, i64 } %3146, 0
  %3148 = extractvalue { i1, i64 } %3146, 1
  br i1 %3147, label %492, label %494

Flow3820:                                         ; preds = %512, %if.then.i.i.i.i240
  %3149 = phi i1 [ false, %512 ], [ true, %if.then.i.i.i.i240 ]
  br i1 %3149, label %if.end14.i.i.i311, label %Flow3821, !amdgpu.uniform !42

ComputeLoop2467:                                  ; preds = %if.then12.i.i.i245, %ComputeLoop2467
  %Accumulator2469 = phi i32 [ %3153, %ComputeLoop2467 ], [ -1, %if.then12.i.i.i245 ]
  %ActiveBits2470 = phi i64 [ %3156, %ComputeLoop2467 ], [ %509, %if.then12.i.i.i245 ]
  %3150 = call i64 @llvm.cttz.i64(i64 %ActiveBits2470, i1 true)
  %3151 = trunc i64 %3150 to i32
  %3152 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i.i.i251, i32 %3151)
  %3153 = and i32 %Accumulator2469, %3152
  %3154 = shl i64 1, %3150
  %3155 = xor i64 %3154, -1
  %3156 = and i64 %ActiveBits2470, %3155
  %3157 = icmp eq i64 %3156, 0
  br i1 %3157, label %ComputeEnd2468, label %ComputeLoop2467, !amdgpu.uniform !42

ComputeEnd2468:                                   ; preds = %ComputeLoop2467
  %.lcssa4186 = phi i32 [ %3153, %ComputeLoop2467 ]
  %3158 = trunc i64 %509 to i32
  %3159 = lshr i64 %509, 32
  %3160 = trunc i64 %3159 to i32
  %3161 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %3158, i32 0)
  %3162 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %3160, i32 %3161)
  %3163 = icmp eq i32 %3162, 0
  %3164 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3163)
  %3165 = extractvalue { i1, i64 } %3164, 0
  %3166 = extractvalue { i1, i64 } %3164, 1
  br i1 %3165, label %510, label %512

ComputeLoop2471:                                  ; preds = %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit319, %ComputeLoop2471
  %Accumulator2473 = phi i32 [ %3170, %ComputeLoop2471 ], [ -1, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit319 ]
  %ActiveBits2474 = phi i64 [ %3173, %ComputeLoop2471 ], [ %589, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit319 ]
  %3167 = call i64 @llvm.cttz.i64(i64 %ActiveBits2474, i1 true)
  %3168 = trunc i64 %3167 to i32
  %3169 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i21.i.i310, i32 %3168)
  %3170 = and i32 %Accumulator2473, %3169
  %3171 = shl i64 1, %3167
  %3172 = xor i64 %3171, -1
  %3173 = and i64 %ActiveBits2474, %3172
  %3174 = icmp eq i64 %3173, 0
  br i1 %3174, label %ComputeEnd2472, label %ComputeLoop2471, !amdgpu.uniform !42

ComputeEnd2472:                                   ; preds = %ComputeLoop2471
  %.lcssa4180 = phi i32 [ %3170, %ComputeLoop2471 ]
  %3175 = trunc i64 %589 to i32
  %3176 = lshr i64 %589, 32
  %3177 = trunc i64 %3176 to i32
  %3178 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %3175, i32 0)
  %3179 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %3177, i32 %3178)
  %3180 = icmp eq i32 %3179, 0
  %3181 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3180)
  %3182 = extractvalue { i1, i64 } %3181, 0
  %3183 = extractvalue { i1, i64 } %3181, 1
  br i1 %3182, label %590, label %593

ComputeLoop2475:                                  ; preds = %for.cond.i.i.i328, %ComputeLoop2475
  %Accumulator2477 = phi i32 [ %3188, %ComputeLoop2475 ], [ 0, %for.cond.i.i.i328 ]
  %OldValuePhi2478 = phi i32 [ %3187, %ComputeLoop2475 ], [ poison, %for.cond.i.i.i328 ]
  %ActiveBits2479 = phi i64 [ %3191, %ComputeLoop2475 ], [ %614, %for.cond.i.i.i328 ]
  %3184 = call i64 @llvm.cttz.i64(i64 %ActiveBits2479, i1 true)
  %3185 = trunc i64 %3184 to i32
  %3186 = call i32 @llvm.amdgcn.readlane.i32(i32 %shl.i.i.i.i.i340, i32 %3185)
  %3187 = call i32 @llvm.amdgcn.writelane.i32(i32 %Accumulator2477, i32 %3185, i32 %OldValuePhi2478)
  %3188 = or i32 %Accumulator2477, %3186
  %3189 = shl i64 1, %3184
  %3190 = xor i64 %3189, -1
  %3191 = and i64 %ActiveBits2479, %3190
  %3192 = icmp eq i64 %3191, 0
  br i1 %3192, label %ComputeEnd2476, label %ComputeLoop2475, !amdgpu.uniform !42

ComputeEnd2476:                                   ; preds = %ComputeLoop2475
  %.lcssa4175 = phi i32 [ %3187, %ComputeLoop2475 ]
  %.lcssa4174 = phi i32 [ %3188, %ComputeLoop2475 ]
  %div8.i.i.i.i.i336 = lshr i32 %spec.store.select.i.i.i335, 5
  %idxprom.i.i.i.i.i337 = zext nneg i32 %div8.i.i.i.i.i336 to i64
  %3193 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i.i.i337
  %3194 = trunc i64 %614 to i32
  %3195 = lshr i64 %614, 32
  %3196 = trunc i64 %3195 to i32
  %3197 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %3194, i32 0)
  %3198 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %3196, i32 %3197)
  %3199 = icmp eq i32 %3198, 0
  %3200 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3199)
  %3201 = extractvalue { i1, i64 } %3200, 0
  %3202 = extractvalue { i1, i64 } %3200, 1
  br i1 %3201, label %615, label %617

Flow3797:                                         ; preds = %635, %if.then.i.i.i.i346
  %3203 = phi i1 [ false, %635 ], [ true, %if.then.i.i.i.i346 ]
  br i1 %3203, label %if.end14.i.i.i417, label %Flow3798, !amdgpu.uniform !42

ComputeLoop2480:                                  ; preds = %if.then12.i.i.i351, %ComputeLoop2480
  %Accumulator2482 = phi i32 [ %3207, %ComputeLoop2480 ], [ -1, %if.then12.i.i.i351 ]
  %ActiveBits2483 = phi i64 [ %3210, %ComputeLoop2480 ], [ %632, %if.then12.i.i.i351 ]
  %3204 = call i64 @llvm.cttz.i64(i64 %ActiveBits2483, i1 true)
  %3205 = trunc i64 %3204 to i32
  %3206 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i.i.i357, i32 %3205)
  %3207 = and i32 %Accumulator2482, %3206
  %3208 = shl i64 1, %3204
  %3209 = xor i64 %3208, -1
  %3210 = and i64 %ActiveBits2483, %3209
  %3211 = icmp eq i64 %3210, 0
  br i1 %3211, label %ComputeEnd2481, label %ComputeLoop2480, !amdgpu.uniform !42

ComputeEnd2481:                                   ; preds = %ComputeLoop2480
  %.lcssa4176 = phi i32 [ %3207, %ComputeLoop2480 ]
  %3212 = trunc i64 %632 to i32
  %3213 = lshr i64 %632, 32
  %3214 = trunc i64 %3213 to i32
  %3215 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %3212, i32 0)
  %3216 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %3214, i32 %3215)
  %3217 = icmp eq i32 %3216, 0
  %3218 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3217)
  %3219 = extractvalue { i1, i64 } %3218, 0
  %3220 = extractvalue { i1, i64 } %3218, 1
  br i1 %3219, label %633, label %635

ComputeLoop2484:                                  ; preds = %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsINS_3cpp6stringEEERS1_T_.exit, %ComputeLoop2484
  %Accumulator2486 = phi i32 [ %3224, %ComputeLoop2484 ], [ -1, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsINS_3cpp6stringEEERS1_T_.exit ]
  %ActiveBits2487 = phi i64 [ %3227, %ComputeLoop2484 ], [ %713, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsINS_3cpp6stringEEERS1_T_.exit ]
  %3221 = call i64 @llvm.cttz.i64(i64 %ActiveBits2487, i1 true)
  %3222 = trunc i64 %3221 to i32
  %3223 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i21.i.i416, i32 %3222)
  %3224 = and i32 %Accumulator2486, %3223
  %3225 = shl i64 1, %3221
  %3226 = xor i64 %3225, -1
  %3227 = and i64 %ActiveBits2487, %3226
  %3228 = icmp eq i64 %3227, 0
  br i1 %3228, label %ComputeEnd2485, label %ComputeLoop2484, !amdgpu.uniform !42

ComputeEnd2485:                                   ; preds = %ComputeLoop2484
  %.lcssa4170 = phi i32 [ %3224, %ComputeLoop2484 ]
  %3229 = trunc i64 %713 to i32
  %3230 = lshr i64 %713, 32
  %3231 = trunc i64 %3230 to i32
  %3232 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %3229, i32 0)
  %3233 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %3231, i32 %3232)
  %3234 = icmp eq i32 %3233, 0
  %3235 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3234)
  %3236 = extractvalue { i1, i64 } %3235, 0
  %3237 = extractvalue { i1, i64 } %3235, 1
  br i1 %3236, label %714, label %717

ComputeLoop2488:                                  ; preds = %for.cond.i.i1.i433, %ComputeLoop2488
  %Accumulator2490 = phi i32 [ %3242, %ComputeLoop2488 ], [ 0, %for.cond.i.i1.i433 ]
  %OldValuePhi2491 = phi i32 [ %3241, %ComputeLoop2488 ], [ poison, %for.cond.i.i1.i433 ]
  %ActiveBits2492 = phi i64 [ %3245, %ComputeLoop2488 ], [ %726, %for.cond.i.i1.i433 ]
  %3238 = call i64 @llvm.cttz.i64(i64 %ActiveBits2492, i1 true)
  %3239 = trunc i64 %3238 to i32
  %3240 = call i32 @llvm.amdgcn.readlane.i32(i32 %shl.i.i.i.i.i445, i32 %3239)
  %3241 = call i32 @llvm.amdgcn.writelane.i32(i32 %Accumulator2490, i32 %3239, i32 %OldValuePhi2491)
  %3242 = or i32 %Accumulator2490, %3240
  %3243 = shl i64 1, %3238
  %3244 = xor i64 %3243, -1
  %3245 = and i64 %ActiveBits2492, %3244
  %3246 = icmp eq i64 %3245, 0
  br i1 %3246, label %ComputeEnd2489, label %ComputeLoop2488, !amdgpu.uniform !42

ComputeEnd2489:                                   ; preds = %ComputeLoop2488
  %.lcssa4165 = phi i32 [ %3241, %ComputeLoop2488 ]
  %.lcssa4164 = phi i32 [ %3242, %ComputeLoop2488 ]
  %div8.i.i.i.i.i441 = lshr i32 %spec.store.select.i.i.i440, 5
  %idxprom.i.i.i.i.i442 = zext nneg i32 %div8.i.i.i.i.i441 to i64
  %3247 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i.i.i442
  %3248 = trunc i64 %726 to i32
  %3249 = lshr i64 %726, 32
  %3250 = trunc i64 %3249 to i32
  %3251 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %3248, i32 0)
  %3252 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %3250, i32 %3251)
  %3253 = icmp eq i32 %3252, 0
  %3254 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3253)
  %3255 = extractvalue { i1, i64 } %3254, 0
  %3256 = extractvalue { i1, i64 } %3254, 1
  br i1 %3255, label %727, label %729

Flow3778:                                         ; preds = %747, %if.then.i.i.i.i451
  %3257 = phi i1 [ false, %747 ], [ true, %if.then.i.i.i.i451 ]
  br i1 %3257, label %if.end14.i.i.i522, label %Flow3779, !amdgpu.uniform !42

ComputeLoop2493:                                  ; preds = %if.then12.i.i.i456, %ComputeLoop2493
  %Accumulator2495 = phi i32 [ %3261, %ComputeLoop2493 ], [ -1, %if.then12.i.i.i456 ]
  %ActiveBits2496 = phi i64 [ %3264, %ComputeLoop2493 ], [ %744, %if.then12.i.i.i456 ]
  %3258 = call i64 @llvm.cttz.i64(i64 %ActiveBits2496, i1 true)
  %3259 = trunc i64 %3258 to i32
  %3260 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i.i.i462, i32 %3259)
  %3261 = and i32 %Accumulator2495, %3260
  %3262 = shl i64 1, %3258
  %3263 = xor i64 %3262, -1
  %3264 = and i64 %ActiveBits2496, %3263
  %3265 = icmp eq i64 %3264, 0
  br i1 %3265, label %ComputeEnd2494, label %ComputeLoop2493, !amdgpu.uniform !42

ComputeEnd2494:                                   ; preds = %ComputeLoop2493
  %.lcssa4166 = phi i32 [ %3261, %ComputeLoop2493 ]
  %3266 = trunc i64 %744 to i32
  %3267 = lshr i64 %744, 32
  %3268 = trunc i64 %3267 to i32
  %3269 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %3266, i32 0)
  %3270 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %3268, i32 %3269)
  %3271 = icmp eq i32 %3270, 0
  %3272 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3271)
  %3273 = extractvalue { i1, i64 } %3272, 0
  %3274 = extractvalue { i1, i64 } %3272, 1
  br i1 %3273, label %745, label %747

ComputeLoop2497:                                  ; preds = %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit530, %ComputeLoop2497
  %Accumulator2499 = phi i32 [ %3278, %ComputeLoop2497 ], [ -1, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit530 ]
  %ActiveBits2500 = phi i64 [ %3281, %ComputeLoop2497 ], [ %824, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit530 ]
  %3275 = call i64 @llvm.cttz.i64(i64 %ActiveBits2500, i1 true)
  %3276 = trunc i64 %3275 to i32
  %3277 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i21.i.i521, i32 %3276)
  %3278 = and i32 %Accumulator2499, %3277
  %3279 = shl i64 1, %3275
  %3280 = xor i64 %3279, -1
  %3281 = and i64 %ActiveBits2500, %3280
  %3282 = icmp eq i64 %3281, 0
  br i1 %3282, label %ComputeEnd2498, label %ComputeLoop2497, !amdgpu.uniform !42

ComputeEnd2498:                                   ; preds = %ComputeLoop2497
  %.lcssa4160 = phi i32 [ %3278, %ComputeLoop2497 ]
  %3283 = trunc i64 %824 to i32
  %3284 = lshr i64 %824, 32
  %3285 = trunc i64 %3284 to i32
  %3286 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %3283, i32 0)
  %3287 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %3285, i32 %3286)
  %3288 = icmp eq i32 %3287, 0
  %3289 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3288)
  %3290 = extractvalue { i1, i64 } %3289, 0
  %3291 = extractvalue { i1, i64 } %3289, 1
  br i1 %3290, label %825, label %828

ComputeLoop2501:                                  ; preds = %for.cond.i.i1.i539, %ComputeLoop2501
  %Accumulator2503 = phi i32 [ %3296, %ComputeLoop2501 ], [ 0, %for.cond.i.i1.i539 ]
  %OldValuePhi2504 = phi i32 [ %3295, %ComputeLoop2501 ], [ poison, %for.cond.i.i1.i539 ]
  %ActiveBits2505 = phi i64 [ %3299, %ComputeLoop2501 ], [ %841, %for.cond.i.i1.i539 ]
  %3292 = call i64 @llvm.cttz.i64(i64 %ActiveBits2505, i1 true)
  %3293 = trunc i64 %3292 to i32
  %3294 = call i32 @llvm.amdgcn.readlane.i32(i32 %shl.i.i.i.i.i551, i32 %3293)
  %3295 = call i32 @llvm.amdgcn.writelane.i32(i32 %Accumulator2503, i32 %3293, i32 %OldValuePhi2504)
  %3296 = or i32 %Accumulator2503, %3294
  %3297 = shl i64 1, %3292
  %3298 = xor i64 %3297, -1
  %3299 = and i64 %ActiveBits2505, %3298
  %3300 = icmp eq i64 %3299, 0
  br i1 %3300, label %ComputeEnd2502, label %ComputeLoop2501, !amdgpu.uniform !42

ComputeEnd2502:                                   ; preds = %ComputeLoop2501
  %.lcssa4154 = phi i32 [ %3295, %ComputeLoop2501 ]
  %.lcssa4153 = phi i32 [ %3296, %ComputeLoop2501 ]
  %div8.i.i.i.i.i547 = lshr i32 %spec.store.select.i.i.i546, 5
  %idxprom.i.i.i.i.i548 = zext nneg i32 %div8.i.i.i.i.i547 to i64
  %3301 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i.i.i548
  %3302 = trunc i64 %841 to i32
  %3303 = lshr i64 %841, 32
  %3304 = trunc i64 %3303 to i32
  %3305 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %3302, i32 0)
  %3306 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %3304, i32 %3305)
  %3307 = icmp eq i32 %3306, 0
  %3308 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3307)
  %3309 = extractvalue { i1, i64 } %3308, 0
  %3310 = extractvalue { i1, i64 } %3308, 1
  br i1 %3309, label %842, label %844

Flow3759:                                         ; preds = %862, %if.then.i.i.i.i557
  %3311 = phi i1 [ false, %862 ], [ true, %if.then.i.i.i.i557 ]
  br i1 %3311, label %if.end14.i.i.i628, label %Flow3760, !amdgpu.uniform !42

ComputeLoop2506:                                  ; preds = %if.then12.i.i.i562, %ComputeLoop2506
  %Accumulator2508 = phi i32 [ %3315, %ComputeLoop2506 ], [ -1, %if.then12.i.i.i562 ]
  %ActiveBits2509 = phi i64 [ %3318, %ComputeLoop2506 ], [ %859, %if.then12.i.i.i562 ]
  %3312 = call i64 @llvm.cttz.i64(i64 %ActiveBits2509, i1 true)
  %3313 = trunc i64 %3312 to i32
  %3314 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i.i.i568, i32 %3313)
  %3315 = and i32 %Accumulator2508, %3314
  %3316 = shl i64 1, %3312
  %3317 = xor i64 %3316, -1
  %3318 = and i64 %ActiveBits2509, %3317
  %3319 = icmp eq i64 %3318, 0
  br i1 %3319, label %ComputeEnd2507, label %ComputeLoop2506, !amdgpu.uniform !42

ComputeEnd2507:                                   ; preds = %ComputeLoop2506
  %.lcssa4155 = phi i32 [ %3315, %ComputeLoop2506 ]
  %3320 = trunc i64 %859 to i32
  %3321 = lshr i64 %859, 32
  %3322 = trunc i64 %3321 to i32
  %3323 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %3320, i32 0)
  %3324 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %3322, i32 %3323)
  %3325 = icmp eq i32 %3324, 0
  %3326 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3325)
  %3327 = extractvalue { i1, i64 } %3326, 0
  %3328 = extractvalue { i1, i64 } %3326, 1
  br i1 %3327, label %860, label %862

ComputeLoop2510:                                  ; preds = %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit636, %ComputeLoop2510
  %Accumulator2512 = phi i32 [ %3332, %ComputeLoop2510 ], [ -1, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit636 ]
  %ActiveBits2513 = phi i64 [ %3335, %ComputeLoop2510 ], [ %957, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit636 ]
  %3329 = call i64 @llvm.cttz.i64(i64 %ActiveBits2513, i1 true)
  %3330 = trunc i64 %3329 to i32
  %3331 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i21.i.i627, i32 %3330)
  %3332 = and i32 %Accumulator2512, %3331
  %3333 = shl i64 1, %3329
  %3334 = xor i64 %3333, -1
  %3335 = and i64 %ActiveBits2513, %3334
  %3336 = icmp eq i64 %3335, 0
  br i1 %3336, label %ComputeEnd2511, label %ComputeLoop2510, !amdgpu.uniform !42

ComputeEnd2511:                                   ; preds = %ComputeLoop2510
  %.lcssa4145 = phi i32 [ %3332, %ComputeLoop2510 ]
  %3337 = trunc i64 %957 to i32
  %3338 = lshr i64 %957, 32
  %3339 = trunc i64 %3338 to i32
  %3340 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %3337, i32 0)
  %3341 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %3339, i32 %3340)
  %3342 = icmp eq i32 %3341, 0
  %3343 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3342)
  %3344 = extractvalue { i1, i64 } %3343, 0
  %3345 = extractvalue { i1, i64 } %3343, 1
  br i1 %3344, label %958, label %961

ComputeLoop2514:                                  ; preds = %for.cond.i.i.i638, %ComputeLoop2514
  %Accumulator2516 = phi i32 [ %3350, %ComputeLoop2514 ], [ 0, %for.cond.i.i.i638 ]
  %OldValuePhi2517 = phi i32 [ %3349, %ComputeLoop2514 ], [ poison, %for.cond.i.i.i638 ]
  %ActiveBits2518 = phi i64 [ %3353, %ComputeLoop2514 ], [ %970, %for.cond.i.i.i638 ]
  %3346 = call i64 @llvm.cttz.i64(i64 %ActiveBits2518, i1 true)
  %3347 = trunc i64 %3346 to i32
  %3348 = call i32 @llvm.amdgcn.readlane.i32(i32 %shl.i.i.i.i.i650, i32 %3347)
  %3349 = call i32 @llvm.amdgcn.writelane.i32(i32 %Accumulator2516, i32 %3347, i32 %OldValuePhi2517)
  %3350 = or i32 %Accumulator2516, %3348
  %3351 = shl i64 1, %3346
  %3352 = xor i64 %3351, -1
  %3353 = and i64 %ActiveBits2518, %3352
  %3354 = icmp eq i64 %3353, 0
  br i1 %3354, label %ComputeEnd2515, label %ComputeLoop2514, !amdgpu.uniform !42

ComputeEnd2515:                                   ; preds = %ComputeLoop2514
  %.lcssa4140 = phi i32 [ %3349, %ComputeLoop2514 ]
  %.lcssa4139 = phi i32 [ %3350, %ComputeLoop2514 ]
  %div8.i.i.i.i.i646 = lshr i32 %spec.store.select.i.i.i645, 5
  %idxprom.i.i.i.i.i647 = zext nneg i32 %div8.i.i.i.i.i646 to i64
  %3355 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i.i.i647
  %3356 = trunc i64 %970 to i32
  %3357 = lshr i64 %970, 32
  %3358 = trunc i64 %3357 to i32
  %3359 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %3356, i32 0)
  %3360 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %3358, i32 %3359)
  %3361 = icmp eq i32 %3360, 0
  %3362 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3361)
  %3363 = extractvalue { i1, i64 } %3362, 0
  %3364 = extractvalue { i1, i64 } %3362, 1
  br i1 %3363, label %971, label %973

Flow3740:                                         ; preds = %991, %if.then.i.i.i.i656
  %3365 = phi i1 [ false, %991 ], [ true, %if.then.i.i.i.i656 ]
  br i1 %3365, label %if.end14.i.i.i725, label %Flow3741, !amdgpu.uniform !42

ComputeLoop2519:                                  ; preds = %if.then12.i.i.i661, %ComputeLoop2519
  %Accumulator2521 = phi i32 [ %3369, %ComputeLoop2519 ], [ -1, %if.then12.i.i.i661 ]
  %ActiveBits2522 = phi i64 [ %3372, %ComputeLoop2519 ], [ %988, %if.then12.i.i.i661 ]
  %3366 = call i64 @llvm.cttz.i64(i64 %ActiveBits2522, i1 true)
  %3367 = trunc i64 %3366 to i32
  %3368 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i.i.i667, i32 %3367)
  %3369 = and i32 %Accumulator2521, %3368
  %3370 = shl i64 1, %3366
  %3371 = xor i64 %3370, -1
  %3372 = and i64 %ActiveBits2522, %3371
  %3373 = icmp eq i64 %3372, 0
  br i1 %3373, label %ComputeEnd2520, label %ComputeLoop2519, !amdgpu.uniform !42

ComputeEnd2520:                                   ; preds = %ComputeLoop2519
  %.lcssa4141 = phi i32 [ %3369, %ComputeLoop2519 ]
  %3374 = trunc i64 %988 to i32
  %3375 = lshr i64 %988, 32
  %3376 = trunc i64 %3375 to i32
  %3377 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %3374, i32 0)
  %3378 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %3376, i32 %3377)
  %3379 = icmp eq i32 %3378, 0
  %3380 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3379)
  %3381 = extractvalue { i1, i64 } %3380, 0
  %3382 = extractvalue { i1, i64 } %3380, 1
  br i1 %3381, label %989, label %991

ComputeLoop2523:                                  ; preds = %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIcEERS1_T_.exit, %ComputeLoop2523
  %Accumulator2525 = phi i32 [ %3386, %ComputeLoop2523 ], [ -1, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIcEERS1_T_.exit ]
  %ActiveBits2526 = phi i64 [ %3389, %ComputeLoop2523 ], [ %1043, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIcEERS1_T_.exit ]
  %3383 = call i64 @llvm.cttz.i64(i64 %ActiveBits2526, i1 true)
  %3384 = trunc i64 %3383 to i32
  %3385 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i21.i.i724, i32 %3384)
  %3386 = and i32 %Accumulator2525, %3385
  %3387 = shl i64 1, %3383
  %3388 = xor i64 %3387, -1
  %3389 = and i64 %ActiveBits2526, %3388
  %3390 = icmp eq i64 %3389, 0
  br i1 %3390, label %ComputeEnd2524, label %ComputeLoop2523, !amdgpu.uniform !42

ComputeEnd2524:                                   ; preds = %ComputeLoop2523
  %.lcssa4135 = phi i32 [ %3386, %ComputeLoop2523 ]
  %3391 = trunc i64 %1043 to i32
  %3392 = lshr i64 %1043, 32
  %3393 = trunc i64 %3392 to i32
  %3394 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %3391, i32 0)
  %3395 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %3393, i32 %3394)
  %3396 = icmp eq i32 %3395, 0
  %3397 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3396)
  %3398 = extractvalue { i1, i64 } %3397, 0
  %3399 = extractvalue { i1, i64 } %3397, 1
  br i1 %3398, label %1044, label %1047

ComputeLoop2527:                                  ; preds = %for.cond.i.i.i741, %ComputeLoop2527
  %Accumulator2529 = phi i32 [ %3404, %ComputeLoop2527 ], [ 0, %for.cond.i.i.i741 ]
  %OldValuePhi2530 = phi i32 [ %3403, %ComputeLoop2527 ], [ poison, %for.cond.i.i.i741 ]
  %ActiveBits2531 = phi i64 [ %3407, %ComputeLoop2527 ], [ %1065, %for.cond.i.i.i741 ]
  %3400 = call i64 @llvm.cttz.i64(i64 %ActiveBits2531, i1 true)
  %3401 = trunc i64 %3400 to i32
  %3402 = call i32 @llvm.amdgcn.readlane.i32(i32 %shl.i.i.i.i.i753, i32 %3401)
  %3403 = call i32 @llvm.amdgcn.writelane.i32(i32 %Accumulator2529, i32 %3401, i32 %OldValuePhi2530)
  %3404 = or i32 %Accumulator2529, %3402
  %3405 = shl i64 1, %3400
  %3406 = xor i64 %3405, -1
  %3407 = and i64 %ActiveBits2531, %3406
  %3408 = icmp eq i64 %3407, 0
  br i1 %3408, label %ComputeEnd2528, label %ComputeLoop2527, !amdgpu.uniform !42

ComputeEnd2528:                                   ; preds = %ComputeLoop2527
  %.lcssa4130 = phi i32 [ %3403, %ComputeLoop2527 ]
  %.lcssa4129 = phi i32 [ %3404, %ComputeLoop2527 ]
  %div8.i.i.i.i.i749 = lshr i32 %spec.store.select.i.i.i748, 5
  %idxprom.i.i.i.i.i750 = zext nneg i32 %div8.i.i.i.i.i749 to i64
  %3409 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i.i.i750
  %3410 = trunc i64 %1065 to i32
  %3411 = lshr i64 %1065, 32
  %3412 = trunc i64 %3411 to i32
  %3413 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %3410, i32 0)
  %3414 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %3412, i32 %3413)
  %3415 = icmp eq i32 %3414, 0
  %3416 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3415)
  %3417 = extractvalue { i1, i64 } %3416, 0
  %3418 = extractvalue { i1, i64 } %3416, 1
  br i1 %3417, label %1066, label %1068

Flow3727:                                         ; preds = %1086, %if.then.i.i.i.i759
  %3419 = phi i1 [ false, %1086 ], [ true, %if.then.i.i.i.i759 ]
  br i1 %3419, label %if.end14.i.i.i830, label %Flow3728, !amdgpu.uniform !42

ComputeLoop2532:                                  ; preds = %if.then12.i.i.i764, %ComputeLoop2532
  %Accumulator2534 = phi i32 [ %3423, %ComputeLoop2532 ], [ -1, %if.then12.i.i.i764 ]
  %ActiveBits2535 = phi i64 [ %3426, %ComputeLoop2532 ], [ %1083, %if.then12.i.i.i764 ]
  %3420 = call i64 @llvm.cttz.i64(i64 %ActiveBits2535, i1 true)
  %3421 = trunc i64 %3420 to i32
  %3422 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i.i.i770, i32 %3421)
  %3423 = and i32 %Accumulator2534, %3422
  %3424 = shl i64 1, %3420
  %3425 = xor i64 %3424, -1
  %3426 = and i64 %ActiveBits2535, %3425
  %3427 = icmp eq i64 %3426, 0
  br i1 %3427, label %ComputeEnd2533, label %ComputeLoop2532, !amdgpu.uniform !42

ComputeEnd2533:                                   ; preds = %ComputeLoop2532
  %.lcssa4131 = phi i32 [ %3423, %ComputeLoop2532 ]
  %3428 = trunc i64 %1083 to i32
  %3429 = lshr i64 %1083, 32
  %3430 = trunc i64 %3429 to i32
  %3431 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %3428, i32 0)
  %3432 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %3430, i32 %3431)
  %3433 = icmp eq i32 %3432, 0
  %3434 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3433)
  %3435 = extractvalue { i1, i64 } %3434, 0
  %3436 = extractvalue { i1, i64 } %3434, 1
  br i1 %3435, label %1084, label %1086

ComputeLoop2536:                                  ; preds = %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsINS_3cpp6stringEEERS1_T_.exit838, %ComputeLoop2536
  %Accumulator2538 = phi i32 [ %3440, %ComputeLoop2536 ], [ -1, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsINS_3cpp6stringEEERS1_T_.exit838 ]
  %ActiveBits2539 = phi i64 [ %3443, %ComputeLoop2536 ], [ %1160, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsINS_3cpp6stringEEERS1_T_.exit838 ]
  %3437 = call i64 @llvm.cttz.i64(i64 %ActiveBits2539, i1 true)
  %3438 = trunc i64 %3437 to i32
  %3439 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i21.i.i829, i32 %3438)
  %3440 = and i32 %Accumulator2538, %3439
  %3441 = shl i64 1, %3437
  %3442 = xor i64 %3441, -1
  %3443 = and i64 %ActiveBits2539, %3442
  %3444 = icmp eq i64 %3443, 0
  br i1 %3444, label %ComputeEnd2537, label %ComputeLoop2536, !amdgpu.uniform !42

ComputeEnd2537:                                   ; preds = %ComputeLoop2536
  %.lcssa4125 = phi i32 [ %3440, %ComputeLoop2536 ]
  %3445 = trunc i64 %1160 to i32
  %3446 = lshr i64 %1160, 32
  %3447 = trunc i64 %3446 to i32
  %3448 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %3445, i32 0)
  %3449 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %3447, i32 %3448)
  %3450 = icmp eq i32 %3449, 0
  %3451 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3450)
  %3452 = extractvalue { i1, i64 } %3451, 0
  %3453 = extractvalue { i1, i64 } %3451, 1
  br i1 %3452, label %1161, label %1164

ComputeLoop2540:                                  ; preds = %for.cond.i.i1.i847, %ComputeLoop2540
  %Accumulator2542 = phi i32 [ %3458, %ComputeLoop2540 ], [ 0, %for.cond.i.i1.i847 ]
  %OldValuePhi2543 = phi i32 [ %3457, %ComputeLoop2540 ], [ poison, %for.cond.i.i1.i847 ]
  %ActiveBits2544 = phi i64 [ %3461, %ComputeLoop2540 ], [ %1173, %for.cond.i.i1.i847 ]
  %3454 = call i64 @llvm.cttz.i64(i64 %ActiveBits2544, i1 true)
  %3455 = trunc i64 %3454 to i32
  %3456 = call i32 @llvm.amdgcn.readlane.i32(i32 %shl.i.i.i.i.i859, i32 %3455)
  %3457 = call i32 @llvm.amdgcn.writelane.i32(i32 %Accumulator2542, i32 %3455, i32 %OldValuePhi2543)
  %3458 = or i32 %Accumulator2542, %3456
  %3459 = shl i64 1, %3454
  %3460 = xor i64 %3459, -1
  %3461 = and i64 %ActiveBits2544, %3460
  %3462 = icmp eq i64 %3461, 0
  br i1 %3462, label %ComputeEnd2541, label %ComputeLoop2540, !amdgpu.uniform !42

ComputeEnd2541:                                   ; preds = %ComputeLoop2540
  %.lcssa4120 = phi i32 [ %3457, %ComputeLoop2540 ]
  %.lcssa4119 = phi i32 [ %3458, %ComputeLoop2540 ]
  %div8.i.i.i.i.i855 = lshr i32 %spec.store.select.i.i.i854, 5
  %idxprom.i.i.i.i.i856 = zext nneg i32 %div8.i.i.i.i.i855 to i64
  %3463 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i.i.i856
  %3464 = trunc i64 %1173 to i32
  %3465 = lshr i64 %1173, 32
  %3466 = trunc i64 %3465 to i32
  %3467 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %3464, i32 0)
  %3468 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %3466, i32 %3467)
  %3469 = icmp eq i32 %3468, 0
  %3470 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3469)
  %3471 = extractvalue { i1, i64 } %3470, 0
  %3472 = extractvalue { i1, i64 } %3470, 1
  br i1 %3471, label %1174, label %1176

Flow3708:                                         ; preds = %1194, %if.then.i.i.i.i865
  %3473 = phi i1 [ false, %1194 ], [ true, %if.then.i.i.i.i865 ]
  br i1 %3473, label %if.end14.i.i.i936, label %Flow3709, !amdgpu.uniform !42

ComputeLoop2545:                                  ; preds = %if.then12.i.i.i870, %ComputeLoop2545
  %Accumulator2547 = phi i32 [ %3477, %ComputeLoop2545 ], [ -1, %if.then12.i.i.i870 ]
  %ActiveBits2548 = phi i64 [ %3480, %ComputeLoop2545 ], [ %1191, %if.then12.i.i.i870 ]
  %3474 = call i64 @llvm.cttz.i64(i64 %ActiveBits2548, i1 true)
  %3475 = trunc i64 %3474 to i32
  %3476 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i.i.i876, i32 %3475)
  %3477 = and i32 %Accumulator2547, %3476
  %3478 = shl i64 1, %3474
  %3479 = xor i64 %3478, -1
  %3480 = and i64 %ActiveBits2548, %3479
  %3481 = icmp eq i64 %3480, 0
  br i1 %3481, label %ComputeEnd2546, label %ComputeLoop2545, !amdgpu.uniform !42

ComputeEnd2546:                                   ; preds = %ComputeLoop2545
  %.lcssa4121 = phi i32 [ %3477, %ComputeLoop2545 ]
  %3482 = trunc i64 %1191 to i32
  %3483 = lshr i64 %1191, 32
  %3484 = trunc i64 %3483 to i32
  %3485 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %3482, i32 0)
  %3486 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %3484, i32 %3485)
  %3487 = icmp eq i32 %3486, 0
  %3488 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3487)
  %3489 = extractvalue { i1, i64 } %3488, 0
  %3490 = extractvalue { i1, i64 } %3488, 1
  br i1 %3489, label %1192, label %1194

ComputeLoop2549:                                  ; preds = %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit944, %ComputeLoop2549
  %Accumulator2551 = phi i32 [ %3494, %ComputeLoop2549 ], [ -1, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit944 ]
  %ActiveBits2552 = phi i64 [ %3497, %ComputeLoop2549 ], [ %1271, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit944 ]
  %3491 = call i64 @llvm.cttz.i64(i64 %ActiveBits2552, i1 true)
  %3492 = trunc i64 %3491 to i32
  %3493 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i21.i.i935, i32 %3492)
  %3494 = and i32 %Accumulator2551, %3493
  %3495 = shl i64 1, %3491
  %3496 = xor i64 %3495, -1
  %3497 = and i64 %ActiveBits2552, %3496
  %3498 = icmp eq i64 %3497, 0
  br i1 %3498, label %ComputeEnd2550, label %ComputeLoop2549, !amdgpu.uniform !42

ComputeEnd2550:                                   ; preds = %ComputeLoop2549
  %.lcssa4115 = phi i32 [ %3494, %ComputeLoop2549 ]
  %3499 = trunc i64 %1271 to i32
  %3500 = lshr i64 %1271, 32
  %3501 = trunc i64 %3500 to i32
  %3502 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %3499, i32 0)
  %3503 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %3501, i32 %3502)
  %3504 = icmp eq i32 %3503, 0
  %3505 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3504)
  %3506 = extractvalue { i1, i64 } %3505, 0
  %3507 = extractvalue { i1, i64 } %3505, 1
  br i1 %3506, label %1272, label %1275

ComputeLoop2553:                                  ; preds = %for.cond.i.i.i953, %ComputeLoop2553
  %Accumulator2555 = phi i32 [ %3512, %ComputeLoop2553 ], [ 0, %for.cond.i.i.i953 ]
  %OldValuePhi2556 = phi i32 [ %3511, %ComputeLoop2553 ], [ poison, %for.cond.i.i.i953 ]
  %ActiveBits2557 = phi i64 [ %3515, %ComputeLoop2553 ], [ %1412, %for.cond.i.i.i953 ]
  %3508 = call i64 @llvm.cttz.i64(i64 %ActiveBits2557, i1 true)
  %3509 = trunc i64 %3508 to i32
  %3510 = call i32 @llvm.amdgcn.readlane.i32(i32 %shl.i.i.i.i.i965, i32 %3509)
  %3511 = call i32 @llvm.amdgcn.writelane.i32(i32 %Accumulator2555, i32 %3509, i32 %OldValuePhi2556)
  %3512 = or i32 %Accumulator2555, %3510
  %3513 = shl i64 1, %3508
  %3514 = xor i64 %3513, -1
  %3515 = and i64 %ActiveBits2557, %3514
  %3516 = icmp eq i64 %3515, 0
  br i1 %3516, label %ComputeEnd2554, label %ComputeLoop2553, !amdgpu.uniform !42

ComputeEnd2554:                                   ; preds = %ComputeLoop2553
  %.lcssa4106 = phi i32 [ %3511, %ComputeLoop2553 ]
  %.lcssa4105 = phi i32 [ %3512, %ComputeLoop2553 ]
  %div8.i.i.i.i.i961 = lshr i32 %spec.store.select.i.i.i960, 5
  %idxprom.i.i.i.i.i962 = zext nneg i32 %div8.i.i.i.i.i961 to i64
  %3517 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i.i.i962
  %3518 = trunc i64 %1412 to i32
  %3519 = lshr i64 %1412, 32
  %3520 = trunc i64 %3519 to i32
  %3521 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %3518, i32 0)
  %3522 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %3520, i32 %3521)
  %3523 = icmp eq i32 %3522, 0
  %3524 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3523)
  %3525 = extractvalue { i1, i64 } %3524, 0
  %3526 = extractvalue { i1, i64 } %3524, 1
  br i1 %3525, label %1413, label %1415

Flow3655:                                         ; preds = %1433, %if.then.i.i.i.i971
  %3527 = phi i1 [ false, %1433 ], [ true, %if.then.i.i.i.i971 ]
  br i1 %3527, label %if.end14.i.i.i1042, label %Flow3656, !amdgpu.uniform !42

ComputeLoop2558:                                  ; preds = %if.then12.i.i.i976, %ComputeLoop2558
  %Accumulator2560 = phi i32 [ %3531, %ComputeLoop2558 ], [ -1, %if.then12.i.i.i976 ]
  %ActiveBits2561 = phi i64 [ %3534, %ComputeLoop2558 ], [ %1430, %if.then12.i.i.i976 ]
  %3528 = call i64 @llvm.cttz.i64(i64 %ActiveBits2561, i1 true)
  %3529 = trunc i64 %3528 to i32
  %3530 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i.i.i982, i32 %3529)
  %3531 = and i32 %Accumulator2560, %3530
  %3532 = shl i64 1, %3528
  %3533 = xor i64 %3532, -1
  %3534 = and i64 %ActiveBits2561, %3533
  %3535 = icmp eq i64 %3534, 0
  br i1 %3535, label %ComputeEnd2559, label %ComputeLoop2558, !amdgpu.uniform !42

ComputeEnd2559:                                   ; preds = %ComputeLoop2558
  %.lcssa4107 = phi i32 [ %3531, %ComputeLoop2558 ]
  %3536 = trunc i64 %1430 to i32
  %3537 = lshr i64 %1430, 32
  %3538 = trunc i64 %3537 to i32
  %3539 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %3536, i32 0)
  %3540 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %3538, i32 %3539)
  %3541 = icmp eq i32 %3540, 0
  %3542 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3541)
  %3543 = extractvalue { i1, i64 } %3542, 0
  %3544 = extractvalue { i1, i64 } %3542, 1
  br i1 %3543, label %1431, label %1433

ComputeLoop2562:                                  ; preds = %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsINS_3cpp6stringEEERS1_T_.exit1050, %ComputeLoop2562
  %Accumulator2564 = phi i32 [ %3548, %ComputeLoop2562 ], [ -1, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsINS_3cpp6stringEEERS1_T_.exit1050 ]
  %ActiveBits2565 = phi i64 [ %3551, %ComputeLoop2562 ], [ %1528, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsINS_3cpp6stringEEERS1_T_.exit1050 ]
  %3545 = call i64 @llvm.cttz.i64(i64 %ActiveBits2565, i1 true)
  %3546 = trunc i64 %3545 to i32
  %3547 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i21.i.i1041, i32 %3546)
  %3548 = and i32 %Accumulator2564, %3547
  %3549 = shl i64 1, %3545
  %3550 = xor i64 %3549, -1
  %3551 = and i64 %ActiveBits2565, %3550
  %3552 = icmp eq i64 %3551, 0
  br i1 %3552, label %ComputeEnd2563, label %ComputeLoop2562, !amdgpu.uniform !42

ComputeEnd2563:                                   ; preds = %ComputeLoop2562
  %.lcssa4097 = phi i32 [ %3548, %ComputeLoop2562 ]
  %3553 = trunc i64 %1528 to i32
  %3554 = lshr i64 %1528, 32
  %3555 = trunc i64 %3554 to i32
  %3556 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %3553, i32 0)
  %3557 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %3555, i32 %3556)
  %3558 = icmp eq i32 %3557, 0
  %3559 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3558)
  %3560 = extractvalue { i1, i64 } %3559, 0
  %3561 = extractvalue { i1, i64 } %3559, 1
  br i1 %3560, label %1529, label %1532

ComputeLoop2566:                                  ; preds = %for.cond.i.i.i1054, %ComputeLoop2566
  %Accumulator2568 = phi i32 [ %3566, %ComputeLoop2566 ], [ 0, %for.cond.i.i.i1054 ]
  %OldValuePhi2569 = phi i32 [ %3565, %ComputeLoop2566 ], [ poison, %for.cond.i.i.i1054 ]
  %ActiveBits2570 = phi i64 [ %3569, %ComputeLoop2566 ], [ %1541, %for.cond.i.i.i1054 ]
  %3562 = call i64 @llvm.cttz.i64(i64 %ActiveBits2570, i1 true)
  %3563 = trunc i64 %3562 to i32
  %3564 = call i32 @llvm.amdgcn.readlane.i32(i32 %shl.i.i.i.i.i1066, i32 %3563)
  %3565 = call i32 @llvm.amdgcn.writelane.i32(i32 %Accumulator2568, i32 %3563, i32 %OldValuePhi2569)
  %3566 = or i32 %Accumulator2568, %3564
  %3567 = shl i64 1, %3562
  %3568 = xor i64 %3567, -1
  %3569 = and i64 %ActiveBits2570, %3568
  %3570 = icmp eq i64 %3569, 0
  br i1 %3570, label %ComputeEnd2567, label %ComputeLoop2566, !amdgpu.uniform !42

ComputeEnd2567:                                   ; preds = %ComputeLoop2566
  %.lcssa4092 = phi i32 [ %3565, %ComputeLoop2566 ]
  %.lcssa4091 = phi i32 [ %3566, %ComputeLoop2566 ]
  %div8.i.i.i.i.i1062 = lshr i32 %spec.store.select.i.i.i1061, 5
  %idxprom.i.i.i.i.i1063 = zext nneg i32 %div8.i.i.i.i.i1062 to i64
  %3571 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i.i.i1063
  %3572 = trunc i64 %1541 to i32
  %3573 = lshr i64 %1541, 32
  %3574 = trunc i64 %3573 to i32
  %3575 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %3572, i32 0)
  %3576 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %3574, i32 %3575)
  %3577 = icmp eq i32 %3576, 0
  %3578 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3577)
  %3579 = extractvalue { i1, i64 } %3578, 0
  %3580 = extractvalue { i1, i64 } %3578, 1
  br i1 %3579, label %1542, label %1544

Flow3629:                                         ; preds = %1562, %if.then.i.i.i.i1072
  %3581 = phi i1 [ false, %1562 ], [ true, %if.then.i.i.i.i1072 ]
  br i1 %3581, label %if.end14.i.i.i1141, label %Flow3630, !amdgpu.uniform !42

ComputeLoop2571:                                  ; preds = %if.then12.i.i.i1077, %ComputeLoop2571
  %Accumulator2573 = phi i32 [ %3585, %ComputeLoop2571 ], [ -1, %if.then12.i.i.i1077 ]
  %ActiveBits2574 = phi i64 [ %3588, %ComputeLoop2571 ], [ %1559, %if.then12.i.i.i1077 ]
  %3582 = call i64 @llvm.cttz.i64(i64 %ActiveBits2574, i1 true)
  %3583 = trunc i64 %3582 to i32
  %3584 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i.i.i1083, i32 %3583)
  %3585 = and i32 %Accumulator2573, %3584
  %3586 = shl i64 1, %3582
  %3587 = xor i64 %3586, -1
  %3588 = and i64 %ActiveBits2574, %3587
  %3589 = icmp eq i64 %3588, 0
  br i1 %3589, label %ComputeEnd2572, label %ComputeLoop2571, !amdgpu.uniform !42

ComputeEnd2572:                                   ; preds = %ComputeLoop2571
  %.lcssa4093 = phi i32 [ %3585, %ComputeLoop2571 ]
  %3590 = trunc i64 %1559 to i32
  %3591 = lshr i64 %1559, 32
  %3592 = trunc i64 %3591 to i32
  %3593 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %3590, i32 0)
  %3594 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %3592, i32 %3593)
  %3595 = icmp eq i32 %3594, 0
  %3596 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3595)
  %3597 = extractvalue { i1, i64 } %3596, 0
  %3598 = extractvalue { i1, i64 } %3596, 1
  br i1 %3597, label %1560, label %1562

ComputeLoop2575:                                  ; preds = %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIcEERS1_T_.exit1149, %ComputeLoop2575
  %Accumulator2577 = phi i32 [ %3602, %ComputeLoop2575 ], [ -1, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIcEERS1_T_.exit1149 ]
  %ActiveBits2578 = phi i64 [ %3605, %ComputeLoop2575 ], [ %1614, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIcEERS1_T_.exit1149 ]
  %3599 = call i64 @llvm.cttz.i64(i64 %ActiveBits2578, i1 true)
  %3600 = trunc i64 %3599 to i32
  %3601 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i21.i.i1140, i32 %3600)
  %3602 = and i32 %Accumulator2577, %3601
  %3603 = shl i64 1, %3599
  %3604 = xor i64 %3603, -1
  %3605 = and i64 %ActiveBits2578, %3604
  %3606 = icmp eq i64 %3605, 0
  br i1 %3606, label %ComputeEnd2576, label %ComputeLoop2575, !amdgpu.uniform !42

ComputeEnd2576:                                   ; preds = %ComputeLoop2575
  %.lcssa4087 = phi i32 [ %3602, %ComputeLoop2575 ]
  %3607 = trunc i64 %1614 to i32
  %3608 = lshr i64 %1614, 32
  %3609 = trunc i64 %3608 to i32
  %3610 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %3607, i32 0)
  %3611 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %3609, i32 %3610)
  %3612 = icmp eq i32 %3611, 0
  %3613 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3612)
  %3614 = extractvalue { i1, i64 } %3613, 0
  %3615 = extractvalue { i1, i64 } %3613, 1
  br i1 %3614, label %1615, label %1618

ComputeLoop2579:                                  ; preds = %for.cond.i.i1.i1158, %ComputeLoop2579
  %Accumulator2581 = phi i32 [ %3620, %ComputeLoop2579 ], [ 0, %for.cond.i.i1.i1158 ]
  %OldValuePhi2582 = phi i32 [ %3619, %ComputeLoop2579 ], [ poison, %for.cond.i.i1.i1158 ]
  %ActiveBits2583 = phi i64 [ %3623, %ComputeLoop2579 ], [ %1627, %for.cond.i.i1.i1158 ]
  %3616 = call i64 @llvm.cttz.i64(i64 %ActiveBits2583, i1 true)
  %3617 = trunc i64 %3616 to i32
  %3618 = call i32 @llvm.amdgcn.readlane.i32(i32 %shl.i.i.i.i.i1170, i32 %3617)
  %3619 = call i32 @llvm.amdgcn.writelane.i32(i32 %Accumulator2581, i32 %3617, i32 %OldValuePhi2582)
  %3620 = or i32 %Accumulator2581, %3618
  %3621 = shl i64 1, %3616
  %3622 = xor i64 %3621, -1
  %3623 = and i64 %ActiveBits2583, %3622
  %3624 = icmp eq i64 %3623, 0
  br i1 %3624, label %ComputeEnd2580, label %ComputeLoop2579, !amdgpu.uniform !42

ComputeEnd2580:                                   ; preds = %ComputeLoop2579
  %.lcssa4082 = phi i32 [ %3619, %ComputeLoop2579 ]
  %.lcssa4081 = phi i32 [ %3620, %ComputeLoop2579 ]
  %div8.i.i.i.i.i1166 = lshr i32 %spec.store.select.i.i.i1165, 5
  %idxprom.i.i.i.i.i1167 = zext nneg i32 %div8.i.i.i.i.i1166 to i64
  %3625 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i.i.i1167
  %3626 = trunc i64 %1627 to i32
  %3627 = lshr i64 %1627, 32
  %3628 = trunc i64 %3627 to i32
  %3629 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %3626, i32 0)
  %3630 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %3628, i32 %3629)
  %3631 = icmp eq i32 %3630, 0
  %3632 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3631)
  %3633 = extractvalue { i1, i64 } %3632, 0
  %3634 = extractvalue { i1, i64 } %3632, 1
  br i1 %3633, label %1628, label %1630

Flow3618:                                         ; preds = %1648, %if.then.i.i.i.i1176
  %3635 = phi i1 [ false, %1648 ], [ true, %if.then.i.i.i.i1176 ]
  br i1 %3635, label %if.end14.i.i.i1247, label %Flow3619, !amdgpu.uniform !42

ComputeLoop2584:                                  ; preds = %if.then12.i.i.i1181, %ComputeLoop2584
  %Accumulator2586 = phi i32 [ %3639, %ComputeLoop2584 ], [ -1, %if.then12.i.i.i1181 ]
  %ActiveBits2587 = phi i64 [ %3642, %ComputeLoop2584 ], [ %1645, %if.then12.i.i.i1181 ]
  %3636 = call i64 @llvm.cttz.i64(i64 %ActiveBits2587, i1 true)
  %3637 = trunc i64 %3636 to i32
  %3638 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i.i.i1187, i32 %3637)
  %3639 = and i32 %Accumulator2586, %3638
  %3640 = shl i64 1, %3636
  %3641 = xor i64 %3640, -1
  %3642 = and i64 %ActiveBits2587, %3641
  %3643 = icmp eq i64 %3642, 0
  br i1 %3643, label %ComputeEnd2585, label %ComputeLoop2584, !amdgpu.uniform !42

ComputeEnd2585:                                   ; preds = %ComputeLoop2584
  %.lcssa4083 = phi i32 [ %3639, %ComputeLoop2584 ]
  %3644 = trunc i64 %1645 to i32
  %3645 = lshr i64 %1645, 32
  %3646 = trunc i64 %3645 to i32
  %3647 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %3644, i32 0)
  %3648 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %3646, i32 %3647)
  %3649 = icmp eq i32 %3648, 0
  %3650 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3649)
  %3651 = extractvalue { i1, i64 } %3650, 0
  %3652 = extractvalue { i1, i64 } %3650, 1
  br i1 %3651, label %1646, label %1648

ComputeLoop2588:                                  ; preds = %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit1255, %ComputeLoop2588
  %Accumulator2590 = phi i32 [ %3656, %ComputeLoop2588 ], [ -1, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit1255 ]
  %ActiveBits2591 = phi i64 [ %3659, %ComputeLoop2588 ], [ %1725, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit1255 ]
  %3653 = call i64 @llvm.cttz.i64(i64 %ActiveBits2591, i1 true)
  %3654 = trunc i64 %3653 to i32
  %3655 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i21.i.i1246, i32 %3654)
  %3656 = and i32 %Accumulator2590, %3655
  %3657 = shl i64 1, %3653
  %3658 = xor i64 %3657, -1
  %3659 = and i64 %ActiveBits2591, %3658
  %3660 = icmp eq i64 %3659, 0
  br i1 %3660, label %ComputeEnd2589, label %ComputeLoop2588, !amdgpu.uniform !42

ComputeEnd2589:                                   ; preds = %ComputeLoop2588
  %.lcssa4077 = phi i32 [ %3656, %ComputeLoop2588 ]
  %3661 = trunc i64 %1725 to i32
  %3662 = lshr i64 %1725, 32
  %3663 = trunc i64 %3662 to i32
  %3664 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %3661, i32 0)
  %3665 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %3663, i32 %3664)
  %3666 = icmp eq i32 %3665, 0
  %3667 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3666)
  %3668 = extractvalue { i1, i64 } %3667, 0
  %3669 = extractvalue { i1, i64 } %3667, 1
  br i1 %3668, label %1726, label %1729

ComputeLoop2592:                                  ; preds = %for.cond.i.i.i1257, %ComputeLoop2592
  %Accumulator2594 = phi i32 [ %3674, %ComputeLoop2592 ], [ 0, %for.cond.i.i.i1257 ]
  %OldValuePhi2595 = phi i32 [ %3673, %ComputeLoop2592 ], [ poison, %for.cond.i.i.i1257 ]
  %ActiveBits2596 = phi i64 [ %3677, %ComputeLoop2592 ], [ %1738, %for.cond.i.i.i1257 ]
  %3670 = call i64 @llvm.cttz.i64(i64 %ActiveBits2596, i1 true)
  %3671 = trunc i64 %3670 to i32
  %3672 = call i32 @llvm.amdgcn.readlane.i32(i32 %shl.i.i.i.i.i1269, i32 %3671)
  %3673 = call i32 @llvm.amdgcn.writelane.i32(i32 %Accumulator2594, i32 %3671, i32 %OldValuePhi2595)
  %3674 = or i32 %Accumulator2594, %3672
  %3675 = shl i64 1, %3670
  %3676 = xor i64 %3675, -1
  %3677 = and i64 %ActiveBits2596, %3676
  %3678 = icmp eq i64 %3677, 0
  br i1 %3678, label %ComputeEnd2593, label %ComputeLoop2592, !amdgpu.uniform !42

ComputeEnd2593:                                   ; preds = %ComputeLoop2592
  %.lcssa4072 = phi i32 [ %3673, %ComputeLoop2592 ]
  %.lcssa4071 = phi i32 [ %3674, %ComputeLoop2592 ]
  %div8.i.i.i.i.i1265 = lshr i32 %spec.store.select.i.i.i1264, 5
  %idxprom.i.i.i.i.i1266 = zext nneg i32 %div8.i.i.i.i.i1265 to i64
  %3679 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i.i.i1266
  %3680 = trunc i64 %1738 to i32
  %3681 = lshr i64 %1738, 32
  %3682 = trunc i64 %3681 to i32
  %3683 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %3680, i32 0)
  %3684 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %3682, i32 %3683)
  %3685 = icmp eq i32 %3684, 0
  %3686 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3685)
  %3687 = extractvalue { i1, i64 } %3686, 0
  %3688 = extractvalue { i1, i64 } %3686, 1
  br i1 %3687, label %1739, label %1741

Flow3592:                                         ; preds = %1759, %if.then.i.i.i.i1275
  %3689 = phi i1 [ false, %1759 ], [ true, %if.then.i.i.i.i1275 ]
  br i1 %3689, label %if.end14.i.i.i1346, label %Flow3593, !amdgpu.uniform !42

ComputeLoop2597:                                  ; preds = %if.then12.i.i.i1280, %ComputeLoop2597
  %Accumulator2599 = phi i32 [ %3693, %ComputeLoop2597 ], [ -1, %if.then12.i.i.i1280 ]
  %ActiveBits2600 = phi i64 [ %3696, %ComputeLoop2597 ], [ %1756, %if.then12.i.i.i1280 ]
  %3690 = call i64 @llvm.cttz.i64(i64 %ActiveBits2600, i1 true)
  %3691 = trunc i64 %3690 to i32
  %3692 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i.i.i1286, i32 %3691)
  %3693 = and i32 %Accumulator2599, %3692
  %3694 = shl i64 1, %3690
  %3695 = xor i64 %3694, -1
  %3696 = and i64 %ActiveBits2600, %3695
  %3697 = icmp eq i64 %3696, 0
  br i1 %3697, label %ComputeEnd2598, label %ComputeLoop2597, !amdgpu.uniform !42

ComputeEnd2598:                                   ; preds = %ComputeLoop2597
  %.lcssa4073 = phi i32 [ %3693, %ComputeLoop2597 ]
  %3698 = trunc i64 %1756 to i32
  %3699 = lshr i64 %1756, 32
  %3700 = trunc i64 %3699 to i32
  %3701 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %3698, i32 0)
  %3702 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %3700, i32 %3701)
  %3703 = icmp eq i32 %3702, 0
  %3704 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3703)
  %3705 = extractvalue { i1, i64 } %3704, 0
  %3706 = extractvalue { i1, i64 } %3704, 1
  br i1 %3705, label %1757, label %1759

ComputeLoop2601:                                  ; preds = %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsINS_3cpp11string_viewEEERS1_T_.exit, %ComputeLoop2601
  %Accumulator2603 = phi i32 [ %3710, %ComputeLoop2601 ], [ -1, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsINS_3cpp11string_viewEEERS1_T_.exit ]
  %ActiveBits2604 = phi i64 [ %3713, %ComputeLoop2601 ], [ %1836, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsINS_3cpp11string_viewEEERS1_T_.exit ]
  %3707 = call i64 @llvm.cttz.i64(i64 %ActiveBits2604, i1 true)
  %3708 = trunc i64 %3707 to i32
  %3709 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i21.i.i1345, i32 %3708)
  %3710 = and i32 %Accumulator2603, %3709
  %3711 = shl i64 1, %3707
  %3712 = xor i64 %3711, -1
  %3713 = and i64 %ActiveBits2604, %3712
  %3714 = icmp eq i64 %3713, 0
  br i1 %3714, label %ComputeEnd2602, label %ComputeLoop2601, !amdgpu.uniform !42

ComputeEnd2602:                                   ; preds = %ComputeLoop2601
  %.lcssa4067 = phi i32 [ %3710, %ComputeLoop2601 ]
  %3715 = trunc i64 %1836 to i32
  %3716 = lshr i64 %1836, 32
  %3717 = trunc i64 %3716 to i32
  %3718 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %3715, i32 0)
  %3719 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %3717, i32 %3718)
  %3720 = icmp eq i32 %3719, 0
  %3721 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3720)
  %3722 = extractvalue { i1, i64 } %3721, 0
  %3723 = extractvalue { i1, i64 } %3721, 1
  br i1 %3722, label %1837, label %1840

ComputeLoop2605:                                  ; preds = %for.cond.i.i1.i1362, %ComputeLoop2605
  %Accumulator2607 = phi i32 [ %3728, %ComputeLoop2605 ], [ 0, %for.cond.i.i1.i1362 ]
  %OldValuePhi2608 = phi i32 [ %3727, %ComputeLoop2605 ], [ poison, %for.cond.i.i1.i1362 ]
  %ActiveBits2609 = phi i64 [ %3731, %ComputeLoop2605 ], [ %1849, %for.cond.i.i1.i1362 ]
  %3724 = call i64 @llvm.cttz.i64(i64 %ActiveBits2609, i1 true)
  %3725 = trunc i64 %3724 to i32
  %3726 = call i32 @llvm.amdgcn.readlane.i32(i32 %shl.i.i.i.i.i1374, i32 %3725)
  %3727 = call i32 @llvm.amdgcn.writelane.i32(i32 %Accumulator2607, i32 %3725, i32 %OldValuePhi2608)
  %3728 = or i32 %Accumulator2607, %3726
  %3729 = shl i64 1, %3724
  %3730 = xor i64 %3729, -1
  %3731 = and i64 %ActiveBits2609, %3730
  %3732 = icmp eq i64 %3731, 0
  br i1 %3732, label %ComputeEnd2606, label %ComputeLoop2605, !amdgpu.uniform !42

ComputeEnd2606:                                   ; preds = %ComputeLoop2605
  %.lcssa4062 = phi i32 [ %3727, %ComputeLoop2605 ]
  %.lcssa4061 = phi i32 [ %3728, %ComputeLoop2605 ]
  %div8.i.i.i.i.i1370 = lshr i32 %spec.store.select.i.i.i1369, 5
  %idxprom.i.i.i.i.i1371 = zext nneg i32 %div8.i.i.i.i.i1370 to i64
  %3733 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i.i.i1371
  %3734 = trunc i64 %1849 to i32
  %3735 = lshr i64 %1849, 32
  %3736 = trunc i64 %3735 to i32
  %3737 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %3734, i32 0)
  %3738 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %3736, i32 %3737)
  %3739 = icmp eq i32 %3738, 0
  %3740 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3739)
  %3741 = extractvalue { i1, i64 } %3740, 0
  %3742 = extractvalue { i1, i64 } %3740, 1
  br i1 %3741, label %1850, label %1852

Flow3566:                                         ; preds = %1870, %if.then.i.i.i.i1380
  %3743 = phi i1 [ false, %1870 ], [ true, %if.then.i.i.i.i1380 ]
  br i1 %3743, label %if.end14.i.i.i1451, label %Flow3567, !amdgpu.uniform !42

ComputeLoop2610:                                  ; preds = %if.then12.i.i.i1385, %ComputeLoop2610
  %Accumulator2612 = phi i32 [ %3747, %ComputeLoop2610 ], [ -1, %if.then12.i.i.i1385 ]
  %ActiveBits2613 = phi i64 [ %3750, %ComputeLoop2610 ], [ %1867, %if.then12.i.i.i1385 ]
  %3744 = call i64 @llvm.cttz.i64(i64 %ActiveBits2613, i1 true)
  %3745 = trunc i64 %3744 to i32
  %3746 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i.i.i1391, i32 %3745)
  %3747 = and i32 %Accumulator2612, %3746
  %3748 = shl i64 1, %3744
  %3749 = xor i64 %3748, -1
  %3750 = and i64 %ActiveBits2613, %3749
  %3751 = icmp eq i64 %3750, 0
  br i1 %3751, label %ComputeEnd2611, label %ComputeLoop2610, !amdgpu.uniform !42

ComputeEnd2611:                                   ; preds = %ComputeLoop2610
  %.lcssa4063 = phi i32 [ %3747, %ComputeLoop2610 ]
  %3752 = trunc i64 %1867 to i32
  %3753 = lshr i64 %1867, 32
  %3754 = trunc i64 %3753 to i32
  %3755 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %3752, i32 0)
  %3756 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %3754, i32 %3755)
  %3757 = icmp eq i32 %3756, 0
  %3758 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3757)
  %3759 = extractvalue { i1, i64 } %3758, 0
  %3760 = extractvalue { i1, i64 } %3758, 1
  br i1 %3759, label %1868, label %1870

ComputeLoop2614:                                  ; preds = %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit1459, %ComputeLoop2614
  %Accumulator2616 = phi i32 [ %3764, %ComputeLoop2614 ], [ -1, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit1459 ]
  %ActiveBits2617 = phi i64 [ %3767, %ComputeLoop2614 ], [ %1947, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit1459 ]
  %3761 = call i64 @llvm.cttz.i64(i64 %ActiveBits2617, i1 true)
  %3762 = trunc i64 %3761 to i32
  %3763 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i21.i.i1450, i32 %3762)
  %3764 = and i32 %Accumulator2616, %3763
  %3765 = shl i64 1, %3761
  %3766 = xor i64 %3765, -1
  %3767 = and i64 %ActiveBits2617, %3766
  %3768 = icmp eq i64 %3767, 0
  br i1 %3768, label %ComputeEnd2615, label %ComputeLoop2614, !amdgpu.uniform !42

ComputeEnd2615:                                   ; preds = %ComputeLoop2614
  %.lcssa4057 = phi i32 [ %3764, %ComputeLoop2614 ]
  %3769 = trunc i64 %1947 to i32
  %3770 = lshr i64 %1947, 32
  %3771 = trunc i64 %3770 to i32
  %3772 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %3769, i32 0)
  %3773 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %3771, i32 %3772)
  %3774 = icmp eq i32 %3773, 0
  %3775 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3774)
  %3776 = extractvalue { i1, i64 } %3775, 0
  %3777 = extractvalue { i1, i64 } %3775, 1
  br i1 %3776, label %1948, label %1951

ComputeLoop2618:                                  ; preds = %for.cond.i.i1.i1469, %ComputeLoop2618
  %Accumulator2620 = phi i32 [ %3782, %ComputeLoop2618 ], [ 0, %for.cond.i.i1.i1469 ]
  %OldValuePhi2621 = phi i32 [ %3781, %ComputeLoop2618 ], [ poison, %for.cond.i.i1.i1469 ]
  %ActiveBits2622 = phi i64 [ %3785, %ComputeLoop2618 ], [ %1964, %for.cond.i.i1.i1469 ]
  %3778 = call i64 @llvm.cttz.i64(i64 %ActiveBits2622, i1 true)
  %3779 = trunc i64 %3778 to i32
  %3780 = call i32 @llvm.amdgcn.readlane.i32(i32 %shl.i.i.i.i.i1481, i32 %3779)
  %3781 = call i32 @llvm.amdgcn.writelane.i32(i32 %Accumulator2620, i32 %3779, i32 %OldValuePhi2621)
  %3782 = or i32 %Accumulator2620, %3780
  %3783 = shl i64 1, %3778
  %3784 = xor i64 %3783, -1
  %3785 = and i64 %ActiveBits2622, %3784
  %3786 = icmp eq i64 %3785, 0
  br i1 %3786, label %ComputeEnd2619, label %ComputeLoop2618, !amdgpu.uniform !42

ComputeEnd2619:                                   ; preds = %ComputeLoop2618
  %.lcssa4051 = phi i32 [ %3781, %ComputeLoop2618 ]
  %.lcssa4050 = phi i32 [ %3782, %ComputeLoop2618 ]
  %div8.i.i.i.i.i1477 = lshr i32 %spec.store.select.i.i.i1476, 5
  %idxprom.i.i.i.i.i1478 = zext nneg i32 %div8.i.i.i.i.i1477 to i64
  %3787 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i.i.i1478
  %3788 = trunc i64 %1964 to i32
  %3789 = lshr i64 %1964, 32
  %3790 = trunc i64 %3789 to i32
  %3791 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %3788, i32 0)
  %3792 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %3790, i32 %3791)
  %3793 = icmp eq i32 %3792, 0
  %3794 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3793)
  %3795 = extractvalue { i1, i64 } %3794, 0
  %3796 = extractvalue { i1, i64 } %3794, 1
  br i1 %3795, label %1965, label %1967

Flow3540:                                         ; preds = %1985, %if.then.i.i.i.i1487
  %3797 = phi i1 [ false, %1985 ], [ true, %if.then.i.i.i.i1487 ]
  br i1 %3797, label %if.end14.i.i.i1558, label %Flow3541, !amdgpu.uniform !42

ComputeLoop2623:                                  ; preds = %if.then12.i.i.i1492, %ComputeLoop2623
  %Accumulator2625 = phi i32 [ %3801, %ComputeLoop2623 ], [ -1, %if.then12.i.i.i1492 ]
  %ActiveBits2626 = phi i64 [ %3804, %ComputeLoop2623 ], [ %1982, %if.then12.i.i.i1492 ]
  %3798 = call i64 @llvm.cttz.i64(i64 %ActiveBits2626, i1 true)
  %3799 = trunc i64 %3798 to i32
  %3800 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i.i.i1498, i32 %3799)
  %3801 = and i32 %Accumulator2625, %3800
  %3802 = shl i64 1, %3798
  %3803 = xor i64 %3802, -1
  %3804 = and i64 %ActiveBits2626, %3803
  %3805 = icmp eq i64 %3804, 0
  br i1 %3805, label %ComputeEnd2624, label %ComputeLoop2623, !amdgpu.uniform !42

ComputeEnd2624:                                   ; preds = %ComputeLoop2623
  %.lcssa4052 = phi i32 [ %3801, %ComputeLoop2623 ]
  %3806 = trunc i64 %1982 to i32
  %3807 = lshr i64 %1982, 32
  %3808 = trunc i64 %3807 to i32
  %3809 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %3806, i32 0)
  %3810 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %3808, i32 %3809)
  %3811 = icmp eq i32 %3810, 0
  %3812 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3811)
  %3813 = extractvalue { i1, i64 } %3812, 0
  %3814 = extractvalue { i1, i64 } %3812, 1
  br i1 %3813, label %1983, label %1985

ComputeLoop2627:                                  ; preds = %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit1566, %ComputeLoop2627
  %Accumulator2629 = phi i32 [ %3818, %ComputeLoop2627 ], [ -1, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit1566 ]
  %ActiveBits2630 = phi i64 [ %3821, %ComputeLoop2627 ], [ %2080, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit1566 ]
  %3815 = call i64 @llvm.cttz.i64(i64 %ActiveBits2630, i1 true)
  %3816 = trunc i64 %3815 to i32
  %3817 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i21.i.i1557, i32 %3816)
  %3818 = and i32 %Accumulator2629, %3817
  %3819 = shl i64 1, %3815
  %3820 = xor i64 %3819, -1
  %3821 = and i64 %ActiveBits2630, %3820
  %3822 = icmp eq i64 %3821, 0
  br i1 %3822, label %ComputeEnd2628, label %ComputeLoop2627, !amdgpu.uniform !42

ComputeEnd2628:                                   ; preds = %ComputeLoop2627
  %.lcssa4042 = phi i32 [ %3818, %ComputeLoop2627 ]
  %3823 = trunc i64 %2080 to i32
  %3824 = lshr i64 %2080, 32
  %3825 = trunc i64 %3824 to i32
  %3826 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %3823, i32 0)
  %3827 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %3825, i32 %3826)
  %3828 = icmp eq i32 %3827, 0
  %3829 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3828)
  %3830 = extractvalue { i1, i64 } %3829, 0
  %3831 = extractvalue { i1, i64 } %3829, 1
  br i1 %3830, label %2081, label %2084

ComputeLoop2631:                                  ; preds = %for.cond.i.i.i1570, %ComputeLoop2631
  %Accumulator2633 = phi i32 [ %3836, %ComputeLoop2631 ], [ 0, %for.cond.i.i.i1570 ]
  %OldValuePhi2634 = phi i32 [ %3835, %ComputeLoop2631 ], [ poison, %for.cond.i.i.i1570 ]
  %ActiveBits2635 = phi i64 [ %3839, %ComputeLoop2631 ], [ %2093, %for.cond.i.i.i1570 ]
  %3832 = call i64 @llvm.cttz.i64(i64 %ActiveBits2635, i1 true)
  %3833 = trunc i64 %3832 to i32
  %3834 = call i32 @llvm.amdgcn.readlane.i32(i32 %shl.i.i.i.i.i1582, i32 %3833)
  %3835 = call i32 @llvm.amdgcn.writelane.i32(i32 %Accumulator2633, i32 %3833, i32 %OldValuePhi2634)
  %3836 = or i32 %Accumulator2633, %3834
  %3837 = shl i64 1, %3832
  %3838 = xor i64 %3837, -1
  %3839 = and i64 %ActiveBits2635, %3838
  %3840 = icmp eq i64 %3839, 0
  br i1 %3840, label %ComputeEnd2632, label %ComputeLoop2631, !amdgpu.uniform !42

ComputeEnd2632:                                   ; preds = %ComputeLoop2631
  %.lcssa4037 = phi i32 [ %3835, %ComputeLoop2631 ]
  %.lcssa4036 = phi i32 [ %3836, %ComputeLoop2631 ]
  %div8.i.i.i.i.i1578 = lshr i32 %spec.store.select.i.i.i1577, 5
  %idxprom.i.i.i.i.i1579 = zext nneg i32 %div8.i.i.i.i.i1578 to i64
  %3841 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i.i.i1579
  %3842 = trunc i64 %2093 to i32
  %3843 = lshr i64 %2093, 32
  %3844 = trunc i64 %3843 to i32
  %3845 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %3842, i32 0)
  %3846 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %3844, i32 %3845)
  %3847 = icmp eq i32 %3846, 0
  %3848 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3847)
  %3849 = extractvalue { i1, i64 } %3848, 0
  %3850 = extractvalue { i1, i64 } %3848, 1
  br i1 %3849, label %2094, label %2096

Flow3514:                                         ; preds = %2114, %if.then.i.i.i.i1588
  %3851 = phi i1 [ false, %2114 ], [ true, %if.then.i.i.i.i1588 ]
  br i1 %3851, label %if.end14.i.i.i1657, label %Flow3515, !amdgpu.uniform !42

ComputeLoop2636:                                  ; preds = %if.then12.i.i.i1593, %ComputeLoop2636
  %Accumulator2638 = phi i32 [ %3855, %ComputeLoop2636 ], [ -1, %if.then12.i.i.i1593 ]
  %ActiveBits2639 = phi i64 [ %3858, %ComputeLoop2636 ], [ %2111, %if.then12.i.i.i1593 ]
  %3852 = call i64 @llvm.cttz.i64(i64 %ActiveBits2639, i1 true)
  %3853 = trunc i64 %3852 to i32
  %3854 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i.i.i1599, i32 %3853)
  %3855 = and i32 %Accumulator2638, %3854
  %3856 = shl i64 1, %3852
  %3857 = xor i64 %3856, -1
  %3858 = and i64 %ActiveBits2639, %3857
  %3859 = icmp eq i64 %3858, 0
  br i1 %3859, label %ComputeEnd2637, label %ComputeLoop2636, !amdgpu.uniform !42

ComputeEnd2637:                                   ; preds = %ComputeLoop2636
  %.lcssa4038 = phi i32 [ %3855, %ComputeLoop2636 ]
  %3860 = trunc i64 %2111 to i32
  %3861 = lshr i64 %2111, 32
  %3862 = trunc i64 %3861 to i32
  %3863 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %3860, i32 0)
  %3864 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %3862, i32 %3863)
  %3865 = icmp eq i32 %3864, 0
  %3866 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3865)
  %3867 = extractvalue { i1, i64 } %3866, 0
  %3868 = extractvalue { i1, i64 } %3866, 1
  br i1 %3867, label %2112, label %2114

ComputeLoop2640:                                  ; preds = %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIcEERS1_T_.exit1665, %ComputeLoop2640
  %Accumulator2642 = phi i32 [ %3872, %ComputeLoop2640 ], [ -1, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIcEERS1_T_.exit1665 ]
  %ActiveBits2643 = phi i64 [ %3875, %ComputeLoop2640 ], [ %2166, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIcEERS1_T_.exit1665 ]
  %3869 = call i64 @llvm.cttz.i64(i64 %ActiveBits2643, i1 true)
  %3870 = trunc i64 %3869 to i32
  %3871 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i21.i.i1656, i32 %3870)
  %3872 = and i32 %Accumulator2642, %3871
  %3873 = shl i64 1, %3869
  %3874 = xor i64 %3873, -1
  %3875 = and i64 %ActiveBits2643, %3874
  %3876 = icmp eq i64 %3875, 0
  br i1 %3876, label %ComputeEnd2641, label %ComputeLoop2640, !amdgpu.uniform !42

ComputeEnd2641:                                   ; preds = %ComputeLoop2640
  %.lcssa4032 = phi i32 [ %3872, %ComputeLoop2640 ]
  %3877 = trunc i64 %2166 to i32
  %3878 = lshr i64 %2166, 32
  %3879 = trunc i64 %3878 to i32
  %3880 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %3877, i32 0)
  %3881 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %3879, i32 %3880)
  %3882 = icmp eq i32 %3881, 0
  %3883 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3882)
  %3884 = extractvalue { i1, i64 } %3883, 0
  %3885 = extractvalue { i1, i64 } %3883, 1
  br i1 %3884, label %2167, label %2170

ComputeLoop2644:                                  ; preds = %for.cond.i.i.i1674, %ComputeLoop2644
  %Accumulator2646 = phi i32 [ %3890, %ComputeLoop2644 ], [ 0, %for.cond.i.i.i1674 ]
  %OldValuePhi2647 = phi i32 [ %3889, %ComputeLoop2644 ], [ poison, %for.cond.i.i.i1674 ]
  %ActiveBits2648 = phi i64 [ %3893, %ComputeLoop2644 ], [ %2188, %for.cond.i.i.i1674 ]
  %3886 = call i64 @llvm.cttz.i64(i64 %ActiveBits2648, i1 true)
  %3887 = trunc i64 %3886 to i32
  %3888 = call i32 @llvm.amdgcn.readlane.i32(i32 %shl.i.i.i.i.i1686, i32 %3887)
  %3889 = call i32 @llvm.amdgcn.writelane.i32(i32 %Accumulator2646, i32 %3887, i32 %OldValuePhi2647)
  %3890 = or i32 %Accumulator2646, %3888
  %3891 = shl i64 1, %3886
  %3892 = xor i64 %3891, -1
  %3893 = and i64 %ActiveBits2648, %3892
  %3894 = icmp eq i64 %3893, 0
  br i1 %3894, label %ComputeEnd2645, label %ComputeLoop2644, !amdgpu.uniform !42

ComputeEnd2645:                                   ; preds = %ComputeLoop2644
  %.lcssa4027 = phi i32 [ %3889, %ComputeLoop2644 ]
  %.lcssa4026 = phi i32 [ %3890, %ComputeLoop2644 ]
  %div8.i.i.i.i.i1682 = lshr i32 %spec.store.select.i.i.i1681, 5
  %idxprom.i.i.i.i.i1683 = zext nneg i32 %div8.i.i.i.i.i1682 to i64
  %3895 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i.i.i1683
  %3896 = trunc i64 %2188 to i32
  %3897 = lshr i64 %2188, 32
  %3898 = trunc i64 %3897 to i32
  %3899 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %3896, i32 0)
  %3900 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %3898, i32 %3899)
  %3901 = icmp eq i32 %3900, 0
  %3902 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3901)
  %3903 = extractvalue { i1, i64 } %3902, 0
  %3904 = extractvalue { i1, i64 } %3902, 1
  br i1 %3903, label %2189, label %2191

Flow3497:                                         ; preds = %2209, %if.then.i.i.i.i1692
  %3905 = phi i1 [ false, %2209 ], [ true, %if.then.i.i.i.i1692 ]
  br i1 %3905, label %if.end14.i.i.i1763, label %Flow3498, !amdgpu.uniform !42

ComputeLoop2649:                                  ; preds = %if.then12.i.i.i1697, %ComputeLoop2649
  %Accumulator2651 = phi i32 [ %3909, %ComputeLoop2649 ], [ -1, %if.then12.i.i.i1697 ]
  %ActiveBits2652 = phi i64 [ %3912, %ComputeLoop2649 ], [ %2206, %if.then12.i.i.i1697 ]
  %3906 = call i64 @llvm.cttz.i64(i64 %ActiveBits2652, i1 true)
  %3907 = trunc i64 %3906 to i32
  %3908 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i.i.i1703, i32 %3907)
  %3909 = and i32 %Accumulator2651, %3908
  %3910 = shl i64 1, %3906
  %3911 = xor i64 %3910, -1
  %3912 = and i64 %ActiveBits2652, %3911
  %3913 = icmp eq i64 %3912, 0
  br i1 %3913, label %ComputeEnd2650, label %ComputeLoop2649, !amdgpu.uniform !42

ComputeEnd2650:                                   ; preds = %ComputeLoop2649
  %.lcssa4028 = phi i32 [ %3909, %ComputeLoop2649 ]
  %3914 = trunc i64 %2206 to i32
  %3915 = lshr i64 %2206, 32
  %3916 = trunc i64 %3915 to i32
  %3917 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %3914, i32 0)
  %3918 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %3916, i32 %3917)
  %3919 = icmp eq i32 %3918, 0
  %3920 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3919)
  %3921 = extractvalue { i1, i64 } %3920, 0
  %3922 = extractvalue { i1, i64 } %3920, 1
  br i1 %3921, label %2207, label %2209

ComputeLoop2653:                                  ; preds = %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsINS_3cpp6stringEEERS1_T_.exit1771, %ComputeLoop2653
  %Accumulator2655 = phi i32 [ %3926, %ComputeLoop2653 ], [ -1, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsINS_3cpp6stringEEERS1_T_.exit1771 ]
  %ActiveBits2656 = phi i64 [ %3929, %ComputeLoop2653 ], [ %2283, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsINS_3cpp6stringEEERS1_T_.exit1771 ]
  %3923 = call i64 @llvm.cttz.i64(i64 %ActiveBits2656, i1 true)
  %3924 = trunc i64 %3923 to i32
  %3925 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i21.i.i1762, i32 %3924)
  %3926 = and i32 %Accumulator2655, %3925
  %3927 = shl i64 1, %3923
  %3928 = xor i64 %3927, -1
  %3929 = and i64 %ActiveBits2656, %3928
  %3930 = icmp eq i64 %3929, 0
  br i1 %3930, label %ComputeEnd2654, label %ComputeLoop2653, !amdgpu.uniform !42

ComputeEnd2654:                                   ; preds = %ComputeLoop2653
  %.lcssa4022 = phi i32 [ %3926, %ComputeLoop2653 ]
  %3931 = trunc i64 %2283 to i32
  %3932 = lshr i64 %2283, 32
  %3933 = trunc i64 %3932 to i32
  %3934 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %3931, i32 0)
  %3935 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %3933, i32 %3934)
  %3936 = icmp eq i32 %3935, 0
  %3937 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3936)
  %3938 = extractvalue { i1, i64 } %3937, 0
  %3939 = extractvalue { i1, i64 } %3937, 1
  br i1 %3938, label %2284, label %2287

ComputeLoop2657:                                  ; preds = %for.cond.i.i1.i1780, %ComputeLoop2657
  %Accumulator2659 = phi i32 [ %3944, %ComputeLoop2657 ], [ 0, %for.cond.i.i1.i1780 ]
  %OldValuePhi2660 = phi i32 [ %3943, %ComputeLoop2657 ], [ poison, %for.cond.i.i1.i1780 ]
  %ActiveBits2661 = phi i64 [ %3947, %ComputeLoop2657 ], [ %2296, %for.cond.i.i1.i1780 ]
  %3940 = call i64 @llvm.cttz.i64(i64 %ActiveBits2661, i1 true)
  %3941 = trunc i64 %3940 to i32
  %3942 = call i32 @llvm.amdgcn.readlane.i32(i32 %shl.i.i.i.i.i1792, i32 %3941)
  %3943 = call i32 @llvm.amdgcn.writelane.i32(i32 %Accumulator2659, i32 %3941, i32 %OldValuePhi2660)
  %3944 = or i32 %Accumulator2659, %3942
  %3945 = shl i64 1, %3940
  %3946 = xor i64 %3945, -1
  %3947 = and i64 %ActiveBits2661, %3946
  %3948 = icmp eq i64 %3947, 0
  br i1 %3948, label %ComputeEnd2658, label %ComputeLoop2657, !amdgpu.uniform !42

ComputeEnd2658:                                   ; preds = %ComputeLoop2657
  %.lcssa4017 = phi i32 [ %3943, %ComputeLoop2657 ]
  %.lcssa4016 = phi i32 [ %3944, %ComputeLoop2657 ]
  %div8.i.i.i.i.i1788 = lshr i32 %spec.store.select.i.i.i1787, 5
  %idxprom.i.i.i.i.i1789 = zext nneg i32 %div8.i.i.i.i.i1788 to i64
  %3949 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i.i.i1789
  %3950 = trunc i64 %2296 to i32
  %3951 = lshr i64 %2296, 32
  %3952 = trunc i64 %3951 to i32
  %3953 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %3950, i32 0)
  %3954 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %3952, i32 %3953)
  %3955 = icmp eq i32 %3954, 0
  %3956 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3955)
  %3957 = extractvalue { i1, i64 } %3956, 0
  %3958 = extractvalue { i1, i64 } %3956, 1
  br i1 %3957, label %2297, label %2299

Flow3471:                                         ; preds = %2317, %if.then.i.i.i.i1798
  %3959 = phi i1 [ false, %2317 ], [ true, %if.then.i.i.i.i1798 ]
  br i1 %3959, label %if.end14.i.i.i1869, label %Flow3472, !amdgpu.uniform !42

ComputeLoop2662:                                  ; preds = %if.then12.i.i.i1803, %ComputeLoop2662
  %Accumulator2664 = phi i32 [ %3963, %ComputeLoop2662 ], [ -1, %if.then12.i.i.i1803 ]
  %ActiveBits2665 = phi i64 [ %3966, %ComputeLoop2662 ], [ %2314, %if.then12.i.i.i1803 ]
  %3960 = call i64 @llvm.cttz.i64(i64 %ActiveBits2665, i1 true)
  %3961 = trunc i64 %3960 to i32
  %3962 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i.i.i1809, i32 %3961)
  %3963 = and i32 %Accumulator2664, %3962
  %3964 = shl i64 1, %3960
  %3965 = xor i64 %3964, -1
  %3966 = and i64 %ActiveBits2665, %3965
  %3967 = icmp eq i64 %3966, 0
  br i1 %3967, label %ComputeEnd2663, label %ComputeLoop2662, !amdgpu.uniform !42

ComputeEnd2663:                                   ; preds = %ComputeLoop2662
  %.lcssa4018 = phi i32 [ %3963, %ComputeLoop2662 ]
  %3968 = trunc i64 %2314 to i32
  %3969 = lshr i64 %2314, 32
  %3970 = trunc i64 %3969 to i32
  %3971 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %3968, i32 0)
  %3972 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %3970, i32 %3971)
  %3973 = icmp eq i32 %3972, 0
  %3974 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3973)
  %3975 = extractvalue { i1, i64 } %3974, 0
  %3976 = extractvalue { i1, i64 } %3974, 1
  br i1 %3975, label %2315, label %2317

ComputeLoop2666:                                  ; preds = %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit1877, %ComputeLoop2666
  %Accumulator2668 = phi i32 [ %3980, %ComputeLoop2666 ], [ -1, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit1877 ]
  %ActiveBits2669 = phi i64 [ %3983, %ComputeLoop2666 ], [ %2391, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit1877 ]
  %3977 = call i64 @llvm.cttz.i64(i64 %ActiveBits2669, i1 true)
  %3978 = trunc i64 %3977 to i32
  %3979 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i21.i.i1868, i32 %3978)
  %3980 = and i32 %Accumulator2668, %3979
  %3981 = shl i64 1, %3977
  %3982 = xor i64 %3981, -1
  %3983 = and i64 %ActiveBits2669, %3982
  %3984 = icmp eq i64 %3983, 0
  br i1 %3984, label %ComputeEnd2667, label %ComputeLoop2666, !amdgpu.uniform !42

ComputeEnd2667:                                   ; preds = %ComputeLoop2666
  %.lcssa4012 = phi i32 [ %3980, %ComputeLoop2666 ]
  %3985 = trunc i64 %2391 to i32
  %3986 = lshr i64 %2391, 32
  %3987 = trunc i64 %3986 to i32
  %3988 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %3985, i32 0)
  %3989 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %3987, i32 %3988)
  %3990 = icmp eq i32 %3989, 0
  %3991 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3990)
  %3992 = extractvalue { i1, i64 } %3991, 0
  %3993 = extractvalue { i1, i64 } %3991, 1
  br i1 %3992, label %2392, label %2395

ComputeLoop2670:                                  ; preds = %for.cond.i.i.i1886, %ComputeLoop2670
  %Accumulator2672 = phi i32 [ %3998, %ComputeLoop2670 ], [ 0, %for.cond.i.i.i1886 ]
  %OldValuePhi2673 = phi i32 [ %3997, %ComputeLoop2670 ], [ poison, %for.cond.i.i.i1886 ]
  %ActiveBits2674 = phi i64 [ %4001, %ComputeLoop2670 ], [ %2532, %for.cond.i.i.i1886 ]
  %3994 = call i64 @llvm.cttz.i64(i64 %ActiveBits2674, i1 true)
  %3995 = trunc i64 %3994 to i32
  %3996 = call i32 @llvm.amdgcn.readlane.i32(i32 %shl.i.i.i.i.i1898, i32 %3995)
  %3997 = call i32 @llvm.amdgcn.writelane.i32(i32 %Accumulator2672, i32 %3995, i32 %OldValuePhi2673)
  %3998 = or i32 %Accumulator2672, %3996
  %3999 = shl i64 1, %3994
  %4000 = xor i64 %3999, -1
  %4001 = and i64 %ActiveBits2674, %4000
  %4002 = icmp eq i64 %4001, 0
  br i1 %4002, label %ComputeEnd2671, label %ComputeLoop2670, !amdgpu.uniform !42

ComputeEnd2671:                                   ; preds = %ComputeLoop2670
  %.lcssa4003 = phi i32 [ %3997, %ComputeLoop2670 ]
  %.lcssa4002 = phi i32 [ %3998, %ComputeLoop2670 ]
  %div8.i.i.i.i.i1894 = lshr i32 %spec.store.select.i.i.i1893, 5
  %idxprom.i.i.i.i.i1895 = zext nneg i32 %div8.i.i.i.i.i1894 to i64
  %4003 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i.i.i1895
  %4004 = trunc i64 %2532 to i32
  %4005 = lshr i64 %2532, 32
  %4006 = trunc i64 %4005 to i32
  %4007 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %4004, i32 0)
  %4008 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %4006, i32 %4007)
  %4009 = icmp eq i32 %4008, 0
  %4010 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %4009)
  %4011 = extractvalue { i1, i64 } %4010, 0
  %4012 = extractvalue { i1, i64 } %4010, 1
  br i1 %4011, label %2533, label %2535

Flow3425:                                         ; preds = %2553, %if.then.i.i.i.i1904
  %4013 = phi i1 [ false, %2553 ], [ true, %if.then.i.i.i.i1904 ]
  br i1 %4013, label %if.end14.i.i.i1975, label %Flow3426, !amdgpu.uniform !42

ComputeLoop2675:                                  ; preds = %if.then12.i.i.i1909, %ComputeLoop2675
  %Accumulator2677 = phi i32 [ %4017, %ComputeLoop2675 ], [ -1, %if.then12.i.i.i1909 ]
  %ActiveBits2678 = phi i64 [ %4020, %ComputeLoop2675 ], [ %2550, %if.then12.i.i.i1909 ]
  %4014 = call i64 @llvm.cttz.i64(i64 %ActiveBits2678, i1 true)
  %4015 = trunc i64 %4014 to i32
  %4016 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i.i.i1915, i32 %4015)
  %4017 = and i32 %Accumulator2677, %4016
  %4018 = shl i64 1, %4014
  %4019 = xor i64 %4018, -1
  %4020 = and i64 %ActiveBits2678, %4019
  %4021 = icmp eq i64 %4020, 0
  br i1 %4021, label %ComputeEnd2676, label %ComputeLoop2675, !amdgpu.uniform !42

ComputeEnd2676:                                   ; preds = %ComputeLoop2675
  %.lcssa4004 = phi i32 [ %4017, %ComputeLoop2675 ]
  %4022 = trunc i64 %2550 to i32
  %4023 = lshr i64 %2550, 32
  %4024 = trunc i64 %4023 to i32
  %4025 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %4022, i32 0)
  %4026 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %4024, i32 %4025)
  %4027 = icmp eq i32 %4026, 0
  %4028 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %4027)
  %4029 = extractvalue { i1, i64 } %4028, 0
  %4030 = extractvalue { i1, i64 } %4028, 1
  br i1 %4029, label %2551, label %2553

ComputeLoop2679:                                  ; preds = %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsINS_3cpp6stringEEERS1_T_.exit1983, %ComputeLoop2679
  %Accumulator2681 = phi i32 [ %4034, %ComputeLoop2679 ], [ -1, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsINS_3cpp6stringEEERS1_T_.exit1983 ]
  %ActiveBits2682 = phi i64 [ %4037, %ComputeLoop2679 ], [ %2648, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsINS_3cpp6stringEEERS1_T_.exit1983 ]
  %4031 = call i64 @llvm.cttz.i64(i64 %ActiveBits2682, i1 true)
  %4032 = trunc i64 %4031 to i32
  %4033 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i21.i.i1974, i32 %4032)
  %4034 = and i32 %Accumulator2681, %4033
  %4035 = shl i64 1, %4031
  %4036 = xor i64 %4035, -1
  %4037 = and i64 %ActiveBits2682, %4036
  %4038 = icmp eq i64 %4037, 0
  br i1 %4038, label %ComputeEnd2680, label %ComputeLoop2679, !amdgpu.uniform !42

ComputeEnd2680:                                   ; preds = %ComputeLoop2679
  %.lcssa3994 = phi i32 [ %4034, %ComputeLoop2679 ]
  %4039 = trunc i64 %2648 to i32
  %4040 = lshr i64 %2648, 32
  %4041 = trunc i64 %4040 to i32
  %4042 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %4039, i32 0)
  %4043 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %4041, i32 %4042)
  %4044 = icmp eq i32 %4043, 0
  %4045 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %4044)
  %4046 = extractvalue { i1, i64 } %4045, 0
  %4047 = extractvalue { i1, i64 } %4045, 1
  br i1 %4046, label %2649, label %2652

ComputeLoop2683:                                  ; preds = %for.cond.i.i.i1987, %ComputeLoop2683
  %Accumulator2685 = phi i32 [ %4052, %ComputeLoop2683 ], [ 0, %for.cond.i.i.i1987 ]
  %OldValuePhi2686 = phi i32 [ %4051, %ComputeLoop2683 ], [ poison, %for.cond.i.i.i1987 ]
  %ActiveBits2687 = phi i64 [ %4055, %ComputeLoop2683 ], [ %2661, %for.cond.i.i.i1987 ]
  %4048 = call i64 @llvm.cttz.i64(i64 %ActiveBits2687, i1 true)
  %4049 = trunc i64 %4048 to i32
  %4050 = call i32 @llvm.amdgcn.readlane.i32(i32 %shl.i.i.i.i.i1999, i32 %4049)
  %4051 = call i32 @llvm.amdgcn.writelane.i32(i32 %Accumulator2685, i32 %4049, i32 %OldValuePhi2686)
  %4052 = or i32 %Accumulator2685, %4050
  %4053 = shl i64 1, %4048
  %4054 = xor i64 %4053, -1
  %4055 = and i64 %ActiveBits2687, %4054
  %4056 = icmp eq i64 %4055, 0
  br i1 %4056, label %ComputeEnd2684, label %ComputeLoop2683, !amdgpu.uniform !42

ComputeEnd2684:                                   ; preds = %ComputeLoop2683
  %.lcssa3989 = phi i32 [ %4051, %ComputeLoop2683 ]
  %.lcssa3988 = phi i32 [ %4052, %ComputeLoop2683 ]
  %div8.i.i.i.i.i1995 = lshr i32 %spec.store.select.i.i.i1994, 5
  %idxprom.i.i.i.i.i1996 = zext nneg i32 %div8.i.i.i.i.i1995 to i64
  %4057 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i.i.i1996
  %4058 = trunc i64 %2661 to i32
  %4059 = lshr i64 %2661, 32
  %4060 = trunc i64 %4059 to i32
  %4061 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %4058, i32 0)
  %4062 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %4060, i32 %4061)
  %4063 = icmp eq i32 %4062, 0
  %4064 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %4063)
  %4065 = extractvalue { i1, i64 } %4064, 0
  %4066 = extractvalue { i1, i64 } %4064, 1
  br i1 %4065, label %2662, label %2664

Flow3406:                                         ; preds = %2682, %if.then.i.i.i.i2005
  %4067 = phi i1 [ false, %2682 ], [ true, %if.then.i.i.i.i2005 ]
  br i1 %4067, label %if.end14.i.i.i2074, label %Flow3407, !amdgpu.uniform !42

ComputeLoop2688:                                  ; preds = %if.then12.i.i.i2010, %ComputeLoop2688
  %Accumulator2690 = phi i32 [ %4071, %ComputeLoop2688 ], [ -1, %if.then12.i.i.i2010 ]
  %ActiveBits2691 = phi i64 [ %4074, %ComputeLoop2688 ], [ %2679, %if.then12.i.i.i2010 ]
  %4068 = call i64 @llvm.cttz.i64(i64 %ActiveBits2691, i1 true)
  %4069 = trunc i64 %4068 to i32
  %4070 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i.i.i2016, i32 %4069)
  %4071 = and i32 %Accumulator2690, %4070
  %4072 = shl i64 1, %4068
  %4073 = xor i64 %4072, -1
  %4074 = and i64 %ActiveBits2691, %4073
  %4075 = icmp eq i64 %4074, 0
  br i1 %4075, label %ComputeEnd2689, label %ComputeLoop2688, !amdgpu.uniform !42

ComputeEnd2689:                                   ; preds = %ComputeLoop2688
  %.lcssa3990 = phi i32 [ %4071, %ComputeLoop2688 ]
  %4076 = trunc i64 %2679 to i32
  %4077 = lshr i64 %2679, 32
  %4078 = trunc i64 %4077 to i32
  %4079 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %4076, i32 0)
  %4080 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %4078, i32 %4079)
  %4081 = icmp eq i32 %4080, 0
  %4082 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %4081)
  %4083 = extractvalue { i1, i64 } %4082, 0
  %4084 = extractvalue { i1, i64 } %4082, 1
  br i1 %4083, label %2680, label %2682

ComputeLoop2692:                                  ; preds = %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIcEERS1_T_.exit2082, %ComputeLoop2692
  %Accumulator2694 = phi i32 [ %4088, %ComputeLoop2692 ], [ -1, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIcEERS1_T_.exit2082 ]
  %ActiveBits2695 = phi i64 [ %4091, %ComputeLoop2692 ], [ %2734, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIcEERS1_T_.exit2082 ]
  %4085 = call i64 @llvm.cttz.i64(i64 %ActiveBits2695, i1 true)
  %4086 = trunc i64 %4085 to i32
  %4087 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i21.i.i2073, i32 %4086)
  %4088 = and i32 %Accumulator2694, %4087
  %4089 = shl i64 1, %4085
  %4090 = xor i64 %4089, -1
  %4091 = and i64 %ActiveBits2695, %4090
  %4092 = icmp eq i64 %4091, 0
  br i1 %4092, label %ComputeEnd2693, label %ComputeLoop2692, !amdgpu.uniform !42

ComputeEnd2693:                                   ; preds = %ComputeLoop2692
  %.lcssa = phi i32 [ %4088, %ComputeLoop2692 ]
  %4093 = trunc i64 %2734 to i32
  %4094 = lshr i64 %2734, 32
  %4095 = trunc i64 %4094 to i32
  %4096 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %4093, i32 0)
  %4097 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %4095, i32 %4096)
  %4098 = icmp eq i32 %4097, 0
  %4099 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %4098)
  %4100 = extractvalue { i1, i64 } %4099, 0
  %4101 = extractvalue { i1, i64 } %4099, 1
  br i1 %4100, label %2735, label %2738
}

; Function Attrs: mustprogress nocallback nofree nosync nounwind speculatable willreturn memory(none)
declare i32 @llvm.abs.i32(i32, i1 immarg) #2

; Function Attrs: convergent mustprogress nofree norecurse nounwind
define internal fastcc noundef zeroext i1 @_ZN22__llvm_libc_22_0_0_git7testing8internal4testIbEEbPNS1_10RunContextENS0_8TestCondET_S6_PKcS8_NS1_8LocationE(ptr noundef writeonly captures(none) %Ctx, i1 noundef zeroext %LHS, i1 noundef zeroext %RHS, ptr noundef %LHSStr, ptr noundef %RHSStr, i32 %Loc.coerce1) unnamed_addr #8 comdat {
entry:
  %LHS.inv = xor i1 %LHS, true
  %RHS.inv = xor i1 %RHS, true
  %buffer.i.i.i = alloca %"class.__llvm_libc_22_0_0_git::IntegerToString", align 8, addrspace(5)
  %buffer.i.i.i93.i = alloca %"class.__llvm_libc_22_0_0_git::IntegerToString", align 8, addrspace(5)
  %buffer.i.i.i.i = alloca %"class.__llvm_libc_22_0_0_git::IntegerToString", align 8, addrspace(5)
  %0 = xor i1 %LHS, %RHS
  %1 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %0)
  %2 = extractvalue { i1, i64 } %1, 0
  %3 = extractvalue { i1, i64 } %1, 1
  br i1 %2, label %if.end.i, label %Flow3776

Flow3775:                                         ; preds = %Flow3774, %if.end.i
  %4 = phi ptr [ %cond.i.i, %Flow3774 ], [ addrspacecast (ptr addrspace(4) @_ZN22__llvm_libc_22_0_0_git3cpp6string14NULL_CHARACTERE to ptr), %if.end.i ]
  br label %if.end.i.i.i, !amdgpu.uniform !42

if.end4.i.i.i.i:                                  ; preds = %if.then3.i.i.i.i, %if.then.i.i.i
  %new_capacity.addr.0.i.i.i.i = phi i64 [ %div18.i.i.i.i, %if.then3.i.i.i.i ], [ %inc.i.i.i.i, %if.then.i.i.i ]
  %sub.i.i = add i64 %new_capacity.addr.0.i.i.i.i, 7
  %div2.i.i = and i64 %sub.i.i, -8
  %5 = load ptr, ptr addrspace(1) @_ZN12_GLOBAL__N_13ptrE, align 8, !tbaa !55
  %add.ptr.i.i = getelementptr inbounds nuw i8, ptr %5, i64 %div2.i.i
  store ptr %add.ptr.i.i, ptr addrspace(1) @_ZN12_GLOBAL__N_13ptrE, align 8, !tbaa !55
  %sub.ptr.lhs.cast.i.i = ptrtoint ptr %add.ptr.i.i to i64
  %sub.ptr.sub.i.i = sub i64 %sub.ptr.lhs.cast.i.i, ptrtoint (ptr addrspacecast (ptr addrspace(1) @_ZN12_GLOBAL__N_16memoryE to ptr) to i64)
  %cmp.i.i = icmp ugt i64 %sub.ptr.sub.i.i, 65335
  %cond.i.i = select i1 %cmp.i.i, ptr null, ptr %5
  %6 = icmp ne i64 0, %spec.select.i
  br i1 %6, label %loadstoreloop2181.preheader, label %Flow3774, !amdgpu.uniform !42

loadstoreloop2181.preheader:                      ; preds = %if.end4.i.i.i.i
  br label %loadstoreloop2181, !amdgpu.uniform !42

Flow3774:                                         ; preds = %Flow3773, %if.end4.i.i.i.i
  br label %Flow3775, !amdgpu.uniform !42

loadstoreloop2181:                                ; preds = %loadstoreloop2181.preheader, %loadstoreloop2181
  %7 = phi i64 [ %9, %loadstoreloop2181 ], [ 0, %loadstoreloop2181.preheader ]
  %8 = getelementptr i8, ptr %5, i64 %7
  store i8 0, ptr %8, align 1
  %9 = add i64 %7, 1
  %10 = icmp uge i64 %9, %spec.select.i
  br i1 %10, label %Flow3773, label %loadstoreloop2181, !amdgpu.uniform !42

Flow3773:                                         ; preds = %loadstoreloop2181
  br label %Flow3774, !amdgpu.uniform !42

Flow3776:                                         ; preds = %2546, %entry
  call void @llvm.amdgcn.end.cf.i64(i64 %3)
  br label %_ZZN22__llvm_libc_22_0_0_git7testing8internal4testIbEEbPNS1_10RunContextENS0_8TestCondET_S6_PKcS8_NS1_8LocationEENKUlbNS_3cpp11string_viewEE_clEbSB_.exit, !amdgpu.uniform !42

if.end.i.i.i:                                     ; preds = %Flow3775
  %cmp.not.i6.i.i.i = icmp ne ptr %4, addrspacecast (ptr addrspace(4) @_ZN22__llvm_libc_22_0_0_git3cpp6string14NULL_CHARACTERE to ptr)
  br i1 %cmp.not.i6.i.i.i, label %if.then.i.i.i.i, label %_ZN22__llvm_libc_22_0_0_git3cpp6stringC2Emc.exit.i, !amdgpu.uniform !42

if.then.i.i.i.i:                                  ; preds = %if.end.i.i.i
  %arrayidx.i.i.i.i = getelementptr inbounds nuw i8, ptr %4, i64 %spec.select.i
  store i8 0, ptr %arrayidx.i.i.i.i, align 1, !tbaa !57
  br label %_ZN22__llvm_libc_22_0_0_git3cpp6stringC2Emc.exit.i, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git3cpp6stringC2Emc.exit.i: ; preds = %if.then.i.i.i.i, %if.end.i.i.i
  %11 = icmp ne i64 0, %spec.select.i
  br i1 %11, label %loadstoreloop2179.preheader, label %Flow3772, !amdgpu.uniform !42

loadstoreloop2179.preheader:                      ; preds = %_ZN22__llvm_libc_22_0_0_git3cpp6stringC2Emc.exit.i
  br label %loadstoreloop2179, !amdgpu.uniform !42

Flow3772:                                         ; preds = %Flow3771, %_ZN22__llvm_libc_22_0_0_git3cpp6stringC2Emc.exit.i
  br label %split2178, !amdgpu.uniform !42

loadstoreloop2179:                                ; preds = %loadstoreloop2179.preheader, %loadstoreloop2179
  %12 = phi i64 [ %14, %loadstoreloop2179 ], [ 0, %loadstoreloop2179.preheader ]
  %13 = getelementptr i8, ptr %4, i64 %12
  store i8 32, ptr %13, align 1
  %14 = add i64 %12, 1
  %15 = icmp uge i64 %14, %spec.select.i
  br i1 %15, label %Flow3771, label %loadstoreloop2179, !amdgpu.uniform !42

Flow3771:                                         ; preds = %loadstoreloop2179
  br label %Flow3772, !amdgpu.uniform !42

split2178:                                        ; preds = %Flow3772
  %16 = tail call i32 @llvm.amdgcn.mbcnt.lo(i32 -1, i32 0)
  %17 = tail call noundef i32 @llvm.amdgcn.mbcnt.hi(i32 -1, i32 %16)
  %sh_prom.i.i.i.i = zext i32 %17 to i64
  br label %for.cond.i.i1.i, !amdgpu.uniform !42

for.cond.i.i1.i:                                  ; preds = %split2178, %Flow3770
  %18 = phi i32 [ %56, %Flow3770 ], [ poison, %split2178 ]
  %19 = phi i64 [ %57, %Flow3770 ], [ poison, %split2178 ]
  %20 = phi i32 [ %58, %Flow3770 ], [ poison, %split2178 ]
  %retval.sroa.6.0.i.i.i = phi i32 [ %61, %Flow3770 ], [ undef, %split2178 ]
  %retval.sroa.2.0.i.i.i = phi i64 [ %60, %Flow3770 ], [ undef, %split2178 ]
  %retval.sroa.8.0.i.i.i = phi i32 [ %59, %Flow3770 ], [ undef, %split2178 ]
  %index.0.i.i.i = phi i32 [ %63, %Flow3770 ], [ 0, %split2178 ]
  %21 = load i32, ptr addrspace(1) @__llvm_rpc_client, align 8, !tbaa !68
  %cmp.not.i.i.i = icmp ult i32 %index.0.i.i.i, %21
  %spec.store.select.i.i.i = select i1 %cmp.not.i.i.i, i32 %index.0.i.i.i, i32 0
  %22 = tail call noundef i64 @llvm.amdgcn.ballot.i64(i1 true)
  %23 = and i32 %spec.store.select.i.i.i, 31
  %24 = lshr i64 %22, %sh_prom.i.i.i.i
  %25 = trunc i64 %24 to i32
  %conv4.i.i.i.i.i = and i32 %25, 1
  %shl.i.i.i.i.i = shl nuw i32 %conv4.i.i.i.i.i, %23
  %26 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop, !amdgpu.uniform !42

27:                                               ; preds = %ComputeEnd
  %sunkaddr = getelementptr inbounds i8, ptr addrspace(1) %2753, i64 40
  %28 = atomicrmw or ptr addrspace(1) %sunkaddr, i32 %.lcssa4081 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %29, !amdgpu.uniform !42

29:                                               ; preds = %27, %ComputeEnd
  %30 = phi i32 [ %28, %27 ], [ poison, %ComputeEnd ]
  call void @llvm.amdgcn.end.cf.i64(i64 %2762)
  %31 = call i32 @llvm.amdgcn.readfirstlane.i32(i32 %30)
  %32 = or i32 %31, %.lcssa4082
  %shl5.i.i.i.i.i = shl nuw i32 1, %23
  %and.i.i.i.i.i = and i32 %shl5.i.i.i.i.i, %32
  %tobool3.i.i.i.i = icmp ne i32 %and.i.i.i.i.i, 0
  %33 = tail call i64 @llvm.amdgcn.ballot.i64(i1 %tobool3.i.i.i.i)
  %and.i.i.i.i.i.i = and i64 %33, %22
  %cmp.i.not.i.i.i = icmp ne i64 %22, %and.i.i.i.i.i.i
  br i1 %cmp.i.not.i.i.i, label %if.then.i.i.i.i1, label %Flow3769, !amdgpu.uniform !42

if.then.i.i.i.i1:                                 ; preds = %29
  fence syncscope("agent") acquire
  %34 = load <2 x i64>, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !75
  %35 = extractelement <2 x i64> %34, i32 0
  %36 = inttoptr i64 %35 to ptr
  %37 = extractelement <2 x i64> %34, i32 1
  %38 = inttoptr i64 %37 to ptr
  %idxprom.i.i.i.i = zext i32 %spec.store.select.i.i.i to i64
  %arrayidx.i.i.i.i2 = getelementptr inbounds nuw i32, ptr %36, i64 %idxprom.i.i.i.i, !amdgpu.uniform !42
  %39 = load atomic i32, ptr %arrayidx.i.i.i.i2 monotonic, align 4
  %40 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %39)
  %arrayidx.i45.i.i.i = getelementptr inbounds nuw i32, ptr %38, i64 %idxprom.i.i.i.i, !amdgpu.uniform !42
  %41 = load atomic i32, ptr %arrayidx.i45.i.i.i monotonic, align 4
  %42 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %41)
  %cmp.i46.not.i.i.i = icmp ne i32 %40, %42
  br i1 %cmp.i46.not.i.i.i, label %if.then12.i.i.i, label %Flow3767, !amdgpu.uniform !42

if.then12.i.i.i:                                  ; preds = %if.then.i.i.i.i1
  fence syncscope("agent") release
  %43 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %22, i1 true)
  %iszero.i.i.i.i.i.i.i = icmp ne i64 %22, 0
  %cmp2.i.i.i.i.i.i = icmp eq i64 %43, %sh_prom.i.i.i.i
  %cmp.i.i.i.i.i.i = select i1 %iszero.i.i.i.i.i.i.i, i1 %cmp2.i.i.i.i.i.i, i1 false
  %conv4.i.i51.i.i.i = zext i1 %cmp.i.i.i.i.i.i to i32
  %shl.i.i52.i.i.i = shl nuw i32 %conv4.i.i51.i.i.i, %23
  %xor.i.i.i.i.i = xor i32 %shl.i.i52.i.i.i, -1
  %44 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2416, !amdgpu.uniform !42

45:                                               ; preds = %ComputeEnd2417
  %sunkaddr3238 = getelementptr inbounds i8, ptr addrspace(1) %2753, i64 40
  %46 = atomicrmw and ptr addrspace(1) %sunkaddr3238, i32 %.lcssa4083 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %47, !amdgpu.uniform !42

47:                                               ; preds = %45, %ComputeEnd2417
  call void @llvm.amdgcn.end.cf.i64(i64 %2780)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %Flow3767, !amdgpu.uniform !42

if.end14.i.i.i:                                   ; preds = %Flow3767
  %48 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %22, i1 true)
  %iszero.i.i.i.i.i.i = icmp ne i64 %22, 0
  %cmp2.i.i.i.i.i = icmp eq i64 %48, %sh_prom.i.i.i.i
  %cmp.i.i.i.i.i = select i1 %iszero.i.i.i.i.i.i, i1 %cmp2.i.i.i.i.i, i1 false
  %49 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.i.i.i.i.i)
  %50 = extractvalue { i1, i64 } %49, 0
  %51 = extractvalue { i1, i64 } %49, 1
  br i1 %50, label %if.then16.i.i.i, label %if.end22.i.i.i

if.then16.i.i.i:                                  ; preds = %if.end14.i.i.i
  %52 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !76
  %opcode.i.i.i = getelementptr inbounds nuw %"struct.rpc::Header", ptr %52, i64 %idxprom.i.i.i.i, i32 1
  store i32 1660944387, ptr %opcode.i.i.i, align 8, !tbaa !77
  %arrayidx21.i.i.i = getelementptr inbounds nuw %"struct.rpc::Header", ptr %52, i64 %idxprom.i.i.i.i
  store i64 %22, ptr %arrayidx21.i.i.i, align 8, !tbaa !80
  br label %if.end22.i.i.i, !amdgpu.uniform !42

Flow3768:                                         ; preds = %if.end22.i.i.i, %Flow3767
  %53 = phi i32 [ %40, %if.end22.i.i.i ], [ %retval.sroa.8.0.i.i.i, %Flow3767 ]
  %54 = phi i64 [ %22, %if.end22.i.i.i ], [ %retval.sroa.2.0.i.i.i, %Flow3767 ]
  %55 = phi i32 [ %spec.store.select.i.i.i, %if.end22.i.i.i ], [ %retval.sroa.6.0.i.i.i, %Flow3767 ]
  br label %cleanup26.i.i.i, !amdgpu.uniform !42

if.end22.i.i.i:                                   ; preds = %if.then16.i.i.i, %if.end14.i.i.i
  call void @llvm.amdgcn.end.cf.i64(i64 %51)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %Flow3768, !amdgpu.uniform !42

Flow3769:                                         ; preds = %cleanup26.i.i.i, %29
  %56 = phi i32 [ %55, %cleanup26.i.i.i ], [ %18, %29 ]
  %57 = phi i64 [ %54, %cleanup26.i.i.i ], [ %19, %29 ]
  %58 = phi i32 [ %53, %cleanup26.i.i.i ], [ %20, %29 ]
  %59 = phi i32 [ %53, %cleanup26.i.i.i ], [ %retval.sroa.8.0.i.i.i, %29 ]
  %60 = phi i64 [ %54, %cleanup26.i.i.i ], [ %retval.sroa.2.0.i.i.i, %29 ]
  %61 = phi i32 [ %55, %cleanup26.i.i.i ], [ %retval.sroa.6.0.i.i.i, %29 ]
  %62 = phi i1 [ %cmp.i46.not.i.i.i, %cleanup26.i.i.i ], [ true, %29 ]
  br i1 %62, label %for.inc.i.i.i, label %Flow3770, !amdgpu.uniform !42

cleanup26.i.i.i:                                  ; preds = %Flow3768
  br label %Flow3769, !amdgpu.uniform !42

for.inc.i.i.i:                                    ; preds = %Flow3769
  %inc.i.i.i = add i32 %spec.store.select.i.i.i, 1
  br label %Flow3770, !amdgpu.uniform !42

Flow3770:                                         ; preds = %for.inc.i.i.i, %Flow3769
  %63 = phi i32 [ %inc.i.i.i, %for.inc.i.i.i ], [ poison, %Flow3769 ]
  %64 = phi i1 [ false, %for.inc.i.i.i ], [ true, %Flow3769 ]
  br i1 %64, label %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i, label %for.cond.i.i1.i, !amdgpu.uniform !42

_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i: ; preds = %Flow3770
  %.lcssa4086 = phi i32 [ %56, %Flow3770 ]
  %.lcssa4085 = phi i64 [ %57, %Flow3770 ]
  %.lcssa4084 = phi i32 [ %58, %Flow3770 ]
  fence acquire
  %65 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %mul.i.i.i.i.i.i = shl i32 %.lcssa4086, 6
  %idxprom.i19.i.i.i.i.i = zext i32 %mul.i.i.i.i.i.i to i64
  %arrayidx.i20.i.i.i.i.i = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %65, i64 %idxprom.i19.i.i.i.i.i
  %arrayidx.i22.i.i.i.i.i = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %arrayidx.i20.i.i.i.i.i, i64 %sh_prom.i.i.i.i
  store i64 sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.2 to ptr), i64 81) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.2 to ptr) to i64)), ptr %arrayidx.i22.i.i.i.i.i, align 8, !tbaa !82
  %spec.select.i.i.i.i.i.i.i = tail call i64 @llvm.umin.i64(i64 sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.2 to ptr), i64 81) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.2 to ptr) to i64)), i64 56)
  %66 = and i64 %spec.select.i.i.i.i.i.i.i, 15
  %67 = sub i64 %spec.select.i.i.i.i.i.i.i, %66
  %68 = icmp ne i64 %67, 0
  br i1 %68, label %loop-memcpy-expansion2309.preheader, label %Flow3766, !amdgpu.uniform !42

loop-memcpy-expansion2309.preheader:              ; preds = %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i
  %69 = shl nuw nsw i64 %sh_prom.i.i.i.i, 6
  %70 = add nuw nsw i64 %69, 8
  %scevgep3205 = getelementptr i8, ptr %65, i64 %70
  %71 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i, 6
  %scevgep3206 = getelementptr i8, ptr %scevgep3205, i64 %71
  br label %loop-memcpy-expansion2309, !amdgpu.uniform !42

Flow3766:                                         ; preds = %Flow3765, %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i
  br label %loop-memcpy-residual-header2312, !amdgpu.uniform !42

loop-memcpy-expansion2309:                        ; preds = %loop-memcpy-expansion2309.preheader, %loop-memcpy-expansion2309
  %loop-index2310 = phi i64 [ %74, %loop-memcpy-expansion2309 ], [ 0, %loop-memcpy-expansion2309.preheader ]
  %72 = getelementptr i8, ptr addrspace(4) @.str.2, i64 %loop-index2310, !amdgpu.uniform !42
  %73 = load <4 x i32>, ptr addrspace(4) %72, align 1
  %scevgep3207 = getelementptr i8, ptr %scevgep3206, i64 %loop-index2310
  store <4 x i32> %73, ptr %scevgep3207, align 1
  %74 = add i64 %loop-index2310, 16
  %75 = icmp uge i64 %74, %67
  br i1 %75, label %Flow3765, label %loop-memcpy-expansion2309, !amdgpu.uniform !42

Flow3764:                                         ; preds = %Flow3762, %Flow3763
  %76 = phi i64 [ %2672, %Flow3762 ], [ %2669, %Flow3763 ]
  %77 = phi i64 [ %2674, %Flow3762 ], [ %2670, %Flow3763 ]
  br label %post-loop-memcpy-expansion2308, !amdgpu.uniform !42

loop-memcpy-residual2311:                         ; preds = %loop-memcpy-residual2311.preheader, %loop-memcpy-residual2311
  %residual-loop-index2313 = phi i64 [ %79, %loop-memcpy-residual2311 ], [ 0, %loop-memcpy-residual2311.preheader ]
  %scevgep3204 = getelementptr i8, ptr addrspace(4) getelementptr inbounds nuw (i8, ptr addrspace(4) @.str.2, i64 48), i64 %residual-loop-index2313, !amdgpu.uniform !42
  %78 = load i8, ptr addrspace(4) %scevgep3204, align 1
  %scevgep3203 = getelementptr i8, ptr %scevgep3202, i64 %residual-loop-index2313
  store i8 %78, ptr %scevgep3203, align 1
  %79 = add i64 %residual-loop-index2313, 1
  %80 = icmp uge i64 %79, %66
  br i1 %80, label %Flow3762, label %loop-memcpy-residual2311, !amdgpu.uniform !42

Flow3762:                                         ; preds = %loop-memcpy-residual2311
  br label %Flow3764, !amdgpu.uniform !42

post-loop-memcpy-expansion2308:                   ; preds = %Flow3764
  %tobool.not.i.i.i.i.i.i = icmp eq i32 %.lcssa4084, 0
  %conv.i.i.i.i.i.i = zext i1 %tobool.not.i.i.i.i.i.i to i32
  fence release
  %81 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %idxprom.i23.i.i.i.i.i = zext i32 %.lcssa4086 to i64
  %arrayidx.i24.i.i.i.i.i = getelementptr inbounds nuw i32, ptr %81, i64 %idxprom.i23.i.i.i.i.i
  store atomic i32 %conv.i.i.i.i.i.i, ptr %arrayidx.i24.i.i.i.i.i monotonic, align 4
  %82 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !84
  %arrayidx.i.i5.i.i = getelementptr inbounds nuw %"struct.rpc::Header", ptr %82, i64 %idxprom.i23.i.i.i.i.i, !amdgpu.uniform !42
  %83 = load i64, ptr %arrayidx.i.i5.i.i, align 8, !tbaa !80
  br label %while.cond.i.i.i.i, !amdgpu.uniform !42

while.cond.i.i.i.i:                               ; preds = %post-loop-memcpy-expansion2308, %Flow3761
  %phi.broken = phi i64 [ 0, %post-loop-memcpy-expansion2308 ], [ %100, %Flow3761 ]
  %lsr.iv3194 = phi i64 [ %98, %Flow3761 ], [ 25, %post-loop-memcpy-expansion2308 ]
  %lsr.iv3189 = phi ptr addrspace(4) [ %97, %Flow3761 ], [ getelementptr inbounds nuw (i8, ptr addrspace(4) @.str.2, i64 56), %post-loop-memcpy-expansion2308 ]
  %port.sroa.37.0.i.i = phi i32 [ %96, %Flow3761 ], [ %conv.i.i.i.i.i.i, %post-loop-memcpy-expansion2308 ]
  %idx.0.i.i.i.i = phi i64 [ %95, %Flow3761 ], [ 56, %post-loop-memcpy-expansion2308 ]
  %cmp.i.i.i.i = icmp ult i64 %idx.0.i.i.i.i, sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.2 to ptr), i64 81) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.2 to ptr) to i64))
  %84 = tail call i64 @llvm.amdgcn.ballot.i64(i1 %cmp.i.i.i.i)
  %and.i.i.i.i6.i.i = and i64 %84, %83
  %tobool.not.i.i.i.i = icmp ne i64 %and.i.i.i.i6.i.i, 0
  %85 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i22.i.i.i = getelementptr inbounds nuw i32, ptr %85, i64 %idxprom.i23.i.i.i.i.i, !amdgpu.uniform !42
  %86 = load atomic i32, ptr %arrayidx.i22.i.i.i monotonic, align 4
  %87 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %tobool.not.i.i.i.i)
  %88 = extractvalue { i1, i64 } %87, 0
  %89 = extractvalue { i1, i64 } %87, 1
  br i1 %88, label %cond.false.i8.i.i.i.i, label %Flow3761

cond.false.i8.i.i.i.i:                            ; preds = %while.cond.i.i.i.i
  %90 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %86)
  %91 = icmp ne i32 %90, %port.sroa.37.0.i.i
  br i1 %91, label %while.body.i.i21.i.i.i.i.preheader, label %Flow3760, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i.preheader:               ; preds = %cond.false.i8.i.i.i.i
  br label %while.body.i.i21.i.i.i.i, !amdgpu.uniform !42

Flow3760:                                         ; preds = %Flow3759, %cond.false.i8.i.i.i.i
  br label %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i:                         ; preds = %while.body.i.i21.i.i.i.i.preheader, %while.body.i.i21.i.i.i.i
  tail call void @llvm.amdgcn.s.sleep(i32 2)
  %92 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i22.i.i.i.i = getelementptr inbounds nuw i32, ptr %92, i64 %idxprom.i23.i.i.i.i.i, !amdgpu.uniform !42
  %93 = load atomic i32, ptr %arrayidx.i.i.i22.i.i.i.i monotonic, align 4
  %94 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %93)
  %cmp.i.not.i.i23.i.i.i.i = icmp eq i32 %94, %port.sroa.37.0.i.i
  br i1 %cmp.i.not.i.i23.i.i.i.i, label %Flow3759, label %while.body.i.i21.i.i.i.i, !amdgpu.uniform !42

Flow3759:                                         ; preds = %while.body.i.i21.i.i.i.i
  br label %Flow3760, !amdgpu.uniform !42

Flow3761:                                         ; preds = %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i, %while.cond.i.i.i.i
  %95 = phi i64 [ %add.i.i.i.i, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i ], [ poison, %while.cond.i.i.i.i ]
  %96 = phi i32 [ %conv.i.i26.i.i.i.i, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i ], [ poison, %while.cond.i.i.i.i ]
  %97 = phi ptr addrspace(4) [ %scevgep3190, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i ], [ poison, %while.cond.i.i.i.i ]
  %98 = phi i64 [ %lsr.iv.next3195, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i ], [ poison, %while.cond.i.i.i.i ]
  %99 = phi i1 [ false, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i ], [ true, %while.cond.i.i.i.i ]
  call void @llvm.amdgcn.end.cf.i64(i64 %89)
  %100 = call i64 @llvm.amdgcn.if.break.i64(i1 %99, i64 %phi.broken)
  %101 = call i1 @llvm.amdgcn.loop.i64(i64 %100)
  br i1 %101, label %cond.false.i.i.i, label %while.cond.i.i.i.i

_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i: ; preds = %Flow3760
  fence acquire
  br i1 %cmp.i.i.i.i, label %if.then.i.i.i.i.i.i.i, label %Flow3758, !amdgpu.uniform !42

if.then.i.i.i.i.i.i.i:                            ; preds = %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i
  %sub.i.i.i.i.i.i.i = sub nuw i64 sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.2 to ptr), i64 81) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.2 to ptr) to i64)), %idx.0.i.i.i.i
  %spec.select.i.i.i36.i.i.i.i = tail call i64 @llvm.umin.i64(i64 %sub.i.i.i.i.i.i.i, i64 64)
  %102 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %103 = and i64 %spec.select.i.i.i36.i.i.i.i, 15
  %104 = sub i64 %spec.select.i.i.i36.i.i.i.i, %103
  %105 = icmp ne i64 %104, 0
  br i1 %105, label %loop-memcpy-expansion2315.preheader, label %Flow3757, !amdgpu.uniform !42

loop-memcpy-expansion2315.preheader:              ; preds = %if.then.i.i.i.i.i.i.i
  %arrayidx.i20.i35.i.i.i.i = getelementptr %"struct.rpc::Buffer", ptr %102, i64 %idxprom.i19.i.i.i.i.i
  %arrayidx.i22.i38.i.i.i.i = getelementptr %"struct.rpc::Buffer", ptr %arrayidx.i20.i35.i.i.i.i, i64 %sh_prom.i.i.i.i
  br label %loop-memcpy-expansion2315, !amdgpu.uniform !42

Flow3757:                                         ; preds = %Flow3756, %if.then.i.i.i.i.i.i.i
  br label %loop-memcpy-residual-header2318, !amdgpu.uniform !42

loop-memcpy-expansion2315:                        ; preds = %loop-memcpy-expansion2315.preheader, %loop-memcpy-expansion2315
  %loop-index2316 = phi i64 [ %108, %loop-memcpy-expansion2315 ], [ 0, %loop-memcpy-expansion2315.preheader ]
  %scevgep3191 = getelementptr i8, ptr addrspace(4) %lsr.iv3189, i64 %loop-index2316, !amdgpu.uniform !42
  %106 = load <4 x i32>, ptr addrspace(4) %scevgep3191, align 1
  %107 = getelementptr i8, ptr %arrayidx.i22.i38.i.i.i.i, i64 %loop-index2316
  store <4 x i32> %106, ptr %107, align 1
  %108 = add i64 %loop-index2316, 16
  %109 = icmp uge i64 %108, %104
  br i1 %109, label %Flow3756, label %loop-memcpy-expansion2315, !amdgpu.uniform !42

Flow3755:                                         ; preds = %Flow3754, %loop-memcpy-residual-header2318
  br label %Flow3758, !amdgpu.uniform !42

loop-memcpy-residual2317:                         ; preds = %loop-memcpy-residual2317.preheader, %loop-memcpy-residual2317
  %residual-loop-index2319 = phi i64 [ %111, %loop-memcpy-residual2317 ], [ 0, %loop-memcpy-residual2317.preheader ]
  %scevgep3200 = getelementptr i8, ptr addrspace(4) %scevgep3199, i64 %residual-loop-index2319, !amdgpu.uniform !42
  %110 = load i8, ptr addrspace(4) %scevgep3200, align 1
  %scevgep3198 = getelementptr i8, ptr %scevgep3197, i64 %residual-loop-index2319
  store i8 %110, ptr %scevgep3198, align 1
  %111 = add i64 %residual-loop-index2319, 1
  %112 = icmp uge i64 %111, %103
  br i1 %112, label %Flow3754, label %loop-memcpy-residual2317, !amdgpu.uniform !42

Flow3754:                                         ; preds = %loop-memcpy-residual2317
  br label %Flow3755, !amdgpu.uniform !42

_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i: ; preds = %Flow3758
  %conv.i.i26.i.i.i.i = xor i32 %port.sroa.37.0.i.i, 1
  fence release
  %113 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %arrayidx.i24.i29.i.i.i.i = getelementptr inbounds nuw i32, ptr %113, i64 %idxprom.i23.i.i.i.i.i
  store atomic i32 %conv.i.i26.i.i.i.i, ptr %arrayidx.i24.i29.i.i.i.i monotonic, align 4
  %add.i.i.i.i = add i64 %idx.0.i.i.i.i, 64
  %scevgep3190 = getelementptr i8, ptr addrspace(4) %lsr.iv3189, i64 64
  %lsr.iv.next3195 = add i64 %lsr.iv3194, -64
  br label %Flow3761, !amdgpu.uniform !42

cond.false.i.i.i:                                 ; preds = %Flow3761
  %.lcssa4080 = phi i64 [ %100, %Flow3761 ]
  %port.sroa.37.0.i.i.lcssa = phi i32 [ %port.sroa.37.0.i.i, %Flow3761 ]
  %.lcssa4079 = phi i32 [ %86, %Flow3761 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa4080)
  %114 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %.lcssa4079)
  %115 = icmp ne i32 %114, %port.sroa.37.0.i.i.lcssa
  %116 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %115)
  %117 = extractvalue { i1, i64 } %116, 0
  %118 = extractvalue { i1, i64 } %116, 1
  br i1 %117, label %while.body.i.i.i.i.preheader, label %Flow3753

while.body.i.i.i.i.preheader:                     ; preds = %cond.false.i.i.i
  br label %while.body.i.i.i.i, !amdgpu.uniform !42

Flow3753:                                         ; preds = %Flow3752, %cond.false.i.i.i
  call void @llvm.amdgcn.end.cf.i64(i64 %118)
  br label %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit, !amdgpu.uniform !42

while.body.i.i.i.i:                               ; preds = %while.body.i.i.i.i.preheader, %while.body.i.i.i.i
  %phi.broken3777 = phi i64 [ 0, %while.body.i.i.i.i.preheader ], [ %122, %while.body.i.i.i.i ]
  tail call void @llvm.amdgcn.s.sleep(i32 2)
  %119 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i8.i.i = getelementptr inbounds nuw i32, ptr %119, i64 %idxprom.i23.i.i.i.i.i, !amdgpu.uniform !42
  %120 = load atomic i32, ptr %arrayidx.i.i.i8.i.i monotonic, align 4
  %121 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %120)
  %cmp.i.not.i.i.i.i = icmp eq i32 %121, %port.sroa.37.0.i.i.lcssa
  %122 = call i64 @llvm.amdgcn.if.break.i64(i1 %cmp.i.not.i.i.i.i, i64 %phi.broken3777)
  %123 = call i1 @llvm.amdgcn.loop.i64(i64 %122)
  br i1 %123, label %Flow3752, label %while.body.i.i.i.i

Flow3752:                                         ; preds = %while.body.i.i.i.i
  %.lcssa4078 = phi i64 [ %122, %while.body.i.i.i.i ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa4078)
  br label %Flow3753, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit: ; preds = %Flow3753
  fence acquire
  tail call void @llvm.amdgcn.wave.barrier()
  fence syncscope("agent") release
  %124 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %.lcssa4085, i1 true)
  %iszero.i.i.i.i.i13.i.i = icmp ne i64 %.lcssa4085, 0
  %cmp2.i.i.i.i14.i.i = icmp eq i64 %124, %sh_prom.i.i.i.i
  %cmp.i.i.i.i15.i.i = select i1 %iszero.i.i.i.i.i13.i.i, i1 %cmp2.i.i.i.i14.i.i, i1 false
  %125 = and i32 %.lcssa4086, 31
  %conv4.i.i.i19.i.i = zext i1 %cmp.i.i.i.i15.i.i to i32
  %shl.i.i.i20.i.i = shl nuw i32 %conv4.i.i.i19.i.i, %125
  %xor.i.i.i21.i.i = xor i32 %shl.i.i.i20.i.i, -1
  %126 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2420, !amdgpu.uniform !42

127:                                              ; preds = %ComputeEnd2421
  %div8.i.i.i16.i.i = lshr i32 %.lcssa4086, 5
  %idxprom.i.i.i17.i.i = zext nneg i32 %div8.i.i.i16.i.i to i64
  %128 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i17.i.i
  %sunkaddr3239 = getelementptr inbounds i8, ptr addrspace(1) %128, i64 40
  %129 = atomicrmw and ptr addrspace(1) %sunkaddr3239, i32 %.lcssa4077 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %130, !amdgpu.uniform !42

130:                                              ; preds = %127, %ComputeEnd2421
  call void @llvm.amdgcn.end.cf.i64(i64 %2797)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %for.cond.i.i1.i11, !amdgpu.uniform !42

for.cond.i.i1.i11:                                ; preds = %130, %Flow3751
  %131 = phi i32 [ %169, %Flow3751 ], [ poison, %130 ]
  %132 = phi i64 [ %170, %Flow3751 ], [ poison, %130 ]
  %133 = phi i32 [ %171, %Flow3751 ], [ poison, %130 ]
  %retval.sroa.6.0.i.i.i12 = phi i32 [ %174, %Flow3751 ], [ undef, %130 ]
  %retval.sroa.2.0.i.i.i14 = phi i64 [ %173, %Flow3751 ], [ undef, %130 ]
  %retval.sroa.8.0.i.i.i15 = phi i32 [ %172, %Flow3751 ], [ undef, %130 ]
  %index.0.i.i.i16 = phi i32 [ %176, %Flow3751 ], [ 0, %130 ]
  %134 = load i32, ptr addrspace(1) @__llvm_rpc_client, align 8, !tbaa !68
  %cmp.not.i.i.i17 = icmp ult i32 %index.0.i.i.i16, %134
  %spec.store.select.i.i.i18 = select i1 %cmp.not.i.i.i17, i32 %index.0.i.i.i16, i32 0
  %135 = tail call noundef i64 @llvm.amdgcn.ballot.i64(i1 true)
  %136 = and i32 %spec.store.select.i.i.i18, 31
  %137 = lshr i64 %135, %sh_prom.i.i.i.i
  %138 = trunc i64 %137 to i32
  %conv4.i.i.i.i.i22 = and i32 %138, 1
  %shl.i.i.i.i.i23 = shl nuw i32 %conv4.i.i.i.i.i22, %136
  %139 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2424, !amdgpu.uniform !42

140:                                              ; preds = %ComputeEnd2425
  %sunkaddr3240 = getelementptr inbounds i8, ptr addrspace(1) %2807, i64 40
  %141 = atomicrmw or ptr addrspace(1) %sunkaddr3240, i32 %.lcssa4071 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %142, !amdgpu.uniform !42

142:                                              ; preds = %140, %ComputeEnd2425
  %143 = phi i32 [ %141, %140 ], [ poison, %ComputeEnd2425 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %2816)
  %144 = call i32 @llvm.amdgcn.readfirstlane.i32(i32 %143)
  %145 = or i32 %144, %.lcssa4072
  %shl5.i.i.i.i.i24 = shl nuw i32 1, %136
  %and.i.i.i.i.i25 = and i32 %shl5.i.i.i.i.i24, %145
  %tobool3.i.i.i.i26 = icmp ne i32 %and.i.i.i.i.i25, 0
  %146 = tail call i64 @llvm.amdgcn.ballot.i64(i1 %tobool3.i.i.i.i26)
  %and.i.i.i.i.i.i27 = and i64 %146, %135
  %cmp.i.not.i.i.i28 = icmp ne i64 %135, %and.i.i.i.i.i.i27
  br i1 %cmp.i.not.i.i.i28, label %if.then.i.i.i.i29, label %Flow3750, !amdgpu.uniform !42

if.then.i.i.i.i29:                                ; preds = %142
  fence syncscope("agent") acquire
  %147 = load <2 x i64>, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !75
  %148 = extractelement <2 x i64> %147, i32 0
  %149 = inttoptr i64 %148 to ptr
  %150 = extractelement <2 x i64> %147, i32 1
  %151 = inttoptr i64 %150 to ptr
  %idxprom.i.i.i.i30 = zext i32 %spec.store.select.i.i.i18 to i64
  %arrayidx.i.i.i.i31 = getelementptr inbounds nuw i32, ptr %149, i64 %idxprom.i.i.i.i30, !amdgpu.uniform !42
  %152 = load atomic i32, ptr %arrayidx.i.i.i.i31 monotonic, align 4
  %153 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %152)
  %arrayidx.i45.i.i.i32 = getelementptr inbounds nuw i32, ptr %151, i64 %idxprom.i.i.i.i30, !amdgpu.uniform !42
  %154 = load atomic i32, ptr %arrayidx.i45.i.i.i32 monotonic, align 4
  %155 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %154)
  %cmp.i46.not.i.i.i33 = icmp ne i32 %153, %155
  br i1 %cmp.i46.not.i.i.i33, label %if.then12.i.i.i34, label %Flow3748, !amdgpu.uniform !42

if.then12.i.i.i34:                                ; preds = %if.then.i.i.i.i29
  fence syncscope("agent") release
  %156 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %135, i1 true)
  %iszero.i.i.i.i.i.i.i35 = icmp ne i64 %135, 0
  %cmp2.i.i.i.i.i.i36 = icmp eq i64 %156, %sh_prom.i.i.i.i
  %cmp.i.i.i.i.i.i37 = select i1 %iszero.i.i.i.i.i.i.i35, i1 %cmp2.i.i.i.i.i.i36, i1 false
  %conv4.i.i51.i.i.i38 = zext i1 %cmp.i.i.i.i.i.i37 to i32
  %shl.i.i52.i.i.i39 = shl nuw i32 %conv4.i.i51.i.i.i38, %136
  %xor.i.i.i.i.i40 = xor i32 %shl.i.i52.i.i.i39, -1
  %157 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2429, !amdgpu.uniform !42

158:                                              ; preds = %ComputeEnd2430
  %sunkaddr3241 = getelementptr inbounds i8, ptr addrspace(1) %2807, i64 40
  %159 = atomicrmw and ptr addrspace(1) %sunkaddr3241, i32 %.lcssa4073 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %160, !amdgpu.uniform !42

160:                                              ; preds = %158, %ComputeEnd2430
  call void @llvm.amdgcn.end.cf.i64(i64 %2834)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %Flow3748, !amdgpu.uniform !42

if.end14.i.i.i99:                                 ; preds = %Flow3748
  %161 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %135, i1 true)
  %iszero.i.i.i.i.i.i100 = icmp ne i64 %135, 0
  %cmp2.i.i.i.i.i101 = icmp eq i64 %161, %sh_prom.i.i.i.i
  %cmp.i.i.i.i.i102 = select i1 %iszero.i.i.i.i.i.i100, i1 %cmp2.i.i.i.i.i101, i1 false
  %162 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.i.i.i.i.i102)
  %163 = extractvalue { i1, i64 } %162, 0
  %164 = extractvalue { i1, i64 } %162, 1
  br i1 %163, label %if.then16.i.i.i104, label %if.end22.i.i.i103

if.then16.i.i.i104:                               ; preds = %if.end14.i.i.i99
  %165 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !76
  %opcode.i.i.i105 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %165, i64 %idxprom.i.i.i.i30, i32 1
  store i32 1660944387, ptr %opcode.i.i.i105, align 8, !tbaa !77
  %arrayidx21.i.i.i106 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %165, i64 %idxprom.i.i.i.i30
  store i64 %135, ptr %arrayidx21.i.i.i106, align 8, !tbaa !80
  br label %if.end22.i.i.i103, !amdgpu.uniform !42

Flow3749:                                         ; preds = %if.end22.i.i.i103, %Flow3748
  %166 = phi i32 [ %153, %if.end22.i.i.i103 ], [ %retval.sroa.8.0.i.i.i15, %Flow3748 ]
  %167 = phi i64 [ %135, %if.end22.i.i.i103 ], [ %retval.sroa.2.0.i.i.i14, %Flow3748 ]
  %168 = phi i32 [ %spec.store.select.i.i.i18, %if.end22.i.i.i103 ], [ %retval.sroa.6.0.i.i.i12, %Flow3748 ]
  br label %cleanup26.i.i.i41, !amdgpu.uniform !42

if.end22.i.i.i103:                                ; preds = %if.then16.i.i.i104, %if.end14.i.i.i99
  call void @llvm.amdgcn.end.cf.i64(i64 %164)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %Flow3749, !amdgpu.uniform !42

Flow3750:                                         ; preds = %cleanup26.i.i.i41, %142
  %169 = phi i32 [ %168, %cleanup26.i.i.i41 ], [ %131, %142 ]
  %170 = phi i64 [ %167, %cleanup26.i.i.i41 ], [ %132, %142 ]
  %171 = phi i32 [ %166, %cleanup26.i.i.i41 ], [ %133, %142 ]
  %172 = phi i32 [ %166, %cleanup26.i.i.i41 ], [ %retval.sroa.8.0.i.i.i15, %142 ]
  %173 = phi i64 [ %167, %cleanup26.i.i.i41 ], [ %retval.sroa.2.0.i.i.i14, %142 ]
  %174 = phi i32 [ %168, %cleanup26.i.i.i41 ], [ %retval.sroa.6.0.i.i.i12, %142 ]
  %175 = phi i1 [ %cmp.i46.not.i.i.i33, %cleanup26.i.i.i41 ], [ true, %142 ]
  br i1 %175, label %for.inc.i.i.i46, label %Flow3751, !amdgpu.uniform !42

cleanup26.i.i.i41:                                ; preds = %Flow3749
  br label %Flow3750, !amdgpu.uniform !42

for.inc.i.i.i46:                                  ; preds = %Flow3750
  %inc.i.i.i51 = add i32 %spec.store.select.i.i.i18, 1
  br label %Flow3751, !amdgpu.uniform !42

Flow3751:                                         ; preds = %for.inc.i.i.i46, %Flow3750
  %176 = phi i32 [ %inc.i.i.i51, %for.inc.i.i.i46 ], [ poison, %Flow3750 ]
  %177 = phi i1 [ false, %for.inc.i.i.i46 ], [ true, %Flow3750 ]
  br i1 %177, label %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i52, label %for.cond.i.i1.i11, !amdgpu.uniform !42

_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i52: ; preds = %Flow3751
  %.lcssa4076 = phi i32 [ %169, %Flow3751 ]
  %.lcssa4075 = phi i64 [ %170, %Flow3751 ]
  %.lcssa4074 = phi i32 [ %171, %Flow3751 ]
  fence acquire
  %178 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %mul.i.i.i.i.i.i53 = shl i32 %.lcssa4076, 6
  %idxprom.i19.i.i.i.i.i54 = zext i32 %mul.i.i.i.i.i.i53 to i64
  %arrayidx.i20.i.i.i.i.i55 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %178, i64 %idxprom.i19.i.i.i.i.i54
  %arrayidx.i22.i.i.i.i.i56 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %arrayidx.i20.i.i.i.i.i55, i64 %sh_prom.i.i.i.i
  store i64 sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.6.10 to ptr), i64 1) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.6.10 to ptr) to i64)), ptr %arrayidx.i22.i.i.i.i.i56, align 8, !tbaa !82
  %spec.select.i.i.i.i.i.i.i57 = tail call i64 @llvm.umin.i64(i64 sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.6.10 to ptr), i64 1) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.6.10 to ptr) to i64)), i64 56)
  %179 = and i64 %spec.select.i.i.i.i.i.i.i57, 15
  %180 = sub i64 %spec.select.i.i.i.i.i.i.i57, %179
  %181 = icmp ne i64 %180, 0
  br i1 %181, label %loop-memcpy-expansion2321.preheader, label %Flow3747, !amdgpu.uniform !42

loop-memcpy-expansion2321.preheader:              ; preds = %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i52
  %182 = add nuw nsw i64 %76, 8
  %scevgep3186 = getelementptr i8, ptr %178, i64 %182
  %183 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i54, 6
  %scevgep3187 = getelementptr i8, ptr %scevgep3186, i64 %183
  br label %loop-memcpy-expansion2321, !amdgpu.uniform !42

Flow3747:                                         ; preds = %Flow3746, %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i52
  br label %loop-memcpy-residual-header2324, !amdgpu.uniform !42

loop-memcpy-expansion2321:                        ; preds = %loop-memcpy-expansion2321.preheader, %loop-memcpy-expansion2321
  %loop-index2322 = phi i64 [ %186, %loop-memcpy-expansion2321 ], [ 0, %loop-memcpy-expansion2321.preheader ]
  %184 = getelementptr i8, ptr addrspace(4) @.str.6.10, i64 %loop-index2322, !amdgpu.uniform !42
  %185 = load <4 x i32>, ptr addrspace(4) %184, align 1
  %scevgep3188 = getelementptr i8, ptr %scevgep3187, i64 %loop-index2322
  store <4 x i32> %185, ptr %scevgep3188, align 1
  %186 = add i64 %loop-index2322, 16
  %187 = icmp uge i64 %186, %180
  br i1 %187, label %Flow3746, label %loop-memcpy-expansion2321, !amdgpu.uniform !42

Flow3745:                                         ; preds = %Flow3743, %Flow3744
  %188 = phi i64 [ %2682, %Flow3743 ], [ %2679, %Flow3744 ]
  br label %post-loop-memcpy-expansion2320, !amdgpu.uniform !42

loop-memcpy-residual2323:                         ; preds = %loop-memcpy-residual2323.preheader, %loop-memcpy-residual2323
  %residual-loop-index2325 = phi i64 [ %190, %loop-memcpy-residual2323 ], [ 0, %loop-memcpy-residual2323.preheader ]
  %scevgep3185 = getelementptr i8, ptr addrspace(4) @.str.6.10, i64 %residual-loop-index2325, !amdgpu.uniform !42
  %189 = load i8, ptr addrspace(4) %scevgep3185, align 1
  %scevgep3184 = getelementptr i8, ptr %scevgep3183, i64 %residual-loop-index2325
  store i8 %189, ptr %scevgep3184, align 1
  %190 = add i64 %residual-loop-index2325, 1
  %191 = icmp uge i64 %190, %179
  br i1 %191, label %Flow3743, label %loop-memcpy-residual2323, !amdgpu.uniform !42

Flow3743:                                         ; preds = %loop-memcpy-residual2323
  br label %Flow3745, !amdgpu.uniform !42

post-loop-memcpy-expansion2320:                   ; preds = %Flow3745
  %tobool.not.i.i.i.i.i.i59 = icmp eq i32 %.lcssa4074, 0
  %conv.i.i.i.i.i.i60 = zext i1 %tobool.not.i.i.i.i.i.i59 to i32
  fence release
  %192 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %idxprom.i23.i.i.i.i.i61 = zext i32 %.lcssa4076 to i64
  %arrayidx.i24.i.i.i.i.i62 = getelementptr inbounds nuw i32, ptr %192, i64 %idxprom.i23.i.i.i.i.i61
  store atomic i32 %conv.i.i.i.i.i.i60, ptr %arrayidx.i24.i.i.i.i.i62 monotonic, align 4
  %193 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !84
  %arrayidx.i.i5.i.i63 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %193, i64 %idxprom.i23.i.i.i.i.i61, !amdgpu.uniform !42
  %194 = load i64, ptr %arrayidx.i.i5.i.i63, align 8, !tbaa !80
  br label %while.cond.i.i.i.i64, !amdgpu.uniform !42

while.cond.i.i.i.i64:                             ; preds = %post-loop-memcpy-expansion2320, %Flow3742
  %phi.broken3778 = phi i64 [ 0, %post-loop-memcpy-expansion2320 ], [ %211, %Flow3742 ]
  %lsr.iv3175 = phi i64 [ %209, %Flow3742 ], [ -55, %post-loop-memcpy-expansion2320 ]
  %lsr.iv3170 = phi ptr addrspace(4) [ %208, %Flow3742 ], [ getelementptr (i8, ptr addrspace(4) @.str.6.10, i64 56), %post-loop-memcpy-expansion2320 ]
  %port.sroa.37.0.i.i65 = phi i32 [ %207, %Flow3742 ], [ %conv.i.i.i.i.i.i60, %post-loop-memcpy-expansion2320 ]
  %idx.0.i.i.i.i66 = phi i64 [ %206, %Flow3742 ], [ 56, %post-loop-memcpy-expansion2320 ]
  %cmp.i.i.i.i67 = icmp ult i64 %idx.0.i.i.i.i66, sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.6.10 to ptr), i64 1) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.6.10 to ptr) to i64))
  %195 = tail call i64 @llvm.amdgcn.ballot.i64(i1 %cmp.i.i.i.i67)
  %and.i.i.i.i6.i.i68 = and i64 %195, %194
  %tobool.not.i.i.i.i69 = icmp ne i64 %and.i.i.i.i6.i.i68, 0
  %196 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i22.i.i.i70 = getelementptr inbounds nuw i32, ptr %196, i64 %idxprom.i23.i.i.i.i.i61, !amdgpu.uniform !42
  %197 = load atomic i32, ptr %arrayidx.i22.i.i.i70 monotonic, align 4
  %198 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %tobool.not.i.i.i.i69)
  %199 = extractvalue { i1, i64 } %198, 0
  %200 = extractvalue { i1, i64 } %198, 1
  br i1 %199, label %cond.false.i8.i.i.i.i71, label %Flow3742

cond.false.i8.i.i.i.i71:                          ; preds = %while.cond.i.i.i.i64
  %201 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %197)
  %202 = icmp ne i32 %201, %port.sroa.37.0.i.i65
  br i1 %202, label %while.body.i.i21.i.i.i.i72.preheader, label %Flow3741, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i72.preheader:             ; preds = %cond.false.i8.i.i.i.i71
  br label %while.body.i.i21.i.i.i.i72, !amdgpu.uniform !42

Flow3741:                                         ; preds = %Flow3740, %cond.false.i8.i.i.i.i71
  br label %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i75, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i72:                       ; preds = %while.body.i.i21.i.i.i.i72.preheader, %while.body.i.i21.i.i.i.i72
  tail call void @llvm.amdgcn.s.sleep(i32 2)
  %203 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i22.i.i.i.i73 = getelementptr inbounds nuw i32, ptr %203, i64 %idxprom.i23.i.i.i.i.i61, !amdgpu.uniform !42
  %204 = load atomic i32, ptr %arrayidx.i.i.i22.i.i.i.i73 monotonic, align 4
  %205 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %204)
  %cmp.i.not.i.i23.i.i.i.i74 = icmp eq i32 %205, %port.sroa.37.0.i.i65
  br i1 %cmp.i.not.i.i23.i.i.i.i74, label %Flow3740, label %while.body.i.i21.i.i.i.i72, !amdgpu.uniform !42

Flow3740:                                         ; preds = %while.body.i.i21.i.i.i.i72
  br label %Flow3741, !amdgpu.uniform !42

Flow3742:                                         ; preds = %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i76, %while.cond.i.i.i.i64
  %206 = phi i64 [ %add.i.i.i.i79, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i76 ], [ poison, %while.cond.i.i.i.i64 ]
  %207 = phi i32 [ %conv.i.i26.i.i.i.i77, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i76 ], [ poison, %while.cond.i.i.i.i64 ]
  %208 = phi ptr addrspace(4) [ %scevgep3171, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i76 ], [ poison, %while.cond.i.i.i.i64 ]
  %209 = phi i64 [ %lsr.iv.next3176, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i76 ], [ poison, %while.cond.i.i.i.i64 ]
  %210 = phi i1 [ false, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i76 ], [ true, %while.cond.i.i.i.i64 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %200)
  %211 = call i64 @llvm.amdgcn.if.break.i64(i1 %210, i64 %phi.broken3778)
  %212 = call i1 @llvm.amdgcn.loop.i64(i64 %211)
  br i1 %212, label %cond.false.i.i.i86, label %while.cond.i.i.i.i64

_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i75: ; preds = %Flow3741
  fence acquire
  br i1 %cmp.i.i.i.i67, label %if.then.i.i.i.i.i.i.i80, label %Flow3739, !amdgpu.uniform !42

if.then.i.i.i.i.i.i.i80:                          ; preds = %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i75
  %sub.i.i.i.i.i.i.i82 = sub nuw i64 sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.6.10 to ptr), i64 1) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.6.10 to ptr) to i64)), %idx.0.i.i.i.i66
  %spec.select.i.i.i36.i.i.i.i83 = tail call i64 @llvm.umin.i64(i64 %sub.i.i.i.i.i.i.i82, i64 64)
  %213 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %214 = and i64 %spec.select.i.i.i36.i.i.i.i83, 15
  %215 = sub i64 %spec.select.i.i.i36.i.i.i.i83, %214
  %216 = icmp ne i64 %215, 0
  br i1 %216, label %loop-memcpy-expansion2327.preheader, label %Flow3738, !amdgpu.uniform !42

loop-memcpy-expansion2327.preheader:              ; preds = %if.then.i.i.i.i.i.i.i80
  %arrayidx.i20.i35.i.i.i.i84 = getelementptr %"struct.rpc::Buffer", ptr %213, i64 %idxprom.i19.i.i.i.i.i54
  %arrayidx.i22.i38.i.i.i.i85 = getelementptr %"struct.rpc::Buffer", ptr %arrayidx.i20.i35.i.i.i.i84, i64 %sh_prom.i.i.i.i
  br label %loop-memcpy-expansion2327, !amdgpu.uniform !42

Flow3738:                                         ; preds = %Flow3737, %if.then.i.i.i.i.i.i.i80
  br label %loop-memcpy-residual-header2330, !amdgpu.uniform !42

loop-memcpy-expansion2327:                        ; preds = %loop-memcpy-expansion2327.preheader, %loop-memcpy-expansion2327
  %loop-index2328 = phi i64 [ %219, %loop-memcpy-expansion2327 ], [ 0, %loop-memcpy-expansion2327.preheader ]
  %scevgep3172 = getelementptr i8, ptr addrspace(4) %lsr.iv3170, i64 %loop-index2328, !amdgpu.uniform !42
  %217 = load <4 x i32>, ptr addrspace(4) %scevgep3172, align 1
  %218 = getelementptr i8, ptr %arrayidx.i22.i38.i.i.i.i85, i64 %loop-index2328
  store <4 x i32> %217, ptr %218, align 1
  %219 = add i64 %loop-index2328, 16
  %220 = icmp uge i64 %219, %215
  br i1 %220, label %Flow3737, label %loop-memcpy-expansion2327, !amdgpu.uniform !42

Flow3736:                                         ; preds = %Flow3735, %loop-memcpy-residual-header2330
  br label %Flow3739, !amdgpu.uniform !42

loop-memcpy-residual2329:                         ; preds = %loop-memcpy-residual2329.preheader, %loop-memcpy-residual2329
  %residual-loop-index2331 = phi i64 [ %222, %loop-memcpy-residual2329 ], [ 0, %loop-memcpy-residual2329.preheader ]
  %scevgep3181 = getelementptr i8, ptr addrspace(4) %scevgep3180, i64 %residual-loop-index2331, !amdgpu.uniform !42
  %221 = load i8, ptr addrspace(4) %scevgep3181, align 1
  %scevgep3179 = getelementptr i8, ptr %scevgep3178, i64 %residual-loop-index2331
  store i8 %221, ptr %scevgep3179, align 1
  %222 = add i64 %residual-loop-index2331, 1
  %223 = icmp uge i64 %222, %214
  br i1 %223, label %Flow3735, label %loop-memcpy-residual2329, !amdgpu.uniform !42

Flow3735:                                         ; preds = %loop-memcpy-residual2329
  br label %Flow3736, !amdgpu.uniform !42

_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i76: ; preds = %Flow3739
  %conv.i.i26.i.i.i.i77 = xor i32 %port.sroa.37.0.i.i65, 1
  fence release
  %224 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %arrayidx.i24.i29.i.i.i.i78 = getelementptr inbounds nuw i32, ptr %224, i64 %idxprom.i23.i.i.i.i.i61
  store atomic i32 %conv.i.i26.i.i.i.i77, ptr %arrayidx.i24.i29.i.i.i.i78 monotonic, align 4
  %add.i.i.i.i79 = add i64 %idx.0.i.i.i.i66, 64
  %scevgep3171 = getelementptr i8, ptr addrspace(4) %lsr.iv3170, i64 64
  %lsr.iv.next3176 = add i64 %lsr.iv3175, -64
  br label %Flow3742, !amdgpu.uniform !42

cond.false.i.i.i86:                               ; preds = %Flow3742
  %.lcssa4070 = phi i64 [ %211, %Flow3742 ]
  %port.sroa.37.0.i.i65.lcssa = phi i32 [ %port.sroa.37.0.i.i65, %Flow3742 ]
  %.lcssa4069 = phi i32 [ %197, %Flow3742 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa4070)
  %225 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %.lcssa4069)
  %226 = icmp ne i32 %225, %port.sroa.37.0.i.i65.lcssa
  %227 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %226)
  %228 = extractvalue { i1, i64 } %227, 0
  %229 = extractvalue { i1, i64 } %227, 1
  br i1 %228, label %while.body.i.i.i.i87.preheader, label %Flow3734

while.body.i.i.i.i87.preheader:                   ; preds = %cond.false.i.i.i86
  br label %while.body.i.i.i.i87, !amdgpu.uniform !42

Flow3734:                                         ; preds = %Flow3733, %cond.false.i.i.i86
  call void @llvm.amdgcn.end.cf.i64(i64 %229)
  br label %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit107, !amdgpu.uniform !42

while.body.i.i.i.i87:                             ; preds = %while.body.i.i.i.i87.preheader, %while.body.i.i.i.i87
  %phi.broken3779 = phi i64 [ 0, %while.body.i.i.i.i87.preheader ], [ %233, %while.body.i.i.i.i87 ]
  tail call void @llvm.amdgcn.s.sleep(i32 2)
  %230 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i8.i.i88 = getelementptr inbounds nuw i32, ptr %230, i64 %idxprom.i23.i.i.i.i.i61, !amdgpu.uniform !42
  %231 = load atomic i32, ptr %arrayidx.i.i.i8.i.i88 monotonic, align 4
  %232 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %231)
  %cmp.i.not.i.i.i.i89 = icmp eq i32 %232, %port.sroa.37.0.i.i65.lcssa
  %233 = call i64 @llvm.amdgcn.if.break.i64(i1 %cmp.i.not.i.i.i.i89, i64 %phi.broken3779)
  %234 = call i1 @llvm.amdgcn.loop.i64(i64 %233)
  br i1 %234, label %Flow3733, label %while.body.i.i.i.i87

Flow3733:                                         ; preds = %while.body.i.i.i.i87
  %.lcssa4068 = phi i64 [ %233, %while.body.i.i.i.i87 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa4068)
  br label %Flow3734, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit107: ; preds = %Flow3734
  fence acquire
  tail call void @llvm.amdgcn.wave.barrier()
  fence syncscope("agent") release
  %235 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %.lcssa4075, i1 true)
  %iszero.i.i.i.i.i13.i.i90 = icmp ne i64 %.lcssa4075, 0
  %cmp2.i.i.i.i14.i.i91 = icmp eq i64 %235, %sh_prom.i.i.i.i
  %cmp.i.i.i.i15.i.i92 = select i1 %iszero.i.i.i.i.i13.i.i90, i1 %cmp2.i.i.i.i14.i.i91, i1 false
  %236 = and i32 %.lcssa4076, 31
  %conv4.i.i.i19.i.i96 = zext i1 %cmp.i.i.i.i15.i.i92 to i32
  %shl.i.i.i20.i.i97 = shl nuw i32 %conv4.i.i.i19.i.i96, %236
  %xor.i.i.i21.i.i98 = xor i32 %shl.i.i.i20.i.i97, -1
  %237 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2433, !amdgpu.uniform !42

238:                                              ; preds = %ComputeEnd2434
  %div8.i.i.i16.i.i93 = lshr i32 %.lcssa4076, 5
  %idxprom.i.i.i17.i.i94 = zext nneg i32 %div8.i.i.i16.i.i93 to i64
  %239 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i17.i.i94
  %sunkaddr3242 = getelementptr inbounds i8, ptr addrspace(1) %239, i64 40
  %240 = atomicrmw and ptr addrspace(1) %sunkaddr3242, i32 %.lcssa4067 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %241, !amdgpu.uniform !42

241:                                              ; preds = %238, %ComputeEnd2434
  call void @llvm.amdgcn.end.cf.i64(i64 %2851)
  tail call void @llvm.amdgcn.wave.barrier()
  call void @llvm.lifetime.start.p5(ptr addrspace(5) %buffer.i.i.i) #34, !noalias !128
  %cmp7.i.i.i.i.i.i = icmp ne i32 %Loc.coerce1, 0
  %242 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp7.i.i.i.i.i.i)
  %243 = extractvalue { i1, i64 } %242, 0
  %244 = extractvalue { i1, i64 } %242, 1
  br i1 %243, label %while.body.lr.ph.i.i.i.i.i.i, label %Flow3732

while.body.lr.ph.i.i.i.i.i.i:                     ; preds = %241
  %retval.0.i.i.i.i.i.i = tail call noundef i32 @llvm.abs.i32(i32 %Loc.coerce1, i1 false)
  br label %while.body.i.i.i.i.i.i, !amdgpu.uniform !42

Flow3732:                                         ; preds = %Flow3731, %241
  %245 = phi i1 [ %cmp.i.i.i.i.i.i.i.i.lcssa, %Flow3731 ], [ false, %241 ]
  %246 = phi i64 [ %writer.sroa.10.1.i.i.i.i.lcssa, %Flow3731 ], [ 0, %241 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %244)
  br label %_ZN22__llvm_libc_22_0_0_git15IntegerToStringIiNS_7details3FmtILh10ELb0ELb0ELb0ELm1EEEE13IntegerWriter25write_unsigned_number_decEjRNS1_22StringBufferWriterImplILb0EEE.exit.i.i.i.i.i, !amdgpu.uniform !42

while.body.i.i.i.i.i.i:                           ; preds = %while.body.lr.ph.i.i.i.i.i.i, %_ZN22__llvm_libc_22_0_0_git7details22StringBufferWriterImplILb0EE4pushEc.exit.i.i.i.i.i.i
  %phi.broken3780 = phi i64 [ 0, %while.body.lr.ph.i.i.i.i.i.i ], [ %314, %_ZN22__llvm_libc_22_0_0_git7details22StringBufferWriterImplILb0EE4pushEc.exit.i.i.i.i.i.i ]
  %writer.sroa.10.0.i.i.i.i = phi i64 [ %writer.sroa.10.1.i.i.i.i, %_ZN22__llvm_libc_22_0_0_git7details22StringBufferWriterImplILb0EE4pushEc.exit.i.i.i.i.i.i ], [ 0, %while.body.lr.ph.i.i.i.i.i.i ]
  %value.addr.09.i.i.i.i.i.i = phi i32 [ %div.i.i.i.i.i.i.i, %_ZN22__llvm_libc_22_0_0_git7details22StringBufferWriterImplILb0EE4pushEc.exit.i.i.i.i.i.i ], [ %retval.0.i.i.i.i.i.i, %while.body.lr.ph.i.i.i.i.i.i ]
  br label %NodeBlock3312, !amdgpu.uniform !42

NodeBlock3312:                                    ; preds = %while.body.i.i.i.i.i.i
  %div.i.i.i.i.i.i.i = udiv i32 %value.addr.09.i.i.i.i.i.i, 10
  %.neg.i = mul i32 %div.i.i.i.i.i.i.i, 246
  %rem.i.i.i.i.i.i.decomposed.i = add i32 %.neg.i, %value.addr.09.i.i.i.i.i.i
  %conv.i.i.i.i.i.i.i = trunc i32 %rem.i.i.i.i.i.i.decomposed.i to i8
  %247 = zext i8 %conv.i.i.i.i.i.i.i to i16
  %Pivot3313 = icmp sge i16 %247, 5
  %248 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %Pivot3313)
  %249 = extractvalue { i1, i64 } %248, 0
  %250 = extractvalue { i1, i64 } %248, 1
  br i1 %249, label %NodeBlock3310, label %Flow3730

NodeBlock3310:                                    ; preds = %NodeBlock3312
  %Pivot3311 = icmp sge i16 %247, 7
  %251 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %Pivot3311)
  %252 = extractvalue { i1, i64 } %251, 0
  %253 = extractvalue { i1, i64 } %251, 1
  br i1 %252, label %NodeBlock3308, label %Flow3721

NodeBlock3308:                                    ; preds = %NodeBlock3310
  %Pivot3309 = icmp sge i16 %247, 8
  %254 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %Pivot3309)
  %255 = extractvalue { i1, i64 } %254, 0
  %256 = extractvalue { i1, i64 } %254, 1
  br i1 %255, label %NodeBlock3306, label %Flow3717

NodeBlock3306:                                    ; preds = %NodeBlock3308
  %Pivot3307 = icmp sge i16 %247, 9
  %257 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %Pivot3307)
  %258 = extractvalue { i1, i64 } %257, 0
  %259 = extractvalue { i1, i64 } %257, 1
  br i1 %258, label %sw.bb9.i.i.i.i.i.i.i.i, label %Flow3715

Flow3721:                                         ; preds = %Flow3718, %NodeBlock3310
  %260 = phi i8 [ %304, %Flow3718 ], [ poison, %NodeBlock3310 ]
  %261 = call { i1, i64 } @llvm.amdgcn.else.i64.i64(i64 %253)
  %262 = extractvalue { i1, i64 } %261, 0
  %263 = extractvalue { i1, i64 } %261, 1
  br i1 %262, label %NodeBlock3304, label %Flow3722

NodeBlock3304:                                    ; preds = %Flow3721
  %Pivot3305 = icmp sge i16 %247, 6
  %264 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %Pivot3305)
  %265 = extractvalue { i1, i64 } %264, 0
  %266 = extractvalue { i1, i64 } %264, 1
  br i1 %265, label %sw.bb6.i.i.i.i.i.i.i.i, label %Flow3719

Flow3730:                                         ; preds = %Flow3722, %NodeBlock3312
  %267 = phi i8 [ %306, %Flow3722 ], [ poison, %NodeBlock3312 ]
  %268 = call { i1, i64 } @llvm.amdgcn.else.i64.i64(i64 %250)
  %269 = extractvalue { i1, i64 } %268, 0
  %270 = extractvalue { i1, i64 } %268, 1
  br i1 %269, label %NodeBlock3302, label %if.then.i.i.i.i.i.i.i108

NodeBlock3302:                                    ; preds = %Flow3730
  %Pivot3303 = icmp sge i16 %247, 3
  %271 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %Pivot3303)
  %272 = extractvalue { i1, i64 } %271, 0
  %273 = extractvalue { i1, i64 } %271, 1
  br i1 %272, label %NodeBlock3300, label %Flow3728

NodeBlock3300:                                    ; preds = %NodeBlock3302
  %Pivot3301 = icmp sge i16 %247, 4
  %274 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %Pivot3301)
  %275 = extractvalue { i1, i64 } %274, 0
  %276 = extractvalue { i1, i64 } %274, 1
  br i1 %275, label %sw.bb4.i.i.i.i.i.i.i.i, label %Flow3723

Flow3728:                                         ; preds = %Flow3724, %NodeBlock3302
  %277 = phi i8 [ %307, %Flow3724 ], [ poison, %NodeBlock3302 ]
  %278 = call { i1, i64 } @llvm.amdgcn.else.i64.i64(i64 %273)
  %279 = extractvalue { i1, i64 } %278, 0
  %280 = extractvalue { i1, i64 } %278, 1
  br i1 %279, label %NodeBlock, label %Flow3729

NodeBlock:                                        ; preds = %Flow3728
  %Pivot = icmp sge i16 %247, 2
  %281 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %Pivot)
  %282 = extractvalue { i1, i64 } %281, 0
  %283 = extractvalue { i1, i64 } %281, 1
  br i1 %282, label %sw.bb2.i.i.i.i.i.i.i.i, label %Flow3726

Flow3726:                                         ; preds = %sw.bb2.i.i.i.i.i.i.i.i, %NodeBlock
  %284 = call { i1, i64 } @llvm.amdgcn.else.i64.i64(i64 %283)
  %285 = extractvalue { i1, i64 } %284, 0
  %286 = extractvalue { i1, i64 } %284, 1
  br i1 %285, label %LeafBlock, label %Flow3727

LeafBlock:                                        ; preds = %Flow3726
  %SwitchLeaf = icmp eq i16 %247, 1
  %287 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %SwitchLeaf)
  %288 = extractvalue { i1, i64 } %287, 0
  %289 = extractvalue { i1, i64 } %287, 1
  br i1 %288, label %sw.bb1.i.i.i.i.i.i.i.i, label %Flow3725

sw.bb1.i.i.i.i.i.i.i.i:                           ; preds = %LeafBlock
  br label %Flow3725, !amdgpu.uniform !42

sw.bb2.i.i.i.i.i.i.i.i:                           ; preds = %NodeBlock
  br label %Flow3726, !amdgpu.uniform !42

Flow3723:                                         ; preds = %sw.bb4.i.i.i.i.i.i.i.i, %NodeBlock3300
  %290 = call { i1, i64 } @llvm.amdgcn.else.i64.i64(i64 %276)
  %291 = extractvalue { i1, i64 } %290, 0
  %292 = extractvalue { i1, i64 } %290, 1
  br i1 %291, label %sw.bb3.i.i.i.i.i.i.i.i, label %Flow3724

sw.bb3.i.i.i.i.i.i.i.i:                           ; preds = %Flow3723
  br label %Flow3724, !amdgpu.uniform !42

sw.bb4.i.i.i.i.i.i.i.i:                           ; preds = %NodeBlock3300
  br label %Flow3723, !amdgpu.uniform !42

Flow3719:                                         ; preds = %sw.bb6.i.i.i.i.i.i.i.i, %NodeBlock3304
  %293 = call { i1, i64 } @llvm.amdgcn.else.i64.i64(i64 %266)
  %294 = extractvalue { i1, i64 } %293, 0
  %295 = extractvalue { i1, i64 } %293, 1
  br i1 %294, label %sw.bb5.i.i.i.i.i.i.i.i, label %Flow3720

sw.bb5.i.i.i.i.i.i.i.i:                           ; preds = %Flow3719
  br label %Flow3720, !amdgpu.uniform !42

sw.bb6.i.i.i.i.i.i.i.i:                           ; preds = %NodeBlock3304
  br label %Flow3719, !amdgpu.uniform !42

Flow3717:                                         ; preds = %Flow3716, %NodeBlock3308
  %296 = phi i8 [ %303, %Flow3716 ], [ poison, %NodeBlock3308 ]
  %297 = call { i1, i64 } @llvm.amdgcn.else.i64.i64(i64 %256)
  %298 = extractvalue { i1, i64 } %297, 0
  %299 = extractvalue { i1, i64 } %297, 1
  br i1 %298, label %sw.bb7.i.i.i.i.i.i.i.i, label %Flow3718

sw.bb7.i.i.i.i.i.i.i.i:                           ; preds = %Flow3717
  br label %Flow3718, !amdgpu.uniform !42

Flow3715:                                         ; preds = %sw.bb9.i.i.i.i.i.i.i.i, %NodeBlock3306
  %300 = call { i1, i64 } @llvm.amdgcn.else.i64.i64(i64 %259)
  %301 = extractvalue { i1, i64 } %300, 0
  %302 = extractvalue { i1, i64 } %300, 1
  br i1 %301, label %sw.bb8.i.i.i.i.i.i.i.i, label %Flow3716

sw.bb8.i.i.i.i.i.i.i.i:                           ; preds = %Flow3715
  br label %Flow3716, !amdgpu.uniform !42

sw.bb9.i.i.i.i.i.i.i.i:                           ; preds = %NodeBlock3306
  br label %Flow3715, !amdgpu.uniform !42

Flow3716:                                         ; preds = %sw.bb8.i.i.i.i.i.i.i.i, %Flow3715
  %303 = phi i8 [ 56, %sw.bb8.i.i.i.i.i.i.i.i ], [ 57, %Flow3715 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %302)
  br label %Flow3717, !amdgpu.uniform !42

Flow3718:                                         ; preds = %sw.bb7.i.i.i.i.i.i.i.i, %Flow3717
  %304 = phi i8 [ 55, %sw.bb7.i.i.i.i.i.i.i.i ], [ %296, %Flow3717 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %299)
  br label %Flow3721, !amdgpu.uniform !42

Flow3720:                                         ; preds = %sw.bb5.i.i.i.i.i.i.i.i, %Flow3719
  %305 = phi i8 [ 53, %sw.bb5.i.i.i.i.i.i.i.i ], [ 54, %Flow3719 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %295)
  br label %Flow3722, !amdgpu.uniform !42

Flow3722:                                         ; preds = %Flow3720, %Flow3721
  %306 = phi i8 [ %305, %Flow3720 ], [ %260, %Flow3721 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %263)
  br label %Flow3730, !amdgpu.uniform !42

Flow3724:                                         ; preds = %sw.bb3.i.i.i.i.i.i.i.i, %Flow3723
  %307 = phi i8 [ 51, %sw.bb3.i.i.i.i.i.i.i.i ], [ 52, %Flow3723 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %292)
  br label %Flow3728, !amdgpu.uniform !42

Flow3725:                                         ; preds = %sw.bb1.i.i.i.i.i.i.i.i, %LeafBlock
  %308 = phi i8 [ 49, %sw.bb1.i.i.i.i.i.i.i.i ], [ 48, %LeafBlock ]
  call void @llvm.amdgcn.end.cf.i64(i64 %289)
  br label %Flow3727, !amdgpu.uniform !42

Flow3727:                                         ; preds = %Flow3725, %Flow3726
  %309 = phi i8 [ %308, %Flow3725 ], [ 50, %Flow3726 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %286)
  br label %Flow3729, !amdgpu.uniform !42

Flow3729:                                         ; preds = %Flow3727, %Flow3728
  %310 = phi i8 [ %309, %Flow3727 ], [ %277, %Flow3728 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %280)
  br label %if.then.i.i.i.i.i.i.i108, !amdgpu.uniform !42

if.then.i.i.i.i.i.i.i108:                         ; preds = %Flow3730, %Flow3729
  %retval.0.i.i.i.i.i.i.i.i = phi i8 [ %267, %Flow3730 ], [ %310, %Flow3729 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %270)
  %cmp.i.i.i.i.i.i.i.i = icmp eq i64 %writer.sroa.10.0.i.i.i.i, 11
  %cmp.i.i.i.i.i.i.i.i.inv = xor i1 %cmp.i.i.i.i.i.i.i.i, true
  br i1 %cmp.i.i.i.i.i.i.i.i.inv, label %if.then3.i.i.i.i.i.i.i, label %_ZN22__llvm_libc_22_0_0_git7details22StringBufferWriterImplILb0EE4pushEc.exit.i.i.i.i.i.i, !amdgpu.uniform !42

if.then3.i.i.i.i.i.i.i:                           ; preds = %if.then.i.i.i.i.i.i.i108
  %311 = trunc i64 %writer.sroa.10.0.i.i.i.i to i32
  %312 = xor i32 %311, -1
  %313 = getelementptr i8, ptr addrspace(5) %buffer.i.i.i, i32 %312
  %arrayidx.i.i.i.i.i.i.i.i1092688 = getelementptr i8, ptr addrspace(5) %313, i32 11
  %inc.i.i.i.i.i.i.i = add i64 %writer.sroa.10.0.i.i.i.i, 1
  store i8 %retval.0.i.i.i.i.i.i.i.i, ptr addrspace(5) %arrayidx.i.i.i.i.i.i.i.i1092688, align 1, !tbaa !57, !noalias !128
  br label %_ZN22__llvm_libc_22_0_0_git7details22StringBufferWriterImplILb0EE4pushEc.exit.i.i.i.i.i.i, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git7details22StringBufferWriterImplILb0EE4pushEc.exit.i.i.i.i.i.i: ; preds = %if.then3.i.i.i.i.i.i.i, %if.then.i.i.i.i.i.i.i108
  %writer.sroa.10.1.i.i.i.i = phi i64 [ %inc.i.i.i.i.i.i.i, %if.then3.i.i.i.i.i.i.i ], [ 11, %if.then.i.i.i.i.i.i.i108 ]
  %cmp.i.i.i.i.i.i110 = icmp ult i32 %value.addr.09.i.i.i.i.i.i, 10
  %.not.i.i.i.i.i.i = or i1 %cmp.i.i.i.i.i.i.i.i, %cmp.i.i.i.i.i.i110
  %314 = call i64 @llvm.amdgcn.if.break.i64(i1 %.not.i.i.i.i.i.i, i64 %phi.broken3780)
  %315 = call i1 @llvm.amdgcn.loop.i64(i64 %314)
  br i1 %315, label %Flow3731, label %while.body.i.i.i.i.i.i

Flow3731:                                         ; preds = %_ZN22__llvm_libc_22_0_0_git7details22StringBufferWriterImplILb0EE4pushEc.exit.i.i.i.i.i.i
  %writer.sroa.10.1.i.i.i.i.lcssa = phi i64 [ %writer.sroa.10.1.i.i.i.i, %_ZN22__llvm_libc_22_0_0_git7details22StringBufferWriterImplILb0EE4pushEc.exit.i.i.i.i.i.i ]
  %.lcssa4066 = phi i64 [ %314, %_ZN22__llvm_libc_22_0_0_git7details22StringBufferWriterImplILb0EE4pushEc.exit.i.i.i.i.i.i ]
  %cmp.i.i.i.i.i.i.i.i.lcssa = phi i1 [ %cmp.i.i.i.i.i.i.i.i, %_ZN22__llvm_libc_22_0_0_git7details22StringBufferWriterImplILb0EE4pushEc.exit.i.i.i.i.i.i ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa4066)
  br label %Flow3732, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git15IntegerToStringIiNS_7details3FmtILh10ELb0ELb0ELb0ELm1EEEE13IntegerWriter25write_unsigned_number_decEjRNS1_22StringBufferWriterImplILb0EEE.exit.i.i.i.i.i: ; preds = %Flow3732
  %cmp.i19.i.i.i.i = icmp ne i64 %246, 0
  %or.cond.not.i.i.i.i = select i1 %245, i1 true, i1 %cmp.i19.i.i.i.i
  %or.cond.not.i.i.i.i.inv = xor i1 %or.cond.not.i.i.i.i, true
  %316 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %or.cond.not.i.i.i.i.inv)
  %317 = extractvalue { i1, i64 } %316, 0
  %318 = extractvalue { i1, i64 } %316, 1
  br i1 %317, label %if.then3.i.i.lr.ph.i.i.i.i, label %while.end.i.i.i.i.i

if.then3.i.i.lr.ph.i.i.i.i:                       ; preds = %_ZN22__llvm_libc_22_0_0_git15IntegerToStringIiNS_7details3FmtILh10ELb0ELb0ELb0ELm1EEEE13IntegerWriter25write_unsigned_number_decEjRNS1_22StringBufferWriterImplILb0EEE.exit.i.i.i.i.i
  %arrayidx.i.i.i.i.i.i.i = getelementptr inbounds nuw i8, ptr addrspace(5) %buffer.i.i.i, i32 10
  store i8 48, ptr addrspace(5) %arrayidx.i.i.i.i.i.i.i, align 2, !tbaa !57, !noalias !128
  br label %while.end.i.i.i.i.i, !amdgpu.uniform !42

while.end.i.i.i.i.i:                              ; preds = %if.then3.i.i.lr.ph.i.i.i.i, %_ZN22__llvm_libc_22_0_0_git15IntegerToStringIiNS_7details3FmtILh10ELb0ELb0ELb0ELm1EEEE13IntegerWriter25write_unsigned_number_decEjRNS1_22StringBufferWriterImplILb0EEE.exit.i.i.i.i.i
  %writer.sroa.10.4.i.i.i.i = phi i64 [ 1, %if.then3.i.i.lr.ph.i.i.i.i ], [ %246, %_ZN22__llvm_libc_22_0_0_git15IntegerToStringIiNS_7details3FmtILh10ELb0ELb0ELb0ELm1EEEE13IntegerWriter25write_unsigned_number_decEjRNS1_22StringBufferWriterImplILb0EEE.exit.i.i.i.i.i ]
  call void @llvm.amdgcn.end.cf.i64(i64 %318)
  %cmp4.i.i.i.i.i = icmp sgt i32 %Loc.coerce1, -1
  %brmerge.i.i.i.i.i = or i1 %cmp4.i.i.i.i.i, %245
  %brmerge.i.i.i.i.i.inv = xor i1 %brmerge.i.i.i.i.i, true
  %319 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %brmerge.i.i.i.i.i.inv)
  %320 = extractvalue { i1, i64 } %319, 0
  %321 = extractvalue { i1, i64 } %319, 1
  br i1 %320, label %if.then.i14.i.i.i.i.i, label %_ZN22__llvm_libc_22_0_0_git15IntegerToStringIiNS_7details3FmtILh10ELb0ELb0ELb0ELm1EEEEC2Ei.exit.i.i.i

if.then.i14.i.i.i.i.i:                            ; preds = %while.end.i.i.i.i.i
  %cmp.i.i17.i.i.i.i.i = icmp ne i64 %writer.sroa.10.4.i.i.i.i, 11
  %322 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.i.i17.i.i.i.i.i)
  %323 = extractvalue { i1, i64 } %322, 0
  %324 = extractvalue { i1, i64 } %322, 1
  br i1 %323, label %if.then3.i18.i.i.i.i.i, label %Flow3714

if.then3.i18.i.i.i.i.i:                           ; preds = %if.then.i14.i.i.i.i.i
  %325 = trunc i64 %writer.sroa.10.4.i.i.i.i to i32
  %326 = xor i32 %325, -1
  %327 = getelementptr i8, ptr addrspace(5) %buffer.i.i.i, i32 %326
  %arrayidx.i.i19.i.i.i.i.i2689 = getelementptr i8, ptr addrspace(5) %327, i32 11
  %inc.i20.i.i.i.i.i = add i64 %writer.sroa.10.4.i.i.i.i, 1
  store i8 45, ptr addrspace(5) %arrayidx.i.i19.i.i.i.i.i2689, align 1, !tbaa !57, !noalias !128
  br label %Flow3714, !amdgpu.uniform !42

Flow3714:                                         ; preds = %if.then3.i18.i.i.i.i.i, %if.then.i14.i.i.i.i.i
  %328 = phi i64 [ %inc.i20.i.i.i.i.i, %if.then3.i18.i.i.i.i.i ], [ 11, %if.then.i14.i.i.i.i.i ]
  call void @llvm.amdgcn.end.cf.i64(i64 %324)
  br label %_ZN22__llvm_libc_22_0_0_git15IntegerToStringIiNS_7details3FmtILh10ELb0ELb0ELb0ELm1EEEEC2Ei.exit.i.i.i, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git15IntegerToStringIiNS_7details3FmtILh10ELb0ELb0ELb0ELm1EEEEC2Ei.exit.i.i.i: ; preds = %while.end.i.i.i.i.i, %Flow3714
  %writer.sroa.10.5.i.i.i.i = phi i64 [ %writer.sroa.10.4.i.i.i.i, %while.end.i.i.i.i.i ], [ %328, %Flow3714 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %321)
  %written.i.i.i.i = getelementptr inbounds nuw i8, ptr addrspace(5) %buffer.i.i.i, i32 16
  store i64 %writer.sroa.10.5.i.i.i.i, ptr addrspace(5) %written.i.i.i.i, align 8, !tbaa !91, !noalias !128
  %cmp.i.not.i.i.i.i.i = icmp ne i64 %writer.sroa.10.5.i.i.i.i, 0
  %329 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.i.not.i.i.i.i.i)
  %330 = extractvalue { i1, i64 } %329, 0
  %331 = extractvalue { i1, i64 } %329, 1
  br i1 %330, label %if.then.i.i.i.i.i.i, label %Flow3713

if.then.i.i.i.i.i.i:                              ; preds = %_ZN22__llvm_libc_22_0_0_git15IntegerToStringIiNS_7details3FmtILh10ELb0ELb0ELb0ELm1EEEEC2Ei.exit.i.i.i
  %332 = call { i64, i1 } @llvm.uadd.with.overflow.i64(i64 %writer.sroa.10.5.i.i.i.i, i64 1)
  %ov = extractvalue { i64, i1 } %332, 1
  %ov.inv = xor i1 %ov, true
  %333 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %ov.inv)
  %334 = extractvalue { i1, i64 } %333, 0
  %335 = extractvalue { i1, i64 } %333, 1
  br i1 %334, label %if.end.i.i.i.i.i.i.i111, label %Flow3712

if.end.i.i.i.i.i.i.i111:                          ; preds = %if.then.i.i.i.i.i.i
  %math = extractvalue { i64, i1 } %332, 0
  %cmp2.i.i.i.i.i.i.i = icmp ult i64 %math, 1676976733973595601
  %336 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp2.i.i.i.i.i.i.i)
  %337 = extractvalue { i1, i64 } %336, 0
  %338 = extractvalue { i1, i64 } %336, 1
  br i1 %337, label %if.then3.i.i.i.i4.i.i.i, label %if.end4.i.i.i.i.i.i.i

if.then3.i.i.i.i4.i.i.i:                          ; preds = %if.end.i.i.i.i.i.i.i111
  %mul.i.i.i.i.i.i.i = mul nuw i64 %math, 11
  %div18.i.i.i.i.i.i.i = lshr i64 %mul.i.i.i.i.i.i.i, 3
  br label %if.end4.i.i.i.i.i.i.i, !amdgpu.uniform !42

Flow3712:                                         ; preds = %if.end4.i.i.i.i.i.i.i, %if.then.i.i.i.i.i.i
  %339 = phi ptr [ %cond.i.i.i, %if.end4.i.i.i.i.i.i.i ], [ addrspacecast (ptr addrspace(4) @_ZN22__llvm_libc_22_0_0_git3cpp6string14NULL_CHARACTERE to ptr), %if.then.i.i.i.i.i.i ]
  call void @llvm.amdgcn.end.cf.i64(i64 %335)
  br label %_ZN22__llvm_libc_22_0_0_git3cpp6string7reserveEm.exit.i.i.i.i.i.i, !amdgpu.uniform !42

if.end4.i.i.i.i.i.i.i:                            ; preds = %if.then3.i.i.i.i4.i.i.i, %if.end.i.i.i.i.i.i.i111
  %new_capacity.addr.0.i.i.i.i.i.i.i = phi i64 [ %div18.i.i.i.i.i.i.i, %if.then3.i.i.i.i4.i.i.i ], [ %math, %if.end.i.i.i.i.i.i.i111 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %338)
  %sub.i.i.i = add i64 %new_capacity.addr.0.i.i.i.i.i.i.i, 7
  %div2.i.i.i = and i64 %sub.i.i.i, -8
  %340 = load ptr, ptr addrspace(1) @_ZN12_GLOBAL__N_13ptrE, align 8, !tbaa !55
  %add.ptr.i.i.i = getelementptr inbounds nuw i8, ptr %340, i64 %div2.i.i.i
  store ptr %add.ptr.i.i.i, ptr addrspace(1) @_ZN12_GLOBAL__N_13ptrE, align 8, !tbaa !55
  %sub.ptr.lhs.cast.i.i.i112 = ptrtoint ptr %add.ptr.i.i.i to i64
  %sub.ptr.sub.i.i.i113 = sub i64 %sub.ptr.lhs.cast.i.i.i112, ptrtoint (ptr addrspacecast (ptr addrspace(1) @_ZN12_GLOBAL__N_16memoryE to ptr) to i64)
  %cmp.i.i.i114 = icmp ugt i64 %sub.ptr.sub.i.i.i113, 65335
  %cond.i.i.i = select i1 %cmp.i.i.i114, ptr null, ptr %340
  br label %Flow3712, !amdgpu.uniform !42

Flow3713:                                         ; preds = %Flow3711, %_ZN22__llvm_libc_22_0_0_git15IntegerToStringIiNS_7details3FmtILh10ELb0ELb0ELb0ELm1EEEEC2Ei.exit.i.i.i
  %341 = phi ptr [ %339, %Flow3711 ], [ addrspacecast (ptr addrspace(4) @_ZN22__llvm_libc_22_0_0_git3cpp6string14NULL_CHARACTERE to ptr), %_ZN22__llvm_libc_22_0_0_git15IntegerToStringIiNS_7details3FmtILh10ELb0ELb0ELb0ELm1EEEEC2Ei.exit.i.i.i ]
  call void @llvm.amdgcn.end.cf.i64(i64 %331)
  br label %if.end.i.i.i.i.i.i, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git3cpp6string7reserveEm.exit.i.i.i.i.i.i: ; preds = %Flow3712
  br label %loadstoreloop, !amdgpu.uniform !42

loadstoreloop:                                    ; preds = %_ZN22__llvm_libc_22_0_0_git3cpp6string7reserveEm.exit.i.i.i.i.i.i, %loadstoreloop
  %phi.broken3781 = phi i64 [ 0, %_ZN22__llvm_libc_22_0_0_git3cpp6string7reserveEm.exit.i.i.i.i.i.i ], [ %346, %loadstoreloop ]
  %342 = phi i64 [ %344, %loadstoreloop ], [ 0, %_ZN22__llvm_libc_22_0_0_git3cpp6string7reserveEm.exit.i.i.i.i.i.i ]
  %343 = getelementptr i8, ptr %339, i64 %342
  store i8 0, ptr %343, align 1
  %344 = add i64 %342, 1
  %345 = icmp uge i64 %344, %writer.sroa.10.5.i.i.i.i
  %346 = call i64 @llvm.amdgcn.if.break.i64(i1 %345, i64 %phi.broken3781)
  %347 = call i1 @llvm.amdgcn.loop.i64(i64 %346)
  br i1 %347, label %Flow3711, label %loadstoreloop

Flow3711:                                         ; preds = %loadstoreloop
  %.lcssa4065 = phi i64 [ %346, %loadstoreloop ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa4065)
  br label %Flow3713, !amdgpu.uniform !42

if.end.i.i.i.i.i.i:                               ; preds = %Flow3713
  %cmp.not.i6.i.i.i.i.i.i = icmp ne ptr %341, addrspacecast (ptr addrspace(4) @_ZN22__llvm_libc_22_0_0_git3cpp6string14NULL_CHARACTERE to ptr)
  %348 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.not.i6.i.i.i.i.i.i)
  %349 = extractvalue { i1, i64 } %348, 0
  %350 = extractvalue { i1, i64 } %348, 1
  br i1 %349, label %if.then.i.i.i.i2.i.i.i, label %_ZN22__llvm_libc_22_0_0_git3cpp9to_stringEi.exit.i

if.then.i.i.i.i2.i.i.i:                           ; preds = %if.end.i.i.i.i.i.i
  %arrayidx.i.i.i.i3.i.i.i = getelementptr inbounds nuw i8, ptr %341, i64 %writer.sroa.10.5.i.i.i.i
  store i8 0, ptr %arrayidx.i.i.i.i3.i.i.i, align 1, !tbaa !57, !noalias !128
  br label %_ZN22__llvm_libc_22_0_0_git3cpp9to_stringEi.exit.i, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git3cpp9to_stringEi.exit.i: ; preds = %if.then.i.i.i.i2.i.i.i, %if.end.i.i.i.i.i.i
  call void @llvm.amdgcn.end.cf.i64(i64 %350)
  %351 = trunc i64 %writer.sroa.10.5.i.i.i.i to i32
  %352 = and i64 %writer.sroa.10.5.i.i.i.i, 15
  %353 = sub i64 %writer.sroa.10.5.i.i.i.i, %352
  %354 = icmp ne i64 %353, 0
  %355 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %354)
  %356 = extractvalue { i1, i64 } %355, 0
  %357 = extractvalue { i1, i64 } %355, 1
  br i1 %356, label %loop-memcpy-expansion2291.preheader, label %Flow3710

loop-memcpy-expansion2291.preheader:              ; preds = %_ZN22__llvm_libc_22_0_0_git3cpp9to_stringEi.exit.i
  %358 = sub i32 11, %351
  %scevgep3167 = getelementptr i8, ptr addrspace(5) %buffer.i.i.i, i32 %358
  br label %loop-memcpy-expansion2291, !amdgpu.uniform !42

Flow3710:                                         ; preds = %Flow3709, %_ZN22__llvm_libc_22_0_0_git3cpp9to_stringEi.exit.i
  call void @llvm.amdgcn.end.cf.i64(i64 %357)
  br label %loop-memcpy-residual-header2294, !amdgpu.uniform !42

loop-memcpy-expansion2291:                        ; preds = %loop-memcpy-expansion2291.preheader, %loop-memcpy-expansion2291
  %phi.broken3782 = phi i64 [ 0, %loop-memcpy-expansion2291.preheader ], [ %363, %loop-memcpy-expansion2291 ]
  %lsr.iv3168 = phi ptr addrspace(5) [ %scevgep3169, %loop-memcpy-expansion2291 ], [ %scevgep3167, %loop-memcpy-expansion2291.preheader ]
  %loop-index2292 = phi i64 [ %361, %loop-memcpy-expansion2291 ], [ 0, %loop-memcpy-expansion2291.preheader ]
  %359 = load <4 x i32>, ptr addrspace(5) %lsr.iv3168, align 1
  %360 = getelementptr i8, ptr %341, i64 %loop-index2292
  store <4 x i32> %359, ptr %360, align 1
  %361 = add i64 %loop-index2292, 16
  %scevgep3169 = getelementptr i8, ptr addrspace(5) %lsr.iv3168, i32 16
  %362 = icmp uge i64 %361, %353
  %363 = call i64 @llvm.amdgcn.if.break.i64(i1 %362, i64 %phi.broken3782)
  %364 = call i1 @llvm.amdgcn.loop.i64(i64 %363)
  br i1 %364, label %Flow3709, label %loop-memcpy-expansion2291

Flow3708:                                         ; preds = %Flow3707, %loop-memcpy-residual-header2294
  call void @llvm.amdgcn.end.cf.i64(i64 %2651)
  br label %post-loop-memcpy-expansion2290, !amdgpu.uniform !42

loop-memcpy-residual2293:                         ; preds = %loop-memcpy-residual2293.preheader, %loop-memcpy-residual2293
  %phi.broken3783 = phi i64 [ 0, %loop-memcpy-residual2293.preheader ], [ %368, %loop-memcpy-residual2293 ]
  %lsr.iv3163 = phi ptr addrspace(5) [ %scevgep3164, %loop-memcpy-residual2293 ], [ %scevgep3162, %loop-memcpy-residual2293.preheader ]
  %residual-loop-index2295 = phi i64 [ %366, %loop-memcpy-residual2293 ], [ 0, %loop-memcpy-residual2293.preheader ]
  %365 = load i8, ptr addrspace(5) %lsr.iv3163, align 1
  %scevgep3166 = getelementptr i8, ptr %scevgep3165, i64 %residual-loop-index2295
  store i8 %365, ptr %scevgep3166, align 1
  %366 = add i64 %residual-loop-index2295, 1
  %scevgep3164 = getelementptr i8, ptr addrspace(5) %lsr.iv3163, i32 1
  %367 = icmp uge i64 %366, %352
  %368 = call i64 @llvm.amdgcn.if.break.i64(i1 %367, i64 %phi.broken3783)
  %369 = call i1 @llvm.amdgcn.loop.i64(i64 %368)
  br i1 %369, label %Flow3707, label %loop-memcpy-residual2293

Flow3707:                                         ; preds = %loop-memcpy-residual2293
  %.lcssa4063 = phi i64 [ %368, %loop-memcpy-residual2293 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa4063)
  br label %Flow3708, !amdgpu.uniform !42

post-loop-memcpy-expansion2290:                   ; preds = %Flow3708
  call void @llvm.lifetime.end.p5(ptr addrspace(5) %buffer.i.i.i) #34, !noalias !128
  br label %for.cond.i.i.i116, !amdgpu.uniform !42

for.cond.i.i.i116:                                ; preds = %post-loop-memcpy-expansion2290, %Flow3706
  %370 = phi i32 [ %408, %Flow3706 ], [ poison, %post-loop-memcpy-expansion2290 ]
  %371 = phi i64 [ %409, %Flow3706 ], [ poison, %post-loop-memcpy-expansion2290 ]
  %372 = phi i32 [ %410, %Flow3706 ], [ poison, %post-loop-memcpy-expansion2290 ]
  %retval.sroa.6.0.i.i.i117 = phi i32 [ %413, %Flow3706 ], [ undef, %post-loop-memcpy-expansion2290 ]
  %retval.sroa.2.0.i.i.i119 = phi i64 [ %412, %Flow3706 ], [ undef, %post-loop-memcpy-expansion2290 ]
  %retval.sroa.8.0.i.i.i120 = phi i32 [ %411, %Flow3706 ], [ undef, %post-loop-memcpy-expansion2290 ]
  %index.0.i.i.i121 = phi i32 [ %415, %Flow3706 ], [ 0, %post-loop-memcpy-expansion2290 ]
  %373 = load i32, ptr addrspace(1) @__llvm_rpc_client, align 8, !tbaa !68
  %cmp.not.i.i.i122 = icmp ult i32 %index.0.i.i.i121, %373
  %spec.store.select.i.i.i123 = select i1 %cmp.not.i.i.i122, i32 %index.0.i.i.i121, i32 0
  %374 = tail call noundef i64 @llvm.amdgcn.ballot.i64(i1 true)
  %375 = and i32 %spec.store.select.i.i.i123, 31
  %376 = lshr i64 %374, %sh_prom.i.i.i.i
  %377 = trunc i64 %376 to i32
  %conv4.i.i.i.i.i127 = and i32 %377, 1
  %shl.i.i.i.i.i128 = shl nuw i32 %conv4.i.i.i.i.i127, %375
  %378 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2437, !amdgpu.uniform !42

379:                                              ; preds = %ComputeEnd2438
  %sunkaddr3243 = getelementptr inbounds i8, ptr addrspace(1) %2861, i64 40
  %380 = atomicrmw or ptr addrspace(1) %sunkaddr3243, i32 %.lcssa4057 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %381, !amdgpu.uniform !42

381:                                              ; preds = %379, %ComputeEnd2438
  %382 = phi i32 [ %380, %379 ], [ poison, %ComputeEnd2438 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %2870)
  %383 = call i32 @llvm.amdgcn.readfirstlane.i32(i32 %382)
  %384 = or i32 %383, %.lcssa4058
  %shl5.i.i.i.i.i129 = shl nuw i32 1, %375
  %and.i.i.i.i.i130 = and i32 %shl5.i.i.i.i.i129, %384
  %tobool3.i.i.i.i131 = icmp ne i32 %and.i.i.i.i.i130, 0
  %385 = tail call i64 @llvm.amdgcn.ballot.i64(i1 %tobool3.i.i.i.i131)
  %and.i.i.i.i.i.i132 = and i64 %385, %374
  %cmp.i.not.i.i.i133 = icmp ne i64 %374, %and.i.i.i.i.i.i132
  br i1 %cmp.i.not.i.i.i133, label %if.then.i.i.i.i134, label %Flow3705, !amdgpu.uniform !42

if.then.i.i.i.i134:                               ; preds = %381
  fence syncscope("agent") acquire
  %386 = load <2 x i64>, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !75
  %387 = extractelement <2 x i64> %386, i32 0
  %388 = inttoptr i64 %387 to ptr
  %389 = extractelement <2 x i64> %386, i32 1
  %390 = inttoptr i64 %389 to ptr
  %idxprom.i.i.i.i135 = zext i32 %spec.store.select.i.i.i123 to i64
  %arrayidx.i.i.i.i136 = getelementptr inbounds nuw i32, ptr %388, i64 %idxprom.i.i.i.i135, !amdgpu.uniform !42
  %391 = load atomic i32, ptr %arrayidx.i.i.i.i136 monotonic, align 4
  %392 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %391)
  %arrayidx.i45.i.i.i137 = getelementptr inbounds nuw i32, ptr %390, i64 %idxprom.i.i.i.i135, !amdgpu.uniform !42
  %393 = load atomic i32, ptr %arrayidx.i45.i.i.i137 monotonic, align 4
  %394 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %393)
  %cmp.i46.not.i.i.i138 = icmp ne i32 %392, %394
  br i1 %cmp.i46.not.i.i.i138, label %if.then12.i.i.i139, label %Flow3703, !amdgpu.uniform !42

if.then12.i.i.i139:                               ; preds = %if.then.i.i.i.i134
  fence syncscope("agent") release
  %395 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %374, i1 true)
  %iszero.i.i.i.i.i.i.i140 = icmp ne i64 %374, 0
  %cmp2.i.i.i.i.i.i141 = icmp eq i64 %395, %sh_prom.i.i.i.i
  %cmp.i.i.i.i.i2.i = select i1 %iszero.i.i.i.i.i.i.i140, i1 %cmp2.i.i.i.i.i.i141, i1 false
  %conv4.i.i51.i.i.i142 = zext i1 %cmp.i.i.i.i.i2.i to i32
  %shl.i.i52.i.i.i143 = shl nuw i32 %conv4.i.i51.i.i.i142, %375
  %xor.i.i.i.i.i144 = xor i32 %shl.i.i52.i.i.i143, -1
  %396 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2442, !amdgpu.uniform !42

397:                                              ; preds = %ComputeEnd2443
  %sunkaddr3244 = getelementptr inbounds i8, ptr addrspace(1) %2861, i64 40
  %398 = atomicrmw and ptr addrspace(1) %sunkaddr3244, i32 %.lcssa4059 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %399, !amdgpu.uniform !42

399:                                              ; preds = %397, %ComputeEnd2443
  call void @llvm.amdgcn.end.cf.i64(i64 %2888)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %Flow3703, !amdgpu.uniform !42

if.end14.i.i.i204:                                ; preds = %Flow3703
  %400 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %374, i1 true)
  %iszero.i.i.i.i.i.i205 = icmp ne i64 %374, 0
  %cmp2.i.i.i.i.i206 = icmp eq i64 %400, %sh_prom.i.i.i.i
  %cmp.i.i.i.i.i207 = select i1 %iszero.i.i.i.i.i.i205, i1 %cmp2.i.i.i.i.i206, i1 false
  %401 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.i.i.i.i.i207)
  %402 = extractvalue { i1, i64 } %401, 0
  %403 = extractvalue { i1, i64 } %401, 1
  br i1 %402, label %if.then16.i.i.i209, label %if.end22.i.i.i208

if.then16.i.i.i209:                               ; preds = %if.end14.i.i.i204
  %404 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !76
  %opcode.i.i.i210 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %404, i64 %idxprom.i.i.i.i135, i32 1
  store i32 1660944387, ptr %opcode.i.i.i210, align 8, !tbaa !77
  %arrayidx21.i.i.i211 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %404, i64 %idxprom.i.i.i.i135
  store i64 %374, ptr %arrayidx21.i.i.i211, align 8, !tbaa !80
  br label %if.end22.i.i.i208, !amdgpu.uniform !42

Flow3704:                                         ; preds = %if.end22.i.i.i208, %Flow3703
  %405 = phi i32 [ %392, %if.end22.i.i.i208 ], [ %retval.sroa.8.0.i.i.i120, %Flow3703 ]
  %406 = phi i64 [ %374, %if.end22.i.i.i208 ], [ %retval.sroa.2.0.i.i.i119, %Flow3703 ]
  %407 = phi i32 [ %spec.store.select.i.i.i123, %if.end22.i.i.i208 ], [ %retval.sroa.6.0.i.i.i117, %Flow3703 ]
  br label %cleanup26.i.i.i145, !amdgpu.uniform !42

if.end22.i.i.i208:                                ; preds = %if.then16.i.i.i209, %if.end14.i.i.i204
  call void @llvm.amdgcn.end.cf.i64(i64 %403)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %Flow3704, !amdgpu.uniform !42

Flow3705:                                         ; preds = %cleanup26.i.i.i145, %381
  %408 = phi i32 [ %407, %cleanup26.i.i.i145 ], [ %370, %381 ]
  %409 = phi i64 [ %406, %cleanup26.i.i.i145 ], [ %371, %381 ]
  %410 = phi i32 [ %405, %cleanup26.i.i.i145 ], [ %372, %381 ]
  %411 = phi i32 [ %405, %cleanup26.i.i.i145 ], [ %retval.sroa.8.0.i.i.i120, %381 ]
  %412 = phi i64 [ %406, %cleanup26.i.i.i145 ], [ %retval.sroa.2.0.i.i.i119, %381 ]
  %413 = phi i32 [ %407, %cleanup26.i.i.i145 ], [ %retval.sroa.6.0.i.i.i117, %381 ]
  %414 = phi i1 [ %cmp.i46.not.i.i.i138, %cleanup26.i.i.i145 ], [ true, %381 ]
  br i1 %414, label %for.inc.i.i.i150, label %Flow3706, !amdgpu.uniform !42

cleanup26.i.i.i145:                               ; preds = %Flow3704
  br label %Flow3705, !amdgpu.uniform !42

for.inc.i.i.i150:                                 ; preds = %Flow3705
  %inc.i.i.i155 = add i32 %spec.store.select.i.i.i123, 1
  br label %Flow3706, !amdgpu.uniform !42

Flow3706:                                         ; preds = %for.inc.i.i.i150, %Flow3705
  %415 = phi i32 [ %inc.i.i.i155, %for.inc.i.i.i150 ], [ poison, %Flow3705 ]
  %416 = phi i1 [ false, %for.inc.i.i.i150 ], [ true, %Flow3705 ]
  br i1 %416, label %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i156, label %for.cond.i.i.i116, !amdgpu.uniform !42

_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i156: ; preds = %Flow3706
  %.lcssa4062 = phi i32 [ %408, %Flow3706 ]
  %.lcssa4061 = phi i64 [ %409, %Flow3706 ]
  %.lcssa4060 = phi i32 [ %410, %Flow3706 ]
  fence acquire
  %417 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %mul.i.i.i.i.i.i157 = shl i32 %.lcssa4062, 6
  %idxprom.i19.i.i.i.i.i158 = zext i32 %mul.i.i.i.i.i.i157 to i64
  %arrayidx.i20.i.i.i.i.i159 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %417, i64 %idxprom.i19.i.i.i.i.i158
  %arrayidx.i22.i.i.i.i.i160 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %arrayidx.i20.i.i.i.i.i159, i64 %sh_prom.i.i.i.i
  store i64 %writer.sroa.10.5.i.i.i.i, ptr %arrayidx.i22.i.i.i.i.i160, align 8, !tbaa !82
  %spec.select.i.i.i.i.i.i.i161 = tail call i64 @llvm.umin.i64(i64 %writer.sroa.10.5.i.i.i.i, i64 56)
  %418 = and i64 %spec.select.i.i.i.i.i.i.i161, 15
  %419 = sub i64 %spec.select.i.i.i.i.i.i.i161, %418
  %420 = icmp ne i64 %419, 0
  %421 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %420)
  %422 = extractvalue { i1, i64 } %421, 0
  %423 = extractvalue { i1, i64 } %421, 1
  br i1 %422, label %loop-memcpy-expansion2267.preheader, label %Flow3702

loop-memcpy-expansion2267.preheader:              ; preds = %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i156
  %424 = add nuw nsw i64 %76, 8
  %scevgep3159 = getelementptr i8, ptr %417, i64 %424
  %425 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i158, 6
  %scevgep3160 = getelementptr i8, ptr %scevgep3159, i64 %425
  br label %loop-memcpy-expansion2267, !amdgpu.uniform !42

Flow3702:                                         ; preds = %Flow3701, %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i156
  call void @llvm.amdgcn.end.cf.i64(i64 %423)
  br label %loop-memcpy-residual-header2270, !amdgpu.uniform !42

loop-memcpy-expansion2267:                        ; preds = %loop-memcpy-expansion2267.preheader, %loop-memcpy-expansion2267
  %phi.broken3784 = phi i64 [ 0, %loop-memcpy-expansion2267.preheader ], [ %430, %loop-memcpy-expansion2267 ]
  %loop-index2268 = phi i64 [ %428, %loop-memcpy-expansion2267 ], [ 0, %loop-memcpy-expansion2267.preheader ]
  %426 = getelementptr i8, ptr %341, i64 %loop-index2268
  %427 = load <4 x i32>, ptr %426, align 1
  %scevgep3161 = getelementptr i8, ptr %scevgep3160, i64 %loop-index2268
  store <4 x i32> %427, ptr %scevgep3161, align 1
  %428 = add i64 %loop-index2268, 16
  %429 = icmp uge i64 %428, %419
  %430 = call i64 @llvm.amdgcn.if.break.i64(i1 %429, i64 %phi.broken3784)
  %431 = call i1 @llvm.amdgcn.loop.i64(i64 %430)
  br i1 %431, label %Flow3701, label %loop-memcpy-expansion2267

Flow3700:                                         ; preds = %Flow3698, %Flow3699
  %432 = phi i64 [ %2644, %Flow3698 ], [ %2636, %Flow3699 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %2639)
  br label %post-loop-memcpy-expansion2266, !amdgpu.uniform !42

loop-memcpy-residual2269:                         ; preds = %loop-memcpy-residual2269.preheader, %loop-memcpy-residual2269
  %phi.broken3785 = phi i64 [ 0, %loop-memcpy-residual2269.preheader ], [ %436, %loop-memcpy-residual2269 ]
  %residual-loop-index2271 = phi i64 [ %434, %loop-memcpy-residual2269 ], [ 0, %loop-memcpy-residual2269.preheader ]
  %scevgep3158 = getelementptr i8, ptr %scevgep3157, i64 %residual-loop-index2271
  %433 = load i8, ptr %scevgep3158, align 1
  %scevgep3156 = getelementptr i8, ptr %scevgep3155, i64 %residual-loop-index2271
  store i8 %433, ptr %scevgep3156, align 1
  %434 = add i64 %residual-loop-index2271, 1
  %435 = icmp uge i64 %434, %418
  %436 = call i64 @llvm.amdgcn.if.break.i64(i1 %435, i64 %phi.broken3785)
  %437 = call i1 @llvm.amdgcn.loop.i64(i64 %436)
  br i1 %437, label %Flow3698, label %loop-memcpy-residual2269

Flow3698:                                         ; preds = %loop-memcpy-residual2269
  %.lcssa4055 = phi i64 [ %436, %loop-memcpy-residual2269 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa4055)
  br label %Flow3700, !amdgpu.uniform !42

post-loop-memcpy-expansion2266:                   ; preds = %Flow3700
  %tobool.not.i.i.i.i.i.i163 = icmp eq i32 %.lcssa4060, 0
  %conv.i.i.i.i.i.i164 = zext i1 %tobool.not.i.i.i.i.i.i163 to i32
  fence release
  %438 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %idxprom.i23.i.i.i.i.i165 = zext i32 %.lcssa4062 to i64
  %arrayidx.i24.i.i.i.i.i166 = getelementptr inbounds nuw i32, ptr %438, i64 %idxprom.i23.i.i.i.i.i165
  store atomic i32 %conv.i.i.i.i.i.i164, ptr %arrayidx.i24.i.i.i.i.i166 monotonic, align 4
  %439 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !84
  %arrayidx.i.i5.i.i167 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %439, i64 %idxprom.i23.i.i.i.i.i165, !amdgpu.uniform !42
  %440 = load i64, ptr %arrayidx.i.i5.i.i167, align 8, !tbaa !80
  %scevgep3141 = getelementptr i8, ptr %341, i64 56
  %441 = add i64 %writer.sroa.10.5.i.i.i.i, -56
  br label %while.cond.i.i.i.i168, !amdgpu.uniform !42

while.cond.i.i.i.i168:                            ; preds = %post-loop-memcpy-expansion2266, %Flow3697
  %phi.broken3788 = phi i64 [ 0, %post-loop-memcpy-expansion2266 ], [ %458, %Flow3697 ]
  %lsr.iv3147 = phi i64 [ %456, %Flow3697 ], [ %441, %post-loop-memcpy-expansion2266 ]
  %lsr.iv3142 = phi ptr [ %455, %Flow3697 ], [ %scevgep3141, %post-loop-memcpy-expansion2266 ]
  %port.sroa.37.0.i.i169 = phi i32 [ %454, %Flow3697 ], [ %conv.i.i.i.i.i.i164, %post-loop-memcpy-expansion2266 ]
  %idx.0.i.i.i.i170 = phi i64 [ %453, %Flow3697 ], [ 56, %post-loop-memcpy-expansion2266 ]
  %cmp.i.i.i.i171 = icmp ugt i64 %writer.sroa.10.5.i.i.i.i, %idx.0.i.i.i.i170
  %442 = tail call i64 @llvm.amdgcn.ballot.i64(i1 %cmp.i.i.i.i171)
  %and.i.i.i.i6.i.i172 = and i64 %442, %440
  %tobool.not.i.i.i.i173 = icmp ne i64 %and.i.i.i.i6.i.i172, 0
  %443 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i22.i.i.i174 = getelementptr inbounds nuw i32, ptr %443, i64 %idxprom.i23.i.i.i.i.i165, !amdgpu.uniform !42
  %444 = load atomic i32, ptr %arrayidx.i22.i.i.i174 monotonic, align 4
  %445 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %tobool.not.i.i.i.i173)
  %446 = extractvalue { i1, i64 } %445, 0
  %447 = extractvalue { i1, i64 } %445, 1
  br i1 %446, label %cond.false.i8.i.i.i.i175, label %Flow3697

cond.false.i8.i.i.i.i175:                         ; preds = %while.cond.i.i.i.i168
  %448 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %444)
  %449 = icmp ne i32 %448, %port.sroa.37.0.i.i169
  br i1 %449, label %while.body.i.i21.i.i.i.i176.preheader, label %Flow3696, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i176.preheader:            ; preds = %cond.false.i8.i.i.i.i175
  br label %while.body.i.i21.i.i.i.i176, !amdgpu.uniform !42

Flow3696:                                         ; preds = %Flow3695, %cond.false.i8.i.i.i.i175
  br label %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i179, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i176:                      ; preds = %while.body.i.i21.i.i.i.i176.preheader, %while.body.i.i21.i.i.i.i176
  tail call void @llvm.amdgcn.s.sleep(i32 2)
  %450 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i22.i.i.i.i177 = getelementptr inbounds nuw i32, ptr %450, i64 %idxprom.i23.i.i.i.i.i165, !amdgpu.uniform !42
  %451 = load atomic i32, ptr %arrayidx.i.i.i22.i.i.i.i177 monotonic, align 4
  %452 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %451)
  %cmp.i.not.i.i23.i.i.i.i178 = icmp eq i32 %452, %port.sroa.37.0.i.i169
  br i1 %cmp.i.not.i.i23.i.i.i.i178, label %Flow3695, label %while.body.i.i21.i.i.i.i176, !amdgpu.uniform !42

Flow3695:                                         ; preds = %while.body.i.i21.i.i.i.i176
  br label %Flow3696, !amdgpu.uniform !42

Flow3697:                                         ; preds = %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i180, %while.cond.i.i.i.i168
  %453 = phi i64 [ %add.i.i.i.i183, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i180 ], [ poison, %while.cond.i.i.i.i168 ]
  %454 = phi i32 [ %conv.i.i26.i.i.i.i181, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i180 ], [ poison, %while.cond.i.i.i.i168 ]
  %455 = phi ptr [ %scevgep3143, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i180 ], [ poison, %while.cond.i.i.i.i168 ]
  %456 = phi i64 [ %lsr.iv.next3148, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i180 ], [ poison, %while.cond.i.i.i.i168 ]
  %457 = phi i1 [ false, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i180 ], [ true, %while.cond.i.i.i.i168 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %447)
  %458 = call i64 @llvm.amdgcn.if.break.i64(i1 %457, i64 %phi.broken3788)
  %459 = call i1 @llvm.amdgcn.loop.i64(i64 %458)
  br i1 %459, label %cond.false.i.i.i189, label %while.cond.i.i.i.i168

_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i179: ; preds = %Flow3696
  fence acquire
  %460 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.i.i.i.i171)
  %461 = extractvalue { i1, i64 } %460, 0
  %462 = extractvalue { i1, i64 } %460, 1
  br i1 %461, label %if.then.i.i.i.i.i.i3.i, label %Flow3694

if.then.i.i.i.i.i.i3.i:                           ; preds = %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i179
  %sub.i.i.i.i.i.i.i185 = sub nuw i64 %writer.sroa.10.5.i.i.i.i, %idx.0.i.i.i.i170
  %spec.select.i.i.i36.i.i.i.i186 = tail call i64 @llvm.umin.i64(i64 %sub.i.i.i.i.i.i.i185, i64 64)
  %463 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %464 = and i64 %spec.select.i.i.i36.i.i.i.i186, 15
  %465 = sub i64 %spec.select.i.i.i36.i.i.i.i186, %464
  %466 = icmp ne i64 %465, 0
  %467 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %466)
  %468 = extractvalue { i1, i64 } %467, 0
  %469 = extractvalue { i1, i64 } %467, 1
  br i1 %468, label %loop-memcpy-expansion2261.preheader, label %Flow3693

loop-memcpy-expansion2261.preheader:              ; preds = %if.then.i.i.i.i.i.i3.i
  %arrayidx.i20.i35.i.i.i.i187 = getelementptr %"struct.rpc::Buffer", ptr %463, i64 %idxprom.i19.i.i.i.i.i158
  %arrayidx.i22.i38.i.i.i.i188 = getelementptr %"struct.rpc::Buffer", ptr %arrayidx.i20.i35.i.i.i.i187, i64 %sh_prom.i.i.i.i
  br label %loop-memcpy-expansion2261, !amdgpu.uniform !42

Flow3693:                                         ; preds = %Flow3692, %if.then.i.i.i.i.i.i3.i
  call void @llvm.amdgcn.end.cf.i64(i64 %469)
  br label %loop-memcpy-residual-header2264, !amdgpu.uniform !42

loop-memcpy-expansion2261:                        ; preds = %loop-memcpy-expansion2261.preheader, %loop-memcpy-expansion2261
  %phi.broken3786 = phi i64 [ 0, %loop-memcpy-expansion2261.preheader ], [ %474, %loop-memcpy-expansion2261 ]
  %loop-index2262 = phi i64 [ %472, %loop-memcpy-expansion2261 ], [ 0, %loop-memcpy-expansion2261.preheader ]
  %scevgep3144 = getelementptr i8, ptr %lsr.iv3142, i64 %loop-index2262
  %470 = load <4 x i32>, ptr %scevgep3144, align 1
  %471 = getelementptr i8, ptr %arrayidx.i22.i38.i.i.i.i188, i64 %loop-index2262
  store <4 x i32> %470, ptr %471, align 1
  %472 = add i64 %loop-index2262, 16
  %473 = icmp uge i64 %472, %465
  %474 = call i64 @llvm.amdgcn.if.break.i64(i1 %473, i64 %phi.broken3786)
  %475 = call i1 @llvm.amdgcn.loop.i64(i64 %474)
  br i1 %475, label %Flow3692, label %loop-memcpy-expansion2261

Flow3691:                                         ; preds = %Flow3690, %loop-memcpy-residual-header2264
  call void @llvm.amdgcn.end.cf.i64(i64 %2629)
  br label %Flow3694, !amdgpu.uniform !42

loop-memcpy-residual2263:                         ; preds = %loop-memcpy-residual2263.preheader, %loop-memcpy-residual2263
  %phi.broken3787 = phi i64 [ 0, %loop-memcpy-residual2263.preheader ], [ %479, %loop-memcpy-residual2263 ]
  %residual-loop-index2265 = phi i64 [ %477, %loop-memcpy-residual2263 ], [ 0, %loop-memcpy-residual2263.preheader ]
  %scevgep3153 = getelementptr i8, ptr %scevgep3152, i64 %residual-loop-index2265
  %476 = load i8, ptr %scevgep3153, align 1
  %scevgep3151 = getelementptr i8, ptr %scevgep3150, i64 %residual-loop-index2265
  store i8 %476, ptr %scevgep3151, align 1
  %477 = add i64 %residual-loop-index2265, 1
  %478 = icmp uge i64 %477, %464
  %479 = call i64 @llvm.amdgcn.if.break.i64(i1 %478, i64 %phi.broken3787)
  %480 = call i1 @llvm.amdgcn.loop.i64(i64 %479)
  br i1 %480, label %Flow3690, label %loop-memcpy-residual2263

Flow3690:                                         ; preds = %loop-memcpy-residual2263
  %.lcssa4052 = phi i64 [ %479, %loop-memcpy-residual2263 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa4052)
  br label %Flow3691, !amdgpu.uniform !42

_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i180: ; preds = %Flow3694
  %conv.i.i26.i.i.i.i181 = xor i32 %port.sroa.37.0.i.i169, 1
  fence release
  %481 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %arrayidx.i24.i29.i.i.i.i182 = getelementptr inbounds nuw i32, ptr %481, i64 %idxprom.i23.i.i.i.i.i165
  store atomic i32 %conv.i.i26.i.i.i.i181, ptr %arrayidx.i24.i29.i.i.i.i182 monotonic, align 4
  %add.i.i.i.i183 = add i64 %idx.0.i.i.i.i170, 64
  %scevgep3143 = getelementptr i8, ptr %lsr.iv3142, i64 64
  %lsr.iv.next3148 = add i64 %lsr.iv3147, -64
  br label %Flow3697, !amdgpu.uniform !42

cond.false.i.i.i189:                              ; preds = %Flow3697
  %.lcssa4054 = phi i64 [ %458, %Flow3697 ]
  %port.sroa.37.0.i.i169.lcssa = phi i32 [ %port.sroa.37.0.i.i169, %Flow3697 ]
  %.lcssa4053 = phi i32 [ %444, %Flow3697 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa4054)
  %482 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %.lcssa4053)
  %483 = icmp ne i32 %482, %port.sroa.37.0.i.i169.lcssa
  %484 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %483)
  %485 = extractvalue { i1, i64 } %484, 0
  %486 = extractvalue { i1, i64 } %484, 1
  br i1 %485, label %while.body.i.i.i.i190.preheader, label %Flow3689

while.body.i.i.i.i190.preheader:                  ; preds = %cond.false.i.i.i189
  br label %while.body.i.i.i.i190, !amdgpu.uniform !42

Flow3689:                                         ; preds = %Flow3688, %cond.false.i.i.i189
  call void @llvm.amdgcn.end.cf.i64(i64 %486)
  br label %_ZN22__llvm_libc_22_0_0_git15write_to_stderrENS_3cpp11string_viewE.exit.i, !amdgpu.uniform !42

while.body.i.i.i.i190:                            ; preds = %while.body.i.i.i.i190.preheader, %while.body.i.i.i.i190
  %phi.broken3789 = phi i64 [ 0, %while.body.i.i.i.i190.preheader ], [ %490, %while.body.i.i.i.i190 ]
  tail call void @llvm.amdgcn.s.sleep(i32 2)
  %487 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i8.i.i191 = getelementptr inbounds nuw i32, ptr %487, i64 %idxprom.i23.i.i.i.i.i165, !amdgpu.uniform !42
  %488 = load atomic i32, ptr %arrayidx.i.i.i8.i.i191 monotonic, align 4
  %489 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %488)
  %cmp.i.not.i.i.i.i192 = icmp eq i32 %489, %port.sroa.37.0.i.i169.lcssa
  %490 = call i64 @llvm.amdgcn.if.break.i64(i1 %cmp.i.not.i.i.i.i192, i64 %phi.broken3789)
  %491 = call i1 @llvm.amdgcn.loop.i64(i64 %490)
  br i1 %491, label %Flow3688, label %while.body.i.i.i.i190

Flow3688:                                         ; preds = %while.body.i.i.i.i190
  %.lcssa4050 = phi i64 [ %490, %while.body.i.i.i.i190 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa4050)
  br label %Flow3689, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git15write_to_stderrENS_3cpp11string_viewE.exit.i: ; preds = %Flow3689
  fence acquire
  tail call void @llvm.amdgcn.wave.barrier()
  fence syncscope("agent") release
  %492 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %.lcssa4061, i1 true)
  %iszero.i.i.i.i.i13.i.i193 = icmp ne i64 %.lcssa4061, 0
  %cmp2.i.i.i.i14.i.i194 = icmp eq i64 %492, %sh_prom.i.i.i.i
  %cmp.i.i.i.i15.i.i195 = select i1 %iszero.i.i.i.i.i13.i.i193, i1 %cmp2.i.i.i.i14.i.i194, i1 false
  %493 = and i32 %.lcssa4062, 31
  %conv4.i.i.i19.i.i199 = zext i1 %cmp.i.i.i.i15.i.i195 to i32
  %shl.i.i.i20.i.i200 = shl nuw i32 %conv4.i.i.i19.i.i199, %493
  %xor.i.i.i21.i.i201 = xor i32 %shl.i.i.i20.i.i200, -1
  %494 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2446, !amdgpu.uniform !42

495:                                              ; preds = %ComputeEnd2447
  %div8.i.i.i16.i.i196 = lshr i32 %.lcssa4062, 5
  %idxprom.i.i.i17.i.i197 = zext nneg i32 %div8.i.i.i16.i.i196 to i64
  %496 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i17.i.i197
  %sunkaddr3245 = getelementptr inbounds i8, ptr addrspace(1) %496, i64 40
  %497 = atomicrmw and ptr addrspace(1) %sunkaddr3245, i32 %.lcssa4049 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %498, !amdgpu.uniform !42

498:                                              ; preds = %495, %ComputeEnd2447
  call void @llvm.amdgcn.end.cf.i64(i64 %2905)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %for.cond.i.i1.i220, !amdgpu.uniform !42

for.cond.i.i1.i220:                               ; preds = %498, %Flow3687
  %499 = phi i32 [ %537, %Flow3687 ], [ poison, %498 ]
  %500 = phi i64 [ %538, %Flow3687 ], [ poison, %498 ]
  %501 = phi i32 [ %539, %Flow3687 ], [ poison, %498 ]
  %retval.sroa.6.0.i.i.i221 = phi i32 [ %542, %Flow3687 ], [ undef, %498 ]
  %retval.sroa.2.0.i.i.i223 = phi i64 [ %541, %Flow3687 ], [ undef, %498 ]
  %retval.sroa.8.0.i.i.i224 = phi i32 [ %540, %Flow3687 ], [ undef, %498 ]
  %index.0.i.i.i225 = phi i32 [ %544, %Flow3687 ], [ 0, %498 ]
  %502 = load i32, ptr addrspace(1) @__llvm_rpc_client, align 8, !tbaa !68
  %cmp.not.i.i.i226 = icmp ult i32 %index.0.i.i.i225, %502
  %spec.store.select.i.i.i227 = select i1 %cmp.not.i.i.i226, i32 %index.0.i.i.i225, i32 0
  %503 = tail call noundef i64 @llvm.amdgcn.ballot.i64(i1 true)
  %504 = and i32 %spec.store.select.i.i.i227, 31
  %505 = lshr i64 %503, %sh_prom.i.i.i.i
  %506 = trunc i64 %505 to i32
  %conv4.i.i.i.i.i231 = and i32 %506, 1
  %shl.i.i.i.i.i232 = shl nuw i32 %conv4.i.i.i.i.i231, %504
  %507 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2450, !amdgpu.uniform !42

508:                                              ; preds = %ComputeEnd2451
  %sunkaddr3246 = getelementptr inbounds i8, ptr addrspace(1) %2915, i64 40
  %509 = atomicrmw or ptr addrspace(1) %sunkaddr3246, i32 %.lcssa4043 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %510, !amdgpu.uniform !42

510:                                              ; preds = %508, %ComputeEnd2451
  %511 = phi i32 [ %509, %508 ], [ poison, %ComputeEnd2451 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %2924)
  %512 = call i32 @llvm.amdgcn.readfirstlane.i32(i32 %511)
  %513 = or i32 %512, %.lcssa4044
  %shl5.i.i.i.i.i233 = shl nuw i32 1, %504
  %and.i.i.i.i.i234 = and i32 %shl5.i.i.i.i.i233, %513
  %tobool3.i.i.i.i235 = icmp ne i32 %and.i.i.i.i.i234, 0
  %514 = tail call i64 @llvm.amdgcn.ballot.i64(i1 %tobool3.i.i.i.i235)
  %and.i.i.i.i.i.i236 = and i64 %514, %503
  %cmp.i.not.i.i.i237 = icmp ne i64 %503, %and.i.i.i.i.i.i236
  br i1 %cmp.i.not.i.i.i237, label %if.then.i.i.i.i238, label %Flow3686, !amdgpu.uniform !42

if.then.i.i.i.i238:                               ; preds = %510
  fence syncscope("agent") acquire
  %515 = load <2 x i64>, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !75
  %516 = extractelement <2 x i64> %515, i32 0
  %517 = inttoptr i64 %516 to ptr
  %518 = extractelement <2 x i64> %515, i32 1
  %519 = inttoptr i64 %518 to ptr
  %idxprom.i.i.i.i239 = zext i32 %spec.store.select.i.i.i227 to i64
  %arrayidx.i.i.i.i240 = getelementptr inbounds nuw i32, ptr %517, i64 %idxprom.i.i.i.i239, !amdgpu.uniform !42
  %520 = load atomic i32, ptr %arrayidx.i.i.i.i240 monotonic, align 4
  %521 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %520)
  %arrayidx.i45.i.i.i241 = getelementptr inbounds nuw i32, ptr %519, i64 %idxprom.i.i.i.i239, !amdgpu.uniform !42
  %522 = load atomic i32, ptr %arrayidx.i45.i.i.i241 monotonic, align 4
  %523 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %522)
  %cmp.i46.not.i.i.i242 = icmp ne i32 %521, %523
  br i1 %cmp.i46.not.i.i.i242, label %if.then12.i.i.i243, label %Flow3684, !amdgpu.uniform !42

if.then12.i.i.i243:                               ; preds = %if.then.i.i.i.i238
  fence syncscope("agent") release
  %524 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %503, i1 true)
  %iszero.i.i.i.i.i.i.i244 = icmp ne i64 %503, 0
  %cmp2.i.i.i.i.i.i245 = icmp eq i64 %524, %sh_prom.i.i.i.i
  %cmp.i.i.i.i.i.i246 = select i1 %iszero.i.i.i.i.i.i.i244, i1 %cmp2.i.i.i.i.i.i245, i1 false
  %conv4.i.i51.i.i.i247 = zext i1 %cmp.i.i.i.i.i.i246 to i32
  %shl.i.i52.i.i.i248 = shl nuw i32 %conv4.i.i51.i.i.i247, %504
  %xor.i.i.i.i.i249 = xor i32 %shl.i.i52.i.i.i248, -1
  %525 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2455, !amdgpu.uniform !42

526:                                              ; preds = %ComputeEnd2456
  %sunkaddr3247 = getelementptr inbounds i8, ptr addrspace(1) %2915, i64 40
  %527 = atomicrmw and ptr addrspace(1) %sunkaddr3247, i32 %.lcssa4045 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %528, !amdgpu.uniform !42

528:                                              ; preds = %526, %ComputeEnd2456
  call void @llvm.amdgcn.end.cf.i64(i64 %2942)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %Flow3684, !amdgpu.uniform !42

if.end14.i.i.i309:                                ; preds = %Flow3684
  %529 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %503, i1 true)
  %iszero.i.i.i.i.i.i310 = icmp ne i64 %503, 0
  %cmp2.i.i.i.i.i311 = icmp eq i64 %529, %sh_prom.i.i.i.i
  %cmp.i.i.i.i.i312 = select i1 %iszero.i.i.i.i.i.i310, i1 %cmp2.i.i.i.i.i311, i1 false
  %530 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.i.i.i.i.i312)
  %531 = extractvalue { i1, i64 } %530, 0
  %532 = extractvalue { i1, i64 } %530, 1
  br i1 %531, label %if.then16.i.i.i314, label %if.end22.i.i.i313

if.then16.i.i.i314:                               ; preds = %if.end14.i.i.i309
  %533 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !76
  %opcode.i.i.i315 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %533, i64 %idxprom.i.i.i.i239, i32 1
  store i32 1660944387, ptr %opcode.i.i.i315, align 8, !tbaa !77
  %arrayidx21.i.i.i316 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %533, i64 %idxprom.i.i.i.i239
  store i64 %503, ptr %arrayidx21.i.i.i316, align 8, !tbaa !80
  br label %if.end22.i.i.i313, !amdgpu.uniform !42

Flow3685:                                         ; preds = %if.end22.i.i.i313, %Flow3684
  %534 = phi i32 [ %521, %if.end22.i.i.i313 ], [ %retval.sroa.8.0.i.i.i224, %Flow3684 ]
  %535 = phi i64 [ %503, %if.end22.i.i.i313 ], [ %retval.sroa.2.0.i.i.i223, %Flow3684 ]
  %536 = phi i32 [ %spec.store.select.i.i.i227, %if.end22.i.i.i313 ], [ %retval.sroa.6.0.i.i.i221, %Flow3684 ]
  br label %cleanup26.i.i.i250, !amdgpu.uniform !42

if.end22.i.i.i313:                                ; preds = %if.then16.i.i.i314, %if.end14.i.i.i309
  call void @llvm.amdgcn.end.cf.i64(i64 %532)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %Flow3685, !amdgpu.uniform !42

Flow3686:                                         ; preds = %cleanup26.i.i.i250, %510
  %537 = phi i32 [ %536, %cleanup26.i.i.i250 ], [ %499, %510 ]
  %538 = phi i64 [ %535, %cleanup26.i.i.i250 ], [ %500, %510 ]
  %539 = phi i32 [ %534, %cleanup26.i.i.i250 ], [ %501, %510 ]
  %540 = phi i32 [ %534, %cleanup26.i.i.i250 ], [ %retval.sroa.8.0.i.i.i224, %510 ]
  %541 = phi i64 [ %535, %cleanup26.i.i.i250 ], [ %retval.sroa.2.0.i.i.i223, %510 ]
  %542 = phi i32 [ %536, %cleanup26.i.i.i250 ], [ %retval.sroa.6.0.i.i.i221, %510 ]
  %543 = phi i1 [ %cmp.i46.not.i.i.i242, %cleanup26.i.i.i250 ], [ true, %510 ]
  br i1 %543, label %for.inc.i.i.i255, label %Flow3687, !amdgpu.uniform !42

cleanup26.i.i.i250:                               ; preds = %Flow3685
  br label %Flow3686, !amdgpu.uniform !42

for.inc.i.i.i255:                                 ; preds = %Flow3686
  %inc.i.i.i260 = add i32 %spec.store.select.i.i.i227, 1
  br label %Flow3687, !amdgpu.uniform !42

Flow3687:                                         ; preds = %for.inc.i.i.i255, %Flow3686
  %544 = phi i32 [ %inc.i.i.i260, %for.inc.i.i.i255 ], [ poison, %Flow3686 ]
  %545 = phi i1 [ false, %for.inc.i.i.i255 ], [ true, %Flow3686 ]
  br i1 %545, label %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i261, label %for.cond.i.i1.i220, !amdgpu.uniform !42

_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i261: ; preds = %Flow3687
  %.lcssa4048 = phi i32 [ %537, %Flow3687 ]
  %.lcssa4047 = phi i64 [ %538, %Flow3687 ]
  %.lcssa4046 = phi i32 [ %539, %Flow3687 ]
  fence acquire
  %546 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %mul.i.i.i.i.i.i262 = shl i32 %.lcssa4048, 6
  %idxprom.i19.i.i.i.i.i263 = zext i32 %mul.i.i.i.i.i.i262 to i64
  %arrayidx.i20.i.i.i.i.i264 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %546, i64 %idxprom.i19.i.i.i.i.i263
  %arrayidx.i22.i.i.i.i.i265 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %arrayidx.i20.i.i.i.i.i264, i64 %sh_prom.i.i.i.i
  store i64 sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.7.11 to ptr), i64 10) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.7.11 to ptr) to i64)), ptr %arrayidx.i22.i.i.i.i.i265, align 8, !tbaa !82
  %spec.select.i.i.i.i.i.i.i266 = tail call i64 @llvm.umin.i64(i64 sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.7.11 to ptr), i64 10) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.7.11 to ptr) to i64)), i64 56)
  %547 = and i64 %spec.select.i.i.i.i.i.i.i266, 15
  %548 = sub i64 %spec.select.i.i.i.i.i.i.i266, %547
  %549 = icmp ne i64 %548, 0
  br i1 %549, label %loop-memcpy-expansion2333.preheader, label %Flow3683, !amdgpu.uniform !42

loop-memcpy-expansion2333.preheader:              ; preds = %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i261
  %550 = add nuw nsw i64 %76, 8
  %scevgep3138 = getelementptr i8, ptr %546, i64 %550
  %551 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i263, 6
  %scevgep3139 = getelementptr i8, ptr %scevgep3138, i64 %551
  br label %loop-memcpy-expansion2333, !amdgpu.uniform !42

Flow3683:                                         ; preds = %Flow3682, %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i261
  br label %loop-memcpy-residual-header2336, !amdgpu.uniform !42

loop-memcpy-expansion2333:                        ; preds = %loop-memcpy-expansion2333.preheader, %loop-memcpy-expansion2333
  %loop-index2334 = phi i64 [ %554, %loop-memcpy-expansion2333 ], [ 0, %loop-memcpy-expansion2333.preheader ]
  %552 = getelementptr i8, ptr addrspace(4) @.str.7.11, i64 %loop-index2334, !amdgpu.uniform !42
  %553 = load <4 x i32>, ptr addrspace(4) %552, align 1
  %scevgep3140 = getelementptr i8, ptr %scevgep3139, i64 %loop-index2334
  store <4 x i32> %553, ptr %scevgep3140, align 1
  %554 = add i64 %loop-index2334, 16
  %555 = icmp uge i64 %554, %548
  br i1 %555, label %Flow3682, label %loop-memcpy-expansion2333, !amdgpu.uniform !42

Flow3681:                                         ; preds = %Flow3679, %Flow3680
  %556 = phi i64 [ %2690, %Flow3679 ], [ %2687, %Flow3680 ]
  br label %post-loop-memcpy-expansion2332, !amdgpu.uniform !42

loop-memcpy-residual2335:                         ; preds = %loop-memcpy-residual2335.preheader, %loop-memcpy-residual2335
  %residual-loop-index2337 = phi i64 [ %558, %loop-memcpy-residual2335 ], [ 0, %loop-memcpy-residual2335.preheader ]
  %scevgep3137 = getelementptr i8, ptr addrspace(4) @.str.7.11, i64 %residual-loop-index2337, !amdgpu.uniform !42
  %557 = load i8, ptr addrspace(4) %scevgep3137, align 1
  %scevgep3136 = getelementptr i8, ptr %scevgep3135, i64 %residual-loop-index2337
  store i8 %557, ptr %scevgep3136, align 1
  %558 = add i64 %residual-loop-index2337, 1
  %559 = icmp uge i64 %558, %547
  br i1 %559, label %Flow3679, label %loop-memcpy-residual2335, !amdgpu.uniform !42

Flow3679:                                         ; preds = %loop-memcpy-residual2335
  br label %Flow3681, !amdgpu.uniform !42

post-loop-memcpy-expansion2332:                   ; preds = %Flow3681
  %tobool.not.i.i.i.i.i.i268 = icmp eq i32 %.lcssa4046, 0
  %conv.i.i.i.i.i.i269 = zext i1 %tobool.not.i.i.i.i.i.i268 to i32
  fence release
  %560 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %idxprom.i23.i.i.i.i.i270 = zext i32 %.lcssa4048 to i64
  %arrayidx.i24.i.i.i.i.i271 = getelementptr inbounds nuw i32, ptr %560, i64 %idxprom.i23.i.i.i.i.i270
  store atomic i32 %conv.i.i.i.i.i.i269, ptr %arrayidx.i24.i.i.i.i.i271 monotonic, align 4
  %561 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !84
  %arrayidx.i.i5.i.i272 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %561, i64 %idxprom.i23.i.i.i.i.i270, !amdgpu.uniform !42
  %562 = load i64, ptr %arrayidx.i.i5.i.i272, align 8, !tbaa !80
  br label %while.cond.i.i.i.i273, !amdgpu.uniform !42

while.cond.i.i.i.i273:                            ; preds = %post-loop-memcpy-expansion2332, %Flow3678
  %phi.broken3790 = phi i64 [ 0, %post-loop-memcpy-expansion2332 ], [ %579, %Flow3678 ]
  %lsr.iv3127 = phi i64 [ %577, %Flow3678 ], [ -46, %post-loop-memcpy-expansion2332 ]
  %lsr.iv3122 = phi ptr addrspace(4) [ %576, %Flow3678 ], [ getelementptr (i8, ptr addrspace(4) @.str.7.11, i64 56), %post-loop-memcpy-expansion2332 ]
  %port.sroa.37.0.i.i274 = phi i32 [ %575, %Flow3678 ], [ %conv.i.i.i.i.i.i269, %post-loop-memcpy-expansion2332 ]
  %idx.0.i.i.i.i275 = phi i64 [ %574, %Flow3678 ], [ 56, %post-loop-memcpy-expansion2332 ]
  %cmp.i.i.i.i276 = icmp ult i64 %idx.0.i.i.i.i275, sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.7.11 to ptr), i64 10) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.7.11 to ptr) to i64))
  %563 = tail call i64 @llvm.amdgcn.ballot.i64(i1 %cmp.i.i.i.i276)
  %and.i.i.i.i6.i.i277 = and i64 %563, %562
  %tobool.not.i.i.i.i278 = icmp ne i64 %and.i.i.i.i6.i.i277, 0
  %564 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i22.i.i.i279 = getelementptr inbounds nuw i32, ptr %564, i64 %idxprom.i23.i.i.i.i.i270, !amdgpu.uniform !42
  %565 = load atomic i32, ptr %arrayidx.i22.i.i.i279 monotonic, align 4
  %566 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %tobool.not.i.i.i.i278)
  %567 = extractvalue { i1, i64 } %566, 0
  %568 = extractvalue { i1, i64 } %566, 1
  br i1 %567, label %cond.false.i8.i.i.i.i280, label %Flow3678

cond.false.i8.i.i.i.i280:                         ; preds = %while.cond.i.i.i.i273
  %569 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %565)
  %570 = icmp ne i32 %569, %port.sroa.37.0.i.i274
  br i1 %570, label %while.body.i.i21.i.i.i.i281.preheader, label %Flow3677, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i281.preheader:            ; preds = %cond.false.i8.i.i.i.i280
  br label %while.body.i.i21.i.i.i.i281, !amdgpu.uniform !42

Flow3677:                                         ; preds = %Flow3676, %cond.false.i8.i.i.i.i280
  br label %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i284, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i281:                      ; preds = %while.body.i.i21.i.i.i.i281.preheader, %while.body.i.i21.i.i.i.i281
  tail call void @llvm.amdgcn.s.sleep(i32 2)
  %571 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i22.i.i.i.i282 = getelementptr inbounds nuw i32, ptr %571, i64 %idxprom.i23.i.i.i.i.i270, !amdgpu.uniform !42
  %572 = load atomic i32, ptr %arrayidx.i.i.i22.i.i.i.i282 monotonic, align 4
  %573 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %572)
  %cmp.i.not.i.i23.i.i.i.i283 = icmp eq i32 %573, %port.sroa.37.0.i.i274
  br i1 %cmp.i.not.i.i23.i.i.i.i283, label %Flow3676, label %while.body.i.i21.i.i.i.i281, !amdgpu.uniform !42

Flow3676:                                         ; preds = %while.body.i.i21.i.i.i.i281
  br label %Flow3677, !amdgpu.uniform !42

Flow3678:                                         ; preds = %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i285, %while.cond.i.i.i.i273
  %574 = phi i64 [ %add.i.i.i.i288, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i285 ], [ poison, %while.cond.i.i.i.i273 ]
  %575 = phi i32 [ %conv.i.i26.i.i.i.i286, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i285 ], [ poison, %while.cond.i.i.i.i273 ]
  %576 = phi ptr addrspace(4) [ %scevgep3123, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i285 ], [ poison, %while.cond.i.i.i.i273 ]
  %577 = phi i64 [ %lsr.iv.next3128, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i285 ], [ poison, %while.cond.i.i.i.i273 ]
  %578 = phi i1 [ false, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i285 ], [ true, %while.cond.i.i.i.i273 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %568)
  %579 = call i64 @llvm.amdgcn.if.break.i64(i1 %578, i64 %phi.broken3790)
  %580 = call i1 @llvm.amdgcn.loop.i64(i64 %579)
  br i1 %580, label %cond.false.i.i.i295, label %while.cond.i.i.i.i273

_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i284: ; preds = %Flow3677
  fence acquire
  br i1 %cmp.i.i.i.i276, label %if.then.i.i.i.i.i.i.i289, label %Flow3675, !amdgpu.uniform !42

if.then.i.i.i.i.i.i.i289:                         ; preds = %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i284
  %sub.i.i.i.i.i.i.i291 = sub nuw i64 sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.7.11 to ptr), i64 10) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.7.11 to ptr) to i64)), %idx.0.i.i.i.i275
  %spec.select.i.i.i36.i.i.i.i292 = tail call i64 @llvm.umin.i64(i64 %sub.i.i.i.i.i.i.i291, i64 64)
  %581 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %582 = and i64 %spec.select.i.i.i36.i.i.i.i292, 15
  %583 = sub i64 %spec.select.i.i.i36.i.i.i.i292, %582
  %584 = icmp ne i64 %583, 0
  br i1 %584, label %loop-memcpy-expansion2339.preheader, label %Flow3674, !amdgpu.uniform !42

loop-memcpy-expansion2339.preheader:              ; preds = %if.then.i.i.i.i.i.i.i289
  %arrayidx.i20.i35.i.i.i.i293 = getelementptr %"struct.rpc::Buffer", ptr %581, i64 %idxprom.i19.i.i.i.i.i263
  %arrayidx.i22.i38.i.i.i.i294 = getelementptr %"struct.rpc::Buffer", ptr %arrayidx.i20.i35.i.i.i.i293, i64 %sh_prom.i.i.i.i
  br label %loop-memcpy-expansion2339, !amdgpu.uniform !42

Flow3674:                                         ; preds = %Flow3673, %if.then.i.i.i.i.i.i.i289
  br label %loop-memcpy-residual-header2342, !amdgpu.uniform !42

loop-memcpy-expansion2339:                        ; preds = %loop-memcpy-expansion2339.preheader, %loop-memcpy-expansion2339
  %loop-index2340 = phi i64 [ %587, %loop-memcpy-expansion2339 ], [ 0, %loop-memcpy-expansion2339.preheader ]
  %scevgep3124 = getelementptr i8, ptr addrspace(4) %lsr.iv3122, i64 %loop-index2340, !amdgpu.uniform !42
  %585 = load <4 x i32>, ptr addrspace(4) %scevgep3124, align 1
  %586 = getelementptr i8, ptr %arrayidx.i22.i38.i.i.i.i294, i64 %loop-index2340
  store <4 x i32> %585, ptr %586, align 1
  %587 = add i64 %loop-index2340, 16
  %588 = icmp uge i64 %587, %583
  br i1 %588, label %Flow3673, label %loop-memcpy-expansion2339, !amdgpu.uniform !42

Flow3672:                                         ; preds = %Flow3671, %loop-memcpy-residual-header2342
  br label %Flow3675, !amdgpu.uniform !42

loop-memcpy-residual2341:                         ; preds = %loop-memcpy-residual2341.preheader, %loop-memcpy-residual2341
  %residual-loop-index2343 = phi i64 [ %590, %loop-memcpy-residual2341 ], [ 0, %loop-memcpy-residual2341.preheader ]
  %scevgep3133 = getelementptr i8, ptr addrspace(4) %scevgep3132, i64 %residual-loop-index2343, !amdgpu.uniform !42
  %589 = load i8, ptr addrspace(4) %scevgep3133, align 1
  %scevgep3131 = getelementptr i8, ptr %scevgep3130, i64 %residual-loop-index2343
  store i8 %589, ptr %scevgep3131, align 1
  %590 = add i64 %residual-loop-index2343, 1
  %591 = icmp uge i64 %590, %582
  br i1 %591, label %Flow3671, label %loop-memcpy-residual2341, !amdgpu.uniform !42

Flow3671:                                         ; preds = %loop-memcpy-residual2341
  br label %Flow3672, !amdgpu.uniform !42

_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i285: ; preds = %Flow3675
  %conv.i.i26.i.i.i.i286 = xor i32 %port.sroa.37.0.i.i274, 1
  fence release
  %592 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %arrayidx.i24.i29.i.i.i.i287 = getelementptr inbounds nuw i32, ptr %592, i64 %idxprom.i23.i.i.i.i.i270
  store atomic i32 %conv.i.i26.i.i.i.i286, ptr %arrayidx.i24.i29.i.i.i.i287 monotonic, align 4
  %add.i.i.i.i288 = add i64 %idx.0.i.i.i.i275, 64
  %scevgep3123 = getelementptr i8, ptr addrspace(4) %lsr.iv3122, i64 64
  %lsr.iv.next3128 = add i64 %lsr.iv3127, -64
  br label %Flow3678, !amdgpu.uniform !42

cond.false.i.i.i295:                              ; preds = %Flow3678
  %.lcssa4042 = phi i64 [ %579, %Flow3678 ]
  %port.sroa.37.0.i.i274.lcssa = phi i32 [ %port.sroa.37.0.i.i274, %Flow3678 ]
  %.lcssa4041 = phi i32 [ %565, %Flow3678 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa4042)
  %593 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %.lcssa4041)
  %594 = icmp ne i32 %593, %port.sroa.37.0.i.i274.lcssa
  %595 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %594)
  %596 = extractvalue { i1, i64 } %595, 0
  %597 = extractvalue { i1, i64 } %595, 1
  br i1 %596, label %while.body.i.i.i.i296.preheader, label %Flow3670

while.body.i.i.i.i296.preheader:                  ; preds = %cond.false.i.i.i295
  br label %while.body.i.i.i.i296, !amdgpu.uniform !42

Flow3670:                                         ; preds = %Flow3669, %cond.false.i.i.i295
  call void @llvm.amdgcn.end.cf.i64(i64 %597)
  br label %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit317, !amdgpu.uniform !42

while.body.i.i.i.i296:                            ; preds = %while.body.i.i.i.i296.preheader, %while.body.i.i.i.i296
  %phi.broken3791 = phi i64 [ 0, %while.body.i.i.i.i296.preheader ], [ %601, %while.body.i.i.i.i296 ]
  tail call void @llvm.amdgcn.s.sleep(i32 2)
  %598 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i8.i.i297 = getelementptr inbounds nuw i32, ptr %598, i64 %idxprom.i23.i.i.i.i.i270, !amdgpu.uniform !42
  %599 = load atomic i32, ptr %arrayidx.i.i.i8.i.i297 monotonic, align 4
  %600 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %599)
  %cmp.i.not.i.i.i.i298 = icmp eq i32 %600, %port.sroa.37.0.i.i274.lcssa
  %601 = call i64 @llvm.amdgcn.if.break.i64(i1 %cmp.i.not.i.i.i.i298, i64 %phi.broken3791)
  %602 = call i1 @llvm.amdgcn.loop.i64(i64 %601)
  br i1 %602, label %Flow3669, label %while.body.i.i.i.i296

Flow3669:                                         ; preds = %while.body.i.i.i.i296
  %.lcssa4040 = phi i64 [ %601, %while.body.i.i.i.i296 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa4040)
  br label %Flow3670, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit317: ; preds = %Flow3670
  fence acquire
  tail call void @llvm.amdgcn.wave.barrier()
  fence syncscope("agent") release
  %603 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %.lcssa4047, i1 true)
  %iszero.i.i.i.i.i13.i.i300 = icmp ne i64 %.lcssa4047, 0
  %cmp2.i.i.i.i14.i.i301 = icmp eq i64 %603, %sh_prom.i.i.i.i
  %cmp.i.i.i.i15.i.i302 = select i1 %iszero.i.i.i.i.i13.i.i300, i1 %cmp2.i.i.i.i14.i.i301, i1 false
  %604 = and i32 %.lcssa4048, 31
  %conv4.i.i.i19.i.i306 = zext i1 %cmp.i.i.i.i15.i.i302 to i32
  %shl.i.i.i20.i.i307 = shl nuw i32 %conv4.i.i.i19.i.i306, %604
  %xor.i.i.i21.i.i308 = xor i32 %shl.i.i.i20.i.i307, -1
  %605 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2459, !amdgpu.uniform !42

606:                                              ; preds = %ComputeEnd2460
  %div8.i.i.i16.i.i303 = lshr i32 %.lcssa4048, 5
  %idxprom.i.i.i17.i.i304 = zext nneg i32 %div8.i.i.i16.i.i303 to i64
  %607 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i17.i.i304
  %sunkaddr3248 = getelementptr inbounds i8, ptr addrspace(1) %607, i64 40
  %608 = atomicrmw and ptr addrspace(1) %sunkaddr3248, i32 %.lcssa4039 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %609, !amdgpu.uniform !42

609:                                              ; preds = %606, %ComputeEnd2460
  call void @llvm.amdgcn.end.cf.i64(i64 %2959)
  tail call void @llvm.amdgcn.wave.barrier()
  %inc.i.i.i34.i = add nuw i64 %spec.select.i, 1
  %cmp2.i.i.i37.i = icmp ult i64 %spec.select.i, 1676976733973595600
  br i1 %cmp2.i.i.i37.i, label %if.then3.i.i.i46.i, label %if.end4.i.i.i38.i, !amdgpu.uniform !42

if.then3.i.i.i46.i:                               ; preds = %609
  %mul.i.i.i47.i = mul nuw i64 %inc.i.i.i34.i, 11
  %div18.i.i.i48.i = lshr i64 %mul.i.i.i47.i, 3
  br label %if.end4.i.i.i38.i, !amdgpu.uniform !42

if.end4.i.i.i38.i:                                ; preds = %if.then3.i.i.i46.i, %609
  %new_capacity.addr.0.i.i.i39.i = phi i64 [ %div18.i.i.i48.i, %if.then3.i.i.i46.i ], [ %inc.i.i.i34.i, %609 ]
  %sub.i.i318 = add i64 %new_capacity.addr.0.i.i.i39.i, 7
  %div2.i.i319 = and i64 %sub.i.i318, -8
  %610 = load ptr, ptr addrspace(1) @_ZN12_GLOBAL__N_13ptrE, align 8, !tbaa !55
  %add.ptr.i.i320 = getelementptr inbounds nuw i8, ptr %610, i64 %div2.i.i319
  store ptr %add.ptr.i.i320, ptr addrspace(1) @_ZN12_GLOBAL__N_13ptrE, align 8, !tbaa !55
  %611 = and i64 %spec.select.i, 15
  %612 = sub i64 %spec.select.i, %611
  %613 = icmp ne i64 %612, 0
  br i1 %613, label %loop-memcpy-expansion2285.preheader, label %Flow3668, !amdgpu.uniform !42

loop-memcpy-expansion2285.preheader:              ; preds = %if.end4.i.i.i38.i
  br label %loop-memcpy-expansion2285, !amdgpu.uniform !42

Flow3668:                                         ; preds = %Flow3667, %if.end4.i.i.i38.i
  br label %loop-memcpy-residual-header2288, !amdgpu.uniform !42

loop-memcpy-expansion2285:                        ; preds = %loop-memcpy-expansion2285.preheader, %loop-memcpy-expansion2285
  %loop-index2286 = phi i64 [ %617, %loop-memcpy-expansion2285 ], [ 0, %loop-memcpy-expansion2285.preheader ]
  %614 = getelementptr i8, ptr %4, i64 %loop-index2286, !amdgpu.uniform !42
  %615 = load <4 x i32>, ptr %614, align 1
  %616 = getelementptr i8, ptr %610, i64 %loop-index2286
  store <4 x i32> %615, ptr %616, align 1
  %617 = add i64 %loop-index2286, 16
  %618 = icmp uge i64 %617, %612
  br i1 %618, label %Flow3667, label %loop-memcpy-expansion2285, !amdgpu.uniform !42

Flow3666:                                         ; preds = %Flow3665, %loop-memcpy-residual-header2288
  br label %post-loop-memcpy-expansion2284, !amdgpu.uniform !42

loop-memcpy-residual2287:                         ; preds = %loop-memcpy-residual2287.preheader, %loop-memcpy-residual2287
  %residual-loop-index2289 = phi i64 [ %620, %loop-memcpy-residual2287 ], [ 0, %loop-memcpy-residual2287.preheader ]
  %scevgep3121 = getelementptr i8, ptr %scevgep3120, i64 %residual-loop-index2289, !amdgpu.uniform !42
  %619 = load i8, ptr %scevgep3121, align 1
  %scevgep3119 = getelementptr i8, ptr %scevgep3118, i64 %residual-loop-index2289
  store i8 %619, ptr %scevgep3119, align 1
  %620 = add i64 %residual-loop-index2289, 1
  %621 = icmp uge i64 %620, %611
  br i1 %621, label %Flow3665, label %loop-memcpy-residual2287, !amdgpu.uniform !42

Flow3665:                                         ; preds = %loop-memcpy-residual2287
  br label %Flow3666, !amdgpu.uniform !42

post-loop-memcpy-expansion2284:                   ; preds = %Flow3666
  %sub.ptr.lhs.cast.i.i321 = ptrtoint ptr %add.ptr.i.i320 to i64
  %sub.ptr.sub.i.i322 = sub i64 %sub.ptr.lhs.cast.i.i321, ptrtoint (ptr addrspacecast (ptr addrspace(1) @_ZN12_GLOBAL__N_16memoryE to ptr) to i64)
  %cmp.i.i323 = icmp ugt i64 %sub.ptr.sub.i.i322, 65335
  %cond.i.i324 = select i1 %cmp.i.i323, ptr null, ptr %610
  %cmp.not.i10.i.i.i = icmp ne ptr %cond.i.i324, addrspacecast (ptr addrspace(4) @_ZN22__llvm_libc_22_0_0_git3cpp6string14NULL_CHARACTERE to ptr)
  br i1 %cmp.not.i10.i.i.i, label %if.then.i.i.i44.i, label %for.cond.i.i.i326.preheader, !amdgpu.uniform !42

if.then.i.i.i44.i:                                ; preds = %post-loop-memcpy-expansion2284
  %arrayidx.i.i.i45.i = getelementptr inbounds nuw i8, ptr %610, i64 %spec.select.i
  store i8 0, ptr %arrayidx.i.i.i45.i, align 1, !tbaa !57
  br label %for.cond.i.i.i326.preheader, !amdgpu.uniform !42

for.cond.i.i.i326.preheader:                      ; preds = %if.then.i.i.i44.i, %post-loop-memcpy-expansion2284
  br label %for.cond.i.i.i326, !amdgpu.uniform !42

for.cond.i.i.i326:                                ; preds = %for.cond.i.i.i326.preheader, %Flow3664
  %622 = phi i32 [ %660, %Flow3664 ], [ poison, %for.cond.i.i.i326.preheader ]
  %623 = phi i64 [ %661, %Flow3664 ], [ poison, %for.cond.i.i.i326.preheader ]
  %624 = phi i32 [ %662, %Flow3664 ], [ poison, %for.cond.i.i.i326.preheader ]
  %retval.sroa.6.0.i.i.i327 = phi i32 [ %665, %Flow3664 ], [ undef, %for.cond.i.i.i326.preheader ]
  %retval.sroa.2.0.i.i.i329 = phi i64 [ %664, %Flow3664 ], [ undef, %for.cond.i.i.i326.preheader ]
  %retval.sroa.8.0.i.i.i330 = phi i32 [ %663, %Flow3664 ], [ undef, %for.cond.i.i.i326.preheader ]
  %index.0.i.i.i331 = phi i32 [ %667, %Flow3664 ], [ 0, %for.cond.i.i.i326.preheader ]
  %625 = load i32, ptr addrspace(1) @__llvm_rpc_client, align 8, !tbaa !68
  %cmp.not.i.i.i332 = icmp ult i32 %index.0.i.i.i331, %625
  %spec.store.select.i.i.i333 = select i1 %cmp.not.i.i.i332, i32 %index.0.i.i.i331, i32 0
  %626 = tail call noundef i64 @llvm.amdgcn.ballot.i64(i1 true)
  %627 = and i32 %spec.store.select.i.i.i333, 31
  %628 = lshr i64 %626, %sh_prom.i.i.i.i
  %629 = trunc i64 %628 to i32
  %conv4.i.i.i.i.i337 = and i32 %629, 1
  %shl.i.i.i.i.i338 = shl nuw i32 %conv4.i.i.i.i.i337, %627
  %630 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2463, !amdgpu.uniform !42

631:                                              ; preds = %ComputeEnd2464
  %sunkaddr3249 = getelementptr inbounds i8, ptr addrspace(1) %2969, i64 40
  %632 = atomicrmw or ptr addrspace(1) %sunkaddr3249, i32 %.lcssa4033 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %633, !amdgpu.uniform !42

633:                                              ; preds = %631, %ComputeEnd2464
  %634 = phi i32 [ %632, %631 ], [ poison, %ComputeEnd2464 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %2978)
  %635 = call i32 @llvm.amdgcn.readfirstlane.i32(i32 %634)
  %636 = or i32 %635, %.lcssa4034
  %shl5.i.i.i.i.i339 = shl nuw i32 1, %627
  %and.i.i.i.i.i340 = and i32 %shl5.i.i.i.i.i339, %636
  %tobool3.i.i.i.i341 = icmp ne i32 %and.i.i.i.i.i340, 0
  %637 = tail call i64 @llvm.amdgcn.ballot.i64(i1 %tobool3.i.i.i.i341)
  %and.i.i.i.i.i.i342 = and i64 %637, %626
  %cmp.i.not.i.i.i343 = icmp ne i64 %626, %and.i.i.i.i.i.i342
  br i1 %cmp.i.not.i.i.i343, label %if.then.i.i.i.i344, label %Flow3663, !amdgpu.uniform !42

if.then.i.i.i.i344:                               ; preds = %633
  fence syncscope("agent") acquire
  %638 = load <2 x i64>, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !75
  %639 = extractelement <2 x i64> %638, i32 0
  %640 = inttoptr i64 %639 to ptr
  %641 = extractelement <2 x i64> %638, i32 1
  %642 = inttoptr i64 %641 to ptr
  %idxprom.i.i.i.i345 = zext i32 %spec.store.select.i.i.i333 to i64
  %arrayidx.i.i.i.i346 = getelementptr inbounds nuw i32, ptr %640, i64 %idxprom.i.i.i.i345, !amdgpu.uniform !42
  %643 = load atomic i32, ptr %arrayidx.i.i.i.i346 monotonic, align 4
  %644 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %643)
  %arrayidx.i45.i.i.i347 = getelementptr inbounds nuw i32, ptr %642, i64 %idxprom.i.i.i.i345, !amdgpu.uniform !42
  %645 = load atomic i32, ptr %arrayidx.i45.i.i.i347 monotonic, align 4
  %646 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %645)
  %cmp.i46.not.i.i.i348 = icmp ne i32 %644, %646
  br i1 %cmp.i46.not.i.i.i348, label %if.then12.i.i.i349, label %Flow3661, !amdgpu.uniform !42

if.then12.i.i.i349:                               ; preds = %if.then.i.i.i.i344
  fence syncscope("agent") release
  %647 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %626, i1 true)
  %iszero.i.i.i.i.i.i.i350 = icmp ne i64 %626, 0
  %cmp2.i.i.i.i.i.i351 = icmp eq i64 %647, %sh_prom.i.i.i.i
  %cmp.i.i.i.i.i.i352 = select i1 %iszero.i.i.i.i.i.i.i350, i1 %cmp2.i.i.i.i.i.i351, i1 false
  %conv4.i.i51.i.i.i353 = zext i1 %cmp.i.i.i.i.i.i352 to i32
  %shl.i.i52.i.i.i354 = shl nuw i32 %conv4.i.i51.i.i.i353, %627
  %xor.i.i.i.i.i355 = xor i32 %shl.i.i52.i.i.i354, -1
  %648 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2468, !amdgpu.uniform !42

649:                                              ; preds = %ComputeEnd2469
  %sunkaddr3250 = getelementptr inbounds i8, ptr addrspace(1) %2969, i64 40
  %650 = atomicrmw and ptr addrspace(1) %sunkaddr3250, i32 %.lcssa4035 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %651, !amdgpu.uniform !42

651:                                              ; preds = %649, %ComputeEnd2469
  call void @llvm.amdgcn.end.cf.i64(i64 %2996)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %Flow3661, !amdgpu.uniform !42

if.end14.i.i.i415:                                ; preds = %Flow3661
  %652 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %626, i1 true)
  %iszero.i.i.i.i.i.i416 = icmp ne i64 %626, 0
  %cmp2.i.i.i.i.i417 = icmp eq i64 %652, %sh_prom.i.i.i.i
  %cmp.i.i.i.i.i418 = select i1 %iszero.i.i.i.i.i.i416, i1 %cmp2.i.i.i.i.i417, i1 false
  %653 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.i.i.i.i.i418)
  %654 = extractvalue { i1, i64 } %653, 0
  %655 = extractvalue { i1, i64 } %653, 1
  br i1 %654, label %if.then16.i.i.i420, label %if.end22.i.i.i419

if.then16.i.i.i420:                               ; preds = %if.end14.i.i.i415
  %656 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !76
  %opcode.i.i.i421 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %656, i64 %idxprom.i.i.i.i345, i32 1
  store i32 1660944387, ptr %opcode.i.i.i421, align 8, !tbaa !77
  %arrayidx21.i.i.i422 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %656, i64 %idxprom.i.i.i.i345
  store i64 %626, ptr %arrayidx21.i.i.i422, align 8, !tbaa !80
  br label %if.end22.i.i.i419, !amdgpu.uniform !42

Flow3662:                                         ; preds = %if.end22.i.i.i419, %Flow3661
  %657 = phi i32 [ %644, %if.end22.i.i.i419 ], [ %retval.sroa.8.0.i.i.i330, %Flow3661 ]
  %658 = phi i64 [ %626, %if.end22.i.i.i419 ], [ %retval.sroa.2.0.i.i.i329, %Flow3661 ]
  %659 = phi i32 [ %spec.store.select.i.i.i333, %if.end22.i.i.i419 ], [ %retval.sroa.6.0.i.i.i327, %Flow3661 ]
  br label %cleanup26.i.i.i356, !amdgpu.uniform !42

if.end22.i.i.i419:                                ; preds = %if.then16.i.i.i420, %if.end14.i.i.i415
  call void @llvm.amdgcn.end.cf.i64(i64 %655)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %Flow3662, !amdgpu.uniform !42

Flow3663:                                         ; preds = %cleanup26.i.i.i356, %633
  %660 = phi i32 [ %659, %cleanup26.i.i.i356 ], [ %622, %633 ]
  %661 = phi i64 [ %658, %cleanup26.i.i.i356 ], [ %623, %633 ]
  %662 = phi i32 [ %657, %cleanup26.i.i.i356 ], [ %624, %633 ]
  %663 = phi i32 [ %657, %cleanup26.i.i.i356 ], [ %retval.sroa.8.0.i.i.i330, %633 ]
  %664 = phi i64 [ %658, %cleanup26.i.i.i356 ], [ %retval.sroa.2.0.i.i.i329, %633 ]
  %665 = phi i32 [ %659, %cleanup26.i.i.i356 ], [ %retval.sroa.6.0.i.i.i327, %633 ]
  %666 = phi i1 [ %cmp.i46.not.i.i.i348, %cleanup26.i.i.i356 ], [ true, %633 ]
  br i1 %666, label %for.inc.i.i.i361, label %Flow3664, !amdgpu.uniform !42

cleanup26.i.i.i356:                               ; preds = %Flow3662
  br label %Flow3663, !amdgpu.uniform !42

for.inc.i.i.i361:                                 ; preds = %Flow3663
  %inc.i.i.i366 = add i32 %spec.store.select.i.i.i333, 1
  br label %Flow3664, !amdgpu.uniform !42

Flow3664:                                         ; preds = %for.inc.i.i.i361, %Flow3663
  %667 = phi i32 [ %inc.i.i.i366, %for.inc.i.i.i361 ], [ poison, %Flow3663 ]
  %668 = phi i1 [ false, %for.inc.i.i.i361 ], [ true, %Flow3663 ]
  br i1 %668, label %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i367, label %for.cond.i.i.i326, !amdgpu.uniform !42

_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i367: ; preds = %Flow3664
  %.lcssa4038 = phi i32 [ %660, %Flow3664 ]
  %.lcssa4037 = phi i64 [ %661, %Flow3664 ]
  %.lcssa4036 = phi i32 [ %662, %Flow3664 ]
  fence acquire
  %669 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %mul.i.i.i.i.i.i368 = shl i32 %.lcssa4038, 6
  %idxprom.i19.i.i.i.i.i369 = zext i32 %mul.i.i.i.i.i.i368 to i64
  %arrayidx.i20.i.i.i.i.i370 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %669, i64 %idxprom.i19.i.i.i.i.i369
  %arrayidx.i22.i.i.i.i.i371 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %arrayidx.i20.i.i.i.i.i370, i64 %sh_prom.i.i.i.i
  store i64 %spec.select.i, ptr %arrayidx.i22.i.i.i.i.i371, align 8, !tbaa !82
  %spec.select.i.i.i.i.i.i.i372 = tail call i64 @llvm.umin.i64(i64 %spec.select.i, i64 56)
  %670 = and i64 %spec.select.i.i.i.i.i.i.i372, 15
  %671 = sub i64 %spec.select.i.i.i.i.i.i.i372, %670
  %672 = icmp ne i64 %671, 0
  br i1 %672, label %loop-memcpy-expansion2255.preheader, label %Flow3660, !amdgpu.uniform !42

loop-memcpy-expansion2255.preheader:              ; preds = %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i367
  %673 = add nuw nsw i64 %76, 8
  %scevgep3115 = getelementptr i8, ptr %669, i64 %673
  %674 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i369, 6
  %scevgep3116 = getelementptr i8, ptr %scevgep3115, i64 %674
  br label %loop-memcpy-expansion2255, !amdgpu.uniform !42

Flow3660:                                         ; preds = %Flow3659, %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i367
  br label %loop-memcpy-residual-header2258, !amdgpu.uniform !42

loop-memcpy-expansion2255:                        ; preds = %loop-memcpy-expansion2255.preheader, %loop-memcpy-expansion2255
  %loop-index2256 = phi i64 [ %677, %loop-memcpy-expansion2255 ], [ 0, %loop-memcpy-expansion2255.preheader ]
  %675 = getelementptr i8, ptr %610, i64 %loop-index2256, !amdgpu.uniform !42
  %676 = load <4 x i32>, ptr %675, align 1
  %scevgep3117 = getelementptr i8, ptr %scevgep3116, i64 %loop-index2256
  store <4 x i32> %676, ptr %scevgep3117, align 1
  %677 = add i64 %loop-index2256, 16
  %678 = icmp uge i64 %677, %671
  br i1 %678, label %Flow3659, label %loop-memcpy-expansion2255, !amdgpu.uniform !42

Flow3658:                                         ; preds = %Flow3656, %Flow3657
  %679 = phi i64 [ %2625, %Flow3656 ], [ %2621, %Flow3657 ]
  br label %post-loop-memcpy-expansion2254, !amdgpu.uniform !42

loop-memcpy-residual2257:                         ; preds = %loop-memcpy-residual2257.preheader, %loop-memcpy-residual2257
  %residual-loop-index2259 = phi i64 [ %681, %loop-memcpy-residual2257 ], [ 0, %loop-memcpy-residual2257.preheader ]
  %scevgep3114 = getelementptr i8, ptr %scevgep3113, i64 %residual-loop-index2259, !amdgpu.uniform !42
  %680 = load i8, ptr %scevgep3114, align 1
  %scevgep3112 = getelementptr i8, ptr %scevgep3111, i64 %residual-loop-index2259
  store i8 %680, ptr %scevgep3112, align 1
  %681 = add i64 %residual-loop-index2259, 1
  %682 = icmp uge i64 %681, %670
  br i1 %682, label %Flow3656, label %loop-memcpy-residual2257, !amdgpu.uniform !42

Flow3656:                                         ; preds = %loop-memcpy-residual2257
  br label %Flow3658, !amdgpu.uniform !42

post-loop-memcpy-expansion2254:                   ; preds = %Flow3658
  %tobool.not.i.i.i.i.i.i374 = icmp eq i32 %.lcssa4036, 0
  %conv.i.i.i.i.i.i375 = zext i1 %tobool.not.i.i.i.i.i.i374 to i32
  fence release
  %683 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %idxprom.i23.i.i.i.i.i376 = zext i32 %.lcssa4038 to i64
  %arrayidx.i24.i.i.i.i.i377 = getelementptr inbounds nuw i32, ptr %683, i64 %idxprom.i23.i.i.i.i.i376
  store atomic i32 %conv.i.i.i.i.i.i375, ptr %arrayidx.i24.i.i.i.i.i377 monotonic, align 4
  %684 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !84
  %arrayidx.i.i5.i.i378 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %684, i64 %idxprom.i23.i.i.i.i.i376, !amdgpu.uniform !42
  %685 = load i64, ptr %arrayidx.i.i5.i.i378, align 8, !tbaa !80
  %scevgep3096 = getelementptr i8, ptr %610, i64 56
  %umin3102 = call i64 @llvm.umin.i64(i64 sub (i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) getelementptr inbounds nuw (i8, ptr addrspace(4) @.str.9, i64 8) to ptr) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.9 to ptr) to i64)), i64 2)
  %686 = sub i64 sub (i64 add (i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) getelementptr inbounds nuw (i8, ptr addrspace(4) @.str.9, i64 8) to ptr) to i64), i64 -56), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.9 to ptr) to i64)), %umin3102
  br label %while.cond.i.i.i.i379, !amdgpu.uniform !42

while.cond.i.i.i.i379:                            ; preds = %post-loop-memcpy-expansion2254, %Flow3655
  %phi.broken3792 = phi i64 [ 0, %post-loop-memcpy-expansion2254 ], [ %703, %Flow3655 ]
  %lsr.iv3103 = phi i64 [ %701, %Flow3655 ], [ %686, %post-loop-memcpy-expansion2254 ]
  %lsr.iv3097 = phi ptr [ %700, %Flow3655 ], [ %scevgep3096, %post-loop-memcpy-expansion2254 ]
  %port.sroa.37.0.i.i380 = phi i32 [ %699, %Flow3655 ], [ %conv.i.i.i.i.i.i375, %post-loop-memcpy-expansion2254 ]
  %idx.0.i.i.i.i381 = phi i64 [ %698, %Flow3655 ], [ 56, %post-loop-memcpy-expansion2254 ]
  %cmp.i.i.i.i382 = icmp ugt i64 %spec.select.i, %idx.0.i.i.i.i381
  %687 = tail call i64 @llvm.amdgcn.ballot.i64(i1 %cmp.i.i.i.i382)
  %and.i.i.i.i6.i.i383 = and i64 %687, %685
  %tobool.not.i.i.i.i384 = icmp ne i64 %and.i.i.i.i6.i.i383, 0
  %688 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i22.i.i.i385 = getelementptr inbounds nuw i32, ptr %688, i64 %idxprom.i23.i.i.i.i.i376, !amdgpu.uniform !42
  %689 = load atomic i32, ptr %arrayidx.i22.i.i.i385 monotonic, align 4
  %690 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %tobool.not.i.i.i.i384)
  %691 = extractvalue { i1, i64 } %690, 0
  %692 = extractvalue { i1, i64 } %690, 1
  br i1 %691, label %cond.false.i8.i.i.i.i386, label %Flow3655

cond.false.i8.i.i.i.i386:                         ; preds = %while.cond.i.i.i.i379
  %693 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %689)
  %694 = icmp ne i32 %693, %port.sroa.37.0.i.i380
  br i1 %694, label %while.body.i.i21.i.i.i.i387.preheader, label %Flow3654, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i387.preheader:            ; preds = %cond.false.i8.i.i.i.i386
  br label %while.body.i.i21.i.i.i.i387, !amdgpu.uniform !42

Flow3654:                                         ; preds = %Flow3653, %cond.false.i8.i.i.i.i386
  br label %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i390, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i387:                      ; preds = %while.body.i.i21.i.i.i.i387.preheader, %while.body.i.i21.i.i.i.i387
  tail call void @llvm.amdgcn.s.sleep(i32 2)
  %695 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i22.i.i.i.i388 = getelementptr inbounds nuw i32, ptr %695, i64 %idxprom.i23.i.i.i.i.i376, !amdgpu.uniform !42
  %696 = load atomic i32, ptr %arrayidx.i.i.i22.i.i.i.i388 monotonic, align 4
  %697 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %696)
  %cmp.i.not.i.i23.i.i.i.i389 = icmp eq i32 %697, %port.sroa.37.0.i.i380
  br i1 %cmp.i.not.i.i23.i.i.i.i389, label %Flow3653, label %while.body.i.i21.i.i.i.i387, !amdgpu.uniform !42

Flow3653:                                         ; preds = %while.body.i.i21.i.i.i.i387
  br label %Flow3654, !amdgpu.uniform !42

Flow3655:                                         ; preds = %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i391, %while.cond.i.i.i.i379
  %698 = phi i64 [ %add.i.i.i.i394, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i391 ], [ poison, %while.cond.i.i.i.i379 ]
  %699 = phi i32 [ %conv.i.i26.i.i.i.i392, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i391 ], [ poison, %while.cond.i.i.i.i379 ]
  %700 = phi ptr [ %scevgep3098, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i391 ], [ poison, %while.cond.i.i.i.i379 ]
  %701 = phi i64 [ %lsr.iv.next3104, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i391 ], [ poison, %while.cond.i.i.i.i379 ]
  %702 = phi i1 [ false, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i391 ], [ true, %while.cond.i.i.i.i379 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %692)
  %703 = call i64 @llvm.amdgcn.if.break.i64(i1 %702, i64 %phi.broken3792)
  %704 = call i1 @llvm.amdgcn.loop.i64(i64 %703)
  br i1 %704, label %cond.false.i.i.i401, label %while.cond.i.i.i.i379

_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i390: ; preds = %Flow3654
  fence acquire
  br i1 %cmp.i.i.i.i382, label %if.then.i.i.i.i.i.i.i395, label %Flow3652, !amdgpu.uniform !42

if.then.i.i.i.i.i.i.i395:                         ; preds = %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i390
  %sub.i.i.i.i.i.i.i397 = sub nuw i64 %spec.select.i, %idx.0.i.i.i.i381
  %spec.select.i.i.i36.i.i.i.i398 = tail call i64 @llvm.umin.i64(i64 %sub.i.i.i.i.i.i.i397, i64 64)
  %705 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %706 = and i64 %spec.select.i.i.i36.i.i.i.i398, 15
  %707 = sub i64 %spec.select.i.i.i36.i.i.i.i398, %706
  %708 = icmp ne i64 %707, 0
  br i1 %708, label %loop-memcpy-expansion2249.preheader, label %Flow3651, !amdgpu.uniform !42

loop-memcpy-expansion2249.preheader:              ; preds = %if.then.i.i.i.i.i.i.i395
  %arrayidx.i20.i35.i.i.i.i399 = getelementptr %"struct.rpc::Buffer", ptr %705, i64 %idxprom.i19.i.i.i.i.i369
  %arrayidx.i22.i38.i.i.i.i400 = getelementptr %"struct.rpc::Buffer", ptr %arrayidx.i20.i35.i.i.i.i399, i64 %sh_prom.i.i.i.i
  br label %loop-memcpy-expansion2249, !amdgpu.uniform !42

Flow3651:                                         ; preds = %Flow3650, %if.then.i.i.i.i.i.i.i395
  br label %loop-memcpy-residual-header2252, !amdgpu.uniform !42

loop-memcpy-expansion2249:                        ; preds = %loop-memcpy-expansion2249.preheader, %loop-memcpy-expansion2249
  %loop-index2250 = phi i64 [ %711, %loop-memcpy-expansion2249 ], [ 0, %loop-memcpy-expansion2249.preheader ]
  %scevgep3099 = getelementptr i8, ptr %lsr.iv3097, i64 %loop-index2250, !amdgpu.uniform !42
  %709 = load <4 x i32>, ptr %scevgep3099, align 1
  %710 = getelementptr i8, ptr %arrayidx.i22.i38.i.i.i.i400, i64 %loop-index2250
  store <4 x i32> %709, ptr %710, align 1
  %711 = add i64 %loop-index2250, 16
  %712 = icmp uge i64 %711, %707
  br i1 %712, label %Flow3650, label %loop-memcpy-expansion2249, !amdgpu.uniform !42

Flow3649:                                         ; preds = %Flow3648, %loop-memcpy-residual-header2252
  br label %Flow3652, !amdgpu.uniform !42

loop-memcpy-residual2251:                         ; preds = %loop-memcpy-residual2251.preheader, %loop-memcpy-residual2251
  %residual-loop-index2253 = phi i64 [ %714, %loop-memcpy-residual2251 ], [ 0, %loop-memcpy-residual2251.preheader ]
  %scevgep3109 = getelementptr i8, ptr %scevgep3108, i64 %residual-loop-index2253, !amdgpu.uniform !42
  %713 = load i8, ptr %scevgep3109, align 1
  %scevgep3107 = getelementptr i8, ptr %scevgep3106, i64 %residual-loop-index2253
  store i8 %713, ptr %scevgep3107, align 1
  %714 = add i64 %residual-loop-index2253, 1
  %715 = icmp uge i64 %714, %706
  br i1 %715, label %Flow3648, label %loop-memcpy-residual2251, !amdgpu.uniform !42

Flow3648:                                         ; preds = %loop-memcpy-residual2251
  br label %Flow3649, !amdgpu.uniform !42

_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i391: ; preds = %Flow3652
  %conv.i.i26.i.i.i.i392 = xor i32 %port.sroa.37.0.i.i380, 1
  fence release
  %716 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %arrayidx.i24.i29.i.i.i.i393 = getelementptr inbounds nuw i32, ptr %716, i64 %idxprom.i23.i.i.i.i.i376
  store atomic i32 %conv.i.i26.i.i.i.i392, ptr %arrayidx.i24.i29.i.i.i.i393 monotonic, align 4
  %add.i.i.i.i394 = add i64 %idx.0.i.i.i.i381, 64
  %scevgep3098 = getelementptr i8, ptr %lsr.iv3097, i64 64
  %lsr.iv.next3104 = add i64 %lsr.iv3103, -64
  br label %Flow3655, !amdgpu.uniform !42

cond.false.i.i.i401:                              ; preds = %Flow3655
  %.lcssa4032 = phi i64 [ %703, %Flow3655 ]
  %port.sroa.37.0.i.i380.lcssa = phi i32 [ %port.sroa.37.0.i.i380, %Flow3655 ]
  %.lcssa4031 = phi i32 [ %689, %Flow3655 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa4032)
  %717 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %.lcssa4031)
  %718 = icmp ne i32 %717, %port.sroa.37.0.i.i380.lcssa
  %719 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %718)
  %720 = extractvalue { i1, i64 } %719, 0
  %721 = extractvalue { i1, i64 } %719, 1
  br i1 %720, label %while.body.i.i.i.i402.preheader, label %Flow3647

while.body.i.i.i.i402.preheader:                  ; preds = %cond.false.i.i.i401
  br label %while.body.i.i.i.i402, !amdgpu.uniform !42

Flow3647:                                         ; preds = %Flow3646, %cond.false.i.i.i401
  call void @llvm.amdgcn.end.cf.i64(i64 %721)
  br label %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsINS_3cpp6stringEEERS1_T_.exit, !amdgpu.uniform !42

while.body.i.i.i.i402:                            ; preds = %while.body.i.i.i.i402.preheader, %while.body.i.i.i.i402
  %phi.broken3793 = phi i64 [ 0, %while.body.i.i.i.i402.preheader ], [ %725, %while.body.i.i.i.i402 ]
  tail call void @llvm.amdgcn.s.sleep(i32 2)
  %722 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i8.i.i403 = getelementptr inbounds nuw i32, ptr %722, i64 %idxprom.i23.i.i.i.i.i376, !amdgpu.uniform !42
  %723 = load atomic i32, ptr %arrayidx.i.i.i8.i.i403 monotonic, align 4
  %724 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %723)
  %cmp.i.not.i.i.i.i404 = icmp eq i32 %724, %port.sroa.37.0.i.i380.lcssa
  %725 = call i64 @llvm.amdgcn.if.break.i64(i1 %cmp.i.not.i.i.i.i404, i64 %phi.broken3793)
  %726 = call i1 @llvm.amdgcn.loop.i64(i64 %725)
  br i1 %726, label %Flow3646, label %while.body.i.i.i.i402

Flow3646:                                         ; preds = %while.body.i.i.i.i402
  %.lcssa4030 = phi i64 [ %725, %while.body.i.i.i.i402 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa4030)
  br label %Flow3647, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsINS_3cpp6stringEEERS1_T_.exit: ; preds = %Flow3647
  fence acquire
  tail call void @llvm.amdgcn.wave.barrier()
  fence syncscope("agent") release
  %727 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %.lcssa4037, i1 true)
  %iszero.i.i.i.i.i13.i.i406 = icmp ne i64 %.lcssa4037, 0
  %cmp2.i.i.i.i14.i.i407 = icmp eq i64 %727, %sh_prom.i.i.i.i
  %cmp.i.i.i.i15.i.i408 = select i1 %iszero.i.i.i.i.i13.i.i406, i1 %cmp2.i.i.i.i14.i.i407, i1 false
  %728 = and i32 %.lcssa4038, 31
  %conv4.i.i.i19.i.i412 = zext i1 %cmp.i.i.i.i15.i.i408 to i32
  %shl.i.i.i20.i.i413 = shl nuw i32 %conv4.i.i.i19.i.i412, %728
  %xor.i.i.i21.i.i414 = xor i32 %shl.i.i.i20.i.i413, -1
  %729 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2472, !amdgpu.uniform !42

730:                                              ; preds = %ComputeEnd2473
  %div8.i.i.i16.i.i409 = lshr i32 %.lcssa4038, 5
  %idxprom.i.i.i17.i.i410 = zext nneg i32 %div8.i.i.i16.i.i409 to i64
  %731 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i17.i.i410
  %sunkaddr3251 = getelementptr inbounds i8, ptr addrspace(1) %731, i64 40
  %732 = atomicrmw and ptr addrspace(1) %sunkaddr3251, i32 %.lcssa4029 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %733, !amdgpu.uniform !42

733:                                              ; preds = %730, %ComputeEnd2473
  call void @llvm.amdgcn.end.cf.i64(i64 %3013)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %for.cond.i.i1.i431, !amdgpu.uniform !42

for.cond.i.i1.i431:                               ; preds = %733, %Flow3645
  %734 = phi i32 [ %772, %Flow3645 ], [ poison, %733 ]
  %735 = phi i64 [ %773, %Flow3645 ], [ poison, %733 ]
  %736 = phi i32 [ %774, %Flow3645 ], [ poison, %733 ]
  %retval.sroa.6.0.i.i.i432 = phi i32 [ %777, %Flow3645 ], [ undef, %733 ]
  %retval.sroa.2.0.i.i.i434 = phi i64 [ %776, %Flow3645 ], [ undef, %733 ]
  %retval.sroa.8.0.i.i.i435 = phi i32 [ %775, %Flow3645 ], [ undef, %733 ]
  %index.0.i.i.i436 = phi i32 [ %779, %Flow3645 ], [ 0, %733 ]
  %737 = load i32, ptr addrspace(1) @__llvm_rpc_client, align 8, !tbaa !68
  %cmp.not.i.i.i437 = icmp ult i32 %index.0.i.i.i436, %737
  %spec.store.select.i.i.i438 = select i1 %cmp.not.i.i.i437, i32 %index.0.i.i.i436, i32 0
  %738 = tail call noundef i64 @llvm.amdgcn.ballot.i64(i1 true)
  %739 = and i32 %spec.store.select.i.i.i438, 31
  %740 = lshr i64 %738, %sh_prom.i.i.i.i
  %741 = trunc i64 %740 to i32
  %conv4.i.i.i.i.i442 = and i32 %741, 1
  %shl.i.i.i.i.i443 = shl nuw i32 %conv4.i.i.i.i.i442, %739
  %742 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2476, !amdgpu.uniform !42

743:                                              ; preds = %ComputeEnd2477
  %sunkaddr3252 = getelementptr inbounds i8, ptr addrspace(1) %3023, i64 40
  %744 = atomicrmw or ptr addrspace(1) %sunkaddr3252, i32 %.lcssa4023 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %745, !amdgpu.uniform !42

745:                                              ; preds = %743, %ComputeEnd2477
  %746 = phi i32 [ %744, %743 ], [ poison, %ComputeEnd2477 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %3032)
  %747 = call i32 @llvm.amdgcn.readfirstlane.i32(i32 %746)
  %748 = or i32 %747, %.lcssa4024
  %shl5.i.i.i.i.i444 = shl nuw i32 1, %739
  %and.i.i.i.i.i445 = and i32 %shl5.i.i.i.i.i444, %748
  %tobool3.i.i.i.i446 = icmp ne i32 %and.i.i.i.i.i445, 0
  %749 = tail call i64 @llvm.amdgcn.ballot.i64(i1 %tobool3.i.i.i.i446)
  %and.i.i.i.i.i.i447 = and i64 %749, %738
  %cmp.i.not.i.i.i448 = icmp ne i64 %738, %and.i.i.i.i.i.i447
  br i1 %cmp.i.not.i.i.i448, label %if.then.i.i.i.i449, label %Flow3644, !amdgpu.uniform !42

if.then.i.i.i.i449:                               ; preds = %745
  fence syncscope("agent") acquire
  %750 = load <2 x i64>, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !75
  %751 = extractelement <2 x i64> %750, i32 0
  %752 = inttoptr i64 %751 to ptr
  %753 = extractelement <2 x i64> %750, i32 1
  %754 = inttoptr i64 %753 to ptr
  %idxprom.i.i.i.i450 = zext i32 %spec.store.select.i.i.i438 to i64
  %arrayidx.i.i.i.i451 = getelementptr inbounds nuw i32, ptr %752, i64 %idxprom.i.i.i.i450, !amdgpu.uniform !42
  %755 = load atomic i32, ptr %arrayidx.i.i.i.i451 monotonic, align 4
  %756 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %755)
  %arrayidx.i45.i.i.i452 = getelementptr inbounds nuw i32, ptr %754, i64 %idxprom.i.i.i.i450, !amdgpu.uniform !42
  %757 = load atomic i32, ptr %arrayidx.i45.i.i.i452 monotonic, align 4
  %758 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %757)
  %cmp.i46.not.i.i.i453 = icmp ne i32 %756, %758
  br i1 %cmp.i46.not.i.i.i453, label %if.then12.i.i.i454, label %Flow3642, !amdgpu.uniform !42

if.then12.i.i.i454:                               ; preds = %if.then.i.i.i.i449
  fence syncscope("agent") release
  %759 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %738, i1 true)
  %iszero.i.i.i.i.i.i.i455 = icmp ne i64 %738, 0
  %cmp2.i.i.i.i.i.i456 = icmp eq i64 %759, %sh_prom.i.i.i.i
  %cmp.i.i.i.i.i.i457 = select i1 %iszero.i.i.i.i.i.i.i455, i1 %cmp2.i.i.i.i.i.i456, i1 false
  %conv4.i.i51.i.i.i458 = zext i1 %cmp.i.i.i.i.i.i457 to i32
  %shl.i.i52.i.i.i459 = shl nuw i32 %conv4.i.i51.i.i.i458, %739
  %xor.i.i.i.i.i460 = xor i32 %shl.i.i52.i.i.i459, -1
  %760 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2481, !amdgpu.uniform !42

761:                                              ; preds = %ComputeEnd2482
  %sunkaddr3253 = getelementptr inbounds i8, ptr addrspace(1) %3023, i64 40
  %762 = atomicrmw and ptr addrspace(1) %sunkaddr3253, i32 %.lcssa4025 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %763, !amdgpu.uniform !42

763:                                              ; preds = %761, %ComputeEnd2482
  call void @llvm.amdgcn.end.cf.i64(i64 %3050)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %Flow3642, !amdgpu.uniform !42

if.end14.i.i.i520:                                ; preds = %Flow3642
  %764 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %738, i1 true)
  %iszero.i.i.i.i.i.i521 = icmp ne i64 %738, 0
  %cmp2.i.i.i.i.i522 = icmp eq i64 %764, %sh_prom.i.i.i.i
  %cmp.i.i.i.i.i523 = select i1 %iszero.i.i.i.i.i.i521, i1 %cmp2.i.i.i.i.i522, i1 false
  %765 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.i.i.i.i.i523)
  %766 = extractvalue { i1, i64 } %765, 0
  %767 = extractvalue { i1, i64 } %765, 1
  br i1 %766, label %if.then16.i.i.i525, label %if.end22.i.i.i524

if.then16.i.i.i525:                               ; preds = %if.end14.i.i.i520
  %768 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !76
  %opcode.i.i.i526 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %768, i64 %idxprom.i.i.i.i450, i32 1
  store i32 1660944387, ptr %opcode.i.i.i526, align 8, !tbaa !77
  %arrayidx21.i.i.i527 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %768, i64 %idxprom.i.i.i.i450
  store i64 %738, ptr %arrayidx21.i.i.i527, align 8, !tbaa !80
  br label %if.end22.i.i.i524, !amdgpu.uniform !42

Flow3643:                                         ; preds = %if.end22.i.i.i524, %Flow3642
  %769 = phi i32 [ %756, %if.end22.i.i.i524 ], [ %retval.sroa.8.0.i.i.i435, %Flow3642 ]
  %770 = phi i64 [ %738, %if.end22.i.i.i524 ], [ %retval.sroa.2.0.i.i.i434, %Flow3642 ]
  %771 = phi i32 [ %spec.store.select.i.i.i438, %if.end22.i.i.i524 ], [ %retval.sroa.6.0.i.i.i432, %Flow3642 ]
  br label %cleanup26.i.i.i461, !amdgpu.uniform !42

if.end22.i.i.i524:                                ; preds = %if.then16.i.i.i525, %if.end14.i.i.i520
  call void @llvm.amdgcn.end.cf.i64(i64 %767)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %Flow3643, !amdgpu.uniform !42

Flow3644:                                         ; preds = %cleanup26.i.i.i461, %745
  %772 = phi i32 [ %771, %cleanup26.i.i.i461 ], [ %734, %745 ]
  %773 = phi i64 [ %770, %cleanup26.i.i.i461 ], [ %735, %745 ]
  %774 = phi i32 [ %769, %cleanup26.i.i.i461 ], [ %736, %745 ]
  %775 = phi i32 [ %769, %cleanup26.i.i.i461 ], [ %retval.sroa.8.0.i.i.i435, %745 ]
  %776 = phi i64 [ %770, %cleanup26.i.i.i461 ], [ %retval.sroa.2.0.i.i.i434, %745 ]
  %777 = phi i32 [ %771, %cleanup26.i.i.i461 ], [ %retval.sroa.6.0.i.i.i432, %745 ]
  %778 = phi i1 [ %cmp.i46.not.i.i.i453, %cleanup26.i.i.i461 ], [ true, %745 ]
  br i1 %778, label %for.inc.i.i.i466, label %Flow3645, !amdgpu.uniform !42

cleanup26.i.i.i461:                               ; preds = %Flow3643
  br label %Flow3644, !amdgpu.uniform !42

for.inc.i.i.i466:                                 ; preds = %Flow3644
  %inc.i.i.i471 = add i32 %spec.store.select.i.i.i438, 1
  br label %Flow3645, !amdgpu.uniform !42

Flow3645:                                         ; preds = %for.inc.i.i.i466, %Flow3644
  %779 = phi i32 [ %inc.i.i.i471, %for.inc.i.i.i466 ], [ poison, %Flow3644 ]
  %780 = phi i1 [ false, %for.inc.i.i.i466 ], [ true, %Flow3644 ]
  br i1 %780, label %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i472, label %for.cond.i.i1.i431, !amdgpu.uniform !42

_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i472: ; preds = %Flow3645
  %.lcssa4028 = phi i32 [ %772, %Flow3645 ]
  %.lcssa4027 = phi i64 [ %773, %Flow3645 ]
  %.lcssa4026 = phi i32 [ %774, %Flow3645 ]
  fence acquire
  %781 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %mul.i.i.i.i.i.i473 = shl i32 %.lcssa4028, 6
  %idxprom.i19.i.i.i.i.i474 = zext i32 %mul.i.i.i.i.i.i473 to i64
  %arrayidx.i20.i.i.i.i.i475 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %781, i64 %idxprom.i19.i.i.i.i.i474
  %arrayidx.i22.i.i.i.i.i476 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %arrayidx.i20.i.i.i.i.i475, i64 %sh_prom.i.i.i.i
  store i64 sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.34.12 to ptr), i64 10) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.34.12 to ptr) to i64)), ptr %arrayidx.i22.i.i.i.i.i476, align 8, !tbaa !82
  %spec.select.i.i.i.i.i.i.i477 = tail call i64 @llvm.umin.i64(i64 sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.34.12 to ptr), i64 10) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.34.12 to ptr) to i64)), i64 56)
  %782 = and i64 %spec.select.i.i.i.i.i.i.i477, 15
  %783 = sub i64 %spec.select.i.i.i.i.i.i.i477, %782
  %784 = icmp ne i64 %783, 0
  br i1 %784, label %loop-memcpy-expansion2345.preheader, label %Flow3641, !amdgpu.uniform !42

loop-memcpy-expansion2345.preheader:              ; preds = %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i472
  %785 = add nuw nsw i64 %76, 8
  %scevgep3093 = getelementptr i8, ptr %781, i64 %785
  %786 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i474, 6
  %scevgep3094 = getelementptr i8, ptr %scevgep3093, i64 %786
  br label %loop-memcpy-expansion2345, !amdgpu.uniform !42

Flow3641:                                         ; preds = %Flow3640, %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i472
  br label %loop-memcpy-residual-header2348, !amdgpu.uniform !42

loop-memcpy-expansion2345:                        ; preds = %loop-memcpy-expansion2345.preheader, %loop-memcpy-expansion2345
  %loop-index2346 = phi i64 [ %789, %loop-memcpy-expansion2345 ], [ 0, %loop-memcpy-expansion2345.preheader ]
  %787 = getelementptr i8, ptr addrspace(4) @.str.34.12, i64 %loop-index2346, !amdgpu.uniform !42
  %788 = load <4 x i32>, ptr addrspace(4) %787, align 1
  %scevgep3095 = getelementptr i8, ptr %scevgep3094, i64 %loop-index2346
  store <4 x i32> %788, ptr %scevgep3095, align 1
  %789 = add i64 %loop-index2346, 16
  %790 = icmp uge i64 %789, %783
  br i1 %790, label %Flow3640, label %loop-memcpy-expansion2345, !amdgpu.uniform !42

Flow3639:                                         ; preds = %Flow3637, %Flow3638
  %791 = phi i64 [ %2698, %Flow3637 ], [ %2695, %Flow3638 ]
  br label %post-loop-memcpy-expansion2344, !amdgpu.uniform !42

loop-memcpy-residual2347:                         ; preds = %loop-memcpy-residual2347.preheader, %loop-memcpy-residual2347
  %residual-loop-index2349 = phi i64 [ %793, %loop-memcpy-residual2347 ], [ 0, %loop-memcpy-residual2347.preheader ]
  %scevgep3092 = getelementptr i8, ptr addrspace(4) @.str.34.12, i64 %residual-loop-index2349, !amdgpu.uniform !42
  %792 = load i8, ptr addrspace(4) %scevgep3092, align 1
  %scevgep3091 = getelementptr i8, ptr %scevgep3090, i64 %residual-loop-index2349
  store i8 %792, ptr %scevgep3091, align 1
  %793 = add i64 %residual-loop-index2349, 1
  %794 = icmp uge i64 %793, %782
  br i1 %794, label %Flow3637, label %loop-memcpy-residual2347, !amdgpu.uniform !42

Flow3637:                                         ; preds = %loop-memcpy-residual2347
  br label %Flow3639, !amdgpu.uniform !42

post-loop-memcpy-expansion2344:                   ; preds = %Flow3639
  %tobool.not.i.i.i.i.i.i479 = icmp eq i32 %.lcssa4026, 0
  %conv.i.i.i.i.i.i480 = zext i1 %tobool.not.i.i.i.i.i.i479 to i32
  fence release
  %795 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %idxprom.i23.i.i.i.i.i481 = zext i32 %.lcssa4028 to i64
  %arrayidx.i24.i.i.i.i.i482 = getelementptr inbounds nuw i32, ptr %795, i64 %idxprom.i23.i.i.i.i.i481
  store atomic i32 %conv.i.i.i.i.i.i480, ptr %arrayidx.i24.i.i.i.i.i482 monotonic, align 4
  %796 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !84
  %arrayidx.i.i5.i.i483 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %796, i64 %idxprom.i23.i.i.i.i.i481, !amdgpu.uniform !42
  %797 = load i64, ptr %arrayidx.i.i5.i.i483, align 8, !tbaa !80
  br label %while.cond.i.i.i.i484, !amdgpu.uniform !42

while.cond.i.i.i.i484:                            ; preds = %post-loop-memcpy-expansion2344, %Flow3636
  %phi.broken3794 = phi i64 [ 0, %post-loop-memcpy-expansion2344 ], [ %814, %Flow3636 ]
  %lsr.iv3082 = phi i64 [ %812, %Flow3636 ], [ -46, %post-loop-memcpy-expansion2344 ]
  %lsr.iv3077 = phi ptr addrspace(4) [ %811, %Flow3636 ], [ getelementptr (i8, ptr addrspace(4) @.str.34.12, i64 56), %post-loop-memcpy-expansion2344 ]
  %port.sroa.37.0.i.i485 = phi i32 [ %810, %Flow3636 ], [ %conv.i.i.i.i.i.i480, %post-loop-memcpy-expansion2344 ]
  %idx.0.i.i.i.i486 = phi i64 [ %809, %Flow3636 ], [ 56, %post-loop-memcpy-expansion2344 ]
  %cmp.i.i.i.i487 = icmp ult i64 %idx.0.i.i.i.i486, sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.34.12 to ptr), i64 10) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.34.12 to ptr) to i64))
  %798 = tail call i64 @llvm.amdgcn.ballot.i64(i1 %cmp.i.i.i.i487)
  %and.i.i.i.i6.i.i488 = and i64 %798, %797
  %tobool.not.i.i.i.i489 = icmp ne i64 %and.i.i.i.i6.i.i488, 0
  %799 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i22.i.i.i490 = getelementptr inbounds nuw i32, ptr %799, i64 %idxprom.i23.i.i.i.i.i481, !amdgpu.uniform !42
  %800 = load atomic i32, ptr %arrayidx.i22.i.i.i490 monotonic, align 4
  %801 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %tobool.not.i.i.i.i489)
  %802 = extractvalue { i1, i64 } %801, 0
  %803 = extractvalue { i1, i64 } %801, 1
  br i1 %802, label %cond.false.i8.i.i.i.i491, label %Flow3636

cond.false.i8.i.i.i.i491:                         ; preds = %while.cond.i.i.i.i484
  %804 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %800)
  %805 = icmp ne i32 %804, %port.sroa.37.0.i.i485
  br i1 %805, label %while.body.i.i21.i.i.i.i492.preheader, label %Flow3635, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i492.preheader:            ; preds = %cond.false.i8.i.i.i.i491
  br label %while.body.i.i21.i.i.i.i492, !amdgpu.uniform !42

Flow3635:                                         ; preds = %Flow3634, %cond.false.i8.i.i.i.i491
  br label %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i495, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i492:                      ; preds = %while.body.i.i21.i.i.i.i492.preheader, %while.body.i.i21.i.i.i.i492
  tail call void @llvm.amdgcn.s.sleep(i32 2)
  %806 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i22.i.i.i.i493 = getelementptr inbounds nuw i32, ptr %806, i64 %idxprom.i23.i.i.i.i.i481, !amdgpu.uniform !42
  %807 = load atomic i32, ptr %arrayidx.i.i.i22.i.i.i.i493 monotonic, align 4
  %808 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %807)
  %cmp.i.not.i.i23.i.i.i.i494 = icmp eq i32 %808, %port.sroa.37.0.i.i485
  br i1 %cmp.i.not.i.i23.i.i.i.i494, label %Flow3634, label %while.body.i.i21.i.i.i.i492, !amdgpu.uniform !42

Flow3634:                                         ; preds = %while.body.i.i21.i.i.i.i492
  br label %Flow3635, !amdgpu.uniform !42

Flow3636:                                         ; preds = %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i496, %while.cond.i.i.i.i484
  %809 = phi i64 [ %add.i.i.i.i499, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i496 ], [ poison, %while.cond.i.i.i.i484 ]
  %810 = phi i32 [ %conv.i.i26.i.i.i.i497, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i496 ], [ poison, %while.cond.i.i.i.i484 ]
  %811 = phi ptr addrspace(4) [ %scevgep3078, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i496 ], [ poison, %while.cond.i.i.i.i484 ]
  %812 = phi i64 [ %lsr.iv.next3083, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i496 ], [ poison, %while.cond.i.i.i.i484 ]
  %813 = phi i1 [ false, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i496 ], [ true, %while.cond.i.i.i.i484 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %803)
  %814 = call i64 @llvm.amdgcn.if.break.i64(i1 %813, i64 %phi.broken3794)
  %815 = call i1 @llvm.amdgcn.loop.i64(i64 %814)
  br i1 %815, label %cond.false.i.i.i506, label %while.cond.i.i.i.i484

_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i495: ; preds = %Flow3635
  fence acquire
  br i1 %cmp.i.i.i.i487, label %if.then.i.i.i.i.i.i.i500, label %Flow3633, !amdgpu.uniform !42

if.then.i.i.i.i.i.i.i500:                         ; preds = %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i495
  %sub.i.i.i.i.i.i.i502 = sub nuw i64 sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.34.12 to ptr), i64 10) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.34.12 to ptr) to i64)), %idx.0.i.i.i.i486
  %spec.select.i.i.i36.i.i.i.i503 = tail call i64 @llvm.umin.i64(i64 %sub.i.i.i.i.i.i.i502, i64 64)
  %816 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %817 = and i64 %spec.select.i.i.i36.i.i.i.i503, 15
  %818 = sub i64 %spec.select.i.i.i36.i.i.i.i503, %817
  %819 = icmp ne i64 %818, 0
  br i1 %819, label %loop-memcpy-expansion2351.preheader, label %Flow3632, !amdgpu.uniform !42

loop-memcpy-expansion2351.preheader:              ; preds = %if.then.i.i.i.i.i.i.i500
  %arrayidx.i20.i35.i.i.i.i504 = getelementptr %"struct.rpc::Buffer", ptr %816, i64 %idxprom.i19.i.i.i.i.i474
  %arrayidx.i22.i38.i.i.i.i505 = getelementptr %"struct.rpc::Buffer", ptr %arrayidx.i20.i35.i.i.i.i504, i64 %sh_prom.i.i.i.i
  br label %loop-memcpy-expansion2351, !amdgpu.uniform !42

Flow3632:                                         ; preds = %Flow3631, %if.then.i.i.i.i.i.i.i500
  br label %loop-memcpy-residual-header2354, !amdgpu.uniform !42

loop-memcpy-expansion2351:                        ; preds = %loop-memcpy-expansion2351.preheader, %loop-memcpy-expansion2351
  %loop-index2352 = phi i64 [ %822, %loop-memcpy-expansion2351 ], [ 0, %loop-memcpy-expansion2351.preheader ]
  %scevgep3079 = getelementptr i8, ptr addrspace(4) %lsr.iv3077, i64 %loop-index2352, !amdgpu.uniform !42
  %820 = load <4 x i32>, ptr addrspace(4) %scevgep3079, align 1
  %821 = getelementptr i8, ptr %arrayidx.i22.i38.i.i.i.i505, i64 %loop-index2352
  store <4 x i32> %820, ptr %821, align 1
  %822 = add i64 %loop-index2352, 16
  %823 = icmp uge i64 %822, %818
  br i1 %823, label %Flow3631, label %loop-memcpy-expansion2351, !amdgpu.uniform !42

Flow3630:                                         ; preds = %Flow3629, %loop-memcpy-residual-header2354
  br label %Flow3633, !amdgpu.uniform !42

loop-memcpy-residual2353:                         ; preds = %loop-memcpy-residual2353.preheader, %loop-memcpy-residual2353
  %residual-loop-index2355 = phi i64 [ %825, %loop-memcpy-residual2353 ], [ 0, %loop-memcpy-residual2353.preheader ]
  %scevgep3088 = getelementptr i8, ptr addrspace(4) %scevgep3087, i64 %residual-loop-index2355, !amdgpu.uniform !42
  %824 = load i8, ptr addrspace(4) %scevgep3088, align 1
  %scevgep3086 = getelementptr i8, ptr %scevgep3085, i64 %residual-loop-index2355
  store i8 %824, ptr %scevgep3086, align 1
  %825 = add i64 %residual-loop-index2355, 1
  %826 = icmp uge i64 %825, %817
  br i1 %826, label %Flow3629, label %loop-memcpy-residual2353, !amdgpu.uniform !42

Flow3629:                                         ; preds = %loop-memcpy-residual2353
  br label %Flow3630, !amdgpu.uniform !42

_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i496: ; preds = %Flow3633
  %conv.i.i26.i.i.i.i497 = xor i32 %port.sroa.37.0.i.i485, 1
  fence release
  %827 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %arrayidx.i24.i29.i.i.i.i498 = getelementptr inbounds nuw i32, ptr %827, i64 %idxprom.i23.i.i.i.i.i481
  store atomic i32 %conv.i.i26.i.i.i.i497, ptr %arrayidx.i24.i29.i.i.i.i498 monotonic, align 4
  %add.i.i.i.i499 = add i64 %idx.0.i.i.i.i486, 64
  %scevgep3078 = getelementptr i8, ptr addrspace(4) %lsr.iv3077, i64 64
  %lsr.iv.next3083 = add i64 %lsr.iv3082, -64
  br label %Flow3636, !amdgpu.uniform !42

cond.false.i.i.i506:                              ; preds = %Flow3636
  %.lcssa4022 = phi i64 [ %814, %Flow3636 ]
  %port.sroa.37.0.i.i485.lcssa = phi i32 [ %port.sroa.37.0.i.i485, %Flow3636 ]
  %.lcssa4021 = phi i32 [ %800, %Flow3636 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa4022)
  %828 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %.lcssa4021)
  %829 = icmp ne i32 %828, %port.sroa.37.0.i.i485.lcssa
  %830 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %829)
  %831 = extractvalue { i1, i64 } %830, 0
  %832 = extractvalue { i1, i64 } %830, 1
  br i1 %831, label %while.body.i.i.i.i507.preheader, label %Flow3628

while.body.i.i.i.i507.preheader:                  ; preds = %cond.false.i.i.i506
  br label %while.body.i.i.i.i507, !amdgpu.uniform !42

Flow3628:                                         ; preds = %Flow3627, %cond.false.i.i.i506
  call void @llvm.amdgcn.end.cf.i64(i64 %832)
  br label %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit528, !amdgpu.uniform !42

while.body.i.i.i.i507:                            ; preds = %while.body.i.i.i.i507.preheader, %while.body.i.i.i.i507
  %phi.broken3795 = phi i64 [ 0, %while.body.i.i.i.i507.preheader ], [ %836, %while.body.i.i.i.i507 ]
  tail call void @llvm.amdgcn.s.sleep(i32 2)
  %833 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i8.i.i508 = getelementptr inbounds nuw i32, ptr %833, i64 %idxprom.i23.i.i.i.i.i481, !amdgpu.uniform !42
  %834 = load atomic i32, ptr %arrayidx.i.i.i8.i.i508 monotonic, align 4
  %835 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %834)
  %cmp.i.not.i.i.i.i509 = icmp eq i32 %835, %port.sroa.37.0.i.i485.lcssa
  %836 = call i64 @llvm.amdgcn.if.break.i64(i1 %cmp.i.not.i.i.i.i509, i64 %phi.broken3795)
  %837 = call i1 @llvm.amdgcn.loop.i64(i64 %836)
  br i1 %837, label %Flow3627, label %while.body.i.i.i.i507

Flow3627:                                         ; preds = %while.body.i.i.i.i507
  %.lcssa4020 = phi i64 [ %836, %while.body.i.i.i.i507 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa4020)
  br label %Flow3628, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit528: ; preds = %Flow3628
  fence acquire
  tail call void @llvm.amdgcn.wave.barrier()
  fence syncscope("agent") release
  %838 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %.lcssa4027, i1 true)
  %iszero.i.i.i.i.i13.i.i511 = icmp ne i64 %.lcssa4027, 0
  %cmp2.i.i.i.i14.i.i512 = icmp eq i64 %838, %sh_prom.i.i.i.i
  %cmp.i.i.i.i15.i.i513 = select i1 %iszero.i.i.i.i.i13.i.i511, i1 %cmp2.i.i.i.i14.i.i512, i1 false
  %839 = and i32 %.lcssa4028, 31
  %conv4.i.i.i19.i.i517 = zext i1 %cmp.i.i.i.i15.i.i513 to i32
  %shl.i.i.i20.i.i518 = shl nuw i32 %conv4.i.i.i19.i.i517, %839
  %xor.i.i.i21.i.i519 = xor i32 %shl.i.i.i20.i.i518, -1
  %840 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2485, !amdgpu.uniform !42

841:                                              ; preds = %ComputeEnd2486
  %div8.i.i.i16.i.i514 = lshr i32 %.lcssa4028, 5
  %idxprom.i.i.i17.i.i515 = zext nneg i32 %div8.i.i.i16.i.i514 to i64
  %842 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i17.i.i515
  %sunkaddr3254 = getelementptr inbounds i8, ptr addrspace(1) %842, i64 40
  %843 = atomicrmw and ptr addrspace(1) %sunkaddr3254, i32 %.lcssa4019 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %844, !amdgpu.uniform !42

844:                                              ; preds = %841, %ComputeEnd2486
  call void @llvm.amdgcn.end.cf.i64(i64 %3067)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %for.cond.i.i.i529, !amdgpu.uniform !42

for.cond.i.i.i529:                                ; preds = %844, %for.cond.i.i.i529
  %phi.broken3796 = phi i64 [ 0, %844 ], [ %847, %for.cond.i.i.i529 ]
  %lsr.iv3063 = phi i64 [ %lsr.iv.next3064, %for.cond.i.i.i529 ], [ -1, %844 ]
  %scevgep3075 = getelementptr i8, ptr %LHSStr, i64 %lsr.iv3063
  %scevgep3076 = getelementptr i8, ptr %scevgep3075, i64 1
  %845 = addrspacecast ptr %scevgep3076 to ptr addrspace(4)
  %846 = load i8, ptr addrspace(4) %845, align 1, !tbaa !57, !noalias.addrspace !101
  %cmp.i.i.i531 = icmp eq i8 %846, 0
  %lsr.iv.next3064 = add i64 %lsr.iv3063, 1
  %847 = call i64 @llvm.amdgcn.if.break.i64(i1 %cmp.i.i.i531, i64 %phi.broken3796)
  %848 = call i1 @llvm.amdgcn.loop.i64(i64 %847)
  br i1 %848, label %for.cond.i.i1.i537.preheader, label %for.cond.i.i.i529

for.cond.i.i1.i537.preheader:                     ; preds = %for.cond.i.i.i529
  %lsr.iv.next3064.lcssa = phi i64 [ %lsr.iv.next3064, %for.cond.i.i.i529 ]
  %.lcssa4018 = phi i64 [ %847, %for.cond.i.i.i529 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa4018)
  br label %for.cond.i.i1.i537, !amdgpu.uniform !42

for.cond.i.i1.i537:                               ; preds = %for.cond.i.i1.i537.preheader, %Flow3626
  %849 = phi i32 [ %887, %Flow3626 ], [ poison, %for.cond.i.i1.i537.preheader ]
  %850 = phi i64 [ %888, %Flow3626 ], [ poison, %for.cond.i.i1.i537.preheader ]
  %851 = phi i32 [ %889, %Flow3626 ], [ poison, %for.cond.i.i1.i537.preheader ]
  %retval.sroa.6.0.i.i.i538 = phi i32 [ %892, %Flow3626 ], [ undef, %for.cond.i.i1.i537.preheader ]
  %retval.sroa.2.0.i.i.i540 = phi i64 [ %891, %Flow3626 ], [ undef, %for.cond.i.i1.i537.preheader ]
  %retval.sroa.8.0.i.i.i541 = phi i32 [ %890, %Flow3626 ], [ undef, %for.cond.i.i1.i537.preheader ]
  %index.0.i.i.i542 = phi i32 [ %894, %Flow3626 ], [ 0, %for.cond.i.i1.i537.preheader ]
  %852 = load i32, ptr addrspace(1) @__llvm_rpc_client, align 8, !tbaa !68
  %cmp.not.i.i.i543 = icmp ult i32 %index.0.i.i.i542, %852
  %spec.store.select.i.i.i544 = select i1 %cmp.not.i.i.i543, i32 %index.0.i.i.i542, i32 0
  %853 = tail call noundef i64 @llvm.amdgcn.ballot.i64(i1 true)
  %854 = and i32 %spec.store.select.i.i.i544, 31
  %855 = lshr i64 %853, %sh_prom.i.i.i.i
  %856 = trunc i64 %855 to i32
  %conv4.i.i.i.i.i548 = and i32 %856, 1
  %shl.i.i.i.i.i549 = shl nuw i32 %conv4.i.i.i.i.i548, %854
  %857 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2489, !amdgpu.uniform !42

858:                                              ; preds = %ComputeEnd2490
  %sunkaddr3255 = getelementptr inbounds i8, ptr addrspace(1) %3077, i64 40
  %859 = atomicrmw or ptr addrspace(1) %sunkaddr3255, i32 %.lcssa4012 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %860, !amdgpu.uniform !42

860:                                              ; preds = %858, %ComputeEnd2490
  %861 = phi i32 [ %859, %858 ], [ poison, %ComputeEnd2490 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %3086)
  %862 = call i32 @llvm.amdgcn.readfirstlane.i32(i32 %861)
  %863 = or i32 %862, %.lcssa4013
  %shl5.i.i.i.i.i550 = shl nuw i32 1, %854
  %and.i.i.i.i.i551 = and i32 %shl5.i.i.i.i.i550, %863
  %tobool3.i.i.i.i552 = icmp ne i32 %and.i.i.i.i.i551, 0
  %864 = tail call i64 @llvm.amdgcn.ballot.i64(i1 %tobool3.i.i.i.i552)
  %and.i.i.i.i.i.i553 = and i64 %864, %853
  %cmp.i.not.i.i.i554 = icmp ne i64 %853, %and.i.i.i.i.i.i553
  br i1 %cmp.i.not.i.i.i554, label %if.then.i.i.i.i555, label %Flow3625, !amdgpu.uniform !42

if.then.i.i.i.i555:                               ; preds = %860
  fence syncscope("agent") acquire
  %865 = load <2 x i64>, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !75
  %866 = extractelement <2 x i64> %865, i32 0
  %867 = inttoptr i64 %866 to ptr
  %868 = extractelement <2 x i64> %865, i32 1
  %869 = inttoptr i64 %868 to ptr
  %idxprom.i.i.i.i556 = zext i32 %spec.store.select.i.i.i544 to i64
  %arrayidx.i.i.i.i557 = getelementptr inbounds nuw i32, ptr %867, i64 %idxprom.i.i.i.i556, !amdgpu.uniform !42
  %870 = load atomic i32, ptr %arrayidx.i.i.i.i557 monotonic, align 4
  %871 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %870)
  %arrayidx.i45.i.i.i558 = getelementptr inbounds nuw i32, ptr %869, i64 %idxprom.i.i.i.i556, !amdgpu.uniform !42
  %872 = load atomic i32, ptr %arrayidx.i45.i.i.i558 monotonic, align 4
  %873 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %872)
  %cmp.i46.not.i.i.i559 = icmp ne i32 %871, %873
  br i1 %cmp.i46.not.i.i.i559, label %if.then12.i.i.i560, label %Flow3623, !amdgpu.uniform !42

if.then12.i.i.i560:                               ; preds = %if.then.i.i.i.i555
  fence syncscope("agent") release
  %874 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %853, i1 true)
  %iszero.i.i.i.i.i.i.i561 = icmp ne i64 %853, 0
  %cmp2.i.i.i.i.i.i562 = icmp eq i64 %874, %sh_prom.i.i.i.i
  %cmp.i.i.i.i.i.i563 = select i1 %iszero.i.i.i.i.i.i.i561, i1 %cmp2.i.i.i.i.i.i562, i1 false
  %conv4.i.i51.i.i.i564 = zext i1 %cmp.i.i.i.i.i.i563 to i32
  %shl.i.i52.i.i.i565 = shl nuw i32 %conv4.i.i51.i.i.i564, %854
  %xor.i.i.i.i.i566 = xor i32 %shl.i.i52.i.i.i565, -1
  %875 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2494, !amdgpu.uniform !42

876:                                              ; preds = %ComputeEnd2495
  %sunkaddr3256 = getelementptr inbounds i8, ptr addrspace(1) %3077, i64 40
  %877 = atomicrmw and ptr addrspace(1) %sunkaddr3256, i32 %.lcssa4014 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %878, !amdgpu.uniform !42

878:                                              ; preds = %876, %ComputeEnd2495
  call void @llvm.amdgcn.end.cf.i64(i64 %3104)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %Flow3623, !amdgpu.uniform !42

if.end14.i.i.i626:                                ; preds = %Flow3623
  %879 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %853, i1 true)
  %iszero.i.i.i.i.i.i627 = icmp ne i64 %853, 0
  %cmp2.i.i.i.i.i628 = icmp eq i64 %879, %sh_prom.i.i.i.i
  %cmp.i.i.i.i.i629 = select i1 %iszero.i.i.i.i.i.i627, i1 %cmp2.i.i.i.i.i628, i1 false
  %880 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.i.i.i.i.i629)
  %881 = extractvalue { i1, i64 } %880, 0
  %882 = extractvalue { i1, i64 } %880, 1
  br i1 %881, label %if.then16.i.i.i631, label %if.end22.i.i.i630

if.then16.i.i.i631:                               ; preds = %if.end14.i.i.i626
  %883 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !76
  %opcode.i.i.i632 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %883, i64 %idxprom.i.i.i.i556, i32 1
  store i32 1660944387, ptr %opcode.i.i.i632, align 8, !tbaa !77
  %arrayidx21.i.i.i633 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %883, i64 %idxprom.i.i.i.i556
  store i64 %853, ptr %arrayidx21.i.i.i633, align 8, !tbaa !80
  br label %if.end22.i.i.i630, !amdgpu.uniform !42

Flow3624:                                         ; preds = %if.end22.i.i.i630, %Flow3623
  %884 = phi i32 [ %871, %if.end22.i.i.i630 ], [ %retval.sroa.8.0.i.i.i541, %Flow3623 ]
  %885 = phi i64 [ %853, %if.end22.i.i.i630 ], [ %retval.sroa.2.0.i.i.i540, %Flow3623 ]
  %886 = phi i32 [ %spec.store.select.i.i.i544, %if.end22.i.i.i630 ], [ %retval.sroa.6.0.i.i.i538, %Flow3623 ]
  br label %cleanup26.i.i.i567, !amdgpu.uniform !42

if.end22.i.i.i630:                                ; preds = %if.then16.i.i.i631, %if.end14.i.i.i626
  call void @llvm.amdgcn.end.cf.i64(i64 %882)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %Flow3624, !amdgpu.uniform !42

Flow3625:                                         ; preds = %cleanup26.i.i.i567, %860
  %887 = phi i32 [ %886, %cleanup26.i.i.i567 ], [ %849, %860 ]
  %888 = phi i64 [ %885, %cleanup26.i.i.i567 ], [ %850, %860 ]
  %889 = phi i32 [ %884, %cleanup26.i.i.i567 ], [ %851, %860 ]
  %890 = phi i32 [ %884, %cleanup26.i.i.i567 ], [ %retval.sroa.8.0.i.i.i541, %860 ]
  %891 = phi i64 [ %885, %cleanup26.i.i.i567 ], [ %retval.sroa.2.0.i.i.i540, %860 ]
  %892 = phi i32 [ %886, %cleanup26.i.i.i567 ], [ %retval.sroa.6.0.i.i.i538, %860 ]
  %893 = phi i1 [ %cmp.i46.not.i.i.i559, %cleanup26.i.i.i567 ], [ true, %860 ]
  br i1 %893, label %for.inc.i.i.i572, label %Flow3626, !amdgpu.uniform !42

cleanup26.i.i.i567:                               ; preds = %Flow3624
  br label %Flow3625, !amdgpu.uniform !42

for.inc.i.i.i572:                                 ; preds = %Flow3625
  %inc.i.i.i577 = add i32 %spec.store.select.i.i.i544, 1
  br label %Flow3626, !amdgpu.uniform !42

Flow3626:                                         ; preds = %for.inc.i.i.i572, %Flow3625
  %894 = phi i32 [ %inc.i.i.i577, %for.inc.i.i.i572 ], [ poison, %Flow3625 ]
  %895 = phi i1 [ false, %for.inc.i.i.i572 ], [ true, %Flow3625 ]
  br i1 %895, label %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i578, label %for.cond.i.i1.i537, !amdgpu.uniform !42

_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i578: ; preds = %Flow3626
  %.lcssa4017 = phi i32 [ %887, %Flow3626 ]
  %.lcssa4016 = phi i64 [ %888, %Flow3626 ]
  %.lcssa4015 = phi i32 [ %889, %Flow3626 ]
  %scevgep3074 = getelementptr i8, ptr %LHSStr, i64 %lsr.iv.next3064.lcssa
  %sub.ptr.lhs.cast.i.i.i534 = ptrtoint ptr %scevgep3074 to i64
  %sub.ptr.rhs.cast.i.i.i = ptrtoint ptr %LHSStr to i64
  %sub.ptr.sub.i.i.i535 = sub i64 %sub.ptr.lhs.cast.i.i.i534, %sub.ptr.rhs.cast.i.i.i
  fence acquire
  %896 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %mul.i.i.i.i.i.i579 = shl i32 %.lcssa4017, 6
  %idxprom.i19.i.i.i.i.i580 = zext i32 %mul.i.i.i.i.i.i579 to i64
  %arrayidx.i20.i.i.i.i.i581 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %896, i64 %idxprom.i19.i.i.i.i.i580
  %arrayidx.i22.i.i.i.i.i582 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %arrayidx.i20.i.i.i.i.i581, i64 %sh_prom.i.i.i.i
  store i64 %sub.ptr.sub.i.i.i535, ptr %arrayidx.i22.i.i.i.i.i582, align 8, !tbaa !82
  %spec.select.i.i.i.i.i.i.i583 = tail call i64 @llvm.umin.i64(i64 %sub.ptr.sub.i.i.i535, i64 56)
  %897 = and i64 %spec.select.i.i.i.i.i.i.i583, 15
  %898 = sub i64 %spec.select.i.i.i.i.i.i.i583, %897
  %899 = icmp ne i64 %898, 0
  %900 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %899)
  %901 = extractvalue { i1, i64 } %900, 0
  %902 = extractvalue { i1, i64 } %900, 1
  br i1 %901, label %loop-memcpy-expansion2243.preheader, label %Flow3622

loop-memcpy-expansion2243.preheader:              ; preds = %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i578
  %903 = add nuw nsw i64 %76, 8
  %scevgep3071 = getelementptr i8, ptr %896, i64 %903
  %904 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i580, 6
  %scevgep3072 = getelementptr i8, ptr %scevgep3071, i64 %904
  br label %loop-memcpy-expansion2243, !amdgpu.uniform !42

Flow3622:                                         ; preds = %Flow3621, %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i578
  call void @llvm.amdgcn.end.cf.i64(i64 %902)
  br label %loop-memcpy-residual-header2246, !amdgpu.uniform !42

loop-memcpy-expansion2243:                        ; preds = %loop-memcpy-expansion2243.preheader, %loop-memcpy-expansion2243
  %phi.broken3797 = phi i64 [ 0, %loop-memcpy-expansion2243.preheader ], [ %909, %loop-memcpy-expansion2243 ]
  %loop-index2244 = phi i64 [ %907, %loop-memcpy-expansion2243 ], [ 0, %loop-memcpy-expansion2243.preheader ]
  %905 = getelementptr i8, ptr %LHSStr, i64 %loop-index2244
  %906 = load <4 x i32>, ptr %905, align 1
  %scevgep3073 = getelementptr i8, ptr %scevgep3072, i64 %loop-index2244
  store <4 x i32> %906, ptr %scevgep3073, align 1
  %907 = add i64 %loop-index2244, 16
  %908 = icmp uge i64 %907, %898
  %909 = call i64 @llvm.amdgcn.if.break.i64(i1 %908, i64 %phi.broken3797)
  %910 = call i1 @llvm.amdgcn.loop.i64(i64 %909)
  br i1 %910, label %Flow3621, label %loop-memcpy-expansion2243

Flow3620:                                         ; preds = %Flow3618, %Flow3619
  %911 = phi i64 [ %2616, %Flow3618 ], [ %2609, %Flow3619 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %2612)
  br label %post-loop-memcpy-expansion2242, !amdgpu.uniform !42

loop-memcpy-residual2245:                         ; preds = %loop-memcpy-residual2245.preheader, %loop-memcpy-residual2245
  %phi.broken3798 = phi i64 [ 0, %loop-memcpy-residual2245.preheader ], [ %915, %loop-memcpy-residual2245 ]
  %residual-loop-index2247 = phi i64 [ %913, %loop-memcpy-residual2245 ], [ 0, %loop-memcpy-residual2245.preheader ]
  %scevgep3070 = getelementptr i8, ptr %scevgep3069, i64 %residual-loop-index2247
  %912 = load i8, ptr %scevgep3070, align 1
  %scevgep3068 = getelementptr i8, ptr %scevgep3067, i64 %residual-loop-index2247
  store i8 %912, ptr %scevgep3068, align 1
  %913 = add i64 %residual-loop-index2247, 1
  %914 = icmp uge i64 %913, %897
  %915 = call i64 @llvm.amdgcn.if.break.i64(i1 %914, i64 %phi.broken3798)
  %916 = call i1 @llvm.amdgcn.loop.i64(i64 %915)
  br i1 %916, label %Flow3618, label %loop-memcpy-residual2245

Flow3618:                                         ; preds = %loop-memcpy-residual2245
  %.lcssa4010 = phi i64 [ %915, %loop-memcpy-residual2245 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa4010)
  br label %Flow3620, !amdgpu.uniform !42

post-loop-memcpy-expansion2242:                   ; preds = %Flow3620
  %917 = add i64 %lsr.iv.next3064.lcssa, -56
  %tobool.not.i.i.i.i.i.i585 = icmp eq i32 %.lcssa4015, 0
  %conv.i.i.i.i.i.i586 = zext i1 %tobool.not.i.i.i.i.i.i585 to i32
  fence release
  %918 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %idxprom.i23.i.i.i.i.i587 = zext i32 %.lcssa4017 to i64
  %arrayidx.i24.i.i.i.i.i588 = getelementptr inbounds nuw i32, ptr %918, i64 %idxprom.i23.i.i.i.i.i587
  store atomic i32 %conv.i.i.i.i.i.i586, ptr %arrayidx.i24.i.i.i.i.i588 monotonic, align 4
  %919 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !84
  %arrayidx.i.i5.i.i589 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %919, i64 %idxprom.i23.i.i.i.i.i587, !amdgpu.uniform !42
  %920 = load i64, ptr %arrayidx.i.i5.i.i589, align 8, !tbaa !80
  %scevgep3046 = getelementptr i8, ptr %LHSStr, i64 56
  br label %while.cond.i.i.i.i590, !amdgpu.uniform !42

while.cond.i.i.i.i590:                            ; preds = %post-loop-memcpy-expansion2242, %Flow3617
  %phi.broken3801 = phi i64 [ 0, %post-loop-memcpy-expansion2242 ], [ %937, %Flow3617 ]
  %lsr.iv3054 = phi i64 [ %935, %Flow3617 ], [ %917, %post-loop-memcpy-expansion2242 ]
  %lsr.iv3047 = phi ptr [ %934, %Flow3617 ], [ %scevgep3046, %post-loop-memcpy-expansion2242 ]
  %port.sroa.37.0.i.i591 = phi i32 [ %933, %Flow3617 ], [ %conv.i.i.i.i.i.i586, %post-loop-memcpy-expansion2242 ]
  %idx.0.i.i.i.i592 = phi i64 [ %932, %Flow3617 ], [ 56, %post-loop-memcpy-expansion2242 ]
  %cmp.i.i.i.i593 = icmp ugt i64 %sub.ptr.sub.i.i.i535, %idx.0.i.i.i.i592
  %921 = tail call i64 @llvm.amdgcn.ballot.i64(i1 %cmp.i.i.i.i593)
  %and.i.i.i.i6.i.i594 = and i64 %921, %920
  %tobool.not.i.i.i.i595 = icmp ne i64 %and.i.i.i.i6.i.i594, 0
  %922 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i22.i.i.i596 = getelementptr inbounds nuw i32, ptr %922, i64 %idxprom.i23.i.i.i.i.i587, !amdgpu.uniform !42
  %923 = load atomic i32, ptr %arrayidx.i22.i.i.i596 monotonic, align 4
  %924 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %tobool.not.i.i.i.i595)
  %925 = extractvalue { i1, i64 } %924, 0
  %926 = extractvalue { i1, i64 } %924, 1
  br i1 %925, label %cond.false.i8.i.i.i.i597, label %Flow3617

cond.false.i8.i.i.i.i597:                         ; preds = %while.cond.i.i.i.i590
  %927 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %923)
  %928 = icmp ne i32 %927, %port.sroa.37.0.i.i591
  br i1 %928, label %while.body.i.i21.i.i.i.i598.preheader, label %Flow3616, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i598.preheader:            ; preds = %cond.false.i8.i.i.i.i597
  br label %while.body.i.i21.i.i.i.i598, !amdgpu.uniform !42

Flow3616:                                         ; preds = %Flow3615, %cond.false.i8.i.i.i.i597
  br label %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i601, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i598:                      ; preds = %while.body.i.i21.i.i.i.i598.preheader, %while.body.i.i21.i.i.i.i598
  tail call void @llvm.amdgcn.s.sleep(i32 2)
  %929 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i22.i.i.i.i599 = getelementptr inbounds nuw i32, ptr %929, i64 %idxprom.i23.i.i.i.i.i587, !amdgpu.uniform !42
  %930 = load atomic i32, ptr %arrayidx.i.i.i22.i.i.i.i599 monotonic, align 4
  %931 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %930)
  %cmp.i.not.i.i23.i.i.i.i600 = icmp eq i32 %931, %port.sroa.37.0.i.i591
  br i1 %cmp.i.not.i.i23.i.i.i.i600, label %Flow3615, label %while.body.i.i21.i.i.i.i598, !amdgpu.uniform !42

Flow3615:                                         ; preds = %while.body.i.i21.i.i.i.i598
  br label %Flow3616, !amdgpu.uniform !42

Flow3617:                                         ; preds = %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i602, %while.cond.i.i.i.i590
  %932 = phi i64 [ %add.i.i.i.i605, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i602 ], [ poison, %while.cond.i.i.i.i590 ]
  %933 = phi i32 [ %conv.i.i26.i.i.i.i603, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i602 ], [ poison, %while.cond.i.i.i.i590 ]
  %934 = phi ptr [ %scevgep3048, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i602 ], [ poison, %while.cond.i.i.i.i590 ]
  %935 = phi i64 [ %lsr.iv.next3055, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i602 ], [ poison, %while.cond.i.i.i.i590 ]
  %936 = phi i1 [ false, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i602 ], [ true, %while.cond.i.i.i.i590 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %926)
  %937 = call i64 @llvm.amdgcn.if.break.i64(i1 %936, i64 %phi.broken3801)
  %938 = call i1 @llvm.amdgcn.loop.i64(i64 %937)
  br i1 %938, label %cond.false.i.i.i612, label %while.cond.i.i.i.i590

_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i601: ; preds = %Flow3616
  fence acquire
  %939 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.i.i.i.i593)
  %940 = extractvalue { i1, i64 } %939, 0
  %941 = extractvalue { i1, i64 } %939, 1
  br i1 %940, label %if.then.i.i.i.i.i.i.i606, label %Flow3614

if.then.i.i.i.i.i.i.i606:                         ; preds = %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i601
  %sub.i.i.i.i.i.i.i608 = sub nuw i64 %sub.ptr.sub.i.i.i535, %idx.0.i.i.i.i592
  %spec.select.i.i.i36.i.i.i.i609 = tail call i64 @llvm.umin.i64(i64 %sub.i.i.i.i.i.i.i608, i64 64)
  %942 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %943 = and i64 %spec.select.i.i.i36.i.i.i.i609, 15
  %944 = sub i64 %spec.select.i.i.i36.i.i.i.i609, %943
  %945 = icmp ne i64 %944, 0
  %946 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %945)
  %947 = extractvalue { i1, i64 } %946, 0
  %948 = extractvalue { i1, i64 } %946, 1
  br i1 %947, label %loop-memcpy-expansion2237.preheader, label %Flow3613

loop-memcpy-expansion2237.preheader:              ; preds = %if.then.i.i.i.i.i.i.i606
  %arrayidx.i20.i35.i.i.i.i610 = getelementptr %"struct.rpc::Buffer", ptr %942, i64 %idxprom.i19.i.i.i.i.i580
  %arrayidx.i22.i38.i.i.i.i611 = getelementptr %"struct.rpc::Buffer", ptr %arrayidx.i20.i35.i.i.i.i610, i64 %sh_prom.i.i.i.i
  br label %loop-memcpy-expansion2237, !amdgpu.uniform !42

Flow3613:                                         ; preds = %Flow3612, %if.then.i.i.i.i.i.i.i606
  call void @llvm.amdgcn.end.cf.i64(i64 %948)
  br label %loop-memcpy-residual-header2240, !amdgpu.uniform !42

loop-memcpy-expansion2237:                        ; preds = %loop-memcpy-expansion2237.preheader, %loop-memcpy-expansion2237
  %phi.broken3799 = phi i64 [ 0, %loop-memcpy-expansion2237.preheader ], [ %953, %loop-memcpy-expansion2237 ]
  %loop-index2238 = phi i64 [ %951, %loop-memcpy-expansion2237 ], [ 0, %loop-memcpy-expansion2237.preheader ]
  %scevgep3049 = getelementptr i8, ptr %lsr.iv3047, i64 %loop-index2238
  %949 = load <4 x i32>, ptr %scevgep3049, align 1
  %950 = getelementptr i8, ptr %arrayidx.i22.i38.i.i.i.i611, i64 %loop-index2238
  store <4 x i32> %949, ptr %950, align 1
  %951 = add i64 %loop-index2238, 16
  %952 = icmp uge i64 %951, %944
  %953 = call i64 @llvm.amdgcn.if.break.i64(i1 %952, i64 %phi.broken3799)
  %954 = call i1 @llvm.amdgcn.loop.i64(i64 %953)
  br i1 %954, label %Flow3612, label %loop-memcpy-expansion2237

Flow3611:                                         ; preds = %Flow3610, %loop-memcpy-residual-header2240
  call void @llvm.amdgcn.end.cf.i64(i64 %2602)
  br label %Flow3614, !amdgpu.uniform !42

loop-memcpy-residual2239:                         ; preds = %loop-memcpy-residual2239.preheader, %loop-memcpy-residual2239
  %phi.broken3800 = phi i64 [ 0, %loop-memcpy-residual2239.preheader ], [ %958, %loop-memcpy-residual2239 ]
  %residual-loop-index2241 = phi i64 [ %956, %loop-memcpy-residual2239 ], [ 0, %loop-memcpy-residual2239.preheader ]
  %scevgep3061 = getelementptr i8, ptr %scevgep3060, i64 %residual-loop-index2241
  %955 = load i8, ptr %scevgep3061, align 1
  %scevgep3058 = getelementptr i8, ptr %scevgep3057, i64 %residual-loop-index2241
  store i8 %955, ptr %scevgep3058, align 1
  %956 = add i64 %residual-loop-index2241, 1
  %957 = icmp uge i64 %956, %943
  %958 = call i64 @llvm.amdgcn.if.break.i64(i1 %957, i64 %phi.broken3800)
  %959 = call i1 @llvm.amdgcn.loop.i64(i64 %958)
  br i1 %959, label %Flow3610, label %loop-memcpy-residual2239

Flow3610:                                         ; preds = %loop-memcpy-residual2239
  %.lcssa4007 = phi i64 [ %958, %loop-memcpy-residual2239 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa4007)
  br label %Flow3611, !amdgpu.uniform !42

_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i602: ; preds = %Flow3614
  %conv.i.i26.i.i.i.i603 = xor i32 %port.sroa.37.0.i.i591, 1
  fence release
  %960 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %arrayidx.i24.i29.i.i.i.i604 = getelementptr inbounds nuw i32, ptr %960, i64 %idxprom.i23.i.i.i.i.i587
  store atomic i32 %conv.i.i26.i.i.i.i603, ptr %arrayidx.i24.i29.i.i.i.i604 monotonic, align 4
  %add.i.i.i.i605 = add i64 %idx.0.i.i.i.i592, 64
  %scevgep3048 = getelementptr i8, ptr %lsr.iv3047, i64 64
  %lsr.iv.next3055 = add i64 %lsr.iv3054, -64
  br label %Flow3617, !amdgpu.uniform !42

cond.false.i.i.i612:                              ; preds = %Flow3617
  %.lcssa4009 = phi i64 [ %937, %Flow3617 ]
  %port.sroa.37.0.i.i591.lcssa = phi i32 [ %port.sroa.37.0.i.i591, %Flow3617 ]
  %.lcssa4008 = phi i32 [ %923, %Flow3617 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa4009)
  %961 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %.lcssa4008)
  %962 = icmp ne i32 %961, %port.sroa.37.0.i.i591.lcssa
  %963 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %962)
  %964 = extractvalue { i1, i64 } %963, 0
  %965 = extractvalue { i1, i64 } %963, 1
  br i1 %964, label %while.body.i.i.i.i613.preheader, label %Flow3609

while.body.i.i.i.i613.preheader:                  ; preds = %cond.false.i.i.i612
  br label %while.body.i.i.i.i613, !amdgpu.uniform !42

Flow3609:                                         ; preds = %Flow3608, %cond.false.i.i.i612
  call void @llvm.amdgcn.end.cf.i64(i64 %965)
  br label %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit634, !amdgpu.uniform !42

while.body.i.i.i.i613:                            ; preds = %while.body.i.i.i.i613.preheader, %while.body.i.i.i.i613
  %phi.broken3802 = phi i64 [ 0, %while.body.i.i.i.i613.preheader ], [ %969, %while.body.i.i.i.i613 ]
  tail call void @llvm.amdgcn.s.sleep(i32 2)
  %966 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i8.i.i614 = getelementptr inbounds nuw i32, ptr %966, i64 %idxprom.i23.i.i.i.i.i587, !amdgpu.uniform !42
  %967 = load atomic i32, ptr %arrayidx.i.i.i8.i.i614 monotonic, align 4
  %968 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %967)
  %cmp.i.not.i.i.i.i615 = icmp eq i32 %968, %port.sroa.37.0.i.i591.lcssa
  %969 = call i64 @llvm.amdgcn.if.break.i64(i1 %cmp.i.not.i.i.i.i615, i64 %phi.broken3802)
  %970 = call i1 @llvm.amdgcn.loop.i64(i64 %969)
  br i1 %970, label %Flow3608, label %while.body.i.i.i.i613

Flow3608:                                         ; preds = %while.body.i.i.i.i613
  %.lcssa4005 = phi i64 [ %969, %while.body.i.i.i.i613 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa4005)
  br label %Flow3609, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit634: ; preds = %Flow3609
  fence acquire
  tail call void @llvm.amdgcn.wave.barrier()
  fence syncscope("agent") release
  %971 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %.lcssa4016, i1 true)
  %iszero.i.i.i.i.i13.i.i617 = icmp ne i64 %.lcssa4016, 0
  %cmp2.i.i.i.i14.i.i618 = icmp eq i64 %971, %sh_prom.i.i.i.i
  %cmp.i.i.i.i15.i.i619 = select i1 %iszero.i.i.i.i.i13.i.i617, i1 %cmp2.i.i.i.i14.i.i618, i1 false
  %972 = and i32 %.lcssa4017, 31
  %conv4.i.i.i19.i.i623 = zext i1 %cmp.i.i.i.i15.i.i619 to i32
  %shl.i.i.i20.i.i624 = shl nuw i32 %conv4.i.i.i19.i.i623, %972
  %xor.i.i.i21.i.i625 = xor i32 %shl.i.i.i20.i.i624, -1
  %973 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2498, !amdgpu.uniform !42

974:                                              ; preds = %ComputeEnd2499
  %div8.i.i.i16.i.i620 = lshr i32 %.lcssa4017, 5
  %idxprom.i.i.i17.i.i621 = zext nneg i32 %div8.i.i.i16.i.i620 to i64
  %975 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i17.i.i621
  %sunkaddr3257 = getelementptr inbounds i8, ptr addrspace(1) %975, i64 40
  %976 = atomicrmw and ptr addrspace(1) %sunkaddr3257, i32 %.lcssa4004 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %977, !amdgpu.uniform !42

977:                                              ; preds = %974, %ComputeEnd2499
  call void @llvm.amdgcn.end.cf.i64(i64 %3121)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %for.cond.i.i.i636, !amdgpu.uniform !42

for.cond.i.i.i636:                                ; preds = %977, %Flow3607
  %978 = phi i32 [ %1016, %Flow3607 ], [ poison, %977 ]
  %979 = phi i64 [ %1017, %Flow3607 ], [ poison, %977 ]
  %980 = phi i32 [ %1018, %Flow3607 ], [ poison, %977 ]
  %retval.sroa.6.0.i.i.i637 = phi i32 [ %1021, %Flow3607 ], [ undef, %977 ]
  %retval.sroa.2.0.i.i.i639 = phi i64 [ %1020, %Flow3607 ], [ undef, %977 ]
  %retval.sroa.8.0.i.i.i640 = phi i32 [ %1019, %Flow3607 ], [ undef, %977 ]
  %index.0.i.i.i641 = phi i32 [ %1023, %Flow3607 ], [ 0, %977 ]
  %981 = load i32, ptr addrspace(1) @__llvm_rpc_client, align 8, !tbaa !68
  %cmp.not.i.i.i642 = icmp ult i32 %index.0.i.i.i641, %981
  %spec.store.select.i.i.i643 = select i1 %cmp.not.i.i.i642, i32 %index.0.i.i.i641, i32 0
  %982 = tail call noundef i64 @llvm.amdgcn.ballot.i64(i1 true)
  %983 = and i32 %spec.store.select.i.i.i643, 31
  %984 = lshr i64 %982, %sh_prom.i.i.i.i
  %985 = trunc i64 %984 to i32
  %conv4.i.i.i.i.i647 = and i32 %985, 1
  %shl.i.i.i.i.i648 = shl nuw i32 %conv4.i.i.i.i.i647, %983
  %986 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2502, !amdgpu.uniform !42

987:                                              ; preds = %ComputeEnd2503
  %sunkaddr3258 = getelementptr inbounds i8, ptr addrspace(1) %3131, i64 40
  %988 = atomicrmw or ptr addrspace(1) %sunkaddr3258, i32 %.lcssa3998 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %989, !amdgpu.uniform !42

989:                                              ; preds = %987, %ComputeEnd2503
  %990 = phi i32 [ %988, %987 ], [ poison, %ComputeEnd2503 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %3140)
  %991 = call i32 @llvm.amdgcn.readfirstlane.i32(i32 %990)
  %992 = or i32 %991, %.lcssa3999
  %shl5.i.i.i.i.i649 = shl nuw i32 1, %983
  %and.i.i.i.i.i650 = and i32 %shl5.i.i.i.i.i649, %992
  %tobool3.i.i.i.i651 = icmp ne i32 %and.i.i.i.i.i650, 0
  %993 = tail call i64 @llvm.amdgcn.ballot.i64(i1 %tobool3.i.i.i.i651)
  %and.i.i.i.i.i.i652 = and i64 %993, %982
  %cmp.i.not.i.i.i653 = icmp ne i64 %982, %and.i.i.i.i.i.i652
  br i1 %cmp.i.not.i.i.i653, label %if.then.i.i.i.i654, label %Flow3606, !amdgpu.uniform !42

if.then.i.i.i.i654:                               ; preds = %989
  fence syncscope("agent") acquire
  %994 = load <2 x i64>, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !75
  %995 = extractelement <2 x i64> %994, i32 0
  %996 = inttoptr i64 %995 to ptr
  %997 = extractelement <2 x i64> %994, i32 1
  %998 = inttoptr i64 %997 to ptr
  %idxprom.i.i.i.i655 = zext i32 %spec.store.select.i.i.i643 to i64
  %arrayidx.i.i.i.i656 = getelementptr inbounds nuw i32, ptr %996, i64 %idxprom.i.i.i.i655, !amdgpu.uniform !42
  %999 = load atomic i32, ptr %arrayidx.i.i.i.i656 monotonic, align 4
  %1000 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %999)
  %arrayidx.i45.i.i.i657 = getelementptr inbounds nuw i32, ptr %998, i64 %idxprom.i.i.i.i655, !amdgpu.uniform !42
  %1001 = load atomic i32, ptr %arrayidx.i45.i.i.i657 monotonic, align 4
  %1002 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %1001)
  %cmp.i46.not.i.i.i658 = icmp ne i32 %1000, %1002
  br i1 %cmp.i46.not.i.i.i658, label %if.then12.i.i.i659, label %Flow3604, !amdgpu.uniform !42

if.then12.i.i.i659:                               ; preds = %if.then.i.i.i.i654
  fence syncscope("agent") release
  %1003 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %982, i1 true)
  %iszero.i.i.i.i.i.i.i660 = icmp ne i64 %982, 0
  %cmp2.i.i.i.i.i.i661 = icmp eq i64 %1003, %sh_prom.i.i.i.i
  %cmp.i.i.i.i.i.i662 = select i1 %iszero.i.i.i.i.i.i.i660, i1 %cmp2.i.i.i.i.i.i661, i1 false
  %conv4.i.i51.i.i.i663 = zext i1 %cmp.i.i.i.i.i.i662 to i32
  %shl.i.i52.i.i.i664 = shl nuw i32 %conv4.i.i51.i.i.i663, %983
  %xor.i.i.i.i.i665 = xor i32 %shl.i.i52.i.i.i664, -1
  %1004 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2507, !amdgpu.uniform !42

1005:                                             ; preds = %ComputeEnd2508
  %sunkaddr3259 = getelementptr inbounds i8, ptr addrspace(1) %3131, i64 40
  %1006 = atomicrmw and ptr addrspace(1) %sunkaddr3259, i32 %.lcssa4000 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %1007, !amdgpu.uniform !42

1007:                                             ; preds = %1005, %ComputeEnd2508
  call void @llvm.amdgcn.end.cf.i64(i64 %3158)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %Flow3604, !amdgpu.uniform !42

if.end14.i.i.i723:                                ; preds = %Flow3604
  %1008 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %982, i1 true)
  %iszero.i.i.i.i.i.i724 = icmp ne i64 %982, 0
  %cmp2.i.i.i.i.i725 = icmp eq i64 %1008, %sh_prom.i.i.i.i
  %cmp.i.i.i.i.i726 = select i1 %iszero.i.i.i.i.i.i724, i1 %cmp2.i.i.i.i.i725, i1 false
  %1009 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.i.i.i.i.i726)
  %1010 = extractvalue { i1, i64 } %1009, 0
  %1011 = extractvalue { i1, i64 } %1009, 1
  br i1 %1010, label %if.then16.i.i.i728, label %if.end22.i.i.i727

if.then16.i.i.i728:                               ; preds = %if.end14.i.i.i723
  %1012 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !76
  %opcode.i.i.i729 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %1012, i64 %idxprom.i.i.i.i655, i32 1
  store i32 1660944387, ptr %opcode.i.i.i729, align 8, !tbaa !77
  %arrayidx21.i.i.i730 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %1012, i64 %idxprom.i.i.i.i655
  store i64 %982, ptr %arrayidx21.i.i.i730, align 8, !tbaa !80
  br label %if.end22.i.i.i727, !amdgpu.uniform !42

Flow3605:                                         ; preds = %if.end22.i.i.i727, %Flow3604
  %1013 = phi i32 [ %1000, %if.end22.i.i.i727 ], [ %retval.sroa.8.0.i.i.i640, %Flow3604 ]
  %1014 = phi i64 [ %982, %if.end22.i.i.i727 ], [ %retval.sroa.2.0.i.i.i639, %Flow3604 ]
  %1015 = phi i32 [ %spec.store.select.i.i.i643, %if.end22.i.i.i727 ], [ %retval.sroa.6.0.i.i.i637, %Flow3604 ]
  br label %cleanup26.i.i.i666, !amdgpu.uniform !42

if.end22.i.i.i727:                                ; preds = %if.then16.i.i.i728, %if.end14.i.i.i723
  call void @llvm.amdgcn.end.cf.i64(i64 %1011)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %Flow3605, !amdgpu.uniform !42

Flow3606:                                         ; preds = %cleanup26.i.i.i666, %989
  %1016 = phi i32 [ %1015, %cleanup26.i.i.i666 ], [ %978, %989 ]
  %1017 = phi i64 [ %1014, %cleanup26.i.i.i666 ], [ %979, %989 ]
  %1018 = phi i32 [ %1013, %cleanup26.i.i.i666 ], [ %980, %989 ]
  %1019 = phi i32 [ %1013, %cleanup26.i.i.i666 ], [ %retval.sroa.8.0.i.i.i640, %989 ]
  %1020 = phi i64 [ %1014, %cleanup26.i.i.i666 ], [ %retval.sroa.2.0.i.i.i639, %989 ]
  %1021 = phi i32 [ %1015, %cleanup26.i.i.i666 ], [ %retval.sroa.6.0.i.i.i637, %989 ]
  %1022 = phi i1 [ %cmp.i46.not.i.i.i658, %cleanup26.i.i.i666 ], [ true, %989 ]
  br i1 %1022, label %for.inc.i.i.i671, label %Flow3607, !amdgpu.uniform !42

cleanup26.i.i.i666:                               ; preds = %Flow3605
  br label %Flow3606, !amdgpu.uniform !42

for.inc.i.i.i671:                                 ; preds = %Flow3606
  %inc.i.i.i676 = add i32 %spec.store.select.i.i.i643, 1
  br label %Flow3607, !amdgpu.uniform !42

Flow3607:                                         ; preds = %for.inc.i.i.i671, %Flow3606
  %1023 = phi i32 [ %inc.i.i.i676, %for.inc.i.i.i671 ], [ poison, %Flow3606 ]
  %1024 = phi i1 [ false, %for.inc.i.i.i671 ], [ true, %Flow3606 ]
  br i1 %1024, label %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i677, label %for.cond.i.i.i636, !amdgpu.uniform !42

_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i677: ; preds = %Flow3607
  %.lcssa4003 = phi i32 [ %1016, %Flow3607 ]
  %.lcssa4002 = phi i64 [ %1017, %Flow3607 ]
  %.lcssa4001 = phi i32 [ %1018, %Flow3607 ]
  fence acquire
  %1025 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %mul.i.i.i.i.i.i678 = shl i32 %.lcssa4003, 6
  %idxprom.i19.i.i.i.i.i679 = zext i32 %mul.i.i.i.i.i.i678 to i64
  %arrayidx.i20.i.i.i.i.i680 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %1025, i64 %idxprom.i19.i.i.i.i.i679
  %arrayidx.i22.i.i.i.i.i681 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %arrayidx.i20.i.i.i.i.i680, i64 %sh_prom.i.i.i.i
  store i64 1, ptr %arrayidx.i22.i.i.i.i.i681, align 8, !tbaa !82
  %arrayidx6.i.i.i.i.i.i.i682 = getelementptr inbounds nuw i8, ptr %arrayidx.i22.i.i.i.i.i681, i64 8
  store i8 10, ptr %arrayidx6.i.i.i.i.i.i.i682, align 8
  %tobool.not.i.i.i.i.i.i683 = icmp eq i32 %.lcssa4001, 0
  %conv.i.i.i.i.i.i684 = zext i1 %tobool.not.i.i.i.i.i.i683 to i32
  fence release
  %1026 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %idxprom.i23.i.i.i.i.i685 = zext i32 %.lcssa4003 to i64
  %arrayidx.i24.i.i.i.i.i686 = getelementptr inbounds nuw i32, ptr %1026, i64 %idxprom.i23.i.i.i.i.i685
  store atomic i32 %conv.i.i.i.i.i.i684, ptr %arrayidx.i24.i.i.i.i.i686 monotonic, align 4
  %1027 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !84
  %arrayidx.i.i5.i.i687 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %1027, i64 %idxprom.i23.i.i.i.i.i685, !amdgpu.uniform !42
  %1028 = load i64, ptr %arrayidx.i.i5.i.i687, align 8, !tbaa !80
  br label %while.cond.i.i.i.i688, !amdgpu.uniform !42

while.cond.i.i.i.i688:                            ; preds = %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i677, %Flow3603
  %phi.broken3803 = phi i64 [ 0, %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i677 ], [ %1043, %Flow3603 ]
  %port.sroa.37.0.i.i689 = phi i32 [ %1041, %Flow3603 ], [ %conv.i.i.i.i.i.i684, %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i677 ]
  %idx.0.i.i.i.i690 = phi i64 [ %1040, %Flow3603 ], [ 56, %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i677 ]
  %cmp.i.i.i.i691 = icmp eq i64 %idx.0.i.i.i.i690, 0
  %1029 = tail call i64 @llvm.amdgcn.ballot.i64(i1 %cmp.i.i.i.i691)
  %and.i.i.i.i6.i.i692 = and i64 %1029, %1028
  %tobool.not.i.i.i.i693 = icmp ne i64 %and.i.i.i.i6.i.i692, 0
  %1030 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i22.i.i.i694 = getelementptr inbounds nuw i32, ptr %1030, i64 %idxprom.i23.i.i.i.i.i685, !amdgpu.uniform !42
  %1031 = load atomic i32, ptr %arrayidx.i22.i.i.i694 monotonic, align 4
  %1032 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %tobool.not.i.i.i.i693)
  %1033 = extractvalue { i1, i64 } %1032, 0
  %1034 = extractvalue { i1, i64 } %1032, 1
  br i1 %1033, label %cond.false.i8.i.i.i.i695, label %Flow3603

cond.false.i8.i.i.i.i695:                         ; preds = %while.cond.i.i.i.i688
  %1035 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %1031)
  %1036 = icmp ne i32 %1035, %port.sroa.37.0.i.i689
  br i1 %1036, label %while.body.i.i21.i.i.i.i696.preheader, label %Flow3602, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i696.preheader:            ; preds = %cond.false.i8.i.i.i.i695
  br label %while.body.i.i21.i.i.i.i696, !amdgpu.uniform !42

Flow3602:                                         ; preds = %Flow3601, %cond.false.i8.i.i.i.i695
  br label %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i699, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i696:                      ; preds = %while.body.i.i21.i.i.i.i696.preheader, %while.body.i.i21.i.i.i.i696
  tail call void @llvm.amdgcn.s.sleep(i32 2)
  %1037 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i22.i.i.i.i697 = getelementptr inbounds nuw i32, ptr %1037, i64 %idxprom.i23.i.i.i.i.i685, !amdgpu.uniform !42
  %1038 = load atomic i32, ptr %arrayidx.i.i.i22.i.i.i.i697 monotonic, align 4
  %1039 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %1038)
  %cmp.i.not.i.i23.i.i.i.i698 = icmp eq i32 %1039, %port.sroa.37.0.i.i689
  br i1 %cmp.i.not.i.i23.i.i.i.i698, label %Flow3601, label %while.body.i.i21.i.i.i.i696, !amdgpu.uniform !42

Flow3601:                                         ; preds = %while.body.i.i21.i.i.i.i696
  br label %Flow3602, !amdgpu.uniform !42

Flow3603:                                         ; preds = %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i700, %while.cond.i.i.i.i688
  %1040 = phi i64 [ %add.i.i.i.i703, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i700 ], [ poison, %while.cond.i.i.i.i688 ]
  %1041 = phi i32 [ %conv.i.i26.i.i.i.i701, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i700 ], [ poison, %while.cond.i.i.i.i688 ]
  %1042 = phi i1 [ false, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i700 ], [ true, %while.cond.i.i.i.i688 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %1034)
  %1043 = call i64 @llvm.amdgcn.if.break.i64(i1 %1042, i64 %phi.broken3803)
  %1044 = call i1 @llvm.amdgcn.loop.i64(i64 %1043)
  br i1 %1044, label %cond.false.i.i.i709, label %while.cond.i.i.i.i688

_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i699: ; preds = %Flow3602
  fence acquire
  br i1 %cmp.i.i.i.i691, label %if.then.i.i.i.i.i.i.i704, label %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i700, !amdgpu.uniform !42

if.then.i.i.i.i.i.i.i704:                         ; preds = %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i699
  %1045 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %arrayidx.i20.i35.i.i.i.i707 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %1045, i64 %idxprom.i19.i.i.i.i.i679
  %arrayidx.i22.i38.i.i.i.i708 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %arrayidx.i20.i35.i.i.i.i707, i64 %sh_prom.i.i.i.i
  store i8 10, ptr %arrayidx.i22.i38.i.i.i.i708, align 1
  br label %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i700, !amdgpu.uniform !42

_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i700: ; preds = %if.then.i.i.i.i.i.i.i704, %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i699
  %conv.i.i26.i.i.i.i701 = xor i32 %port.sroa.37.0.i.i689, 1
  fence release
  %1046 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %arrayidx.i24.i29.i.i.i.i702 = getelementptr inbounds nuw i32, ptr %1046, i64 %idxprom.i23.i.i.i.i.i685
  store atomic i32 %conv.i.i26.i.i.i.i701, ptr %arrayidx.i24.i29.i.i.i.i702 monotonic, align 4
  %add.i.i.i.i703 = add i64 %idx.0.i.i.i.i690, 64
  br label %Flow3603, !amdgpu.uniform !42

cond.false.i.i.i709:                              ; preds = %Flow3603
  %.lcssa3997 = phi i64 [ %1043, %Flow3603 ]
  %port.sroa.37.0.i.i689.lcssa = phi i32 [ %port.sroa.37.0.i.i689, %Flow3603 ]
  %.lcssa3996 = phi i32 [ %1031, %Flow3603 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa3997)
  %1047 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %.lcssa3996)
  %1048 = icmp ne i32 %1047, %port.sroa.37.0.i.i689.lcssa
  %1049 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %1048)
  %1050 = extractvalue { i1, i64 } %1049, 0
  %1051 = extractvalue { i1, i64 } %1049, 1
  br i1 %1050, label %while.body.i.i.i.i710.preheader, label %Flow3600

while.body.i.i.i.i710.preheader:                  ; preds = %cond.false.i.i.i709
  br label %while.body.i.i.i.i710, !amdgpu.uniform !42

Flow3600:                                         ; preds = %Flow3599, %cond.false.i.i.i709
  call void @llvm.amdgcn.end.cf.i64(i64 %1051)
  br label %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIcEERS1_T_.exit, !amdgpu.uniform !42

while.body.i.i.i.i710:                            ; preds = %while.body.i.i.i.i710.preheader, %while.body.i.i.i.i710
  %phi.broken3804 = phi i64 [ 0, %while.body.i.i.i.i710.preheader ], [ %1055, %while.body.i.i.i.i710 ]
  tail call void @llvm.amdgcn.s.sleep(i32 2)
  %1052 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i8.i.i711 = getelementptr inbounds nuw i32, ptr %1052, i64 %idxprom.i23.i.i.i.i.i685, !amdgpu.uniform !42
  %1053 = load atomic i32, ptr %arrayidx.i.i.i8.i.i711 monotonic, align 4
  %1054 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %1053)
  %cmp.i.not.i.i.i.i712 = icmp eq i32 %1054, %port.sroa.37.0.i.i689.lcssa
  %1055 = call i64 @llvm.amdgcn.if.break.i64(i1 %cmp.i.not.i.i.i.i712, i64 %phi.broken3804)
  %1056 = call i1 @llvm.amdgcn.loop.i64(i64 %1055)
  br i1 %1056, label %Flow3599, label %while.body.i.i.i.i710

Flow3599:                                         ; preds = %while.body.i.i.i.i710
  %.lcssa3995 = phi i64 [ %1055, %while.body.i.i.i.i710 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa3995)
  br label %Flow3600, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIcEERS1_T_.exit: ; preds = %Flow3600
  fence acquire
  tail call void @llvm.amdgcn.wave.barrier()
  fence syncscope("agent") release
  %1057 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %.lcssa4002, i1 true)
  %iszero.i.i.i.i.i13.i.i714 = icmp ne i64 %.lcssa4002, 0
  %cmp2.i.i.i.i14.i.i715 = icmp eq i64 %1057, %sh_prom.i.i.i.i
  %cmp.i.i.i.i15.i.i716 = select i1 %iszero.i.i.i.i.i13.i.i714, i1 %cmp2.i.i.i.i14.i.i715, i1 false
  %1058 = and i32 %.lcssa4003, 31
  %conv4.i.i.i19.i.i720 = zext i1 %cmp.i.i.i.i15.i.i716 to i32
  %shl.i.i.i20.i.i721 = shl nuw i32 %conv4.i.i.i19.i.i720, %1058
  %xor.i.i.i21.i.i722 = xor i32 %shl.i.i.i20.i.i721, -1
  %1059 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2511, !amdgpu.uniform !42

1060:                                             ; preds = %ComputeEnd2512
  %div8.i.i.i16.i.i717 = lshr i32 %.lcssa4003, 5
  %idxprom.i.i.i17.i.i718 = zext nneg i32 %div8.i.i.i16.i.i717 to i64
  %1061 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i17.i.i718
  %sunkaddr3260 = getelementptr inbounds i8, ptr addrspace(1) %1061, i64 40
  %1062 = atomicrmw and ptr addrspace(1) %sunkaddr3260, i32 %.lcssa3994 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %1063, !amdgpu.uniform !42

1063:                                             ; preds = %1060, %ComputeEnd2512
  call void @llvm.amdgcn.end.cf.i64(i64 %3175)
  tail call void @llvm.amdgcn.wave.barrier()
  br i1 %cmp2.i.i.i37.i, label %if.then3.i.i.i67.i, label %if.end4.i.i.i56.i, !amdgpu.uniform !42

if.then3.i.i.i67.i:                               ; preds = %1063
  %mul.i.i.i68.i = mul nuw i64 %inc.i.i.i34.i, 11
  %div18.i.i.i69.i = lshr i64 %mul.i.i.i68.i, 3
  br label %if.end4.i.i.i56.i, !amdgpu.uniform !42

if.end4.i.i.i56.i:                                ; preds = %if.then3.i.i.i67.i, %1063
  %new_capacity.addr.0.i.i.i57.i = phi i64 [ %div18.i.i.i69.i, %if.then3.i.i.i67.i ], [ %inc.i.i.i34.i, %1063 ]
  %sub.i.i731 = add i64 %new_capacity.addr.0.i.i.i57.i, 7
  %div2.i.i732 = and i64 %sub.i.i731, -8
  %1064 = load ptr, ptr addrspace(1) @_ZN12_GLOBAL__N_13ptrE, align 8, !tbaa !55
  %add.ptr.i.i733 = getelementptr inbounds nuw i8, ptr %1064, i64 %div2.i.i732
  store ptr %add.ptr.i.i733, ptr addrspace(1) @_ZN12_GLOBAL__N_13ptrE, align 8, !tbaa !55
  br i1 %613, label %loop-memcpy-expansion2279.preheader, label %Flow3598, !amdgpu.uniform !42

loop-memcpy-expansion2279.preheader:              ; preds = %if.end4.i.i.i56.i
  br label %loop-memcpy-expansion2279, !amdgpu.uniform !42

Flow3598:                                         ; preds = %Flow3597, %if.end4.i.i.i56.i
  br label %loop-memcpy-residual-header2282, !amdgpu.uniform !42

loop-memcpy-expansion2279:                        ; preds = %loop-memcpy-expansion2279.preheader, %loop-memcpy-expansion2279
  %loop-index2280 = phi i64 [ %1068, %loop-memcpy-expansion2279 ], [ 0, %loop-memcpy-expansion2279.preheader ]
  %1065 = getelementptr i8, ptr %4, i64 %loop-index2280, !amdgpu.uniform !42
  %1066 = load <4 x i32>, ptr %1065, align 1
  %1067 = getelementptr i8, ptr %1064, i64 %loop-index2280
  store <4 x i32> %1066, ptr %1067, align 1
  %1068 = add i64 %loop-index2280, 16
  %1069 = icmp uge i64 %1068, %612
  br i1 %1069, label %Flow3597, label %loop-memcpy-expansion2279, !amdgpu.uniform !42

Flow3596:                                         ; preds = %Flow3595, %loop-memcpy-residual-header2282
  br label %post-loop-memcpy-expansion2278, !amdgpu.uniform !42

loop-memcpy-residual2281:                         ; preds = %loop-memcpy-residual2281.preheader, %loop-memcpy-residual2281
  %residual-loop-index2283 = phi i64 [ %1071, %loop-memcpy-residual2281 ], [ 0, %loop-memcpy-residual2281.preheader ]
  %scevgep3045 = getelementptr i8, ptr %scevgep3044, i64 %residual-loop-index2283, !amdgpu.uniform !42
  %1070 = load i8, ptr %scevgep3045, align 1
  %scevgep3043 = getelementptr i8, ptr %scevgep3042, i64 %residual-loop-index2283
  store i8 %1070, ptr %scevgep3043, align 1
  %1071 = add i64 %residual-loop-index2283, 1
  %1072 = icmp uge i64 %1071, %611
  br i1 %1072, label %Flow3595, label %loop-memcpy-residual2281, !amdgpu.uniform !42

Flow3595:                                         ; preds = %loop-memcpy-residual2281
  br label %Flow3596, !amdgpu.uniform !42

post-loop-memcpy-expansion2278:                   ; preds = %Flow3596
  %sub.ptr.lhs.cast.i.i734 = ptrtoint ptr %add.ptr.i.i733 to i64
  %sub.ptr.sub.i.i735 = sub i64 %sub.ptr.lhs.cast.i.i734, ptrtoint (ptr addrspacecast (ptr addrspace(1) @_ZN12_GLOBAL__N_16memoryE to ptr) to i64)
  %cmp.i.i736 = icmp ugt i64 %sub.ptr.sub.i.i735, 65335
  %cond.i.i737 = select i1 %cmp.i.i736, ptr null, ptr %1064
  %cmp.not.i10.i.i64.i = icmp ne ptr %cond.i.i737, addrspacecast (ptr addrspace(4) @_ZN22__llvm_libc_22_0_0_git3cpp6string14NULL_CHARACTERE to ptr)
  br i1 %cmp.not.i10.i.i64.i, label %if.then.i.i.i65.i, label %for.cond.i.i.i739.preheader, !amdgpu.uniform !42

if.then.i.i.i65.i:                                ; preds = %post-loop-memcpy-expansion2278
  %arrayidx.i.i.i66.i = getelementptr inbounds nuw i8, ptr %1064, i64 %spec.select.i
  store i8 0, ptr %arrayidx.i.i.i66.i, align 1, !tbaa !57
  br label %for.cond.i.i.i739.preheader, !amdgpu.uniform !42

for.cond.i.i.i739.preheader:                      ; preds = %if.then.i.i.i65.i, %post-loop-memcpy-expansion2278
  br label %for.cond.i.i.i739, !amdgpu.uniform !42

for.cond.i.i.i739:                                ; preds = %for.cond.i.i.i739.preheader, %Flow3594
  %1073 = phi i32 [ %1111, %Flow3594 ], [ poison, %for.cond.i.i.i739.preheader ]
  %1074 = phi i64 [ %1112, %Flow3594 ], [ poison, %for.cond.i.i.i739.preheader ]
  %1075 = phi i32 [ %1113, %Flow3594 ], [ poison, %for.cond.i.i.i739.preheader ]
  %retval.sroa.6.0.i.i.i740 = phi i32 [ %1116, %Flow3594 ], [ undef, %for.cond.i.i.i739.preheader ]
  %retval.sroa.2.0.i.i.i742 = phi i64 [ %1115, %Flow3594 ], [ undef, %for.cond.i.i.i739.preheader ]
  %retval.sroa.8.0.i.i.i743 = phi i32 [ %1114, %Flow3594 ], [ undef, %for.cond.i.i.i739.preheader ]
  %index.0.i.i.i744 = phi i32 [ %1118, %Flow3594 ], [ 0, %for.cond.i.i.i739.preheader ]
  %1076 = load i32, ptr addrspace(1) @__llvm_rpc_client, align 8, !tbaa !68
  %cmp.not.i.i.i745 = icmp ult i32 %index.0.i.i.i744, %1076
  %spec.store.select.i.i.i746 = select i1 %cmp.not.i.i.i745, i32 %index.0.i.i.i744, i32 0
  %1077 = tail call noundef i64 @llvm.amdgcn.ballot.i64(i1 true)
  %1078 = and i32 %spec.store.select.i.i.i746, 31
  %1079 = lshr i64 %1077, %sh_prom.i.i.i.i
  %1080 = trunc i64 %1079 to i32
  %conv4.i.i.i.i.i750 = and i32 %1080, 1
  %shl.i.i.i.i.i751 = shl nuw i32 %conv4.i.i.i.i.i750, %1078
  %1081 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2515, !amdgpu.uniform !42

1082:                                             ; preds = %ComputeEnd2516
  %sunkaddr3261 = getelementptr inbounds i8, ptr addrspace(1) %3185, i64 40
  %1083 = atomicrmw or ptr addrspace(1) %sunkaddr3261, i32 %.lcssa3988 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %1084, !amdgpu.uniform !42

1084:                                             ; preds = %1082, %ComputeEnd2516
  %1085 = phi i32 [ %1083, %1082 ], [ poison, %ComputeEnd2516 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %3194)
  %1086 = call i32 @llvm.amdgcn.readfirstlane.i32(i32 %1085)
  %1087 = or i32 %1086, %.lcssa3989
  %shl5.i.i.i.i.i752 = shl nuw i32 1, %1078
  %and.i.i.i.i.i753 = and i32 %shl5.i.i.i.i.i752, %1087
  %tobool3.i.i.i.i754 = icmp ne i32 %and.i.i.i.i.i753, 0
  %1088 = tail call i64 @llvm.amdgcn.ballot.i64(i1 %tobool3.i.i.i.i754)
  %and.i.i.i.i.i.i755 = and i64 %1088, %1077
  %cmp.i.not.i.i.i756 = icmp ne i64 %1077, %and.i.i.i.i.i.i755
  br i1 %cmp.i.not.i.i.i756, label %if.then.i.i.i.i757, label %Flow3593, !amdgpu.uniform !42

if.then.i.i.i.i757:                               ; preds = %1084
  fence syncscope("agent") acquire
  %1089 = load <2 x i64>, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !75
  %1090 = extractelement <2 x i64> %1089, i32 0
  %1091 = inttoptr i64 %1090 to ptr
  %1092 = extractelement <2 x i64> %1089, i32 1
  %1093 = inttoptr i64 %1092 to ptr
  %idxprom.i.i.i.i758 = zext i32 %spec.store.select.i.i.i746 to i64
  %arrayidx.i.i.i.i759 = getelementptr inbounds nuw i32, ptr %1091, i64 %idxprom.i.i.i.i758, !amdgpu.uniform !42
  %1094 = load atomic i32, ptr %arrayidx.i.i.i.i759 monotonic, align 4
  %1095 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %1094)
  %arrayidx.i45.i.i.i760 = getelementptr inbounds nuw i32, ptr %1093, i64 %idxprom.i.i.i.i758, !amdgpu.uniform !42
  %1096 = load atomic i32, ptr %arrayidx.i45.i.i.i760 monotonic, align 4
  %1097 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %1096)
  %cmp.i46.not.i.i.i761 = icmp ne i32 %1095, %1097
  br i1 %cmp.i46.not.i.i.i761, label %if.then12.i.i.i762, label %Flow3591, !amdgpu.uniform !42

if.then12.i.i.i762:                               ; preds = %if.then.i.i.i.i757
  fence syncscope("agent") release
  %1098 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %1077, i1 true)
  %iszero.i.i.i.i.i.i.i763 = icmp ne i64 %1077, 0
  %cmp2.i.i.i.i.i.i764 = icmp eq i64 %1098, %sh_prom.i.i.i.i
  %cmp.i.i.i.i.i.i765 = select i1 %iszero.i.i.i.i.i.i.i763, i1 %cmp2.i.i.i.i.i.i764, i1 false
  %conv4.i.i51.i.i.i766 = zext i1 %cmp.i.i.i.i.i.i765 to i32
  %shl.i.i52.i.i.i767 = shl nuw i32 %conv4.i.i51.i.i.i766, %1078
  %xor.i.i.i.i.i768 = xor i32 %shl.i.i52.i.i.i767, -1
  %1099 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2520, !amdgpu.uniform !42

1100:                                             ; preds = %ComputeEnd2521
  %sunkaddr3262 = getelementptr inbounds i8, ptr addrspace(1) %3185, i64 40
  %1101 = atomicrmw and ptr addrspace(1) %sunkaddr3262, i32 %.lcssa3990 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %1102, !amdgpu.uniform !42

1102:                                             ; preds = %1100, %ComputeEnd2521
  call void @llvm.amdgcn.end.cf.i64(i64 %3212)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %Flow3591, !amdgpu.uniform !42

if.end14.i.i.i828:                                ; preds = %Flow3591
  %1103 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %1077, i1 true)
  %iszero.i.i.i.i.i.i829 = icmp ne i64 %1077, 0
  %cmp2.i.i.i.i.i830 = icmp eq i64 %1103, %sh_prom.i.i.i.i
  %cmp.i.i.i.i.i831 = select i1 %iszero.i.i.i.i.i.i829, i1 %cmp2.i.i.i.i.i830, i1 false
  %1104 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.i.i.i.i.i831)
  %1105 = extractvalue { i1, i64 } %1104, 0
  %1106 = extractvalue { i1, i64 } %1104, 1
  br i1 %1105, label %if.then16.i.i.i833, label %if.end22.i.i.i832

if.then16.i.i.i833:                               ; preds = %if.end14.i.i.i828
  %1107 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !76
  %opcode.i.i.i834 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %1107, i64 %idxprom.i.i.i.i758, i32 1
  store i32 1660944387, ptr %opcode.i.i.i834, align 8, !tbaa !77
  %arrayidx21.i.i.i835 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %1107, i64 %idxprom.i.i.i.i758
  store i64 %1077, ptr %arrayidx21.i.i.i835, align 8, !tbaa !80
  br label %if.end22.i.i.i832, !amdgpu.uniform !42

Flow3592:                                         ; preds = %if.end22.i.i.i832, %Flow3591
  %1108 = phi i32 [ %1095, %if.end22.i.i.i832 ], [ %retval.sroa.8.0.i.i.i743, %Flow3591 ]
  %1109 = phi i64 [ %1077, %if.end22.i.i.i832 ], [ %retval.sroa.2.0.i.i.i742, %Flow3591 ]
  %1110 = phi i32 [ %spec.store.select.i.i.i746, %if.end22.i.i.i832 ], [ %retval.sroa.6.0.i.i.i740, %Flow3591 ]
  br label %cleanup26.i.i.i769, !amdgpu.uniform !42

if.end22.i.i.i832:                                ; preds = %if.then16.i.i.i833, %if.end14.i.i.i828
  call void @llvm.amdgcn.end.cf.i64(i64 %1106)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %Flow3592, !amdgpu.uniform !42

Flow3593:                                         ; preds = %cleanup26.i.i.i769, %1084
  %1111 = phi i32 [ %1110, %cleanup26.i.i.i769 ], [ %1073, %1084 ]
  %1112 = phi i64 [ %1109, %cleanup26.i.i.i769 ], [ %1074, %1084 ]
  %1113 = phi i32 [ %1108, %cleanup26.i.i.i769 ], [ %1075, %1084 ]
  %1114 = phi i32 [ %1108, %cleanup26.i.i.i769 ], [ %retval.sroa.8.0.i.i.i743, %1084 ]
  %1115 = phi i64 [ %1109, %cleanup26.i.i.i769 ], [ %retval.sroa.2.0.i.i.i742, %1084 ]
  %1116 = phi i32 [ %1110, %cleanup26.i.i.i769 ], [ %retval.sroa.6.0.i.i.i740, %1084 ]
  %1117 = phi i1 [ %cmp.i46.not.i.i.i761, %cleanup26.i.i.i769 ], [ true, %1084 ]
  br i1 %1117, label %for.inc.i.i.i774, label %Flow3594, !amdgpu.uniform !42

cleanup26.i.i.i769:                               ; preds = %Flow3592
  br label %Flow3593, !amdgpu.uniform !42

for.inc.i.i.i774:                                 ; preds = %Flow3593
  %inc.i.i.i779 = add i32 %spec.store.select.i.i.i746, 1
  br label %Flow3594, !amdgpu.uniform !42

Flow3594:                                         ; preds = %for.inc.i.i.i774, %Flow3593
  %1118 = phi i32 [ %inc.i.i.i779, %for.inc.i.i.i774 ], [ poison, %Flow3593 ]
  %1119 = phi i1 [ false, %for.inc.i.i.i774 ], [ true, %Flow3593 ]
  br i1 %1119, label %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i780, label %for.cond.i.i.i739, !amdgpu.uniform !42

_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i780: ; preds = %Flow3594
  %.lcssa3993 = phi i32 [ %1111, %Flow3594 ]
  %.lcssa3992 = phi i64 [ %1112, %Flow3594 ]
  %.lcssa3991 = phi i32 [ %1113, %Flow3594 ]
  fence acquire
  %1120 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %mul.i.i.i.i.i.i781 = shl i32 %.lcssa3993, 6
  %idxprom.i19.i.i.i.i.i782 = zext i32 %mul.i.i.i.i.i.i781 to i64
  %arrayidx.i20.i.i.i.i.i783 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %1120, i64 %idxprom.i19.i.i.i.i.i782
  %arrayidx.i22.i.i.i.i.i784 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %arrayidx.i20.i.i.i.i.i783, i64 %sh_prom.i.i.i.i
  store i64 %spec.select.i, ptr %arrayidx.i22.i.i.i.i.i784, align 8, !tbaa !82
  br i1 %672, label %loop-memcpy-expansion2231.preheader, label %Flow3590, !amdgpu.uniform !42

loop-memcpy-expansion2231.preheader:              ; preds = %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i780
  %1121 = add nuw nsw i64 %76, 8
  %scevgep3039 = getelementptr i8, ptr %1120, i64 %1121
  %1122 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i782, 6
  %scevgep3040 = getelementptr i8, ptr %scevgep3039, i64 %1122
  br label %loop-memcpy-expansion2231, !amdgpu.uniform !42

Flow3590:                                         ; preds = %Flow3589, %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i780
  br label %loop-memcpy-residual-header2234, !amdgpu.uniform !42

loop-memcpy-expansion2231:                        ; preds = %loop-memcpy-expansion2231.preheader, %loop-memcpy-expansion2231
  %loop-index2232 = phi i64 [ %1125, %loop-memcpy-expansion2231 ], [ 0, %loop-memcpy-expansion2231.preheader ]
  %1123 = getelementptr i8, ptr %1064, i64 %loop-index2232, !amdgpu.uniform !42
  %1124 = load <4 x i32>, ptr %1123, align 1
  %scevgep3041 = getelementptr i8, ptr %scevgep3040, i64 %loop-index2232
  store <4 x i32> %1124, ptr %scevgep3041, align 1
  %1125 = add i64 %loop-index2232, 16
  %1126 = icmp uge i64 %1125, %671
  br i1 %1126, label %Flow3589, label %loop-memcpy-expansion2231, !amdgpu.uniform !42

Flow3588:                                         ; preds = %Flow3586, %Flow3587
  %1127 = phi i64 [ %2598, %Flow3586 ], [ %2594, %Flow3587 ]
  br label %post-loop-memcpy-expansion2230, !amdgpu.uniform !42

loop-memcpy-residual2233:                         ; preds = %loop-memcpy-residual2233.preheader, %loop-memcpy-residual2233
  %residual-loop-index2235 = phi i64 [ %1129, %loop-memcpy-residual2233 ], [ 0, %loop-memcpy-residual2233.preheader ]
  %scevgep3038 = getelementptr i8, ptr %scevgep3037, i64 %residual-loop-index2235, !amdgpu.uniform !42
  %1128 = load i8, ptr %scevgep3038, align 1
  %scevgep3036 = getelementptr i8, ptr %scevgep3035, i64 %residual-loop-index2235
  store i8 %1128, ptr %scevgep3036, align 1
  %1129 = add i64 %residual-loop-index2235, 1
  %1130 = icmp uge i64 %1129, %670
  br i1 %1130, label %Flow3586, label %loop-memcpy-residual2233, !amdgpu.uniform !42

Flow3586:                                         ; preds = %loop-memcpy-residual2233
  br label %Flow3588, !amdgpu.uniform !42

post-loop-memcpy-expansion2230:                   ; preds = %Flow3588
  %tobool.not.i.i.i.i.i.i787 = icmp eq i32 %.lcssa3991, 0
  %conv.i.i.i.i.i.i788 = zext i1 %tobool.not.i.i.i.i.i.i787 to i32
  fence release
  %1131 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %idxprom.i23.i.i.i.i.i789 = zext i32 %.lcssa3993 to i64
  %arrayidx.i24.i.i.i.i.i790 = getelementptr inbounds nuw i32, ptr %1131, i64 %idxprom.i23.i.i.i.i.i789
  store atomic i32 %conv.i.i.i.i.i.i788, ptr %arrayidx.i24.i.i.i.i.i790 monotonic, align 4
  %1132 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !84
  %arrayidx.i.i5.i.i791 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %1132, i64 %idxprom.i23.i.i.i.i.i789, !amdgpu.uniform !42
  %1133 = load i64, ptr %arrayidx.i.i5.i.i791, align 8, !tbaa !80
  %scevgep3020 = getelementptr i8, ptr %1064, i64 56
  br label %while.cond.i.i.i.i792, !amdgpu.uniform !42

while.cond.i.i.i.i792:                            ; preds = %post-loop-memcpy-expansion2230, %Flow3585
  %phi.broken3805 = phi i64 [ 0, %post-loop-memcpy-expansion2230 ], [ %1150, %Flow3585 ]
  %lsr.iv3027 = phi i64 [ %1148, %Flow3585 ], [ %686, %post-loop-memcpy-expansion2230 ]
  %lsr.iv3021 = phi ptr [ %1147, %Flow3585 ], [ %scevgep3020, %post-loop-memcpy-expansion2230 ]
  %port.sroa.37.0.i.i793 = phi i32 [ %1146, %Flow3585 ], [ %conv.i.i.i.i.i.i788, %post-loop-memcpy-expansion2230 ]
  %idx.0.i.i.i.i794 = phi i64 [ %1145, %Flow3585 ], [ 56, %post-loop-memcpy-expansion2230 ]
  %cmp.i.i.i.i795 = icmp ugt i64 %spec.select.i, %idx.0.i.i.i.i794
  %1134 = tail call i64 @llvm.amdgcn.ballot.i64(i1 %cmp.i.i.i.i795)
  %and.i.i.i.i6.i.i796 = and i64 %1134, %1133
  %tobool.not.i.i.i.i797 = icmp ne i64 %and.i.i.i.i6.i.i796, 0
  %1135 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i22.i.i.i798 = getelementptr inbounds nuw i32, ptr %1135, i64 %idxprom.i23.i.i.i.i.i789, !amdgpu.uniform !42
  %1136 = load atomic i32, ptr %arrayidx.i22.i.i.i798 monotonic, align 4
  %1137 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %tobool.not.i.i.i.i797)
  %1138 = extractvalue { i1, i64 } %1137, 0
  %1139 = extractvalue { i1, i64 } %1137, 1
  br i1 %1138, label %cond.false.i8.i.i.i.i799, label %Flow3585

cond.false.i8.i.i.i.i799:                         ; preds = %while.cond.i.i.i.i792
  %1140 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %1136)
  %1141 = icmp ne i32 %1140, %port.sroa.37.0.i.i793
  br i1 %1141, label %while.body.i.i21.i.i.i.i800.preheader, label %Flow3584, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i800.preheader:            ; preds = %cond.false.i8.i.i.i.i799
  br label %while.body.i.i21.i.i.i.i800, !amdgpu.uniform !42

Flow3584:                                         ; preds = %Flow3583, %cond.false.i8.i.i.i.i799
  br label %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i803, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i800:                      ; preds = %while.body.i.i21.i.i.i.i800.preheader, %while.body.i.i21.i.i.i.i800
  tail call void @llvm.amdgcn.s.sleep(i32 2)
  %1142 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i22.i.i.i.i801 = getelementptr inbounds nuw i32, ptr %1142, i64 %idxprom.i23.i.i.i.i.i789, !amdgpu.uniform !42
  %1143 = load atomic i32, ptr %arrayidx.i.i.i22.i.i.i.i801 monotonic, align 4
  %1144 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %1143)
  %cmp.i.not.i.i23.i.i.i.i802 = icmp eq i32 %1144, %port.sroa.37.0.i.i793
  br i1 %cmp.i.not.i.i23.i.i.i.i802, label %Flow3583, label %while.body.i.i21.i.i.i.i800, !amdgpu.uniform !42

Flow3583:                                         ; preds = %while.body.i.i21.i.i.i.i800
  br label %Flow3584, !amdgpu.uniform !42

Flow3585:                                         ; preds = %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i804, %while.cond.i.i.i.i792
  %1145 = phi i64 [ %add.i.i.i.i807, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i804 ], [ poison, %while.cond.i.i.i.i792 ]
  %1146 = phi i32 [ %conv.i.i26.i.i.i.i805, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i804 ], [ poison, %while.cond.i.i.i.i792 ]
  %1147 = phi ptr [ %scevgep3022, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i804 ], [ poison, %while.cond.i.i.i.i792 ]
  %1148 = phi i64 [ %lsr.iv.next3028, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i804 ], [ poison, %while.cond.i.i.i.i792 ]
  %1149 = phi i1 [ false, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i804 ], [ true, %while.cond.i.i.i.i792 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %1139)
  %1150 = call i64 @llvm.amdgcn.if.break.i64(i1 %1149, i64 %phi.broken3805)
  %1151 = call i1 @llvm.amdgcn.loop.i64(i64 %1150)
  br i1 %1151, label %cond.false.i.i.i814, label %while.cond.i.i.i.i792

_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i803: ; preds = %Flow3584
  fence acquire
  br i1 %cmp.i.i.i.i795, label %if.then.i.i.i.i.i.i.i808, label %Flow3582, !amdgpu.uniform !42

if.then.i.i.i.i.i.i.i808:                         ; preds = %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i803
  %sub.i.i.i.i.i.i.i810 = sub nuw i64 %spec.select.i, %idx.0.i.i.i.i794
  %spec.select.i.i.i36.i.i.i.i811 = tail call i64 @llvm.umin.i64(i64 %sub.i.i.i.i.i.i.i810, i64 64)
  %1152 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %1153 = and i64 %spec.select.i.i.i36.i.i.i.i811, 15
  %1154 = sub i64 %spec.select.i.i.i36.i.i.i.i811, %1153
  %1155 = icmp ne i64 %1154, 0
  br i1 %1155, label %loop-memcpy-expansion2225.preheader, label %Flow3581, !amdgpu.uniform !42

loop-memcpy-expansion2225.preheader:              ; preds = %if.then.i.i.i.i.i.i.i808
  %arrayidx.i20.i35.i.i.i.i812 = getelementptr %"struct.rpc::Buffer", ptr %1152, i64 %idxprom.i19.i.i.i.i.i782
  %arrayidx.i22.i38.i.i.i.i813 = getelementptr %"struct.rpc::Buffer", ptr %arrayidx.i20.i35.i.i.i.i812, i64 %sh_prom.i.i.i.i
  br label %loop-memcpy-expansion2225, !amdgpu.uniform !42

Flow3581:                                         ; preds = %Flow3580, %if.then.i.i.i.i.i.i.i808
  br label %loop-memcpy-residual-header2228, !amdgpu.uniform !42

loop-memcpy-expansion2225:                        ; preds = %loop-memcpy-expansion2225.preheader, %loop-memcpy-expansion2225
  %loop-index2226 = phi i64 [ %1158, %loop-memcpy-expansion2225 ], [ 0, %loop-memcpy-expansion2225.preheader ]
  %scevgep3023 = getelementptr i8, ptr %lsr.iv3021, i64 %loop-index2226, !amdgpu.uniform !42
  %1156 = load <4 x i32>, ptr %scevgep3023, align 1
  %1157 = getelementptr i8, ptr %arrayidx.i22.i38.i.i.i.i813, i64 %loop-index2226
  store <4 x i32> %1156, ptr %1157, align 1
  %1158 = add i64 %loop-index2226, 16
  %1159 = icmp uge i64 %1158, %1154
  br i1 %1159, label %Flow3580, label %loop-memcpy-expansion2225, !amdgpu.uniform !42

Flow3579:                                         ; preds = %Flow3578, %loop-memcpy-residual-header2228
  br label %Flow3582, !amdgpu.uniform !42

loop-memcpy-residual2227:                         ; preds = %loop-memcpy-residual2227.preheader, %loop-memcpy-residual2227
  %residual-loop-index2229 = phi i64 [ %1161, %loop-memcpy-residual2227 ], [ 0, %loop-memcpy-residual2227.preheader ]
  %scevgep3033 = getelementptr i8, ptr %scevgep3032, i64 %residual-loop-index2229, !amdgpu.uniform !42
  %1160 = load i8, ptr %scevgep3033, align 1
  %scevgep3031 = getelementptr i8, ptr %scevgep3030, i64 %residual-loop-index2229
  store i8 %1160, ptr %scevgep3031, align 1
  %1161 = add i64 %residual-loop-index2229, 1
  %1162 = icmp uge i64 %1161, %1153
  br i1 %1162, label %Flow3578, label %loop-memcpy-residual2227, !amdgpu.uniform !42

Flow3578:                                         ; preds = %loop-memcpy-residual2227
  br label %Flow3579, !amdgpu.uniform !42

_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i804: ; preds = %Flow3582
  %conv.i.i26.i.i.i.i805 = xor i32 %port.sroa.37.0.i.i793, 1
  fence release
  %1163 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %arrayidx.i24.i29.i.i.i.i806 = getelementptr inbounds nuw i32, ptr %1163, i64 %idxprom.i23.i.i.i.i.i789
  store atomic i32 %conv.i.i26.i.i.i.i805, ptr %arrayidx.i24.i29.i.i.i.i806 monotonic, align 4
  %add.i.i.i.i807 = add i64 %idx.0.i.i.i.i794, 64
  %scevgep3022 = getelementptr i8, ptr %lsr.iv3021, i64 64
  %lsr.iv.next3028 = add i64 %lsr.iv3027, -64
  br label %Flow3585, !amdgpu.uniform !42

cond.false.i.i.i814:                              ; preds = %Flow3585
  %.lcssa3987 = phi i64 [ %1150, %Flow3585 ]
  %port.sroa.37.0.i.i793.lcssa = phi i32 [ %port.sroa.37.0.i.i793, %Flow3585 ]
  %.lcssa3986 = phi i32 [ %1136, %Flow3585 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa3987)
  %1164 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %.lcssa3986)
  %1165 = icmp ne i32 %1164, %port.sroa.37.0.i.i793.lcssa
  %1166 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %1165)
  %1167 = extractvalue { i1, i64 } %1166, 0
  %1168 = extractvalue { i1, i64 } %1166, 1
  br i1 %1167, label %while.body.i.i.i.i815.preheader, label %Flow3577

while.body.i.i.i.i815.preheader:                  ; preds = %cond.false.i.i.i814
  br label %while.body.i.i.i.i815, !amdgpu.uniform !42

Flow3577:                                         ; preds = %Flow3576, %cond.false.i.i.i814
  call void @llvm.amdgcn.end.cf.i64(i64 %1168)
  br label %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsINS_3cpp6stringEEERS1_T_.exit836, !amdgpu.uniform !42

while.body.i.i.i.i815:                            ; preds = %while.body.i.i.i.i815.preheader, %while.body.i.i.i.i815
  %phi.broken3806 = phi i64 [ 0, %while.body.i.i.i.i815.preheader ], [ %1172, %while.body.i.i.i.i815 ]
  tail call void @llvm.amdgcn.s.sleep(i32 2)
  %1169 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i8.i.i816 = getelementptr inbounds nuw i32, ptr %1169, i64 %idxprom.i23.i.i.i.i.i789, !amdgpu.uniform !42
  %1170 = load atomic i32, ptr %arrayidx.i.i.i8.i.i816 monotonic, align 4
  %1171 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %1170)
  %cmp.i.not.i.i.i.i817 = icmp eq i32 %1171, %port.sroa.37.0.i.i793.lcssa
  %1172 = call i64 @llvm.amdgcn.if.break.i64(i1 %cmp.i.not.i.i.i.i817, i64 %phi.broken3806)
  %1173 = call i1 @llvm.amdgcn.loop.i64(i64 %1172)
  br i1 %1173, label %Flow3576, label %while.body.i.i.i.i815

Flow3576:                                         ; preds = %while.body.i.i.i.i815
  %.lcssa3985 = phi i64 [ %1172, %while.body.i.i.i.i815 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa3985)
  br label %Flow3577, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsINS_3cpp6stringEEERS1_T_.exit836: ; preds = %Flow3577
  fence acquire
  tail call void @llvm.amdgcn.wave.barrier()
  fence syncscope("agent") release
  %1174 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %.lcssa3992, i1 true)
  %iszero.i.i.i.i.i13.i.i819 = icmp ne i64 %.lcssa3992, 0
  %cmp2.i.i.i.i14.i.i820 = icmp eq i64 %1174, %sh_prom.i.i.i.i
  %cmp.i.i.i.i15.i.i821 = select i1 %iszero.i.i.i.i.i13.i.i819, i1 %cmp2.i.i.i.i14.i.i820, i1 false
  %1175 = and i32 %.lcssa3993, 31
  %conv4.i.i.i19.i.i825 = zext i1 %cmp.i.i.i.i15.i.i821 to i32
  %shl.i.i.i20.i.i826 = shl nuw i32 %conv4.i.i.i19.i.i825, %1175
  %xor.i.i.i21.i.i827 = xor i32 %shl.i.i.i20.i.i826, -1
  %1176 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2524, !amdgpu.uniform !42

1177:                                             ; preds = %ComputeEnd2525
  %div8.i.i.i16.i.i822 = lshr i32 %.lcssa3993, 5
  %idxprom.i.i.i17.i.i823 = zext nneg i32 %div8.i.i.i16.i.i822 to i64
  %1178 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i17.i.i823
  %sunkaddr3263 = getelementptr inbounds i8, ptr addrspace(1) %1178, i64 40
  %1179 = atomicrmw and ptr addrspace(1) %sunkaddr3263, i32 %.lcssa3984 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %1180, !amdgpu.uniform !42

1180:                                             ; preds = %1177, %ComputeEnd2525
  call void @llvm.amdgcn.end.cf.i64(i64 %3229)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %for.cond.i.i1.i845, !amdgpu.uniform !42

for.cond.i.i1.i845:                               ; preds = %1180, %Flow3575
  %1181 = phi i32 [ %1219, %Flow3575 ], [ poison, %1180 ]
  %1182 = phi i64 [ %1220, %Flow3575 ], [ poison, %1180 ]
  %1183 = phi i32 [ %1221, %Flow3575 ], [ poison, %1180 ]
  %retval.sroa.6.0.i.i.i846 = phi i32 [ %1224, %Flow3575 ], [ undef, %1180 ]
  %retval.sroa.2.0.i.i.i848 = phi i64 [ %1223, %Flow3575 ], [ undef, %1180 ]
  %retval.sroa.8.0.i.i.i849 = phi i32 [ %1222, %Flow3575 ], [ undef, %1180 ]
  %index.0.i.i.i850 = phi i32 [ %1226, %Flow3575 ], [ 0, %1180 ]
  %1184 = load i32, ptr addrspace(1) @__llvm_rpc_client, align 8, !tbaa !68
  %cmp.not.i.i.i851 = icmp ult i32 %index.0.i.i.i850, %1184
  %spec.store.select.i.i.i852 = select i1 %cmp.not.i.i.i851, i32 %index.0.i.i.i850, i32 0
  %1185 = tail call noundef i64 @llvm.amdgcn.ballot.i64(i1 true)
  %1186 = and i32 %spec.store.select.i.i.i852, 31
  %1187 = lshr i64 %1185, %sh_prom.i.i.i.i
  %1188 = trunc i64 %1187 to i32
  %conv4.i.i.i.i.i856 = and i32 %1188, 1
  %shl.i.i.i.i.i857 = shl nuw i32 %conv4.i.i.i.i.i856, %1186
  %1189 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2528, !amdgpu.uniform !42

1190:                                             ; preds = %ComputeEnd2529
  %sunkaddr3264 = getelementptr inbounds i8, ptr addrspace(1) %3239, i64 40
  %1191 = atomicrmw or ptr addrspace(1) %sunkaddr3264, i32 %.lcssa3978 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %1192, !amdgpu.uniform !42

1192:                                             ; preds = %1190, %ComputeEnd2529
  %1193 = phi i32 [ %1191, %1190 ], [ poison, %ComputeEnd2529 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %3248)
  %1194 = call i32 @llvm.amdgcn.readfirstlane.i32(i32 %1193)
  %1195 = or i32 %1194, %.lcssa3979
  %shl5.i.i.i.i.i858 = shl nuw i32 1, %1186
  %and.i.i.i.i.i859 = and i32 %shl5.i.i.i.i.i858, %1195
  %tobool3.i.i.i.i860 = icmp ne i32 %and.i.i.i.i.i859, 0
  %1196 = tail call i64 @llvm.amdgcn.ballot.i64(i1 %tobool3.i.i.i.i860)
  %and.i.i.i.i.i.i861 = and i64 %1196, %1185
  %cmp.i.not.i.i.i862 = icmp ne i64 %1185, %and.i.i.i.i.i.i861
  br i1 %cmp.i.not.i.i.i862, label %if.then.i.i.i.i863, label %Flow3574, !amdgpu.uniform !42

if.then.i.i.i.i863:                               ; preds = %1192
  fence syncscope("agent") acquire
  %1197 = load <2 x i64>, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !75
  %1198 = extractelement <2 x i64> %1197, i32 0
  %1199 = inttoptr i64 %1198 to ptr
  %1200 = extractelement <2 x i64> %1197, i32 1
  %1201 = inttoptr i64 %1200 to ptr
  %idxprom.i.i.i.i864 = zext i32 %spec.store.select.i.i.i852 to i64
  %arrayidx.i.i.i.i865 = getelementptr inbounds nuw i32, ptr %1199, i64 %idxprom.i.i.i.i864, !amdgpu.uniform !42
  %1202 = load atomic i32, ptr %arrayidx.i.i.i.i865 monotonic, align 4
  %1203 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %1202)
  %arrayidx.i45.i.i.i866 = getelementptr inbounds nuw i32, ptr %1201, i64 %idxprom.i.i.i.i864, !amdgpu.uniform !42
  %1204 = load atomic i32, ptr %arrayidx.i45.i.i.i866 monotonic, align 4
  %1205 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %1204)
  %cmp.i46.not.i.i.i867 = icmp ne i32 %1203, %1205
  br i1 %cmp.i46.not.i.i.i867, label %if.then12.i.i.i868, label %Flow3572, !amdgpu.uniform !42

if.then12.i.i.i868:                               ; preds = %if.then.i.i.i.i863
  fence syncscope("agent") release
  %1206 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %1185, i1 true)
  %iszero.i.i.i.i.i.i.i869 = icmp ne i64 %1185, 0
  %cmp2.i.i.i.i.i.i870 = icmp eq i64 %1206, %sh_prom.i.i.i.i
  %cmp.i.i.i.i.i.i871 = select i1 %iszero.i.i.i.i.i.i.i869, i1 %cmp2.i.i.i.i.i.i870, i1 false
  %conv4.i.i51.i.i.i872 = zext i1 %cmp.i.i.i.i.i.i871 to i32
  %shl.i.i52.i.i.i873 = shl nuw i32 %conv4.i.i51.i.i.i872, %1186
  %xor.i.i.i.i.i874 = xor i32 %shl.i.i52.i.i.i873, -1
  %1207 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2533, !amdgpu.uniform !42

1208:                                             ; preds = %ComputeEnd2534
  %sunkaddr3265 = getelementptr inbounds i8, ptr addrspace(1) %3239, i64 40
  %1209 = atomicrmw and ptr addrspace(1) %sunkaddr3265, i32 %.lcssa3980 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %1210, !amdgpu.uniform !42

1210:                                             ; preds = %1208, %ComputeEnd2534
  call void @llvm.amdgcn.end.cf.i64(i64 %3266)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %Flow3572, !amdgpu.uniform !42

if.end14.i.i.i934:                                ; preds = %Flow3572
  %1211 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %1185, i1 true)
  %iszero.i.i.i.i.i.i935 = icmp ne i64 %1185, 0
  %cmp2.i.i.i.i.i936 = icmp eq i64 %1211, %sh_prom.i.i.i.i
  %cmp.i.i.i.i.i937 = select i1 %iszero.i.i.i.i.i.i935, i1 %cmp2.i.i.i.i.i936, i1 false
  %1212 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.i.i.i.i.i937)
  %1213 = extractvalue { i1, i64 } %1212, 0
  %1214 = extractvalue { i1, i64 } %1212, 1
  br i1 %1213, label %if.then16.i.i.i939, label %if.end22.i.i.i938

if.then16.i.i.i939:                               ; preds = %if.end14.i.i.i934
  %1215 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !76
  %opcode.i.i.i940 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %1215, i64 %idxprom.i.i.i.i864, i32 1
  store i32 1660944387, ptr %opcode.i.i.i940, align 8, !tbaa !77
  %arrayidx21.i.i.i941 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %1215, i64 %idxprom.i.i.i.i864
  store i64 %1185, ptr %arrayidx21.i.i.i941, align 8, !tbaa !80
  br label %if.end22.i.i.i938, !amdgpu.uniform !42

Flow3573:                                         ; preds = %if.end22.i.i.i938, %Flow3572
  %1216 = phi i32 [ %1203, %if.end22.i.i.i938 ], [ %retval.sroa.8.0.i.i.i849, %Flow3572 ]
  %1217 = phi i64 [ %1185, %if.end22.i.i.i938 ], [ %retval.sroa.2.0.i.i.i848, %Flow3572 ]
  %1218 = phi i32 [ %spec.store.select.i.i.i852, %if.end22.i.i.i938 ], [ %retval.sroa.6.0.i.i.i846, %Flow3572 ]
  br label %cleanup26.i.i.i875, !amdgpu.uniform !42

if.end22.i.i.i938:                                ; preds = %if.then16.i.i.i939, %if.end14.i.i.i934
  call void @llvm.amdgcn.end.cf.i64(i64 %1214)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %Flow3573, !amdgpu.uniform !42

Flow3574:                                         ; preds = %cleanup26.i.i.i875, %1192
  %1219 = phi i32 [ %1218, %cleanup26.i.i.i875 ], [ %1181, %1192 ]
  %1220 = phi i64 [ %1217, %cleanup26.i.i.i875 ], [ %1182, %1192 ]
  %1221 = phi i32 [ %1216, %cleanup26.i.i.i875 ], [ %1183, %1192 ]
  %1222 = phi i32 [ %1216, %cleanup26.i.i.i875 ], [ %retval.sroa.8.0.i.i.i849, %1192 ]
  %1223 = phi i64 [ %1217, %cleanup26.i.i.i875 ], [ %retval.sroa.2.0.i.i.i848, %1192 ]
  %1224 = phi i32 [ %1218, %cleanup26.i.i.i875 ], [ %retval.sroa.6.0.i.i.i846, %1192 ]
  %1225 = phi i1 [ %cmp.i46.not.i.i.i867, %cleanup26.i.i.i875 ], [ true, %1192 ]
  br i1 %1225, label %for.inc.i.i.i880, label %Flow3575, !amdgpu.uniform !42

cleanup26.i.i.i875:                               ; preds = %Flow3573
  br label %Flow3574, !amdgpu.uniform !42

for.inc.i.i.i880:                                 ; preds = %Flow3574
  %inc.i.i.i885 = add i32 %spec.store.select.i.i.i852, 1
  br label %Flow3575, !amdgpu.uniform !42

Flow3575:                                         ; preds = %for.inc.i.i.i880, %Flow3574
  %1226 = phi i32 [ %inc.i.i.i885, %for.inc.i.i.i880 ], [ poison, %Flow3574 ]
  %1227 = phi i1 [ false, %for.inc.i.i.i880 ], [ true, %Flow3574 ]
  br i1 %1227, label %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i886, label %for.cond.i.i1.i845, !amdgpu.uniform !42

_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i886: ; preds = %Flow3575
  %.lcssa3983 = phi i32 [ %1219, %Flow3575 ]
  %.lcssa3982 = phi i64 [ %1220, %Flow3575 ]
  %.lcssa3981 = phi i32 [ %1221, %Flow3575 ]
  fence acquire
  %1228 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %mul.i.i.i.i.i.i887 = shl i32 %.lcssa3983, 6
  %idxprom.i19.i.i.i.i.i888 = zext i32 %mul.i.i.i.i.i.i887 to i64
  %arrayidx.i20.i.i.i.i.i889 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %1228, i64 %idxprom.i19.i.i.i.i.i888
  %arrayidx.i22.i.i.i.i.i890 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %arrayidx.i20.i.i.i.i.i889, i64 %sh_prom.i.i.i.i
  store i64 sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.35 to ptr), i64 10) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.35 to ptr) to i64)), ptr %arrayidx.i22.i.i.i.i.i890, align 8, !tbaa !82
  %spec.select.i.i.i.i.i.i.i891 = tail call i64 @llvm.umin.i64(i64 sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.35 to ptr), i64 10) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.35 to ptr) to i64)), i64 56)
  %1229 = and i64 %spec.select.i.i.i.i.i.i.i891, 15
  %1230 = sub i64 %spec.select.i.i.i.i.i.i.i891, %1229
  %1231 = icmp ne i64 %1230, 0
  br i1 %1231, label %loop-memcpy-expansion2357.preheader, label %Flow3571, !amdgpu.uniform !42

loop-memcpy-expansion2357.preheader:              ; preds = %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i886
  %1232 = add nuw nsw i64 %76, 8
  %scevgep3017 = getelementptr i8, ptr %1228, i64 %1232
  %1233 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i888, 6
  %scevgep3018 = getelementptr i8, ptr %scevgep3017, i64 %1233
  br label %loop-memcpy-expansion2357, !amdgpu.uniform !42

Flow3571:                                         ; preds = %Flow3570, %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i886
  br label %loop-memcpy-residual-header2360, !amdgpu.uniform !42

loop-memcpy-expansion2357:                        ; preds = %loop-memcpy-expansion2357.preheader, %loop-memcpy-expansion2357
  %loop-index2358 = phi i64 [ %1236, %loop-memcpy-expansion2357 ], [ 0, %loop-memcpy-expansion2357.preheader ]
  %1234 = getelementptr i8, ptr addrspace(4) @.str.35, i64 %loop-index2358, !amdgpu.uniform !42
  %1235 = load <4 x i32>, ptr addrspace(4) %1234, align 1
  %scevgep3019 = getelementptr i8, ptr %scevgep3018, i64 %loop-index2358
  store <4 x i32> %1235, ptr %scevgep3019, align 1
  %1236 = add i64 %loop-index2358, 16
  %1237 = icmp uge i64 %1236, %1230
  br i1 %1237, label %Flow3570, label %loop-memcpy-expansion2357, !amdgpu.uniform !42

Flow3569:                                         ; preds = %Flow3567, %Flow3568
  %1238 = phi i64 [ %2706, %Flow3567 ], [ %2703, %Flow3568 ]
  br label %post-loop-memcpy-expansion2356, !amdgpu.uniform !42

loop-memcpy-residual2359:                         ; preds = %loop-memcpy-residual2359.preheader, %loop-memcpy-residual2359
  %residual-loop-index2361 = phi i64 [ %1240, %loop-memcpy-residual2359 ], [ 0, %loop-memcpy-residual2359.preheader ]
  %scevgep3016 = getelementptr i8, ptr addrspace(4) @.str.35, i64 %residual-loop-index2361, !amdgpu.uniform !42
  %1239 = load i8, ptr addrspace(4) %scevgep3016, align 1
  %scevgep3015 = getelementptr i8, ptr %scevgep3014, i64 %residual-loop-index2361
  store i8 %1239, ptr %scevgep3015, align 1
  %1240 = add i64 %residual-loop-index2361, 1
  %1241 = icmp uge i64 %1240, %1229
  br i1 %1241, label %Flow3567, label %loop-memcpy-residual2359, !amdgpu.uniform !42

Flow3567:                                         ; preds = %loop-memcpy-residual2359
  br label %Flow3569, !amdgpu.uniform !42

post-loop-memcpy-expansion2356:                   ; preds = %Flow3569
  %tobool.not.i.i.i.i.i.i893 = icmp eq i32 %.lcssa3981, 0
  %conv.i.i.i.i.i.i894 = zext i1 %tobool.not.i.i.i.i.i.i893 to i32
  fence release
  %1242 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %idxprom.i23.i.i.i.i.i895 = zext i32 %.lcssa3983 to i64
  %arrayidx.i24.i.i.i.i.i896 = getelementptr inbounds nuw i32, ptr %1242, i64 %idxprom.i23.i.i.i.i.i895
  store atomic i32 %conv.i.i.i.i.i.i894, ptr %arrayidx.i24.i.i.i.i.i896 monotonic, align 4
  %1243 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !84
  %arrayidx.i.i5.i.i897 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %1243, i64 %idxprom.i23.i.i.i.i.i895, !amdgpu.uniform !42
  %1244 = load i64, ptr %arrayidx.i.i5.i.i897, align 8, !tbaa !80
  br label %while.cond.i.i.i.i898, !amdgpu.uniform !42

while.cond.i.i.i.i898:                            ; preds = %post-loop-memcpy-expansion2356, %Flow3565
  %phi.broken3807 = phi i64 [ 0, %post-loop-memcpy-expansion2356 ], [ %1261, %Flow3565 ]
  %lsr.iv3006 = phi i64 [ %1259, %Flow3565 ], [ -46, %post-loop-memcpy-expansion2356 ]
  %lsr.iv3001 = phi ptr addrspace(4) [ %1258, %Flow3565 ], [ getelementptr (i8, ptr addrspace(4) @.str.35, i64 56), %post-loop-memcpy-expansion2356 ]
  %port.sroa.37.0.i.i899 = phi i32 [ %1257, %Flow3565 ], [ %conv.i.i.i.i.i.i894, %post-loop-memcpy-expansion2356 ]
  %idx.0.i.i.i.i900 = phi i64 [ %1256, %Flow3565 ], [ 56, %post-loop-memcpy-expansion2356 ]
  %cmp.i.i.i.i901 = icmp ult i64 %idx.0.i.i.i.i900, sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.35 to ptr), i64 10) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.35 to ptr) to i64))
  %1245 = tail call i64 @llvm.amdgcn.ballot.i64(i1 %cmp.i.i.i.i901)
  %and.i.i.i.i6.i.i902 = and i64 %1245, %1244
  %tobool.not.i.i.i.i903 = icmp ne i64 %and.i.i.i.i6.i.i902, 0
  %1246 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i22.i.i.i904 = getelementptr inbounds nuw i32, ptr %1246, i64 %idxprom.i23.i.i.i.i.i895, !amdgpu.uniform !42
  %1247 = load atomic i32, ptr %arrayidx.i22.i.i.i904 monotonic, align 4
  %1248 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %tobool.not.i.i.i.i903)
  %1249 = extractvalue { i1, i64 } %1248, 0
  %1250 = extractvalue { i1, i64 } %1248, 1
  br i1 %1249, label %cond.false.i8.i.i.i.i905, label %Flow3565

cond.false.i8.i.i.i.i905:                         ; preds = %while.cond.i.i.i.i898
  %1251 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %1247)
  %1252 = icmp ne i32 %1251, %port.sroa.37.0.i.i899
  br i1 %1252, label %while.body.i.i21.i.i.i.i906.preheader, label %Flow3564, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i906.preheader:            ; preds = %cond.false.i8.i.i.i.i905
  br label %while.body.i.i21.i.i.i.i906, !amdgpu.uniform !42

Flow3564:                                         ; preds = %Flow3562, %cond.false.i8.i.i.i.i905
  br label %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i909, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i906:                      ; preds = %while.body.i.i21.i.i.i.i906.preheader, %while.body.i.i21.i.i.i.i906
  tail call void @llvm.amdgcn.s.sleep(i32 2)
  %1253 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i22.i.i.i.i907 = getelementptr inbounds nuw i32, ptr %1253, i64 %idxprom.i23.i.i.i.i.i895, !amdgpu.uniform !42
  %1254 = load atomic i32, ptr %arrayidx.i.i.i22.i.i.i.i907 monotonic, align 4
  %1255 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %1254)
  %cmp.i.not.i.i23.i.i.i.i908 = icmp eq i32 %1255, %port.sroa.37.0.i.i899
  br i1 %cmp.i.not.i.i23.i.i.i.i908, label %Flow3562, label %while.body.i.i21.i.i.i.i906, !amdgpu.uniform !42

Flow3562:                                         ; preds = %while.body.i.i21.i.i.i.i906
  br label %Flow3564, !amdgpu.uniform !42

Flow3565:                                         ; preds = %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i910, %while.cond.i.i.i.i898
  %1256 = phi i64 [ %add.i.i.i.i913, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i910 ], [ poison, %while.cond.i.i.i.i898 ]
  %1257 = phi i32 [ %conv.i.i26.i.i.i.i911, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i910 ], [ poison, %while.cond.i.i.i.i898 ]
  %1258 = phi ptr addrspace(4) [ %scevgep3002, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i910 ], [ poison, %while.cond.i.i.i.i898 ]
  %1259 = phi i64 [ %lsr.iv.next3007, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i910 ], [ poison, %while.cond.i.i.i.i898 ]
  %1260 = phi i1 [ false, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i910 ], [ true, %while.cond.i.i.i.i898 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %1250)
  %1261 = call i64 @llvm.amdgcn.if.break.i64(i1 %1260, i64 %phi.broken3807)
  %1262 = call i1 @llvm.amdgcn.loop.i64(i64 %1261)
  br i1 %1262, label %cond.false.i.i.i920, label %while.cond.i.i.i.i898

_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i909: ; preds = %Flow3564
  fence acquire
  br i1 %cmp.i.i.i.i901, label %if.then.i.i.i.i.i.i.i914, label %Flow3561, !amdgpu.uniform !42

if.then.i.i.i.i.i.i.i914:                         ; preds = %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i909
  %sub.i.i.i.i.i.i.i916 = sub nuw i64 sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.35 to ptr), i64 10) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.35 to ptr) to i64)), %idx.0.i.i.i.i900
  %spec.select.i.i.i36.i.i.i.i917 = tail call i64 @llvm.umin.i64(i64 %sub.i.i.i.i.i.i.i916, i64 64)
  %1263 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %1264 = and i64 %spec.select.i.i.i36.i.i.i.i917, 15
  %1265 = sub i64 %spec.select.i.i.i36.i.i.i.i917, %1264
  %1266 = icmp ne i64 %1265, 0
  br i1 %1266, label %loop-memcpy-expansion2369.preheader, label %Flow3560, !amdgpu.uniform !42

loop-memcpy-expansion2369.preheader:              ; preds = %if.then.i.i.i.i.i.i.i914
  %arrayidx.i20.i35.i.i.i.i918 = getelementptr %"struct.rpc::Buffer", ptr %1263, i64 %idxprom.i19.i.i.i.i.i888
  %arrayidx.i22.i38.i.i.i.i919 = getelementptr %"struct.rpc::Buffer", ptr %arrayidx.i20.i35.i.i.i.i918, i64 %sh_prom.i.i.i.i
  br label %loop-memcpy-expansion2369, !amdgpu.uniform !42

Flow3560:                                         ; preds = %Flow3559, %if.then.i.i.i.i.i.i.i914
  br label %loop-memcpy-residual-header2372, !amdgpu.uniform !42

loop-memcpy-expansion2369:                        ; preds = %loop-memcpy-expansion2369.preheader, %loop-memcpy-expansion2369
  %loop-index2370 = phi i64 [ %1269, %loop-memcpy-expansion2369 ], [ 0, %loop-memcpy-expansion2369.preheader ]
  %scevgep3003 = getelementptr i8, ptr addrspace(4) %lsr.iv3001, i64 %loop-index2370, !amdgpu.uniform !42
  %1267 = load <4 x i32>, ptr addrspace(4) %scevgep3003, align 1
  %1268 = getelementptr i8, ptr %arrayidx.i22.i38.i.i.i.i919, i64 %loop-index2370
  store <4 x i32> %1267, ptr %1268, align 1
  %1269 = add i64 %loop-index2370, 16
  %1270 = icmp uge i64 %1269, %1265
  br i1 %1270, label %Flow3559, label %loop-memcpy-expansion2369, !amdgpu.uniform !42

Flow3557:                                         ; preds = %Flow3556, %loop-memcpy-residual-header2372
  br label %Flow3561, !amdgpu.uniform !42

loop-memcpy-residual2371:                         ; preds = %loop-memcpy-residual2371.preheader, %loop-memcpy-residual2371
  %residual-loop-index2373 = phi i64 [ %1272, %loop-memcpy-residual2371 ], [ 0, %loop-memcpy-residual2371.preheader ]
  %scevgep3012 = getelementptr i8, ptr addrspace(4) %scevgep3011, i64 %residual-loop-index2373, !amdgpu.uniform !42
  %1271 = load i8, ptr addrspace(4) %scevgep3012, align 1
  %scevgep3010 = getelementptr i8, ptr %scevgep3009, i64 %residual-loop-index2373
  store i8 %1271, ptr %scevgep3010, align 1
  %1272 = add i64 %residual-loop-index2373, 1
  %1273 = icmp uge i64 %1272, %1264
  br i1 %1273, label %Flow3556, label %loop-memcpy-residual2371, !amdgpu.uniform !42

Flow3556:                                         ; preds = %loop-memcpy-residual2371
  br label %Flow3557, !amdgpu.uniform !42

_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i910: ; preds = %Flow3561
  %conv.i.i26.i.i.i.i911 = xor i32 %port.sroa.37.0.i.i899, 1
  fence release
  %1274 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %arrayidx.i24.i29.i.i.i.i912 = getelementptr inbounds nuw i32, ptr %1274, i64 %idxprom.i23.i.i.i.i.i895
  store atomic i32 %conv.i.i26.i.i.i.i911, ptr %arrayidx.i24.i29.i.i.i.i912 monotonic, align 4
  %add.i.i.i.i913 = add i64 %idx.0.i.i.i.i900, 64
  %scevgep3002 = getelementptr i8, ptr addrspace(4) %lsr.iv3001, i64 64
  %lsr.iv.next3007 = add i64 %lsr.iv3006, -64
  br label %Flow3565, !amdgpu.uniform !42

cond.false.i.i.i920:                              ; preds = %Flow3565
  %.lcssa3977 = phi i64 [ %1261, %Flow3565 ]
  %port.sroa.37.0.i.i899.lcssa = phi i32 [ %port.sroa.37.0.i.i899, %Flow3565 ]
  %.lcssa3976 = phi i32 [ %1247, %Flow3565 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa3977)
  %1275 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %.lcssa3976)
  %1276 = icmp ne i32 %1275, %port.sroa.37.0.i.i899.lcssa
  %1277 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %1276)
  %1278 = extractvalue { i1, i64 } %1277, 0
  %1279 = extractvalue { i1, i64 } %1277, 1
  br i1 %1278, label %while.body.i.i.i.i921.preheader, label %Flow3554

while.body.i.i.i.i921.preheader:                  ; preds = %cond.false.i.i.i920
  br label %while.body.i.i.i.i921, !amdgpu.uniform !42

Flow3554:                                         ; preds = %Flow3552, %cond.false.i.i.i920
  call void @llvm.amdgcn.end.cf.i64(i64 %1279)
  br label %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit942, !amdgpu.uniform !42

while.body.i.i.i.i921:                            ; preds = %while.body.i.i.i.i921.preheader, %while.body.i.i.i.i921
  %phi.broken3808 = phi i64 [ 0, %while.body.i.i.i.i921.preheader ], [ %1283, %while.body.i.i.i.i921 ]
  tail call void @llvm.amdgcn.s.sleep(i32 2)
  %1280 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i8.i.i922 = getelementptr inbounds nuw i32, ptr %1280, i64 %idxprom.i23.i.i.i.i.i895, !amdgpu.uniform !42
  %1281 = load atomic i32, ptr %arrayidx.i.i.i8.i.i922 monotonic, align 4
  %1282 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %1281)
  %cmp.i.not.i.i.i.i923 = icmp eq i32 %1282, %port.sroa.37.0.i.i899.lcssa
  %1283 = call i64 @llvm.amdgcn.if.break.i64(i1 %cmp.i.not.i.i.i.i923, i64 %phi.broken3808)
  %1284 = call i1 @llvm.amdgcn.loop.i64(i64 %1283)
  br i1 %1284, label %Flow3552, label %while.body.i.i.i.i921

Flow3552:                                         ; preds = %while.body.i.i.i.i921
  %.lcssa3975 = phi i64 [ %1283, %while.body.i.i.i.i921 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa3975)
  br label %Flow3554, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit942: ; preds = %Flow3554
  fence acquire
  tail call void @llvm.amdgcn.wave.barrier()
  fence syncscope("agent") release
  %1285 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %.lcssa3982, i1 true)
  %iszero.i.i.i.i.i13.i.i925 = icmp ne i64 %.lcssa3982, 0
  %cmp2.i.i.i.i14.i.i926 = icmp eq i64 %1285, %sh_prom.i.i.i.i
  %cmp.i.i.i.i15.i.i927 = select i1 %iszero.i.i.i.i.i13.i.i925, i1 %cmp2.i.i.i.i14.i.i926, i1 false
  %1286 = and i32 %.lcssa3983, 31
  %conv4.i.i.i19.i.i931 = zext i1 %cmp.i.i.i.i15.i.i927 to i32
  %shl.i.i.i20.i.i932 = shl nuw i32 %conv4.i.i.i19.i.i931, %1286
  %xor.i.i.i21.i.i933 = xor i32 %shl.i.i.i20.i.i932, -1
  %1287 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2537, !amdgpu.uniform !42

1288:                                             ; preds = %ComputeEnd2538
  %div8.i.i.i16.i.i928 = lshr i32 %.lcssa3983, 5
  %idxprom.i.i.i17.i.i929 = zext nneg i32 %div8.i.i.i16.i.i928 to i64
  %1289 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i17.i.i929
  %sunkaddr3266 = getelementptr inbounds i8, ptr addrspace(1) %1289, i64 40
  %1290 = atomicrmw and ptr addrspace(1) %sunkaddr3266, i32 %.lcssa3974 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %1291, !amdgpu.uniform !42

1291:                                             ; preds = %1288, %ComputeEnd2538
  call void @llvm.amdgcn.end.cf.i64(i64 %3283)
  tail call void @llvm.amdgcn.wave.barrier()
  call void @llvm.lifetime.start.p5(ptr addrspace(5) %buffer.i.i.i) #34, !noalias !133
  %1292 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %LHS)
  %1293 = extractvalue { i1, i64 } %1292, 0
  %1294 = extractvalue { i1, i64 } %1292, 1
  br i1 %1293, label %if.then3.i.i.i.i.i.i.i.i, label %_ZN22__llvm_libc_22_0_0_git15IntegerToStringIiNS_7details3FmtILh10ELb0ELb0ELb0ELm1EEEE13IntegerWriter25write_unsigned_number_decEjRNS1_22StringBufferWriterImplILb0EEE.exit.i.i.i.i.i.i

if.then3.i.i.i.i.i.i.i.i:                         ; preds = %1291
  %arrayidx.i.i.i.i.i.i.i.i.i = getelementptr inbounds nuw i8, ptr addrspace(5) %buffer.i.i.i, i32 10
  store i8 49, ptr addrspace(5) %arrayidx.i.i.i.i.i.i.i.i.i, align 2, !tbaa !57, !noalias !133
  br label %_ZN22__llvm_libc_22_0_0_git15IntegerToStringIiNS_7details3FmtILh10ELb0ELb0ELb0ELm1EEEE13IntegerWriter25write_unsigned_number_decEjRNS1_22StringBufferWriterImplILb0EEE.exit.i.i.i.i.i.i, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git15IntegerToStringIiNS_7details3FmtILh10ELb0ELb0ELb0ELm1EEEE13IntegerWriter25write_unsigned_number_decEjRNS1_22StringBufferWriterImplILb0EEE.exit.i.i.i.i.i.i: ; preds = %if.then3.i.i.i.i.i.i.i.i, %1291
  %writer.sroa.10.2.i.i.i.i.i = phi i64 [ 1, %if.then3.i.i.i.i.i.i.i.i ], [ 0, %1291 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %1294)
  %1295 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %LHS.inv)
  %1296 = extractvalue { i1, i64 } %1295, 0
  %1297 = extractvalue { i1, i64 } %1295, 1
  br i1 %1296, label %if.then3.i.i.lr.ph.i.i.i.i.i, label %while.end.i.i.i.i.i.i

if.then3.i.i.lr.ph.i.i.i.i.i:                     ; preds = %_ZN22__llvm_libc_22_0_0_git15IntegerToStringIiNS_7details3FmtILh10ELb0ELb0ELb0ELm1EEEE13IntegerWriter25write_unsigned_number_decEjRNS1_22StringBufferWriterImplILb0EEE.exit.i.i.i.i.i.i
  %arrayidx.i.i.i.i.i.i.i.i = getelementptr inbounds nuw i8, ptr addrspace(5) %buffer.i.i.i, i32 10
  store i8 48, ptr addrspace(5) %arrayidx.i.i.i.i.i.i.i.i, align 2, !tbaa !57, !noalias !133
  br label %while.end.i.i.i.i.i.i, !amdgpu.uniform !42

while.end.i.i.i.i.i.i:                            ; preds = %if.then3.i.i.lr.ph.i.i.i.i.i, %_ZN22__llvm_libc_22_0_0_git15IntegerToStringIiNS_7details3FmtILh10ELb0ELb0ELb0ELm1EEEE13IntegerWriter25write_unsigned_number_decEjRNS1_22StringBufferWriterImplILb0EEE.exit.i.i.i.i.i.i
  %writer.sroa.10.4.i.i.i.i.i = phi i64 [ 1, %if.then3.i.i.lr.ph.i.i.i.i.i ], [ %writer.sroa.10.2.i.i.i.i.i, %_ZN22__llvm_libc_22_0_0_git15IntegerToStringIiNS_7details3FmtILh10ELb0ELb0ELb0ELm1EEEE13IntegerWriter25write_unsigned_number_decEjRNS1_22StringBufferWriterImplILb0EEE.exit.i.i.i.i.i.i ]
  call void @llvm.amdgcn.end.cf.i64(i64 %1297)
  %written.i.i.i.i.i = getelementptr inbounds nuw i8, ptr addrspace(5) %buffer.i.i.i, i32 16
  store i64 %writer.sroa.10.4.i.i.i.i.i, ptr addrspace(5) %written.i.i.i.i.i, align 8, !tbaa !91, !noalias !133
  %cmp.i.not.i.i.i.i.i.i = icmp ne i64 %writer.sroa.10.4.i.i.i.i.i, 0
  %1298 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.i.not.i.i.i.i.i.i)
  %1299 = extractvalue { i1, i64 } %1298, 0
  %1300 = extractvalue { i1, i64 } %1298, 1
  br i1 %1299, label %if.end4.i.i.i.i.i.i.i.i, label %if.end.i.i.i.i.i.i.i

if.end4.i.i.i.i.i.i.i.i:                          ; preds = %while.end.i.i.i.i.i.i
  %1301 = load ptr, ptr addrspace(1) @_ZN12_GLOBAL__N_13ptrE, align 8, !tbaa !55
  %add.ptr.i.i943 = getelementptr inbounds nuw i8, ptr %1301, i64 8
  store ptr %add.ptr.i.i943, ptr addrspace(1) @_ZN12_GLOBAL__N_13ptrE, align 8, !tbaa !55
  %sub.ptr.lhs.cast.i.i944 = ptrtoint ptr %add.ptr.i.i943 to i64
  %sub.ptr.sub.i.i945 = sub i64 %sub.ptr.lhs.cast.i.i944, ptrtoint (ptr addrspacecast (ptr addrspace(1) @_ZN12_GLOBAL__N_16memoryE to ptr) to i64)
  %cmp.i.i946 = icmp ugt i64 %sub.ptr.sub.i.i945, 65335
  %cond.i.i947 = select i1 %cmp.i.i946, ptr null, ptr %1301
  store i8 0, ptr %1301, align 1, !noalias !140
  br label %if.end.i.i.i.i.i.i.i, !amdgpu.uniform !42

if.end.i.i.i.i.i.i.i:                             ; preds = %if.end4.i.i.i.i.i.i.i.i, %while.end.i.i.i.i.i.i
  %agg.tmp13.i.sroa.0.0 = phi ptr [ %cond.i.i947, %if.end4.i.i.i.i.i.i.i.i ], [ addrspacecast (ptr addrspace(4) @_ZN22__llvm_libc_22_0_0_git3cpp6string14NULL_CHARACTERE to ptr), %while.end.i.i.i.i.i.i ]
  call void @llvm.amdgcn.end.cf.i64(i64 %1300)
  %cmp.not.i6.i.i.i.i.i.i.i = icmp ne ptr %agg.tmp13.i.sroa.0.0, addrspacecast (ptr addrspace(4) @_ZN22__llvm_libc_22_0_0_git3cpp6string14NULL_CHARACTERE to ptr)
  %1302 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.not.i6.i.i.i.i.i.i.i)
  %1303 = extractvalue { i1, i64 } %1302, 0
  %1304 = extractvalue { i1, i64 } %1302, 1
  br i1 %1303, label %if.then.i.i.i.i2.i.i.i.i, label %_ZN22__llvm_libc_22_0_0_git7testing8internal13describeValueIbEENS_3cpp9enable_ifIXaasr3cppE13is_integral_vIT_ElestS5_Lm8EENS3_6stringEE4typeES5_.exit.i

if.then.i.i.i.i2.i.i.i.i:                         ; preds = %if.end.i.i.i.i.i.i.i
  %arrayidx.i.i.i.i3.i.i.i.i = getelementptr inbounds nuw i8, ptr %agg.tmp13.i.sroa.0.0, i64 %writer.sroa.10.4.i.i.i.i.i
  store i8 0, ptr %arrayidx.i.i.i.i3.i.i.i.i, align 1, !tbaa !57, !noalias !133
  br label %_ZN22__llvm_libc_22_0_0_git7testing8internal13describeValueIbEENS_3cpp9enable_ifIXaasr3cppE13is_integral_vIT_ElestS5_Lm8EENS3_6stringEE4typeES5_.exit.i, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git7testing8internal13describeValueIbEENS_3cpp9enable_ifIXaasr3cppE13is_integral_vIT_ElestS5_Lm8EENS3_6stringEE4typeES5_.exit.i: ; preds = %if.then.i.i.i.i2.i.i.i.i, %if.end.i.i.i.i.i.i.i
  call void @llvm.amdgcn.end.cf.i64(i64 %1304)
  %1305 = trunc i64 %writer.sroa.10.4.i.i.i.i.i to i32
  %1306 = and i64 %writer.sroa.10.4.i.i.i.i.i, 15
  %1307 = sub i64 %writer.sroa.10.4.i.i.i.i.i, %1306
  %1308 = icmp ne i64 %1307, 0
  %1309 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %1308)
  %1310 = extractvalue { i1, i64 } %1309, 0
  %1311 = extractvalue { i1, i64 } %1309, 1
  br i1 %1310, label %loop-memcpy-expansion2303.preheader, label %Flow3551

loop-memcpy-expansion2303.preheader:              ; preds = %_ZN22__llvm_libc_22_0_0_git7testing8internal13describeValueIbEENS_3cpp9enable_ifIXaasr3cppE13is_integral_vIT_ElestS5_Lm8EENS3_6stringEE4typeES5_.exit.i
  %1312 = sub i32 11, %1305
  %scevgep2998 = getelementptr i8, ptr addrspace(5) %buffer.i.i.i, i32 %1312
  br label %loop-memcpy-expansion2303, !amdgpu.uniform !42

Flow3551:                                         ; preds = %Flow3550, %_ZN22__llvm_libc_22_0_0_git7testing8internal13describeValueIbEENS_3cpp9enable_ifIXaasr3cppE13is_integral_vIT_ElestS5_Lm8EENS3_6stringEE4typeES5_.exit.i
  call void @llvm.amdgcn.end.cf.i64(i64 %1311)
  br label %loop-memcpy-residual-header2306, !amdgpu.uniform !42

loop-memcpy-expansion2303:                        ; preds = %loop-memcpy-expansion2303.preheader, %loop-memcpy-expansion2303
  %phi.broken3809 = phi i64 [ 0, %loop-memcpy-expansion2303.preheader ], [ %1317, %loop-memcpy-expansion2303 ]
  %lsr.iv2999 = phi ptr addrspace(5) [ %scevgep3000, %loop-memcpy-expansion2303 ], [ %scevgep2998, %loop-memcpy-expansion2303.preheader ]
  %loop-index2304 = phi i64 [ %1315, %loop-memcpy-expansion2303 ], [ 0, %loop-memcpy-expansion2303.preheader ]
  %1313 = load <4 x i32>, ptr addrspace(5) %lsr.iv2999, align 1
  %1314 = getelementptr i8, ptr %agg.tmp13.i.sroa.0.0, i64 %loop-index2304
  store <4 x i32> %1313, ptr %1314, align 1
  %1315 = add i64 %loop-index2304, 16
  %scevgep3000 = getelementptr i8, ptr addrspace(5) %lsr.iv2999, i32 16
  %1316 = icmp uge i64 %1315, %1307
  %1317 = call i64 @llvm.amdgcn.if.break.i64(i1 %1316, i64 %phi.broken3809)
  %1318 = call i1 @llvm.amdgcn.loop.i64(i64 %1317)
  br i1 %1318, label %Flow3550, label %loop-memcpy-expansion2303

Flow3548:                                         ; preds = %Flow3547, %loop-memcpy-residual-header2306
  call void @llvm.amdgcn.end.cf.i64(i64 %2666)
  br label %post-loop-memcpy-expansion2302, !amdgpu.uniform !42

loop-memcpy-residual2305:                         ; preds = %loop-memcpy-residual2305.preheader, %loop-memcpy-residual2305
  %phi.broken3810 = phi i64 [ 0, %loop-memcpy-residual2305.preheader ], [ %1322, %loop-memcpy-residual2305 ]
  %lsr.iv2995 = phi ptr addrspace(5) [ %scevgep2996, %loop-memcpy-residual2305 ], [ %scevgep2994, %loop-memcpy-residual2305.preheader ]
  %residual-loop-index2307 = phi i64 [ %1320, %loop-memcpy-residual2305 ], [ 0, %loop-memcpy-residual2305.preheader ]
  %1319 = load i8, ptr addrspace(5) %lsr.iv2995, align 1
  %scevgep2997 = getelementptr i8, ptr %agg.tmp13.i.sroa.0.0, i64 %residual-loop-index2307
  store i8 %1319, ptr %scevgep2997, align 1
  %1320 = add i64 %residual-loop-index2307, 1
  %scevgep2996 = getelementptr i8, ptr addrspace(5) %lsr.iv2995, i32 1
  %1321 = icmp uge i64 %1320, %1306
  %1322 = call i64 @llvm.amdgcn.if.break.i64(i1 %1321, i64 %phi.broken3810)
  %1323 = call i1 @llvm.amdgcn.loop.i64(i64 %1322)
  br i1 %1323, label %Flow3547, label %loop-memcpy-residual2305

Flow3547:                                         ; preds = %loop-memcpy-residual2305
  %.lcssa3972 = phi i64 [ %1322, %loop-memcpy-residual2305 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa3972)
  br label %Flow3548, !amdgpu.uniform !42

post-loop-memcpy-expansion2302:                   ; preds = %Flow3548
  call void @llvm.lifetime.end.p5(ptr addrspace(5) %buffer.i.i.i) #34, !noalias !133
  br label %for.cond.i.i.i949, !amdgpu.uniform !42

for.cond.i.i.i949:                                ; preds = %post-loop-memcpy-expansion2302, %Flow3545
  %1324 = phi i32 [ %1362, %Flow3545 ], [ poison, %post-loop-memcpy-expansion2302 ]
  %1325 = phi i64 [ %1363, %Flow3545 ], [ poison, %post-loop-memcpy-expansion2302 ]
  %1326 = phi i32 [ %1364, %Flow3545 ], [ poison, %post-loop-memcpy-expansion2302 ]
  %retval.sroa.6.0.i.i.i950 = phi i32 [ %1367, %Flow3545 ], [ undef, %post-loop-memcpy-expansion2302 ]
  %retval.sroa.2.0.i.i.i952 = phi i64 [ %1366, %Flow3545 ], [ undef, %post-loop-memcpy-expansion2302 ]
  %retval.sroa.8.0.i.i.i953 = phi i32 [ %1365, %Flow3545 ], [ undef, %post-loop-memcpy-expansion2302 ]
  %index.0.i.i.i954 = phi i32 [ %1369, %Flow3545 ], [ 0, %post-loop-memcpy-expansion2302 ]
  %1327 = load i32, ptr addrspace(1) @__llvm_rpc_client, align 8, !tbaa !68
  %cmp.not.i.i.i955 = icmp ult i32 %index.0.i.i.i954, %1327
  %spec.store.select.i.i.i956 = select i1 %cmp.not.i.i.i955, i32 %index.0.i.i.i954, i32 0
  %1328 = tail call noundef i64 @llvm.amdgcn.ballot.i64(i1 true)
  %1329 = and i32 %spec.store.select.i.i.i956, 31
  %1330 = lshr i64 %1328, %sh_prom.i.i.i.i
  %1331 = trunc i64 %1330 to i32
  %conv4.i.i.i.i.i960 = and i32 %1331, 1
  %shl.i.i.i.i.i961 = shl nuw i32 %conv4.i.i.i.i.i960, %1329
  %1332 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2541, !amdgpu.uniform !42

1333:                                             ; preds = %ComputeEnd2542
  %sunkaddr3267 = getelementptr inbounds i8, ptr addrspace(1) %3293, i64 40
  %1334 = atomicrmw or ptr addrspace(1) %sunkaddr3267, i32 %.lcssa3966 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %1335, !amdgpu.uniform !42

1335:                                             ; preds = %1333, %ComputeEnd2542
  %1336 = phi i32 [ %1334, %1333 ], [ poison, %ComputeEnd2542 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %3302)
  %1337 = call i32 @llvm.amdgcn.readfirstlane.i32(i32 %1336)
  %1338 = or i32 %1337, %.lcssa3967
  %shl5.i.i.i.i.i962 = shl nuw i32 1, %1329
  %and.i.i.i.i.i963 = and i32 %shl5.i.i.i.i.i962, %1338
  %tobool3.i.i.i.i964 = icmp ne i32 %and.i.i.i.i.i963, 0
  %1339 = tail call i64 @llvm.amdgcn.ballot.i64(i1 %tobool3.i.i.i.i964)
  %and.i.i.i.i.i.i965 = and i64 %1339, %1328
  %cmp.i.not.i.i.i966 = icmp ne i64 %1328, %and.i.i.i.i.i.i965
  br i1 %cmp.i.not.i.i.i966, label %if.then.i.i.i.i967, label %Flow3544, !amdgpu.uniform !42

if.then.i.i.i.i967:                               ; preds = %1335
  fence syncscope("agent") acquire
  %1340 = load <2 x i64>, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !75
  %1341 = extractelement <2 x i64> %1340, i32 0
  %1342 = inttoptr i64 %1341 to ptr
  %1343 = extractelement <2 x i64> %1340, i32 1
  %1344 = inttoptr i64 %1343 to ptr
  %idxprom.i.i.i.i968 = zext i32 %spec.store.select.i.i.i956 to i64
  %arrayidx.i.i.i.i969 = getelementptr inbounds nuw i32, ptr %1342, i64 %idxprom.i.i.i.i968, !amdgpu.uniform !42
  %1345 = load atomic i32, ptr %arrayidx.i.i.i.i969 monotonic, align 4
  %1346 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %1345)
  %arrayidx.i45.i.i.i970 = getelementptr inbounds nuw i32, ptr %1344, i64 %idxprom.i.i.i.i968, !amdgpu.uniform !42
  %1347 = load atomic i32, ptr %arrayidx.i45.i.i.i970 monotonic, align 4
  %1348 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %1347)
  %cmp.i46.not.i.i.i971 = icmp ne i32 %1346, %1348
  br i1 %cmp.i46.not.i.i.i971, label %if.then12.i.i.i972, label %Flow3542, !amdgpu.uniform !42

if.then12.i.i.i972:                               ; preds = %if.then.i.i.i.i967
  fence syncscope("agent") release
  %1349 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %1328, i1 true)
  %iszero.i.i.i.i.i.i.i973 = icmp ne i64 %1328, 0
  %cmp2.i.i.i.i.i.i974 = icmp eq i64 %1349, %sh_prom.i.i.i.i
  %cmp.i.i.i.i.i.i975 = select i1 %iszero.i.i.i.i.i.i.i973, i1 %cmp2.i.i.i.i.i.i974, i1 false
  %conv4.i.i51.i.i.i976 = zext i1 %cmp.i.i.i.i.i.i975 to i32
  %shl.i.i52.i.i.i977 = shl nuw i32 %conv4.i.i51.i.i.i976, %1329
  %xor.i.i.i.i.i978 = xor i32 %shl.i.i52.i.i.i977, -1
  %1350 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2546, !amdgpu.uniform !42

1351:                                             ; preds = %ComputeEnd2547
  %sunkaddr3268 = getelementptr inbounds i8, ptr addrspace(1) %3293, i64 40
  %1352 = atomicrmw and ptr addrspace(1) %sunkaddr3268, i32 %.lcssa3968 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %1353, !amdgpu.uniform !42

1353:                                             ; preds = %1351, %ComputeEnd2547
  call void @llvm.amdgcn.end.cf.i64(i64 %3320)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %Flow3542, !amdgpu.uniform !42

if.end14.i.i.i1038:                               ; preds = %Flow3542
  %1354 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %1328, i1 true)
  %iszero.i.i.i.i.i.i1039 = icmp ne i64 %1328, 0
  %cmp2.i.i.i.i.i1040 = icmp eq i64 %1354, %sh_prom.i.i.i.i
  %cmp.i.i.i.i.i1041 = select i1 %iszero.i.i.i.i.i.i1039, i1 %cmp2.i.i.i.i.i1040, i1 false
  %1355 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.i.i.i.i.i1041)
  %1356 = extractvalue { i1, i64 } %1355, 0
  %1357 = extractvalue { i1, i64 } %1355, 1
  br i1 %1356, label %if.then16.i.i.i1043, label %if.end22.i.i.i1042

if.then16.i.i.i1043:                              ; preds = %if.end14.i.i.i1038
  %1358 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !76
  %opcode.i.i.i1044 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %1358, i64 %idxprom.i.i.i.i968, i32 1
  store i32 1660944387, ptr %opcode.i.i.i1044, align 8, !tbaa !77
  %arrayidx21.i.i.i1045 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %1358, i64 %idxprom.i.i.i.i968
  store i64 %1328, ptr %arrayidx21.i.i.i1045, align 8, !tbaa !80
  br label %if.end22.i.i.i1042, !amdgpu.uniform !42

Flow3543:                                         ; preds = %if.end22.i.i.i1042, %Flow3542
  %1359 = phi i32 [ %1346, %if.end22.i.i.i1042 ], [ %retval.sroa.8.0.i.i.i953, %Flow3542 ]
  %1360 = phi i64 [ %1328, %if.end22.i.i.i1042 ], [ %retval.sroa.2.0.i.i.i952, %Flow3542 ]
  %1361 = phi i32 [ %spec.store.select.i.i.i956, %if.end22.i.i.i1042 ], [ %retval.sroa.6.0.i.i.i950, %Flow3542 ]
  br label %cleanup26.i.i.i979, !amdgpu.uniform !42

if.end22.i.i.i1042:                               ; preds = %if.then16.i.i.i1043, %if.end14.i.i.i1038
  call void @llvm.amdgcn.end.cf.i64(i64 %1357)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %Flow3543, !amdgpu.uniform !42

Flow3544:                                         ; preds = %cleanup26.i.i.i979, %1335
  %1362 = phi i32 [ %1361, %cleanup26.i.i.i979 ], [ %1324, %1335 ]
  %1363 = phi i64 [ %1360, %cleanup26.i.i.i979 ], [ %1325, %1335 ]
  %1364 = phi i32 [ %1359, %cleanup26.i.i.i979 ], [ %1326, %1335 ]
  %1365 = phi i32 [ %1359, %cleanup26.i.i.i979 ], [ %retval.sroa.8.0.i.i.i953, %1335 ]
  %1366 = phi i64 [ %1360, %cleanup26.i.i.i979 ], [ %retval.sroa.2.0.i.i.i952, %1335 ]
  %1367 = phi i32 [ %1361, %cleanup26.i.i.i979 ], [ %retval.sroa.6.0.i.i.i950, %1335 ]
  %1368 = phi i1 [ %cmp.i46.not.i.i.i971, %cleanup26.i.i.i979 ], [ true, %1335 ]
  br i1 %1368, label %for.inc.i.i.i984, label %Flow3545, !amdgpu.uniform !42

cleanup26.i.i.i979:                               ; preds = %Flow3543
  br label %Flow3544, !amdgpu.uniform !42

for.inc.i.i.i984:                                 ; preds = %Flow3544
  %inc.i.i.i989 = add i32 %spec.store.select.i.i.i956, 1
  br label %Flow3545, !amdgpu.uniform !42

Flow3545:                                         ; preds = %for.inc.i.i.i984, %Flow3544
  %1369 = phi i32 [ %inc.i.i.i989, %for.inc.i.i.i984 ], [ poison, %Flow3544 ]
  %1370 = phi i1 [ false, %for.inc.i.i.i984 ], [ true, %Flow3544 ]
  br i1 %1370, label %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i990, label %for.cond.i.i.i949, !amdgpu.uniform !42

_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i990: ; preds = %Flow3545
  %.lcssa3971 = phi i32 [ %1362, %Flow3545 ]
  %.lcssa3970 = phi i64 [ %1363, %Flow3545 ]
  %.lcssa3969 = phi i32 [ %1364, %Flow3545 ]
  fence acquire
  %1371 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %mul.i.i.i.i.i.i991 = shl i32 %.lcssa3971, 6
  %idxprom.i19.i.i.i.i.i992 = zext i32 %mul.i.i.i.i.i.i991 to i64
  %arrayidx.i20.i.i.i.i.i993 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %1371, i64 %idxprom.i19.i.i.i.i.i992
  %arrayidx.i22.i.i.i.i.i994 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %arrayidx.i20.i.i.i.i.i993, i64 %sh_prom.i.i.i.i
  store i64 %writer.sroa.10.4.i.i.i.i.i, ptr %arrayidx.i22.i.i.i.i.i994, align 8, !tbaa !82
  %1372 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %1308)
  %1373 = extractvalue { i1, i64 } %1372, 0
  %1374 = extractvalue { i1, i64 } %1372, 1
  br i1 %1373, label %loop-memcpy-expansion2219.preheader, label %Flow3541

loop-memcpy-expansion2219.preheader:              ; preds = %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i990
  %1375 = add nuw nsw i64 %76, 8
  %scevgep2991 = getelementptr i8, ptr %1371, i64 %1375
  %1376 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i992, 6
  %scevgep2992 = getelementptr i8, ptr %scevgep2991, i64 %1376
  br label %loop-memcpy-expansion2219, !amdgpu.uniform !42

Flow3541:                                         ; preds = %Flow3540, %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i990
  call void @llvm.amdgcn.end.cf.i64(i64 %1374)
  br label %loop-memcpy-residual-header2222, !amdgpu.uniform !42

loop-memcpy-expansion2219:                        ; preds = %loop-memcpy-expansion2219.preheader, %loop-memcpy-expansion2219
  %phi.broken3811 = phi i64 [ 0, %loop-memcpy-expansion2219.preheader ], [ %1381, %loop-memcpy-expansion2219 ]
  %loop-index2220 = phi i64 [ %1379, %loop-memcpy-expansion2219 ], [ 0, %loop-memcpy-expansion2219.preheader ]
  %1377 = getelementptr i8, ptr %agg.tmp13.i.sroa.0.0, i64 %loop-index2220
  %1378 = load <4 x i32>, ptr %1377, align 1
  %scevgep2993 = getelementptr i8, ptr %scevgep2992, i64 %loop-index2220
  store <4 x i32> %1378, ptr %scevgep2993, align 1
  %1379 = add i64 %loop-index2220, 16
  %1380 = icmp uge i64 %1379, %1307
  %1381 = call i64 @llvm.amdgcn.if.break.i64(i1 %1380, i64 %phi.broken3811)
  %1382 = call i1 @llvm.amdgcn.loop.i64(i64 %1381)
  br i1 %1382, label %Flow3540, label %loop-memcpy-expansion2219

Flow3538:                                         ; preds = %Flow3537, %loop-memcpy-residual-header2222
  call void @llvm.amdgcn.end.cf.i64(i64 %2588)
  br label %post-loop-memcpy-expansion2218, !amdgpu.uniform !42

loop-memcpy-residual2221:                         ; preds = %loop-memcpy-residual2221.preheader, %loop-memcpy-residual2221
  %phi.broken3812 = phi i64 [ 0, %loop-memcpy-residual2221.preheader ], [ %1386, %loop-memcpy-residual2221 ]
  %residual-loop-index2223 = phi i64 [ %1384, %loop-memcpy-residual2221 ], [ 0, %loop-memcpy-residual2221.preheader ]
  %scevgep2990 = getelementptr i8, ptr %agg.tmp13.i.sroa.0.0, i64 %residual-loop-index2223
  %1383 = load i8, ptr %scevgep2990, align 1
  %scevgep2989 = getelementptr i8, ptr %scevgep2988, i64 %residual-loop-index2223
  store i8 %1383, ptr %scevgep2989, align 1
  %1384 = add i64 %residual-loop-index2223, 1
  %1385 = icmp uge i64 %1384, %1306
  %1386 = call i64 @llvm.amdgcn.if.break.i64(i1 %1385, i64 %phi.broken3812)
  %1387 = call i1 @llvm.amdgcn.loop.i64(i64 %1386)
  br i1 %1387, label %Flow3537, label %loop-memcpy-residual2221

Flow3537:                                         ; preds = %loop-memcpy-residual2221
  %.lcssa3964 = phi i64 [ %1386, %loop-memcpy-residual2221 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa3964)
  br label %Flow3538, !amdgpu.uniform !42

post-loop-memcpy-expansion2218:                   ; preds = %Flow3538
  %tobool.not.i.i.i.i.i.i997 = icmp eq i32 %.lcssa3969, 0
  %conv.i.i.i.i.i.i998 = zext i1 %tobool.not.i.i.i.i.i.i997 to i32
  fence release
  %1388 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %idxprom.i23.i.i.i.i.i999 = zext i32 %.lcssa3971 to i64
  %arrayidx.i24.i.i.i.i.i1000 = getelementptr inbounds nuw i32, ptr %1388, i64 %idxprom.i23.i.i.i.i.i999
  store atomic i32 %conv.i.i.i.i.i.i998, ptr %arrayidx.i24.i.i.i.i.i1000 monotonic, align 4
  %1389 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !84
  %arrayidx.i.i5.i.i1001 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %1389, i64 %idxprom.i23.i.i.i.i.i999, !amdgpu.uniform !42
  %1390 = load i64, ptr %arrayidx.i.i5.i.i1001, align 8, !tbaa !80
  br label %while.cond.i.i.i.i1002, !amdgpu.uniform !42

while.cond.i.i.i.i1002:                           ; preds = %post-loop-memcpy-expansion2218, %Flow3535
  %phi.broken3815 = phi i64 [ 0, %post-loop-memcpy-expansion2218 ], [ %1405, %Flow3535 ]
  %port.sroa.37.0.i.i1003 = phi i32 [ %1403, %Flow3535 ], [ %conv.i.i.i.i.i.i998, %post-loop-memcpy-expansion2218 ]
  %idx.0.i.i.i.i1004 = phi i64 [ %1402, %Flow3535 ], [ 56, %post-loop-memcpy-expansion2218 ]
  %cmp.i.i.i.i1005 = icmp ult i64 %idx.0.i.i.i.i1004, %writer.sroa.10.4.i.i.i.i.i
  %1391 = tail call i64 @llvm.amdgcn.ballot.i64(i1 %cmp.i.i.i.i1005)
  %and.i.i.i.i6.i.i1006 = and i64 %1391, %1390
  %tobool.not.i.i.i.i1007 = icmp ne i64 %and.i.i.i.i6.i.i1006, 0
  %1392 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i22.i.i.i1008 = getelementptr inbounds nuw i32, ptr %1392, i64 %idxprom.i23.i.i.i.i.i999, !amdgpu.uniform !42
  %1393 = load atomic i32, ptr %arrayidx.i22.i.i.i1008 monotonic, align 4
  %1394 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %tobool.not.i.i.i.i1007)
  %1395 = extractvalue { i1, i64 } %1394, 0
  %1396 = extractvalue { i1, i64 } %1394, 1
  br i1 %1395, label %cond.false.i8.i.i.i.i1009, label %Flow3535

cond.false.i8.i.i.i.i1009:                        ; preds = %while.cond.i.i.i.i1002
  %1397 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %1393)
  %1398 = icmp ne i32 %1397, %port.sroa.37.0.i.i1003
  br i1 %1398, label %while.body.i.i21.i.i.i.i1010.preheader, label %Flow3534, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i1010.preheader:           ; preds = %cond.false.i8.i.i.i.i1009
  br label %while.body.i.i21.i.i.i.i1010, !amdgpu.uniform !42

Flow3534:                                         ; preds = %Flow3532, %cond.false.i8.i.i.i.i1009
  br label %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i1013, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i1010:                     ; preds = %while.body.i.i21.i.i.i.i1010.preheader, %while.body.i.i21.i.i.i.i1010
  tail call void @llvm.amdgcn.s.sleep(i32 2)
  %1399 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i22.i.i.i.i1011 = getelementptr inbounds nuw i32, ptr %1399, i64 %idxprom.i23.i.i.i.i.i999, !amdgpu.uniform !42
  %1400 = load atomic i32, ptr %arrayidx.i.i.i22.i.i.i.i1011 monotonic, align 4
  %1401 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %1400)
  %cmp.i.not.i.i23.i.i.i.i1012 = icmp eq i32 %1401, %port.sroa.37.0.i.i1003
  br i1 %cmp.i.not.i.i23.i.i.i.i1012, label %Flow3532, label %while.body.i.i21.i.i.i.i1010, !amdgpu.uniform !42

Flow3532:                                         ; preds = %while.body.i.i21.i.i.i.i1010
  br label %Flow3534, !amdgpu.uniform !42

Flow3535:                                         ; preds = %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1014, %while.cond.i.i.i.i1002
  %1402 = phi i64 [ %add.i.i.i.i1017, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1014 ], [ poison, %while.cond.i.i.i.i1002 ]
  %1403 = phi i32 [ %conv.i.i26.i.i.i.i1015, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1014 ], [ poison, %while.cond.i.i.i.i1002 ]
  %1404 = phi i1 [ false, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1014 ], [ true, %while.cond.i.i.i.i1002 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %1396)
  %1405 = call i64 @llvm.amdgcn.if.break.i64(i1 %1404, i64 %phi.broken3815)
  %1406 = call i1 @llvm.amdgcn.loop.i64(i64 %1405)
  br i1 %1406, label %cond.false.i.i.i1024, label %while.cond.i.i.i.i1002

_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i1013: ; preds = %Flow3534
  fence acquire
  %1407 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.i.i.i.i1005)
  %1408 = extractvalue { i1, i64 } %1407, 0
  %1409 = extractvalue { i1, i64 } %1407, 1
  br i1 %1408, label %if.then.i.i.i.i.i.i.i1018, label %Flow3531

if.then.i.i.i.i.i.i.i1018:                        ; preds = %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i1013
  %1410 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %arrayidx.i20.i35.i.i.i.i1022 = getelementptr %"struct.rpc::Buffer", ptr %1410, i64 %idxprom.i19.i.i.i.i.i992
  %arrayidx.i22.i38.i.i.i.i1023 = getelementptr %"struct.rpc::Buffer", ptr %arrayidx.i20.i35.i.i.i.i1022, i64 %sh_prom.i.i.i.i
  %1411 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %1308)
  %1412 = extractvalue { i1, i64 } %1411, 0
  %1413 = extractvalue { i1, i64 } %1411, 1
  br i1 %1412, label %loop-memcpy-expansion2213.preheader, label %Flow3530

loop-memcpy-expansion2213.preheader:              ; preds = %if.then.i.i.i.i.i.i.i1018
  br label %loop-memcpy-expansion2213, !amdgpu.uniform !42

Flow3530:                                         ; preds = %Flow3529, %if.then.i.i.i.i.i.i.i1018
  call void @llvm.amdgcn.end.cf.i64(i64 %1413)
  br label %loop-memcpy-residual-header2216, !amdgpu.uniform !42

loop-memcpy-expansion2213:                        ; preds = %loop-memcpy-expansion2213.preheader, %loop-memcpy-expansion2213
  %phi.broken3813 = phi i64 [ 0, %loop-memcpy-expansion2213.preheader ], [ %1419, %loop-memcpy-expansion2213 ]
  %loop-index2214 = phi i64 [ %1417, %loop-memcpy-expansion2213 ], [ 0, %loop-memcpy-expansion2213.preheader ]
  %1414 = getelementptr i8, ptr %agg.tmp13.i.sroa.0.0, i64 %loop-index2214
  %1415 = load <4 x i32>, ptr %1414, align 1
  %1416 = getelementptr i8, ptr %arrayidx.i22.i38.i.i.i.i1023, i64 %loop-index2214
  store <4 x i32> %1415, ptr %1416, align 1
  %1417 = add i64 %loop-index2214, 16
  %1418 = icmp uge i64 %1417, %1307
  %1419 = call i64 @llvm.amdgcn.if.break.i64(i1 %1418, i64 %phi.broken3813)
  %1420 = call i1 @llvm.amdgcn.loop.i64(i64 %1419)
  br i1 %1420, label %Flow3529, label %loop-memcpy-expansion2213

Flow3527:                                         ; preds = %Flow3526, %loop-memcpy-residual-header2216
  call void @llvm.amdgcn.end.cf.i64(i64 %2585)
  br label %Flow3531, !amdgpu.uniform !42

loop-memcpy-residual2215:                         ; preds = %loop-memcpy-residual2215.preheader, %loop-memcpy-residual2215
  %phi.broken3814 = phi i64 [ 0, %loop-memcpy-residual2215.preheader ], [ %1425, %loop-memcpy-residual2215 ]
  %residual-loop-index2217 = phi i64 [ %1423, %loop-memcpy-residual2215 ], [ 0, %loop-memcpy-residual2215.preheader ]
  %scevgep2986 = getelementptr i8, ptr %agg.tmp13.i.sroa.0.0, i64 %residual-loop-index2217
  %1421 = load i8, ptr %scevgep2986, align 1
  %1422 = getelementptr i8, ptr %arrayidx.i22.i38.i.i.i.i1023, i64 %residual-loop-index2217
  store i8 %1421, ptr %1422, align 1
  %1423 = add i64 %residual-loop-index2217, 1
  %1424 = icmp uge i64 %1423, %1306
  %1425 = call i64 @llvm.amdgcn.if.break.i64(i1 %1424, i64 %phi.broken3814)
  %1426 = call i1 @llvm.amdgcn.loop.i64(i64 %1425)
  br i1 %1426, label %Flow3526, label %loop-memcpy-residual2215

Flow3526:                                         ; preds = %loop-memcpy-residual2215
  %.lcssa3961 = phi i64 [ %1425, %loop-memcpy-residual2215 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa3961)
  br label %Flow3527, !amdgpu.uniform !42

_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1014: ; preds = %Flow3531
  %conv.i.i26.i.i.i.i1015 = xor i32 %port.sroa.37.0.i.i1003, 1
  fence release
  %1427 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %arrayidx.i24.i29.i.i.i.i1016 = getelementptr inbounds nuw i32, ptr %1427, i64 %idxprom.i23.i.i.i.i.i999
  store atomic i32 %conv.i.i26.i.i.i.i1015, ptr %arrayidx.i24.i29.i.i.i.i1016 monotonic, align 4
  %add.i.i.i.i1017 = add i64 %idx.0.i.i.i.i1004, 64
  br label %Flow3535, !amdgpu.uniform !42

cond.false.i.i.i1024:                             ; preds = %Flow3535
  %.lcssa3963 = phi i64 [ %1405, %Flow3535 ]
  %port.sroa.37.0.i.i1003.lcssa = phi i32 [ %port.sroa.37.0.i.i1003, %Flow3535 ]
  %.lcssa3962 = phi i32 [ %1393, %Flow3535 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa3963)
  %1428 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %.lcssa3962)
  %1429 = icmp ne i32 %1428, %port.sroa.37.0.i.i1003.lcssa
  %1430 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %1429)
  %1431 = extractvalue { i1, i64 } %1430, 0
  %1432 = extractvalue { i1, i64 } %1430, 1
  br i1 %1431, label %while.body.i.i.i.i1025.preheader, label %Flow3524

while.body.i.i.i.i1025.preheader:                 ; preds = %cond.false.i.i.i1024
  br label %while.body.i.i.i.i1025, !amdgpu.uniform !42

Flow3524:                                         ; preds = %Flow3522, %cond.false.i.i.i1024
  call void @llvm.amdgcn.end.cf.i64(i64 %1432)
  br label %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsINS_3cpp6stringEEERS1_T_.exit1046, !amdgpu.uniform !42

while.body.i.i.i.i1025:                           ; preds = %while.body.i.i.i.i1025.preheader, %while.body.i.i.i.i1025
  %phi.broken3816 = phi i64 [ 0, %while.body.i.i.i.i1025.preheader ], [ %1436, %while.body.i.i.i.i1025 ]
  tail call void @llvm.amdgcn.s.sleep(i32 2)
  %1433 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i8.i.i1026 = getelementptr inbounds nuw i32, ptr %1433, i64 %idxprom.i23.i.i.i.i.i999, !amdgpu.uniform !42
  %1434 = load atomic i32, ptr %arrayidx.i.i.i8.i.i1026 monotonic, align 4
  %1435 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %1434)
  %cmp.i.not.i.i.i.i1027 = icmp eq i32 %1435, %port.sroa.37.0.i.i1003.lcssa
  %1436 = call i64 @llvm.amdgcn.if.break.i64(i1 %cmp.i.not.i.i.i.i1027, i64 %phi.broken3816)
  %1437 = call i1 @llvm.amdgcn.loop.i64(i64 %1436)
  br i1 %1437, label %Flow3522, label %while.body.i.i.i.i1025

Flow3522:                                         ; preds = %while.body.i.i.i.i1025
  %.lcssa3959 = phi i64 [ %1436, %while.body.i.i.i.i1025 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa3959)
  br label %Flow3524, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsINS_3cpp6stringEEERS1_T_.exit1046: ; preds = %Flow3524
  fence acquire
  tail call void @llvm.amdgcn.wave.barrier()
  fence syncscope("agent") release
  %1438 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %.lcssa3970, i1 true)
  %iszero.i.i.i.i.i13.i.i1029 = icmp ne i64 %.lcssa3970, 0
  %cmp2.i.i.i.i14.i.i1030 = icmp eq i64 %1438, %sh_prom.i.i.i.i
  %cmp.i.i.i.i15.i.i1031 = select i1 %iszero.i.i.i.i.i13.i.i1029, i1 %cmp2.i.i.i.i14.i.i1030, i1 false
  %1439 = and i32 %.lcssa3971, 31
  %conv4.i.i.i19.i.i1035 = zext i1 %cmp.i.i.i.i15.i.i1031 to i32
  %shl.i.i.i20.i.i1036 = shl nuw i32 %conv4.i.i.i19.i.i1035, %1439
  %xor.i.i.i21.i.i1037 = xor i32 %shl.i.i.i20.i.i1036, -1
  %1440 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2550, !amdgpu.uniform !42

1441:                                             ; preds = %ComputeEnd2551
  %div8.i.i.i16.i.i1032 = lshr i32 %.lcssa3971, 5
  %idxprom.i.i.i17.i.i1033 = zext nneg i32 %div8.i.i.i16.i.i1032 to i64
  %1442 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i17.i.i1033
  %sunkaddr3269 = getelementptr inbounds i8, ptr addrspace(1) %1442, i64 40
  %1443 = atomicrmw and ptr addrspace(1) %sunkaddr3269, i32 %.lcssa3958 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %1444, !amdgpu.uniform !42

1444:                                             ; preds = %1441, %ComputeEnd2551
  call void @llvm.amdgcn.end.cf.i64(i64 %3337)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %for.cond.i.i.i1050, !amdgpu.uniform !42

for.cond.i.i.i1050:                               ; preds = %1444, %Flow3521
  %1445 = phi i32 [ %1483, %Flow3521 ], [ poison, %1444 ]
  %1446 = phi i64 [ %1484, %Flow3521 ], [ poison, %1444 ]
  %1447 = phi i32 [ %1485, %Flow3521 ], [ poison, %1444 ]
  %retval.sroa.6.0.i.i.i1051 = phi i32 [ %1488, %Flow3521 ], [ undef, %1444 ]
  %retval.sroa.2.0.i.i.i1053 = phi i64 [ %1487, %Flow3521 ], [ undef, %1444 ]
  %retval.sroa.8.0.i.i.i1054 = phi i32 [ %1486, %Flow3521 ], [ undef, %1444 ]
  %index.0.i.i.i1055 = phi i32 [ %1490, %Flow3521 ], [ 0, %1444 ]
  %1448 = load i32, ptr addrspace(1) @__llvm_rpc_client, align 8, !tbaa !68
  %cmp.not.i.i.i1056 = icmp ult i32 %index.0.i.i.i1055, %1448
  %spec.store.select.i.i.i1057 = select i1 %cmp.not.i.i.i1056, i32 %index.0.i.i.i1055, i32 0
  %1449 = tail call noundef i64 @llvm.amdgcn.ballot.i64(i1 true)
  %1450 = and i32 %spec.store.select.i.i.i1057, 31
  %1451 = lshr i64 %1449, %sh_prom.i.i.i.i
  %1452 = trunc i64 %1451 to i32
  %conv4.i.i.i.i.i1061 = and i32 %1452, 1
  %shl.i.i.i.i.i1062 = shl nuw i32 %conv4.i.i.i.i.i1061, %1450
  %1453 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2554, !amdgpu.uniform !42

1454:                                             ; preds = %ComputeEnd2555
  %sunkaddr3270 = getelementptr inbounds i8, ptr addrspace(1) %3347, i64 40
  %1455 = atomicrmw or ptr addrspace(1) %sunkaddr3270, i32 %.lcssa3952 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %1456, !amdgpu.uniform !42

1456:                                             ; preds = %1454, %ComputeEnd2555
  %1457 = phi i32 [ %1455, %1454 ], [ poison, %ComputeEnd2555 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %3356)
  %1458 = call i32 @llvm.amdgcn.readfirstlane.i32(i32 %1457)
  %1459 = or i32 %1458, %.lcssa3953
  %shl5.i.i.i.i.i1063 = shl nuw i32 1, %1450
  %and.i.i.i.i.i1064 = and i32 %shl5.i.i.i.i.i1063, %1459
  %tobool3.i.i.i.i1065 = icmp ne i32 %and.i.i.i.i.i1064, 0
  %1460 = tail call i64 @llvm.amdgcn.ballot.i64(i1 %tobool3.i.i.i.i1065)
  %and.i.i.i.i.i.i1066 = and i64 %1460, %1449
  %cmp.i.not.i.i.i1067 = icmp ne i64 %1449, %and.i.i.i.i.i.i1066
  br i1 %cmp.i.not.i.i.i1067, label %if.then.i.i.i.i1068, label %Flow3520, !amdgpu.uniform !42

if.then.i.i.i.i1068:                              ; preds = %1456
  fence syncscope("agent") acquire
  %1461 = load <2 x i64>, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !75
  %1462 = extractelement <2 x i64> %1461, i32 0
  %1463 = inttoptr i64 %1462 to ptr
  %1464 = extractelement <2 x i64> %1461, i32 1
  %1465 = inttoptr i64 %1464 to ptr
  %idxprom.i.i.i.i1069 = zext i32 %spec.store.select.i.i.i1057 to i64
  %arrayidx.i.i.i.i1070 = getelementptr inbounds nuw i32, ptr %1463, i64 %idxprom.i.i.i.i1069, !amdgpu.uniform !42
  %1466 = load atomic i32, ptr %arrayidx.i.i.i.i1070 monotonic, align 4
  %1467 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %1466)
  %arrayidx.i45.i.i.i1071 = getelementptr inbounds nuw i32, ptr %1465, i64 %idxprom.i.i.i.i1069, !amdgpu.uniform !42
  %1468 = load atomic i32, ptr %arrayidx.i45.i.i.i1071 monotonic, align 4
  %1469 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %1468)
  %cmp.i46.not.i.i.i1072 = icmp ne i32 %1467, %1469
  br i1 %cmp.i46.not.i.i.i1072, label %if.then12.i.i.i1073, label %Flow3518, !amdgpu.uniform !42

if.then12.i.i.i1073:                              ; preds = %if.then.i.i.i.i1068
  fence syncscope("agent") release
  %1470 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %1449, i1 true)
  %iszero.i.i.i.i.i.i.i1074 = icmp ne i64 %1449, 0
  %cmp2.i.i.i.i.i.i1075 = icmp eq i64 %1470, %sh_prom.i.i.i.i
  %cmp.i.i.i.i.i.i1076 = select i1 %iszero.i.i.i.i.i.i.i1074, i1 %cmp2.i.i.i.i.i.i1075, i1 false
  %conv4.i.i51.i.i.i1077 = zext i1 %cmp.i.i.i.i.i.i1076 to i32
  %shl.i.i52.i.i.i1078 = shl nuw i32 %conv4.i.i51.i.i.i1077, %1450
  %xor.i.i.i.i.i1079 = xor i32 %shl.i.i52.i.i.i1078, -1
  %1471 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2559, !amdgpu.uniform !42

1472:                                             ; preds = %ComputeEnd2560
  %sunkaddr3271 = getelementptr inbounds i8, ptr addrspace(1) %3347, i64 40
  %1473 = atomicrmw and ptr addrspace(1) %sunkaddr3271, i32 %.lcssa3954 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %1474, !amdgpu.uniform !42

1474:                                             ; preds = %1472, %ComputeEnd2560
  call void @llvm.amdgcn.end.cf.i64(i64 %3374)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %Flow3518, !amdgpu.uniform !42

if.end14.i.i.i1137:                               ; preds = %Flow3518
  %1475 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %1449, i1 true)
  %iszero.i.i.i.i.i.i1138 = icmp ne i64 %1449, 0
  %cmp2.i.i.i.i.i1139 = icmp eq i64 %1475, %sh_prom.i.i.i.i
  %cmp.i.i.i.i.i1140 = select i1 %iszero.i.i.i.i.i.i1138, i1 %cmp2.i.i.i.i.i1139, i1 false
  %1476 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.i.i.i.i.i1140)
  %1477 = extractvalue { i1, i64 } %1476, 0
  %1478 = extractvalue { i1, i64 } %1476, 1
  br i1 %1477, label %if.then16.i.i.i1142, label %if.end22.i.i.i1141

if.then16.i.i.i1142:                              ; preds = %if.end14.i.i.i1137
  %1479 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !76
  %opcode.i.i.i1143 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %1479, i64 %idxprom.i.i.i.i1069, i32 1
  store i32 1660944387, ptr %opcode.i.i.i1143, align 8, !tbaa !77
  %arrayidx21.i.i.i1144 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %1479, i64 %idxprom.i.i.i.i1069
  store i64 %1449, ptr %arrayidx21.i.i.i1144, align 8, !tbaa !80
  br label %if.end22.i.i.i1141, !amdgpu.uniform !42

Flow3519:                                         ; preds = %if.end22.i.i.i1141, %Flow3518
  %1480 = phi i32 [ %1467, %if.end22.i.i.i1141 ], [ %retval.sroa.8.0.i.i.i1054, %Flow3518 ]
  %1481 = phi i64 [ %1449, %if.end22.i.i.i1141 ], [ %retval.sroa.2.0.i.i.i1053, %Flow3518 ]
  %1482 = phi i32 [ %spec.store.select.i.i.i1057, %if.end22.i.i.i1141 ], [ %retval.sroa.6.0.i.i.i1051, %Flow3518 ]
  br label %cleanup26.i.i.i1080, !amdgpu.uniform !42

if.end22.i.i.i1141:                               ; preds = %if.then16.i.i.i1142, %if.end14.i.i.i1137
  call void @llvm.amdgcn.end.cf.i64(i64 %1478)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %Flow3519, !amdgpu.uniform !42

Flow3520:                                         ; preds = %cleanup26.i.i.i1080, %1456
  %1483 = phi i32 [ %1482, %cleanup26.i.i.i1080 ], [ %1445, %1456 ]
  %1484 = phi i64 [ %1481, %cleanup26.i.i.i1080 ], [ %1446, %1456 ]
  %1485 = phi i32 [ %1480, %cleanup26.i.i.i1080 ], [ %1447, %1456 ]
  %1486 = phi i32 [ %1480, %cleanup26.i.i.i1080 ], [ %retval.sroa.8.0.i.i.i1054, %1456 ]
  %1487 = phi i64 [ %1481, %cleanup26.i.i.i1080 ], [ %retval.sroa.2.0.i.i.i1053, %1456 ]
  %1488 = phi i32 [ %1482, %cleanup26.i.i.i1080 ], [ %retval.sroa.6.0.i.i.i1051, %1456 ]
  %1489 = phi i1 [ %cmp.i46.not.i.i.i1072, %cleanup26.i.i.i1080 ], [ true, %1456 ]
  br i1 %1489, label %for.inc.i.i.i1085, label %Flow3521, !amdgpu.uniform !42

cleanup26.i.i.i1080:                              ; preds = %Flow3519
  br label %Flow3520, !amdgpu.uniform !42

for.inc.i.i.i1085:                                ; preds = %Flow3520
  %inc.i.i.i1090 = add i32 %spec.store.select.i.i.i1057, 1
  br label %Flow3521, !amdgpu.uniform !42

Flow3521:                                         ; preds = %for.inc.i.i.i1085, %Flow3520
  %1490 = phi i32 [ %inc.i.i.i1090, %for.inc.i.i.i1085 ], [ poison, %Flow3520 ]
  %1491 = phi i1 [ false, %for.inc.i.i.i1085 ], [ true, %Flow3520 ]
  br i1 %1491, label %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1091, label %for.cond.i.i.i1050, !amdgpu.uniform !42

_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1091: ; preds = %Flow3521
  %.lcssa3957 = phi i32 [ %1483, %Flow3521 ]
  %.lcssa3956 = phi i64 [ %1484, %Flow3521 ]
  %.lcssa3955 = phi i32 [ %1485, %Flow3521 ]
  fence acquire
  %1492 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %mul.i.i.i.i.i.i1092 = shl i32 %.lcssa3957, 6
  %idxprom.i19.i.i.i.i.i1093 = zext i32 %mul.i.i.i.i.i.i1092 to i64
  %arrayidx.i20.i.i.i.i.i1094 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %1492, i64 %idxprom.i19.i.i.i.i.i1093
  %arrayidx.i22.i.i.i.i.i1095 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %arrayidx.i20.i.i.i.i.i1094, i64 %sh_prom.i.i.i.i
  store i64 1, ptr %arrayidx.i22.i.i.i.i.i1095, align 8, !tbaa !82
  %arrayidx6.i.i.i.i.i.i.i1096 = getelementptr inbounds nuw i8, ptr %arrayidx.i22.i.i.i.i.i1095, i64 8
  store i8 10, ptr %arrayidx6.i.i.i.i.i.i.i1096, align 8
  %tobool.not.i.i.i.i.i.i1097 = icmp eq i32 %.lcssa3955, 0
  %conv.i.i.i.i.i.i1098 = zext i1 %tobool.not.i.i.i.i.i.i1097 to i32
  fence release
  %1493 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %idxprom.i23.i.i.i.i.i1099 = zext i32 %.lcssa3957 to i64
  %arrayidx.i24.i.i.i.i.i1100 = getelementptr inbounds nuw i32, ptr %1493, i64 %idxprom.i23.i.i.i.i.i1099
  store atomic i32 %conv.i.i.i.i.i.i1098, ptr %arrayidx.i24.i.i.i.i.i1100 monotonic, align 4
  %1494 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !84
  %arrayidx.i.i5.i.i1101 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %1494, i64 %idxprom.i23.i.i.i.i.i1099, !amdgpu.uniform !42
  %1495 = load i64, ptr %arrayidx.i.i5.i.i1101, align 8, !tbaa !80
  br label %while.cond.i.i.i.i1102, !amdgpu.uniform !42

while.cond.i.i.i.i1102:                           ; preds = %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1091, %Flow3517
  %phi.broken3817 = phi i64 [ 0, %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1091 ], [ %1510, %Flow3517 ]
  %port.sroa.37.0.i.i1103 = phi i32 [ %1508, %Flow3517 ], [ %conv.i.i.i.i.i.i1098, %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1091 ]
  %idx.0.i.i.i.i1104 = phi i64 [ %1507, %Flow3517 ], [ 56, %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1091 ]
  %cmp.i.i.i.i1105 = icmp eq i64 %idx.0.i.i.i.i1104, 0
  %1496 = tail call i64 @llvm.amdgcn.ballot.i64(i1 %cmp.i.i.i.i1105)
  %and.i.i.i.i6.i.i1106 = and i64 %1496, %1495
  %tobool.not.i.i.i.i1107 = icmp ne i64 %and.i.i.i.i6.i.i1106, 0
  %1497 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i22.i.i.i1108 = getelementptr inbounds nuw i32, ptr %1497, i64 %idxprom.i23.i.i.i.i.i1099, !amdgpu.uniform !42
  %1498 = load atomic i32, ptr %arrayidx.i22.i.i.i1108 monotonic, align 4
  %1499 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %tobool.not.i.i.i.i1107)
  %1500 = extractvalue { i1, i64 } %1499, 0
  %1501 = extractvalue { i1, i64 } %1499, 1
  br i1 %1500, label %cond.false.i8.i.i.i.i1109, label %Flow3517

cond.false.i8.i.i.i.i1109:                        ; preds = %while.cond.i.i.i.i1102
  %1502 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %1498)
  %1503 = icmp ne i32 %1502, %port.sroa.37.0.i.i1103
  br i1 %1503, label %while.body.i.i21.i.i.i.i1110.preheader, label %Flow3516, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i1110.preheader:           ; preds = %cond.false.i8.i.i.i.i1109
  br label %while.body.i.i21.i.i.i.i1110, !amdgpu.uniform !42

Flow3516:                                         ; preds = %Flow3514, %cond.false.i8.i.i.i.i1109
  br label %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i1113, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i1110:                     ; preds = %while.body.i.i21.i.i.i.i1110.preheader, %while.body.i.i21.i.i.i.i1110
  tail call void @llvm.amdgcn.s.sleep(i32 2)
  %1504 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i22.i.i.i.i1111 = getelementptr inbounds nuw i32, ptr %1504, i64 %idxprom.i23.i.i.i.i.i1099, !amdgpu.uniform !42
  %1505 = load atomic i32, ptr %arrayidx.i.i.i22.i.i.i.i1111 monotonic, align 4
  %1506 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %1505)
  %cmp.i.not.i.i23.i.i.i.i1112 = icmp eq i32 %1506, %port.sroa.37.0.i.i1103
  br i1 %cmp.i.not.i.i23.i.i.i.i1112, label %Flow3514, label %while.body.i.i21.i.i.i.i1110, !amdgpu.uniform !42

Flow3514:                                         ; preds = %while.body.i.i21.i.i.i.i1110
  br label %Flow3516, !amdgpu.uniform !42

Flow3517:                                         ; preds = %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1114, %while.cond.i.i.i.i1102
  %1507 = phi i64 [ %add.i.i.i.i1117, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1114 ], [ poison, %while.cond.i.i.i.i1102 ]
  %1508 = phi i32 [ %conv.i.i26.i.i.i.i1115, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1114 ], [ poison, %while.cond.i.i.i.i1102 ]
  %1509 = phi i1 [ false, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1114 ], [ true, %while.cond.i.i.i.i1102 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %1501)
  %1510 = call i64 @llvm.amdgcn.if.break.i64(i1 %1509, i64 %phi.broken3817)
  %1511 = call i1 @llvm.amdgcn.loop.i64(i64 %1510)
  br i1 %1511, label %cond.false.i.i.i1123, label %while.cond.i.i.i.i1102

_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i1113: ; preds = %Flow3516
  fence acquire
  br i1 %cmp.i.i.i.i1105, label %if.then.i.i.i.i.i.i.i1118, label %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1114, !amdgpu.uniform !42

if.then.i.i.i.i.i.i.i1118:                        ; preds = %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i1113
  %1512 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %arrayidx.i20.i35.i.i.i.i1121 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %1512, i64 %idxprom.i19.i.i.i.i.i1093
  %arrayidx.i22.i38.i.i.i.i1122 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %arrayidx.i20.i35.i.i.i.i1121, i64 %sh_prom.i.i.i.i
  store i8 10, ptr %arrayidx.i22.i38.i.i.i.i1122, align 1
  br label %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1114, !amdgpu.uniform !42

_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1114: ; preds = %if.then.i.i.i.i.i.i.i1118, %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i1113
  %conv.i.i26.i.i.i.i1115 = xor i32 %port.sroa.37.0.i.i1103, 1
  fence release
  %1513 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %arrayidx.i24.i29.i.i.i.i1116 = getelementptr inbounds nuw i32, ptr %1513, i64 %idxprom.i23.i.i.i.i.i1099
  store atomic i32 %conv.i.i26.i.i.i.i1115, ptr %arrayidx.i24.i29.i.i.i.i1116 monotonic, align 4
  %add.i.i.i.i1117 = add i64 %idx.0.i.i.i.i1104, 64
  br label %Flow3517, !amdgpu.uniform !42

cond.false.i.i.i1123:                             ; preds = %Flow3517
  %.lcssa3951 = phi i64 [ %1510, %Flow3517 ]
  %port.sroa.37.0.i.i1103.lcssa = phi i32 [ %port.sroa.37.0.i.i1103, %Flow3517 ]
  %.lcssa3950 = phi i32 [ %1498, %Flow3517 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa3951)
  %1514 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %.lcssa3950)
  %1515 = icmp ne i32 %1514, %port.sroa.37.0.i.i1103.lcssa
  %1516 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %1515)
  %1517 = extractvalue { i1, i64 } %1516, 0
  %1518 = extractvalue { i1, i64 } %1516, 1
  br i1 %1517, label %while.body.i.i.i.i1124.preheader, label %Flow3513

while.body.i.i.i.i1124.preheader:                 ; preds = %cond.false.i.i.i1123
  br label %while.body.i.i.i.i1124, !amdgpu.uniform !42

Flow3513:                                         ; preds = %Flow3511, %cond.false.i.i.i1123
  call void @llvm.amdgcn.end.cf.i64(i64 %1518)
  br label %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIcEERS1_T_.exit1145, !amdgpu.uniform !42

while.body.i.i.i.i1124:                           ; preds = %while.body.i.i.i.i1124.preheader, %while.body.i.i.i.i1124
  %phi.broken3818 = phi i64 [ 0, %while.body.i.i.i.i1124.preheader ], [ %1522, %while.body.i.i.i.i1124 ]
  tail call void @llvm.amdgcn.s.sleep(i32 2)
  %1519 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i8.i.i1125 = getelementptr inbounds nuw i32, ptr %1519, i64 %idxprom.i23.i.i.i.i.i1099, !amdgpu.uniform !42
  %1520 = load atomic i32, ptr %arrayidx.i.i.i8.i.i1125 monotonic, align 4
  %1521 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %1520)
  %cmp.i.not.i.i.i.i1126 = icmp eq i32 %1521, %port.sroa.37.0.i.i1103.lcssa
  %1522 = call i64 @llvm.amdgcn.if.break.i64(i1 %cmp.i.not.i.i.i.i1126, i64 %phi.broken3818)
  %1523 = call i1 @llvm.amdgcn.loop.i64(i64 %1522)
  br i1 %1523, label %Flow3511, label %while.body.i.i.i.i1124

Flow3511:                                         ; preds = %while.body.i.i.i.i1124
  %.lcssa3949 = phi i64 [ %1522, %while.body.i.i.i.i1124 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa3949)
  br label %Flow3513, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIcEERS1_T_.exit1145: ; preds = %Flow3513
  fence acquire
  tail call void @llvm.amdgcn.wave.barrier()
  fence syncscope("agent") release
  %1524 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %.lcssa3956, i1 true)
  %iszero.i.i.i.i.i13.i.i1128 = icmp ne i64 %.lcssa3956, 0
  %cmp2.i.i.i.i14.i.i1129 = icmp eq i64 %1524, %sh_prom.i.i.i.i
  %cmp.i.i.i.i15.i.i1130 = select i1 %iszero.i.i.i.i.i13.i.i1128, i1 %cmp2.i.i.i.i14.i.i1129, i1 false
  %1525 = and i32 %.lcssa3957, 31
  %conv4.i.i.i19.i.i1134 = zext i1 %cmp.i.i.i.i15.i.i1130 to i32
  %shl.i.i.i20.i.i1135 = shl nuw i32 %conv4.i.i.i19.i.i1134, %1525
  %xor.i.i.i21.i.i1136 = xor i32 %shl.i.i.i20.i.i1135, -1
  %1526 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2563, !amdgpu.uniform !42

1527:                                             ; preds = %ComputeEnd2564
  %div8.i.i.i16.i.i1131 = lshr i32 %.lcssa3957, 5
  %idxprom.i.i.i17.i.i1132 = zext nneg i32 %div8.i.i.i16.i.i1131 to i64
  %1528 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i17.i.i1132
  %sunkaddr3272 = getelementptr inbounds i8, ptr addrspace(1) %1528, i64 40
  %1529 = atomicrmw and ptr addrspace(1) %sunkaddr3272, i32 %.lcssa3948 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %1530, !amdgpu.uniform !42

1530:                                             ; preds = %1527, %ComputeEnd2564
  call void @llvm.amdgcn.end.cf.i64(i64 %3391)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %for.cond.i.i1.i1154, !amdgpu.uniform !42

for.cond.i.i1.i1154:                              ; preds = %1530, %Flow3510
  %1531 = phi i32 [ %1569, %Flow3510 ], [ poison, %1530 ]
  %1532 = phi i64 [ %1570, %Flow3510 ], [ poison, %1530 ]
  %1533 = phi i32 [ %1571, %Flow3510 ], [ poison, %1530 ]
  %retval.sroa.6.0.i.i.i1155 = phi i32 [ %1574, %Flow3510 ], [ undef, %1530 ]
  %retval.sroa.2.0.i.i.i1157 = phi i64 [ %1573, %Flow3510 ], [ undef, %1530 ]
  %retval.sroa.8.0.i.i.i1158 = phi i32 [ %1572, %Flow3510 ], [ undef, %1530 ]
  %index.0.i.i.i1159 = phi i32 [ %1576, %Flow3510 ], [ 0, %1530 ]
  %1534 = load i32, ptr addrspace(1) @__llvm_rpc_client, align 8, !tbaa !68
  %cmp.not.i.i.i1160 = icmp ult i32 %index.0.i.i.i1159, %1534
  %spec.store.select.i.i.i1161 = select i1 %cmp.not.i.i.i1160, i32 %index.0.i.i.i1159, i32 0
  %1535 = tail call noundef i64 @llvm.amdgcn.ballot.i64(i1 true)
  %1536 = and i32 %spec.store.select.i.i.i1161, 31
  %1537 = lshr i64 %1535, %sh_prom.i.i.i.i
  %1538 = trunc i64 %1537 to i32
  %conv4.i.i.i.i.i1165 = and i32 %1538, 1
  %shl.i.i.i.i.i1166 = shl nuw i32 %conv4.i.i.i.i.i1165, %1536
  %1539 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2567, !amdgpu.uniform !42

1540:                                             ; preds = %ComputeEnd2568
  %sunkaddr3273 = getelementptr inbounds i8, ptr addrspace(1) %3401, i64 40
  %1541 = atomicrmw or ptr addrspace(1) %sunkaddr3273, i32 %.lcssa3942 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %1542, !amdgpu.uniform !42

1542:                                             ; preds = %1540, %ComputeEnd2568
  %1543 = phi i32 [ %1541, %1540 ], [ poison, %ComputeEnd2568 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %3410)
  %1544 = call i32 @llvm.amdgcn.readfirstlane.i32(i32 %1543)
  %1545 = or i32 %1544, %.lcssa3943
  %shl5.i.i.i.i.i1167 = shl nuw i32 1, %1536
  %and.i.i.i.i.i1168 = and i32 %shl5.i.i.i.i.i1167, %1545
  %tobool3.i.i.i.i1169 = icmp ne i32 %and.i.i.i.i.i1168, 0
  %1546 = tail call i64 @llvm.amdgcn.ballot.i64(i1 %tobool3.i.i.i.i1169)
  %and.i.i.i.i.i.i1170 = and i64 %1546, %1535
  %cmp.i.not.i.i.i1171 = icmp ne i64 %1535, %and.i.i.i.i.i.i1170
  br i1 %cmp.i.not.i.i.i1171, label %if.then.i.i.i.i1172, label %Flow3509, !amdgpu.uniform !42

if.then.i.i.i.i1172:                              ; preds = %1542
  fence syncscope("agent") acquire
  %1547 = load <2 x i64>, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !75
  %1548 = extractelement <2 x i64> %1547, i32 0
  %1549 = inttoptr i64 %1548 to ptr
  %1550 = extractelement <2 x i64> %1547, i32 1
  %1551 = inttoptr i64 %1550 to ptr
  %idxprom.i.i.i.i1173 = zext i32 %spec.store.select.i.i.i1161 to i64
  %arrayidx.i.i.i.i1174 = getelementptr inbounds nuw i32, ptr %1549, i64 %idxprom.i.i.i.i1173, !amdgpu.uniform !42
  %1552 = load atomic i32, ptr %arrayidx.i.i.i.i1174 monotonic, align 4
  %1553 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %1552)
  %arrayidx.i45.i.i.i1175 = getelementptr inbounds nuw i32, ptr %1551, i64 %idxprom.i.i.i.i1173, !amdgpu.uniform !42
  %1554 = load atomic i32, ptr %arrayidx.i45.i.i.i1175 monotonic, align 4
  %1555 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %1554)
  %cmp.i46.not.i.i.i1176 = icmp ne i32 %1553, %1555
  br i1 %cmp.i46.not.i.i.i1176, label %if.then12.i.i.i1177, label %Flow3507, !amdgpu.uniform !42

if.then12.i.i.i1177:                              ; preds = %if.then.i.i.i.i1172
  fence syncscope("agent") release
  %1556 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %1535, i1 true)
  %iszero.i.i.i.i.i.i.i1178 = icmp ne i64 %1535, 0
  %cmp2.i.i.i.i.i.i1179 = icmp eq i64 %1556, %sh_prom.i.i.i.i
  %cmp.i.i.i.i.i.i1180 = select i1 %iszero.i.i.i.i.i.i.i1178, i1 %cmp2.i.i.i.i.i.i1179, i1 false
  %conv4.i.i51.i.i.i1181 = zext i1 %cmp.i.i.i.i.i.i1180 to i32
  %shl.i.i52.i.i.i1182 = shl nuw i32 %conv4.i.i51.i.i.i1181, %1536
  %xor.i.i.i.i.i1183 = xor i32 %shl.i.i52.i.i.i1182, -1
  %1557 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2572, !amdgpu.uniform !42

1558:                                             ; preds = %ComputeEnd2573
  %sunkaddr3274 = getelementptr inbounds i8, ptr addrspace(1) %3401, i64 40
  %1559 = atomicrmw and ptr addrspace(1) %sunkaddr3274, i32 %.lcssa3944 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %1560, !amdgpu.uniform !42

1560:                                             ; preds = %1558, %ComputeEnd2573
  call void @llvm.amdgcn.end.cf.i64(i64 %3428)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %Flow3507, !amdgpu.uniform !42

if.end14.i.i.i1243:                               ; preds = %Flow3507
  %1561 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %1535, i1 true)
  %iszero.i.i.i.i.i.i1244 = icmp ne i64 %1535, 0
  %cmp2.i.i.i.i.i1245 = icmp eq i64 %1561, %sh_prom.i.i.i.i
  %cmp.i.i.i.i.i1246 = select i1 %iszero.i.i.i.i.i.i1244, i1 %cmp2.i.i.i.i.i1245, i1 false
  %1562 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.i.i.i.i.i1246)
  %1563 = extractvalue { i1, i64 } %1562, 0
  %1564 = extractvalue { i1, i64 } %1562, 1
  br i1 %1563, label %if.then16.i.i.i1248, label %if.end22.i.i.i1247

if.then16.i.i.i1248:                              ; preds = %if.end14.i.i.i1243
  %1565 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !76
  %opcode.i.i.i1249 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %1565, i64 %idxprom.i.i.i.i1173, i32 1
  store i32 1660944387, ptr %opcode.i.i.i1249, align 8, !tbaa !77
  %arrayidx21.i.i.i1250 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %1565, i64 %idxprom.i.i.i.i1173
  store i64 %1535, ptr %arrayidx21.i.i.i1250, align 8, !tbaa !80
  br label %if.end22.i.i.i1247, !amdgpu.uniform !42

Flow3508:                                         ; preds = %if.end22.i.i.i1247, %Flow3507
  %1566 = phi i32 [ %1553, %if.end22.i.i.i1247 ], [ %retval.sroa.8.0.i.i.i1158, %Flow3507 ]
  %1567 = phi i64 [ %1535, %if.end22.i.i.i1247 ], [ %retval.sroa.2.0.i.i.i1157, %Flow3507 ]
  %1568 = phi i32 [ %spec.store.select.i.i.i1161, %if.end22.i.i.i1247 ], [ %retval.sroa.6.0.i.i.i1155, %Flow3507 ]
  br label %cleanup26.i.i.i1184, !amdgpu.uniform !42

if.end22.i.i.i1247:                               ; preds = %if.then16.i.i.i1248, %if.end14.i.i.i1243
  call void @llvm.amdgcn.end.cf.i64(i64 %1564)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %Flow3508, !amdgpu.uniform !42

Flow3509:                                         ; preds = %cleanup26.i.i.i1184, %1542
  %1569 = phi i32 [ %1568, %cleanup26.i.i.i1184 ], [ %1531, %1542 ]
  %1570 = phi i64 [ %1567, %cleanup26.i.i.i1184 ], [ %1532, %1542 ]
  %1571 = phi i32 [ %1566, %cleanup26.i.i.i1184 ], [ %1533, %1542 ]
  %1572 = phi i32 [ %1566, %cleanup26.i.i.i1184 ], [ %retval.sroa.8.0.i.i.i1158, %1542 ]
  %1573 = phi i64 [ %1567, %cleanup26.i.i.i1184 ], [ %retval.sroa.2.0.i.i.i1157, %1542 ]
  %1574 = phi i32 [ %1568, %cleanup26.i.i.i1184 ], [ %retval.sroa.6.0.i.i.i1155, %1542 ]
  %1575 = phi i1 [ %cmp.i46.not.i.i.i1176, %cleanup26.i.i.i1184 ], [ true, %1542 ]
  br i1 %1575, label %for.inc.i.i.i1189, label %Flow3510, !amdgpu.uniform !42

cleanup26.i.i.i1184:                              ; preds = %Flow3508
  br label %Flow3509, !amdgpu.uniform !42

for.inc.i.i.i1189:                                ; preds = %Flow3509
  %inc.i.i.i1194 = add i32 %spec.store.select.i.i.i1161, 1
  br label %Flow3510, !amdgpu.uniform !42

Flow3510:                                         ; preds = %for.inc.i.i.i1189, %Flow3509
  %1576 = phi i32 [ %inc.i.i.i1194, %for.inc.i.i.i1189 ], [ poison, %Flow3509 ]
  %1577 = phi i1 [ false, %for.inc.i.i.i1189 ], [ true, %Flow3509 ]
  br i1 %1577, label %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1195, label %for.cond.i.i1.i1154, !amdgpu.uniform !42

_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1195: ; preds = %Flow3510
  %.lcssa3947 = phi i32 [ %1569, %Flow3510 ]
  %.lcssa3946 = phi i64 [ %1570, %Flow3510 ]
  %.lcssa3945 = phi i32 [ %1571, %Flow3510 ]
  fence acquire
  %1578 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %mul.i.i.i.i.i.i1196 = shl i32 %.lcssa3947, 6
  %idxprom.i19.i.i.i.i.i1197 = zext i32 %mul.i.i.i.i.i.i1196 to i64
  %arrayidx.i20.i.i.i.i.i1198 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %1578, i64 %idxprom.i19.i.i.i.i.i1197
  %arrayidx.i22.i.i.i.i.i1199 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %arrayidx.i20.i.i.i.i.i1198, i64 %sh_prom.i.i.i.i
  store i64 sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.36.13 to ptr), i64 6) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.36.13 to ptr) to i64)), ptr %arrayidx.i22.i.i.i.i.i1199, align 8, !tbaa !82
  %spec.select.i.i.i.i.i.i.i1200 = tail call i64 @llvm.umin.i64(i64 sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.36.13 to ptr), i64 6) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.36.13 to ptr) to i64)), i64 56)
  %1579 = and i64 %spec.select.i.i.i.i.i.i.i1200, 15
  %1580 = sub i64 %spec.select.i.i.i.i.i.i.i1200, %1579
  %1581 = icmp ne i64 %1580, 0
  br i1 %1581, label %loop-memcpy-expansion2375.preheader, label %Flow3506, !amdgpu.uniform !42

loop-memcpy-expansion2375.preheader:              ; preds = %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1195
  %1582 = add nuw nsw i64 %76, 8
  %scevgep2983 = getelementptr i8, ptr %1578, i64 %1582
  %1583 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i1197, 6
  %scevgep2984 = getelementptr i8, ptr %scevgep2983, i64 %1583
  br label %loop-memcpy-expansion2375, !amdgpu.uniform !42

Flow3506:                                         ; preds = %Flow3505, %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1195
  br label %loop-memcpy-residual-header2378, !amdgpu.uniform !42

loop-memcpy-expansion2375:                        ; preds = %loop-memcpy-expansion2375.preheader, %loop-memcpy-expansion2375
  %loop-index2376 = phi i64 [ %1586, %loop-memcpy-expansion2375 ], [ 0, %loop-memcpy-expansion2375.preheader ]
  %1584 = getelementptr i8, ptr addrspace(4) @.str.36.13, i64 %loop-index2376, !amdgpu.uniform !42
  %1585 = load <4 x i32>, ptr addrspace(4) %1584, align 1
  %scevgep2985 = getelementptr i8, ptr %scevgep2984, i64 %loop-index2376
  store <4 x i32> %1585, ptr %scevgep2985, align 1
  %1586 = add i64 %loop-index2376, 16
  %1587 = icmp uge i64 %1586, %1580
  br i1 %1587, label %Flow3505, label %loop-memcpy-expansion2375, !amdgpu.uniform !42

Flow3503:                                         ; preds = %Flow3500, %Flow3502
  %1588 = phi i64 [ %2718, %Flow3500 ], [ %2715, %Flow3502 ]
  br label %post-loop-memcpy-expansion2374, !amdgpu.uniform !42

loop-memcpy-residual2377:                         ; preds = %loop-memcpy-residual2377.preheader, %loop-memcpy-residual2377
  %residual-loop-index2379 = phi i64 [ %1590, %loop-memcpy-residual2377 ], [ 0, %loop-memcpy-residual2377.preheader ]
  %scevgep2982 = getelementptr i8, ptr addrspace(4) @.str.36.13, i64 %residual-loop-index2379, !amdgpu.uniform !42
  %1589 = load i8, ptr addrspace(4) %scevgep2982, align 1
  %scevgep2981 = getelementptr i8, ptr %scevgep2980, i64 %residual-loop-index2379
  store i8 %1589, ptr %scevgep2981, align 1
  %1590 = add i64 %residual-loop-index2379, 1
  %1591 = icmp uge i64 %1590, %1579
  br i1 %1591, label %Flow3500, label %loop-memcpy-residual2377, !amdgpu.uniform !42

Flow3500:                                         ; preds = %loop-memcpy-residual2377
  br label %Flow3503, !amdgpu.uniform !42

post-loop-memcpy-expansion2374:                   ; preds = %Flow3503
  %tobool.not.i.i.i.i.i.i1202 = icmp eq i32 %.lcssa3945, 0
  %conv.i.i.i.i.i.i1203 = zext i1 %tobool.not.i.i.i.i.i.i1202 to i32
  fence release
  %1592 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %idxprom.i23.i.i.i.i.i1204 = zext i32 %.lcssa3947 to i64
  %arrayidx.i24.i.i.i.i.i1205 = getelementptr inbounds nuw i32, ptr %1592, i64 %idxprom.i23.i.i.i.i.i1204
  store atomic i32 %conv.i.i.i.i.i.i1203, ptr %arrayidx.i24.i.i.i.i.i1205 monotonic, align 4
  %1593 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !84
  %arrayidx.i.i5.i.i1206 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %1593, i64 %idxprom.i23.i.i.i.i.i1204, !amdgpu.uniform !42
  %1594 = load i64, ptr %arrayidx.i.i5.i.i1206, align 8, !tbaa !80
  br label %while.cond.i.i.i.i1207, !amdgpu.uniform !42

while.cond.i.i.i.i1207:                           ; preds = %post-loop-memcpy-expansion2374, %Flow3498
  %phi.broken3819 = phi i64 [ 0, %post-loop-memcpy-expansion2374 ], [ %1611, %Flow3498 ]
  %lsr.iv2972 = phi i64 [ %1609, %Flow3498 ], [ -50, %post-loop-memcpy-expansion2374 ]
  %lsr.iv2967 = phi ptr addrspace(4) [ %1608, %Flow3498 ], [ getelementptr (i8, ptr addrspace(4) @.str.36.13, i64 56), %post-loop-memcpy-expansion2374 ]
  %port.sroa.37.0.i.i1208 = phi i32 [ %1607, %Flow3498 ], [ %conv.i.i.i.i.i.i1203, %post-loop-memcpy-expansion2374 ]
  %idx.0.i.i.i.i1209 = phi i64 [ %1606, %Flow3498 ], [ 56, %post-loop-memcpy-expansion2374 ]
  %cmp.i.i.i.i1210 = icmp ult i64 %idx.0.i.i.i.i1209, sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.36.13 to ptr), i64 6) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.36.13 to ptr) to i64))
  %1595 = tail call i64 @llvm.amdgcn.ballot.i64(i1 %cmp.i.i.i.i1210)
  %and.i.i.i.i6.i.i1211 = and i64 %1595, %1594
  %tobool.not.i.i.i.i1212 = icmp ne i64 %and.i.i.i.i6.i.i1211, 0
  %1596 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i22.i.i.i1213 = getelementptr inbounds nuw i32, ptr %1596, i64 %idxprom.i23.i.i.i.i.i1204, !amdgpu.uniform !42
  %1597 = load atomic i32, ptr %arrayidx.i22.i.i.i1213 monotonic, align 4
  %1598 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %tobool.not.i.i.i.i1212)
  %1599 = extractvalue { i1, i64 } %1598, 0
  %1600 = extractvalue { i1, i64 } %1598, 1
  br i1 %1599, label %cond.false.i8.i.i.i.i1214, label %Flow3498

cond.false.i8.i.i.i.i1214:                        ; preds = %while.cond.i.i.i.i1207
  %1601 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %1597)
  %1602 = icmp ne i32 %1601, %port.sroa.37.0.i.i1208
  br i1 %1602, label %while.body.i.i21.i.i.i.i1215.preheader, label %Flow3497, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i1215.preheader:           ; preds = %cond.false.i8.i.i.i.i1214
  br label %while.body.i.i21.i.i.i.i1215, !amdgpu.uniform !42

Flow3497:                                         ; preds = %Flow3495, %cond.false.i8.i.i.i.i1214
  br label %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i1218, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i1215:                     ; preds = %while.body.i.i21.i.i.i.i1215.preheader, %while.body.i.i21.i.i.i.i1215
  tail call void @llvm.amdgcn.s.sleep(i32 2)
  %1603 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i22.i.i.i.i1216 = getelementptr inbounds nuw i32, ptr %1603, i64 %idxprom.i23.i.i.i.i.i1204, !amdgpu.uniform !42
  %1604 = load atomic i32, ptr %arrayidx.i.i.i22.i.i.i.i1216 monotonic, align 4
  %1605 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %1604)
  %cmp.i.not.i.i23.i.i.i.i1217 = icmp eq i32 %1605, %port.sroa.37.0.i.i1208
  br i1 %cmp.i.not.i.i23.i.i.i.i1217, label %Flow3495, label %while.body.i.i21.i.i.i.i1215, !amdgpu.uniform !42

Flow3495:                                         ; preds = %while.body.i.i21.i.i.i.i1215
  br label %Flow3497, !amdgpu.uniform !42

Flow3498:                                         ; preds = %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1219, %while.cond.i.i.i.i1207
  %1606 = phi i64 [ %add.i.i.i.i1222, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1219 ], [ poison, %while.cond.i.i.i.i1207 ]
  %1607 = phi i32 [ %conv.i.i26.i.i.i.i1220, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1219 ], [ poison, %while.cond.i.i.i.i1207 ]
  %1608 = phi ptr addrspace(4) [ %scevgep2968, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1219 ], [ poison, %while.cond.i.i.i.i1207 ]
  %1609 = phi i64 [ %lsr.iv.next2973, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1219 ], [ poison, %while.cond.i.i.i.i1207 ]
  %1610 = phi i1 [ false, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1219 ], [ true, %while.cond.i.i.i.i1207 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %1600)
  %1611 = call i64 @llvm.amdgcn.if.break.i64(i1 %1610, i64 %phi.broken3819)
  %1612 = call i1 @llvm.amdgcn.loop.i64(i64 %1611)
  br i1 %1612, label %cond.false.i.i.i1229, label %while.cond.i.i.i.i1207

_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i1218: ; preds = %Flow3497
  fence acquire
  br i1 %cmp.i.i.i.i1210, label %if.then.i.i.i.i.i.i.i1223, label %Flow3494, !amdgpu.uniform !42

if.then.i.i.i.i.i.i.i1223:                        ; preds = %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i1218
  %sub.i.i.i.i.i.i.i1225 = sub nuw i64 sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.36.13 to ptr), i64 6) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.36.13 to ptr) to i64)), %idx.0.i.i.i.i1209
  %spec.select.i.i.i36.i.i.i.i1226 = tail call i64 @llvm.umin.i64(i64 %sub.i.i.i.i.i.i.i1225, i64 64)
  %1613 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %1614 = and i64 %spec.select.i.i.i36.i.i.i.i1226, 15
  %1615 = sub i64 %spec.select.i.i.i36.i.i.i.i1226, %1614
  %1616 = icmp ne i64 %1615, 0
  br i1 %1616, label %loop-memcpy-expansion2381.preheader, label %Flow3493, !amdgpu.uniform !42

loop-memcpy-expansion2381.preheader:              ; preds = %if.then.i.i.i.i.i.i.i1223
  %arrayidx.i20.i35.i.i.i.i1227 = getelementptr %"struct.rpc::Buffer", ptr %1613, i64 %idxprom.i19.i.i.i.i.i1197
  %arrayidx.i22.i38.i.i.i.i1228 = getelementptr %"struct.rpc::Buffer", ptr %arrayidx.i20.i35.i.i.i.i1227, i64 %sh_prom.i.i.i.i
  br label %loop-memcpy-expansion2381, !amdgpu.uniform !42

Flow3493:                                         ; preds = %Flow3492, %if.then.i.i.i.i.i.i.i1223
  br label %loop-memcpy-residual-header2384, !amdgpu.uniform !42

loop-memcpy-expansion2381:                        ; preds = %loop-memcpy-expansion2381.preheader, %loop-memcpy-expansion2381
  %loop-index2382 = phi i64 [ %1619, %loop-memcpy-expansion2381 ], [ 0, %loop-memcpy-expansion2381.preheader ]
  %scevgep2969 = getelementptr i8, ptr addrspace(4) %lsr.iv2967, i64 %loop-index2382, !amdgpu.uniform !42
  %1617 = load <4 x i32>, ptr addrspace(4) %scevgep2969, align 1
  %1618 = getelementptr i8, ptr %arrayidx.i22.i38.i.i.i.i1228, i64 %loop-index2382
  store <4 x i32> %1617, ptr %1618, align 1
  %1619 = add i64 %loop-index2382, 16
  %1620 = icmp uge i64 %1619, %1615
  br i1 %1620, label %Flow3492, label %loop-memcpy-expansion2381, !amdgpu.uniform !42

Flow3490:                                         ; preds = %Flow3489, %loop-memcpy-residual-header2384
  br label %Flow3494, !amdgpu.uniform !42

loop-memcpy-residual2383:                         ; preds = %loop-memcpy-residual2383.preheader, %loop-memcpy-residual2383
  %residual-loop-index2385 = phi i64 [ %1622, %loop-memcpy-residual2383 ], [ 0, %loop-memcpy-residual2383.preheader ]
  %scevgep2978 = getelementptr i8, ptr addrspace(4) %scevgep2977, i64 %residual-loop-index2385, !amdgpu.uniform !42
  %1621 = load i8, ptr addrspace(4) %scevgep2978, align 1
  %scevgep2976 = getelementptr i8, ptr %scevgep2975, i64 %residual-loop-index2385
  store i8 %1621, ptr %scevgep2976, align 1
  %1622 = add i64 %residual-loop-index2385, 1
  %1623 = icmp uge i64 %1622, %1614
  br i1 %1623, label %Flow3489, label %loop-memcpy-residual2383, !amdgpu.uniform !42

Flow3489:                                         ; preds = %loop-memcpy-residual2383
  br label %Flow3490, !amdgpu.uniform !42

_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1219: ; preds = %Flow3494
  %conv.i.i26.i.i.i.i1220 = xor i32 %port.sroa.37.0.i.i1208, 1
  fence release
  %1624 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %arrayidx.i24.i29.i.i.i.i1221 = getelementptr inbounds nuw i32, ptr %1624, i64 %idxprom.i23.i.i.i.i.i1204
  store atomic i32 %conv.i.i26.i.i.i.i1220, ptr %arrayidx.i24.i29.i.i.i.i1221 monotonic, align 4
  %add.i.i.i.i1222 = add i64 %idx.0.i.i.i.i1209, 64
  %scevgep2968 = getelementptr i8, ptr addrspace(4) %lsr.iv2967, i64 64
  %lsr.iv.next2973 = add i64 %lsr.iv2972, -64
  br label %Flow3498, !amdgpu.uniform !42

cond.false.i.i.i1229:                             ; preds = %Flow3498
  %.lcssa3941 = phi i64 [ %1611, %Flow3498 ]
  %port.sroa.37.0.i.i1208.lcssa = phi i32 [ %port.sroa.37.0.i.i1208, %Flow3498 ]
  %.lcssa3940 = phi i32 [ %1597, %Flow3498 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa3941)
  %1625 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %.lcssa3940)
  %1626 = icmp ne i32 %1625, %port.sroa.37.0.i.i1208.lcssa
  %1627 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %1626)
  %1628 = extractvalue { i1, i64 } %1627, 0
  %1629 = extractvalue { i1, i64 } %1627, 1
  br i1 %1628, label %while.body.i.i.i.i1230.preheader, label %Flow3487

while.body.i.i.i.i1230.preheader:                 ; preds = %cond.false.i.i.i1229
  br label %while.body.i.i.i.i1230, !amdgpu.uniform !42

Flow3487:                                         ; preds = %Flow3485, %cond.false.i.i.i1229
  call void @llvm.amdgcn.end.cf.i64(i64 %1629)
  br label %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit1251, !amdgpu.uniform !42

while.body.i.i.i.i1230:                           ; preds = %while.body.i.i.i.i1230.preheader, %while.body.i.i.i.i1230
  %phi.broken3820 = phi i64 [ 0, %while.body.i.i.i.i1230.preheader ], [ %1633, %while.body.i.i.i.i1230 ]
  tail call void @llvm.amdgcn.s.sleep(i32 2)
  %1630 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i8.i.i1231 = getelementptr inbounds nuw i32, ptr %1630, i64 %idxprom.i23.i.i.i.i.i1204, !amdgpu.uniform !42
  %1631 = load atomic i32, ptr %arrayidx.i.i.i8.i.i1231 monotonic, align 4
  %1632 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %1631)
  %cmp.i.not.i.i.i.i1232 = icmp eq i32 %1632, %port.sroa.37.0.i.i1208.lcssa
  %1633 = call i64 @llvm.amdgcn.if.break.i64(i1 %cmp.i.not.i.i.i.i1232, i64 %phi.broken3820)
  %1634 = call i1 @llvm.amdgcn.loop.i64(i64 %1633)
  br i1 %1634, label %Flow3485, label %while.body.i.i.i.i1230

Flow3485:                                         ; preds = %while.body.i.i.i.i1230
  %.lcssa3939 = phi i64 [ %1633, %while.body.i.i.i.i1230 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa3939)
  br label %Flow3487, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit1251: ; preds = %Flow3487
  fence acquire
  tail call void @llvm.amdgcn.wave.barrier()
  fence syncscope("agent") release
  %1635 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %.lcssa3946, i1 true)
  %iszero.i.i.i.i.i13.i.i1234 = icmp ne i64 %.lcssa3946, 0
  %cmp2.i.i.i.i14.i.i1235 = icmp eq i64 %1635, %sh_prom.i.i.i.i
  %cmp.i.i.i.i15.i.i1236 = select i1 %iszero.i.i.i.i.i13.i.i1234, i1 %cmp2.i.i.i.i14.i.i1235, i1 false
  %1636 = and i32 %.lcssa3947, 31
  %conv4.i.i.i19.i.i1240 = zext i1 %cmp.i.i.i.i15.i.i1236 to i32
  %shl.i.i.i20.i.i1241 = shl nuw i32 %conv4.i.i.i19.i.i1240, %1636
  %xor.i.i.i21.i.i1242 = xor i32 %shl.i.i.i20.i.i1241, -1
  %1637 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2576, !amdgpu.uniform !42

1638:                                             ; preds = %ComputeEnd2577
  %div8.i.i.i16.i.i1237 = lshr i32 %.lcssa3947, 5
  %idxprom.i.i.i17.i.i1238 = zext nneg i32 %div8.i.i.i16.i.i1237 to i64
  %1639 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i17.i.i1238
  %sunkaddr3275 = getelementptr inbounds i8, ptr addrspace(1) %1639, i64 40
  %1640 = atomicrmw and ptr addrspace(1) %sunkaddr3275, i32 %.lcssa3938 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %1641, !amdgpu.uniform !42

1641:                                             ; preds = %1638, %ComputeEnd2577
  call void @llvm.amdgcn.end.cf.i64(i64 %3445)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %for.cond.i.i.i1253, !amdgpu.uniform !42

for.cond.i.i.i1253:                               ; preds = %1641, %Flow3484
  %1642 = phi i32 [ %1680, %Flow3484 ], [ poison, %1641 ]
  %1643 = phi i64 [ %1681, %Flow3484 ], [ poison, %1641 ]
  %1644 = phi i32 [ %1682, %Flow3484 ], [ poison, %1641 ]
  %retval.sroa.6.0.i.i.i1254 = phi i32 [ %1685, %Flow3484 ], [ undef, %1641 ]
  %retval.sroa.2.0.i.i.i1256 = phi i64 [ %1684, %Flow3484 ], [ undef, %1641 ]
  %retval.sroa.8.0.i.i.i1257 = phi i32 [ %1683, %Flow3484 ], [ undef, %1641 ]
  %index.0.i.i.i1258 = phi i32 [ %1687, %Flow3484 ], [ 0, %1641 ]
  %1645 = load i32, ptr addrspace(1) @__llvm_rpc_client, align 8, !tbaa !68
  %cmp.not.i.i.i1259 = icmp ult i32 %index.0.i.i.i1258, %1645
  %spec.store.select.i.i.i1260 = select i1 %cmp.not.i.i.i1259, i32 %index.0.i.i.i1258, i32 0
  %1646 = tail call noundef i64 @llvm.amdgcn.ballot.i64(i1 true)
  %1647 = and i32 %spec.store.select.i.i.i1260, 31
  %1648 = lshr i64 %1646, %sh_prom.i.i.i.i
  %1649 = trunc i64 %1648 to i32
  %conv4.i.i.i.i.i1264 = and i32 %1649, 1
  %shl.i.i.i.i.i1265 = shl nuw i32 %conv4.i.i.i.i.i1264, %1647
  %1650 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2580, !amdgpu.uniform !42

1651:                                             ; preds = %ComputeEnd2581
  %sunkaddr3276 = getelementptr inbounds i8, ptr addrspace(1) %3455, i64 40
  %1652 = atomicrmw or ptr addrspace(1) %sunkaddr3276, i32 %.lcssa3932 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %1653, !amdgpu.uniform !42

1653:                                             ; preds = %1651, %ComputeEnd2581
  %1654 = phi i32 [ %1652, %1651 ], [ poison, %ComputeEnd2581 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %3464)
  %1655 = call i32 @llvm.amdgcn.readfirstlane.i32(i32 %1654)
  %1656 = or i32 %1655, %.lcssa3933
  %shl5.i.i.i.i.i1266 = shl nuw i32 1, %1647
  %and.i.i.i.i.i1267 = and i32 %shl5.i.i.i.i.i1266, %1656
  %tobool3.i.i.i.i1268 = icmp ne i32 %and.i.i.i.i.i1267, 0
  %1657 = tail call i64 @llvm.amdgcn.ballot.i64(i1 %tobool3.i.i.i.i1268)
  %and.i.i.i.i.i.i1269 = and i64 %1657, %1646
  %cmp.i.not.i.i.i1270 = icmp ne i64 %1646, %and.i.i.i.i.i.i1269
  br i1 %cmp.i.not.i.i.i1270, label %if.then.i.i.i.i1271, label %Flow3483, !amdgpu.uniform !42

if.then.i.i.i.i1271:                              ; preds = %1653
  fence syncscope("agent") acquire
  %1658 = load <2 x i64>, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !75
  %1659 = extractelement <2 x i64> %1658, i32 0
  %1660 = inttoptr i64 %1659 to ptr
  %1661 = extractelement <2 x i64> %1658, i32 1
  %1662 = inttoptr i64 %1661 to ptr
  %idxprom.i.i.i.i1272 = zext i32 %spec.store.select.i.i.i1260 to i64
  %arrayidx.i.i.i.i1273 = getelementptr inbounds nuw i32, ptr %1660, i64 %idxprom.i.i.i.i1272, !amdgpu.uniform !42
  %1663 = load atomic i32, ptr %arrayidx.i.i.i.i1273 monotonic, align 4
  %1664 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %1663)
  %arrayidx.i45.i.i.i1274 = getelementptr inbounds nuw i32, ptr %1662, i64 %idxprom.i.i.i.i1272, !amdgpu.uniform !42
  %1665 = load atomic i32, ptr %arrayidx.i45.i.i.i1274 monotonic, align 4
  %1666 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %1665)
  %cmp.i46.not.i.i.i1275 = icmp ne i32 %1664, %1666
  br i1 %cmp.i46.not.i.i.i1275, label %if.then12.i.i.i1276, label %Flow3481, !amdgpu.uniform !42

if.then12.i.i.i1276:                              ; preds = %if.then.i.i.i.i1271
  fence syncscope("agent") release
  %1667 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %1646, i1 true)
  %iszero.i.i.i.i.i.i.i1277 = icmp ne i64 %1646, 0
  %cmp2.i.i.i.i.i.i1278 = icmp eq i64 %1667, %sh_prom.i.i.i.i
  %cmp.i.i.i.i.i.i1279 = select i1 %iszero.i.i.i.i.i.i.i1277, i1 %cmp2.i.i.i.i.i.i1278, i1 false
  %conv4.i.i51.i.i.i1280 = zext i1 %cmp.i.i.i.i.i.i1279 to i32
  %shl.i.i52.i.i.i1281 = shl nuw i32 %conv4.i.i51.i.i.i1280, %1647
  %xor.i.i.i.i.i1282 = xor i32 %shl.i.i52.i.i.i1281, -1
  %1668 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2585, !amdgpu.uniform !42

1669:                                             ; preds = %ComputeEnd2586
  %sunkaddr3277 = getelementptr inbounds i8, ptr addrspace(1) %3455, i64 40
  %1670 = atomicrmw and ptr addrspace(1) %sunkaddr3277, i32 %.lcssa3934 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %1671, !amdgpu.uniform !42

1671:                                             ; preds = %1669, %ComputeEnd2586
  call void @llvm.amdgcn.end.cf.i64(i64 %3482)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %Flow3481, !amdgpu.uniform !42

if.end14.i.i.i1342:                               ; preds = %Flow3481
  %1672 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %1646, i1 true)
  %iszero.i.i.i.i.i.i1343 = icmp ne i64 %1646, 0
  %cmp2.i.i.i.i.i1344 = icmp eq i64 %1672, %sh_prom.i.i.i.i
  %cmp.i.i.i.i.i1345 = select i1 %iszero.i.i.i.i.i.i1343, i1 %cmp2.i.i.i.i.i1344, i1 false
  %1673 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.i.i.i.i.i1345)
  %1674 = extractvalue { i1, i64 } %1673, 0
  %1675 = extractvalue { i1, i64 } %1673, 1
  br i1 %1674, label %if.then16.i.i.i1347, label %if.end22.i.i.i1346

if.then16.i.i.i1347:                              ; preds = %if.end14.i.i.i1342
  %1676 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !76
  %opcode.i.i.i1348 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %1676, i64 %idxprom.i.i.i.i1272, i32 1
  store i32 1660944387, ptr %opcode.i.i.i1348, align 8, !tbaa !77
  %arrayidx21.i.i.i1349 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %1676, i64 %idxprom.i.i.i.i1272
  store i64 %1646, ptr %arrayidx21.i.i.i1349, align 8, !tbaa !80
  br label %if.end22.i.i.i1346, !amdgpu.uniform !42

Flow3482:                                         ; preds = %if.end22.i.i.i1346, %Flow3481
  %1677 = phi i32 [ %1664, %if.end22.i.i.i1346 ], [ %retval.sroa.8.0.i.i.i1257, %Flow3481 ]
  %1678 = phi i64 [ %1646, %if.end22.i.i.i1346 ], [ %retval.sroa.2.0.i.i.i1256, %Flow3481 ]
  %1679 = phi i32 [ %spec.store.select.i.i.i1260, %if.end22.i.i.i1346 ], [ %retval.sroa.6.0.i.i.i1254, %Flow3481 ]
  br label %cleanup26.i.i.i1283, !amdgpu.uniform !42

if.end22.i.i.i1346:                               ; preds = %if.then16.i.i.i1347, %if.end14.i.i.i1342
  call void @llvm.amdgcn.end.cf.i64(i64 %1675)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %Flow3482, !amdgpu.uniform !42

Flow3483:                                         ; preds = %cleanup26.i.i.i1283, %1653
  %1680 = phi i32 [ %1679, %cleanup26.i.i.i1283 ], [ %1642, %1653 ]
  %1681 = phi i64 [ %1678, %cleanup26.i.i.i1283 ], [ %1643, %1653 ]
  %1682 = phi i32 [ %1677, %cleanup26.i.i.i1283 ], [ %1644, %1653 ]
  %1683 = phi i32 [ %1677, %cleanup26.i.i.i1283 ], [ %retval.sroa.8.0.i.i.i1257, %1653 ]
  %1684 = phi i64 [ %1678, %cleanup26.i.i.i1283 ], [ %retval.sroa.2.0.i.i.i1256, %1653 ]
  %1685 = phi i32 [ %1679, %cleanup26.i.i.i1283 ], [ %retval.sroa.6.0.i.i.i1254, %1653 ]
  %1686 = phi i1 [ %cmp.i46.not.i.i.i1275, %cleanup26.i.i.i1283 ], [ true, %1653 ]
  br i1 %1686, label %for.inc.i.i.i1288, label %Flow3484, !amdgpu.uniform !42

cleanup26.i.i.i1283:                              ; preds = %Flow3482
  br label %Flow3483, !amdgpu.uniform !42

for.inc.i.i.i1288:                                ; preds = %Flow3483
  %inc.i.i.i1293 = add i32 %spec.store.select.i.i.i1260, 1
  br label %Flow3484, !amdgpu.uniform !42

Flow3484:                                         ; preds = %for.inc.i.i.i1288, %Flow3483
  %1687 = phi i32 [ %inc.i.i.i1293, %for.inc.i.i.i1288 ], [ poison, %Flow3483 ]
  %1688 = phi i1 [ false, %for.inc.i.i.i1288 ], [ true, %Flow3483 ]
  br i1 %1688, label %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1294, label %for.cond.i.i.i1253, !amdgpu.uniform !42

_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1294: ; preds = %Flow3484
  %.lcssa3937 = phi i32 [ %1680, %Flow3484 ]
  %.lcssa3936 = phi i64 [ %1681, %Flow3484 ]
  %.lcssa3935 = phi i32 [ %1682, %Flow3484 ]
  fence acquire
  %1689 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %mul.i.i.i.i.i.i1295 = shl i32 %.lcssa3937, 6
  %idxprom.i19.i.i.i.i.i1296 = zext i32 %mul.i.i.i.i.i.i1295 to i64
  %arrayidx.i20.i.i.i.i.i1297 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %1689, i64 %idxprom.i19.i.i.i.i.i1296
  %arrayidx.i22.i.i.i.i.i1298 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %arrayidx.i20.i.i.i.i.i1297, i64 %sh_prom.i.i.i.i
  store i64 sub (i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) getelementptr inbounds nuw (i8, ptr addrspace(4) @.str.9, i64 8) to ptr) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.9 to ptr) to i64)), ptr %arrayidx.i22.i.i.i.i.i1298, align 8, !tbaa !82
  %spec.select.i.i.i.i.i.i.i1299 = tail call i64 @llvm.umin.i64(i64 sub (i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) getelementptr inbounds nuw (i8, ptr addrspace(4) @.str.9, i64 8) to ptr) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.9 to ptr) to i64)), i64 56)
  %1690 = and i64 %spec.select.i.i.i.i.i.i.i1299, 15
  %1691 = sub i64 %spec.select.i.i.i.i.i.i.i1299, %1690
  %1692 = icmp ne i64 %1691, 0
  br i1 %1692, label %loop-memcpy-expansion2387.preheader, label %Flow3480, !amdgpu.uniform !42

loop-memcpy-expansion2387.preheader:              ; preds = %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1294
  %1693 = add nuw nsw i64 %76, 8
  %scevgep2964 = getelementptr i8, ptr %1689, i64 %1693
  %1694 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i1296, 6
  %scevgep2965 = getelementptr i8, ptr %scevgep2964, i64 %1694
  br label %loop-memcpy-expansion2387, !amdgpu.uniform !42

Flow3480:                                         ; preds = %Flow3479, %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1294
  br label %loop-memcpy-residual-header2390, !amdgpu.uniform !42

loop-memcpy-expansion2387:                        ; preds = %loop-memcpy-expansion2387.preheader, %loop-memcpy-expansion2387
  %loop-index2388 = phi i64 [ %1697, %loop-memcpy-expansion2387 ], [ 0, %loop-memcpy-expansion2387.preheader ]
  %1695 = getelementptr i8, ptr addrspace(4) @.str.9, i64 %loop-index2388, !amdgpu.uniform !42
  %1696 = load <4 x i32>, ptr addrspace(4) %1695, align 1
  %scevgep2966 = getelementptr i8, ptr %scevgep2965, i64 %loop-index2388
  store <4 x i32> %1696, ptr %scevgep2966, align 1
  %1697 = add i64 %loop-index2388, 16
  %1698 = icmp uge i64 %1697, %1691
  br i1 %1698, label %Flow3479, label %loop-memcpy-expansion2387, !amdgpu.uniform !42

Flow3477:                                         ; preds = %Flow3474, %Flow3476
  %1699 = phi i64 [ %2729, %Flow3474 ], [ %2723, %Flow3476 ]
  br label %post-loop-memcpy-expansion2386, !amdgpu.uniform !42

loop-memcpy-residual2389:                         ; preds = %loop-memcpy-residual2389.preheader, %loop-memcpy-residual2389
  %residual-loop-index2391 = phi i64 [ %1701, %loop-memcpy-residual2389 ], [ 0, %loop-memcpy-residual2389.preheader ]
  %scevgep2963 = getelementptr i8, ptr addrspace(4) %scevgep2962, i64 %residual-loop-index2391, !amdgpu.uniform !42
  %1700 = load i8, ptr addrspace(4) %scevgep2963, align 1
  %scevgep2961 = getelementptr i8, ptr %scevgep2960, i64 %residual-loop-index2391
  store i8 %1700, ptr %scevgep2961, align 1
  %1701 = add i64 %residual-loop-index2391, 1
  %1702 = icmp uge i64 %1701, %1690
  br i1 %1702, label %Flow3474, label %loop-memcpy-residual2389, !amdgpu.uniform !42

Flow3474:                                         ; preds = %loop-memcpy-residual2389
  br label %Flow3477, !amdgpu.uniform !42

post-loop-memcpy-expansion2386:                   ; preds = %Flow3477
  %tobool.not.i.i.i.i.i.i1301 = icmp eq i32 %.lcssa3935, 0
  %conv.i.i.i.i.i.i1302 = zext i1 %tobool.not.i.i.i.i.i.i1301 to i32
  fence release
  %1703 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %idxprom.i23.i.i.i.i.i1303 = zext i32 %.lcssa3937 to i64
  %arrayidx.i24.i.i.i.i.i1304 = getelementptr inbounds nuw i32, ptr %1703, i64 %idxprom.i23.i.i.i.i.i1303
  store atomic i32 %conv.i.i.i.i.i.i1302, ptr %arrayidx.i24.i.i.i.i.i1304 monotonic, align 4
  %1704 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !84
  %arrayidx.i.i5.i.i1305 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %1704, i64 %idxprom.i23.i.i.i.i.i1303, !amdgpu.uniform !42
  %1705 = load i64, ptr %arrayidx.i.i5.i.i1305, align 8, !tbaa !80
  br label %while.cond.i.i.i.i1306, !amdgpu.uniform !42

while.cond.i.i.i.i1306:                           ; preds = %post-loop-memcpy-expansion2386, %Flow3472
  %phi.broken3821 = phi i64 [ 0, %post-loop-memcpy-expansion2386 ], [ %1722, %Flow3472 ]
  %lsr.iv2952 = phi i64 [ %1720, %Flow3472 ], [ sub (i64 add (i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) getelementptr inbounds nuw (i8, ptr addrspace(4) @.str.9, i64 8) to ptr) to i64), i64 -56), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.9 to ptr) to i64)), %post-loop-memcpy-expansion2386 ]
  %lsr.iv2947 = phi ptr addrspace(4) [ %1719, %Flow3472 ], [ getelementptr (i8, ptr addrspace(4) @.str.9, i64 56), %post-loop-memcpy-expansion2386 ]
  %port.sroa.37.0.i.i1307 = phi i32 [ %1718, %Flow3472 ], [ %conv.i.i.i.i.i.i1302, %post-loop-memcpy-expansion2386 ]
  %idx.0.i.i.i.i1308 = phi i64 [ %1717, %Flow3472 ], [ 56, %post-loop-memcpy-expansion2386 ]
  %cmp.i.i.i.i1309 = icmp ult i64 %idx.0.i.i.i.i1308, sub (i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) getelementptr inbounds nuw (i8, ptr addrspace(4) @.str.9, i64 8) to ptr) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.9 to ptr) to i64))
  %1706 = tail call i64 @llvm.amdgcn.ballot.i64(i1 %cmp.i.i.i.i1309)
  %and.i.i.i.i6.i.i1310 = and i64 %1706, %1705
  %tobool.not.i.i.i.i1311 = icmp ne i64 %and.i.i.i.i6.i.i1310, 0
  %1707 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i22.i.i.i1312 = getelementptr inbounds nuw i32, ptr %1707, i64 %idxprom.i23.i.i.i.i.i1303, !amdgpu.uniform !42
  %1708 = load atomic i32, ptr %arrayidx.i22.i.i.i1312 monotonic, align 4
  %1709 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %tobool.not.i.i.i.i1311)
  %1710 = extractvalue { i1, i64 } %1709, 0
  %1711 = extractvalue { i1, i64 } %1709, 1
  br i1 %1710, label %cond.false.i8.i.i.i.i1313, label %Flow3472

cond.false.i8.i.i.i.i1313:                        ; preds = %while.cond.i.i.i.i1306
  %1712 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %1708)
  %1713 = icmp ne i32 %1712, %port.sroa.37.0.i.i1307
  br i1 %1713, label %while.body.i.i21.i.i.i.i1314.preheader, label %Flow3471, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i1314.preheader:           ; preds = %cond.false.i8.i.i.i.i1313
  br label %while.body.i.i21.i.i.i.i1314, !amdgpu.uniform !42

Flow3471:                                         ; preds = %Flow3469, %cond.false.i8.i.i.i.i1313
  br label %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i1317, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i1314:                     ; preds = %while.body.i.i21.i.i.i.i1314.preheader, %while.body.i.i21.i.i.i.i1314
  tail call void @llvm.amdgcn.s.sleep(i32 2)
  %1714 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i22.i.i.i.i1315 = getelementptr inbounds nuw i32, ptr %1714, i64 %idxprom.i23.i.i.i.i.i1303, !amdgpu.uniform !42
  %1715 = load atomic i32, ptr %arrayidx.i.i.i22.i.i.i.i1315 monotonic, align 4
  %1716 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %1715)
  %cmp.i.not.i.i23.i.i.i.i1316 = icmp eq i32 %1716, %port.sroa.37.0.i.i1307
  br i1 %cmp.i.not.i.i23.i.i.i.i1316, label %Flow3469, label %while.body.i.i21.i.i.i.i1314, !amdgpu.uniform !42

Flow3469:                                         ; preds = %while.body.i.i21.i.i.i.i1314
  br label %Flow3471, !amdgpu.uniform !42

Flow3472:                                         ; preds = %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1318, %while.cond.i.i.i.i1306
  %1717 = phi i64 [ %add.i.i.i.i1321, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1318 ], [ poison, %while.cond.i.i.i.i1306 ]
  %1718 = phi i32 [ %conv.i.i26.i.i.i.i1319, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1318 ], [ poison, %while.cond.i.i.i.i1306 ]
  %1719 = phi ptr addrspace(4) [ %scevgep2948, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1318 ], [ poison, %while.cond.i.i.i.i1306 ]
  %1720 = phi i64 [ %lsr.iv.next2953, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1318 ], [ poison, %while.cond.i.i.i.i1306 ]
  %1721 = phi i1 [ false, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1318 ], [ true, %while.cond.i.i.i.i1306 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %1711)
  %1722 = call i64 @llvm.amdgcn.if.break.i64(i1 %1721, i64 %phi.broken3821)
  %1723 = call i1 @llvm.amdgcn.loop.i64(i64 %1722)
  br i1 %1723, label %cond.false.i.i.i1328, label %while.cond.i.i.i.i1306

_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i1317: ; preds = %Flow3471
  fence acquire
  br i1 %cmp.i.i.i.i1309, label %if.then.i.i.i.i.i.i.i1322, label %Flow3468, !amdgpu.uniform !42

if.then.i.i.i.i.i.i.i1322:                        ; preds = %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i1317
  %sub.i.i.i.i.i.i.i1324 = sub nuw i64 sub (i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) getelementptr inbounds nuw (i8, ptr addrspace(4) @.str.9, i64 8) to ptr) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.9 to ptr) to i64)), %idx.0.i.i.i.i1308
  %spec.select.i.i.i36.i.i.i.i1325 = tail call i64 @llvm.umin.i64(i64 %sub.i.i.i.i.i.i.i1324, i64 64)
  %1724 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %1725 = and i64 %spec.select.i.i.i36.i.i.i.i1325, 15
  %1726 = sub i64 %spec.select.i.i.i36.i.i.i.i1325, %1725
  %1727 = icmp ne i64 %1726, 0
  br i1 %1727, label %loop-memcpy-expansion2393.preheader, label %Flow3467, !amdgpu.uniform !42

loop-memcpy-expansion2393.preheader:              ; preds = %if.then.i.i.i.i.i.i.i1322
  %arrayidx.i20.i35.i.i.i.i1326 = getelementptr %"struct.rpc::Buffer", ptr %1724, i64 %idxprom.i19.i.i.i.i.i1296
  %arrayidx.i22.i38.i.i.i.i1327 = getelementptr %"struct.rpc::Buffer", ptr %arrayidx.i20.i35.i.i.i.i1326, i64 %sh_prom.i.i.i.i
  br label %loop-memcpy-expansion2393, !amdgpu.uniform !42

Flow3467:                                         ; preds = %Flow3466, %if.then.i.i.i.i.i.i.i1322
  br label %loop-memcpy-residual-header2396, !amdgpu.uniform !42

loop-memcpy-expansion2393:                        ; preds = %loop-memcpy-expansion2393.preheader, %loop-memcpy-expansion2393
  %loop-index2394 = phi i64 [ %1730, %loop-memcpy-expansion2393 ], [ 0, %loop-memcpy-expansion2393.preheader ]
  %scevgep2949 = getelementptr i8, ptr addrspace(4) %lsr.iv2947, i64 %loop-index2394, !amdgpu.uniform !42
  %1728 = load <4 x i32>, ptr addrspace(4) %scevgep2949, align 1
  %1729 = getelementptr i8, ptr %arrayidx.i22.i38.i.i.i.i1327, i64 %loop-index2394
  store <4 x i32> %1728, ptr %1729, align 1
  %1730 = add i64 %loop-index2394, 16
  %1731 = icmp uge i64 %1730, %1726
  br i1 %1731, label %Flow3466, label %loop-memcpy-expansion2393, !amdgpu.uniform !42

Flow3464:                                         ; preds = %Flow3463, %loop-memcpy-residual-header2396
  br label %Flow3468, !amdgpu.uniform !42

loop-memcpy-residual2395:                         ; preds = %loop-memcpy-residual2395.preheader, %loop-memcpy-residual2395
  %residual-loop-index2397 = phi i64 [ %1733, %loop-memcpy-residual2395 ], [ 0, %loop-memcpy-residual2395.preheader ]
  %scevgep2958 = getelementptr i8, ptr addrspace(4) %scevgep2957, i64 %residual-loop-index2397, !amdgpu.uniform !42
  %1732 = load i8, ptr addrspace(4) %scevgep2958, align 1
  %scevgep2956 = getelementptr i8, ptr %scevgep2955, i64 %residual-loop-index2397
  store i8 %1732, ptr %scevgep2956, align 1
  %1733 = add i64 %residual-loop-index2397, 1
  %1734 = icmp uge i64 %1733, %1725
  br i1 %1734, label %Flow3463, label %loop-memcpy-residual2395, !amdgpu.uniform !42

Flow3463:                                         ; preds = %loop-memcpy-residual2395
  br label %Flow3464, !amdgpu.uniform !42

_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1318: ; preds = %Flow3468
  %conv.i.i26.i.i.i.i1319 = xor i32 %port.sroa.37.0.i.i1307, 1
  fence release
  %1735 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %arrayidx.i24.i29.i.i.i.i1320 = getelementptr inbounds nuw i32, ptr %1735, i64 %idxprom.i23.i.i.i.i.i1303
  store atomic i32 %conv.i.i26.i.i.i.i1319, ptr %arrayidx.i24.i29.i.i.i.i1320 monotonic, align 4
  %add.i.i.i.i1321 = add i64 %idx.0.i.i.i.i1308, 64
  %scevgep2948 = getelementptr i8, ptr addrspace(4) %lsr.iv2947, i64 64
  %lsr.iv.next2953 = add i64 %lsr.iv2952, -64
  br label %Flow3472, !amdgpu.uniform !42

cond.false.i.i.i1328:                             ; preds = %Flow3472
  %.lcssa3931 = phi i64 [ %1722, %Flow3472 ]
  %port.sroa.37.0.i.i1307.lcssa = phi i32 [ %port.sroa.37.0.i.i1307, %Flow3472 ]
  %.lcssa3930 = phi i32 [ %1708, %Flow3472 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa3931)
  %1736 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %.lcssa3930)
  %1737 = icmp ne i32 %1736, %port.sroa.37.0.i.i1307.lcssa
  %1738 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %1737)
  %1739 = extractvalue { i1, i64 } %1738, 0
  %1740 = extractvalue { i1, i64 } %1738, 1
  br i1 %1739, label %while.body.i.i.i.i1329.preheader, label %Flow3461

while.body.i.i.i.i1329.preheader:                 ; preds = %cond.false.i.i.i1328
  br label %while.body.i.i.i.i1329, !amdgpu.uniform !42

Flow3461:                                         ; preds = %Flow3459, %cond.false.i.i.i1328
  call void @llvm.amdgcn.end.cf.i64(i64 %1740)
  br label %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsINS_3cpp11string_viewEEERS1_T_.exit, !amdgpu.uniform !42

while.body.i.i.i.i1329:                           ; preds = %while.body.i.i.i.i1329.preheader, %while.body.i.i.i.i1329
  %phi.broken3822 = phi i64 [ 0, %while.body.i.i.i.i1329.preheader ], [ %1744, %while.body.i.i.i.i1329 ]
  tail call void @llvm.amdgcn.s.sleep(i32 2)
  %1741 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i8.i.i1330 = getelementptr inbounds nuw i32, ptr %1741, i64 %idxprom.i23.i.i.i.i.i1303, !amdgpu.uniform !42
  %1742 = load atomic i32, ptr %arrayidx.i.i.i8.i.i1330 monotonic, align 4
  %1743 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %1742)
  %cmp.i.not.i.i.i.i1331 = icmp eq i32 %1743, %port.sroa.37.0.i.i1307.lcssa
  %1744 = call i64 @llvm.amdgcn.if.break.i64(i1 %cmp.i.not.i.i.i.i1331, i64 %phi.broken3822)
  %1745 = call i1 @llvm.amdgcn.loop.i64(i64 %1744)
  br i1 %1745, label %Flow3459, label %while.body.i.i.i.i1329

Flow3459:                                         ; preds = %while.body.i.i.i.i1329
  %.lcssa3929 = phi i64 [ %1744, %while.body.i.i.i.i1329 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa3929)
  br label %Flow3461, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsINS_3cpp11string_viewEEERS1_T_.exit: ; preds = %Flow3461
  fence acquire
  tail call void @llvm.amdgcn.wave.barrier()
  fence syncscope("agent") release
  %1746 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %.lcssa3936, i1 true)
  %iszero.i.i.i.i.i13.i.i1333 = icmp ne i64 %.lcssa3936, 0
  %cmp2.i.i.i.i14.i.i1334 = icmp eq i64 %1746, %sh_prom.i.i.i.i
  %cmp.i.i.i.i15.i.i1335 = select i1 %iszero.i.i.i.i.i13.i.i1333, i1 %cmp2.i.i.i.i14.i.i1334, i1 false
  %1747 = and i32 %.lcssa3937, 31
  %conv4.i.i.i19.i.i1339 = zext i1 %cmp.i.i.i.i15.i.i1335 to i32
  %shl.i.i.i20.i.i1340 = shl nuw i32 %conv4.i.i.i19.i.i1339, %1747
  %xor.i.i.i21.i.i1341 = xor i32 %shl.i.i.i20.i.i1340, -1
  %1748 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2589, !amdgpu.uniform !42

1749:                                             ; preds = %ComputeEnd2590
  %div8.i.i.i16.i.i1336 = lshr i32 %.lcssa3937, 5
  %idxprom.i.i.i17.i.i1337 = zext nneg i32 %div8.i.i.i16.i.i1336 to i64
  %1750 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i17.i.i1337
  %sunkaddr3278 = getelementptr inbounds i8, ptr addrspace(1) %1750, i64 40
  %1751 = atomicrmw and ptr addrspace(1) %sunkaddr3278, i32 %.lcssa3928 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %1752, !amdgpu.uniform !42

1752:                                             ; preds = %1749, %ComputeEnd2590
  call void @llvm.amdgcn.end.cf.i64(i64 %3499)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %for.cond.i.i1.i1358, !amdgpu.uniform !42

for.cond.i.i1.i1358:                              ; preds = %1752, %Flow3458
  %1753 = phi i32 [ %1791, %Flow3458 ], [ poison, %1752 ]
  %1754 = phi i64 [ %1792, %Flow3458 ], [ poison, %1752 ]
  %1755 = phi i32 [ %1793, %Flow3458 ], [ poison, %1752 ]
  %retval.sroa.6.0.i.i.i1359 = phi i32 [ %1796, %Flow3458 ], [ undef, %1752 ]
  %retval.sroa.2.0.i.i.i1361 = phi i64 [ %1795, %Flow3458 ], [ undef, %1752 ]
  %retval.sroa.8.0.i.i.i1362 = phi i32 [ %1794, %Flow3458 ], [ undef, %1752 ]
  %index.0.i.i.i1363 = phi i32 [ %1798, %Flow3458 ], [ 0, %1752 ]
  %1756 = load i32, ptr addrspace(1) @__llvm_rpc_client, align 8, !tbaa !68
  %cmp.not.i.i.i1364 = icmp ult i32 %index.0.i.i.i1363, %1756
  %spec.store.select.i.i.i1365 = select i1 %cmp.not.i.i.i1364, i32 %index.0.i.i.i1363, i32 0
  %1757 = tail call noundef i64 @llvm.amdgcn.ballot.i64(i1 true)
  %1758 = and i32 %spec.store.select.i.i.i1365, 31
  %1759 = lshr i64 %1757, %sh_prom.i.i.i.i
  %1760 = trunc i64 %1759 to i32
  %conv4.i.i.i.i.i1369 = and i32 %1760, 1
  %shl.i.i.i.i.i1370 = shl nuw i32 %conv4.i.i.i.i.i1369, %1758
  %1761 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2593, !amdgpu.uniform !42

1762:                                             ; preds = %ComputeEnd2594
  %sunkaddr3279 = getelementptr inbounds i8, ptr addrspace(1) %3509, i64 40
  %1763 = atomicrmw or ptr addrspace(1) %sunkaddr3279, i32 %.lcssa3922 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %1764, !amdgpu.uniform !42

1764:                                             ; preds = %1762, %ComputeEnd2594
  %1765 = phi i32 [ %1763, %1762 ], [ poison, %ComputeEnd2594 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %3518)
  %1766 = call i32 @llvm.amdgcn.readfirstlane.i32(i32 %1765)
  %1767 = or i32 %1766, %.lcssa3923
  %shl5.i.i.i.i.i1371 = shl nuw i32 1, %1758
  %and.i.i.i.i.i1372 = and i32 %shl5.i.i.i.i.i1371, %1767
  %tobool3.i.i.i.i1373 = icmp ne i32 %and.i.i.i.i.i1372, 0
  %1768 = tail call i64 @llvm.amdgcn.ballot.i64(i1 %tobool3.i.i.i.i1373)
  %and.i.i.i.i.i.i1374 = and i64 %1768, %1757
  %cmp.i.not.i.i.i1375 = icmp ne i64 %1757, %and.i.i.i.i.i.i1374
  br i1 %cmp.i.not.i.i.i1375, label %if.then.i.i.i.i1376, label %Flow3457, !amdgpu.uniform !42

if.then.i.i.i.i1376:                              ; preds = %1764
  fence syncscope("agent") acquire
  %1769 = load <2 x i64>, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !75
  %1770 = extractelement <2 x i64> %1769, i32 0
  %1771 = inttoptr i64 %1770 to ptr
  %1772 = extractelement <2 x i64> %1769, i32 1
  %1773 = inttoptr i64 %1772 to ptr
  %idxprom.i.i.i.i1377 = zext i32 %spec.store.select.i.i.i1365 to i64
  %arrayidx.i.i.i.i1378 = getelementptr inbounds nuw i32, ptr %1771, i64 %idxprom.i.i.i.i1377, !amdgpu.uniform !42
  %1774 = load atomic i32, ptr %arrayidx.i.i.i.i1378 monotonic, align 4
  %1775 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %1774)
  %arrayidx.i45.i.i.i1379 = getelementptr inbounds nuw i32, ptr %1773, i64 %idxprom.i.i.i.i1377, !amdgpu.uniform !42
  %1776 = load atomic i32, ptr %arrayidx.i45.i.i.i1379 monotonic, align 4
  %1777 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %1776)
  %cmp.i46.not.i.i.i1380 = icmp ne i32 %1775, %1777
  br i1 %cmp.i46.not.i.i.i1380, label %if.then12.i.i.i1381, label %Flow3455, !amdgpu.uniform !42

if.then12.i.i.i1381:                              ; preds = %if.then.i.i.i.i1376
  fence syncscope("agent") release
  %1778 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %1757, i1 true)
  %iszero.i.i.i.i.i.i.i1382 = icmp ne i64 %1757, 0
  %cmp2.i.i.i.i.i.i1383 = icmp eq i64 %1778, %sh_prom.i.i.i.i
  %cmp.i.i.i.i.i.i1384 = select i1 %iszero.i.i.i.i.i.i.i1382, i1 %cmp2.i.i.i.i.i.i1383, i1 false
  %conv4.i.i51.i.i.i1385 = zext i1 %cmp.i.i.i.i.i.i1384 to i32
  %shl.i.i52.i.i.i1386 = shl nuw i32 %conv4.i.i51.i.i.i1385, %1758
  %xor.i.i.i.i.i1387 = xor i32 %shl.i.i52.i.i.i1386, -1
  %1779 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2598, !amdgpu.uniform !42

1780:                                             ; preds = %ComputeEnd2599
  %sunkaddr3280 = getelementptr inbounds i8, ptr addrspace(1) %3509, i64 40
  %1781 = atomicrmw and ptr addrspace(1) %sunkaddr3280, i32 %.lcssa3924 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %1782, !amdgpu.uniform !42

1782:                                             ; preds = %1780, %ComputeEnd2599
  call void @llvm.amdgcn.end.cf.i64(i64 %3536)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %Flow3455, !amdgpu.uniform !42

if.end14.i.i.i1447:                               ; preds = %Flow3455
  %1783 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %1757, i1 true)
  %iszero.i.i.i.i.i.i1448 = icmp ne i64 %1757, 0
  %cmp2.i.i.i.i.i1449 = icmp eq i64 %1783, %sh_prom.i.i.i.i
  %cmp.i.i.i.i.i1450 = select i1 %iszero.i.i.i.i.i.i1448, i1 %cmp2.i.i.i.i.i1449, i1 false
  %1784 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.i.i.i.i.i1450)
  %1785 = extractvalue { i1, i64 } %1784, 0
  %1786 = extractvalue { i1, i64 } %1784, 1
  br i1 %1785, label %if.then16.i.i.i1452, label %if.end22.i.i.i1451

if.then16.i.i.i1452:                              ; preds = %if.end14.i.i.i1447
  %1787 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !76
  %opcode.i.i.i1453 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %1787, i64 %idxprom.i.i.i.i1377, i32 1
  store i32 1660944387, ptr %opcode.i.i.i1453, align 8, !tbaa !77
  %arrayidx21.i.i.i1454 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %1787, i64 %idxprom.i.i.i.i1377
  store i64 %1757, ptr %arrayidx21.i.i.i1454, align 8, !tbaa !80
  br label %if.end22.i.i.i1451, !amdgpu.uniform !42

Flow3456:                                         ; preds = %if.end22.i.i.i1451, %Flow3455
  %1788 = phi i32 [ %1775, %if.end22.i.i.i1451 ], [ %retval.sroa.8.0.i.i.i1362, %Flow3455 ]
  %1789 = phi i64 [ %1757, %if.end22.i.i.i1451 ], [ %retval.sroa.2.0.i.i.i1361, %Flow3455 ]
  %1790 = phi i32 [ %spec.store.select.i.i.i1365, %if.end22.i.i.i1451 ], [ %retval.sroa.6.0.i.i.i1359, %Flow3455 ]
  br label %cleanup26.i.i.i1388, !amdgpu.uniform !42

if.end22.i.i.i1451:                               ; preds = %if.then16.i.i.i1452, %if.end14.i.i.i1447
  call void @llvm.amdgcn.end.cf.i64(i64 %1786)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %Flow3456, !amdgpu.uniform !42

Flow3457:                                         ; preds = %cleanup26.i.i.i1388, %1764
  %1791 = phi i32 [ %1790, %cleanup26.i.i.i1388 ], [ %1753, %1764 ]
  %1792 = phi i64 [ %1789, %cleanup26.i.i.i1388 ], [ %1754, %1764 ]
  %1793 = phi i32 [ %1788, %cleanup26.i.i.i1388 ], [ %1755, %1764 ]
  %1794 = phi i32 [ %1788, %cleanup26.i.i.i1388 ], [ %retval.sroa.8.0.i.i.i1362, %1764 ]
  %1795 = phi i64 [ %1789, %cleanup26.i.i.i1388 ], [ %retval.sroa.2.0.i.i.i1361, %1764 ]
  %1796 = phi i32 [ %1790, %cleanup26.i.i.i1388 ], [ %retval.sroa.6.0.i.i.i1359, %1764 ]
  %1797 = phi i1 [ %cmp.i46.not.i.i.i1380, %cleanup26.i.i.i1388 ], [ true, %1764 ]
  br i1 %1797, label %for.inc.i.i.i1393, label %Flow3458, !amdgpu.uniform !42

cleanup26.i.i.i1388:                              ; preds = %Flow3456
  br label %Flow3457, !amdgpu.uniform !42

for.inc.i.i.i1393:                                ; preds = %Flow3457
  %inc.i.i.i1398 = add i32 %spec.store.select.i.i.i1365, 1
  br label %Flow3458, !amdgpu.uniform !42

Flow3458:                                         ; preds = %for.inc.i.i.i1393, %Flow3457
  %1798 = phi i32 [ %inc.i.i.i1398, %for.inc.i.i.i1393 ], [ poison, %Flow3457 ]
  %1799 = phi i1 [ false, %for.inc.i.i.i1393 ], [ true, %Flow3457 ]
  br i1 %1799, label %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1399, label %for.cond.i.i1.i1358, !amdgpu.uniform !42

_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1399: ; preds = %Flow3458
  %.lcssa3927 = phi i32 [ %1791, %Flow3458 ]
  %.lcssa3926 = phi i64 [ %1792, %Flow3458 ]
  %.lcssa3925 = phi i32 [ %1793, %Flow3458 ]
  fence acquire
  %1800 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %mul.i.i.i.i.i.i1400 = shl i32 %.lcssa3927, 6
  %idxprom.i19.i.i.i.i.i1401 = zext i32 %mul.i.i.i.i.i.i1400 to i64
  %arrayidx.i20.i.i.i.i.i1402 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %1800, i64 %idxprom.i19.i.i.i.i.i1401
  %arrayidx.i22.i.i.i.i.i1403 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %arrayidx.i20.i.i.i.i.i1402, i64 %sh_prom.i.i.i.i
  store i64 sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.37.14 to ptr), i64 2) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.37.14 to ptr) to i64)), ptr %arrayidx.i22.i.i.i.i.i1403, align 8, !tbaa !82
  %spec.select.i.i.i.i.i.i.i1404 = tail call i64 @llvm.umin.i64(i64 sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.37.14 to ptr), i64 2) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.37.14 to ptr) to i64)), i64 56)
  %1801 = and i64 %spec.select.i.i.i.i.i.i.i1404, 15
  %1802 = sub i64 %spec.select.i.i.i.i.i.i.i1404, %1801
  %1803 = icmp ne i64 %1802, 0
  br i1 %1803, label %loop-memcpy-expansion2399.preheader, label %Flow3454, !amdgpu.uniform !42

loop-memcpy-expansion2399.preheader:              ; preds = %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1399
  %1804 = add nuw nsw i64 %76, 8
  %scevgep2944 = getelementptr i8, ptr %1800, i64 %1804
  %1805 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i1401, 6
  %scevgep2945 = getelementptr i8, ptr %scevgep2944, i64 %1805
  br label %loop-memcpy-expansion2399, !amdgpu.uniform !42

Flow3454:                                         ; preds = %Flow3453, %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1399
  br label %loop-memcpy-residual-header2402, !amdgpu.uniform !42

loop-memcpy-expansion2399:                        ; preds = %loop-memcpy-expansion2399.preheader, %loop-memcpy-expansion2399
  %loop-index2400 = phi i64 [ %1808, %loop-memcpy-expansion2399 ], [ 0, %loop-memcpy-expansion2399.preheader ]
  %1806 = getelementptr i8, ptr addrspace(4) @.str.37.14, i64 %loop-index2400, !amdgpu.uniform !42
  %1807 = load <4 x i32>, ptr addrspace(4) %1806, align 1
  %scevgep2946 = getelementptr i8, ptr %scevgep2945, i64 %loop-index2400
  store <4 x i32> %1807, ptr %scevgep2946, align 1
  %1808 = add i64 %loop-index2400, 16
  %1809 = icmp uge i64 %1808, %1802
  br i1 %1809, label %Flow3453, label %loop-memcpy-expansion2399, !amdgpu.uniform !42

Flow3451:                                         ; preds = %Flow3448, %Flow3450
  %1810 = phi i64 [ %2737, %Flow3448 ], [ %2734, %Flow3450 ]
  br label %post-loop-memcpy-expansion2398, !amdgpu.uniform !42

loop-memcpy-residual2401:                         ; preds = %loop-memcpy-residual2401.preheader, %loop-memcpy-residual2401
  %residual-loop-index2403 = phi i64 [ %1812, %loop-memcpy-residual2401 ], [ 0, %loop-memcpy-residual2401.preheader ]
  %scevgep2943 = getelementptr i8, ptr addrspace(4) @.str.37.14, i64 %residual-loop-index2403, !amdgpu.uniform !42
  %1811 = load i8, ptr addrspace(4) %scevgep2943, align 1
  %scevgep2942 = getelementptr i8, ptr %scevgep2941, i64 %residual-loop-index2403
  store i8 %1811, ptr %scevgep2942, align 1
  %1812 = add i64 %residual-loop-index2403, 1
  %1813 = icmp uge i64 %1812, %1801
  br i1 %1813, label %Flow3448, label %loop-memcpy-residual2401, !amdgpu.uniform !42

Flow3448:                                         ; preds = %loop-memcpy-residual2401
  br label %Flow3451, !amdgpu.uniform !42

post-loop-memcpy-expansion2398:                   ; preds = %Flow3451
  %tobool.not.i.i.i.i.i.i1406 = icmp eq i32 %.lcssa3925, 0
  %conv.i.i.i.i.i.i1407 = zext i1 %tobool.not.i.i.i.i.i.i1406 to i32
  fence release
  %1814 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %idxprom.i23.i.i.i.i.i1408 = zext i32 %.lcssa3927 to i64
  %arrayidx.i24.i.i.i.i.i1409 = getelementptr inbounds nuw i32, ptr %1814, i64 %idxprom.i23.i.i.i.i.i1408
  store atomic i32 %conv.i.i.i.i.i.i1407, ptr %arrayidx.i24.i.i.i.i.i1409 monotonic, align 4
  %1815 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !84
  %arrayidx.i.i5.i.i1410 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %1815, i64 %idxprom.i23.i.i.i.i.i1408, !amdgpu.uniform !42
  %1816 = load i64, ptr %arrayidx.i.i5.i.i1410, align 8, !tbaa !80
  br label %while.cond.i.i.i.i1411, !amdgpu.uniform !42

while.cond.i.i.i.i1411:                           ; preds = %post-loop-memcpy-expansion2398, %Flow3446
  %phi.broken3823 = phi i64 [ 0, %post-loop-memcpy-expansion2398 ], [ %1833, %Flow3446 ]
  %lsr.iv2933 = phi i64 [ %1831, %Flow3446 ], [ -54, %post-loop-memcpy-expansion2398 ]
  %lsr.iv2928 = phi ptr addrspace(4) [ %1830, %Flow3446 ], [ getelementptr (i8, ptr addrspace(4) @.str.37.14, i64 56), %post-loop-memcpy-expansion2398 ]
  %port.sroa.37.0.i.i1412 = phi i32 [ %1829, %Flow3446 ], [ %conv.i.i.i.i.i.i1407, %post-loop-memcpy-expansion2398 ]
  %idx.0.i.i.i.i1413 = phi i64 [ %1828, %Flow3446 ], [ 56, %post-loop-memcpy-expansion2398 ]
  %cmp.i.i.i.i1414 = icmp ult i64 %idx.0.i.i.i.i1413, sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.37.14 to ptr), i64 2) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.37.14 to ptr) to i64))
  %1817 = tail call i64 @llvm.amdgcn.ballot.i64(i1 %cmp.i.i.i.i1414)
  %and.i.i.i.i6.i.i1415 = and i64 %1817, %1816
  %tobool.not.i.i.i.i1416 = icmp ne i64 %and.i.i.i.i6.i.i1415, 0
  %1818 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i22.i.i.i1417 = getelementptr inbounds nuw i32, ptr %1818, i64 %idxprom.i23.i.i.i.i.i1408, !amdgpu.uniform !42
  %1819 = load atomic i32, ptr %arrayidx.i22.i.i.i1417 monotonic, align 4
  %1820 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %tobool.not.i.i.i.i1416)
  %1821 = extractvalue { i1, i64 } %1820, 0
  %1822 = extractvalue { i1, i64 } %1820, 1
  br i1 %1821, label %cond.false.i8.i.i.i.i1418, label %Flow3446

cond.false.i8.i.i.i.i1418:                        ; preds = %while.cond.i.i.i.i1411
  %1823 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %1819)
  %1824 = icmp ne i32 %1823, %port.sroa.37.0.i.i1412
  br i1 %1824, label %while.body.i.i21.i.i.i.i1419.preheader, label %Flow3445, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i1419.preheader:           ; preds = %cond.false.i8.i.i.i.i1418
  br label %while.body.i.i21.i.i.i.i1419, !amdgpu.uniform !42

Flow3445:                                         ; preds = %Flow3443, %cond.false.i8.i.i.i.i1418
  br label %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i1422, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i1419:                     ; preds = %while.body.i.i21.i.i.i.i1419.preheader, %while.body.i.i21.i.i.i.i1419
  tail call void @llvm.amdgcn.s.sleep(i32 2)
  %1825 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i22.i.i.i.i1420 = getelementptr inbounds nuw i32, ptr %1825, i64 %idxprom.i23.i.i.i.i.i1408, !amdgpu.uniform !42
  %1826 = load atomic i32, ptr %arrayidx.i.i.i22.i.i.i.i1420 monotonic, align 4
  %1827 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %1826)
  %cmp.i.not.i.i23.i.i.i.i1421 = icmp eq i32 %1827, %port.sroa.37.0.i.i1412
  br i1 %cmp.i.not.i.i23.i.i.i.i1421, label %Flow3443, label %while.body.i.i21.i.i.i.i1419, !amdgpu.uniform !42

Flow3443:                                         ; preds = %while.body.i.i21.i.i.i.i1419
  br label %Flow3445, !amdgpu.uniform !42

Flow3446:                                         ; preds = %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1423, %while.cond.i.i.i.i1411
  %1828 = phi i64 [ %add.i.i.i.i1426, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1423 ], [ poison, %while.cond.i.i.i.i1411 ]
  %1829 = phi i32 [ %conv.i.i26.i.i.i.i1424, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1423 ], [ poison, %while.cond.i.i.i.i1411 ]
  %1830 = phi ptr addrspace(4) [ %scevgep2929, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1423 ], [ poison, %while.cond.i.i.i.i1411 ]
  %1831 = phi i64 [ %lsr.iv.next2934, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1423 ], [ poison, %while.cond.i.i.i.i1411 ]
  %1832 = phi i1 [ false, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1423 ], [ true, %while.cond.i.i.i.i1411 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %1822)
  %1833 = call i64 @llvm.amdgcn.if.break.i64(i1 %1832, i64 %phi.broken3823)
  %1834 = call i1 @llvm.amdgcn.loop.i64(i64 %1833)
  br i1 %1834, label %cond.false.i.i.i1433, label %while.cond.i.i.i.i1411

_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i1422: ; preds = %Flow3445
  fence acquire
  br i1 %cmp.i.i.i.i1414, label %if.then.i.i.i.i.i.i.i1427, label %Flow3442, !amdgpu.uniform !42

if.then.i.i.i.i.i.i.i1427:                        ; preds = %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i1422
  %sub.i.i.i.i.i.i.i1429 = sub nuw i64 sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.37.14 to ptr), i64 2) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.37.14 to ptr) to i64)), %idx.0.i.i.i.i1413
  %spec.select.i.i.i36.i.i.i.i1430 = tail call i64 @llvm.umin.i64(i64 %sub.i.i.i.i.i.i.i1429, i64 64)
  %1835 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %1836 = and i64 %spec.select.i.i.i36.i.i.i.i1430, 15
  %1837 = sub i64 %spec.select.i.i.i36.i.i.i.i1430, %1836
  %1838 = icmp ne i64 %1837, 0
  br i1 %1838, label %loop-memcpy-expansion2405.preheader, label %Flow3441, !amdgpu.uniform !42

loop-memcpy-expansion2405.preheader:              ; preds = %if.then.i.i.i.i.i.i.i1427
  %arrayidx.i20.i35.i.i.i.i1431 = getelementptr %"struct.rpc::Buffer", ptr %1835, i64 %idxprom.i19.i.i.i.i.i1401
  %arrayidx.i22.i38.i.i.i.i1432 = getelementptr %"struct.rpc::Buffer", ptr %arrayidx.i20.i35.i.i.i.i1431, i64 %sh_prom.i.i.i.i
  br label %loop-memcpy-expansion2405, !amdgpu.uniform !42

Flow3441:                                         ; preds = %Flow3440, %if.then.i.i.i.i.i.i.i1427
  br label %loop-memcpy-residual-header2408, !amdgpu.uniform !42

loop-memcpy-expansion2405:                        ; preds = %loop-memcpy-expansion2405.preheader, %loop-memcpy-expansion2405
  %loop-index2406 = phi i64 [ %1841, %loop-memcpy-expansion2405 ], [ 0, %loop-memcpy-expansion2405.preheader ]
  %scevgep2930 = getelementptr i8, ptr addrspace(4) %lsr.iv2928, i64 %loop-index2406, !amdgpu.uniform !42
  %1839 = load <4 x i32>, ptr addrspace(4) %scevgep2930, align 1
  %1840 = getelementptr i8, ptr %arrayidx.i22.i38.i.i.i.i1432, i64 %loop-index2406
  store <4 x i32> %1839, ptr %1840, align 1
  %1841 = add i64 %loop-index2406, 16
  %1842 = icmp uge i64 %1841, %1837
  br i1 %1842, label %Flow3440, label %loop-memcpy-expansion2405, !amdgpu.uniform !42

Flow3438:                                         ; preds = %Flow3437, %loop-memcpy-residual-header2408
  br label %Flow3442, !amdgpu.uniform !42

loop-memcpy-residual2407:                         ; preds = %loop-memcpy-residual2407.preheader, %loop-memcpy-residual2407
  %residual-loop-index2409 = phi i64 [ %1844, %loop-memcpy-residual2407 ], [ 0, %loop-memcpy-residual2407.preheader ]
  %scevgep2939 = getelementptr i8, ptr addrspace(4) %scevgep2938, i64 %residual-loop-index2409, !amdgpu.uniform !42
  %1843 = load i8, ptr addrspace(4) %scevgep2939, align 1
  %scevgep2937 = getelementptr i8, ptr %scevgep2936, i64 %residual-loop-index2409
  store i8 %1843, ptr %scevgep2937, align 1
  %1844 = add i64 %residual-loop-index2409, 1
  %1845 = icmp uge i64 %1844, %1836
  br i1 %1845, label %Flow3437, label %loop-memcpy-residual2407, !amdgpu.uniform !42

Flow3437:                                         ; preds = %loop-memcpy-residual2407
  br label %Flow3438, !amdgpu.uniform !42

_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1423: ; preds = %Flow3442
  %conv.i.i26.i.i.i.i1424 = xor i32 %port.sroa.37.0.i.i1412, 1
  fence release
  %1846 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %arrayidx.i24.i29.i.i.i.i1425 = getelementptr inbounds nuw i32, ptr %1846, i64 %idxprom.i23.i.i.i.i.i1408
  store atomic i32 %conv.i.i26.i.i.i.i1424, ptr %arrayidx.i24.i29.i.i.i.i1425 monotonic, align 4
  %add.i.i.i.i1426 = add i64 %idx.0.i.i.i.i1413, 64
  %scevgep2929 = getelementptr i8, ptr addrspace(4) %lsr.iv2928, i64 64
  %lsr.iv.next2934 = add i64 %lsr.iv2933, -64
  br label %Flow3446, !amdgpu.uniform !42

cond.false.i.i.i1433:                             ; preds = %Flow3446
  %.lcssa3921 = phi i64 [ %1833, %Flow3446 ]
  %port.sroa.37.0.i.i1412.lcssa = phi i32 [ %port.sroa.37.0.i.i1412, %Flow3446 ]
  %.lcssa3920 = phi i32 [ %1819, %Flow3446 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa3921)
  %1847 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %.lcssa3920)
  %1848 = icmp ne i32 %1847, %port.sroa.37.0.i.i1412.lcssa
  %1849 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %1848)
  %1850 = extractvalue { i1, i64 } %1849, 0
  %1851 = extractvalue { i1, i64 } %1849, 1
  br i1 %1850, label %while.body.i.i.i.i1434.preheader, label %Flow3435

while.body.i.i.i.i1434.preheader:                 ; preds = %cond.false.i.i.i1433
  br label %while.body.i.i.i.i1434, !amdgpu.uniform !42

Flow3435:                                         ; preds = %Flow3433, %cond.false.i.i.i1433
  call void @llvm.amdgcn.end.cf.i64(i64 %1851)
  br label %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit1455, !amdgpu.uniform !42

while.body.i.i.i.i1434:                           ; preds = %while.body.i.i.i.i1434.preheader, %while.body.i.i.i.i1434
  %phi.broken3824 = phi i64 [ 0, %while.body.i.i.i.i1434.preheader ], [ %1855, %while.body.i.i.i.i1434 ]
  tail call void @llvm.amdgcn.s.sleep(i32 2)
  %1852 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i8.i.i1435 = getelementptr inbounds nuw i32, ptr %1852, i64 %idxprom.i23.i.i.i.i.i1408, !amdgpu.uniform !42
  %1853 = load atomic i32, ptr %arrayidx.i.i.i8.i.i1435 monotonic, align 4
  %1854 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %1853)
  %cmp.i.not.i.i.i.i1436 = icmp eq i32 %1854, %port.sroa.37.0.i.i1412.lcssa
  %1855 = call i64 @llvm.amdgcn.if.break.i64(i1 %cmp.i.not.i.i.i.i1436, i64 %phi.broken3824)
  %1856 = call i1 @llvm.amdgcn.loop.i64(i64 %1855)
  br i1 %1856, label %Flow3433, label %while.body.i.i.i.i1434

Flow3433:                                         ; preds = %while.body.i.i.i.i1434
  %.lcssa3919 = phi i64 [ %1855, %while.body.i.i.i.i1434 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa3919)
  br label %Flow3435, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit1455: ; preds = %Flow3435
  fence acquire
  tail call void @llvm.amdgcn.wave.barrier()
  fence syncscope("agent") release
  %1857 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %.lcssa3926, i1 true)
  %iszero.i.i.i.i.i13.i.i1438 = icmp ne i64 %.lcssa3926, 0
  %cmp2.i.i.i.i14.i.i1439 = icmp eq i64 %1857, %sh_prom.i.i.i.i
  %cmp.i.i.i.i15.i.i1440 = select i1 %iszero.i.i.i.i.i13.i.i1438, i1 %cmp2.i.i.i.i14.i.i1439, i1 false
  %1858 = and i32 %.lcssa3927, 31
  %conv4.i.i.i19.i.i1444 = zext i1 %cmp.i.i.i.i15.i.i1440 to i32
  %shl.i.i.i20.i.i1445 = shl nuw i32 %conv4.i.i.i19.i.i1444, %1858
  %xor.i.i.i21.i.i1446 = xor i32 %shl.i.i.i20.i.i1445, -1
  %1859 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2602, !amdgpu.uniform !42

1860:                                             ; preds = %ComputeEnd2603
  %div8.i.i.i16.i.i1441 = lshr i32 %.lcssa3927, 5
  %idxprom.i.i.i17.i.i1442 = zext nneg i32 %div8.i.i.i16.i.i1441 to i64
  %1861 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i17.i.i1442
  %sunkaddr3281 = getelementptr inbounds i8, ptr addrspace(1) %1861, i64 40
  %1862 = atomicrmw and ptr addrspace(1) %sunkaddr3281, i32 %.lcssa3918 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %1863, !amdgpu.uniform !42

1863:                                             ; preds = %1860, %ComputeEnd2603
  call void @llvm.amdgcn.end.cf.i64(i64 %3553)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %for.cond.i.i.i1456, !amdgpu.uniform !42

for.cond.i.i.i1456:                               ; preds = %1863, %for.cond.i.i.i1456
  %phi.broken3825 = phi i64 [ 0, %1863 ], [ %1866, %for.cond.i.i.i1456 ]
  %lsr.iv2914 = phi i64 [ %lsr.iv.next2915, %for.cond.i.i.i1456 ], [ -1, %1863 ]
  %scevgep2926 = getelementptr i8, ptr %RHSStr, i64 %lsr.iv2914
  %scevgep2927 = getelementptr i8, ptr %scevgep2926, i64 1
  %1864 = addrspacecast ptr %scevgep2927 to ptr addrspace(4)
  %1865 = load i8, ptr addrspace(4) %1864, align 1, !tbaa !57, !noalias.addrspace !101
  %cmp.i.i.i1458 = icmp eq i8 %1865, 0
  %lsr.iv.next2915 = add i64 %lsr.iv2914, 1
  %1866 = call i64 @llvm.amdgcn.if.break.i64(i1 %cmp.i.i.i1458, i64 %phi.broken3825)
  %1867 = call i1 @llvm.amdgcn.loop.i64(i64 %1866)
  br i1 %1867, label %for.cond.i.i1.i1465.preheader, label %for.cond.i.i.i1456

for.cond.i.i1.i1465.preheader:                    ; preds = %for.cond.i.i.i1456
  %lsr.iv.next2915.lcssa = phi i64 [ %lsr.iv.next2915, %for.cond.i.i.i1456 ]
  %.lcssa3917 = phi i64 [ %1866, %for.cond.i.i.i1456 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa3917)
  br label %for.cond.i.i1.i1465, !amdgpu.uniform !42

for.cond.i.i1.i1465:                              ; preds = %for.cond.i.i1.i1465.preheader, %Flow3432
  %1868 = phi i32 [ %1906, %Flow3432 ], [ poison, %for.cond.i.i1.i1465.preheader ]
  %1869 = phi i64 [ %1907, %Flow3432 ], [ poison, %for.cond.i.i1.i1465.preheader ]
  %1870 = phi i32 [ %1908, %Flow3432 ], [ poison, %for.cond.i.i1.i1465.preheader ]
  %retval.sroa.6.0.i.i.i1466 = phi i32 [ %1911, %Flow3432 ], [ undef, %for.cond.i.i1.i1465.preheader ]
  %retval.sroa.2.0.i.i.i1468 = phi i64 [ %1910, %Flow3432 ], [ undef, %for.cond.i.i1.i1465.preheader ]
  %retval.sroa.8.0.i.i.i1469 = phi i32 [ %1909, %Flow3432 ], [ undef, %for.cond.i.i1.i1465.preheader ]
  %index.0.i.i.i1470 = phi i32 [ %1913, %Flow3432 ], [ 0, %for.cond.i.i1.i1465.preheader ]
  %1871 = load i32, ptr addrspace(1) @__llvm_rpc_client, align 8, !tbaa !68
  %cmp.not.i.i.i1471 = icmp ult i32 %index.0.i.i.i1470, %1871
  %spec.store.select.i.i.i1472 = select i1 %cmp.not.i.i.i1471, i32 %index.0.i.i.i1470, i32 0
  %1872 = tail call noundef i64 @llvm.amdgcn.ballot.i64(i1 true)
  %1873 = and i32 %spec.store.select.i.i.i1472, 31
  %1874 = lshr i64 %1872, %sh_prom.i.i.i.i
  %1875 = trunc i64 %1874 to i32
  %conv4.i.i.i.i.i1476 = and i32 %1875, 1
  %shl.i.i.i.i.i1477 = shl nuw i32 %conv4.i.i.i.i.i1476, %1873
  %1876 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2606, !amdgpu.uniform !42

1877:                                             ; preds = %ComputeEnd2607
  %sunkaddr3282 = getelementptr inbounds i8, ptr addrspace(1) %3563, i64 40
  %1878 = atomicrmw or ptr addrspace(1) %sunkaddr3282, i32 %.lcssa3911 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %1879, !amdgpu.uniform !42

1879:                                             ; preds = %1877, %ComputeEnd2607
  %1880 = phi i32 [ %1878, %1877 ], [ poison, %ComputeEnd2607 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %3572)
  %1881 = call i32 @llvm.amdgcn.readfirstlane.i32(i32 %1880)
  %1882 = or i32 %1881, %.lcssa3912
  %shl5.i.i.i.i.i1478 = shl nuw i32 1, %1873
  %and.i.i.i.i.i1479 = and i32 %shl5.i.i.i.i.i1478, %1882
  %tobool3.i.i.i.i1480 = icmp ne i32 %and.i.i.i.i.i1479, 0
  %1883 = tail call i64 @llvm.amdgcn.ballot.i64(i1 %tobool3.i.i.i.i1480)
  %and.i.i.i.i.i.i1481 = and i64 %1883, %1872
  %cmp.i.not.i.i.i1482 = icmp ne i64 %1872, %and.i.i.i.i.i.i1481
  br i1 %cmp.i.not.i.i.i1482, label %if.then.i.i.i.i1483, label %Flow3431, !amdgpu.uniform !42

if.then.i.i.i.i1483:                              ; preds = %1879
  fence syncscope("agent") acquire
  %1884 = load <2 x i64>, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !75
  %1885 = extractelement <2 x i64> %1884, i32 0
  %1886 = inttoptr i64 %1885 to ptr
  %1887 = extractelement <2 x i64> %1884, i32 1
  %1888 = inttoptr i64 %1887 to ptr
  %idxprom.i.i.i.i1484 = zext i32 %spec.store.select.i.i.i1472 to i64
  %arrayidx.i.i.i.i1485 = getelementptr inbounds nuw i32, ptr %1886, i64 %idxprom.i.i.i.i1484, !amdgpu.uniform !42
  %1889 = load atomic i32, ptr %arrayidx.i.i.i.i1485 monotonic, align 4
  %1890 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %1889)
  %arrayidx.i45.i.i.i1486 = getelementptr inbounds nuw i32, ptr %1888, i64 %idxprom.i.i.i.i1484, !amdgpu.uniform !42
  %1891 = load atomic i32, ptr %arrayidx.i45.i.i.i1486 monotonic, align 4
  %1892 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %1891)
  %cmp.i46.not.i.i.i1487 = icmp ne i32 %1890, %1892
  br i1 %cmp.i46.not.i.i.i1487, label %if.then12.i.i.i1488, label %Flow3429, !amdgpu.uniform !42

if.then12.i.i.i1488:                              ; preds = %if.then.i.i.i.i1483
  fence syncscope("agent") release
  %1893 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %1872, i1 true)
  %iszero.i.i.i.i.i.i.i1489 = icmp ne i64 %1872, 0
  %cmp2.i.i.i.i.i.i1490 = icmp eq i64 %1893, %sh_prom.i.i.i.i
  %cmp.i.i.i.i.i.i1491 = select i1 %iszero.i.i.i.i.i.i.i1489, i1 %cmp2.i.i.i.i.i.i1490, i1 false
  %conv4.i.i51.i.i.i1492 = zext i1 %cmp.i.i.i.i.i.i1491 to i32
  %shl.i.i52.i.i.i1493 = shl nuw i32 %conv4.i.i51.i.i.i1492, %1873
  %xor.i.i.i.i.i1494 = xor i32 %shl.i.i52.i.i.i1493, -1
  %1894 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2611, !amdgpu.uniform !42

1895:                                             ; preds = %ComputeEnd2612
  %sunkaddr3283 = getelementptr inbounds i8, ptr addrspace(1) %3563, i64 40
  %1896 = atomicrmw and ptr addrspace(1) %sunkaddr3283, i32 %.lcssa3913 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %1897, !amdgpu.uniform !42

1897:                                             ; preds = %1895, %ComputeEnd2612
  call void @llvm.amdgcn.end.cf.i64(i64 %3590)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %Flow3429, !amdgpu.uniform !42

if.end14.i.i.i1554:                               ; preds = %Flow3429
  %1898 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %1872, i1 true)
  %iszero.i.i.i.i.i.i1555 = icmp ne i64 %1872, 0
  %cmp2.i.i.i.i.i1556 = icmp eq i64 %1898, %sh_prom.i.i.i.i
  %cmp.i.i.i.i.i1557 = select i1 %iszero.i.i.i.i.i.i1555, i1 %cmp2.i.i.i.i.i1556, i1 false
  %1899 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.i.i.i.i.i1557)
  %1900 = extractvalue { i1, i64 } %1899, 0
  %1901 = extractvalue { i1, i64 } %1899, 1
  br i1 %1900, label %if.then16.i.i.i1559, label %if.end22.i.i.i1558

if.then16.i.i.i1559:                              ; preds = %if.end14.i.i.i1554
  %1902 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !76
  %opcode.i.i.i1560 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %1902, i64 %idxprom.i.i.i.i1484, i32 1
  store i32 1660944387, ptr %opcode.i.i.i1560, align 8, !tbaa !77
  %arrayidx21.i.i.i1561 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %1902, i64 %idxprom.i.i.i.i1484
  store i64 %1872, ptr %arrayidx21.i.i.i1561, align 8, !tbaa !80
  br label %if.end22.i.i.i1558, !amdgpu.uniform !42

Flow3430:                                         ; preds = %if.end22.i.i.i1558, %Flow3429
  %1903 = phi i32 [ %1890, %if.end22.i.i.i1558 ], [ %retval.sroa.8.0.i.i.i1469, %Flow3429 ]
  %1904 = phi i64 [ %1872, %if.end22.i.i.i1558 ], [ %retval.sroa.2.0.i.i.i1468, %Flow3429 ]
  %1905 = phi i32 [ %spec.store.select.i.i.i1472, %if.end22.i.i.i1558 ], [ %retval.sroa.6.0.i.i.i1466, %Flow3429 ]
  br label %cleanup26.i.i.i1495, !amdgpu.uniform !42

if.end22.i.i.i1558:                               ; preds = %if.then16.i.i.i1559, %if.end14.i.i.i1554
  call void @llvm.amdgcn.end.cf.i64(i64 %1901)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %Flow3430, !amdgpu.uniform !42

Flow3431:                                         ; preds = %cleanup26.i.i.i1495, %1879
  %1906 = phi i32 [ %1905, %cleanup26.i.i.i1495 ], [ %1868, %1879 ]
  %1907 = phi i64 [ %1904, %cleanup26.i.i.i1495 ], [ %1869, %1879 ]
  %1908 = phi i32 [ %1903, %cleanup26.i.i.i1495 ], [ %1870, %1879 ]
  %1909 = phi i32 [ %1903, %cleanup26.i.i.i1495 ], [ %retval.sroa.8.0.i.i.i1469, %1879 ]
  %1910 = phi i64 [ %1904, %cleanup26.i.i.i1495 ], [ %retval.sroa.2.0.i.i.i1468, %1879 ]
  %1911 = phi i32 [ %1905, %cleanup26.i.i.i1495 ], [ %retval.sroa.6.0.i.i.i1466, %1879 ]
  %1912 = phi i1 [ %cmp.i46.not.i.i.i1487, %cleanup26.i.i.i1495 ], [ true, %1879 ]
  br i1 %1912, label %for.inc.i.i.i1500, label %Flow3432, !amdgpu.uniform !42

cleanup26.i.i.i1495:                              ; preds = %Flow3430
  br label %Flow3431, !amdgpu.uniform !42

for.inc.i.i.i1500:                                ; preds = %Flow3431
  %inc.i.i.i1505 = add i32 %spec.store.select.i.i.i1472, 1
  br label %Flow3432, !amdgpu.uniform !42

Flow3432:                                         ; preds = %for.inc.i.i.i1500, %Flow3431
  %1913 = phi i32 [ %inc.i.i.i1505, %for.inc.i.i.i1500 ], [ poison, %Flow3431 ]
  %1914 = phi i1 [ false, %for.inc.i.i.i1500 ], [ true, %Flow3431 ]
  br i1 %1914, label %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1506, label %for.cond.i.i1.i1465, !amdgpu.uniform !42

_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1506: ; preds = %Flow3432
  %.lcssa3916 = phi i32 [ %1906, %Flow3432 ]
  %.lcssa3915 = phi i64 [ %1907, %Flow3432 ]
  %.lcssa3914 = phi i32 [ %1908, %Flow3432 ]
  %scevgep2925 = getelementptr i8, ptr %RHSStr, i64 %lsr.iv.next2915.lcssa
  %sub.ptr.lhs.cast.i.i.i1461 = ptrtoint ptr %scevgep2925 to i64
  %sub.ptr.rhs.cast.i.i.i1462 = ptrtoint ptr %RHSStr to i64
  %sub.ptr.sub.i.i.i1463 = sub i64 %sub.ptr.lhs.cast.i.i.i1461, %sub.ptr.rhs.cast.i.i.i1462
  fence acquire
  %1915 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %mul.i.i.i.i.i.i1507 = shl i32 %.lcssa3916, 6
  %idxprom.i19.i.i.i.i.i1508 = zext i32 %mul.i.i.i.i.i.i1507 to i64
  %arrayidx.i20.i.i.i.i.i1509 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %1915, i64 %idxprom.i19.i.i.i.i.i1508
  %arrayidx.i22.i.i.i.i.i1510 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %arrayidx.i20.i.i.i.i.i1509, i64 %sh_prom.i.i.i.i
  store i64 %sub.ptr.sub.i.i.i1463, ptr %arrayidx.i22.i.i.i.i.i1510, align 8, !tbaa !82
  %spec.select.i.i.i.i.i.i.i1511 = tail call i64 @llvm.umin.i64(i64 %sub.ptr.sub.i.i.i1463, i64 56)
  %1916 = and i64 %spec.select.i.i.i.i.i.i.i1511, 15
  %1917 = sub i64 %spec.select.i.i.i.i.i.i.i1511, %1916
  %1918 = icmp ne i64 %1917, 0
  %1919 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %1918)
  %1920 = extractvalue { i1, i64 } %1919, 0
  %1921 = extractvalue { i1, i64 } %1919, 1
  br i1 %1920, label %loop-memcpy-expansion2207.preheader, label %Flow3428

loop-memcpy-expansion2207.preheader:              ; preds = %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1506
  %1922 = add nuw nsw i64 %76, 8
  %scevgep2922 = getelementptr i8, ptr %1915, i64 %1922
  %1923 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i1508, 6
  %scevgep2923 = getelementptr i8, ptr %scevgep2922, i64 %1923
  br label %loop-memcpy-expansion2207, !amdgpu.uniform !42

Flow3428:                                         ; preds = %Flow3427, %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1506
  call void @llvm.amdgcn.end.cf.i64(i64 %1921)
  br label %loop-memcpy-residual-header2210, !amdgpu.uniform !42

loop-memcpy-expansion2207:                        ; preds = %loop-memcpy-expansion2207.preheader, %loop-memcpy-expansion2207
  %phi.broken3826 = phi i64 [ 0, %loop-memcpy-expansion2207.preheader ], [ %1928, %loop-memcpy-expansion2207 ]
  %loop-index2208 = phi i64 [ %1926, %loop-memcpy-expansion2207 ], [ 0, %loop-memcpy-expansion2207.preheader ]
  %1924 = getelementptr i8, ptr %RHSStr, i64 %loop-index2208
  %1925 = load <4 x i32>, ptr %1924, align 1
  %scevgep2924 = getelementptr i8, ptr %scevgep2923, i64 %loop-index2208
  store <4 x i32> %1925, ptr %scevgep2924, align 1
  %1926 = add i64 %loop-index2208, 16
  %1927 = icmp uge i64 %1926, %1917
  %1928 = call i64 @llvm.amdgcn.if.break.i64(i1 %1927, i64 %phi.broken3826)
  %1929 = call i1 @llvm.amdgcn.loop.i64(i64 %1928)
  br i1 %1929, label %Flow3427, label %loop-memcpy-expansion2207

Flow3425:                                         ; preds = %Flow3422, %Flow3424
  %1930 = phi i64 [ %2582, %Flow3422 ], [ %2575, %Flow3424 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %2578)
  br label %post-loop-memcpy-expansion2206, !amdgpu.uniform !42

loop-memcpy-residual2209:                         ; preds = %loop-memcpy-residual2209.preheader, %loop-memcpy-residual2209
  %phi.broken3827 = phi i64 [ 0, %loop-memcpy-residual2209.preheader ], [ %1934, %loop-memcpy-residual2209 ]
  %residual-loop-index2211 = phi i64 [ %1932, %loop-memcpy-residual2209 ], [ 0, %loop-memcpy-residual2209.preheader ]
  %scevgep2921 = getelementptr i8, ptr %scevgep2920, i64 %residual-loop-index2211
  %1931 = load i8, ptr %scevgep2921, align 1
  %scevgep2919 = getelementptr i8, ptr %scevgep2918, i64 %residual-loop-index2211
  store i8 %1931, ptr %scevgep2919, align 1
  %1932 = add i64 %residual-loop-index2211, 1
  %1933 = icmp uge i64 %1932, %1916
  %1934 = call i64 @llvm.amdgcn.if.break.i64(i1 %1933, i64 %phi.broken3827)
  %1935 = call i1 @llvm.amdgcn.loop.i64(i64 %1934)
  br i1 %1935, label %Flow3422, label %loop-memcpy-residual2209

Flow3422:                                         ; preds = %loop-memcpy-residual2209
  %.lcssa3909 = phi i64 [ %1934, %loop-memcpy-residual2209 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa3909)
  br label %Flow3425, !amdgpu.uniform !42

post-loop-memcpy-expansion2206:                   ; preds = %Flow3425
  %1936 = add i64 %lsr.iv.next2915.lcssa, -56
  %tobool.not.i.i.i.i.i.i1513 = icmp eq i32 %.lcssa3914, 0
  %conv.i.i.i.i.i.i1514 = zext i1 %tobool.not.i.i.i.i.i.i1513 to i32
  fence release
  %1937 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %idxprom.i23.i.i.i.i.i1515 = zext i32 %.lcssa3916 to i64
  %arrayidx.i24.i.i.i.i.i1516 = getelementptr inbounds nuw i32, ptr %1937, i64 %idxprom.i23.i.i.i.i.i1515
  store atomic i32 %conv.i.i.i.i.i.i1514, ptr %arrayidx.i24.i.i.i.i.i1516 monotonic, align 4
  %1938 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !84
  %arrayidx.i.i5.i.i1517 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %1938, i64 %idxprom.i23.i.i.i.i.i1515, !amdgpu.uniform !42
  %1939 = load i64, ptr %arrayidx.i.i5.i.i1517, align 8, !tbaa !80
  %scevgep2897 = getelementptr i8, ptr %RHSStr, i64 56
  br label %while.cond.i.i.i.i1518, !amdgpu.uniform !42

while.cond.i.i.i.i1518:                           ; preds = %post-loop-memcpy-expansion2206, %Flow3420
  %phi.broken3830 = phi i64 [ 0, %post-loop-memcpy-expansion2206 ], [ %1956, %Flow3420 ]
  %lsr.iv2905 = phi i64 [ %1954, %Flow3420 ], [ %1936, %post-loop-memcpy-expansion2206 ]
  %lsr.iv2898 = phi ptr [ %1953, %Flow3420 ], [ %scevgep2897, %post-loop-memcpy-expansion2206 ]
  %port.sroa.37.0.i.i1519 = phi i32 [ %1952, %Flow3420 ], [ %conv.i.i.i.i.i.i1514, %post-loop-memcpy-expansion2206 ]
  %idx.0.i.i.i.i1520 = phi i64 [ %1951, %Flow3420 ], [ 56, %post-loop-memcpy-expansion2206 ]
  %cmp.i.i.i.i1521 = icmp ugt i64 %sub.ptr.sub.i.i.i1463, %idx.0.i.i.i.i1520
  %1940 = tail call i64 @llvm.amdgcn.ballot.i64(i1 %cmp.i.i.i.i1521)
  %and.i.i.i.i6.i.i1522 = and i64 %1940, %1939
  %tobool.not.i.i.i.i1523 = icmp ne i64 %and.i.i.i.i6.i.i1522, 0
  %1941 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i22.i.i.i1524 = getelementptr inbounds nuw i32, ptr %1941, i64 %idxprom.i23.i.i.i.i.i1515, !amdgpu.uniform !42
  %1942 = load atomic i32, ptr %arrayidx.i22.i.i.i1524 monotonic, align 4
  %1943 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %tobool.not.i.i.i.i1523)
  %1944 = extractvalue { i1, i64 } %1943, 0
  %1945 = extractvalue { i1, i64 } %1943, 1
  br i1 %1944, label %cond.false.i8.i.i.i.i1525, label %Flow3420

cond.false.i8.i.i.i.i1525:                        ; preds = %while.cond.i.i.i.i1518
  %1946 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %1942)
  %1947 = icmp ne i32 %1946, %port.sroa.37.0.i.i1519
  br i1 %1947, label %while.body.i.i21.i.i.i.i1526.preheader, label %Flow3419, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i1526.preheader:           ; preds = %cond.false.i8.i.i.i.i1525
  br label %while.body.i.i21.i.i.i.i1526, !amdgpu.uniform !42

Flow3419:                                         ; preds = %Flow3417, %cond.false.i8.i.i.i.i1525
  br label %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i1529, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i1526:                     ; preds = %while.body.i.i21.i.i.i.i1526.preheader, %while.body.i.i21.i.i.i.i1526
  tail call void @llvm.amdgcn.s.sleep(i32 2)
  %1948 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i22.i.i.i.i1527 = getelementptr inbounds nuw i32, ptr %1948, i64 %idxprom.i23.i.i.i.i.i1515, !amdgpu.uniform !42
  %1949 = load atomic i32, ptr %arrayidx.i.i.i22.i.i.i.i1527 monotonic, align 4
  %1950 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %1949)
  %cmp.i.not.i.i23.i.i.i.i1528 = icmp eq i32 %1950, %port.sroa.37.0.i.i1519
  br i1 %cmp.i.not.i.i23.i.i.i.i1528, label %Flow3417, label %while.body.i.i21.i.i.i.i1526, !amdgpu.uniform !42

Flow3417:                                         ; preds = %while.body.i.i21.i.i.i.i1526
  br label %Flow3419, !amdgpu.uniform !42

Flow3420:                                         ; preds = %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1530, %while.cond.i.i.i.i1518
  %1951 = phi i64 [ %add.i.i.i.i1533, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1530 ], [ poison, %while.cond.i.i.i.i1518 ]
  %1952 = phi i32 [ %conv.i.i26.i.i.i.i1531, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1530 ], [ poison, %while.cond.i.i.i.i1518 ]
  %1953 = phi ptr [ %scevgep2899, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1530 ], [ poison, %while.cond.i.i.i.i1518 ]
  %1954 = phi i64 [ %lsr.iv.next2906, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1530 ], [ poison, %while.cond.i.i.i.i1518 ]
  %1955 = phi i1 [ false, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1530 ], [ true, %while.cond.i.i.i.i1518 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %1945)
  %1956 = call i64 @llvm.amdgcn.if.break.i64(i1 %1955, i64 %phi.broken3830)
  %1957 = call i1 @llvm.amdgcn.loop.i64(i64 %1956)
  br i1 %1957, label %cond.false.i.i.i1540, label %while.cond.i.i.i.i1518

_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i1529: ; preds = %Flow3419
  fence acquire
  %1958 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.i.i.i.i1521)
  %1959 = extractvalue { i1, i64 } %1958, 0
  %1960 = extractvalue { i1, i64 } %1958, 1
  br i1 %1959, label %if.then.i.i.i.i.i.i.i1534, label %Flow3416

if.then.i.i.i.i.i.i.i1534:                        ; preds = %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i1529
  %sub.i.i.i.i.i.i.i1536 = sub nuw i64 %sub.ptr.sub.i.i.i1463, %idx.0.i.i.i.i1520
  %spec.select.i.i.i36.i.i.i.i1537 = tail call i64 @llvm.umin.i64(i64 %sub.i.i.i.i.i.i.i1536, i64 64)
  %1961 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %1962 = and i64 %spec.select.i.i.i36.i.i.i.i1537, 15
  %1963 = sub i64 %spec.select.i.i.i36.i.i.i.i1537, %1962
  %1964 = icmp ne i64 %1963, 0
  %1965 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %1964)
  %1966 = extractvalue { i1, i64 } %1965, 0
  %1967 = extractvalue { i1, i64 } %1965, 1
  br i1 %1966, label %loop-memcpy-expansion2201.preheader, label %Flow3415

loop-memcpy-expansion2201.preheader:              ; preds = %if.then.i.i.i.i.i.i.i1534
  %arrayidx.i20.i35.i.i.i.i1538 = getelementptr %"struct.rpc::Buffer", ptr %1961, i64 %idxprom.i19.i.i.i.i.i1508
  %arrayidx.i22.i38.i.i.i.i1539 = getelementptr %"struct.rpc::Buffer", ptr %arrayidx.i20.i35.i.i.i.i1538, i64 %sh_prom.i.i.i.i
  br label %loop-memcpy-expansion2201, !amdgpu.uniform !42

Flow3415:                                         ; preds = %Flow3414, %if.then.i.i.i.i.i.i.i1534
  call void @llvm.amdgcn.end.cf.i64(i64 %1967)
  br label %loop-memcpy-residual-header2204, !amdgpu.uniform !42

loop-memcpy-expansion2201:                        ; preds = %loop-memcpy-expansion2201.preheader, %loop-memcpy-expansion2201
  %phi.broken3828 = phi i64 [ 0, %loop-memcpy-expansion2201.preheader ], [ %1972, %loop-memcpy-expansion2201 ]
  %loop-index2202 = phi i64 [ %1970, %loop-memcpy-expansion2201 ], [ 0, %loop-memcpy-expansion2201.preheader ]
  %scevgep2900 = getelementptr i8, ptr %lsr.iv2898, i64 %loop-index2202
  %1968 = load <4 x i32>, ptr %scevgep2900, align 1
  %1969 = getelementptr i8, ptr %arrayidx.i22.i38.i.i.i.i1539, i64 %loop-index2202
  store <4 x i32> %1968, ptr %1969, align 1
  %1970 = add i64 %loop-index2202, 16
  %1971 = icmp uge i64 %1970, %1963
  %1972 = call i64 @llvm.amdgcn.if.break.i64(i1 %1971, i64 %phi.broken3828)
  %1973 = call i1 @llvm.amdgcn.loop.i64(i64 %1972)
  br i1 %1973, label %Flow3414, label %loop-memcpy-expansion2201

Flow3412:                                         ; preds = %Flow3411, %loop-memcpy-residual-header2204
  call void @llvm.amdgcn.end.cf.i64(i64 %2568)
  br label %Flow3416, !amdgpu.uniform !42

loop-memcpy-residual2203:                         ; preds = %loop-memcpy-residual2203.preheader, %loop-memcpy-residual2203
  %phi.broken3829 = phi i64 [ 0, %loop-memcpy-residual2203.preheader ], [ %1977, %loop-memcpy-residual2203 ]
  %residual-loop-index2205 = phi i64 [ %1975, %loop-memcpy-residual2203 ], [ 0, %loop-memcpy-residual2203.preheader ]
  %scevgep2912 = getelementptr i8, ptr %scevgep2911, i64 %residual-loop-index2205
  %1974 = load i8, ptr %scevgep2912, align 1
  %scevgep2909 = getelementptr i8, ptr %scevgep2908, i64 %residual-loop-index2205
  store i8 %1974, ptr %scevgep2909, align 1
  %1975 = add i64 %residual-loop-index2205, 1
  %1976 = icmp uge i64 %1975, %1962
  %1977 = call i64 @llvm.amdgcn.if.break.i64(i1 %1976, i64 %phi.broken3829)
  %1978 = call i1 @llvm.amdgcn.loop.i64(i64 %1977)
  br i1 %1978, label %Flow3411, label %loop-memcpy-residual2203

Flow3411:                                         ; preds = %loop-memcpy-residual2203
  %.lcssa3906 = phi i64 [ %1977, %loop-memcpy-residual2203 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa3906)
  br label %Flow3412, !amdgpu.uniform !42

_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1530: ; preds = %Flow3416
  %conv.i.i26.i.i.i.i1531 = xor i32 %port.sroa.37.0.i.i1519, 1
  fence release
  %1979 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %arrayidx.i24.i29.i.i.i.i1532 = getelementptr inbounds nuw i32, ptr %1979, i64 %idxprom.i23.i.i.i.i.i1515
  store atomic i32 %conv.i.i26.i.i.i.i1531, ptr %arrayidx.i24.i29.i.i.i.i1532 monotonic, align 4
  %add.i.i.i.i1533 = add i64 %idx.0.i.i.i.i1520, 64
  %scevgep2899 = getelementptr i8, ptr %lsr.iv2898, i64 64
  %lsr.iv.next2906 = add i64 %lsr.iv2905, -64
  br label %Flow3420, !amdgpu.uniform !42

cond.false.i.i.i1540:                             ; preds = %Flow3420
  %.lcssa3908 = phi i64 [ %1956, %Flow3420 ]
  %port.sroa.37.0.i.i1519.lcssa = phi i32 [ %port.sroa.37.0.i.i1519, %Flow3420 ]
  %.lcssa3907 = phi i32 [ %1942, %Flow3420 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa3908)
  %1980 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %.lcssa3907)
  %1981 = icmp ne i32 %1980, %port.sroa.37.0.i.i1519.lcssa
  %1982 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %1981)
  %1983 = extractvalue { i1, i64 } %1982, 0
  %1984 = extractvalue { i1, i64 } %1982, 1
  br i1 %1983, label %while.body.i.i.i.i1541.preheader, label %Flow3409

while.body.i.i.i.i1541.preheader:                 ; preds = %cond.false.i.i.i1540
  br label %while.body.i.i.i.i1541, !amdgpu.uniform !42

Flow3409:                                         ; preds = %Flow3407, %cond.false.i.i.i1540
  call void @llvm.amdgcn.end.cf.i64(i64 %1984)
  br label %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit1562, !amdgpu.uniform !42

while.body.i.i.i.i1541:                           ; preds = %while.body.i.i.i.i1541.preheader, %while.body.i.i.i.i1541
  %phi.broken3831 = phi i64 [ 0, %while.body.i.i.i.i1541.preheader ], [ %1988, %while.body.i.i.i.i1541 ]
  tail call void @llvm.amdgcn.s.sleep(i32 2)
  %1985 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i8.i.i1542 = getelementptr inbounds nuw i32, ptr %1985, i64 %idxprom.i23.i.i.i.i.i1515, !amdgpu.uniform !42
  %1986 = load atomic i32, ptr %arrayidx.i.i.i8.i.i1542 monotonic, align 4
  %1987 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %1986)
  %cmp.i.not.i.i.i.i1543 = icmp eq i32 %1987, %port.sroa.37.0.i.i1519.lcssa
  %1988 = call i64 @llvm.amdgcn.if.break.i64(i1 %cmp.i.not.i.i.i.i1543, i64 %phi.broken3831)
  %1989 = call i1 @llvm.amdgcn.loop.i64(i64 %1988)
  br i1 %1989, label %Flow3407, label %while.body.i.i.i.i1541

Flow3407:                                         ; preds = %while.body.i.i.i.i1541
  %.lcssa3904 = phi i64 [ %1988, %while.body.i.i.i.i1541 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa3904)
  br label %Flow3409, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit1562: ; preds = %Flow3409
  fence acquire
  tail call void @llvm.amdgcn.wave.barrier()
  fence syncscope("agent") release
  %1990 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %.lcssa3915, i1 true)
  %iszero.i.i.i.i.i13.i.i1545 = icmp ne i64 %.lcssa3915, 0
  %cmp2.i.i.i.i14.i.i1546 = icmp eq i64 %1990, %sh_prom.i.i.i.i
  %cmp.i.i.i.i15.i.i1547 = select i1 %iszero.i.i.i.i.i13.i.i1545, i1 %cmp2.i.i.i.i14.i.i1546, i1 false
  %1991 = and i32 %.lcssa3916, 31
  %conv4.i.i.i19.i.i1551 = zext i1 %cmp.i.i.i.i15.i.i1547 to i32
  %shl.i.i.i20.i.i1552 = shl nuw i32 %conv4.i.i.i19.i.i1551, %1991
  %xor.i.i.i21.i.i1553 = xor i32 %shl.i.i.i20.i.i1552, -1
  %1992 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2615, !amdgpu.uniform !42

1993:                                             ; preds = %ComputeEnd2616
  %div8.i.i.i16.i.i1548 = lshr i32 %.lcssa3916, 5
  %idxprom.i.i.i17.i.i1549 = zext nneg i32 %div8.i.i.i16.i.i1548 to i64
  %1994 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i17.i.i1549
  %sunkaddr3284 = getelementptr inbounds i8, ptr addrspace(1) %1994, i64 40
  %1995 = atomicrmw and ptr addrspace(1) %sunkaddr3284, i32 %.lcssa3903 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %1996, !amdgpu.uniform !42

1996:                                             ; preds = %1993, %ComputeEnd2616
  call void @llvm.amdgcn.end.cf.i64(i64 %3607)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %for.cond.i.i.i1566, !amdgpu.uniform !42

for.cond.i.i.i1566:                               ; preds = %1996, %Flow3406
  %1997 = phi i32 [ %2035, %Flow3406 ], [ poison, %1996 ]
  %1998 = phi i64 [ %2036, %Flow3406 ], [ poison, %1996 ]
  %1999 = phi i32 [ %2037, %Flow3406 ], [ poison, %1996 ]
  %retval.sroa.6.0.i.i.i1567 = phi i32 [ %2040, %Flow3406 ], [ undef, %1996 ]
  %retval.sroa.2.0.i.i.i1569 = phi i64 [ %2039, %Flow3406 ], [ undef, %1996 ]
  %retval.sroa.8.0.i.i.i1570 = phi i32 [ %2038, %Flow3406 ], [ undef, %1996 ]
  %index.0.i.i.i1571 = phi i32 [ %2042, %Flow3406 ], [ 0, %1996 ]
  %2000 = load i32, ptr addrspace(1) @__llvm_rpc_client, align 8, !tbaa !68
  %cmp.not.i.i.i1572 = icmp ult i32 %index.0.i.i.i1571, %2000
  %spec.store.select.i.i.i1573 = select i1 %cmp.not.i.i.i1572, i32 %index.0.i.i.i1571, i32 0
  %2001 = tail call noundef i64 @llvm.amdgcn.ballot.i64(i1 true)
  %2002 = and i32 %spec.store.select.i.i.i1573, 31
  %2003 = lshr i64 %2001, %sh_prom.i.i.i.i
  %2004 = trunc i64 %2003 to i32
  %conv4.i.i.i.i.i1577 = and i32 %2004, 1
  %shl.i.i.i.i.i1578 = shl nuw i32 %conv4.i.i.i.i.i1577, %2002
  %2005 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2619, !amdgpu.uniform !42

2006:                                             ; preds = %ComputeEnd2620
  %sunkaddr3285 = getelementptr inbounds i8, ptr addrspace(1) %3617, i64 40
  %2007 = atomicrmw or ptr addrspace(1) %sunkaddr3285, i32 %.lcssa3897 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %2008, !amdgpu.uniform !42

2008:                                             ; preds = %2006, %ComputeEnd2620
  %2009 = phi i32 [ %2007, %2006 ], [ poison, %ComputeEnd2620 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %3626)
  %2010 = call i32 @llvm.amdgcn.readfirstlane.i32(i32 %2009)
  %2011 = or i32 %2010, %.lcssa3898
  %shl5.i.i.i.i.i1579 = shl nuw i32 1, %2002
  %and.i.i.i.i.i1580 = and i32 %shl5.i.i.i.i.i1579, %2011
  %tobool3.i.i.i.i1581 = icmp ne i32 %and.i.i.i.i.i1580, 0
  %2012 = tail call i64 @llvm.amdgcn.ballot.i64(i1 %tobool3.i.i.i.i1581)
  %and.i.i.i.i.i.i1582 = and i64 %2012, %2001
  %cmp.i.not.i.i.i1583 = icmp ne i64 %2001, %and.i.i.i.i.i.i1582
  br i1 %cmp.i.not.i.i.i1583, label %if.then.i.i.i.i1584, label %Flow3405, !amdgpu.uniform !42

if.then.i.i.i.i1584:                              ; preds = %2008
  fence syncscope("agent") acquire
  %2013 = load <2 x i64>, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !75
  %2014 = extractelement <2 x i64> %2013, i32 0
  %2015 = inttoptr i64 %2014 to ptr
  %2016 = extractelement <2 x i64> %2013, i32 1
  %2017 = inttoptr i64 %2016 to ptr
  %idxprom.i.i.i.i1585 = zext i32 %spec.store.select.i.i.i1573 to i64
  %arrayidx.i.i.i.i1586 = getelementptr inbounds nuw i32, ptr %2015, i64 %idxprom.i.i.i.i1585, !amdgpu.uniform !42
  %2018 = load atomic i32, ptr %arrayidx.i.i.i.i1586 monotonic, align 4
  %2019 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %2018)
  %arrayidx.i45.i.i.i1587 = getelementptr inbounds nuw i32, ptr %2017, i64 %idxprom.i.i.i.i1585, !amdgpu.uniform !42
  %2020 = load atomic i32, ptr %arrayidx.i45.i.i.i1587 monotonic, align 4
  %2021 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %2020)
  %cmp.i46.not.i.i.i1588 = icmp ne i32 %2019, %2021
  br i1 %cmp.i46.not.i.i.i1588, label %if.then12.i.i.i1589, label %Flow3403, !amdgpu.uniform !42

if.then12.i.i.i1589:                              ; preds = %if.then.i.i.i.i1584
  fence syncscope("agent") release
  %2022 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %2001, i1 true)
  %iszero.i.i.i.i.i.i.i1590 = icmp ne i64 %2001, 0
  %cmp2.i.i.i.i.i.i1591 = icmp eq i64 %2022, %sh_prom.i.i.i.i
  %cmp.i.i.i.i.i.i1592 = select i1 %iszero.i.i.i.i.i.i.i1590, i1 %cmp2.i.i.i.i.i.i1591, i1 false
  %conv4.i.i51.i.i.i1593 = zext i1 %cmp.i.i.i.i.i.i1592 to i32
  %shl.i.i52.i.i.i1594 = shl nuw i32 %conv4.i.i51.i.i.i1593, %2002
  %xor.i.i.i.i.i1595 = xor i32 %shl.i.i52.i.i.i1594, -1
  %2023 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2624, !amdgpu.uniform !42

2024:                                             ; preds = %ComputeEnd2625
  %sunkaddr3286 = getelementptr inbounds i8, ptr addrspace(1) %3617, i64 40
  %2025 = atomicrmw and ptr addrspace(1) %sunkaddr3286, i32 %.lcssa3899 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %2026, !amdgpu.uniform !42

2026:                                             ; preds = %2024, %ComputeEnd2625
  call void @llvm.amdgcn.end.cf.i64(i64 %3644)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %Flow3403, !amdgpu.uniform !42

if.end14.i.i.i1653:                               ; preds = %Flow3403
  %2027 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %2001, i1 true)
  %iszero.i.i.i.i.i.i1654 = icmp ne i64 %2001, 0
  %cmp2.i.i.i.i.i1655 = icmp eq i64 %2027, %sh_prom.i.i.i.i
  %cmp.i.i.i.i.i1656 = select i1 %iszero.i.i.i.i.i.i1654, i1 %cmp2.i.i.i.i.i1655, i1 false
  %2028 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.i.i.i.i.i1656)
  %2029 = extractvalue { i1, i64 } %2028, 0
  %2030 = extractvalue { i1, i64 } %2028, 1
  br i1 %2029, label %if.then16.i.i.i1658, label %if.end22.i.i.i1657

if.then16.i.i.i1658:                              ; preds = %if.end14.i.i.i1653
  %2031 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !76
  %opcode.i.i.i1659 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %2031, i64 %idxprom.i.i.i.i1585, i32 1
  store i32 1660944387, ptr %opcode.i.i.i1659, align 8, !tbaa !77
  %arrayidx21.i.i.i1660 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %2031, i64 %idxprom.i.i.i.i1585
  store i64 %2001, ptr %arrayidx21.i.i.i1660, align 8, !tbaa !80
  br label %if.end22.i.i.i1657, !amdgpu.uniform !42

Flow3404:                                         ; preds = %if.end22.i.i.i1657, %Flow3403
  %2032 = phi i32 [ %2019, %if.end22.i.i.i1657 ], [ %retval.sroa.8.0.i.i.i1570, %Flow3403 ]
  %2033 = phi i64 [ %2001, %if.end22.i.i.i1657 ], [ %retval.sroa.2.0.i.i.i1569, %Flow3403 ]
  %2034 = phi i32 [ %spec.store.select.i.i.i1573, %if.end22.i.i.i1657 ], [ %retval.sroa.6.0.i.i.i1567, %Flow3403 ]
  br label %cleanup26.i.i.i1596, !amdgpu.uniform !42

if.end22.i.i.i1657:                               ; preds = %if.then16.i.i.i1658, %if.end14.i.i.i1653
  call void @llvm.amdgcn.end.cf.i64(i64 %2030)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %Flow3404, !amdgpu.uniform !42

Flow3405:                                         ; preds = %cleanup26.i.i.i1596, %2008
  %2035 = phi i32 [ %2034, %cleanup26.i.i.i1596 ], [ %1997, %2008 ]
  %2036 = phi i64 [ %2033, %cleanup26.i.i.i1596 ], [ %1998, %2008 ]
  %2037 = phi i32 [ %2032, %cleanup26.i.i.i1596 ], [ %1999, %2008 ]
  %2038 = phi i32 [ %2032, %cleanup26.i.i.i1596 ], [ %retval.sroa.8.0.i.i.i1570, %2008 ]
  %2039 = phi i64 [ %2033, %cleanup26.i.i.i1596 ], [ %retval.sroa.2.0.i.i.i1569, %2008 ]
  %2040 = phi i32 [ %2034, %cleanup26.i.i.i1596 ], [ %retval.sroa.6.0.i.i.i1567, %2008 ]
  %2041 = phi i1 [ %cmp.i46.not.i.i.i1588, %cleanup26.i.i.i1596 ], [ true, %2008 ]
  br i1 %2041, label %for.inc.i.i.i1601, label %Flow3406, !amdgpu.uniform !42

cleanup26.i.i.i1596:                              ; preds = %Flow3404
  br label %Flow3405, !amdgpu.uniform !42

for.inc.i.i.i1601:                                ; preds = %Flow3405
  %inc.i.i.i1606 = add i32 %spec.store.select.i.i.i1573, 1
  br label %Flow3406, !amdgpu.uniform !42

Flow3406:                                         ; preds = %for.inc.i.i.i1601, %Flow3405
  %2042 = phi i32 [ %inc.i.i.i1606, %for.inc.i.i.i1601 ], [ poison, %Flow3405 ]
  %2043 = phi i1 [ false, %for.inc.i.i.i1601 ], [ true, %Flow3405 ]
  br i1 %2043, label %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1607, label %for.cond.i.i.i1566, !amdgpu.uniform !42

_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1607: ; preds = %Flow3406
  %.lcssa3902 = phi i32 [ %2035, %Flow3406 ]
  %.lcssa3901 = phi i64 [ %2036, %Flow3406 ]
  %.lcssa3900 = phi i32 [ %2037, %Flow3406 ]
  fence acquire
  %2044 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %mul.i.i.i.i.i.i1608 = shl i32 %.lcssa3902, 6
  %idxprom.i19.i.i.i.i.i1609 = zext i32 %mul.i.i.i.i.i.i1608 to i64
  %arrayidx.i20.i.i.i.i.i1610 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %2044, i64 %idxprom.i19.i.i.i.i.i1609
  %arrayidx.i22.i.i.i.i.i1611 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %arrayidx.i20.i.i.i.i.i1610, i64 %sh_prom.i.i.i.i
  store i64 1, ptr %arrayidx.i22.i.i.i.i.i1611, align 8, !tbaa !82
  %arrayidx6.i.i.i.i.i.i.i1612 = getelementptr inbounds nuw i8, ptr %arrayidx.i22.i.i.i.i.i1611, i64 8
  store i8 10, ptr %arrayidx6.i.i.i.i.i.i.i1612, align 8
  %tobool.not.i.i.i.i.i.i1613 = icmp eq i32 %.lcssa3900, 0
  %conv.i.i.i.i.i.i1614 = zext i1 %tobool.not.i.i.i.i.i.i1613 to i32
  fence release
  %2045 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %idxprom.i23.i.i.i.i.i1615 = zext i32 %.lcssa3902 to i64
  %arrayidx.i24.i.i.i.i.i1616 = getelementptr inbounds nuw i32, ptr %2045, i64 %idxprom.i23.i.i.i.i.i1615
  store atomic i32 %conv.i.i.i.i.i.i1614, ptr %arrayidx.i24.i.i.i.i.i1616 monotonic, align 4
  %2046 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !84
  %arrayidx.i.i5.i.i1617 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %2046, i64 %idxprom.i23.i.i.i.i.i1615, !amdgpu.uniform !42
  %2047 = load i64, ptr %arrayidx.i.i5.i.i1617, align 8, !tbaa !80
  br label %while.cond.i.i.i.i1618, !amdgpu.uniform !42

while.cond.i.i.i.i1618:                           ; preds = %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1607, %Flow3402
  %phi.broken3832 = phi i64 [ 0, %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1607 ], [ %2062, %Flow3402 ]
  %port.sroa.37.0.i.i1619 = phi i32 [ %2060, %Flow3402 ], [ %conv.i.i.i.i.i.i1614, %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1607 ]
  %idx.0.i.i.i.i1620 = phi i64 [ %2059, %Flow3402 ], [ 56, %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1607 ]
  %cmp.i.i.i.i1621 = icmp eq i64 %idx.0.i.i.i.i1620, 0
  %2048 = tail call i64 @llvm.amdgcn.ballot.i64(i1 %cmp.i.i.i.i1621)
  %and.i.i.i.i6.i.i1622 = and i64 %2048, %2047
  %tobool.not.i.i.i.i1623 = icmp ne i64 %and.i.i.i.i6.i.i1622, 0
  %2049 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i22.i.i.i1624 = getelementptr inbounds nuw i32, ptr %2049, i64 %idxprom.i23.i.i.i.i.i1615, !amdgpu.uniform !42
  %2050 = load atomic i32, ptr %arrayidx.i22.i.i.i1624 monotonic, align 4
  %2051 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %tobool.not.i.i.i.i1623)
  %2052 = extractvalue { i1, i64 } %2051, 0
  %2053 = extractvalue { i1, i64 } %2051, 1
  br i1 %2052, label %cond.false.i8.i.i.i.i1625, label %Flow3402

cond.false.i8.i.i.i.i1625:                        ; preds = %while.cond.i.i.i.i1618
  %2054 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %2050)
  %2055 = icmp ne i32 %2054, %port.sroa.37.0.i.i1619
  br i1 %2055, label %while.body.i.i21.i.i.i.i1626.preheader, label %Flow3401, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i1626.preheader:           ; preds = %cond.false.i8.i.i.i.i1625
  br label %while.body.i.i21.i.i.i.i1626, !amdgpu.uniform !42

Flow3401:                                         ; preds = %Flow3399, %cond.false.i8.i.i.i.i1625
  br label %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i1629, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i1626:                     ; preds = %while.body.i.i21.i.i.i.i1626.preheader, %while.body.i.i21.i.i.i.i1626
  tail call void @llvm.amdgcn.s.sleep(i32 2)
  %2056 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i22.i.i.i.i1627 = getelementptr inbounds nuw i32, ptr %2056, i64 %idxprom.i23.i.i.i.i.i1615, !amdgpu.uniform !42
  %2057 = load atomic i32, ptr %arrayidx.i.i.i22.i.i.i.i1627 monotonic, align 4
  %2058 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %2057)
  %cmp.i.not.i.i23.i.i.i.i1628 = icmp eq i32 %2058, %port.sroa.37.0.i.i1619
  br i1 %cmp.i.not.i.i23.i.i.i.i1628, label %Flow3399, label %while.body.i.i21.i.i.i.i1626, !amdgpu.uniform !42

Flow3399:                                         ; preds = %while.body.i.i21.i.i.i.i1626
  br label %Flow3401, !amdgpu.uniform !42

Flow3402:                                         ; preds = %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1630, %while.cond.i.i.i.i1618
  %2059 = phi i64 [ %add.i.i.i.i1633, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1630 ], [ poison, %while.cond.i.i.i.i1618 ]
  %2060 = phi i32 [ %conv.i.i26.i.i.i.i1631, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1630 ], [ poison, %while.cond.i.i.i.i1618 ]
  %2061 = phi i1 [ false, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1630 ], [ true, %while.cond.i.i.i.i1618 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %2053)
  %2062 = call i64 @llvm.amdgcn.if.break.i64(i1 %2061, i64 %phi.broken3832)
  %2063 = call i1 @llvm.amdgcn.loop.i64(i64 %2062)
  br i1 %2063, label %cond.false.i.i.i1639, label %while.cond.i.i.i.i1618

_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i1629: ; preds = %Flow3401
  fence acquire
  br i1 %cmp.i.i.i.i1621, label %if.then.i.i.i.i.i.i.i1634, label %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1630, !amdgpu.uniform !42

if.then.i.i.i.i.i.i.i1634:                        ; preds = %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i1629
  %2064 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %arrayidx.i20.i35.i.i.i.i1637 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %2064, i64 %idxprom.i19.i.i.i.i.i1609
  %arrayidx.i22.i38.i.i.i.i1638 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %arrayidx.i20.i35.i.i.i.i1637, i64 %sh_prom.i.i.i.i
  store i8 10, ptr %arrayidx.i22.i38.i.i.i.i1638, align 1
  br label %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1630, !amdgpu.uniform !42

_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1630: ; preds = %if.then.i.i.i.i.i.i.i1634, %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i1629
  %conv.i.i26.i.i.i.i1631 = xor i32 %port.sroa.37.0.i.i1619, 1
  fence release
  %2065 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %arrayidx.i24.i29.i.i.i.i1632 = getelementptr inbounds nuw i32, ptr %2065, i64 %idxprom.i23.i.i.i.i.i1615
  store atomic i32 %conv.i.i26.i.i.i.i1631, ptr %arrayidx.i24.i29.i.i.i.i1632 monotonic, align 4
  %add.i.i.i.i1633 = add i64 %idx.0.i.i.i.i1620, 64
  br label %Flow3402, !amdgpu.uniform !42

cond.false.i.i.i1639:                             ; preds = %Flow3402
  %.lcssa3896 = phi i64 [ %2062, %Flow3402 ]
  %port.sroa.37.0.i.i1619.lcssa = phi i32 [ %port.sroa.37.0.i.i1619, %Flow3402 ]
  %.lcssa3895 = phi i32 [ %2050, %Flow3402 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa3896)
  %2066 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %.lcssa3895)
  %2067 = icmp ne i32 %2066, %port.sroa.37.0.i.i1619.lcssa
  %2068 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %2067)
  %2069 = extractvalue { i1, i64 } %2068, 0
  %2070 = extractvalue { i1, i64 } %2068, 1
  br i1 %2069, label %while.body.i.i.i.i1640.preheader, label %Flow3398

while.body.i.i.i.i1640.preheader:                 ; preds = %cond.false.i.i.i1639
  br label %while.body.i.i.i.i1640, !amdgpu.uniform !42

Flow3398:                                         ; preds = %Flow3396, %cond.false.i.i.i1639
  call void @llvm.amdgcn.end.cf.i64(i64 %2070)
  br label %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIcEERS1_T_.exit1661, !amdgpu.uniform !42

while.body.i.i.i.i1640:                           ; preds = %while.body.i.i.i.i1640.preheader, %while.body.i.i.i.i1640
  %phi.broken3833 = phi i64 [ 0, %while.body.i.i.i.i1640.preheader ], [ %2074, %while.body.i.i.i.i1640 ]
  tail call void @llvm.amdgcn.s.sleep(i32 2)
  %2071 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i8.i.i1641 = getelementptr inbounds nuw i32, ptr %2071, i64 %idxprom.i23.i.i.i.i.i1615, !amdgpu.uniform !42
  %2072 = load atomic i32, ptr %arrayidx.i.i.i8.i.i1641 monotonic, align 4
  %2073 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %2072)
  %cmp.i.not.i.i.i.i1642 = icmp eq i32 %2073, %port.sroa.37.0.i.i1619.lcssa
  %2074 = call i64 @llvm.amdgcn.if.break.i64(i1 %cmp.i.not.i.i.i.i1642, i64 %phi.broken3833)
  %2075 = call i1 @llvm.amdgcn.loop.i64(i64 %2074)
  br i1 %2075, label %Flow3396, label %while.body.i.i.i.i1640

Flow3396:                                         ; preds = %while.body.i.i.i.i1640
  %.lcssa3894 = phi i64 [ %2074, %while.body.i.i.i.i1640 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa3894)
  br label %Flow3398, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIcEERS1_T_.exit1661: ; preds = %Flow3398
  fence acquire
  tail call void @llvm.amdgcn.wave.barrier()
  fence syncscope("agent") release
  %2076 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %.lcssa3901, i1 true)
  %iszero.i.i.i.i.i13.i.i1644 = icmp ne i64 %.lcssa3901, 0
  %cmp2.i.i.i.i14.i.i1645 = icmp eq i64 %2076, %sh_prom.i.i.i.i
  %cmp.i.i.i.i15.i.i1646 = select i1 %iszero.i.i.i.i.i13.i.i1644, i1 %cmp2.i.i.i.i14.i.i1645, i1 false
  %2077 = and i32 %.lcssa3902, 31
  %conv4.i.i.i19.i.i1650 = zext i1 %cmp.i.i.i.i15.i.i1646 to i32
  %shl.i.i.i20.i.i1651 = shl nuw i32 %conv4.i.i.i19.i.i1650, %2077
  %xor.i.i.i21.i.i1652 = xor i32 %shl.i.i.i20.i.i1651, -1
  %2078 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2628, !amdgpu.uniform !42

2079:                                             ; preds = %ComputeEnd2629
  %div8.i.i.i16.i.i1647 = lshr i32 %.lcssa3902, 5
  %idxprom.i.i.i17.i.i1648 = zext nneg i32 %div8.i.i.i16.i.i1647 to i64
  %2080 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i17.i.i1648
  %sunkaddr3287 = getelementptr inbounds i8, ptr addrspace(1) %2080, i64 40
  %2081 = atomicrmw and ptr addrspace(1) %sunkaddr3287, i32 %.lcssa3893 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %2082, !amdgpu.uniform !42

2082:                                             ; preds = %2079, %ComputeEnd2629
  call void @llvm.amdgcn.end.cf.i64(i64 %3661)
  tail call void @llvm.amdgcn.wave.barrier()
  br i1 %cmp2.i.i.i37.i, label %if.then3.i.i.i89.i, label %if.end4.i.i.i78.i, !amdgpu.uniform !42

if.then3.i.i.i89.i:                               ; preds = %2082
  %mul.i.i.i90.i = mul nuw i64 %inc.i.i.i34.i, 11
  %div18.i.i.i91.i = lshr i64 %mul.i.i.i90.i, 3
  br label %if.end4.i.i.i78.i, !amdgpu.uniform !42

if.end4.i.i.i78.i:                                ; preds = %if.then3.i.i.i89.i, %2082
  %new_capacity.addr.0.i.i.i79.i = phi i64 [ %div18.i.i.i91.i, %if.then3.i.i.i89.i ], [ %inc.i.i.i34.i, %2082 ]
  %sub.i.i1662 = add i64 %new_capacity.addr.0.i.i.i79.i, 7
  %div2.i.i1663 = and i64 %sub.i.i1662, -8
  %2083 = load ptr, ptr addrspace(1) @_ZN12_GLOBAL__N_13ptrE, align 8, !tbaa !55
  %add.ptr.i.i1664 = getelementptr inbounds nuw i8, ptr %2083, i64 %div2.i.i1663
  store ptr %add.ptr.i.i1664, ptr addrspace(1) @_ZN12_GLOBAL__N_13ptrE, align 8, !tbaa !55
  br i1 %613, label %loop-memcpy-expansion2273.preheader, label %Flow3395, !amdgpu.uniform !42

loop-memcpy-expansion2273.preheader:              ; preds = %if.end4.i.i.i78.i
  br label %loop-memcpy-expansion2273, !amdgpu.uniform !42

Flow3395:                                         ; preds = %Flow3394, %if.end4.i.i.i78.i
  br label %loop-memcpy-residual-header2276, !amdgpu.uniform !42

loop-memcpy-expansion2273:                        ; preds = %loop-memcpy-expansion2273.preheader, %loop-memcpy-expansion2273
  %loop-index2274 = phi i64 [ %2087, %loop-memcpy-expansion2273 ], [ 0, %loop-memcpy-expansion2273.preheader ]
  %2084 = getelementptr i8, ptr %4, i64 %loop-index2274, !amdgpu.uniform !42
  %2085 = load <4 x i32>, ptr %2084, align 1
  %2086 = getelementptr i8, ptr %2083, i64 %loop-index2274
  store <4 x i32> %2085, ptr %2086, align 1
  %2087 = add i64 %loop-index2274, 16
  %2088 = icmp uge i64 %2087, %612
  br i1 %2088, label %Flow3394, label %loop-memcpy-expansion2273, !amdgpu.uniform !42

Flow3392:                                         ; preds = %Flow3391, %loop-memcpy-residual-header2276
  br label %post-loop-memcpy-expansion2272, !amdgpu.uniform !42

loop-memcpy-residual2275:                         ; preds = %loop-memcpy-residual2275.preheader, %loop-memcpy-residual2275
  %residual-loop-index2277 = phi i64 [ %2090, %loop-memcpy-residual2275 ], [ 0, %loop-memcpy-residual2275.preheader ]
  %scevgep2896 = getelementptr i8, ptr %scevgep2895, i64 %residual-loop-index2277, !amdgpu.uniform !42
  %2089 = load i8, ptr %scevgep2896, align 1
  %scevgep2894 = getelementptr i8, ptr %scevgep2893, i64 %residual-loop-index2277
  store i8 %2089, ptr %scevgep2894, align 1
  %2090 = add i64 %residual-loop-index2277, 1
  %2091 = icmp uge i64 %2090, %611
  br i1 %2091, label %Flow3391, label %loop-memcpy-residual2275, !amdgpu.uniform !42

Flow3391:                                         ; preds = %loop-memcpy-residual2275
  br label %Flow3392, !amdgpu.uniform !42

post-loop-memcpy-expansion2272:                   ; preds = %Flow3392
  %sub.ptr.lhs.cast.i.i1665 = ptrtoint ptr %add.ptr.i.i1664 to i64
  %sub.ptr.sub.i.i1666 = sub i64 %sub.ptr.lhs.cast.i.i1665, ptrtoint (ptr addrspacecast (ptr addrspace(1) @_ZN12_GLOBAL__N_16memoryE to ptr) to i64)
  %cmp.i.i1667 = icmp ugt i64 %sub.ptr.sub.i.i1666, 65335
  %cond.i.i1668 = select i1 %cmp.i.i1667, ptr null, ptr %2083
  %cmp.not.i10.i.i86.i = icmp ne ptr %cond.i.i1668, addrspacecast (ptr addrspace(4) @_ZN22__llvm_libc_22_0_0_git3cpp6string14NULL_CHARACTERE to ptr)
  br i1 %cmp.not.i10.i.i86.i, label %if.then.i.i.i87.i, label %for.cond.i.i.i1670.preheader, !amdgpu.uniform !42

if.then.i.i.i87.i:                                ; preds = %post-loop-memcpy-expansion2272
  %arrayidx.i.i.i88.i = getelementptr inbounds nuw i8, ptr %2083, i64 %spec.select.i
  store i8 0, ptr %arrayidx.i.i.i88.i, align 1, !tbaa !57
  br label %for.cond.i.i.i1670.preheader, !amdgpu.uniform !42

for.cond.i.i.i1670.preheader:                     ; preds = %if.then.i.i.i87.i, %post-loop-memcpy-expansion2272
  br label %for.cond.i.i.i1670, !amdgpu.uniform !42

for.cond.i.i.i1670:                               ; preds = %for.cond.i.i.i1670.preheader, %Flow3389
  %2092 = phi i32 [ %2130, %Flow3389 ], [ poison, %for.cond.i.i.i1670.preheader ]
  %2093 = phi i64 [ %2131, %Flow3389 ], [ poison, %for.cond.i.i.i1670.preheader ]
  %2094 = phi i32 [ %2132, %Flow3389 ], [ poison, %for.cond.i.i.i1670.preheader ]
  %retval.sroa.6.0.i.i.i1671 = phi i32 [ %2135, %Flow3389 ], [ undef, %for.cond.i.i.i1670.preheader ]
  %retval.sroa.2.0.i.i.i1673 = phi i64 [ %2134, %Flow3389 ], [ undef, %for.cond.i.i.i1670.preheader ]
  %retval.sroa.8.0.i.i.i1674 = phi i32 [ %2133, %Flow3389 ], [ undef, %for.cond.i.i.i1670.preheader ]
  %index.0.i.i.i1675 = phi i32 [ %2137, %Flow3389 ], [ 0, %for.cond.i.i.i1670.preheader ]
  %2095 = load i32, ptr addrspace(1) @__llvm_rpc_client, align 8, !tbaa !68
  %cmp.not.i.i.i1676 = icmp ult i32 %index.0.i.i.i1675, %2095
  %spec.store.select.i.i.i1677 = select i1 %cmp.not.i.i.i1676, i32 %index.0.i.i.i1675, i32 0
  %2096 = tail call noundef i64 @llvm.amdgcn.ballot.i64(i1 true)
  %2097 = and i32 %spec.store.select.i.i.i1677, 31
  %2098 = lshr i64 %2096, %sh_prom.i.i.i.i
  %2099 = trunc i64 %2098 to i32
  %conv4.i.i.i.i.i1681 = and i32 %2099, 1
  %shl.i.i.i.i.i1682 = shl nuw i32 %conv4.i.i.i.i.i1681, %2097
  %2100 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2632, !amdgpu.uniform !42

2101:                                             ; preds = %ComputeEnd2633
  %sunkaddr3288 = getelementptr inbounds i8, ptr addrspace(1) %3671, i64 40
  %2102 = atomicrmw or ptr addrspace(1) %sunkaddr3288, i32 %.lcssa3887 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %2103, !amdgpu.uniform !42

2103:                                             ; preds = %2101, %ComputeEnd2633
  %2104 = phi i32 [ %2102, %2101 ], [ poison, %ComputeEnd2633 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %3680)
  %2105 = call i32 @llvm.amdgcn.readfirstlane.i32(i32 %2104)
  %2106 = or i32 %2105, %.lcssa3888
  %shl5.i.i.i.i.i1683 = shl nuw i32 1, %2097
  %and.i.i.i.i.i1684 = and i32 %shl5.i.i.i.i.i1683, %2106
  %tobool3.i.i.i.i1685 = icmp ne i32 %and.i.i.i.i.i1684, 0
  %2107 = tail call i64 @llvm.amdgcn.ballot.i64(i1 %tobool3.i.i.i.i1685)
  %and.i.i.i.i.i.i1686 = and i64 %2107, %2096
  %cmp.i.not.i.i.i1687 = icmp ne i64 %2096, %and.i.i.i.i.i.i1686
  br i1 %cmp.i.not.i.i.i1687, label %if.then.i.i.i.i1688, label %Flow3388, !amdgpu.uniform !42

if.then.i.i.i.i1688:                              ; preds = %2103
  fence syncscope("agent") acquire
  %2108 = load <2 x i64>, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !75
  %2109 = extractelement <2 x i64> %2108, i32 0
  %2110 = inttoptr i64 %2109 to ptr
  %2111 = extractelement <2 x i64> %2108, i32 1
  %2112 = inttoptr i64 %2111 to ptr
  %idxprom.i.i.i.i1689 = zext i32 %spec.store.select.i.i.i1677 to i64
  %arrayidx.i.i.i.i1690 = getelementptr inbounds nuw i32, ptr %2110, i64 %idxprom.i.i.i.i1689, !amdgpu.uniform !42
  %2113 = load atomic i32, ptr %arrayidx.i.i.i.i1690 monotonic, align 4
  %2114 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %2113)
  %arrayidx.i45.i.i.i1691 = getelementptr inbounds nuw i32, ptr %2112, i64 %idxprom.i.i.i.i1689, !amdgpu.uniform !42
  %2115 = load atomic i32, ptr %arrayidx.i45.i.i.i1691 monotonic, align 4
  %2116 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %2115)
  %cmp.i46.not.i.i.i1692 = icmp ne i32 %2114, %2116
  br i1 %cmp.i46.not.i.i.i1692, label %if.then12.i.i.i1693, label %Flow3386, !amdgpu.uniform !42

if.then12.i.i.i1693:                              ; preds = %if.then.i.i.i.i1688
  fence syncscope("agent") release
  %2117 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %2096, i1 true)
  %iszero.i.i.i.i.i.i.i1694 = icmp ne i64 %2096, 0
  %cmp2.i.i.i.i.i.i1695 = icmp eq i64 %2117, %sh_prom.i.i.i.i
  %cmp.i.i.i.i.i.i1696 = select i1 %iszero.i.i.i.i.i.i.i1694, i1 %cmp2.i.i.i.i.i.i1695, i1 false
  %conv4.i.i51.i.i.i1697 = zext i1 %cmp.i.i.i.i.i.i1696 to i32
  %shl.i.i52.i.i.i1698 = shl nuw i32 %conv4.i.i51.i.i.i1697, %2097
  %xor.i.i.i.i.i1699 = xor i32 %shl.i.i52.i.i.i1698, -1
  %2118 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2637, !amdgpu.uniform !42

2119:                                             ; preds = %ComputeEnd2638
  %sunkaddr3289 = getelementptr inbounds i8, ptr addrspace(1) %3671, i64 40
  %2120 = atomicrmw and ptr addrspace(1) %sunkaddr3289, i32 %.lcssa3889 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %2121, !amdgpu.uniform !42

2121:                                             ; preds = %2119, %ComputeEnd2638
  call void @llvm.amdgcn.end.cf.i64(i64 %3698)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %Flow3386, !amdgpu.uniform !42

if.end14.i.i.i1759:                               ; preds = %Flow3386
  %2122 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %2096, i1 true)
  %iszero.i.i.i.i.i.i1760 = icmp ne i64 %2096, 0
  %cmp2.i.i.i.i.i1761 = icmp eq i64 %2122, %sh_prom.i.i.i.i
  %cmp.i.i.i.i.i1762 = select i1 %iszero.i.i.i.i.i.i1760, i1 %cmp2.i.i.i.i.i1761, i1 false
  %2123 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.i.i.i.i.i1762)
  %2124 = extractvalue { i1, i64 } %2123, 0
  %2125 = extractvalue { i1, i64 } %2123, 1
  br i1 %2124, label %if.then16.i.i.i1764, label %if.end22.i.i.i1763

if.then16.i.i.i1764:                              ; preds = %if.end14.i.i.i1759
  %2126 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !76
  %opcode.i.i.i1765 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %2126, i64 %idxprom.i.i.i.i1689, i32 1
  store i32 1660944387, ptr %opcode.i.i.i1765, align 8, !tbaa !77
  %arrayidx21.i.i.i1766 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %2126, i64 %idxprom.i.i.i.i1689
  store i64 %2096, ptr %arrayidx21.i.i.i1766, align 8, !tbaa !80
  br label %if.end22.i.i.i1763, !amdgpu.uniform !42

Flow3387:                                         ; preds = %if.end22.i.i.i1763, %Flow3386
  %2127 = phi i32 [ %2114, %if.end22.i.i.i1763 ], [ %retval.sroa.8.0.i.i.i1674, %Flow3386 ]
  %2128 = phi i64 [ %2096, %if.end22.i.i.i1763 ], [ %retval.sroa.2.0.i.i.i1673, %Flow3386 ]
  %2129 = phi i32 [ %spec.store.select.i.i.i1677, %if.end22.i.i.i1763 ], [ %retval.sroa.6.0.i.i.i1671, %Flow3386 ]
  br label %cleanup26.i.i.i1700, !amdgpu.uniform !42

if.end22.i.i.i1763:                               ; preds = %if.then16.i.i.i1764, %if.end14.i.i.i1759
  call void @llvm.amdgcn.end.cf.i64(i64 %2125)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %Flow3387, !amdgpu.uniform !42

Flow3388:                                         ; preds = %cleanup26.i.i.i1700, %2103
  %2130 = phi i32 [ %2129, %cleanup26.i.i.i1700 ], [ %2092, %2103 ]
  %2131 = phi i64 [ %2128, %cleanup26.i.i.i1700 ], [ %2093, %2103 ]
  %2132 = phi i32 [ %2127, %cleanup26.i.i.i1700 ], [ %2094, %2103 ]
  %2133 = phi i32 [ %2127, %cleanup26.i.i.i1700 ], [ %retval.sroa.8.0.i.i.i1674, %2103 ]
  %2134 = phi i64 [ %2128, %cleanup26.i.i.i1700 ], [ %retval.sroa.2.0.i.i.i1673, %2103 ]
  %2135 = phi i32 [ %2129, %cleanup26.i.i.i1700 ], [ %retval.sroa.6.0.i.i.i1671, %2103 ]
  %2136 = phi i1 [ %cmp.i46.not.i.i.i1692, %cleanup26.i.i.i1700 ], [ true, %2103 ]
  br i1 %2136, label %for.inc.i.i.i1705, label %Flow3389, !amdgpu.uniform !42

cleanup26.i.i.i1700:                              ; preds = %Flow3387
  br label %Flow3388, !amdgpu.uniform !42

for.inc.i.i.i1705:                                ; preds = %Flow3388
  %inc.i.i.i1710 = add i32 %spec.store.select.i.i.i1677, 1
  br label %Flow3389, !amdgpu.uniform !42

Flow3389:                                         ; preds = %for.inc.i.i.i1705, %Flow3388
  %2137 = phi i32 [ %inc.i.i.i1710, %for.inc.i.i.i1705 ], [ poison, %Flow3388 ]
  %2138 = phi i1 [ false, %for.inc.i.i.i1705 ], [ true, %Flow3388 ]
  br i1 %2138, label %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1711, label %for.cond.i.i.i1670, !amdgpu.uniform !42

_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1711: ; preds = %Flow3389
  %.lcssa3892 = phi i32 [ %2130, %Flow3389 ]
  %.lcssa3891 = phi i64 [ %2131, %Flow3389 ]
  %.lcssa3890 = phi i32 [ %2132, %Flow3389 ]
  fence acquire
  %2139 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %mul.i.i.i.i.i.i1712 = shl i32 %.lcssa3892, 6
  %idxprom.i19.i.i.i.i.i1713 = zext i32 %mul.i.i.i.i.i.i1712 to i64
  %arrayidx.i20.i.i.i.i.i1714 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %2139, i64 %idxprom.i19.i.i.i.i.i1713
  %arrayidx.i22.i.i.i.i.i1715 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %arrayidx.i20.i.i.i.i.i1714, i64 %sh_prom.i.i.i.i
  store i64 %spec.select.i, ptr %arrayidx.i22.i.i.i.i.i1715, align 8, !tbaa !82
  br i1 %672, label %loop-memcpy-expansion2195.preheader, label %Flow3385, !amdgpu.uniform !42

loop-memcpy-expansion2195.preheader:              ; preds = %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1711
  %2140 = add nuw nsw i64 %76, 8
  %scevgep2889 = getelementptr i8, ptr %2139, i64 %2140
  %2141 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i1713, 6
  %scevgep2890 = getelementptr i8, ptr %scevgep2889, i64 %2141
  br label %loop-memcpy-expansion2195, !amdgpu.uniform !42

Flow3385:                                         ; preds = %Flow3384, %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1711
  br label %loop-memcpy-residual-header2198, !amdgpu.uniform !42

loop-memcpy-expansion2195:                        ; preds = %loop-memcpy-expansion2195.preheader, %loop-memcpy-expansion2195
  %loop-index2196 = phi i64 [ %2144, %loop-memcpy-expansion2195 ], [ 0, %loop-memcpy-expansion2195.preheader ]
  %2142 = getelementptr i8, ptr %2083, i64 %loop-index2196, !amdgpu.uniform !42
  %2143 = load <4 x i32>, ptr %2142, align 1
  %scevgep2891 = getelementptr i8, ptr %scevgep2890, i64 %loop-index2196
  store <4 x i32> %2143, ptr %scevgep2891, align 1
  %2144 = add i64 %loop-index2196, 16
  %2145 = icmp uge i64 %2144, %671
  br i1 %2145, label %Flow3384, label %loop-memcpy-expansion2195, !amdgpu.uniform !42

Flow3382:                                         ; preds = %Flow3379, %Flow3381
  %2146 = phi i64 [ %2564, %Flow3379 ], [ %2558, %Flow3381 ]
  br label %post-loop-memcpy-expansion2194, !amdgpu.uniform !42

loop-memcpy-residual2197:                         ; preds = %loop-memcpy-residual2197.preheader, %loop-memcpy-residual2197
  %residual-loop-index2199 = phi i64 [ %2148, %loop-memcpy-residual2197 ], [ 0, %loop-memcpy-residual2197.preheader ]
  %scevgep2888 = getelementptr i8, ptr %scevgep2887, i64 %residual-loop-index2199, !amdgpu.uniform !42
  %2147 = load i8, ptr %scevgep2888, align 1
  %scevgep2886 = getelementptr i8, ptr %scevgep2885, i64 %residual-loop-index2199
  store i8 %2147, ptr %scevgep2886, align 1
  %2148 = add i64 %residual-loop-index2199, 1
  %2149 = icmp uge i64 %2148, %670
  br i1 %2149, label %Flow3379, label %loop-memcpy-residual2197, !amdgpu.uniform !42

Flow3379:                                         ; preds = %loop-memcpy-residual2197
  br label %Flow3382, !amdgpu.uniform !42

post-loop-memcpy-expansion2194:                   ; preds = %Flow3382
  %tobool.not.i.i.i.i.i.i1718 = icmp eq i32 %.lcssa3890, 0
  %conv.i.i.i.i.i.i1719 = zext i1 %tobool.not.i.i.i.i.i.i1718 to i32
  fence release
  %2150 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %idxprom.i23.i.i.i.i.i1720 = zext i32 %.lcssa3892 to i64
  %arrayidx.i24.i.i.i.i.i1721 = getelementptr inbounds nuw i32, ptr %2150, i64 %idxprom.i23.i.i.i.i.i1720
  store atomic i32 %conv.i.i.i.i.i.i1719, ptr %arrayidx.i24.i.i.i.i.i1721 monotonic, align 4
  %2151 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !84
  %arrayidx.i.i5.i.i1722 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %2151, i64 %idxprom.i23.i.i.i.i.i1720, !amdgpu.uniform !42
  %2152 = load i64, ptr %arrayidx.i.i5.i.i1722, align 8, !tbaa !80
  %scevgep2870 = getelementptr i8, ptr %2083, i64 56
  br label %while.cond.i.i.i.i1723, !amdgpu.uniform !42

while.cond.i.i.i.i1723:                           ; preds = %post-loop-memcpy-expansion2194, %Flow3377
  %phi.broken3834 = phi i64 [ 0, %post-loop-memcpy-expansion2194 ], [ %2169, %Flow3377 ]
  %lsr.iv2877 = phi i64 [ %2167, %Flow3377 ], [ %686, %post-loop-memcpy-expansion2194 ]
  %lsr.iv2871 = phi ptr [ %2166, %Flow3377 ], [ %scevgep2870, %post-loop-memcpy-expansion2194 ]
  %port.sroa.37.0.i.i1724 = phi i32 [ %2165, %Flow3377 ], [ %conv.i.i.i.i.i.i1719, %post-loop-memcpy-expansion2194 ]
  %idx.0.i.i.i.i1725 = phi i64 [ %2164, %Flow3377 ], [ 56, %post-loop-memcpy-expansion2194 ]
  %cmp.i.i.i.i1726 = icmp ugt i64 %spec.select.i, %idx.0.i.i.i.i1725
  %2153 = tail call i64 @llvm.amdgcn.ballot.i64(i1 %cmp.i.i.i.i1726)
  %and.i.i.i.i6.i.i1727 = and i64 %2153, %2152
  %tobool.not.i.i.i.i1728 = icmp ne i64 %and.i.i.i.i6.i.i1727, 0
  %2154 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i22.i.i.i1729 = getelementptr inbounds nuw i32, ptr %2154, i64 %idxprom.i23.i.i.i.i.i1720, !amdgpu.uniform !42
  %2155 = load atomic i32, ptr %arrayidx.i22.i.i.i1729 monotonic, align 4
  %2156 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %tobool.not.i.i.i.i1728)
  %2157 = extractvalue { i1, i64 } %2156, 0
  %2158 = extractvalue { i1, i64 } %2156, 1
  br i1 %2157, label %cond.false.i8.i.i.i.i1730, label %Flow3377

cond.false.i8.i.i.i.i1730:                        ; preds = %while.cond.i.i.i.i1723
  %2159 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %2155)
  %2160 = icmp ne i32 %2159, %port.sroa.37.0.i.i1724
  br i1 %2160, label %while.body.i.i21.i.i.i.i1731.preheader, label %Flow3376, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i1731.preheader:           ; preds = %cond.false.i8.i.i.i.i1730
  br label %while.body.i.i21.i.i.i.i1731, !amdgpu.uniform !42

Flow3376:                                         ; preds = %Flow3374, %cond.false.i8.i.i.i.i1730
  br label %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i1734, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i1731:                     ; preds = %while.body.i.i21.i.i.i.i1731.preheader, %while.body.i.i21.i.i.i.i1731
  tail call void @llvm.amdgcn.s.sleep(i32 2)
  %2161 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i22.i.i.i.i1732 = getelementptr inbounds nuw i32, ptr %2161, i64 %idxprom.i23.i.i.i.i.i1720, !amdgpu.uniform !42
  %2162 = load atomic i32, ptr %arrayidx.i.i.i22.i.i.i.i1732 monotonic, align 4
  %2163 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %2162)
  %cmp.i.not.i.i23.i.i.i.i1733 = icmp eq i32 %2163, %port.sroa.37.0.i.i1724
  br i1 %cmp.i.not.i.i23.i.i.i.i1733, label %Flow3374, label %while.body.i.i21.i.i.i.i1731, !amdgpu.uniform !42

Flow3374:                                         ; preds = %while.body.i.i21.i.i.i.i1731
  br label %Flow3376, !amdgpu.uniform !42

Flow3377:                                         ; preds = %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1735, %while.cond.i.i.i.i1723
  %2164 = phi i64 [ %add.i.i.i.i1738, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1735 ], [ poison, %while.cond.i.i.i.i1723 ]
  %2165 = phi i32 [ %conv.i.i26.i.i.i.i1736, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1735 ], [ poison, %while.cond.i.i.i.i1723 ]
  %2166 = phi ptr [ %scevgep2872, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1735 ], [ poison, %while.cond.i.i.i.i1723 ]
  %2167 = phi i64 [ %lsr.iv.next2878, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1735 ], [ poison, %while.cond.i.i.i.i1723 ]
  %2168 = phi i1 [ false, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1735 ], [ true, %while.cond.i.i.i.i1723 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %2158)
  %2169 = call i64 @llvm.amdgcn.if.break.i64(i1 %2168, i64 %phi.broken3834)
  %2170 = call i1 @llvm.amdgcn.loop.i64(i64 %2169)
  br i1 %2170, label %cond.false.i.i.i1745, label %while.cond.i.i.i.i1723

_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i1734: ; preds = %Flow3376
  fence acquire
  br i1 %cmp.i.i.i.i1726, label %if.then.i.i.i.i.i.i.i1739, label %Flow3373, !amdgpu.uniform !42

if.then.i.i.i.i.i.i.i1739:                        ; preds = %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i1734
  %sub.i.i.i.i.i.i.i1741 = sub nuw i64 %spec.select.i, %idx.0.i.i.i.i1725
  %spec.select.i.i.i36.i.i.i.i1742 = tail call i64 @llvm.umin.i64(i64 %sub.i.i.i.i.i.i.i1741, i64 64)
  %2171 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %2172 = and i64 %spec.select.i.i.i36.i.i.i.i1742, 15
  %2173 = sub i64 %spec.select.i.i.i36.i.i.i.i1742, %2172
  %2174 = icmp ne i64 %2173, 0
  br i1 %2174, label %loop-memcpy-expansion2189.preheader, label %Flow3372, !amdgpu.uniform !42

loop-memcpy-expansion2189.preheader:              ; preds = %if.then.i.i.i.i.i.i.i1739
  %arrayidx.i20.i35.i.i.i.i1743 = getelementptr %"struct.rpc::Buffer", ptr %2171, i64 %idxprom.i19.i.i.i.i.i1713
  %arrayidx.i22.i38.i.i.i.i1744 = getelementptr %"struct.rpc::Buffer", ptr %arrayidx.i20.i35.i.i.i.i1743, i64 %sh_prom.i.i.i.i
  br label %loop-memcpy-expansion2189, !amdgpu.uniform !42

Flow3372:                                         ; preds = %Flow3371, %if.then.i.i.i.i.i.i.i1739
  br label %loop-memcpy-residual-header2192, !amdgpu.uniform !42

loop-memcpy-expansion2189:                        ; preds = %loop-memcpy-expansion2189.preheader, %loop-memcpy-expansion2189
  %loop-index2190 = phi i64 [ %2177, %loop-memcpy-expansion2189 ], [ 0, %loop-memcpy-expansion2189.preheader ]
  %scevgep2873 = getelementptr i8, ptr %lsr.iv2871, i64 %loop-index2190, !amdgpu.uniform !42
  %2175 = load <4 x i32>, ptr %scevgep2873, align 1
  %2176 = getelementptr i8, ptr %arrayidx.i22.i38.i.i.i.i1744, i64 %loop-index2190
  store <4 x i32> %2175, ptr %2176, align 1
  %2177 = add i64 %loop-index2190, 16
  %2178 = icmp uge i64 %2177, %2173
  br i1 %2178, label %Flow3371, label %loop-memcpy-expansion2189, !amdgpu.uniform !42

Flow3369:                                         ; preds = %Flow3368, %loop-memcpy-residual-header2192
  br label %Flow3373, !amdgpu.uniform !42

loop-memcpy-residual2191:                         ; preds = %loop-memcpy-residual2191.preheader, %loop-memcpy-residual2191
  %residual-loop-index2193 = phi i64 [ %2180, %loop-memcpy-residual2191 ], [ 0, %loop-memcpy-residual2191.preheader ]
  %scevgep2883 = getelementptr i8, ptr %scevgep2882, i64 %residual-loop-index2193, !amdgpu.uniform !42
  %2179 = load i8, ptr %scevgep2883, align 1
  %scevgep2881 = getelementptr i8, ptr %scevgep2880, i64 %residual-loop-index2193
  store i8 %2179, ptr %scevgep2881, align 1
  %2180 = add i64 %residual-loop-index2193, 1
  %2181 = icmp uge i64 %2180, %2172
  br i1 %2181, label %Flow3368, label %loop-memcpy-residual2191, !amdgpu.uniform !42

Flow3368:                                         ; preds = %loop-memcpy-residual2191
  br label %Flow3369, !amdgpu.uniform !42

_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1735: ; preds = %Flow3373
  %conv.i.i26.i.i.i.i1736 = xor i32 %port.sroa.37.0.i.i1724, 1
  fence release
  %2182 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %arrayidx.i24.i29.i.i.i.i1737 = getelementptr inbounds nuw i32, ptr %2182, i64 %idxprom.i23.i.i.i.i.i1720
  store atomic i32 %conv.i.i26.i.i.i.i1736, ptr %arrayidx.i24.i29.i.i.i.i1737 monotonic, align 4
  %add.i.i.i.i1738 = add i64 %idx.0.i.i.i.i1725, 64
  %scevgep2872 = getelementptr i8, ptr %lsr.iv2871, i64 64
  %lsr.iv.next2878 = add i64 %lsr.iv2877, -64
  br label %Flow3377, !amdgpu.uniform !42

cond.false.i.i.i1745:                             ; preds = %Flow3377
  %.lcssa3886 = phi i64 [ %2169, %Flow3377 ]
  %port.sroa.37.0.i.i1724.lcssa = phi i32 [ %port.sroa.37.0.i.i1724, %Flow3377 ]
  %.lcssa3885 = phi i32 [ %2155, %Flow3377 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa3886)
  %2183 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %.lcssa3885)
  %2184 = icmp ne i32 %2183, %port.sroa.37.0.i.i1724.lcssa
  %2185 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %2184)
  %2186 = extractvalue { i1, i64 } %2185, 0
  %2187 = extractvalue { i1, i64 } %2185, 1
  br i1 %2186, label %while.body.i.i.i.i1746.preheader, label %Flow3366

while.body.i.i.i.i1746.preheader:                 ; preds = %cond.false.i.i.i1745
  br label %while.body.i.i.i.i1746, !amdgpu.uniform !42

Flow3366:                                         ; preds = %Flow3364, %cond.false.i.i.i1745
  call void @llvm.amdgcn.end.cf.i64(i64 %2187)
  br label %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsINS_3cpp6stringEEERS1_T_.exit1767, !amdgpu.uniform !42

while.body.i.i.i.i1746:                           ; preds = %while.body.i.i.i.i1746.preheader, %while.body.i.i.i.i1746
  %phi.broken3835 = phi i64 [ 0, %while.body.i.i.i.i1746.preheader ], [ %2191, %while.body.i.i.i.i1746 ]
  tail call void @llvm.amdgcn.s.sleep(i32 2)
  %2188 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i8.i.i1747 = getelementptr inbounds nuw i32, ptr %2188, i64 %idxprom.i23.i.i.i.i.i1720, !amdgpu.uniform !42
  %2189 = load atomic i32, ptr %arrayidx.i.i.i8.i.i1747 monotonic, align 4
  %2190 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %2189)
  %cmp.i.not.i.i.i.i1748 = icmp eq i32 %2190, %port.sroa.37.0.i.i1724.lcssa
  %2191 = call i64 @llvm.amdgcn.if.break.i64(i1 %cmp.i.not.i.i.i.i1748, i64 %phi.broken3835)
  %2192 = call i1 @llvm.amdgcn.loop.i64(i64 %2191)
  br i1 %2192, label %Flow3364, label %while.body.i.i.i.i1746

Flow3364:                                         ; preds = %while.body.i.i.i.i1746
  %.lcssa3884 = phi i64 [ %2191, %while.body.i.i.i.i1746 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa3884)
  br label %Flow3366, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsINS_3cpp6stringEEERS1_T_.exit1767: ; preds = %Flow3366
  fence acquire
  tail call void @llvm.amdgcn.wave.barrier()
  fence syncscope("agent") release
  %2193 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %.lcssa3891, i1 true)
  %iszero.i.i.i.i.i13.i.i1750 = icmp ne i64 %.lcssa3891, 0
  %cmp2.i.i.i.i14.i.i1751 = icmp eq i64 %2193, %sh_prom.i.i.i.i
  %cmp.i.i.i.i15.i.i1752 = select i1 %iszero.i.i.i.i.i13.i.i1750, i1 %cmp2.i.i.i.i14.i.i1751, i1 false
  %2194 = and i32 %.lcssa3892, 31
  %conv4.i.i.i19.i.i1756 = zext i1 %cmp.i.i.i.i15.i.i1752 to i32
  %shl.i.i.i20.i.i1757 = shl nuw i32 %conv4.i.i.i19.i.i1756, %2194
  %xor.i.i.i21.i.i1758 = xor i32 %shl.i.i.i20.i.i1757, -1
  %2195 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2641, !amdgpu.uniform !42

2196:                                             ; preds = %ComputeEnd2642
  %div8.i.i.i16.i.i1753 = lshr i32 %.lcssa3892, 5
  %idxprom.i.i.i17.i.i1754 = zext nneg i32 %div8.i.i.i16.i.i1753 to i64
  %2197 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i17.i.i1754
  %sunkaddr3290 = getelementptr inbounds i8, ptr addrspace(1) %2197, i64 40
  %2198 = atomicrmw and ptr addrspace(1) %sunkaddr3290, i32 %.lcssa3883 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %2199, !amdgpu.uniform !42

2199:                                             ; preds = %2196, %ComputeEnd2642
  call void @llvm.amdgcn.end.cf.i64(i64 %3715)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %for.cond.i.i1.i1776, !amdgpu.uniform !42

for.cond.i.i1.i1776:                              ; preds = %2199, %Flow3363
  %2200 = phi i32 [ %2238, %Flow3363 ], [ poison, %2199 ]
  %2201 = phi i64 [ %2239, %Flow3363 ], [ poison, %2199 ]
  %2202 = phi i32 [ %2240, %Flow3363 ], [ poison, %2199 ]
  %retval.sroa.6.0.i.i.i1777 = phi i32 [ %2243, %Flow3363 ], [ undef, %2199 ]
  %retval.sroa.2.0.i.i.i1779 = phi i64 [ %2242, %Flow3363 ], [ undef, %2199 ]
  %retval.sroa.8.0.i.i.i1780 = phi i32 [ %2241, %Flow3363 ], [ undef, %2199 ]
  %index.0.i.i.i1781 = phi i32 [ %2245, %Flow3363 ], [ 0, %2199 ]
  %2203 = load i32, ptr addrspace(1) @__llvm_rpc_client, align 8, !tbaa !68
  %cmp.not.i.i.i1782 = icmp ult i32 %index.0.i.i.i1781, %2203
  %spec.store.select.i.i.i1783 = select i1 %cmp.not.i.i.i1782, i32 %index.0.i.i.i1781, i32 0
  %2204 = tail call noundef i64 @llvm.amdgcn.ballot.i64(i1 true)
  %2205 = and i32 %spec.store.select.i.i.i1783, 31
  %2206 = lshr i64 %2204, %sh_prom.i.i.i.i
  %2207 = trunc i64 %2206 to i32
  %conv4.i.i.i.i.i1787 = and i32 %2207, 1
  %shl.i.i.i.i.i1788 = shl nuw i32 %conv4.i.i.i.i.i1787, %2205
  %2208 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2645, !amdgpu.uniform !42

2209:                                             ; preds = %ComputeEnd2646
  %sunkaddr3291 = getelementptr inbounds i8, ptr addrspace(1) %3725, i64 40
  %2210 = atomicrmw or ptr addrspace(1) %sunkaddr3291, i32 %.lcssa3877 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %2211, !amdgpu.uniform !42

2211:                                             ; preds = %2209, %ComputeEnd2646
  %2212 = phi i32 [ %2210, %2209 ], [ poison, %ComputeEnd2646 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %3734)
  %2213 = call i32 @llvm.amdgcn.readfirstlane.i32(i32 %2212)
  %2214 = or i32 %2213, %.lcssa3878
  %shl5.i.i.i.i.i1789 = shl nuw i32 1, %2205
  %and.i.i.i.i.i1790 = and i32 %shl5.i.i.i.i.i1789, %2214
  %tobool3.i.i.i.i1791 = icmp ne i32 %and.i.i.i.i.i1790, 0
  %2215 = tail call i64 @llvm.amdgcn.ballot.i64(i1 %tobool3.i.i.i.i1791)
  %and.i.i.i.i.i.i1792 = and i64 %2215, %2204
  %cmp.i.not.i.i.i1793 = icmp ne i64 %2204, %and.i.i.i.i.i.i1792
  br i1 %cmp.i.not.i.i.i1793, label %if.then.i.i.i.i1794, label %Flow3362, !amdgpu.uniform !42

if.then.i.i.i.i1794:                              ; preds = %2211
  fence syncscope("agent") acquire
  %2216 = load <2 x i64>, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !75
  %2217 = extractelement <2 x i64> %2216, i32 0
  %2218 = inttoptr i64 %2217 to ptr
  %2219 = extractelement <2 x i64> %2216, i32 1
  %2220 = inttoptr i64 %2219 to ptr
  %idxprom.i.i.i.i1795 = zext i32 %spec.store.select.i.i.i1783 to i64
  %arrayidx.i.i.i.i1796 = getelementptr inbounds nuw i32, ptr %2218, i64 %idxprom.i.i.i.i1795, !amdgpu.uniform !42
  %2221 = load atomic i32, ptr %arrayidx.i.i.i.i1796 monotonic, align 4
  %2222 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %2221)
  %arrayidx.i45.i.i.i1797 = getelementptr inbounds nuw i32, ptr %2220, i64 %idxprom.i.i.i.i1795, !amdgpu.uniform !42
  %2223 = load atomic i32, ptr %arrayidx.i45.i.i.i1797 monotonic, align 4
  %2224 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %2223)
  %cmp.i46.not.i.i.i1798 = icmp ne i32 %2222, %2224
  br i1 %cmp.i46.not.i.i.i1798, label %if.then12.i.i.i1799, label %Flow3360, !amdgpu.uniform !42

if.then12.i.i.i1799:                              ; preds = %if.then.i.i.i.i1794
  fence syncscope("agent") release
  %2225 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %2204, i1 true)
  %iszero.i.i.i.i.i.i.i1800 = icmp ne i64 %2204, 0
  %cmp2.i.i.i.i.i.i1801 = icmp eq i64 %2225, %sh_prom.i.i.i.i
  %cmp.i.i.i.i.i.i1802 = select i1 %iszero.i.i.i.i.i.i.i1800, i1 %cmp2.i.i.i.i.i.i1801, i1 false
  %conv4.i.i51.i.i.i1803 = zext i1 %cmp.i.i.i.i.i.i1802 to i32
  %shl.i.i52.i.i.i1804 = shl nuw i32 %conv4.i.i51.i.i.i1803, %2205
  %xor.i.i.i.i.i1805 = xor i32 %shl.i.i52.i.i.i1804, -1
  %2226 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2650, !amdgpu.uniform !42

2227:                                             ; preds = %ComputeEnd2651
  %sunkaddr3292 = getelementptr inbounds i8, ptr addrspace(1) %3725, i64 40
  %2228 = atomicrmw and ptr addrspace(1) %sunkaddr3292, i32 %.lcssa3879 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %2229, !amdgpu.uniform !42

2229:                                             ; preds = %2227, %ComputeEnd2651
  call void @llvm.amdgcn.end.cf.i64(i64 %3752)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %Flow3360, !amdgpu.uniform !42

if.end14.i.i.i1865:                               ; preds = %Flow3360
  %2230 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %2204, i1 true)
  %iszero.i.i.i.i.i.i1866 = icmp ne i64 %2204, 0
  %cmp2.i.i.i.i.i1867 = icmp eq i64 %2230, %sh_prom.i.i.i.i
  %cmp.i.i.i.i.i1868 = select i1 %iszero.i.i.i.i.i.i1866, i1 %cmp2.i.i.i.i.i1867, i1 false
  %2231 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.i.i.i.i.i1868)
  %2232 = extractvalue { i1, i64 } %2231, 0
  %2233 = extractvalue { i1, i64 } %2231, 1
  br i1 %2232, label %if.then16.i.i.i1870, label %if.end22.i.i.i1869

if.then16.i.i.i1870:                              ; preds = %if.end14.i.i.i1865
  %2234 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !76
  %opcode.i.i.i1871 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %2234, i64 %idxprom.i.i.i.i1795, i32 1
  store i32 1660944387, ptr %opcode.i.i.i1871, align 8, !tbaa !77
  %arrayidx21.i.i.i1872 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %2234, i64 %idxprom.i.i.i.i1795
  store i64 %2204, ptr %arrayidx21.i.i.i1872, align 8, !tbaa !80
  br label %if.end22.i.i.i1869, !amdgpu.uniform !42

Flow3361:                                         ; preds = %if.end22.i.i.i1869, %Flow3360
  %2235 = phi i32 [ %2222, %if.end22.i.i.i1869 ], [ %retval.sroa.8.0.i.i.i1780, %Flow3360 ]
  %2236 = phi i64 [ %2204, %if.end22.i.i.i1869 ], [ %retval.sroa.2.0.i.i.i1779, %Flow3360 ]
  %2237 = phi i32 [ %spec.store.select.i.i.i1783, %if.end22.i.i.i1869 ], [ %retval.sroa.6.0.i.i.i1777, %Flow3360 ]
  br label %cleanup26.i.i.i1806, !amdgpu.uniform !42

if.end22.i.i.i1869:                               ; preds = %if.then16.i.i.i1870, %if.end14.i.i.i1865
  call void @llvm.amdgcn.end.cf.i64(i64 %2233)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %Flow3361, !amdgpu.uniform !42

Flow3362:                                         ; preds = %cleanup26.i.i.i1806, %2211
  %2238 = phi i32 [ %2237, %cleanup26.i.i.i1806 ], [ %2200, %2211 ]
  %2239 = phi i64 [ %2236, %cleanup26.i.i.i1806 ], [ %2201, %2211 ]
  %2240 = phi i32 [ %2235, %cleanup26.i.i.i1806 ], [ %2202, %2211 ]
  %2241 = phi i32 [ %2235, %cleanup26.i.i.i1806 ], [ %retval.sroa.8.0.i.i.i1780, %2211 ]
  %2242 = phi i64 [ %2236, %cleanup26.i.i.i1806 ], [ %retval.sroa.2.0.i.i.i1779, %2211 ]
  %2243 = phi i32 [ %2237, %cleanup26.i.i.i1806 ], [ %retval.sroa.6.0.i.i.i1777, %2211 ]
  %2244 = phi i1 [ %cmp.i46.not.i.i.i1798, %cleanup26.i.i.i1806 ], [ true, %2211 ]
  br i1 %2244, label %for.inc.i.i.i1811, label %Flow3363, !amdgpu.uniform !42

cleanup26.i.i.i1806:                              ; preds = %Flow3361
  br label %Flow3362, !amdgpu.uniform !42

for.inc.i.i.i1811:                                ; preds = %Flow3362
  %inc.i.i.i1816 = add i32 %spec.store.select.i.i.i1783, 1
  br label %Flow3363, !amdgpu.uniform !42

Flow3363:                                         ; preds = %for.inc.i.i.i1811, %Flow3362
  %2245 = phi i32 [ %inc.i.i.i1816, %for.inc.i.i.i1811 ], [ poison, %Flow3362 ]
  %2246 = phi i1 [ false, %for.inc.i.i.i1811 ], [ true, %Flow3362 ]
  br i1 %2246, label %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1817, label %for.cond.i.i1.i1776, !amdgpu.uniform !42

_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1817: ; preds = %Flow3363
  %.lcssa3882 = phi i32 [ %2238, %Flow3363 ]
  %.lcssa3881 = phi i64 [ %2239, %Flow3363 ]
  %.lcssa3880 = phi i32 [ %2240, %Flow3363 ]
  fence acquire
  %2247 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %mul.i.i.i.i.i.i1818 = shl i32 %.lcssa3882, 6
  %idxprom.i19.i.i.i.i.i1819 = zext i32 %mul.i.i.i.i.i.i1818 to i64
  %arrayidx.i20.i.i.i.i.i1820 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %2247, i64 %idxprom.i19.i.i.i.i.i1819
  %arrayidx.i22.i.i.i.i.i1821 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %arrayidx.i20.i.i.i.i.i1820, i64 %sh_prom.i.i.i.i
  store i64 sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.35 to ptr), i64 10) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.35 to ptr) to i64)), ptr %arrayidx.i22.i.i.i.i.i1821, align 8, !tbaa !82
  br i1 %1231, label %loop-memcpy-expansion2363.preheader, label %Flow3359, !amdgpu.uniform !42

loop-memcpy-expansion2363.preheader:              ; preds = %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1817
  %2248 = add nuw nsw i64 %76, 8
  %scevgep2867 = getelementptr i8, ptr %2247, i64 %2248
  %2249 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i1819, 6
  %scevgep2868 = getelementptr i8, ptr %scevgep2867, i64 %2249
  br label %loop-memcpy-expansion2363, !amdgpu.uniform !42

Flow3359:                                         ; preds = %Flow3358, %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1817
  br label %loop-memcpy-residual-header2366, !amdgpu.uniform !42

loop-memcpy-expansion2363:                        ; preds = %loop-memcpy-expansion2363.preheader, %loop-memcpy-expansion2363
  %loop-index2364 = phi i64 [ %2252, %loop-memcpy-expansion2363 ], [ 0, %loop-memcpy-expansion2363.preheader ]
  %2250 = getelementptr i8, ptr addrspace(4) @.str.35, i64 %loop-index2364, !amdgpu.uniform !42
  %2251 = load <4 x i32>, ptr addrspace(4) %2250, align 1
  %scevgep2869 = getelementptr i8, ptr %scevgep2868, i64 %loop-index2364
  store <4 x i32> %2251, ptr %scevgep2869, align 1
  %2252 = add i64 %loop-index2364, 16
  %2253 = icmp uge i64 %2252, %1230
  br i1 %2253, label %Flow3358, label %loop-memcpy-expansion2363, !amdgpu.uniform !42

Flow3356:                                         ; preds = %Flow3354, %Flow3355
  %2254 = phi i64 [ %2710, %Flow3354 ], [ %2707, %Flow3355 ]
  br label %post-loop-memcpy-expansion2362, !amdgpu.uniform !42

loop-memcpy-residual2365:                         ; preds = %loop-memcpy-residual2365.preheader, %loop-memcpy-residual2365
  %residual-loop-index2367 = phi i64 [ %2256, %loop-memcpy-residual2365 ], [ 0, %loop-memcpy-residual2365.preheader ]
  %scevgep2866 = getelementptr i8, ptr addrspace(4) @.str.35, i64 %residual-loop-index2367, !amdgpu.uniform !42
  %2255 = load i8, ptr addrspace(4) %scevgep2866, align 1
  %scevgep2865 = getelementptr i8, ptr %scevgep2864, i64 %residual-loop-index2367
  store i8 %2255, ptr %scevgep2865, align 1
  %2256 = add i64 %residual-loop-index2367, 1
  %2257 = icmp uge i64 %2256, %1229
  br i1 %2257, label %Flow3354, label %loop-memcpy-residual2365, !amdgpu.uniform !42

Flow3354:                                         ; preds = %loop-memcpy-residual2365
  br label %Flow3356, !amdgpu.uniform !42

post-loop-memcpy-expansion2362:                   ; preds = %Flow3356
  %tobool.not.i.i.i.i.i.i1824 = icmp eq i32 %.lcssa3880, 0
  %conv.i.i.i.i.i.i1825 = zext i1 %tobool.not.i.i.i.i.i.i1824 to i32
  fence release
  %2258 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %idxprom.i23.i.i.i.i.i1826 = zext i32 %.lcssa3882 to i64
  %arrayidx.i24.i.i.i.i.i1827 = getelementptr inbounds nuw i32, ptr %2258, i64 %idxprom.i23.i.i.i.i.i1826
  store atomic i32 %conv.i.i.i.i.i.i1825, ptr %arrayidx.i24.i.i.i.i.i1827 monotonic, align 4
  %2259 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !84
  %arrayidx.i.i5.i.i1828 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %2259, i64 %idxprom.i23.i.i.i.i.i1826, !amdgpu.uniform !42
  %2260 = load i64, ptr %arrayidx.i.i5.i.i1828, align 8, !tbaa !80
  br label %while.cond.i.i.i.i1829, !amdgpu.uniform !42

while.cond.i.i.i.i1829:                           ; preds = %post-loop-memcpy-expansion2362, %Flow3353
  %phi.broken3836 = phi i64 [ 0, %post-loop-memcpy-expansion2362 ], [ %2277, %Flow3353 ]
  %lsr.iv2858 = phi i64 [ %2275, %Flow3353 ], [ -46, %post-loop-memcpy-expansion2362 ]
  %lsr.iv2853 = phi ptr addrspace(4) [ %2274, %Flow3353 ], [ getelementptr (i8, ptr addrspace(4) @.str.35, i64 56), %post-loop-memcpy-expansion2362 ]
  %port.sroa.37.0.i.i1830 = phi i32 [ %2273, %Flow3353 ], [ %conv.i.i.i.i.i.i1825, %post-loop-memcpy-expansion2362 ]
  %idx.0.i.i.i.i1831 = phi i64 [ %2272, %Flow3353 ], [ 56, %post-loop-memcpy-expansion2362 ]
  %cmp.i.i.i.i1832 = icmp ult i64 %idx.0.i.i.i.i1831, sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.35 to ptr), i64 10) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.35 to ptr) to i64))
  %2261 = tail call i64 @llvm.amdgcn.ballot.i64(i1 %cmp.i.i.i.i1832)
  %and.i.i.i.i6.i.i1833 = and i64 %2261, %2260
  %tobool.not.i.i.i.i1834 = icmp ne i64 %and.i.i.i.i6.i.i1833, 0
  %2262 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i22.i.i.i1835 = getelementptr inbounds nuw i32, ptr %2262, i64 %idxprom.i23.i.i.i.i.i1826, !amdgpu.uniform !42
  %2263 = load atomic i32, ptr %arrayidx.i22.i.i.i1835 monotonic, align 4
  %2264 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %tobool.not.i.i.i.i1834)
  %2265 = extractvalue { i1, i64 } %2264, 0
  %2266 = extractvalue { i1, i64 } %2264, 1
  br i1 %2265, label %cond.false.i8.i.i.i.i1836, label %Flow3353

cond.false.i8.i.i.i.i1836:                        ; preds = %while.cond.i.i.i.i1829
  %2267 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %2263)
  %2268 = icmp ne i32 %2267, %port.sroa.37.0.i.i1830
  br i1 %2268, label %while.body.i.i21.i.i.i.i1837.preheader, label %Flow3352, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i1837.preheader:           ; preds = %cond.false.i8.i.i.i.i1836
  br label %while.body.i.i21.i.i.i.i1837, !amdgpu.uniform !42

Flow3352:                                         ; preds = %Flow3351, %cond.false.i8.i.i.i.i1836
  br label %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i1840, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i1837:                     ; preds = %while.body.i.i21.i.i.i.i1837.preheader, %while.body.i.i21.i.i.i.i1837
  tail call void @llvm.amdgcn.s.sleep(i32 2)
  %2269 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i22.i.i.i.i1838 = getelementptr inbounds nuw i32, ptr %2269, i64 %idxprom.i23.i.i.i.i.i1826, !amdgpu.uniform !42
  %2270 = load atomic i32, ptr %arrayidx.i.i.i22.i.i.i.i1838 monotonic, align 4
  %2271 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %2270)
  %cmp.i.not.i.i23.i.i.i.i1839 = icmp eq i32 %2271, %port.sroa.37.0.i.i1830
  br i1 %cmp.i.not.i.i23.i.i.i.i1839, label %Flow3351, label %while.body.i.i21.i.i.i.i1837, !amdgpu.uniform !42

Flow3351:                                         ; preds = %while.body.i.i21.i.i.i.i1837
  br label %Flow3352, !amdgpu.uniform !42

Flow3353:                                         ; preds = %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1841, %while.cond.i.i.i.i1829
  %2272 = phi i64 [ %add.i.i.i.i1844, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1841 ], [ poison, %while.cond.i.i.i.i1829 ]
  %2273 = phi i32 [ %conv.i.i26.i.i.i.i1842, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1841 ], [ poison, %while.cond.i.i.i.i1829 ]
  %2274 = phi ptr addrspace(4) [ %scevgep2854, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1841 ], [ poison, %while.cond.i.i.i.i1829 ]
  %2275 = phi i64 [ %lsr.iv.next, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1841 ], [ poison, %while.cond.i.i.i.i1829 ]
  %2276 = phi i1 [ false, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1841 ], [ true, %while.cond.i.i.i.i1829 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %2266)
  %2277 = call i64 @llvm.amdgcn.if.break.i64(i1 %2276, i64 %phi.broken3836)
  %2278 = call i1 @llvm.amdgcn.loop.i64(i64 %2277)
  br i1 %2278, label %cond.false.i.i.i1851, label %while.cond.i.i.i.i1829

_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i1840: ; preds = %Flow3352
  fence acquire
  br i1 %cmp.i.i.i.i1832, label %if.then.i.i.i.i.i.i.i1845, label %Flow3350, !amdgpu.uniform !42

if.then.i.i.i.i.i.i.i1845:                        ; preds = %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i1840
  %sub.i.i.i.i.i.i.i1847 = sub nuw i64 sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.35 to ptr), i64 10) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.35 to ptr) to i64)), %idx.0.i.i.i.i1831
  %spec.select.i.i.i36.i.i.i.i1848 = tail call i64 @llvm.umin.i64(i64 %sub.i.i.i.i.i.i.i1847, i64 64)
  %2279 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %2280 = and i64 %spec.select.i.i.i36.i.i.i.i1848, 15
  %2281 = sub i64 %spec.select.i.i.i36.i.i.i.i1848, %2280
  %2282 = icmp ne i64 %2281, 0
  br i1 %2282, label %loop-memcpy-expansion2411.preheader, label %Flow3349, !amdgpu.uniform !42

loop-memcpy-expansion2411.preheader:              ; preds = %if.then.i.i.i.i.i.i.i1845
  %arrayidx.i20.i35.i.i.i.i1849 = getelementptr %"struct.rpc::Buffer", ptr %2279, i64 %idxprom.i19.i.i.i.i.i1819
  %arrayidx.i22.i38.i.i.i.i1850 = getelementptr %"struct.rpc::Buffer", ptr %arrayidx.i20.i35.i.i.i.i1849, i64 %sh_prom.i.i.i.i
  br label %loop-memcpy-expansion2411, !amdgpu.uniform !42

Flow3349:                                         ; preds = %Flow3348, %if.then.i.i.i.i.i.i.i1845
  br label %loop-memcpy-residual-header2414, !amdgpu.uniform !42

loop-memcpy-expansion2411:                        ; preds = %loop-memcpy-expansion2411.preheader, %loop-memcpy-expansion2411
  %loop-index2412 = phi i64 [ %2285, %loop-memcpy-expansion2411 ], [ 0, %loop-memcpy-expansion2411.preheader ]
  %scevgep2855 = getelementptr i8, ptr addrspace(4) %lsr.iv2853, i64 %loop-index2412, !amdgpu.uniform !42
  %2283 = load <4 x i32>, ptr addrspace(4) %scevgep2855, align 1
  %2284 = getelementptr i8, ptr %arrayidx.i22.i38.i.i.i.i1850, i64 %loop-index2412
  store <4 x i32> %2283, ptr %2284, align 1
  %2285 = add i64 %loop-index2412, 16
  %2286 = icmp uge i64 %2285, %2281
  br i1 %2286, label %Flow3348, label %loop-memcpy-expansion2411, !amdgpu.uniform !42

Flow3347:                                         ; preds = %Flow3346, %loop-memcpy-residual-header2414
  br label %Flow3350, !amdgpu.uniform !42

loop-memcpy-residual2413:                         ; preds = %loop-memcpy-residual2413.preheader, %loop-memcpy-residual2413
  %residual-loop-index2415 = phi i64 [ %2288, %loop-memcpy-residual2413 ], [ 0, %loop-memcpy-residual2413.preheader ]
  %scevgep2862 = getelementptr i8, ptr addrspace(4) %scevgep2861, i64 %residual-loop-index2415, !amdgpu.uniform !42
  %2287 = load i8, ptr addrspace(4) %scevgep2862, align 1
  %scevgep2860 = getelementptr i8, ptr %scevgep2859, i64 %residual-loop-index2415
  store i8 %2287, ptr %scevgep2860, align 1
  %2288 = add i64 %residual-loop-index2415, 1
  %2289 = icmp uge i64 %2288, %2280
  br i1 %2289, label %Flow3346, label %loop-memcpy-residual2413, !amdgpu.uniform !42

Flow3346:                                         ; preds = %loop-memcpy-residual2413
  br label %Flow3347, !amdgpu.uniform !42

_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1841: ; preds = %Flow3350
  %conv.i.i26.i.i.i.i1842 = xor i32 %port.sroa.37.0.i.i1830, 1
  fence release
  %2290 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %arrayidx.i24.i29.i.i.i.i1843 = getelementptr inbounds nuw i32, ptr %2290, i64 %idxprom.i23.i.i.i.i.i1826
  store atomic i32 %conv.i.i26.i.i.i.i1842, ptr %arrayidx.i24.i29.i.i.i.i1843 monotonic, align 4
  %add.i.i.i.i1844 = add i64 %idx.0.i.i.i.i1831, 64
  %scevgep2854 = getelementptr i8, ptr addrspace(4) %lsr.iv2853, i64 64
  %lsr.iv.next = add i64 %lsr.iv2858, -64
  br label %Flow3353, !amdgpu.uniform !42

cond.false.i.i.i1851:                             ; preds = %Flow3353
  %.lcssa3876 = phi i64 [ %2277, %Flow3353 ]
  %port.sroa.37.0.i.i1830.lcssa = phi i32 [ %port.sroa.37.0.i.i1830, %Flow3353 ]
  %.lcssa3875 = phi i32 [ %2263, %Flow3353 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa3876)
  %2291 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %.lcssa3875)
  %2292 = icmp ne i32 %2291, %port.sroa.37.0.i.i1830.lcssa
  %2293 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %2292)
  %2294 = extractvalue { i1, i64 } %2293, 0
  %2295 = extractvalue { i1, i64 } %2293, 1
  br i1 %2294, label %while.body.i.i.i.i1852.preheader, label %Flow3345

while.body.i.i.i.i1852.preheader:                 ; preds = %cond.false.i.i.i1851
  br label %while.body.i.i.i.i1852, !amdgpu.uniform !42

Flow3345:                                         ; preds = %Flow3344, %cond.false.i.i.i1851
  call void @llvm.amdgcn.end.cf.i64(i64 %2295)
  br label %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit1873, !amdgpu.uniform !42

while.body.i.i.i.i1852:                           ; preds = %while.body.i.i.i.i1852.preheader, %while.body.i.i.i.i1852
  %phi.broken3837 = phi i64 [ 0, %while.body.i.i.i.i1852.preheader ], [ %2299, %while.body.i.i.i.i1852 ]
  tail call void @llvm.amdgcn.s.sleep(i32 2)
  %2296 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i8.i.i1853 = getelementptr inbounds nuw i32, ptr %2296, i64 %idxprom.i23.i.i.i.i.i1826, !amdgpu.uniform !42
  %2297 = load atomic i32, ptr %arrayidx.i.i.i8.i.i1853 monotonic, align 4
  %2298 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %2297)
  %cmp.i.not.i.i.i.i1854 = icmp eq i32 %2298, %port.sroa.37.0.i.i1830.lcssa
  %2299 = call i64 @llvm.amdgcn.if.break.i64(i1 %cmp.i.not.i.i.i.i1854, i64 %phi.broken3837)
  %2300 = call i1 @llvm.amdgcn.loop.i64(i64 %2299)
  br i1 %2300, label %Flow3344, label %while.body.i.i.i.i1852

Flow3344:                                         ; preds = %while.body.i.i.i.i1852
  %.lcssa3874 = phi i64 [ %2299, %while.body.i.i.i.i1852 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa3874)
  br label %Flow3345, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit1873: ; preds = %Flow3345
  fence acquire
  tail call void @llvm.amdgcn.wave.barrier()
  fence syncscope("agent") release
  %2301 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %.lcssa3881, i1 true)
  %iszero.i.i.i.i.i13.i.i1856 = icmp ne i64 %.lcssa3881, 0
  %cmp2.i.i.i.i14.i.i1857 = icmp eq i64 %2301, %sh_prom.i.i.i.i
  %cmp.i.i.i.i15.i.i1858 = select i1 %iszero.i.i.i.i.i13.i.i1856, i1 %cmp2.i.i.i.i14.i.i1857, i1 false
  %2302 = and i32 %.lcssa3882, 31
  %conv4.i.i.i19.i.i1862 = zext i1 %cmp.i.i.i.i15.i.i1858 to i32
  %shl.i.i.i20.i.i1863 = shl nuw i32 %conv4.i.i.i19.i.i1862, %2302
  %xor.i.i.i21.i.i1864 = xor i32 %shl.i.i.i20.i.i1863, -1
  %2303 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2654, !amdgpu.uniform !42

2304:                                             ; preds = %ComputeEnd2655
  %div8.i.i.i16.i.i1859 = lshr i32 %.lcssa3882, 5
  %idxprom.i.i.i17.i.i1860 = zext nneg i32 %div8.i.i.i16.i.i1859 to i64
  %2305 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i17.i.i1860
  %sunkaddr3293 = getelementptr inbounds i8, ptr addrspace(1) %2305, i64 40
  %2306 = atomicrmw and ptr addrspace(1) %sunkaddr3293, i32 %.lcssa3873 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %2307, !amdgpu.uniform !42

2307:                                             ; preds = %2304, %ComputeEnd2655
  call void @llvm.amdgcn.end.cf.i64(i64 %3769)
  tail call void @llvm.amdgcn.wave.barrier()
  call void @llvm.lifetime.start.p5(ptr addrspace(5) %buffer.i.i.i) #34, !noalias !141
  %2308 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %RHS)
  %2309 = extractvalue { i1, i64 } %2308, 0
  %2310 = extractvalue { i1, i64 } %2308, 1
  br i1 %2309, label %if.then3.i.i.i.i.i.i.i114.i, label %_ZN22__llvm_libc_22_0_0_git15IntegerToStringIiNS_7details3FmtILh10ELb0ELb0ELb0ELm1EEEE13IntegerWriter25write_unsigned_number_decEjRNS1_22StringBufferWriterImplILb0EEE.exit.i.i.i.i.i94.i

if.then3.i.i.i.i.i.i.i114.i:                      ; preds = %2307
  %arrayidx.i.i.i.i.i.i.i.i115.i = getelementptr inbounds nuw i8, ptr addrspace(5) %buffer.i.i.i, i32 10
  store i8 49, ptr addrspace(5) %arrayidx.i.i.i.i.i.i.i.i115.i, align 2, !tbaa !57, !noalias !141
  br label %_ZN22__llvm_libc_22_0_0_git15IntegerToStringIiNS_7details3FmtILh10ELb0ELb0ELb0ELm1EEEE13IntegerWriter25write_unsigned_number_decEjRNS1_22StringBufferWriterImplILb0EEE.exit.i.i.i.i.i94.i, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git15IntegerToStringIiNS_7details3FmtILh10ELb0ELb0ELb0ELm1EEEE13IntegerWriter25write_unsigned_number_decEjRNS1_22StringBufferWriterImplILb0EEE.exit.i.i.i.i.i94.i: ; preds = %if.then3.i.i.i.i.i.i.i114.i, %2307
  %writer.sroa.10.2.i.i.i.i95.i = phi i64 [ 1, %if.then3.i.i.i.i.i.i.i114.i ], [ 0, %2307 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %2310)
  %2311 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %RHS.inv)
  %2312 = extractvalue { i1, i64 } %2311, 0
  %2313 = extractvalue { i1, i64 } %2311, 1
  br i1 %2312, label %if.then3.i.i.lr.ph.i.i.i.i96.i, label %while.end.i.i.i.i.i98.i

if.then3.i.i.lr.ph.i.i.i.i96.i:                   ; preds = %_ZN22__llvm_libc_22_0_0_git15IntegerToStringIiNS_7details3FmtILh10ELb0ELb0ELb0ELm1EEEE13IntegerWriter25write_unsigned_number_decEjRNS1_22StringBufferWriterImplILb0EEE.exit.i.i.i.i.i94.i
  %arrayidx.i.i.i.i.i.i.i97.i = getelementptr inbounds nuw i8, ptr addrspace(5) %buffer.i.i.i, i32 10
  store i8 48, ptr addrspace(5) %arrayidx.i.i.i.i.i.i.i97.i, align 2, !tbaa !57, !noalias !141
  br label %while.end.i.i.i.i.i98.i, !amdgpu.uniform !42

while.end.i.i.i.i.i98.i:                          ; preds = %if.then3.i.i.lr.ph.i.i.i.i96.i, %_ZN22__llvm_libc_22_0_0_git15IntegerToStringIiNS_7details3FmtILh10ELb0ELb0ELb0ELm1EEEE13IntegerWriter25write_unsigned_number_decEjRNS1_22StringBufferWriterImplILb0EEE.exit.i.i.i.i.i94.i
  %writer.sroa.10.4.i.i.i.i99.i = phi i64 [ 1, %if.then3.i.i.lr.ph.i.i.i.i96.i ], [ %writer.sroa.10.2.i.i.i.i95.i, %_ZN22__llvm_libc_22_0_0_git15IntegerToStringIiNS_7details3FmtILh10ELb0ELb0ELb0ELm1EEEE13IntegerWriter25write_unsigned_number_decEjRNS1_22StringBufferWriterImplILb0EEE.exit.i.i.i.i.i94.i ]
  call void @llvm.amdgcn.end.cf.i64(i64 %2313)
  %written.i.i.i.i100.i = getelementptr inbounds nuw i8, ptr addrspace(5) %buffer.i.i.i, i32 16
  store i64 %writer.sroa.10.4.i.i.i.i99.i, ptr addrspace(5) %written.i.i.i.i100.i, align 8, !tbaa !91, !noalias !141
  %cmp.i.not.i.i.i.i.i103.i = icmp ne i64 %writer.sroa.10.4.i.i.i.i99.i, 0
  %2314 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.i.not.i.i.i.i.i103.i)
  %2315 = extractvalue { i1, i64 } %2314, 0
  %2316 = extractvalue { i1, i64 } %2314, 1
  br i1 %2315, label %if.end4.i.i.i.i.i.i.i104.i, label %if.end.i.i.i.i.i.i107.i

if.end4.i.i.i.i.i.i.i104.i:                       ; preds = %while.end.i.i.i.i.i98.i
  %2317 = load ptr, ptr addrspace(1) @_ZN12_GLOBAL__N_13ptrE, align 8, !tbaa !55
  %add.ptr.i.i1874 = getelementptr inbounds nuw i8, ptr %2317, i64 8
  store ptr %add.ptr.i.i1874, ptr addrspace(1) @_ZN12_GLOBAL__N_13ptrE, align 8, !tbaa !55
  %sub.ptr.lhs.cast.i.i1875 = ptrtoint ptr %add.ptr.i.i1874 to i64
  %sub.ptr.sub.i.i1876 = sub i64 %sub.ptr.lhs.cast.i.i1875, ptrtoint (ptr addrspacecast (ptr addrspace(1) @_ZN12_GLOBAL__N_16memoryE to ptr) to i64)
  %cmp.i.i1877 = icmp ugt i64 %sub.ptr.sub.i.i1876, 65335
  %cond.i.i1878 = select i1 %cmp.i.i1877, ptr null, ptr %2317
  store i8 0, ptr %2317, align 1, !noalias !148
  br label %if.end.i.i.i.i.i.i107.i, !amdgpu.uniform !42

if.end.i.i.i.i.i.i107.i:                          ; preds = %if.end4.i.i.i.i.i.i.i104.i, %while.end.i.i.i.i.i98.i
  %agg.tmp27.i.sroa.0.0 = phi ptr [ %cond.i.i1878, %if.end4.i.i.i.i.i.i.i104.i ], [ addrspacecast (ptr addrspace(4) @_ZN22__llvm_libc_22_0_0_git3cpp6string14NULL_CHARACTERE to ptr), %while.end.i.i.i.i.i98.i ]
  call void @llvm.amdgcn.end.cf.i64(i64 %2316)
  %cmp.not.i6.i.i.i.i.i.i109.i = icmp ne ptr %agg.tmp27.i.sroa.0.0, addrspacecast (ptr addrspace(4) @_ZN22__llvm_libc_22_0_0_git3cpp6string14NULL_CHARACTERE to ptr)
  %2318 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.not.i6.i.i.i.i.i.i109.i)
  %2319 = extractvalue { i1, i64 } %2318, 0
  %2320 = extractvalue { i1, i64 } %2318, 1
  br i1 %2319, label %if.then.i.i.i.i2.i.i.i110.i, label %_ZN22__llvm_libc_22_0_0_git7testing8internal13describeValueIbEENS_3cpp9enable_ifIXaasr3cppE13is_integral_vIT_ElestS5_Lm8EENS3_6stringEE4typeES5_.exit116.i

if.then.i.i.i.i2.i.i.i110.i:                      ; preds = %if.end.i.i.i.i.i.i107.i
  %arrayidx.i.i.i.i3.i.i.i111.i = getelementptr inbounds nuw i8, ptr %agg.tmp27.i.sroa.0.0, i64 %writer.sroa.10.4.i.i.i.i99.i
  store i8 0, ptr %arrayidx.i.i.i.i3.i.i.i111.i, align 1, !tbaa !57, !noalias !141
  br label %_ZN22__llvm_libc_22_0_0_git7testing8internal13describeValueIbEENS_3cpp9enable_ifIXaasr3cppE13is_integral_vIT_ElestS5_Lm8EENS3_6stringEE4typeES5_.exit116.i, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git7testing8internal13describeValueIbEENS_3cpp9enable_ifIXaasr3cppE13is_integral_vIT_ElestS5_Lm8EENS3_6stringEE4typeES5_.exit116.i: ; preds = %if.then.i.i.i.i2.i.i.i110.i, %if.end.i.i.i.i.i.i107.i
  call void @llvm.amdgcn.end.cf.i64(i64 %2320)
  %2321 = trunc i64 %writer.sroa.10.4.i.i.i.i99.i to i32
  %2322 = and i64 %writer.sroa.10.4.i.i.i.i99.i, 15
  %2323 = sub i64 %writer.sroa.10.4.i.i.i.i99.i, %2322
  %2324 = icmp ne i64 %2323, 0
  %2325 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %2324)
  %2326 = extractvalue { i1, i64 } %2325, 0
  %2327 = extractvalue { i1, i64 } %2325, 1
  br i1 %2326, label %loop-memcpy-expansion2297.preheader, label %Flow3343

loop-memcpy-expansion2297.preheader:              ; preds = %_ZN22__llvm_libc_22_0_0_git7testing8internal13describeValueIbEENS_3cpp9enable_ifIXaasr3cppE13is_integral_vIT_ElestS5_Lm8EENS3_6stringEE4typeES5_.exit116.i
  %2328 = sub i32 11, %2321
  %scevgep2850 = getelementptr i8, ptr addrspace(5) %buffer.i.i.i, i32 %2328
  br label %loop-memcpy-expansion2297, !amdgpu.uniform !42

Flow3343:                                         ; preds = %Flow3342, %_ZN22__llvm_libc_22_0_0_git7testing8internal13describeValueIbEENS_3cpp9enable_ifIXaasr3cppE13is_integral_vIT_ElestS5_Lm8EENS3_6stringEE4typeES5_.exit116.i
  call void @llvm.amdgcn.end.cf.i64(i64 %2327)
  br label %loop-memcpy-residual-header2300, !amdgpu.uniform !42

loop-memcpy-expansion2297:                        ; preds = %loop-memcpy-expansion2297.preheader, %loop-memcpy-expansion2297
  %phi.broken3838 = phi i64 [ 0, %loop-memcpy-expansion2297.preheader ], [ %2333, %loop-memcpy-expansion2297 ]
  %lsr.iv2851 = phi ptr addrspace(5) [ %scevgep2852, %loop-memcpy-expansion2297 ], [ %scevgep2850, %loop-memcpy-expansion2297.preheader ]
  %loop-index2298 = phi i64 [ %2331, %loop-memcpy-expansion2297 ], [ 0, %loop-memcpy-expansion2297.preheader ]
  %2329 = load <4 x i32>, ptr addrspace(5) %lsr.iv2851, align 1
  %2330 = getelementptr i8, ptr %agg.tmp27.i.sroa.0.0, i64 %loop-index2298
  store <4 x i32> %2329, ptr %2330, align 1
  %2331 = add i64 %loop-index2298, 16
  %scevgep2852 = getelementptr i8, ptr addrspace(5) %lsr.iv2851, i32 16
  %2332 = icmp uge i64 %2331, %2323
  %2333 = call i64 @llvm.amdgcn.if.break.i64(i1 %2332, i64 %phi.broken3838)
  %2334 = call i1 @llvm.amdgcn.loop.i64(i64 %2333)
  br i1 %2334, label %Flow3342, label %loop-memcpy-expansion2297

Flow3341:                                         ; preds = %Flow3340, %loop-memcpy-residual-header2300
  call void @llvm.amdgcn.end.cf.i64(i64 %2661)
  br label %post-loop-memcpy-expansion2296, !amdgpu.uniform !42

loop-memcpy-residual2299:                         ; preds = %loop-memcpy-residual2299.preheader, %loop-memcpy-residual2299
  %phi.broken3839 = phi i64 [ 0, %loop-memcpy-residual2299.preheader ], [ %2338, %loop-memcpy-residual2299 ]
  %lsr.iv = phi ptr addrspace(5) [ %scevgep2848, %loop-memcpy-residual2299 ], [ %scevgep2847, %loop-memcpy-residual2299.preheader ]
  %residual-loop-index2301 = phi i64 [ %2336, %loop-memcpy-residual2299 ], [ 0, %loop-memcpy-residual2299.preheader ]
  %2335 = load i8, ptr addrspace(5) %lsr.iv, align 1
  %scevgep2849 = getelementptr i8, ptr %agg.tmp27.i.sroa.0.0, i64 %residual-loop-index2301
  store i8 %2335, ptr %scevgep2849, align 1
  %2336 = add i64 %residual-loop-index2301, 1
  %scevgep2848 = getelementptr i8, ptr addrspace(5) %lsr.iv, i32 1
  %2337 = icmp uge i64 %2336, %2322
  %2338 = call i64 @llvm.amdgcn.if.break.i64(i1 %2337, i64 %phi.broken3839)
  %2339 = call i1 @llvm.amdgcn.loop.i64(i64 %2338)
  br i1 %2339, label %Flow3340, label %loop-memcpy-residual2299

Flow3340:                                         ; preds = %loop-memcpy-residual2299
  %.lcssa3871 = phi i64 [ %2338, %loop-memcpy-residual2299 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa3871)
  br label %Flow3341, !amdgpu.uniform !42

post-loop-memcpy-expansion2296:                   ; preds = %Flow3341
  call void @llvm.lifetime.end.p5(ptr addrspace(5) %buffer.i.i.i) #34, !noalias !141
  br label %for.cond.i.i.i1880, !amdgpu.uniform !42

for.cond.i.i.i1880:                               ; preds = %post-loop-memcpy-expansion2296, %Flow3339
  %2340 = phi i32 [ %2378, %Flow3339 ], [ poison, %post-loop-memcpy-expansion2296 ]
  %2341 = phi i64 [ %2379, %Flow3339 ], [ poison, %post-loop-memcpy-expansion2296 ]
  %2342 = phi i32 [ %2380, %Flow3339 ], [ poison, %post-loop-memcpy-expansion2296 ]
  %retval.sroa.6.0.i.i.i1881 = phi i32 [ %2383, %Flow3339 ], [ undef, %post-loop-memcpy-expansion2296 ]
  %retval.sroa.2.0.i.i.i1883 = phi i64 [ %2382, %Flow3339 ], [ undef, %post-loop-memcpy-expansion2296 ]
  %retval.sroa.8.0.i.i.i1884 = phi i32 [ %2381, %Flow3339 ], [ undef, %post-loop-memcpy-expansion2296 ]
  %index.0.i.i.i1885 = phi i32 [ %2385, %Flow3339 ], [ 0, %post-loop-memcpy-expansion2296 ]
  %2343 = load i32, ptr addrspace(1) @__llvm_rpc_client, align 8, !tbaa !68
  %cmp.not.i.i.i1886 = icmp ult i32 %index.0.i.i.i1885, %2343
  %spec.store.select.i.i.i1887 = select i1 %cmp.not.i.i.i1886, i32 %index.0.i.i.i1885, i32 0
  %2344 = tail call noundef i64 @llvm.amdgcn.ballot.i64(i1 true)
  %2345 = and i32 %spec.store.select.i.i.i1887, 31
  %2346 = lshr i64 %2344, %sh_prom.i.i.i.i
  %2347 = trunc i64 %2346 to i32
  %conv4.i.i.i.i.i1891 = and i32 %2347, 1
  %shl.i.i.i.i.i1892 = shl nuw i32 %conv4.i.i.i.i.i1891, %2345
  %2348 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2658, !amdgpu.uniform !42

2349:                                             ; preds = %ComputeEnd2659
  %sunkaddr3294 = getelementptr inbounds i8, ptr addrspace(1) %3779, i64 40
  %2350 = atomicrmw or ptr addrspace(1) %sunkaddr3294, i32 %.lcssa3865 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %2351, !amdgpu.uniform !42

2351:                                             ; preds = %2349, %ComputeEnd2659
  %2352 = phi i32 [ %2350, %2349 ], [ poison, %ComputeEnd2659 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %3788)
  %2353 = call i32 @llvm.amdgcn.readfirstlane.i32(i32 %2352)
  %2354 = or i32 %2353, %.lcssa3866
  %shl5.i.i.i.i.i1893 = shl nuw i32 1, %2345
  %and.i.i.i.i.i1894 = and i32 %shl5.i.i.i.i.i1893, %2354
  %tobool3.i.i.i.i1895 = icmp ne i32 %and.i.i.i.i.i1894, 0
  %2355 = tail call i64 @llvm.amdgcn.ballot.i64(i1 %tobool3.i.i.i.i1895)
  %and.i.i.i.i.i.i1896 = and i64 %2355, %2344
  %cmp.i.not.i.i.i1897 = icmp ne i64 %2344, %and.i.i.i.i.i.i1896
  br i1 %cmp.i.not.i.i.i1897, label %if.then.i.i.i.i1898, label %Flow3338, !amdgpu.uniform !42

if.then.i.i.i.i1898:                              ; preds = %2351
  fence syncscope("agent") acquire
  %2356 = load <2 x i64>, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !75
  %2357 = extractelement <2 x i64> %2356, i32 0
  %2358 = inttoptr i64 %2357 to ptr
  %2359 = extractelement <2 x i64> %2356, i32 1
  %2360 = inttoptr i64 %2359 to ptr
  %idxprom.i.i.i.i1899 = zext i32 %spec.store.select.i.i.i1887 to i64
  %arrayidx.i.i.i.i1900 = getelementptr inbounds nuw i32, ptr %2358, i64 %idxprom.i.i.i.i1899, !amdgpu.uniform !42
  %2361 = load atomic i32, ptr %arrayidx.i.i.i.i1900 monotonic, align 4
  %2362 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %2361)
  %arrayidx.i45.i.i.i1901 = getelementptr inbounds nuw i32, ptr %2360, i64 %idxprom.i.i.i.i1899, !amdgpu.uniform !42
  %2363 = load atomic i32, ptr %arrayidx.i45.i.i.i1901 monotonic, align 4
  %2364 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %2363)
  %cmp.i46.not.i.i.i1902 = icmp ne i32 %2362, %2364
  br i1 %cmp.i46.not.i.i.i1902, label %if.then12.i.i.i1903, label %Flow3336, !amdgpu.uniform !42

if.then12.i.i.i1903:                              ; preds = %if.then.i.i.i.i1898
  fence syncscope("agent") release
  %2365 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %2344, i1 true)
  %iszero.i.i.i.i.i.i.i1904 = icmp ne i64 %2344, 0
  %cmp2.i.i.i.i.i.i1905 = icmp eq i64 %2365, %sh_prom.i.i.i.i
  %cmp.i.i.i.i.i.i1906 = select i1 %iszero.i.i.i.i.i.i.i1904, i1 %cmp2.i.i.i.i.i.i1905, i1 false
  %conv4.i.i51.i.i.i1907 = zext i1 %cmp.i.i.i.i.i.i1906 to i32
  %shl.i.i52.i.i.i1908 = shl nuw i32 %conv4.i.i51.i.i.i1907, %2345
  %xor.i.i.i.i.i1909 = xor i32 %shl.i.i52.i.i.i1908, -1
  %2366 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2663, !amdgpu.uniform !42

2367:                                             ; preds = %ComputeEnd2664
  %sunkaddr3295 = getelementptr inbounds i8, ptr addrspace(1) %3779, i64 40
  %2368 = atomicrmw and ptr addrspace(1) %sunkaddr3295, i32 %.lcssa3867 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %2369, !amdgpu.uniform !42

2369:                                             ; preds = %2367, %ComputeEnd2664
  call void @llvm.amdgcn.end.cf.i64(i64 %3806)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %Flow3336, !amdgpu.uniform !42

if.end14.i.i.i1969:                               ; preds = %Flow3336
  %2370 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %2344, i1 true)
  %iszero.i.i.i.i.i.i1970 = icmp ne i64 %2344, 0
  %cmp2.i.i.i.i.i1971 = icmp eq i64 %2370, %sh_prom.i.i.i.i
  %cmp.i.i.i.i.i1972 = select i1 %iszero.i.i.i.i.i.i1970, i1 %cmp2.i.i.i.i.i1971, i1 false
  %2371 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.i.i.i.i.i1972)
  %2372 = extractvalue { i1, i64 } %2371, 0
  %2373 = extractvalue { i1, i64 } %2371, 1
  br i1 %2372, label %if.then16.i.i.i1974, label %if.end22.i.i.i1973

if.then16.i.i.i1974:                              ; preds = %if.end14.i.i.i1969
  %2374 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !76
  %opcode.i.i.i1975 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %2374, i64 %idxprom.i.i.i.i1899, i32 1
  store i32 1660944387, ptr %opcode.i.i.i1975, align 8, !tbaa !77
  %arrayidx21.i.i.i1976 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %2374, i64 %idxprom.i.i.i.i1899
  store i64 %2344, ptr %arrayidx21.i.i.i1976, align 8, !tbaa !80
  br label %if.end22.i.i.i1973, !amdgpu.uniform !42

Flow3337:                                         ; preds = %if.end22.i.i.i1973, %Flow3336
  %2375 = phi i32 [ %2362, %if.end22.i.i.i1973 ], [ %retval.sroa.8.0.i.i.i1884, %Flow3336 ]
  %2376 = phi i64 [ %2344, %if.end22.i.i.i1973 ], [ %retval.sroa.2.0.i.i.i1883, %Flow3336 ]
  %2377 = phi i32 [ %spec.store.select.i.i.i1887, %if.end22.i.i.i1973 ], [ %retval.sroa.6.0.i.i.i1881, %Flow3336 ]
  br label %cleanup26.i.i.i1910, !amdgpu.uniform !42

if.end22.i.i.i1973:                               ; preds = %if.then16.i.i.i1974, %if.end14.i.i.i1969
  call void @llvm.amdgcn.end.cf.i64(i64 %2373)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %Flow3337, !amdgpu.uniform !42

Flow3338:                                         ; preds = %cleanup26.i.i.i1910, %2351
  %2378 = phi i32 [ %2377, %cleanup26.i.i.i1910 ], [ %2340, %2351 ]
  %2379 = phi i64 [ %2376, %cleanup26.i.i.i1910 ], [ %2341, %2351 ]
  %2380 = phi i32 [ %2375, %cleanup26.i.i.i1910 ], [ %2342, %2351 ]
  %2381 = phi i32 [ %2375, %cleanup26.i.i.i1910 ], [ %retval.sroa.8.0.i.i.i1884, %2351 ]
  %2382 = phi i64 [ %2376, %cleanup26.i.i.i1910 ], [ %retval.sroa.2.0.i.i.i1883, %2351 ]
  %2383 = phi i32 [ %2377, %cleanup26.i.i.i1910 ], [ %retval.sroa.6.0.i.i.i1881, %2351 ]
  %2384 = phi i1 [ %cmp.i46.not.i.i.i1902, %cleanup26.i.i.i1910 ], [ true, %2351 ]
  br i1 %2384, label %for.inc.i.i.i1915, label %Flow3339, !amdgpu.uniform !42

cleanup26.i.i.i1910:                              ; preds = %Flow3337
  br label %Flow3338, !amdgpu.uniform !42

for.inc.i.i.i1915:                                ; preds = %Flow3338
  %inc.i.i.i1920 = add i32 %spec.store.select.i.i.i1887, 1
  br label %Flow3339, !amdgpu.uniform !42

Flow3339:                                         ; preds = %for.inc.i.i.i1915, %Flow3338
  %2385 = phi i32 [ %inc.i.i.i1920, %for.inc.i.i.i1915 ], [ poison, %Flow3338 ]
  %2386 = phi i1 [ false, %for.inc.i.i.i1915 ], [ true, %Flow3338 ]
  br i1 %2386, label %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1921, label %for.cond.i.i.i1880, !amdgpu.uniform !42

_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1921: ; preds = %Flow3339
  %.lcssa3870 = phi i32 [ %2378, %Flow3339 ]
  %.lcssa3869 = phi i64 [ %2379, %Flow3339 ]
  %.lcssa3868 = phi i32 [ %2380, %Flow3339 ]
  fence acquire
  %2387 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %mul.i.i.i.i.i.i1922 = shl i32 %.lcssa3870, 6
  %idxprom.i19.i.i.i.i.i1923 = zext i32 %mul.i.i.i.i.i.i1922 to i64
  %arrayidx.i20.i.i.i.i.i1924 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %2387, i64 %idxprom.i19.i.i.i.i.i1923
  %arrayidx.i22.i.i.i.i.i1925 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %arrayidx.i20.i.i.i.i.i1924, i64 %sh_prom.i.i.i.i
  store i64 %writer.sroa.10.4.i.i.i.i99.i, ptr %arrayidx.i22.i.i.i.i.i1925, align 8, !tbaa !82
  %2388 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %2324)
  %2389 = extractvalue { i1, i64 } %2388, 0
  %2390 = extractvalue { i1, i64 } %2388, 1
  br i1 %2389, label %loop-memcpy-expansion2183.preheader, label %Flow3335

loop-memcpy-expansion2183.preheader:              ; preds = %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1921
  %2391 = add nuw nsw i64 %76, 8
  %scevgep2844 = getelementptr i8, ptr %2387, i64 %2391
  %2392 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i1923, 6
  %scevgep2845 = getelementptr i8, ptr %scevgep2844, i64 %2392
  br label %loop-memcpy-expansion2183, !amdgpu.uniform !42

Flow3335:                                         ; preds = %Flow3334, %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1921
  call void @llvm.amdgcn.end.cf.i64(i64 %2390)
  br label %loop-memcpy-residual-header2186, !amdgpu.uniform !42

loop-memcpy-expansion2183:                        ; preds = %loop-memcpy-expansion2183.preheader, %loop-memcpy-expansion2183
  %phi.broken3840 = phi i64 [ 0, %loop-memcpy-expansion2183.preheader ], [ %2397, %loop-memcpy-expansion2183 ]
  %loop-index2184 = phi i64 [ %2395, %loop-memcpy-expansion2183 ], [ 0, %loop-memcpy-expansion2183.preheader ]
  %2393 = getelementptr i8, ptr %agg.tmp27.i.sroa.0.0, i64 %loop-index2184
  %2394 = load <4 x i32>, ptr %2393, align 1
  %scevgep2846 = getelementptr i8, ptr %scevgep2845, i64 %loop-index2184
  store <4 x i32> %2394, ptr %scevgep2846, align 1
  %2395 = add i64 %loop-index2184, 16
  %2396 = icmp uge i64 %2395, %2323
  %2397 = call i64 @llvm.amdgcn.if.break.i64(i1 %2396, i64 %phi.broken3840)
  %2398 = call i1 @llvm.amdgcn.loop.i64(i64 %2397)
  br i1 %2398, label %Flow3334, label %loop-memcpy-expansion2183

Flow3333:                                         ; preds = %Flow3332, %loop-memcpy-residual-header2186
  call void @llvm.amdgcn.end.cf.i64(i64 %2552)
  br label %post-loop-memcpy-expansion2182, !amdgpu.uniform !42

loop-memcpy-residual2185:                         ; preds = %loop-memcpy-residual2185.preheader, %loop-memcpy-residual2185
  %phi.broken3841 = phi i64 [ 0, %loop-memcpy-residual2185.preheader ], [ %2402, %loop-memcpy-residual2185 ]
  %residual-loop-index2187 = phi i64 [ %2400, %loop-memcpy-residual2185 ], [ 0, %loop-memcpy-residual2185.preheader ]
  %scevgep2843 = getelementptr i8, ptr %agg.tmp27.i.sroa.0.0, i64 %residual-loop-index2187
  %2399 = load i8, ptr %scevgep2843, align 1
  %scevgep2842 = getelementptr i8, ptr %scevgep2841, i64 %residual-loop-index2187
  store i8 %2399, ptr %scevgep2842, align 1
  %2400 = add i64 %residual-loop-index2187, 1
  %2401 = icmp uge i64 %2400, %2322
  %2402 = call i64 @llvm.amdgcn.if.break.i64(i1 %2401, i64 %phi.broken3841)
  %2403 = call i1 @llvm.amdgcn.loop.i64(i64 %2402)
  br i1 %2403, label %Flow3332, label %loop-memcpy-residual2185

Flow3332:                                         ; preds = %loop-memcpy-residual2185
  %.lcssa3863 = phi i64 [ %2402, %loop-memcpy-residual2185 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa3863)
  br label %Flow3333, !amdgpu.uniform !42

post-loop-memcpy-expansion2182:                   ; preds = %Flow3333
  %tobool.not.i.i.i.i.i.i1928 = icmp eq i32 %.lcssa3868, 0
  %conv.i.i.i.i.i.i1929 = zext i1 %tobool.not.i.i.i.i.i.i1928 to i32
  fence release
  %2404 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %idxprom.i23.i.i.i.i.i1930 = zext i32 %.lcssa3870 to i64
  %arrayidx.i24.i.i.i.i.i1931 = getelementptr inbounds nuw i32, ptr %2404, i64 %idxprom.i23.i.i.i.i.i1930
  store atomic i32 %conv.i.i.i.i.i.i1929, ptr %arrayidx.i24.i.i.i.i.i1931 monotonic, align 4
  %2405 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !84
  %arrayidx.i.i5.i.i1932 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %2405, i64 %idxprom.i23.i.i.i.i.i1930, !amdgpu.uniform !42
  %2406 = load i64, ptr %arrayidx.i.i5.i.i1932, align 8, !tbaa !80
  br label %while.cond.i.i.i.i1933, !amdgpu.uniform !42

while.cond.i.i.i.i1933:                           ; preds = %post-loop-memcpy-expansion2182, %Flow3331
  %phi.broken3844 = phi i64 [ 0, %post-loop-memcpy-expansion2182 ], [ %2421, %Flow3331 ]
  %port.sroa.37.0.i.i1934 = phi i32 [ %2419, %Flow3331 ], [ %conv.i.i.i.i.i.i1929, %post-loop-memcpy-expansion2182 ]
  %idx.0.i.i.i.i1935 = phi i64 [ %2418, %Flow3331 ], [ 56, %post-loop-memcpy-expansion2182 ]
  %cmp.i.i.i.i1936 = icmp ult i64 %idx.0.i.i.i.i1935, %writer.sroa.10.4.i.i.i.i99.i
  %2407 = tail call i64 @llvm.amdgcn.ballot.i64(i1 %cmp.i.i.i.i1936)
  %and.i.i.i.i6.i.i1937 = and i64 %2407, %2406
  %tobool.not.i.i.i.i1938 = icmp ne i64 %and.i.i.i.i6.i.i1937, 0
  %2408 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i22.i.i.i1939 = getelementptr inbounds nuw i32, ptr %2408, i64 %idxprom.i23.i.i.i.i.i1930, !amdgpu.uniform !42
  %2409 = load atomic i32, ptr %arrayidx.i22.i.i.i1939 monotonic, align 4
  %2410 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %tobool.not.i.i.i.i1938)
  %2411 = extractvalue { i1, i64 } %2410, 0
  %2412 = extractvalue { i1, i64 } %2410, 1
  br i1 %2411, label %cond.false.i8.i.i.i.i1940, label %Flow3331

cond.false.i8.i.i.i.i1940:                        ; preds = %while.cond.i.i.i.i1933
  %2413 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %2409)
  %2414 = icmp ne i32 %2413, %port.sroa.37.0.i.i1934
  br i1 %2414, label %while.body.i.i21.i.i.i.i1941.preheader, label %Flow3330, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i1941.preheader:           ; preds = %cond.false.i8.i.i.i.i1940
  br label %while.body.i.i21.i.i.i.i1941, !amdgpu.uniform !42

Flow3330:                                         ; preds = %Flow3329, %cond.false.i8.i.i.i.i1940
  br label %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i1944, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i1941:                     ; preds = %while.body.i.i21.i.i.i.i1941.preheader, %while.body.i.i21.i.i.i.i1941
  tail call void @llvm.amdgcn.s.sleep(i32 2)
  %2415 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i22.i.i.i.i1942 = getelementptr inbounds nuw i32, ptr %2415, i64 %idxprom.i23.i.i.i.i.i1930, !amdgpu.uniform !42
  %2416 = load atomic i32, ptr %arrayidx.i.i.i22.i.i.i.i1942 monotonic, align 4
  %2417 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %2416)
  %cmp.i.not.i.i23.i.i.i.i1943 = icmp eq i32 %2417, %port.sroa.37.0.i.i1934
  br i1 %cmp.i.not.i.i23.i.i.i.i1943, label %Flow3329, label %while.body.i.i21.i.i.i.i1941, !amdgpu.uniform !42

Flow3329:                                         ; preds = %while.body.i.i21.i.i.i.i1941
  br label %Flow3330, !amdgpu.uniform !42

Flow3331:                                         ; preds = %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1945, %while.cond.i.i.i.i1933
  %2418 = phi i64 [ %add.i.i.i.i1948, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1945 ], [ poison, %while.cond.i.i.i.i1933 ]
  %2419 = phi i32 [ %conv.i.i26.i.i.i.i1946, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1945 ], [ poison, %while.cond.i.i.i.i1933 ]
  %2420 = phi i1 [ false, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1945 ], [ true, %while.cond.i.i.i.i1933 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %2412)
  %2421 = call i64 @llvm.amdgcn.if.break.i64(i1 %2420, i64 %phi.broken3844)
  %2422 = call i1 @llvm.amdgcn.loop.i64(i64 %2421)
  br i1 %2422, label %cond.false.i.i.i1955, label %while.cond.i.i.i.i1933

_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i1944: ; preds = %Flow3330
  fence acquire
  %2423 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.i.i.i.i1936)
  %2424 = extractvalue { i1, i64 } %2423, 0
  %2425 = extractvalue { i1, i64 } %2423, 1
  br i1 %2424, label %if.then.i.i.i.i.i.i.i1949, label %Flow3328

if.then.i.i.i.i.i.i.i1949:                        ; preds = %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i1944
  %2426 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %arrayidx.i20.i35.i.i.i.i1953 = getelementptr %"struct.rpc::Buffer", ptr %2426, i64 %idxprom.i19.i.i.i.i.i1923
  %arrayidx.i22.i38.i.i.i.i1954 = getelementptr %"struct.rpc::Buffer", ptr %arrayidx.i20.i35.i.i.i.i1953, i64 %sh_prom.i.i.i.i
  %2427 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %2324)
  %2428 = extractvalue { i1, i64 } %2427, 0
  %2429 = extractvalue { i1, i64 } %2427, 1
  br i1 %2428, label %loop-memcpy-expansion.preheader, label %Flow3327

loop-memcpy-expansion.preheader:                  ; preds = %if.then.i.i.i.i.i.i.i1949
  br label %loop-memcpy-expansion, !amdgpu.uniform !42

Flow3327:                                         ; preds = %Flow3326, %if.then.i.i.i.i.i.i.i1949
  call void @llvm.amdgcn.end.cf.i64(i64 %2429)
  br label %loop-memcpy-residual-header, !amdgpu.uniform !42

loop-memcpy-expansion:                            ; preds = %loop-memcpy-expansion.preheader, %loop-memcpy-expansion
  %phi.broken3842 = phi i64 [ 0, %loop-memcpy-expansion.preheader ], [ %2435, %loop-memcpy-expansion ]
  %loop-index = phi i64 [ %2433, %loop-memcpy-expansion ], [ 0, %loop-memcpy-expansion.preheader ]
  %2430 = getelementptr i8, ptr %agg.tmp27.i.sroa.0.0, i64 %loop-index
  %2431 = load <4 x i32>, ptr %2430, align 1
  %2432 = getelementptr i8, ptr %arrayidx.i22.i38.i.i.i.i1954, i64 %loop-index
  store <4 x i32> %2431, ptr %2432, align 1
  %2433 = add i64 %loop-index, 16
  %2434 = icmp uge i64 %2433, %2323
  %2435 = call i64 @llvm.amdgcn.if.break.i64(i1 %2434, i64 %phi.broken3842)
  %2436 = call i1 @llvm.amdgcn.loop.i64(i64 %2435)
  br i1 %2436, label %Flow3326, label %loop-memcpy-expansion

Flow3325:                                         ; preds = %Flow3324, %loop-memcpy-residual-header
  call void @llvm.amdgcn.end.cf.i64(i64 %2549)
  br label %Flow3328, !amdgpu.uniform !42

loop-memcpy-residual:                             ; preds = %loop-memcpy-residual.preheader, %loop-memcpy-residual
  %phi.broken3843 = phi i64 [ 0, %loop-memcpy-residual.preheader ], [ %2441, %loop-memcpy-residual ]
  %residual-loop-index = phi i64 [ %2439, %loop-memcpy-residual ], [ 0, %loop-memcpy-residual.preheader ]
  %scevgep = getelementptr i8, ptr %agg.tmp27.i.sroa.0.0, i64 %residual-loop-index
  %2437 = load i8, ptr %scevgep, align 1
  %2438 = getelementptr i8, ptr %arrayidx.i22.i38.i.i.i.i1954, i64 %residual-loop-index
  store i8 %2437, ptr %2438, align 1
  %2439 = add i64 %residual-loop-index, 1
  %2440 = icmp uge i64 %2439, %2322
  %2441 = call i64 @llvm.amdgcn.if.break.i64(i1 %2440, i64 %phi.broken3843)
  %2442 = call i1 @llvm.amdgcn.loop.i64(i64 %2441)
  br i1 %2442, label %Flow3324, label %loop-memcpy-residual

Flow3324:                                         ; preds = %loop-memcpy-residual
  %.lcssa3860 = phi i64 [ %2441, %loop-memcpy-residual ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa3860)
  br label %Flow3325, !amdgpu.uniform !42

_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1945: ; preds = %Flow3328
  %conv.i.i26.i.i.i.i1946 = xor i32 %port.sroa.37.0.i.i1934, 1
  fence release
  %2443 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %arrayidx.i24.i29.i.i.i.i1947 = getelementptr inbounds nuw i32, ptr %2443, i64 %idxprom.i23.i.i.i.i.i1930
  store atomic i32 %conv.i.i26.i.i.i.i1946, ptr %arrayidx.i24.i29.i.i.i.i1947 monotonic, align 4
  %add.i.i.i.i1948 = add i64 %idx.0.i.i.i.i1935, 64
  br label %Flow3331, !amdgpu.uniform !42

cond.false.i.i.i1955:                             ; preds = %Flow3331
  %.lcssa3862 = phi i64 [ %2421, %Flow3331 ]
  %port.sroa.37.0.i.i1934.lcssa = phi i32 [ %port.sroa.37.0.i.i1934, %Flow3331 ]
  %.lcssa3861 = phi i32 [ %2409, %Flow3331 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa3862)
  %2444 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %.lcssa3861)
  %2445 = icmp ne i32 %2444, %port.sroa.37.0.i.i1934.lcssa
  %2446 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %2445)
  %2447 = extractvalue { i1, i64 } %2446, 0
  %2448 = extractvalue { i1, i64 } %2446, 1
  br i1 %2447, label %while.body.i.i.i.i1956.preheader, label %Flow3323

while.body.i.i.i.i1956.preheader:                 ; preds = %cond.false.i.i.i1955
  br label %while.body.i.i.i.i1956, !amdgpu.uniform !42

Flow3323:                                         ; preds = %Flow3322, %cond.false.i.i.i1955
  call void @llvm.amdgcn.end.cf.i64(i64 %2448)
  br label %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsINS_3cpp6stringEEERS1_T_.exit1977, !amdgpu.uniform !42

while.body.i.i.i.i1956:                           ; preds = %while.body.i.i.i.i1956.preheader, %while.body.i.i.i.i1956
  %phi.broken3845 = phi i64 [ 0, %while.body.i.i.i.i1956.preheader ], [ %2452, %while.body.i.i.i.i1956 ]
  tail call void @llvm.amdgcn.s.sleep(i32 2)
  %2449 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i8.i.i1957 = getelementptr inbounds nuw i32, ptr %2449, i64 %idxprom.i23.i.i.i.i.i1930, !amdgpu.uniform !42
  %2450 = load atomic i32, ptr %arrayidx.i.i.i8.i.i1957 monotonic, align 4
  %2451 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %2450)
  %cmp.i.not.i.i.i.i1958 = icmp eq i32 %2451, %port.sroa.37.0.i.i1934.lcssa
  %2452 = call i64 @llvm.amdgcn.if.break.i64(i1 %cmp.i.not.i.i.i.i1958, i64 %phi.broken3845)
  %2453 = call i1 @llvm.amdgcn.loop.i64(i64 %2452)
  br i1 %2453, label %Flow3322, label %while.body.i.i.i.i1956

Flow3322:                                         ; preds = %while.body.i.i.i.i1956
  %.lcssa3858 = phi i64 [ %2452, %while.body.i.i.i.i1956 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa3858)
  br label %Flow3323, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsINS_3cpp6stringEEERS1_T_.exit1977: ; preds = %Flow3323
  fence acquire
  tail call void @llvm.amdgcn.wave.barrier()
  fence syncscope("agent") release
  %2454 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %.lcssa3869, i1 true)
  %iszero.i.i.i.i.i13.i.i1960 = icmp ne i64 %.lcssa3869, 0
  %cmp2.i.i.i.i14.i.i1961 = icmp eq i64 %2454, %sh_prom.i.i.i.i
  %cmp.i.i.i.i15.i.i1962 = select i1 %iszero.i.i.i.i.i13.i.i1960, i1 %cmp2.i.i.i.i14.i.i1961, i1 false
  %2455 = and i32 %.lcssa3870, 31
  %conv4.i.i.i19.i.i1966 = zext i1 %cmp.i.i.i.i15.i.i1962 to i32
  %shl.i.i.i20.i.i1967 = shl nuw i32 %conv4.i.i.i19.i.i1966, %2455
  %xor.i.i.i21.i.i1968 = xor i32 %shl.i.i.i20.i.i1967, -1
  %2456 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2667, !amdgpu.uniform !42

2457:                                             ; preds = %ComputeEnd2668
  %div8.i.i.i16.i.i1963 = lshr i32 %.lcssa3870, 5
  %idxprom.i.i.i17.i.i1964 = zext nneg i32 %div8.i.i.i16.i.i1963 to i64
  %2458 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i17.i.i1964
  %sunkaddr3296 = getelementptr inbounds i8, ptr addrspace(1) %2458, i64 40
  %2459 = atomicrmw and ptr addrspace(1) %sunkaddr3296, i32 %.lcssa3857 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %2460, !amdgpu.uniform !42

2460:                                             ; preds = %2457, %ComputeEnd2668
  call void @llvm.amdgcn.end.cf.i64(i64 %3823)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %for.cond.i.i.i1981, !amdgpu.uniform !42

for.cond.i.i.i1981:                               ; preds = %2460, %Flow3321
  %2461 = phi i32 [ %2499, %Flow3321 ], [ poison, %2460 ]
  %2462 = phi i64 [ %2500, %Flow3321 ], [ poison, %2460 ]
  %2463 = phi i32 [ %2501, %Flow3321 ], [ poison, %2460 ]
  %retval.sroa.6.0.i.i.i1982 = phi i32 [ %2504, %Flow3321 ], [ undef, %2460 ]
  %retval.sroa.2.0.i.i.i1984 = phi i64 [ %2503, %Flow3321 ], [ undef, %2460 ]
  %retval.sroa.8.0.i.i.i1985 = phi i32 [ %2502, %Flow3321 ], [ undef, %2460 ]
  %index.0.i.i.i1986 = phi i32 [ %2506, %Flow3321 ], [ 0, %2460 ]
  %2464 = load i32, ptr addrspace(1) @__llvm_rpc_client, align 8, !tbaa !68
  %cmp.not.i.i.i1987 = icmp ult i32 %index.0.i.i.i1986, %2464
  %spec.store.select.i.i.i1988 = select i1 %cmp.not.i.i.i1987, i32 %index.0.i.i.i1986, i32 0
  %2465 = tail call noundef i64 @llvm.amdgcn.ballot.i64(i1 true)
  %2466 = and i32 %spec.store.select.i.i.i1988, 31
  %2467 = lshr i64 %2465, %sh_prom.i.i.i.i
  %2468 = trunc i64 %2467 to i32
  %conv4.i.i.i.i.i1992 = and i32 %2468, 1
  %shl.i.i.i.i.i1993 = shl nuw i32 %conv4.i.i.i.i.i1992, %2466
  %2469 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2671, !amdgpu.uniform !42

2470:                                             ; preds = %ComputeEnd2672
  %sunkaddr3297 = getelementptr inbounds i8, ptr addrspace(1) %3833, i64 40
  %2471 = atomicrmw or ptr addrspace(1) %sunkaddr3297, i32 %.lcssa3851 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %2472, !amdgpu.uniform !42

2472:                                             ; preds = %2470, %ComputeEnd2672
  %2473 = phi i32 [ %2471, %2470 ], [ poison, %ComputeEnd2672 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %3842)
  %2474 = call i32 @llvm.amdgcn.readfirstlane.i32(i32 %2473)
  %2475 = or i32 %2474, %.lcssa3852
  %shl5.i.i.i.i.i1994 = shl nuw i32 1, %2466
  %and.i.i.i.i.i1995 = and i32 %shl5.i.i.i.i.i1994, %2475
  %tobool3.i.i.i.i1996 = icmp ne i32 %and.i.i.i.i.i1995, 0
  %2476 = tail call i64 @llvm.amdgcn.ballot.i64(i1 %tobool3.i.i.i.i1996)
  %and.i.i.i.i.i.i1997 = and i64 %2476, %2465
  %cmp.i.not.i.i.i1998 = icmp ne i64 %2465, %and.i.i.i.i.i.i1997
  br i1 %cmp.i.not.i.i.i1998, label %if.then.i.i.i.i1999, label %Flow3320, !amdgpu.uniform !42

if.then.i.i.i.i1999:                              ; preds = %2472
  fence syncscope("agent") acquire
  %2477 = load <2 x i64>, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !75
  %2478 = extractelement <2 x i64> %2477, i32 0
  %2479 = inttoptr i64 %2478 to ptr
  %2480 = extractelement <2 x i64> %2477, i32 1
  %2481 = inttoptr i64 %2480 to ptr
  %idxprom.i.i.i.i2000 = zext i32 %spec.store.select.i.i.i1988 to i64
  %arrayidx.i.i.i.i2001 = getelementptr inbounds nuw i32, ptr %2479, i64 %idxprom.i.i.i.i2000, !amdgpu.uniform !42
  %2482 = load atomic i32, ptr %arrayidx.i.i.i.i2001 monotonic, align 4
  %2483 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %2482)
  %arrayidx.i45.i.i.i2002 = getelementptr inbounds nuw i32, ptr %2481, i64 %idxprom.i.i.i.i2000, !amdgpu.uniform !42
  %2484 = load atomic i32, ptr %arrayidx.i45.i.i.i2002 monotonic, align 4
  %2485 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %2484)
  %cmp.i46.not.i.i.i2003 = icmp ne i32 %2483, %2485
  br i1 %cmp.i46.not.i.i.i2003, label %if.then12.i.i.i2004, label %Flow3318, !amdgpu.uniform !42

if.then12.i.i.i2004:                              ; preds = %if.then.i.i.i.i1999
  fence syncscope("agent") release
  %2486 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %2465, i1 true)
  %iszero.i.i.i.i.i.i.i2005 = icmp ne i64 %2465, 0
  %cmp2.i.i.i.i.i.i2006 = icmp eq i64 %2486, %sh_prom.i.i.i.i
  %cmp.i.i.i.i.i.i2007 = select i1 %iszero.i.i.i.i.i.i.i2005, i1 %cmp2.i.i.i.i.i.i2006, i1 false
  %conv4.i.i51.i.i.i2008 = zext i1 %cmp.i.i.i.i.i.i2007 to i32
  %shl.i.i52.i.i.i2009 = shl nuw i32 %conv4.i.i51.i.i.i2008, %2466
  %xor.i.i.i.i.i2010 = xor i32 %shl.i.i52.i.i.i2009, -1
  %2487 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2676, !amdgpu.uniform !42

2488:                                             ; preds = %ComputeEnd2677
  %sunkaddr3298 = getelementptr inbounds i8, ptr addrspace(1) %3833, i64 40
  %2489 = atomicrmw and ptr addrspace(1) %sunkaddr3298, i32 %.lcssa3853 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %2490, !amdgpu.uniform !42

2490:                                             ; preds = %2488, %ComputeEnd2677
  call void @llvm.amdgcn.end.cf.i64(i64 %3860)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %Flow3318, !amdgpu.uniform !42

if.end14.i.i.i2068:                               ; preds = %Flow3318
  %2491 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %2465, i1 true)
  %iszero.i.i.i.i.i.i2069 = icmp ne i64 %2465, 0
  %cmp2.i.i.i.i.i2070 = icmp eq i64 %2491, %sh_prom.i.i.i.i
  %cmp.i.i.i.i.i2071 = select i1 %iszero.i.i.i.i.i.i2069, i1 %cmp2.i.i.i.i.i2070, i1 false
  %2492 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.i.i.i.i.i2071)
  %2493 = extractvalue { i1, i64 } %2492, 0
  %2494 = extractvalue { i1, i64 } %2492, 1
  br i1 %2493, label %if.then16.i.i.i2073, label %if.end22.i.i.i2072

if.then16.i.i.i2073:                              ; preds = %if.end14.i.i.i2068
  %2495 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !76
  %opcode.i.i.i2074 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %2495, i64 %idxprom.i.i.i.i2000, i32 1
  store i32 1660944387, ptr %opcode.i.i.i2074, align 8, !tbaa !77
  %arrayidx21.i.i.i2075 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %2495, i64 %idxprom.i.i.i.i2000
  store i64 %2465, ptr %arrayidx21.i.i.i2075, align 8, !tbaa !80
  br label %if.end22.i.i.i2072, !amdgpu.uniform !42

Flow3319:                                         ; preds = %if.end22.i.i.i2072, %Flow3318
  %2496 = phi i32 [ %2483, %if.end22.i.i.i2072 ], [ %retval.sroa.8.0.i.i.i1985, %Flow3318 ]
  %2497 = phi i64 [ %2465, %if.end22.i.i.i2072 ], [ %retval.sroa.2.0.i.i.i1984, %Flow3318 ]
  %2498 = phi i32 [ %spec.store.select.i.i.i1988, %if.end22.i.i.i2072 ], [ %retval.sroa.6.0.i.i.i1982, %Flow3318 ]
  br label %cleanup26.i.i.i2011, !amdgpu.uniform !42

if.end22.i.i.i2072:                               ; preds = %if.then16.i.i.i2073, %if.end14.i.i.i2068
  call void @llvm.amdgcn.end.cf.i64(i64 %2494)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %Flow3319, !amdgpu.uniform !42

Flow3320:                                         ; preds = %cleanup26.i.i.i2011, %2472
  %2499 = phi i32 [ %2498, %cleanup26.i.i.i2011 ], [ %2461, %2472 ]
  %2500 = phi i64 [ %2497, %cleanup26.i.i.i2011 ], [ %2462, %2472 ]
  %2501 = phi i32 [ %2496, %cleanup26.i.i.i2011 ], [ %2463, %2472 ]
  %2502 = phi i32 [ %2496, %cleanup26.i.i.i2011 ], [ %retval.sroa.8.0.i.i.i1985, %2472 ]
  %2503 = phi i64 [ %2497, %cleanup26.i.i.i2011 ], [ %retval.sroa.2.0.i.i.i1984, %2472 ]
  %2504 = phi i32 [ %2498, %cleanup26.i.i.i2011 ], [ %retval.sroa.6.0.i.i.i1982, %2472 ]
  %2505 = phi i1 [ %cmp.i46.not.i.i.i2003, %cleanup26.i.i.i2011 ], [ true, %2472 ]
  br i1 %2505, label %for.inc.i.i.i2016, label %Flow3321, !amdgpu.uniform !42

cleanup26.i.i.i2011:                              ; preds = %Flow3319
  br label %Flow3320, !amdgpu.uniform !42

for.inc.i.i.i2016:                                ; preds = %Flow3320
  %inc.i.i.i2021 = add i32 %spec.store.select.i.i.i1988, 1
  br label %Flow3321, !amdgpu.uniform !42

Flow3321:                                         ; preds = %for.inc.i.i.i2016, %Flow3320
  %2506 = phi i32 [ %inc.i.i.i2021, %for.inc.i.i.i2016 ], [ poison, %Flow3320 ]
  %2507 = phi i1 [ false, %for.inc.i.i.i2016 ], [ true, %Flow3320 ]
  br i1 %2507, label %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i2022, label %for.cond.i.i.i1981, !amdgpu.uniform !42

_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i2022: ; preds = %Flow3321
  %.lcssa3856 = phi i32 [ %2499, %Flow3321 ]
  %.lcssa3855 = phi i64 [ %2500, %Flow3321 ]
  %.lcssa3854 = phi i32 [ %2501, %Flow3321 ]
  fence acquire
  %2508 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %mul.i.i.i.i.i.i2023 = shl i32 %.lcssa3856, 6
  %idxprom.i19.i.i.i.i.i2024 = zext i32 %mul.i.i.i.i.i.i2023 to i64
  %arrayidx.i20.i.i.i.i.i2025 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %2508, i64 %idxprom.i19.i.i.i.i.i2024
  %arrayidx.i22.i.i.i.i.i2026 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %arrayidx.i20.i.i.i.i.i2025, i64 %sh_prom.i.i.i.i
  store i64 1, ptr %arrayidx.i22.i.i.i.i.i2026, align 8, !tbaa !82
  %arrayidx6.i.i.i.i.i.i.i2027 = getelementptr inbounds nuw i8, ptr %arrayidx.i22.i.i.i.i.i2026, i64 8
  store i8 10, ptr %arrayidx6.i.i.i.i.i.i.i2027, align 8
  %tobool.not.i.i.i.i.i.i2028 = icmp eq i32 %.lcssa3854, 0
  %conv.i.i.i.i.i.i2029 = zext i1 %tobool.not.i.i.i.i.i.i2028 to i32
  fence release
  %2509 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %idxprom.i23.i.i.i.i.i2030 = zext i32 %.lcssa3856 to i64
  %arrayidx.i24.i.i.i.i.i2031 = getelementptr inbounds nuw i32, ptr %2509, i64 %idxprom.i23.i.i.i.i.i2030
  store atomic i32 %conv.i.i.i.i.i.i2029, ptr %arrayidx.i24.i.i.i.i.i2031 monotonic, align 4
  %2510 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !84
  %arrayidx.i.i5.i.i2032 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %2510, i64 %idxprom.i23.i.i.i.i.i2030, !amdgpu.uniform !42
  %2511 = load i64, ptr %arrayidx.i.i5.i.i2032, align 8, !tbaa !80
  br label %while.cond.i.i.i.i2033, !amdgpu.uniform !42

while.cond.i.i.i.i2033:                           ; preds = %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i2022, %Flow3317
  %phi.broken3846 = phi i64 [ 0, %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i2022 ], [ %2526, %Flow3317 ]
  %port.sroa.37.0.i.i2034 = phi i32 [ %2524, %Flow3317 ], [ %conv.i.i.i.i.i.i2029, %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i2022 ]
  %idx.0.i.i.i.i2035 = phi i64 [ %2523, %Flow3317 ], [ 56, %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i2022 ]
  %cmp.i.i.i.i2036 = icmp eq i64 %idx.0.i.i.i.i2035, 0
  %2512 = tail call i64 @llvm.amdgcn.ballot.i64(i1 %cmp.i.i.i.i2036)
  %and.i.i.i.i6.i.i2037 = and i64 %2512, %2511
  %tobool.not.i.i.i.i2038 = icmp ne i64 %and.i.i.i.i6.i.i2037, 0
  %2513 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i22.i.i.i2039 = getelementptr inbounds nuw i32, ptr %2513, i64 %idxprom.i23.i.i.i.i.i2030, !amdgpu.uniform !42
  %2514 = load atomic i32, ptr %arrayidx.i22.i.i.i2039 monotonic, align 4
  %2515 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %tobool.not.i.i.i.i2038)
  %2516 = extractvalue { i1, i64 } %2515, 0
  %2517 = extractvalue { i1, i64 } %2515, 1
  br i1 %2516, label %cond.false.i8.i.i.i.i2040, label %Flow3317

cond.false.i8.i.i.i.i2040:                        ; preds = %while.cond.i.i.i.i2033
  %2518 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %2514)
  %2519 = icmp ne i32 %2518, %port.sroa.37.0.i.i2034
  br i1 %2519, label %while.body.i.i21.i.i.i.i2041.preheader, label %Flow3316, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i2041.preheader:           ; preds = %cond.false.i8.i.i.i.i2040
  br label %while.body.i.i21.i.i.i.i2041, !amdgpu.uniform !42

Flow3316:                                         ; preds = %Flow3315, %cond.false.i8.i.i.i.i2040
  br label %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i2044, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i2041:                     ; preds = %while.body.i.i21.i.i.i.i2041.preheader, %while.body.i.i21.i.i.i.i2041
  tail call void @llvm.amdgcn.s.sleep(i32 2)
  %2520 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i22.i.i.i.i2042 = getelementptr inbounds nuw i32, ptr %2520, i64 %idxprom.i23.i.i.i.i.i2030, !amdgpu.uniform !42
  %2521 = load atomic i32, ptr %arrayidx.i.i.i22.i.i.i.i2042 monotonic, align 4
  %2522 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %2521)
  %cmp.i.not.i.i23.i.i.i.i2043 = icmp eq i32 %2522, %port.sroa.37.0.i.i2034
  br i1 %cmp.i.not.i.i23.i.i.i.i2043, label %Flow3315, label %while.body.i.i21.i.i.i.i2041, !amdgpu.uniform !42

Flow3315:                                         ; preds = %while.body.i.i21.i.i.i.i2041
  br label %Flow3316, !amdgpu.uniform !42

Flow3317:                                         ; preds = %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i2045, %while.cond.i.i.i.i2033
  %2523 = phi i64 [ %add.i.i.i.i2048, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i2045 ], [ poison, %while.cond.i.i.i.i2033 ]
  %2524 = phi i32 [ %conv.i.i26.i.i.i.i2046, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i2045 ], [ poison, %while.cond.i.i.i.i2033 ]
  %2525 = phi i1 [ false, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i2045 ], [ true, %while.cond.i.i.i.i2033 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %2517)
  %2526 = call i64 @llvm.amdgcn.if.break.i64(i1 %2525, i64 %phi.broken3846)
  %2527 = call i1 @llvm.amdgcn.loop.i64(i64 %2526)
  br i1 %2527, label %cond.false.i.i.i2054, label %while.cond.i.i.i.i2033

_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i2044: ; preds = %Flow3316
  fence acquire
  br i1 %cmp.i.i.i.i2036, label %if.then.i.i.i.i.i.i.i2049, label %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i2045, !amdgpu.uniform !42

if.then.i.i.i.i.i.i.i2049:                        ; preds = %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i2044
  %2528 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %arrayidx.i20.i35.i.i.i.i2052 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %2528, i64 %idxprom.i19.i.i.i.i.i2024
  %arrayidx.i22.i38.i.i.i.i2053 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %arrayidx.i20.i35.i.i.i.i2052, i64 %sh_prom.i.i.i.i
  store i8 10, ptr %arrayidx.i22.i38.i.i.i.i2053, align 1
  br label %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i2045, !amdgpu.uniform !42

_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i2045: ; preds = %if.then.i.i.i.i.i.i.i2049, %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i2044
  %conv.i.i26.i.i.i.i2046 = xor i32 %port.sroa.37.0.i.i2034, 1
  fence release
  %2529 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %arrayidx.i24.i29.i.i.i.i2047 = getelementptr inbounds nuw i32, ptr %2529, i64 %idxprom.i23.i.i.i.i.i2030
  store atomic i32 %conv.i.i26.i.i.i.i2046, ptr %arrayidx.i24.i29.i.i.i.i2047 monotonic, align 4
  %add.i.i.i.i2048 = add i64 %idx.0.i.i.i.i2035, 64
  br label %Flow3317, !amdgpu.uniform !42

cond.false.i.i.i2054:                             ; preds = %Flow3317
  %.lcssa3850 = phi i64 [ %2526, %Flow3317 ]
  %port.sroa.37.0.i.i2034.lcssa = phi i32 [ %port.sroa.37.0.i.i2034, %Flow3317 ]
  %.lcssa3849 = phi i32 [ %2514, %Flow3317 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa3850)
  %2530 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %.lcssa3849)
  %2531 = icmp ne i32 %2530, %port.sroa.37.0.i.i2034.lcssa
  %2532 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %2531)
  %2533 = extractvalue { i1, i64 } %2532, 0
  %2534 = extractvalue { i1, i64 } %2532, 1
  br i1 %2533, label %while.body.i.i.i.i2055.preheader, label %Flow3314

while.body.i.i.i.i2055.preheader:                 ; preds = %cond.false.i.i.i2054
  br label %while.body.i.i.i.i2055, !amdgpu.uniform !42

Flow3314:                                         ; preds = %Flow, %cond.false.i.i.i2054
  call void @llvm.amdgcn.end.cf.i64(i64 %2534)
  br label %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIcEERS1_T_.exit2076, !amdgpu.uniform !42

while.body.i.i.i.i2055:                           ; preds = %while.body.i.i.i.i2055.preheader, %while.body.i.i.i.i2055
  %phi.broken3847 = phi i64 [ 0, %while.body.i.i.i.i2055.preheader ], [ %2538, %while.body.i.i.i.i2055 ]
  tail call void @llvm.amdgcn.s.sleep(i32 2)
  %2535 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i8.i.i2056 = getelementptr inbounds nuw i32, ptr %2535, i64 %idxprom.i23.i.i.i.i.i2030, !amdgpu.uniform !42
  %2536 = load atomic i32, ptr %arrayidx.i.i.i8.i.i2056 monotonic, align 4
  %2537 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %2536)
  %cmp.i.not.i.i.i.i2057 = icmp eq i32 %2537, %port.sroa.37.0.i.i2034.lcssa
  %2538 = call i64 @llvm.amdgcn.if.break.i64(i1 %cmp.i.not.i.i.i.i2057, i64 %phi.broken3847)
  %2539 = call i1 @llvm.amdgcn.loop.i64(i64 %2538)
  br i1 %2539, label %Flow, label %while.body.i.i.i.i2055

Flow:                                             ; preds = %while.body.i.i.i.i2055
  %.lcssa3848 = phi i64 [ %2538, %while.body.i.i.i.i2055 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa3848)
  br label %Flow3314, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIcEERS1_T_.exit2076: ; preds = %Flow3314
  fence acquire
  tail call void @llvm.amdgcn.wave.barrier()
  fence syncscope("agent") release
  %2540 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %.lcssa3855, i1 true)
  %iszero.i.i.i.i.i13.i.i2059 = icmp ne i64 %.lcssa3855, 0
  %cmp2.i.i.i.i14.i.i2060 = icmp eq i64 %2540, %sh_prom.i.i.i.i
  %cmp.i.i.i.i15.i.i2061 = select i1 %iszero.i.i.i.i.i13.i.i2059, i1 %cmp2.i.i.i.i14.i.i2060, i1 false
  %2541 = and i32 %.lcssa3856, 31
  %conv4.i.i.i19.i.i2065 = zext i1 %cmp.i.i.i.i15.i.i2061 to i32
  %shl.i.i.i20.i.i2066 = shl nuw i32 %conv4.i.i.i19.i.i2065, %2541
  %xor.i.i.i21.i.i2067 = xor i32 %shl.i.i.i20.i.i2066, -1
  %2542 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2680, !amdgpu.uniform !42

2543:                                             ; preds = %ComputeEnd2681
  %div8.i.i.i16.i.i2062 = lshr i32 %.lcssa3856, 5
  %idxprom.i.i.i17.i.i2063 = zext nneg i32 %div8.i.i.i16.i.i2062 to i64
  %2544 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i17.i.i2063
  %sunkaddr3299 = getelementptr inbounds i8, ptr addrspace(1) %2544, i64 40
  %2545 = atomicrmw and ptr addrspace(1) %sunkaddr3299, i32 %.lcssa syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %2546, !amdgpu.uniform !42

2546:                                             ; preds = %2543, %ComputeEnd2681
  call void @llvm.amdgcn.end.cf.i64(i64 %3877)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %Flow3776, !amdgpu.uniform !42

_ZZN22__llvm_libc_22_0_0_git7testing8internal4testIbEEbPNS1_10RunContextENS0_8TestCondET_S6_PKcS8_NS1_8LocationEENKUlbNS_3cpp11string_viewEE_clEbSB_.exit: ; preds = %Flow3776
  %cmp = xor i1 %0, true
  ret i1 %cmp

Flow3326:                                         ; preds = %loop-memcpy-expansion
  %.lcssa3859 = phi i64 [ %2435, %loop-memcpy-expansion ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa3859)
  br label %Flow3327, !amdgpu.uniform !42

Flow3328:                                         ; preds = %Flow3325, %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i1944
  call void @llvm.amdgcn.end.cf.i64(i64 %2425)
  br label %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1945, !amdgpu.uniform !42

loop-memcpy-residual-header:                      ; preds = %Flow3327
  %2547 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %2658)
  %2548 = extractvalue { i1, i64 } %2547, 0
  %2549 = extractvalue { i1, i64 } %2547, 1
  br i1 %2548, label %loop-memcpy-residual.preheader, label %Flow3325

loop-memcpy-residual.preheader:                   ; preds = %loop-memcpy-residual-header
  br label %loop-memcpy-residual, !amdgpu.uniform !42

Flow3334:                                         ; preds = %loop-memcpy-expansion2183
  %.lcssa3864 = phi i64 [ %2397, %loop-memcpy-expansion2183 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa3864)
  br label %Flow3335, !amdgpu.uniform !42

loop-memcpy-residual-header2186:                  ; preds = %Flow3335
  %2550 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %2658)
  %2551 = extractvalue { i1, i64 } %2550, 0
  %2552 = extractvalue { i1, i64 } %2550, 1
  br i1 %2551, label %loop-memcpy-residual2185.preheader, label %Flow3333

loop-memcpy-residual2185.preheader:               ; preds = %loop-memcpy-residual-header2186
  %2553 = add nuw nsw i64 %76, 8
  %scevgep2840 = getelementptr i8, ptr %2387, i64 %2553
  %2554 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i1923, 6
  %scevgep2841 = getelementptr i8, ptr %scevgep2840, i64 %2554
  br label %loop-memcpy-residual2185, !amdgpu.uniform !42

Flow3371:                                         ; preds = %loop-memcpy-expansion2189
  br label %Flow3372, !amdgpu.uniform !42

Flow3373:                                         ; preds = %Flow3369, %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i1734
  br label %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1735, !amdgpu.uniform !42

loop-memcpy-residual-header2192:                  ; preds = %Flow3372
  %2555 = icmp ne i64 %2172, 0
  br i1 %2555, label %loop-memcpy-residual2191.preheader, label %Flow3369, !amdgpu.uniform !42

loop-memcpy-residual2191.preheader:               ; preds = %loop-memcpy-residual-header2192
  %umin2879 = call i64 @llvm.umin.i64(i64 %lsr.iv2877, i64 64)
  %2556 = lshr i64 %umin2879, 4
  %2557 = shl nuw nsw i64 %2556, 4
  %scevgep2882 = getelementptr i8, ptr %lsr.iv2871, i64 %2557
  %scevgep2874 = getelementptr i8, ptr %2171, i64 %76
  %scevgep2875 = getelementptr i8, ptr %scevgep2874, i64 %2146
  %scevgep2880 = getelementptr i8, ptr %scevgep2875, i64 %2557
  br label %loop-memcpy-residual2191, !amdgpu.uniform !42

Flow3384:                                         ; preds = %loop-memcpy-expansion2195
  br label %Flow3385, !amdgpu.uniform !42

loop-memcpy-residual-header2198:                  ; preds = %Flow3385
  br i1 %2620, label %loop-memcpy-residual-header2198.post-loop-memcpy-expansion2194_crit_edge, label %Flow3381, !amdgpu.uniform !42

loop-memcpy-residual-header2198.post-loop-memcpy-expansion2194_crit_edge: ; preds = %loop-memcpy-residual-header2198
  %.pre3234 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i1713, 6
  br label %Flow3381, !amdgpu.uniform !42

Flow3381:                                         ; preds = %loop-memcpy-residual-header2198.post-loop-memcpy-expansion2194_crit_edge, %loop-memcpy-residual-header2198
  %2558 = phi i64 [ %.pre3234, %loop-memcpy-residual-header2198.post-loop-memcpy-expansion2194_crit_edge ], [ poison, %loop-memcpy-residual-header2198 ]
  %2559 = phi i1 [ false, %loop-memcpy-residual-header2198.post-loop-memcpy-expansion2194_crit_edge ], [ true, %loop-memcpy-residual-header2198 ]
  br i1 %2559, label %loop-memcpy-residual2197.preheader, label %Flow3382, !amdgpu.uniform !42

loop-memcpy-residual2197.preheader:               ; preds = %Flow3381
  %2560 = lshr i64 %spec.select.i.i.i.i.i.i.i372, 4
  %2561 = shl nuw nsw i64 %2560, 4
  %2562 = add i64 %76, %2561
  %2563 = add i64 %2562, 8
  %scevgep2884 = getelementptr i8, ptr %2139, i64 %2563
  %2564 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i1713, 6
  %scevgep2885 = getelementptr i8, ptr %scevgep2884, i64 %2564
  %scevgep2887 = getelementptr i8, ptr %2083, i64 %2561
  br label %loop-memcpy-residual2197, !amdgpu.uniform !42

Flow3414:                                         ; preds = %loop-memcpy-expansion2201
  %.lcssa3905 = phi i64 [ %1972, %loop-memcpy-expansion2201 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa3905)
  br label %Flow3415, !amdgpu.uniform !42

Flow3416:                                         ; preds = %Flow3412, %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i1529
  call void @llvm.amdgcn.end.cf.i64(i64 %1960)
  br label %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1530, !amdgpu.uniform !42

loop-memcpy-residual-header2204:                  ; preds = %Flow3415
  %2565 = icmp ne i64 %1962, 0
  %2566 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %2565)
  %2567 = extractvalue { i1, i64 } %2566, 0
  %2568 = extractvalue { i1, i64 } %2566, 1
  br i1 %2567, label %loop-memcpy-residual2203.preheader, label %Flow3412

loop-memcpy-residual2203.preheader:               ; preds = %loop-memcpy-residual-header2204
  %umin2910 = call i64 @llvm.umin.i64(i64 %lsr.iv2905, i64 64)
  %2569 = lshr i64 %umin2910, 4
  %2570 = shl nuw nsw i64 %2569, 4
  %scevgep2911 = getelementptr i8, ptr %lsr.iv2898, i64 %2570
  %scevgep2901 = getelementptr i8, ptr %1961, i64 %76
  %scevgep2902 = getelementptr i8, ptr %scevgep2901, i64 %1930
  %scevgep2908 = getelementptr i8, ptr %scevgep2902, i64 %2570
  br label %loop-memcpy-residual2203, !amdgpu.uniform !42

Flow3427:                                         ; preds = %loop-memcpy-expansion2207
  %.lcssa3910 = phi i64 [ %1928, %loop-memcpy-expansion2207 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa3910)
  br label %Flow3428, !amdgpu.uniform !42

loop-memcpy-residual-header2210:                  ; preds = %Flow3428
  %2571 = icmp eq i64 %1916, 0
  %2572 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %2571)
  %2573 = extractvalue { i1, i64 } %2572, 0
  %2574 = extractvalue { i1, i64 } %2572, 1
  br i1 %2573, label %loop-memcpy-residual-header2210.post-loop-memcpy-expansion2206_crit_edge, label %Flow3424

loop-memcpy-residual-header2210.post-loop-memcpy-expansion2206_crit_edge: ; preds = %loop-memcpy-residual-header2210
  %.pre3232 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i1508, 6
  br label %Flow3424, !amdgpu.uniform !42

Flow3424:                                         ; preds = %loop-memcpy-residual-header2210.post-loop-memcpy-expansion2206_crit_edge, %loop-memcpy-residual-header2210
  %2575 = phi i64 [ %.pre3232, %loop-memcpy-residual-header2210.post-loop-memcpy-expansion2206_crit_edge ], [ poison, %loop-memcpy-residual-header2210 ]
  %2576 = call { i1, i64 } @llvm.amdgcn.else.i64.i64(i64 %2574)
  %2577 = extractvalue { i1, i64 } %2576, 0
  %2578 = extractvalue { i1, i64 } %2576, 1
  br i1 %2577, label %loop-memcpy-residual2209.preheader, label %Flow3425

loop-memcpy-residual2209.preheader:               ; preds = %Flow3424
  %2579 = add nuw nsw i64 %76, 8
  %scevgep2913 = getelementptr i8, ptr %1915, i64 %2579
  %umin2916 = call i64 @llvm.umin.i64(i64 %lsr.iv.next2915.lcssa, i64 56)
  %2580 = lshr i64 %umin2916, 4
  %2581 = shl nuw nsw i64 %2580, 4
  %scevgep2917 = getelementptr i8, ptr %scevgep2913, i64 %2581
  %2582 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i1508, 6
  %scevgep2918 = getelementptr i8, ptr %scevgep2917, i64 %2582
  %scevgep2920 = getelementptr i8, ptr %RHSStr, i64 %2581
  br label %loop-memcpy-residual2209, !amdgpu.uniform !42

Flow3529:                                         ; preds = %loop-memcpy-expansion2213
  %.lcssa3960 = phi i64 [ %1419, %loop-memcpy-expansion2213 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa3960)
  br label %Flow3530, !amdgpu.uniform !42

Flow3531:                                         ; preds = %Flow3527, %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i1013
  call void @llvm.amdgcn.end.cf.i64(i64 %1409)
  br label %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1014, !amdgpu.uniform !42

loop-memcpy-residual-header2216:                  ; preds = %Flow3530
  %2583 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %2663)
  %2584 = extractvalue { i1, i64 } %2583, 0
  %2585 = extractvalue { i1, i64 } %2583, 1
  br i1 %2584, label %loop-memcpy-residual2215.preheader, label %Flow3527

loop-memcpy-residual2215.preheader:               ; preds = %loop-memcpy-residual-header2216
  br label %loop-memcpy-residual2215, !amdgpu.uniform !42

Flow3540:                                         ; preds = %loop-memcpy-expansion2219
  %.lcssa3965 = phi i64 [ %1381, %loop-memcpy-expansion2219 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa3965)
  br label %Flow3541, !amdgpu.uniform !42

loop-memcpy-residual-header2222:                  ; preds = %Flow3541
  %2586 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %2663)
  %2587 = extractvalue { i1, i64 } %2586, 0
  %2588 = extractvalue { i1, i64 } %2586, 1
  br i1 %2587, label %loop-memcpy-residual2221.preheader, label %Flow3538

loop-memcpy-residual2221.preheader:               ; preds = %loop-memcpy-residual-header2222
  %2589 = add nuw nsw i64 %76, 8
  %scevgep2987 = getelementptr i8, ptr %1371, i64 %2589
  %2590 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i992, 6
  %scevgep2988 = getelementptr i8, ptr %scevgep2987, i64 %2590
  br label %loop-memcpy-residual2221, !amdgpu.uniform !42

Flow3580:                                         ; preds = %loop-memcpy-expansion2225
  br label %Flow3581, !amdgpu.uniform !42

Flow3582:                                         ; preds = %Flow3579, %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i803
  br label %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i804, !amdgpu.uniform !42

loop-memcpy-residual-header2228:                  ; preds = %Flow3581
  %2591 = icmp ne i64 %1153, 0
  br i1 %2591, label %loop-memcpy-residual2227.preheader, label %Flow3579, !amdgpu.uniform !42

loop-memcpy-residual2227.preheader:               ; preds = %loop-memcpy-residual-header2228
  %umin3029 = call i64 @llvm.umin.i64(i64 %lsr.iv3027, i64 64)
  %2592 = lshr i64 %umin3029, 4
  %2593 = shl nuw nsw i64 %2592, 4
  %scevgep3032 = getelementptr i8, ptr %lsr.iv3021, i64 %2593
  %scevgep3024 = getelementptr i8, ptr %1152, i64 %76
  %scevgep3025 = getelementptr i8, ptr %scevgep3024, i64 %1127
  %scevgep3030 = getelementptr i8, ptr %scevgep3025, i64 %2593
  br label %loop-memcpy-residual2227, !amdgpu.uniform !42

Flow3589:                                         ; preds = %loop-memcpy-expansion2231
  br label %Flow3590, !amdgpu.uniform !42

loop-memcpy-residual-header2234:                  ; preds = %Flow3590
  br i1 %2620, label %loop-memcpy-residual-header2234.post-loop-memcpy-expansion2230_crit_edge, label %Flow3587, !amdgpu.uniform !42

loop-memcpy-residual-header2234.post-loop-memcpy-expansion2230_crit_edge: ; preds = %loop-memcpy-residual-header2234
  %.pre3222 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i782, 6
  br label %Flow3587, !amdgpu.uniform !42

Flow3587:                                         ; preds = %loop-memcpy-residual-header2234.post-loop-memcpy-expansion2230_crit_edge, %loop-memcpy-residual-header2234
  %2594 = phi i64 [ %.pre3222, %loop-memcpy-residual-header2234.post-loop-memcpy-expansion2230_crit_edge ], [ poison, %loop-memcpy-residual-header2234 ]
  %2595 = phi i1 [ false, %loop-memcpy-residual-header2234.post-loop-memcpy-expansion2230_crit_edge ], [ true, %loop-memcpy-residual-header2234 ]
  br i1 %2595, label %loop-memcpy-residual2233.preheader, label %Flow3588, !amdgpu.uniform !42

loop-memcpy-residual2233.preheader:               ; preds = %Flow3587
  %2596 = add i64 %76, %671
  %2597 = add i64 %2596, 8
  %scevgep3034 = getelementptr i8, ptr %1120, i64 %2597
  %2598 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i782, 6
  %scevgep3035 = getelementptr i8, ptr %scevgep3034, i64 %2598
  %scevgep3037 = getelementptr i8, ptr %1064, i64 %671
  br label %loop-memcpy-residual2233, !amdgpu.uniform !42

Flow3612:                                         ; preds = %loop-memcpy-expansion2237
  %.lcssa4006 = phi i64 [ %953, %loop-memcpy-expansion2237 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa4006)
  br label %Flow3613, !amdgpu.uniform !42

Flow3614:                                         ; preds = %Flow3611, %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i601
  call void @llvm.amdgcn.end.cf.i64(i64 %941)
  br label %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i602, !amdgpu.uniform !42

loop-memcpy-residual-header2240:                  ; preds = %Flow3613
  %2599 = icmp ne i64 %943, 0
  %2600 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %2599)
  %2601 = extractvalue { i1, i64 } %2600, 0
  %2602 = extractvalue { i1, i64 } %2600, 1
  br i1 %2601, label %loop-memcpy-residual2239.preheader, label %Flow3611

loop-memcpy-residual2239.preheader:               ; preds = %loop-memcpy-residual-header2240
  %umin3059 = call i64 @llvm.umin.i64(i64 %lsr.iv3054, i64 64)
  %2603 = lshr i64 %umin3059, 4
  %2604 = shl nuw nsw i64 %2603, 4
  %scevgep3060 = getelementptr i8, ptr %lsr.iv3047, i64 %2604
  %scevgep3050 = getelementptr i8, ptr %942, i64 %76
  %scevgep3051 = getelementptr i8, ptr %scevgep3050, i64 %911
  %scevgep3057 = getelementptr i8, ptr %scevgep3051, i64 %2604
  br label %loop-memcpy-residual2239, !amdgpu.uniform !42

Flow3621:                                         ; preds = %loop-memcpy-expansion2243
  %.lcssa4011 = phi i64 [ %909, %loop-memcpy-expansion2243 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa4011)
  br label %Flow3622, !amdgpu.uniform !42

loop-memcpy-residual-header2246:                  ; preds = %Flow3622
  %2605 = icmp eq i64 %897, 0
  %2606 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %2605)
  %2607 = extractvalue { i1, i64 } %2606, 0
  %2608 = extractvalue { i1, i64 } %2606, 1
  br i1 %2607, label %loop-memcpy-residual-header2246.post-loop-memcpy-expansion2242_crit_edge, label %Flow3619

loop-memcpy-residual-header2246.post-loop-memcpy-expansion2242_crit_edge: ; preds = %loop-memcpy-residual-header2246
  %.pre3220 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i580, 6
  br label %Flow3619, !amdgpu.uniform !42

Flow3619:                                         ; preds = %loop-memcpy-residual-header2246.post-loop-memcpy-expansion2242_crit_edge, %loop-memcpy-residual-header2246
  %2609 = phi i64 [ %.pre3220, %loop-memcpy-residual-header2246.post-loop-memcpy-expansion2242_crit_edge ], [ poison, %loop-memcpy-residual-header2246 ]
  %2610 = call { i1, i64 } @llvm.amdgcn.else.i64.i64(i64 %2608)
  %2611 = extractvalue { i1, i64 } %2610, 0
  %2612 = extractvalue { i1, i64 } %2610, 1
  br i1 %2611, label %loop-memcpy-residual2245.preheader, label %Flow3620

loop-memcpy-residual2245.preheader:               ; preds = %Flow3619
  %2613 = add nuw nsw i64 %76, 8
  %scevgep3062 = getelementptr i8, ptr %896, i64 %2613
  %umin3065 = call i64 @llvm.umin.i64(i64 %lsr.iv.next3064.lcssa, i64 56)
  %2614 = lshr i64 %umin3065, 4
  %2615 = shl nuw nsw i64 %2614, 4
  %scevgep3066 = getelementptr i8, ptr %scevgep3062, i64 %2615
  %2616 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i580, 6
  %scevgep3067 = getelementptr i8, ptr %scevgep3066, i64 %2616
  %scevgep3069 = getelementptr i8, ptr %LHSStr, i64 %2615
  br label %loop-memcpy-residual2245, !amdgpu.uniform !42

Flow3650:                                         ; preds = %loop-memcpy-expansion2249
  br label %Flow3651, !amdgpu.uniform !42

Flow3652:                                         ; preds = %Flow3649, %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i390
  br label %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i391, !amdgpu.uniform !42

loop-memcpy-residual-header2252:                  ; preds = %Flow3651
  %2617 = icmp ne i64 %706, 0
  br i1 %2617, label %loop-memcpy-residual2251.preheader, label %Flow3649, !amdgpu.uniform !42

loop-memcpy-residual2251.preheader:               ; preds = %loop-memcpy-residual-header2252
  %umin3105 = call i64 @llvm.umin.i64(i64 %lsr.iv3103, i64 64)
  %2618 = lshr i64 %umin3105, 4
  %2619 = shl nuw nsw i64 %2618, 4
  %scevgep3108 = getelementptr i8, ptr %lsr.iv3097, i64 %2619
  %scevgep3100 = getelementptr i8, ptr %705, i64 %76
  %scevgep3101 = getelementptr i8, ptr %scevgep3100, i64 %679
  %scevgep3106 = getelementptr i8, ptr %scevgep3101, i64 %2619
  br label %loop-memcpy-residual2251, !amdgpu.uniform !42

Flow3659:                                         ; preds = %loop-memcpy-expansion2255
  br label %Flow3660, !amdgpu.uniform !42

loop-memcpy-residual-header2258:                  ; preds = %Flow3660
  %2620 = icmp eq i64 %670, 0
  br i1 %2620, label %loop-memcpy-residual-header2258.post-loop-memcpy-expansion2254_crit_edge, label %Flow3657, !amdgpu.uniform !42

loop-memcpy-residual-header2258.post-loop-memcpy-expansion2254_crit_edge: ; preds = %loop-memcpy-residual-header2258
  %.pre3216 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i369, 6
  br label %Flow3657, !amdgpu.uniform !42

Flow3657:                                         ; preds = %loop-memcpy-residual-header2258.post-loop-memcpy-expansion2254_crit_edge, %loop-memcpy-residual-header2258
  %2621 = phi i64 [ %.pre3216, %loop-memcpy-residual-header2258.post-loop-memcpy-expansion2254_crit_edge ], [ poison, %loop-memcpy-residual-header2258 ]
  %2622 = phi i1 [ false, %loop-memcpy-residual-header2258.post-loop-memcpy-expansion2254_crit_edge ], [ true, %loop-memcpy-residual-header2258 ]
  br i1 %2622, label %loop-memcpy-residual2257.preheader, label %Flow3658, !amdgpu.uniform !42

loop-memcpy-residual2257.preheader:               ; preds = %Flow3657
  %2623 = add i64 %76, %671
  %2624 = add i64 %2623, 8
  %scevgep3110 = getelementptr i8, ptr %669, i64 %2624
  %2625 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i369, 6
  %scevgep3111 = getelementptr i8, ptr %scevgep3110, i64 %2625
  %scevgep3113 = getelementptr i8, ptr %610, i64 %671
  br label %loop-memcpy-residual2257, !amdgpu.uniform !42

Flow3692:                                         ; preds = %loop-memcpy-expansion2261
  %.lcssa4051 = phi i64 [ %474, %loop-memcpy-expansion2261 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa4051)
  br label %Flow3693, !amdgpu.uniform !42

Flow3694:                                         ; preds = %Flow3691, %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i179
  call void @llvm.amdgcn.end.cf.i64(i64 %462)
  br label %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i180, !amdgpu.uniform !42

loop-memcpy-residual-header2264:                  ; preds = %Flow3693
  %2626 = icmp ne i64 %464, 0
  %2627 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %2626)
  %2628 = extractvalue { i1, i64 } %2627, 0
  %2629 = extractvalue { i1, i64 } %2627, 1
  br i1 %2628, label %loop-memcpy-residual2263.preheader, label %Flow3691

loop-memcpy-residual2263.preheader:               ; preds = %loop-memcpy-residual-header2264
  %umin3149 = call i64 @llvm.umin.i64(i64 %lsr.iv3147, i64 64)
  %2630 = lshr i64 %umin3149, 4
  %2631 = shl nuw nsw i64 %2630, 4
  %scevgep3152 = getelementptr i8, ptr %lsr.iv3142, i64 %2631
  %scevgep3145 = getelementptr i8, ptr %463, i64 %76
  %scevgep3146 = getelementptr i8, ptr %scevgep3145, i64 %432
  %scevgep3150 = getelementptr i8, ptr %scevgep3146, i64 %2631
  br label %loop-memcpy-residual2263, !amdgpu.uniform !42

Flow3701:                                         ; preds = %loop-memcpy-expansion2267
  %.lcssa4056 = phi i64 [ %430, %loop-memcpy-expansion2267 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa4056)
  br label %Flow3702, !amdgpu.uniform !42

loop-memcpy-residual-header2270:                  ; preds = %Flow3702
  %2632 = icmp eq i64 %418, 0
  %2633 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %2632)
  %2634 = extractvalue { i1, i64 } %2633, 0
  %2635 = extractvalue { i1, i64 } %2633, 1
  br i1 %2634, label %loop-memcpy-residual-header2270.post-loop-memcpy-expansion2266_crit_edge, label %Flow3699

loop-memcpy-residual-header2270.post-loop-memcpy-expansion2266_crit_edge: ; preds = %loop-memcpy-residual-header2270
  %.pre3212 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i158, 6
  br label %Flow3699, !amdgpu.uniform !42

Flow3699:                                         ; preds = %loop-memcpy-residual-header2270.post-loop-memcpy-expansion2266_crit_edge, %loop-memcpy-residual-header2270
  %2636 = phi i64 [ %.pre3212, %loop-memcpy-residual-header2270.post-loop-memcpy-expansion2266_crit_edge ], [ poison, %loop-memcpy-residual-header2270 ]
  %2637 = call { i1, i64 } @llvm.amdgcn.else.i64.i64(i64 %2635)
  %2638 = extractvalue { i1, i64 } %2637, 0
  %2639 = extractvalue { i1, i64 } %2637, 1
  br i1 %2638, label %loop-memcpy-residual2269.preheader, label %Flow3700

loop-memcpy-residual2269.preheader:               ; preds = %Flow3699
  %2640 = lshr i64 %spec.select.i.i.i.i.i.i.i161, 4
  %2641 = shl nuw nsw i64 %2640, 4
  %2642 = add i64 %76, %2641
  %2643 = add i64 %2642, 8
  %scevgep3154 = getelementptr i8, ptr %417, i64 %2643
  %2644 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i158, 6
  %scevgep3155 = getelementptr i8, ptr %scevgep3154, i64 %2644
  %scevgep3157 = getelementptr i8, ptr %341, i64 %2641
  br label %loop-memcpy-residual2269, !amdgpu.uniform !42

Flow3394:                                         ; preds = %loop-memcpy-expansion2273
  br label %Flow3395, !amdgpu.uniform !42

loop-memcpy-residual-header2276:                  ; preds = %Flow3395
  br i1 %2647, label %loop-memcpy-residual2275.preheader, label %Flow3392, !amdgpu.uniform !42

loop-memcpy-residual2275.preheader:               ; preds = %loop-memcpy-residual-header2276
  %2645 = sub i64 sub (i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) getelementptr inbounds nuw (i8, ptr addrspace(4) @.str.9, i64 8) to ptr) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.9 to ptr) to i64)), %umin3102
  %2646 = sub i64 %2645, %611
  %scevgep2893 = getelementptr i8, ptr %2083, i64 %2646
  %scevgep2895 = getelementptr i8, ptr %4, i64 %2646
  br label %loop-memcpy-residual2275, !amdgpu.uniform !42

Flow3597:                                         ; preds = %loop-memcpy-expansion2279
  br label %Flow3598, !amdgpu.uniform !42

loop-memcpy-residual-header2282:                  ; preds = %Flow3598
  br i1 %2647, label %loop-memcpy-residual2281.preheader, label %Flow3596, !amdgpu.uniform !42

loop-memcpy-residual2281.preheader:               ; preds = %loop-memcpy-residual-header2282
  %scevgep3042 = getelementptr i8, ptr %1064, i64 %612
  %scevgep3044 = getelementptr i8, ptr %4, i64 %612
  br label %loop-memcpy-residual2281, !amdgpu.uniform !42

Flow3667:                                         ; preds = %loop-memcpy-expansion2285
  br label %Flow3668, !amdgpu.uniform !42

loop-memcpy-residual-header2288:                  ; preds = %Flow3668
  %2647 = icmp ne i64 %611, 0
  br i1 %2647, label %loop-memcpy-residual2287.preheader, label %Flow3666, !amdgpu.uniform !42

loop-memcpy-residual2287.preheader:               ; preds = %loop-memcpy-residual-header2288
  %scevgep3118 = getelementptr i8, ptr %610, i64 %612
  %scevgep3120 = getelementptr i8, ptr %4, i64 %612
  br label %loop-memcpy-residual2287, !amdgpu.uniform !42

Flow3709:                                         ; preds = %loop-memcpy-expansion2291
  %.lcssa4064 = phi i64 [ %363, %loop-memcpy-expansion2291 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa4064)
  br label %Flow3710, !amdgpu.uniform !42

loop-memcpy-residual-header2294:                  ; preds = %Flow3710
  %2648 = icmp ne i64 %352, 0
  %2649 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %2648)
  %2650 = extractvalue { i1, i64 } %2649, 0
  %2651 = extractvalue { i1, i64 } %2649, 1
  br i1 %2650, label %loop-memcpy-residual2293.preheader, label %Flow3708

loop-memcpy-residual2293.preheader:               ; preds = %loop-memcpy-residual-header2294
  %2652 = lshr i64 %writer.sroa.10.5.i.i.i.i, 4
  %2653 = trunc i64 %2652 to i32
  %2654 = shl i32 %2653, 4
  %2655 = add i32 %2654, 11
  %2656 = sub i32 %2655, %351
  %scevgep3162 = getelementptr i8, ptr addrspace(5) %buffer.i.i.i, i32 %2656
  %2657 = shl nuw i64 %2652, 4
  %scevgep3165 = getelementptr i8, ptr %341, i64 %2657
  br label %loop-memcpy-residual2293, !amdgpu.uniform !42

Flow3342:                                         ; preds = %loop-memcpy-expansion2297
  %.lcssa3872 = phi i64 [ %2333, %loop-memcpy-expansion2297 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa3872)
  br label %Flow3343, !amdgpu.uniform !42

loop-memcpy-residual-header2300:                  ; preds = %Flow3343
  %2658 = icmp ne i64 %2322, 0
  %2659 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %2658)
  %2660 = extractvalue { i1, i64 } %2659, 0
  %2661 = extractvalue { i1, i64 } %2659, 1
  br i1 %2660, label %loop-memcpy-residual2299.preheader, label %Flow3341

loop-memcpy-residual2299.preheader:               ; preds = %loop-memcpy-residual-header2300
  %2662 = sub i32 11, %2321
  %scevgep2847 = getelementptr i8, ptr addrspace(5) %buffer.i.i.i, i32 %2662
  br label %loop-memcpy-residual2299, !amdgpu.uniform !42

Flow3550:                                         ; preds = %loop-memcpy-expansion2303
  %.lcssa3973 = phi i64 [ %1317, %loop-memcpy-expansion2303 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa3973)
  br label %Flow3551, !amdgpu.uniform !42

loop-memcpy-residual-header2306:                  ; preds = %Flow3551
  %2663 = icmp ne i64 %1306, 0
  %2664 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %2663)
  %2665 = extractvalue { i1, i64 } %2664, 0
  %2666 = extractvalue { i1, i64 } %2664, 1
  br i1 %2665, label %loop-memcpy-residual2305.preheader, label %Flow3548

loop-memcpy-residual2305.preheader:               ; preds = %loop-memcpy-residual-header2306
  %2667 = sub i32 11, %1305
  %scevgep2994 = getelementptr i8, ptr addrspace(5) %buffer.i.i.i, i32 %2667
  br label %loop-memcpy-residual2305, !amdgpu.uniform !42

Flow3765:                                         ; preds = %loop-memcpy-expansion2309
  br label %Flow3766, !amdgpu.uniform !42

loop-memcpy-residual-header2312:                  ; preds = %Flow3766
  %2668 = icmp eq i64 %66, 0
  br i1 %2668, label %loop-memcpy-residual-header2312.post-loop-memcpy-expansion2308_crit_edge, label %Flow3763, !amdgpu.uniform !42

loop-memcpy-residual-header2312.post-loop-memcpy-expansion2308_crit_edge: ; preds = %loop-memcpy-residual-header2312
  %.pre = shl nuw nsw i64 %sh_prom.i.i.i.i, 6
  %.pre3208 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i, 6
  br label %Flow3763, !amdgpu.uniform !42

Flow3763:                                         ; preds = %loop-memcpy-residual-header2312.post-loop-memcpy-expansion2308_crit_edge, %loop-memcpy-residual-header2312
  %2669 = phi i64 [ %.pre, %loop-memcpy-residual-header2312.post-loop-memcpy-expansion2308_crit_edge ], [ poison, %loop-memcpy-residual-header2312 ]
  %2670 = phi i64 [ %.pre3208, %loop-memcpy-residual-header2312.post-loop-memcpy-expansion2308_crit_edge ], [ poison, %loop-memcpy-residual-header2312 ]
  %2671 = phi i1 [ false, %loop-memcpy-residual-header2312.post-loop-memcpy-expansion2308_crit_edge ], [ true, %loop-memcpy-residual-header2312 ]
  br i1 %2671, label %loop-memcpy-residual2311.preheader, label %Flow3764, !amdgpu.uniform !42

loop-memcpy-residual2311.preheader:               ; preds = %Flow3763
  %2672 = shl nuw nsw i64 %sh_prom.i.i.i.i, 6
  %2673 = add nuw nsw i64 %2672, 56
  %scevgep3201 = getelementptr i8, ptr %65, i64 %2673
  %2674 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i, 6
  %scevgep3202 = getelementptr i8, ptr %scevgep3201, i64 %2674
  br label %loop-memcpy-residual2311, !amdgpu.uniform !42

Flow3756:                                         ; preds = %loop-memcpy-expansion2315
  br label %Flow3757, !amdgpu.uniform !42

Flow3758:                                         ; preds = %Flow3755, %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i
  br label %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i, !amdgpu.uniform !42

loop-memcpy-residual-header2318:                  ; preds = %Flow3757
  %2675 = icmp ne i64 %103, 0
  br i1 %2675, label %loop-memcpy-residual2317.preheader, label %Flow3755, !amdgpu.uniform !42

loop-memcpy-residual2317.preheader:               ; preds = %loop-memcpy-residual-header2318
  %umin3196 = call i64 @llvm.umin.i64(i64 %lsr.iv3194, i64 64)
  %2676 = lshr i64 %umin3196, 4
  %2677 = shl nuw nsw i64 %2676, 4
  %scevgep3199 = getelementptr i8, ptr addrspace(4) %lsr.iv3189, i64 %2677
  %scevgep3192 = getelementptr i8, ptr %102, i64 %76
  %scevgep3193 = getelementptr i8, ptr %scevgep3192, i64 %77
  %scevgep3197 = getelementptr i8, ptr %scevgep3193, i64 %2677
  br label %loop-memcpy-residual2317, !amdgpu.uniform !42

Flow3746:                                         ; preds = %loop-memcpy-expansion2321
  br label %Flow3747, !amdgpu.uniform !42

loop-memcpy-residual-header2324:                  ; preds = %Flow3747
  %2678 = icmp eq i64 %179, 0
  br i1 %2678, label %loop-memcpy-residual-header2324.post-loop-memcpy-expansion2320_crit_edge, label %Flow3744, !amdgpu.uniform !42

loop-memcpy-residual-header2324.post-loop-memcpy-expansion2320_crit_edge: ; preds = %loop-memcpy-residual-header2324
  %.pre3210 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i54, 6
  br label %Flow3744, !amdgpu.uniform !42

Flow3744:                                         ; preds = %loop-memcpy-residual-header2324.post-loop-memcpy-expansion2320_crit_edge, %loop-memcpy-residual-header2324
  %2679 = phi i64 [ %.pre3210, %loop-memcpy-residual-header2324.post-loop-memcpy-expansion2320_crit_edge ], [ poison, %loop-memcpy-residual-header2324 ]
  %2680 = phi i1 [ false, %loop-memcpy-residual-header2324.post-loop-memcpy-expansion2320_crit_edge ], [ true, %loop-memcpy-residual-header2324 ]
  br i1 %2680, label %loop-memcpy-residual2323.preheader, label %Flow3745, !amdgpu.uniform !42

loop-memcpy-residual2323.preheader:               ; preds = %Flow3744
  %2681 = add nuw nsw i64 %76, 8
  %scevgep3182 = getelementptr i8, ptr %178, i64 %2681
  %2682 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i54, 6
  %scevgep3183 = getelementptr i8, ptr %scevgep3182, i64 %2682
  br label %loop-memcpy-residual2323, !amdgpu.uniform !42

Flow3737:                                         ; preds = %loop-memcpy-expansion2327
  br label %Flow3738, !amdgpu.uniform !42

Flow3739:                                         ; preds = %Flow3736, %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i75
  br label %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i76, !amdgpu.uniform !42

loop-memcpy-residual-header2330:                  ; preds = %Flow3738
  %2683 = icmp ne i64 %214, 0
  br i1 %2683, label %loop-memcpy-residual2329.preheader, label %Flow3736, !amdgpu.uniform !42

loop-memcpy-residual2329.preheader:               ; preds = %loop-memcpy-residual-header2330
  %umin3177 = call i64 @llvm.umin.i64(i64 %lsr.iv3175, i64 64)
  %2684 = lshr i64 %umin3177, 4
  %2685 = shl nuw nsw i64 %2684, 4
  %scevgep3180 = getelementptr i8, ptr addrspace(4) %lsr.iv3170, i64 %2685
  %scevgep3173 = getelementptr i8, ptr %213, i64 %76
  %scevgep3174 = getelementptr i8, ptr %scevgep3173, i64 %188
  %scevgep3178 = getelementptr i8, ptr %scevgep3174, i64 %2685
  br label %loop-memcpy-residual2329, !amdgpu.uniform !42

Flow3682:                                         ; preds = %loop-memcpy-expansion2333
  br label %Flow3683, !amdgpu.uniform !42

loop-memcpy-residual-header2336:                  ; preds = %Flow3683
  %2686 = icmp eq i64 %547, 0
  br i1 %2686, label %loop-memcpy-residual-header2336.post-loop-memcpy-expansion2332_crit_edge, label %Flow3680, !amdgpu.uniform !42

loop-memcpy-residual-header2336.post-loop-memcpy-expansion2332_crit_edge: ; preds = %loop-memcpy-residual-header2336
  %.pre3214 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i263, 6
  br label %Flow3680, !amdgpu.uniform !42

Flow3680:                                         ; preds = %loop-memcpy-residual-header2336.post-loop-memcpy-expansion2332_crit_edge, %loop-memcpy-residual-header2336
  %2687 = phi i64 [ %.pre3214, %loop-memcpy-residual-header2336.post-loop-memcpy-expansion2332_crit_edge ], [ poison, %loop-memcpy-residual-header2336 ]
  %2688 = phi i1 [ false, %loop-memcpy-residual-header2336.post-loop-memcpy-expansion2332_crit_edge ], [ true, %loop-memcpy-residual-header2336 ]
  br i1 %2688, label %loop-memcpy-residual2335.preheader, label %Flow3681, !amdgpu.uniform !42

loop-memcpy-residual2335.preheader:               ; preds = %Flow3680
  %2689 = add nuw nsw i64 %76, 8
  %scevgep3134 = getelementptr i8, ptr %546, i64 %2689
  %2690 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i263, 6
  %scevgep3135 = getelementptr i8, ptr %scevgep3134, i64 %2690
  br label %loop-memcpy-residual2335, !amdgpu.uniform !42

Flow3673:                                         ; preds = %loop-memcpy-expansion2339
  br label %Flow3674, !amdgpu.uniform !42

Flow3675:                                         ; preds = %Flow3672, %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i284
  br label %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i285, !amdgpu.uniform !42

loop-memcpy-residual-header2342:                  ; preds = %Flow3674
  %2691 = icmp ne i64 %582, 0
  br i1 %2691, label %loop-memcpy-residual2341.preheader, label %Flow3672, !amdgpu.uniform !42

loop-memcpy-residual2341.preheader:               ; preds = %loop-memcpy-residual-header2342
  %umin3129 = call i64 @llvm.umin.i64(i64 %lsr.iv3127, i64 64)
  %2692 = lshr i64 %umin3129, 4
  %2693 = shl nuw nsw i64 %2692, 4
  %scevgep3132 = getelementptr i8, ptr addrspace(4) %lsr.iv3122, i64 %2693
  %scevgep3125 = getelementptr i8, ptr %581, i64 %76
  %scevgep3126 = getelementptr i8, ptr %scevgep3125, i64 %556
  %scevgep3130 = getelementptr i8, ptr %scevgep3126, i64 %2693
  br label %loop-memcpy-residual2341, !amdgpu.uniform !42

Flow3640:                                         ; preds = %loop-memcpy-expansion2345
  br label %Flow3641, !amdgpu.uniform !42

loop-memcpy-residual-header2348:                  ; preds = %Flow3641
  %2694 = icmp eq i64 %782, 0
  br i1 %2694, label %loop-memcpy-residual-header2348.post-loop-memcpy-expansion2344_crit_edge, label %Flow3638, !amdgpu.uniform !42

loop-memcpy-residual-header2348.post-loop-memcpy-expansion2344_crit_edge: ; preds = %loop-memcpy-residual-header2348
  %.pre3218 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i474, 6
  br label %Flow3638, !amdgpu.uniform !42

Flow3638:                                         ; preds = %loop-memcpy-residual-header2348.post-loop-memcpy-expansion2344_crit_edge, %loop-memcpy-residual-header2348
  %2695 = phi i64 [ %.pre3218, %loop-memcpy-residual-header2348.post-loop-memcpy-expansion2344_crit_edge ], [ poison, %loop-memcpy-residual-header2348 ]
  %2696 = phi i1 [ false, %loop-memcpy-residual-header2348.post-loop-memcpy-expansion2344_crit_edge ], [ true, %loop-memcpy-residual-header2348 ]
  br i1 %2696, label %loop-memcpy-residual2347.preheader, label %Flow3639, !amdgpu.uniform !42

loop-memcpy-residual2347.preheader:               ; preds = %Flow3638
  %2697 = add nuw nsw i64 %76, 8
  %scevgep3089 = getelementptr i8, ptr %781, i64 %2697
  %2698 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i474, 6
  %scevgep3090 = getelementptr i8, ptr %scevgep3089, i64 %2698
  br label %loop-memcpy-residual2347, !amdgpu.uniform !42

Flow3631:                                         ; preds = %loop-memcpy-expansion2351
  br label %Flow3632, !amdgpu.uniform !42

Flow3633:                                         ; preds = %Flow3630, %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i495
  br label %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i496, !amdgpu.uniform !42

loop-memcpy-residual-header2354:                  ; preds = %Flow3632
  %2699 = icmp ne i64 %817, 0
  br i1 %2699, label %loop-memcpy-residual2353.preheader, label %Flow3630, !amdgpu.uniform !42

loop-memcpy-residual2353.preheader:               ; preds = %loop-memcpy-residual-header2354
  %umin3084 = call i64 @llvm.umin.i64(i64 %lsr.iv3082, i64 64)
  %2700 = lshr i64 %umin3084, 4
  %2701 = shl nuw nsw i64 %2700, 4
  %scevgep3087 = getelementptr i8, ptr addrspace(4) %lsr.iv3077, i64 %2701
  %scevgep3080 = getelementptr i8, ptr %816, i64 %76
  %scevgep3081 = getelementptr i8, ptr %scevgep3080, i64 %791
  %scevgep3085 = getelementptr i8, ptr %scevgep3081, i64 %2701
  br label %loop-memcpy-residual2353, !amdgpu.uniform !42

Flow3570:                                         ; preds = %loop-memcpy-expansion2357
  br label %Flow3571, !amdgpu.uniform !42

loop-memcpy-residual-header2360:                  ; preds = %Flow3571
  %2702 = icmp eq i64 %1229, 0
  br i1 %2702, label %loop-memcpy-residual-header2360.post-loop-memcpy-expansion2356_crit_edge, label %Flow3568, !amdgpu.uniform !42

loop-memcpy-residual-header2360.post-loop-memcpy-expansion2356_crit_edge: ; preds = %loop-memcpy-residual-header2360
  %.pre3224 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i888, 6
  br label %Flow3568, !amdgpu.uniform !42

Flow3568:                                         ; preds = %loop-memcpy-residual-header2360.post-loop-memcpy-expansion2356_crit_edge, %loop-memcpy-residual-header2360
  %2703 = phi i64 [ %.pre3224, %loop-memcpy-residual-header2360.post-loop-memcpy-expansion2356_crit_edge ], [ poison, %loop-memcpy-residual-header2360 ]
  %2704 = phi i1 [ false, %loop-memcpy-residual-header2360.post-loop-memcpy-expansion2356_crit_edge ], [ true, %loop-memcpy-residual-header2360 ]
  br i1 %2704, label %loop-memcpy-residual2359.preheader, label %Flow3569, !amdgpu.uniform !42

loop-memcpy-residual2359.preheader:               ; preds = %Flow3568
  %2705 = add nuw nsw i64 %76, 8
  %scevgep3013 = getelementptr i8, ptr %1228, i64 %2705
  %2706 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i888, 6
  %scevgep3014 = getelementptr i8, ptr %scevgep3013, i64 %2706
  br label %loop-memcpy-residual2359, !amdgpu.uniform !42

Flow3358:                                         ; preds = %loop-memcpy-expansion2363
  br label %Flow3359, !amdgpu.uniform !42

loop-memcpy-residual-header2366:                  ; preds = %Flow3359
  br i1 %2702, label %loop-memcpy-residual-header2366.post-loop-memcpy-expansion2362_crit_edge, label %Flow3355, !amdgpu.uniform !42

loop-memcpy-residual-header2366.post-loop-memcpy-expansion2362_crit_edge: ; preds = %loop-memcpy-residual-header2366
  %.pre3236 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i1819, 6
  br label %Flow3355, !amdgpu.uniform !42

Flow3355:                                         ; preds = %loop-memcpy-residual-header2366.post-loop-memcpy-expansion2362_crit_edge, %loop-memcpy-residual-header2366
  %2707 = phi i64 [ %.pre3236, %loop-memcpy-residual-header2366.post-loop-memcpy-expansion2362_crit_edge ], [ poison, %loop-memcpy-residual-header2366 ]
  %2708 = phi i1 [ false, %loop-memcpy-residual-header2366.post-loop-memcpy-expansion2362_crit_edge ], [ true, %loop-memcpy-residual-header2366 ]
  br i1 %2708, label %loop-memcpy-residual2365.preheader, label %Flow3356, !amdgpu.uniform !42

loop-memcpy-residual2365.preheader:               ; preds = %Flow3355
  %2709 = add nuw nsw i64 %76, 8
  %scevgep2863 = getelementptr i8, ptr %2247, i64 %2709
  %2710 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i1819, 6
  %scevgep2864 = getelementptr i8, ptr %scevgep2863, i64 %2710
  br label %loop-memcpy-residual2365, !amdgpu.uniform !42

Flow3559:                                         ; preds = %loop-memcpy-expansion2369
  br label %Flow3560, !amdgpu.uniform !42

Flow3561:                                         ; preds = %Flow3557, %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i909
  br label %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i910, !amdgpu.uniform !42

loop-memcpy-residual-header2372:                  ; preds = %Flow3560
  %2711 = icmp ne i64 %1264, 0
  br i1 %2711, label %loop-memcpy-residual2371.preheader, label %Flow3557, !amdgpu.uniform !42

loop-memcpy-residual2371.preheader:               ; preds = %loop-memcpy-residual-header2372
  %umin3008 = call i64 @llvm.umin.i64(i64 %lsr.iv3006, i64 64)
  %2712 = lshr i64 %umin3008, 4
  %2713 = shl nuw nsw i64 %2712, 4
  %scevgep3011 = getelementptr i8, ptr addrspace(4) %lsr.iv3001, i64 %2713
  %scevgep3004 = getelementptr i8, ptr %1263, i64 %76
  %scevgep3005 = getelementptr i8, ptr %scevgep3004, i64 %1238
  %scevgep3009 = getelementptr i8, ptr %scevgep3005, i64 %2713
  br label %loop-memcpy-residual2371, !amdgpu.uniform !42

Flow3505:                                         ; preds = %loop-memcpy-expansion2375
  br label %Flow3506, !amdgpu.uniform !42

loop-memcpy-residual-header2378:                  ; preds = %Flow3506
  %2714 = icmp eq i64 %1579, 0
  br i1 %2714, label %loop-memcpy-residual-header2378.post-loop-memcpy-expansion2374_crit_edge, label %Flow3502, !amdgpu.uniform !42

loop-memcpy-residual-header2378.post-loop-memcpy-expansion2374_crit_edge: ; preds = %loop-memcpy-residual-header2378
  %.pre3226 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i1197, 6
  br label %Flow3502, !amdgpu.uniform !42

Flow3502:                                         ; preds = %loop-memcpy-residual-header2378.post-loop-memcpy-expansion2374_crit_edge, %loop-memcpy-residual-header2378
  %2715 = phi i64 [ %.pre3226, %loop-memcpy-residual-header2378.post-loop-memcpy-expansion2374_crit_edge ], [ poison, %loop-memcpy-residual-header2378 ]
  %2716 = phi i1 [ false, %loop-memcpy-residual-header2378.post-loop-memcpy-expansion2374_crit_edge ], [ true, %loop-memcpy-residual-header2378 ]
  br i1 %2716, label %loop-memcpy-residual2377.preheader, label %Flow3503, !amdgpu.uniform !42

loop-memcpy-residual2377.preheader:               ; preds = %Flow3502
  %2717 = add nuw nsw i64 %76, 8
  %scevgep2979 = getelementptr i8, ptr %1578, i64 %2717
  %2718 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i1197, 6
  %scevgep2980 = getelementptr i8, ptr %scevgep2979, i64 %2718
  br label %loop-memcpy-residual2377, !amdgpu.uniform !42

Flow3492:                                         ; preds = %loop-memcpy-expansion2381
  br label %Flow3493, !amdgpu.uniform !42

Flow3494:                                         ; preds = %Flow3490, %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i1218
  br label %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1219, !amdgpu.uniform !42

loop-memcpy-residual-header2384:                  ; preds = %Flow3493
  %2719 = icmp ne i64 %1614, 0
  br i1 %2719, label %loop-memcpy-residual2383.preheader, label %Flow3490, !amdgpu.uniform !42

loop-memcpy-residual2383.preheader:               ; preds = %loop-memcpy-residual-header2384
  %umin2974 = call i64 @llvm.umin.i64(i64 %lsr.iv2972, i64 64)
  %2720 = lshr i64 %umin2974, 4
  %2721 = shl nuw nsw i64 %2720, 4
  %scevgep2977 = getelementptr i8, ptr addrspace(4) %lsr.iv2967, i64 %2721
  %scevgep2970 = getelementptr i8, ptr %1613, i64 %76
  %scevgep2971 = getelementptr i8, ptr %scevgep2970, i64 %1588
  %scevgep2975 = getelementptr i8, ptr %scevgep2971, i64 %2721
  br label %loop-memcpy-residual2383, !amdgpu.uniform !42

Flow3479:                                         ; preds = %loop-memcpy-expansion2387
  br label %Flow3480, !amdgpu.uniform !42

loop-memcpy-residual-header2390:                  ; preds = %Flow3480
  %2722 = icmp eq i64 %1690, 0
  br i1 %2722, label %loop-memcpy-residual-header2390.post-loop-memcpy-expansion2386_crit_edge, label %Flow3476, !amdgpu.uniform !42

loop-memcpy-residual-header2390.post-loop-memcpy-expansion2386_crit_edge: ; preds = %loop-memcpy-residual-header2390
  %.pre3228 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i1296, 6
  br label %Flow3476, !amdgpu.uniform !42

Flow3476:                                         ; preds = %loop-memcpy-residual-header2390.post-loop-memcpy-expansion2386_crit_edge, %loop-memcpy-residual-header2390
  %2723 = phi i64 [ %.pre3228, %loop-memcpy-residual-header2390.post-loop-memcpy-expansion2386_crit_edge ], [ poison, %loop-memcpy-residual-header2390 ]
  %2724 = phi i1 [ false, %loop-memcpy-residual-header2390.post-loop-memcpy-expansion2386_crit_edge ], [ true, %loop-memcpy-residual-header2390 ]
  br i1 %2724, label %loop-memcpy-residual2389.preheader, label %Flow3477, !amdgpu.uniform !42

loop-memcpy-residual2389.preheader:               ; preds = %Flow3476
  %2725 = lshr i64 %spec.select.i.i.i.i.i.i.i1299, 4
  %2726 = shl nuw nsw i64 %2725, 4
  %2727 = add i64 %76, %2726
  %2728 = add i64 %2727, 8
  %scevgep2959 = getelementptr i8, ptr %1689, i64 %2728
  %2729 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i1296, 6
  %scevgep2960 = getelementptr i8, ptr %scevgep2959, i64 %2729
  %scevgep2962 = getelementptr i8, ptr addrspace(4) @.str.9, i64 %2726
  br label %loop-memcpy-residual2389, !amdgpu.uniform !42

Flow3466:                                         ; preds = %loop-memcpy-expansion2393
  br label %Flow3467, !amdgpu.uniform !42

Flow3468:                                         ; preds = %Flow3464, %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i1317
  br label %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1318, !amdgpu.uniform !42

loop-memcpy-residual-header2396:                  ; preds = %Flow3467
  %2730 = icmp ne i64 %1725, 0
  br i1 %2730, label %loop-memcpy-residual2395.preheader, label %Flow3464, !amdgpu.uniform !42

loop-memcpy-residual2395.preheader:               ; preds = %loop-memcpy-residual-header2396
  %umin2954 = call i64 @llvm.umin.i64(i64 %lsr.iv2952, i64 64)
  %2731 = lshr i64 %umin2954, 4
  %2732 = shl nuw nsw i64 %2731, 4
  %scevgep2957 = getelementptr i8, ptr addrspace(4) %lsr.iv2947, i64 %2732
  %scevgep2950 = getelementptr i8, ptr %1724, i64 %76
  %scevgep2951 = getelementptr i8, ptr %scevgep2950, i64 %1699
  %scevgep2955 = getelementptr i8, ptr %scevgep2951, i64 %2732
  br label %loop-memcpy-residual2395, !amdgpu.uniform !42

Flow3453:                                         ; preds = %loop-memcpy-expansion2399
  br label %Flow3454, !amdgpu.uniform !42

loop-memcpy-residual-header2402:                  ; preds = %Flow3454
  %2733 = icmp eq i64 %1801, 0
  br i1 %2733, label %loop-memcpy-residual-header2402.post-loop-memcpy-expansion2398_crit_edge, label %Flow3450, !amdgpu.uniform !42

loop-memcpy-residual-header2402.post-loop-memcpy-expansion2398_crit_edge: ; preds = %loop-memcpy-residual-header2402
  %.pre3230 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i1401, 6
  br label %Flow3450, !amdgpu.uniform !42

Flow3450:                                         ; preds = %loop-memcpy-residual-header2402.post-loop-memcpy-expansion2398_crit_edge, %loop-memcpy-residual-header2402
  %2734 = phi i64 [ %.pre3230, %loop-memcpy-residual-header2402.post-loop-memcpy-expansion2398_crit_edge ], [ poison, %loop-memcpy-residual-header2402 ]
  %2735 = phi i1 [ false, %loop-memcpy-residual-header2402.post-loop-memcpy-expansion2398_crit_edge ], [ true, %loop-memcpy-residual-header2402 ]
  br i1 %2735, label %loop-memcpy-residual2401.preheader, label %Flow3451, !amdgpu.uniform !42

loop-memcpy-residual2401.preheader:               ; preds = %Flow3450
  %2736 = add nuw nsw i64 %76, 8
  %scevgep2940 = getelementptr i8, ptr %1800, i64 %2736
  %2737 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i1401, 6
  %scevgep2941 = getelementptr i8, ptr %scevgep2940, i64 %2737
  br label %loop-memcpy-residual2401, !amdgpu.uniform !42

Flow3440:                                         ; preds = %loop-memcpy-expansion2405
  br label %Flow3441, !amdgpu.uniform !42

Flow3442:                                         ; preds = %Flow3438, %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i1422
  br label %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1423, !amdgpu.uniform !42

loop-memcpy-residual-header2408:                  ; preds = %Flow3441
  %2738 = icmp ne i64 %1836, 0
  br i1 %2738, label %loop-memcpy-residual2407.preheader, label %Flow3438, !amdgpu.uniform !42

loop-memcpy-residual2407.preheader:               ; preds = %loop-memcpy-residual-header2408
  %umin2935 = call i64 @llvm.umin.i64(i64 %lsr.iv2933, i64 64)
  %2739 = lshr i64 %umin2935, 4
  %2740 = shl nuw nsw i64 %2739, 4
  %scevgep2938 = getelementptr i8, ptr addrspace(4) %lsr.iv2928, i64 %2740
  %scevgep2931 = getelementptr i8, ptr %1835, i64 %76
  %scevgep2932 = getelementptr i8, ptr %scevgep2931, i64 %1810
  %scevgep2936 = getelementptr i8, ptr %scevgep2932, i64 %2740
  br label %loop-memcpy-residual2407, !amdgpu.uniform !42

Flow3348:                                         ; preds = %loop-memcpy-expansion2411
  br label %Flow3349, !amdgpu.uniform !42

Flow3350:                                         ; preds = %Flow3347, %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i1840
  br label %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1841, !amdgpu.uniform !42

loop-memcpy-residual-header2414:                  ; preds = %Flow3349
  %2741 = icmp ne i64 %2280, 0
  br i1 %2741, label %loop-memcpy-residual2413.preheader, label %Flow3347, !amdgpu.uniform !42

loop-memcpy-residual2413.preheader:               ; preds = %loop-memcpy-residual-header2414
  %umin = call i64 @llvm.umin.i64(i64 %lsr.iv2858, i64 64)
  %2742 = lshr i64 %umin, 4
  %2743 = shl nuw nsw i64 %2742, 4
  %scevgep2861 = getelementptr i8, ptr addrspace(4) %lsr.iv2853, i64 %2743
  %scevgep2856 = getelementptr i8, ptr %2279, i64 %76
  %scevgep2857 = getelementptr i8, ptr %scevgep2856, i64 %2254
  %scevgep2859 = getelementptr i8, ptr %scevgep2857, i64 %2743
  br label %loop-memcpy-residual2413, !amdgpu.uniform !42

ComputeLoop:                                      ; preds = %for.cond.i.i1.i, %ComputeLoop
  %Accumulator = phi i32 [ %2748, %ComputeLoop ], [ 0, %for.cond.i.i1.i ]
  %OldValuePhi = phi i32 [ %2747, %ComputeLoop ], [ poison, %for.cond.i.i1.i ]
  %ActiveBits = phi i64 [ %2751, %ComputeLoop ], [ %26, %for.cond.i.i1.i ]
  %2744 = call i64 @llvm.cttz.i64(i64 %ActiveBits, i1 true)
  %2745 = trunc i64 %2744 to i32
  %2746 = call i32 @llvm.amdgcn.readlane.i32(i32 %shl.i.i.i.i.i, i32 %2745)
  %2747 = call i32 @llvm.amdgcn.writelane.i32(i32 %Accumulator, i32 %2745, i32 %OldValuePhi)
  %2748 = or i32 %Accumulator, %2746
  %2749 = shl i64 1, %2744
  %2750 = xor i64 %2749, -1
  %2751 = and i64 %ActiveBits, %2750
  %2752 = icmp eq i64 %2751, 0
  br i1 %2752, label %ComputeEnd, label %ComputeLoop, !amdgpu.uniform !42

ComputeEnd:                                       ; preds = %ComputeLoop
  %.lcssa4082 = phi i32 [ %2747, %ComputeLoop ]
  %.lcssa4081 = phi i32 [ %2748, %ComputeLoop ]
  %div8.i.i.i.i.i = lshr i32 %spec.store.select.i.i.i, 5
  %idxprom.i.i.i.i.i = zext nneg i32 %div8.i.i.i.i.i to i64
  %2753 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i.i.i
  %2754 = trunc i64 %26 to i32
  %2755 = lshr i64 %26, 32
  %2756 = trunc i64 %2755 to i32
  %2757 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %2754, i32 0)
  %2758 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %2756, i32 %2757)
  %2759 = icmp eq i32 %2758, 0
  %2760 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %2759)
  %2761 = extractvalue { i1, i64 } %2760, 0
  %2762 = extractvalue { i1, i64 } %2760, 1
  br i1 %2761, label %27, label %29

Flow3767:                                         ; preds = %47, %if.then.i.i.i.i1
  %2763 = phi i1 [ false, %47 ], [ true, %if.then.i.i.i.i1 ]
  br i1 %2763, label %if.end14.i.i.i, label %Flow3768, !amdgpu.uniform !42

ComputeLoop2416:                                  ; preds = %if.then12.i.i.i, %ComputeLoop2416
  %Accumulator2418 = phi i32 [ %2767, %ComputeLoop2416 ], [ -1, %if.then12.i.i.i ]
  %ActiveBits2419 = phi i64 [ %2770, %ComputeLoop2416 ], [ %44, %if.then12.i.i.i ]
  %2764 = call i64 @llvm.cttz.i64(i64 %ActiveBits2419, i1 true)
  %2765 = trunc i64 %2764 to i32
  %2766 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i.i.i, i32 %2765)
  %2767 = and i32 %Accumulator2418, %2766
  %2768 = shl i64 1, %2764
  %2769 = xor i64 %2768, -1
  %2770 = and i64 %ActiveBits2419, %2769
  %2771 = icmp eq i64 %2770, 0
  br i1 %2771, label %ComputeEnd2417, label %ComputeLoop2416, !amdgpu.uniform !42

ComputeEnd2417:                                   ; preds = %ComputeLoop2416
  %.lcssa4083 = phi i32 [ %2767, %ComputeLoop2416 ]
  %2772 = trunc i64 %44 to i32
  %2773 = lshr i64 %44, 32
  %2774 = trunc i64 %2773 to i32
  %2775 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %2772, i32 0)
  %2776 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %2774, i32 %2775)
  %2777 = icmp eq i32 %2776, 0
  %2778 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %2777)
  %2779 = extractvalue { i1, i64 } %2778, 0
  %2780 = extractvalue { i1, i64 } %2778, 1
  br i1 %2779, label %45, label %47

ComputeLoop2420:                                  ; preds = %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit, %ComputeLoop2420
  %Accumulator2422 = phi i32 [ %2784, %ComputeLoop2420 ], [ -1, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit ]
  %ActiveBits2423 = phi i64 [ %2787, %ComputeLoop2420 ], [ %126, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit ]
  %2781 = call i64 @llvm.cttz.i64(i64 %ActiveBits2423, i1 true)
  %2782 = trunc i64 %2781 to i32
  %2783 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i21.i.i, i32 %2782)
  %2784 = and i32 %Accumulator2422, %2783
  %2785 = shl i64 1, %2781
  %2786 = xor i64 %2785, -1
  %2787 = and i64 %ActiveBits2423, %2786
  %2788 = icmp eq i64 %2787, 0
  br i1 %2788, label %ComputeEnd2421, label %ComputeLoop2420, !amdgpu.uniform !42

ComputeEnd2421:                                   ; preds = %ComputeLoop2420
  %.lcssa4077 = phi i32 [ %2784, %ComputeLoop2420 ]
  %2789 = trunc i64 %126 to i32
  %2790 = lshr i64 %126, 32
  %2791 = trunc i64 %2790 to i32
  %2792 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %2789, i32 0)
  %2793 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %2791, i32 %2792)
  %2794 = icmp eq i32 %2793, 0
  %2795 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %2794)
  %2796 = extractvalue { i1, i64 } %2795, 0
  %2797 = extractvalue { i1, i64 } %2795, 1
  br i1 %2796, label %127, label %130

ComputeLoop2424:                                  ; preds = %for.cond.i.i1.i11, %ComputeLoop2424
  %Accumulator2426 = phi i32 [ %2802, %ComputeLoop2424 ], [ 0, %for.cond.i.i1.i11 ]
  %OldValuePhi2427 = phi i32 [ %2801, %ComputeLoop2424 ], [ poison, %for.cond.i.i1.i11 ]
  %ActiveBits2428 = phi i64 [ %2805, %ComputeLoop2424 ], [ %139, %for.cond.i.i1.i11 ]
  %2798 = call i64 @llvm.cttz.i64(i64 %ActiveBits2428, i1 true)
  %2799 = trunc i64 %2798 to i32
  %2800 = call i32 @llvm.amdgcn.readlane.i32(i32 %shl.i.i.i.i.i23, i32 %2799)
  %2801 = call i32 @llvm.amdgcn.writelane.i32(i32 %Accumulator2426, i32 %2799, i32 %OldValuePhi2427)
  %2802 = or i32 %Accumulator2426, %2800
  %2803 = shl i64 1, %2798
  %2804 = xor i64 %2803, -1
  %2805 = and i64 %ActiveBits2428, %2804
  %2806 = icmp eq i64 %2805, 0
  br i1 %2806, label %ComputeEnd2425, label %ComputeLoop2424, !amdgpu.uniform !42

ComputeEnd2425:                                   ; preds = %ComputeLoop2424
  %.lcssa4072 = phi i32 [ %2801, %ComputeLoop2424 ]
  %.lcssa4071 = phi i32 [ %2802, %ComputeLoop2424 ]
  %div8.i.i.i.i.i19 = lshr i32 %spec.store.select.i.i.i18, 5
  %idxprom.i.i.i.i.i20 = zext nneg i32 %div8.i.i.i.i.i19 to i64
  %2807 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i.i.i20
  %2808 = trunc i64 %139 to i32
  %2809 = lshr i64 %139, 32
  %2810 = trunc i64 %2809 to i32
  %2811 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %2808, i32 0)
  %2812 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %2810, i32 %2811)
  %2813 = icmp eq i32 %2812, 0
  %2814 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %2813)
  %2815 = extractvalue { i1, i64 } %2814, 0
  %2816 = extractvalue { i1, i64 } %2814, 1
  br i1 %2815, label %140, label %142

Flow3748:                                         ; preds = %160, %if.then.i.i.i.i29
  %2817 = phi i1 [ false, %160 ], [ true, %if.then.i.i.i.i29 ]
  br i1 %2817, label %if.end14.i.i.i99, label %Flow3749, !amdgpu.uniform !42

ComputeLoop2429:                                  ; preds = %if.then12.i.i.i34, %ComputeLoop2429
  %Accumulator2431 = phi i32 [ %2821, %ComputeLoop2429 ], [ -1, %if.then12.i.i.i34 ]
  %ActiveBits2432 = phi i64 [ %2824, %ComputeLoop2429 ], [ %157, %if.then12.i.i.i34 ]
  %2818 = call i64 @llvm.cttz.i64(i64 %ActiveBits2432, i1 true)
  %2819 = trunc i64 %2818 to i32
  %2820 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i.i.i40, i32 %2819)
  %2821 = and i32 %Accumulator2431, %2820
  %2822 = shl i64 1, %2818
  %2823 = xor i64 %2822, -1
  %2824 = and i64 %ActiveBits2432, %2823
  %2825 = icmp eq i64 %2824, 0
  br i1 %2825, label %ComputeEnd2430, label %ComputeLoop2429, !amdgpu.uniform !42

ComputeEnd2430:                                   ; preds = %ComputeLoop2429
  %.lcssa4073 = phi i32 [ %2821, %ComputeLoop2429 ]
  %2826 = trunc i64 %157 to i32
  %2827 = lshr i64 %157, 32
  %2828 = trunc i64 %2827 to i32
  %2829 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %2826, i32 0)
  %2830 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %2828, i32 %2829)
  %2831 = icmp eq i32 %2830, 0
  %2832 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %2831)
  %2833 = extractvalue { i1, i64 } %2832, 0
  %2834 = extractvalue { i1, i64 } %2832, 1
  br i1 %2833, label %158, label %160

ComputeLoop2433:                                  ; preds = %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit107, %ComputeLoop2433
  %Accumulator2435 = phi i32 [ %2838, %ComputeLoop2433 ], [ -1, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit107 ]
  %ActiveBits2436 = phi i64 [ %2841, %ComputeLoop2433 ], [ %237, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit107 ]
  %2835 = call i64 @llvm.cttz.i64(i64 %ActiveBits2436, i1 true)
  %2836 = trunc i64 %2835 to i32
  %2837 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i21.i.i98, i32 %2836)
  %2838 = and i32 %Accumulator2435, %2837
  %2839 = shl i64 1, %2835
  %2840 = xor i64 %2839, -1
  %2841 = and i64 %ActiveBits2436, %2840
  %2842 = icmp eq i64 %2841, 0
  br i1 %2842, label %ComputeEnd2434, label %ComputeLoop2433, !amdgpu.uniform !42

ComputeEnd2434:                                   ; preds = %ComputeLoop2433
  %.lcssa4067 = phi i32 [ %2838, %ComputeLoop2433 ]
  %2843 = trunc i64 %237 to i32
  %2844 = lshr i64 %237, 32
  %2845 = trunc i64 %2844 to i32
  %2846 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %2843, i32 0)
  %2847 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %2845, i32 %2846)
  %2848 = icmp eq i32 %2847, 0
  %2849 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %2848)
  %2850 = extractvalue { i1, i64 } %2849, 0
  %2851 = extractvalue { i1, i64 } %2849, 1
  br i1 %2850, label %238, label %241

ComputeLoop2437:                                  ; preds = %for.cond.i.i.i116, %ComputeLoop2437
  %Accumulator2439 = phi i32 [ %2856, %ComputeLoop2437 ], [ 0, %for.cond.i.i.i116 ]
  %OldValuePhi2440 = phi i32 [ %2855, %ComputeLoop2437 ], [ poison, %for.cond.i.i.i116 ]
  %ActiveBits2441 = phi i64 [ %2859, %ComputeLoop2437 ], [ %378, %for.cond.i.i.i116 ]
  %2852 = call i64 @llvm.cttz.i64(i64 %ActiveBits2441, i1 true)
  %2853 = trunc i64 %2852 to i32
  %2854 = call i32 @llvm.amdgcn.readlane.i32(i32 %shl.i.i.i.i.i128, i32 %2853)
  %2855 = call i32 @llvm.amdgcn.writelane.i32(i32 %Accumulator2439, i32 %2853, i32 %OldValuePhi2440)
  %2856 = or i32 %Accumulator2439, %2854
  %2857 = shl i64 1, %2852
  %2858 = xor i64 %2857, -1
  %2859 = and i64 %ActiveBits2441, %2858
  %2860 = icmp eq i64 %2859, 0
  br i1 %2860, label %ComputeEnd2438, label %ComputeLoop2437, !amdgpu.uniform !42

ComputeEnd2438:                                   ; preds = %ComputeLoop2437
  %.lcssa4058 = phi i32 [ %2855, %ComputeLoop2437 ]
  %.lcssa4057 = phi i32 [ %2856, %ComputeLoop2437 ]
  %div8.i.i.i.i.i124 = lshr i32 %spec.store.select.i.i.i123, 5
  %idxprom.i.i.i.i.i125 = zext nneg i32 %div8.i.i.i.i.i124 to i64
  %2861 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i.i.i125
  %2862 = trunc i64 %378 to i32
  %2863 = lshr i64 %378, 32
  %2864 = trunc i64 %2863 to i32
  %2865 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %2862, i32 0)
  %2866 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %2864, i32 %2865)
  %2867 = icmp eq i32 %2866, 0
  %2868 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %2867)
  %2869 = extractvalue { i1, i64 } %2868, 0
  %2870 = extractvalue { i1, i64 } %2868, 1
  br i1 %2869, label %379, label %381

Flow3703:                                         ; preds = %399, %if.then.i.i.i.i134
  %2871 = phi i1 [ false, %399 ], [ true, %if.then.i.i.i.i134 ]
  br i1 %2871, label %if.end14.i.i.i204, label %Flow3704, !amdgpu.uniform !42

ComputeLoop2442:                                  ; preds = %if.then12.i.i.i139, %ComputeLoop2442
  %Accumulator2444 = phi i32 [ %2875, %ComputeLoop2442 ], [ -1, %if.then12.i.i.i139 ]
  %ActiveBits2445 = phi i64 [ %2878, %ComputeLoop2442 ], [ %396, %if.then12.i.i.i139 ]
  %2872 = call i64 @llvm.cttz.i64(i64 %ActiveBits2445, i1 true)
  %2873 = trunc i64 %2872 to i32
  %2874 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i.i.i144, i32 %2873)
  %2875 = and i32 %Accumulator2444, %2874
  %2876 = shl i64 1, %2872
  %2877 = xor i64 %2876, -1
  %2878 = and i64 %ActiveBits2445, %2877
  %2879 = icmp eq i64 %2878, 0
  br i1 %2879, label %ComputeEnd2443, label %ComputeLoop2442, !amdgpu.uniform !42

ComputeEnd2443:                                   ; preds = %ComputeLoop2442
  %.lcssa4059 = phi i32 [ %2875, %ComputeLoop2442 ]
  %2880 = trunc i64 %396 to i32
  %2881 = lshr i64 %396, 32
  %2882 = trunc i64 %2881 to i32
  %2883 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %2880, i32 0)
  %2884 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %2882, i32 %2883)
  %2885 = icmp eq i32 %2884, 0
  %2886 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %2885)
  %2887 = extractvalue { i1, i64 } %2886, 0
  %2888 = extractvalue { i1, i64 } %2886, 1
  br i1 %2887, label %397, label %399

ComputeLoop2446:                                  ; preds = %_ZN22__llvm_libc_22_0_0_git15write_to_stderrENS_3cpp11string_viewE.exit.i, %ComputeLoop2446
  %Accumulator2448 = phi i32 [ %2892, %ComputeLoop2446 ], [ -1, %_ZN22__llvm_libc_22_0_0_git15write_to_stderrENS_3cpp11string_viewE.exit.i ]
  %ActiveBits2449 = phi i64 [ %2895, %ComputeLoop2446 ], [ %494, %_ZN22__llvm_libc_22_0_0_git15write_to_stderrENS_3cpp11string_viewE.exit.i ]
  %2889 = call i64 @llvm.cttz.i64(i64 %ActiveBits2449, i1 true)
  %2890 = trunc i64 %2889 to i32
  %2891 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i21.i.i201, i32 %2890)
  %2892 = and i32 %Accumulator2448, %2891
  %2893 = shl i64 1, %2889
  %2894 = xor i64 %2893, -1
  %2895 = and i64 %ActiveBits2449, %2894
  %2896 = icmp eq i64 %2895, 0
  br i1 %2896, label %ComputeEnd2447, label %ComputeLoop2446, !amdgpu.uniform !42

ComputeEnd2447:                                   ; preds = %ComputeLoop2446
  %.lcssa4049 = phi i32 [ %2892, %ComputeLoop2446 ]
  %2897 = trunc i64 %494 to i32
  %2898 = lshr i64 %494, 32
  %2899 = trunc i64 %2898 to i32
  %2900 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %2897, i32 0)
  %2901 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %2899, i32 %2900)
  %2902 = icmp eq i32 %2901, 0
  %2903 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %2902)
  %2904 = extractvalue { i1, i64 } %2903, 0
  %2905 = extractvalue { i1, i64 } %2903, 1
  br i1 %2904, label %495, label %498

ComputeLoop2450:                                  ; preds = %for.cond.i.i1.i220, %ComputeLoop2450
  %Accumulator2452 = phi i32 [ %2910, %ComputeLoop2450 ], [ 0, %for.cond.i.i1.i220 ]
  %OldValuePhi2453 = phi i32 [ %2909, %ComputeLoop2450 ], [ poison, %for.cond.i.i1.i220 ]
  %ActiveBits2454 = phi i64 [ %2913, %ComputeLoop2450 ], [ %507, %for.cond.i.i1.i220 ]
  %2906 = call i64 @llvm.cttz.i64(i64 %ActiveBits2454, i1 true)
  %2907 = trunc i64 %2906 to i32
  %2908 = call i32 @llvm.amdgcn.readlane.i32(i32 %shl.i.i.i.i.i232, i32 %2907)
  %2909 = call i32 @llvm.amdgcn.writelane.i32(i32 %Accumulator2452, i32 %2907, i32 %OldValuePhi2453)
  %2910 = or i32 %Accumulator2452, %2908
  %2911 = shl i64 1, %2906
  %2912 = xor i64 %2911, -1
  %2913 = and i64 %ActiveBits2454, %2912
  %2914 = icmp eq i64 %2913, 0
  br i1 %2914, label %ComputeEnd2451, label %ComputeLoop2450, !amdgpu.uniform !42

ComputeEnd2451:                                   ; preds = %ComputeLoop2450
  %.lcssa4044 = phi i32 [ %2909, %ComputeLoop2450 ]
  %.lcssa4043 = phi i32 [ %2910, %ComputeLoop2450 ]
  %div8.i.i.i.i.i228 = lshr i32 %spec.store.select.i.i.i227, 5
  %idxprom.i.i.i.i.i229 = zext nneg i32 %div8.i.i.i.i.i228 to i64
  %2915 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i.i.i229
  %2916 = trunc i64 %507 to i32
  %2917 = lshr i64 %507, 32
  %2918 = trunc i64 %2917 to i32
  %2919 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %2916, i32 0)
  %2920 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %2918, i32 %2919)
  %2921 = icmp eq i32 %2920, 0
  %2922 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %2921)
  %2923 = extractvalue { i1, i64 } %2922, 0
  %2924 = extractvalue { i1, i64 } %2922, 1
  br i1 %2923, label %508, label %510

Flow3684:                                         ; preds = %528, %if.then.i.i.i.i238
  %2925 = phi i1 [ false, %528 ], [ true, %if.then.i.i.i.i238 ]
  br i1 %2925, label %if.end14.i.i.i309, label %Flow3685, !amdgpu.uniform !42

ComputeLoop2455:                                  ; preds = %if.then12.i.i.i243, %ComputeLoop2455
  %Accumulator2457 = phi i32 [ %2929, %ComputeLoop2455 ], [ -1, %if.then12.i.i.i243 ]
  %ActiveBits2458 = phi i64 [ %2932, %ComputeLoop2455 ], [ %525, %if.then12.i.i.i243 ]
  %2926 = call i64 @llvm.cttz.i64(i64 %ActiveBits2458, i1 true)
  %2927 = trunc i64 %2926 to i32
  %2928 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i.i.i249, i32 %2927)
  %2929 = and i32 %Accumulator2457, %2928
  %2930 = shl i64 1, %2926
  %2931 = xor i64 %2930, -1
  %2932 = and i64 %ActiveBits2458, %2931
  %2933 = icmp eq i64 %2932, 0
  br i1 %2933, label %ComputeEnd2456, label %ComputeLoop2455, !amdgpu.uniform !42

ComputeEnd2456:                                   ; preds = %ComputeLoop2455
  %.lcssa4045 = phi i32 [ %2929, %ComputeLoop2455 ]
  %2934 = trunc i64 %525 to i32
  %2935 = lshr i64 %525, 32
  %2936 = trunc i64 %2935 to i32
  %2937 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %2934, i32 0)
  %2938 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %2936, i32 %2937)
  %2939 = icmp eq i32 %2938, 0
  %2940 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %2939)
  %2941 = extractvalue { i1, i64 } %2940, 0
  %2942 = extractvalue { i1, i64 } %2940, 1
  br i1 %2941, label %526, label %528

ComputeLoop2459:                                  ; preds = %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit317, %ComputeLoop2459
  %Accumulator2461 = phi i32 [ %2946, %ComputeLoop2459 ], [ -1, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit317 ]
  %ActiveBits2462 = phi i64 [ %2949, %ComputeLoop2459 ], [ %605, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit317 ]
  %2943 = call i64 @llvm.cttz.i64(i64 %ActiveBits2462, i1 true)
  %2944 = trunc i64 %2943 to i32
  %2945 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i21.i.i308, i32 %2944)
  %2946 = and i32 %Accumulator2461, %2945
  %2947 = shl i64 1, %2943
  %2948 = xor i64 %2947, -1
  %2949 = and i64 %ActiveBits2462, %2948
  %2950 = icmp eq i64 %2949, 0
  br i1 %2950, label %ComputeEnd2460, label %ComputeLoop2459, !amdgpu.uniform !42

ComputeEnd2460:                                   ; preds = %ComputeLoop2459
  %.lcssa4039 = phi i32 [ %2946, %ComputeLoop2459 ]
  %2951 = trunc i64 %605 to i32
  %2952 = lshr i64 %605, 32
  %2953 = trunc i64 %2952 to i32
  %2954 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %2951, i32 0)
  %2955 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %2953, i32 %2954)
  %2956 = icmp eq i32 %2955, 0
  %2957 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %2956)
  %2958 = extractvalue { i1, i64 } %2957, 0
  %2959 = extractvalue { i1, i64 } %2957, 1
  br i1 %2958, label %606, label %609

ComputeLoop2463:                                  ; preds = %for.cond.i.i.i326, %ComputeLoop2463
  %Accumulator2465 = phi i32 [ %2964, %ComputeLoop2463 ], [ 0, %for.cond.i.i.i326 ]
  %OldValuePhi2466 = phi i32 [ %2963, %ComputeLoop2463 ], [ poison, %for.cond.i.i.i326 ]
  %ActiveBits2467 = phi i64 [ %2967, %ComputeLoop2463 ], [ %630, %for.cond.i.i.i326 ]
  %2960 = call i64 @llvm.cttz.i64(i64 %ActiveBits2467, i1 true)
  %2961 = trunc i64 %2960 to i32
  %2962 = call i32 @llvm.amdgcn.readlane.i32(i32 %shl.i.i.i.i.i338, i32 %2961)
  %2963 = call i32 @llvm.amdgcn.writelane.i32(i32 %Accumulator2465, i32 %2961, i32 %OldValuePhi2466)
  %2964 = or i32 %Accumulator2465, %2962
  %2965 = shl i64 1, %2960
  %2966 = xor i64 %2965, -1
  %2967 = and i64 %ActiveBits2467, %2966
  %2968 = icmp eq i64 %2967, 0
  br i1 %2968, label %ComputeEnd2464, label %ComputeLoop2463, !amdgpu.uniform !42

ComputeEnd2464:                                   ; preds = %ComputeLoop2463
  %.lcssa4034 = phi i32 [ %2963, %ComputeLoop2463 ]
  %.lcssa4033 = phi i32 [ %2964, %ComputeLoop2463 ]
  %div8.i.i.i.i.i334 = lshr i32 %spec.store.select.i.i.i333, 5
  %idxprom.i.i.i.i.i335 = zext nneg i32 %div8.i.i.i.i.i334 to i64
  %2969 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i.i.i335
  %2970 = trunc i64 %630 to i32
  %2971 = lshr i64 %630, 32
  %2972 = trunc i64 %2971 to i32
  %2973 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %2970, i32 0)
  %2974 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %2972, i32 %2973)
  %2975 = icmp eq i32 %2974, 0
  %2976 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %2975)
  %2977 = extractvalue { i1, i64 } %2976, 0
  %2978 = extractvalue { i1, i64 } %2976, 1
  br i1 %2977, label %631, label %633

Flow3661:                                         ; preds = %651, %if.then.i.i.i.i344
  %2979 = phi i1 [ false, %651 ], [ true, %if.then.i.i.i.i344 ]
  br i1 %2979, label %if.end14.i.i.i415, label %Flow3662, !amdgpu.uniform !42

ComputeLoop2468:                                  ; preds = %if.then12.i.i.i349, %ComputeLoop2468
  %Accumulator2470 = phi i32 [ %2983, %ComputeLoop2468 ], [ -1, %if.then12.i.i.i349 ]
  %ActiveBits2471 = phi i64 [ %2986, %ComputeLoop2468 ], [ %648, %if.then12.i.i.i349 ]
  %2980 = call i64 @llvm.cttz.i64(i64 %ActiveBits2471, i1 true)
  %2981 = trunc i64 %2980 to i32
  %2982 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i.i.i355, i32 %2981)
  %2983 = and i32 %Accumulator2470, %2982
  %2984 = shl i64 1, %2980
  %2985 = xor i64 %2984, -1
  %2986 = and i64 %ActiveBits2471, %2985
  %2987 = icmp eq i64 %2986, 0
  br i1 %2987, label %ComputeEnd2469, label %ComputeLoop2468, !amdgpu.uniform !42

ComputeEnd2469:                                   ; preds = %ComputeLoop2468
  %.lcssa4035 = phi i32 [ %2983, %ComputeLoop2468 ]
  %2988 = trunc i64 %648 to i32
  %2989 = lshr i64 %648, 32
  %2990 = trunc i64 %2989 to i32
  %2991 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %2988, i32 0)
  %2992 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %2990, i32 %2991)
  %2993 = icmp eq i32 %2992, 0
  %2994 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %2993)
  %2995 = extractvalue { i1, i64 } %2994, 0
  %2996 = extractvalue { i1, i64 } %2994, 1
  br i1 %2995, label %649, label %651

ComputeLoop2472:                                  ; preds = %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsINS_3cpp6stringEEERS1_T_.exit, %ComputeLoop2472
  %Accumulator2474 = phi i32 [ %3000, %ComputeLoop2472 ], [ -1, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsINS_3cpp6stringEEERS1_T_.exit ]
  %ActiveBits2475 = phi i64 [ %3003, %ComputeLoop2472 ], [ %729, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsINS_3cpp6stringEEERS1_T_.exit ]
  %2997 = call i64 @llvm.cttz.i64(i64 %ActiveBits2475, i1 true)
  %2998 = trunc i64 %2997 to i32
  %2999 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i21.i.i414, i32 %2998)
  %3000 = and i32 %Accumulator2474, %2999
  %3001 = shl i64 1, %2997
  %3002 = xor i64 %3001, -1
  %3003 = and i64 %ActiveBits2475, %3002
  %3004 = icmp eq i64 %3003, 0
  br i1 %3004, label %ComputeEnd2473, label %ComputeLoop2472, !amdgpu.uniform !42

ComputeEnd2473:                                   ; preds = %ComputeLoop2472
  %.lcssa4029 = phi i32 [ %3000, %ComputeLoop2472 ]
  %3005 = trunc i64 %729 to i32
  %3006 = lshr i64 %729, 32
  %3007 = trunc i64 %3006 to i32
  %3008 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %3005, i32 0)
  %3009 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %3007, i32 %3008)
  %3010 = icmp eq i32 %3009, 0
  %3011 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3010)
  %3012 = extractvalue { i1, i64 } %3011, 0
  %3013 = extractvalue { i1, i64 } %3011, 1
  br i1 %3012, label %730, label %733

ComputeLoop2476:                                  ; preds = %for.cond.i.i1.i431, %ComputeLoop2476
  %Accumulator2478 = phi i32 [ %3018, %ComputeLoop2476 ], [ 0, %for.cond.i.i1.i431 ]
  %OldValuePhi2479 = phi i32 [ %3017, %ComputeLoop2476 ], [ poison, %for.cond.i.i1.i431 ]
  %ActiveBits2480 = phi i64 [ %3021, %ComputeLoop2476 ], [ %742, %for.cond.i.i1.i431 ]
  %3014 = call i64 @llvm.cttz.i64(i64 %ActiveBits2480, i1 true)
  %3015 = trunc i64 %3014 to i32
  %3016 = call i32 @llvm.amdgcn.readlane.i32(i32 %shl.i.i.i.i.i443, i32 %3015)
  %3017 = call i32 @llvm.amdgcn.writelane.i32(i32 %Accumulator2478, i32 %3015, i32 %OldValuePhi2479)
  %3018 = or i32 %Accumulator2478, %3016
  %3019 = shl i64 1, %3014
  %3020 = xor i64 %3019, -1
  %3021 = and i64 %ActiveBits2480, %3020
  %3022 = icmp eq i64 %3021, 0
  br i1 %3022, label %ComputeEnd2477, label %ComputeLoop2476, !amdgpu.uniform !42

ComputeEnd2477:                                   ; preds = %ComputeLoop2476
  %.lcssa4024 = phi i32 [ %3017, %ComputeLoop2476 ]
  %.lcssa4023 = phi i32 [ %3018, %ComputeLoop2476 ]
  %div8.i.i.i.i.i439 = lshr i32 %spec.store.select.i.i.i438, 5
  %idxprom.i.i.i.i.i440 = zext nneg i32 %div8.i.i.i.i.i439 to i64
  %3023 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i.i.i440
  %3024 = trunc i64 %742 to i32
  %3025 = lshr i64 %742, 32
  %3026 = trunc i64 %3025 to i32
  %3027 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %3024, i32 0)
  %3028 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %3026, i32 %3027)
  %3029 = icmp eq i32 %3028, 0
  %3030 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3029)
  %3031 = extractvalue { i1, i64 } %3030, 0
  %3032 = extractvalue { i1, i64 } %3030, 1
  br i1 %3031, label %743, label %745

Flow3642:                                         ; preds = %763, %if.then.i.i.i.i449
  %3033 = phi i1 [ false, %763 ], [ true, %if.then.i.i.i.i449 ]
  br i1 %3033, label %if.end14.i.i.i520, label %Flow3643, !amdgpu.uniform !42

ComputeLoop2481:                                  ; preds = %if.then12.i.i.i454, %ComputeLoop2481
  %Accumulator2483 = phi i32 [ %3037, %ComputeLoop2481 ], [ -1, %if.then12.i.i.i454 ]
  %ActiveBits2484 = phi i64 [ %3040, %ComputeLoop2481 ], [ %760, %if.then12.i.i.i454 ]
  %3034 = call i64 @llvm.cttz.i64(i64 %ActiveBits2484, i1 true)
  %3035 = trunc i64 %3034 to i32
  %3036 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i.i.i460, i32 %3035)
  %3037 = and i32 %Accumulator2483, %3036
  %3038 = shl i64 1, %3034
  %3039 = xor i64 %3038, -1
  %3040 = and i64 %ActiveBits2484, %3039
  %3041 = icmp eq i64 %3040, 0
  br i1 %3041, label %ComputeEnd2482, label %ComputeLoop2481, !amdgpu.uniform !42

ComputeEnd2482:                                   ; preds = %ComputeLoop2481
  %.lcssa4025 = phi i32 [ %3037, %ComputeLoop2481 ]
  %3042 = trunc i64 %760 to i32
  %3043 = lshr i64 %760, 32
  %3044 = trunc i64 %3043 to i32
  %3045 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %3042, i32 0)
  %3046 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %3044, i32 %3045)
  %3047 = icmp eq i32 %3046, 0
  %3048 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3047)
  %3049 = extractvalue { i1, i64 } %3048, 0
  %3050 = extractvalue { i1, i64 } %3048, 1
  br i1 %3049, label %761, label %763

ComputeLoop2485:                                  ; preds = %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit528, %ComputeLoop2485
  %Accumulator2487 = phi i32 [ %3054, %ComputeLoop2485 ], [ -1, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit528 ]
  %ActiveBits2488 = phi i64 [ %3057, %ComputeLoop2485 ], [ %840, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit528 ]
  %3051 = call i64 @llvm.cttz.i64(i64 %ActiveBits2488, i1 true)
  %3052 = trunc i64 %3051 to i32
  %3053 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i21.i.i519, i32 %3052)
  %3054 = and i32 %Accumulator2487, %3053
  %3055 = shl i64 1, %3051
  %3056 = xor i64 %3055, -1
  %3057 = and i64 %ActiveBits2488, %3056
  %3058 = icmp eq i64 %3057, 0
  br i1 %3058, label %ComputeEnd2486, label %ComputeLoop2485, !amdgpu.uniform !42

ComputeEnd2486:                                   ; preds = %ComputeLoop2485
  %.lcssa4019 = phi i32 [ %3054, %ComputeLoop2485 ]
  %3059 = trunc i64 %840 to i32
  %3060 = lshr i64 %840, 32
  %3061 = trunc i64 %3060 to i32
  %3062 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %3059, i32 0)
  %3063 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %3061, i32 %3062)
  %3064 = icmp eq i32 %3063, 0
  %3065 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3064)
  %3066 = extractvalue { i1, i64 } %3065, 0
  %3067 = extractvalue { i1, i64 } %3065, 1
  br i1 %3066, label %841, label %844

ComputeLoop2489:                                  ; preds = %for.cond.i.i1.i537, %ComputeLoop2489
  %Accumulator2491 = phi i32 [ %3072, %ComputeLoop2489 ], [ 0, %for.cond.i.i1.i537 ]
  %OldValuePhi2492 = phi i32 [ %3071, %ComputeLoop2489 ], [ poison, %for.cond.i.i1.i537 ]
  %ActiveBits2493 = phi i64 [ %3075, %ComputeLoop2489 ], [ %857, %for.cond.i.i1.i537 ]
  %3068 = call i64 @llvm.cttz.i64(i64 %ActiveBits2493, i1 true)
  %3069 = trunc i64 %3068 to i32
  %3070 = call i32 @llvm.amdgcn.readlane.i32(i32 %shl.i.i.i.i.i549, i32 %3069)
  %3071 = call i32 @llvm.amdgcn.writelane.i32(i32 %Accumulator2491, i32 %3069, i32 %OldValuePhi2492)
  %3072 = or i32 %Accumulator2491, %3070
  %3073 = shl i64 1, %3068
  %3074 = xor i64 %3073, -1
  %3075 = and i64 %ActiveBits2493, %3074
  %3076 = icmp eq i64 %3075, 0
  br i1 %3076, label %ComputeEnd2490, label %ComputeLoop2489, !amdgpu.uniform !42

ComputeEnd2490:                                   ; preds = %ComputeLoop2489
  %.lcssa4013 = phi i32 [ %3071, %ComputeLoop2489 ]
  %.lcssa4012 = phi i32 [ %3072, %ComputeLoop2489 ]
  %div8.i.i.i.i.i545 = lshr i32 %spec.store.select.i.i.i544, 5
  %idxprom.i.i.i.i.i546 = zext nneg i32 %div8.i.i.i.i.i545 to i64
  %3077 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i.i.i546
  %3078 = trunc i64 %857 to i32
  %3079 = lshr i64 %857, 32
  %3080 = trunc i64 %3079 to i32
  %3081 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %3078, i32 0)
  %3082 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %3080, i32 %3081)
  %3083 = icmp eq i32 %3082, 0
  %3084 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3083)
  %3085 = extractvalue { i1, i64 } %3084, 0
  %3086 = extractvalue { i1, i64 } %3084, 1
  br i1 %3085, label %858, label %860

Flow3623:                                         ; preds = %878, %if.then.i.i.i.i555
  %3087 = phi i1 [ false, %878 ], [ true, %if.then.i.i.i.i555 ]
  br i1 %3087, label %if.end14.i.i.i626, label %Flow3624, !amdgpu.uniform !42

ComputeLoop2494:                                  ; preds = %if.then12.i.i.i560, %ComputeLoop2494
  %Accumulator2496 = phi i32 [ %3091, %ComputeLoop2494 ], [ -1, %if.then12.i.i.i560 ]
  %ActiveBits2497 = phi i64 [ %3094, %ComputeLoop2494 ], [ %875, %if.then12.i.i.i560 ]
  %3088 = call i64 @llvm.cttz.i64(i64 %ActiveBits2497, i1 true)
  %3089 = trunc i64 %3088 to i32
  %3090 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i.i.i566, i32 %3089)
  %3091 = and i32 %Accumulator2496, %3090
  %3092 = shl i64 1, %3088
  %3093 = xor i64 %3092, -1
  %3094 = and i64 %ActiveBits2497, %3093
  %3095 = icmp eq i64 %3094, 0
  br i1 %3095, label %ComputeEnd2495, label %ComputeLoop2494, !amdgpu.uniform !42

ComputeEnd2495:                                   ; preds = %ComputeLoop2494
  %.lcssa4014 = phi i32 [ %3091, %ComputeLoop2494 ]
  %3096 = trunc i64 %875 to i32
  %3097 = lshr i64 %875, 32
  %3098 = trunc i64 %3097 to i32
  %3099 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %3096, i32 0)
  %3100 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %3098, i32 %3099)
  %3101 = icmp eq i32 %3100, 0
  %3102 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3101)
  %3103 = extractvalue { i1, i64 } %3102, 0
  %3104 = extractvalue { i1, i64 } %3102, 1
  br i1 %3103, label %876, label %878

ComputeLoop2498:                                  ; preds = %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit634, %ComputeLoop2498
  %Accumulator2500 = phi i32 [ %3108, %ComputeLoop2498 ], [ -1, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit634 ]
  %ActiveBits2501 = phi i64 [ %3111, %ComputeLoop2498 ], [ %973, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit634 ]
  %3105 = call i64 @llvm.cttz.i64(i64 %ActiveBits2501, i1 true)
  %3106 = trunc i64 %3105 to i32
  %3107 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i21.i.i625, i32 %3106)
  %3108 = and i32 %Accumulator2500, %3107
  %3109 = shl i64 1, %3105
  %3110 = xor i64 %3109, -1
  %3111 = and i64 %ActiveBits2501, %3110
  %3112 = icmp eq i64 %3111, 0
  br i1 %3112, label %ComputeEnd2499, label %ComputeLoop2498, !amdgpu.uniform !42

ComputeEnd2499:                                   ; preds = %ComputeLoop2498
  %.lcssa4004 = phi i32 [ %3108, %ComputeLoop2498 ]
  %3113 = trunc i64 %973 to i32
  %3114 = lshr i64 %973, 32
  %3115 = trunc i64 %3114 to i32
  %3116 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %3113, i32 0)
  %3117 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %3115, i32 %3116)
  %3118 = icmp eq i32 %3117, 0
  %3119 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3118)
  %3120 = extractvalue { i1, i64 } %3119, 0
  %3121 = extractvalue { i1, i64 } %3119, 1
  br i1 %3120, label %974, label %977

ComputeLoop2502:                                  ; preds = %for.cond.i.i.i636, %ComputeLoop2502
  %Accumulator2504 = phi i32 [ %3126, %ComputeLoop2502 ], [ 0, %for.cond.i.i.i636 ]
  %OldValuePhi2505 = phi i32 [ %3125, %ComputeLoop2502 ], [ poison, %for.cond.i.i.i636 ]
  %ActiveBits2506 = phi i64 [ %3129, %ComputeLoop2502 ], [ %986, %for.cond.i.i.i636 ]
  %3122 = call i64 @llvm.cttz.i64(i64 %ActiveBits2506, i1 true)
  %3123 = trunc i64 %3122 to i32
  %3124 = call i32 @llvm.amdgcn.readlane.i32(i32 %shl.i.i.i.i.i648, i32 %3123)
  %3125 = call i32 @llvm.amdgcn.writelane.i32(i32 %Accumulator2504, i32 %3123, i32 %OldValuePhi2505)
  %3126 = or i32 %Accumulator2504, %3124
  %3127 = shl i64 1, %3122
  %3128 = xor i64 %3127, -1
  %3129 = and i64 %ActiveBits2506, %3128
  %3130 = icmp eq i64 %3129, 0
  br i1 %3130, label %ComputeEnd2503, label %ComputeLoop2502, !amdgpu.uniform !42

ComputeEnd2503:                                   ; preds = %ComputeLoop2502
  %.lcssa3999 = phi i32 [ %3125, %ComputeLoop2502 ]
  %.lcssa3998 = phi i32 [ %3126, %ComputeLoop2502 ]
  %div8.i.i.i.i.i644 = lshr i32 %spec.store.select.i.i.i643, 5
  %idxprom.i.i.i.i.i645 = zext nneg i32 %div8.i.i.i.i.i644 to i64
  %3131 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i.i.i645
  %3132 = trunc i64 %986 to i32
  %3133 = lshr i64 %986, 32
  %3134 = trunc i64 %3133 to i32
  %3135 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %3132, i32 0)
  %3136 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %3134, i32 %3135)
  %3137 = icmp eq i32 %3136, 0
  %3138 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3137)
  %3139 = extractvalue { i1, i64 } %3138, 0
  %3140 = extractvalue { i1, i64 } %3138, 1
  br i1 %3139, label %987, label %989

Flow3604:                                         ; preds = %1007, %if.then.i.i.i.i654
  %3141 = phi i1 [ false, %1007 ], [ true, %if.then.i.i.i.i654 ]
  br i1 %3141, label %if.end14.i.i.i723, label %Flow3605, !amdgpu.uniform !42

ComputeLoop2507:                                  ; preds = %if.then12.i.i.i659, %ComputeLoop2507
  %Accumulator2509 = phi i32 [ %3145, %ComputeLoop2507 ], [ -1, %if.then12.i.i.i659 ]
  %ActiveBits2510 = phi i64 [ %3148, %ComputeLoop2507 ], [ %1004, %if.then12.i.i.i659 ]
  %3142 = call i64 @llvm.cttz.i64(i64 %ActiveBits2510, i1 true)
  %3143 = trunc i64 %3142 to i32
  %3144 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i.i.i665, i32 %3143)
  %3145 = and i32 %Accumulator2509, %3144
  %3146 = shl i64 1, %3142
  %3147 = xor i64 %3146, -1
  %3148 = and i64 %ActiveBits2510, %3147
  %3149 = icmp eq i64 %3148, 0
  br i1 %3149, label %ComputeEnd2508, label %ComputeLoop2507, !amdgpu.uniform !42

ComputeEnd2508:                                   ; preds = %ComputeLoop2507
  %.lcssa4000 = phi i32 [ %3145, %ComputeLoop2507 ]
  %3150 = trunc i64 %1004 to i32
  %3151 = lshr i64 %1004, 32
  %3152 = trunc i64 %3151 to i32
  %3153 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %3150, i32 0)
  %3154 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %3152, i32 %3153)
  %3155 = icmp eq i32 %3154, 0
  %3156 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3155)
  %3157 = extractvalue { i1, i64 } %3156, 0
  %3158 = extractvalue { i1, i64 } %3156, 1
  br i1 %3157, label %1005, label %1007

ComputeLoop2511:                                  ; preds = %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIcEERS1_T_.exit, %ComputeLoop2511
  %Accumulator2513 = phi i32 [ %3162, %ComputeLoop2511 ], [ -1, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIcEERS1_T_.exit ]
  %ActiveBits2514 = phi i64 [ %3165, %ComputeLoop2511 ], [ %1059, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIcEERS1_T_.exit ]
  %3159 = call i64 @llvm.cttz.i64(i64 %ActiveBits2514, i1 true)
  %3160 = trunc i64 %3159 to i32
  %3161 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i21.i.i722, i32 %3160)
  %3162 = and i32 %Accumulator2513, %3161
  %3163 = shl i64 1, %3159
  %3164 = xor i64 %3163, -1
  %3165 = and i64 %ActiveBits2514, %3164
  %3166 = icmp eq i64 %3165, 0
  br i1 %3166, label %ComputeEnd2512, label %ComputeLoop2511, !amdgpu.uniform !42

ComputeEnd2512:                                   ; preds = %ComputeLoop2511
  %.lcssa3994 = phi i32 [ %3162, %ComputeLoop2511 ]
  %3167 = trunc i64 %1059 to i32
  %3168 = lshr i64 %1059, 32
  %3169 = trunc i64 %3168 to i32
  %3170 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %3167, i32 0)
  %3171 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %3169, i32 %3170)
  %3172 = icmp eq i32 %3171, 0
  %3173 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3172)
  %3174 = extractvalue { i1, i64 } %3173, 0
  %3175 = extractvalue { i1, i64 } %3173, 1
  br i1 %3174, label %1060, label %1063

ComputeLoop2515:                                  ; preds = %for.cond.i.i.i739, %ComputeLoop2515
  %Accumulator2517 = phi i32 [ %3180, %ComputeLoop2515 ], [ 0, %for.cond.i.i.i739 ]
  %OldValuePhi2518 = phi i32 [ %3179, %ComputeLoop2515 ], [ poison, %for.cond.i.i.i739 ]
  %ActiveBits2519 = phi i64 [ %3183, %ComputeLoop2515 ], [ %1081, %for.cond.i.i.i739 ]
  %3176 = call i64 @llvm.cttz.i64(i64 %ActiveBits2519, i1 true)
  %3177 = trunc i64 %3176 to i32
  %3178 = call i32 @llvm.amdgcn.readlane.i32(i32 %shl.i.i.i.i.i751, i32 %3177)
  %3179 = call i32 @llvm.amdgcn.writelane.i32(i32 %Accumulator2517, i32 %3177, i32 %OldValuePhi2518)
  %3180 = or i32 %Accumulator2517, %3178
  %3181 = shl i64 1, %3176
  %3182 = xor i64 %3181, -1
  %3183 = and i64 %ActiveBits2519, %3182
  %3184 = icmp eq i64 %3183, 0
  br i1 %3184, label %ComputeEnd2516, label %ComputeLoop2515, !amdgpu.uniform !42

ComputeEnd2516:                                   ; preds = %ComputeLoop2515
  %.lcssa3989 = phi i32 [ %3179, %ComputeLoop2515 ]
  %.lcssa3988 = phi i32 [ %3180, %ComputeLoop2515 ]
  %div8.i.i.i.i.i747 = lshr i32 %spec.store.select.i.i.i746, 5
  %idxprom.i.i.i.i.i748 = zext nneg i32 %div8.i.i.i.i.i747 to i64
  %3185 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i.i.i748
  %3186 = trunc i64 %1081 to i32
  %3187 = lshr i64 %1081, 32
  %3188 = trunc i64 %3187 to i32
  %3189 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %3186, i32 0)
  %3190 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %3188, i32 %3189)
  %3191 = icmp eq i32 %3190, 0
  %3192 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3191)
  %3193 = extractvalue { i1, i64 } %3192, 0
  %3194 = extractvalue { i1, i64 } %3192, 1
  br i1 %3193, label %1082, label %1084

Flow3591:                                         ; preds = %1102, %if.then.i.i.i.i757
  %3195 = phi i1 [ false, %1102 ], [ true, %if.then.i.i.i.i757 ]
  br i1 %3195, label %if.end14.i.i.i828, label %Flow3592, !amdgpu.uniform !42

ComputeLoop2520:                                  ; preds = %if.then12.i.i.i762, %ComputeLoop2520
  %Accumulator2522 = phi i32 [ %3199, %ComputeLoop2520 ], [ -1, %if.then12.i.i.i762 ]
  %ActiveBits2523 = phi i64 [ %3202, %ComputeLoop2520 ], [ %1099, %if.then12.i.i.i762 ]
  %3196 = call i64 @llvm.cttz.i64(i64 %ActiveBits2523, i1 true)
  %3197 = trunc i64 %3196 to i32
  %3198 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i.i.i768, i32 %3197)
  %3199 = and i32 %Accumulator2522, %3198
  %3200 = shl i64 1, %3196
  %3201 = xor i64 %3200, -1
  %3202 = and i64 %ActiveBits2523, %3201
  %3203 = icmp eq i64 %3202, 0
  br i1 %3203, label %ComputeEnd2521, label %ComputeLoop2520, !amdgpu.uniform !42

ComputeEnd2521:                                   ; preds = %ComputeLoop2520
  %.lcssa3990 = phi i32 [ %3199, %ComputeLoop2520 ]
  %3204 = trunc i64 %1099 to i32
  %3205 = lshr i64 %1099, 32
  %3206 = trunc i64 %3205 to i32
  %3207 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %3204, i32 0)
  %3208 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %3206, i32 %3207)
  %3209 = icmp eq i32 %3208, 0
  %3210 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3209)
  %3211 = extractvalue { i1, i64 } %3210, 0
  %3212 = extractvalue { i1, i64 } %3210, 1
  br i1 %3211, label %1100, label %1102

ComputeLoop2524:                                  ; preds = %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsINS_3cpp6stringEEERS1_T_.exit836, %ComputeLoop2524
  %Accumulator2526 = phi i32 [ %3216, %ComputeLoop2524 ], [ -1, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsINS_3cpp6stringEEERS1_T_.exit836 ]
  %ActiveBits2527 = phi i64 [ %3219, %ComputeLoop2524 ], [ %1176, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsINS_3cpp6stringEEERS1_T_.exit836 ]
  %3213 = call i64 @llvm.cttz.i64(i64 %ActiveBits2527, i1 true)
  %3214 = trunc i64 %3213 to i32
  %3215 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i21.i.i827, i32 %3214)
  %3216 = and i32 %Accumulator2526, %3215
  %3217 = shl i64 1, %3213
  %3218 = xor i64 %3217, -1
  %3219 = and i64 %ActiveBits2527, %3218
  %3220 = icmp eq i64 %3219, 0
  br i1 %3220, label %ComputeEnd2525, label %ComputeLoop2524, !amdgpu.uniform !42

ComputeEnd2525:                                   ; preds = %ComputeLoop2524
  %.lcssa3984 = phi i32 [ %3216, %ComputeLoop2524 ]
  %3221 = trunc i64 %1176 to i32
  %3222 = lshr i64 %1176, 32
  %3223 = trunc i64 %3222 to i32
  %3224 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %3221, i32 0)
  %3225 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %3223, i32 %3224)
  %3226 = icmp eq i32 %3225, 0
  %3227 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3226)
  %3228 = extractvalue { i1, i64 } %3227, 0
  %3229 = extractvalue { i1, i64 } %3227, 1
  br i1 %3228, label %1177, label %1180

ComputeLoop2528:                                  ; preds = %for.cond.i.i1.i845, %ComputeLoop2528
  %Accumulator2530 = phi i32 [ %3234, %ComputeLoop2528 ], [ 0, %for.cond.i.i1.i845 ]
  %OldValuePhi2531 = phi i32 [ %3233, %ComputeLoop2528 ], [ poison, %for.cond.i.i1.i845 ]
  %ActiveBits2532 = phi i64 [ %3237, %ComputeLoop2528 ], [ %1189, %for.cond.i.i1.i845 ]
  %3230 = call i64 @llvm.cttz.i64(i64 %ActiveBits2532, i1 true)
  %3231 = trunc i64 %3230 to i32
  %3232 = call i32 @llvm.amdgcn.readlane.i32(i32 %shl.i.i.i.i.i857, i32 %3231)
  %3233 = call i32 @llvm.amdgcn.writelane.i32(i32 %Accumulator2530, i32 %3231, i32 %OldValuePhi2531)
  %3234 = or i32 %Accumulator2530, %3232
  %3235 = shl i64 1, %3230
  %3236 = xor i64 %3235, -1
  %3237 = and i64 %ActiveBits2532, %3236
  %3238 = icmp eq i64 %3237, 0
  br i1 %3238, label %ComputeEnd2529, label %ComputeLoop2528, !amdgpu.uniform !42

ComputeEnd2529:                                   ; preds = %ComputeLoop2528
  %.lcssa3979 = phi i32 [ %3233, %ComputeLoop2528 ]
  %.lcssa3978 = phi i32 [ %3234, %ComputeLoop2528 ]
  %div8.i.i.i.i.i853 = lshr i32 %spec.store.select.i.i.i852, 5
  %idxprom.i.i.i.i.i854 = zext nneg i32 %div8.i.i.i.i.i853 to i64
  %3239 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i.i.i854
  %3240 = trunc i64 %1189 to i32
  %3241 = lshr i64 %1189, 32
  %3242 = trunc i64 %3241 to i32
  %3243 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %3240, i32 0)
  %3244 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %3242, i32 %3243)
  %3245 = icmp eq i32 %3244, 0
  %3246 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3245)
  %3247 = extractvalue { i1, i64 } %3246, 0
  %3248 = extractvalue { i1, i64 } %3246, 1
  br i1 %3247, label %1190, label %1192

Flow3572:                                         ; preds = %1210, %if.then.i.i.i.i863
  %3249 = phi i1 [ false, %1210 ], [ true, %if.then.i.i.i.i863 ]
  br i1 %3249, label %if.end14.i.i.i934, label %Flow3573, !amdgpu.uniform !42

ComputeLoop2533:                                  ; preds = %if.then12.i.i.i868, %ComputeLoop2533
  %Accumulator2535 = phi i32 [ %3253, %ComputeLoop2533 ], [ -1, %if.then12.i.i.i868 ]
  %ActiveBits2536 = phi i64 [ %3256, %ComputeLoop2533 ], [ %1207, %if.then12.i.i.i868 ]
  %3250 = call i64 @llvm.cttz.i64(i64 %ActiveBits2536, i1 true)
  %3251 = trunc i64 %3250 to i32
  %3252 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i.i.i874, i32 %3251)
  %3253 = and i32 %Accumulator2535, %3252
  %3254 = shl i64 1, %3250
  %3255 = xor i64 %3254, -1
  %3256 = and i64 %ActiveBits2536, %3255
  %3257 = icmp eq i64 %3256, 0
  br i1 %3257, label %ComputeEnd2534, label %ComputeLoop2533, !amdgpu.uniform !42

ComputeEnd2534:                                   ; preds = %ComputeLoop2533
  %.lcssa3980 = phi i32 [ %3253, %ComputeLoop2533 ]
  %3258 = trunc i64 %1207 to i32
  %3259 = lshr i64 %1207, 32
  %3260 = trunc i64 %3259 to i32
  %3261 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %3258, i32 0)
  %3262 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %3260, i32 %3261)
  %3263 = icmp eq i32 %3262, 0
  %3264 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3263)
  %3265 = extractvalue { i1, i64 } %3264, 0
  %3266 = extractvalue { i1, i64 } %3264, 1
  br i1 %3265, label %1208, label %1210

ComputeLoop2537:                                  ; preds = %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit942, %ComputeLoop2537
  %Accumulator2539 = phi i32 [ %3270, %ComputeLoop2537 ], [ -1, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit942 ]
  %ActiveBits2540 = phi i64 [ %3273, %ComputeLoop2537 ], [ %1287, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit942 ]
  %3267 = call i64 @llvm.cttz.i64(i64 %ActiveBits2540, i1 true)
  %3268 = trunc i64 %3267 to i32
  %3269 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i21.i.i933, i32 %3268)
  %3270 = and i32 %Accumulator2539, %3269
  %3271 = shl i64 1, %3267
  %3272 = xor i64 %3271, -1
  %3273 = and i64 %ActiveBits2540, %3272
  %3274 = icmp eq i64 %3273, 0
  br i1 %3274, label %ComputeEnd2538, label %ComputeLoop2537, !amdgpu.uniform !42

ComputeEnd2538:                                   ; preds = %ComputeLoop2537
  %.lcssa3974 = phi i32 [ %3270, %ComputeLoop2537 ]
  %3275 = trunc i64 %1287 to i32
  %3276 = lshr i64 %1287, 32
  %3277 = trunc i64 %3276 to i32
  %3278 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %3275, i32 0)
  %3279 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %3277, i32 %3278)
  %3280 = icmp eq i32 %3279, 0
  %3281 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3280)
  %3282 = extractvalue { i1, i64 } %3281, 0
  %3283 = extractvalue { i1, i64 } %3281, 1
  br i1 %3282, label %1288, label %1291

ComputeLoop2541:                                  ; preds = %for.cond.i.i.i949, %ComputeLoop2541
  %Accumulator2543 = phi i32 [ %3288, %ComputeLoop2541 ], [ 0, %for.cond.i.i.i949 ]
  %OldValuePhi2544 = phi i32 [ %3287, %ComputeLoop2541 ], [ poison, %for.cond.i.i.i949 ]
  %ActiveBits2545 = phi i64 [ %3291, %ComputeLoop2541 ], [ %1332, %for.cond.i.i.i949 ]
  %3284 = call i64 @llvm.cttz.i64(i64 %ActiveBits2545, i1 true)
  %3285 = trunc i64 %3284 to i32
  %3286 = call i32 @llvm.amdgcn.readlane.i32(i32 %shl.i.i.i.i.i961, i32 %3285)
  %3287 = call i32 @llvm.amdgcn.writelane.i32(i32 %Accumulator2543, i32 %3285, i32 %OldValuePhi2544)
  %3288 = or i32 %Accumulator2543, %3286
  %3289 = shl i64 1, %3284
  %3290 = xor i64 %3289, -1
  %3291 = and i64 %ActiveBits2545, %3290
  %3292 = icmp eq i64 %3291, 0
  br i1 %3292, label %ComputeEnd2542, label %ComputeLoop2541, !amdgpu.uniform !42

ComputeEnd2542:                                   ; preds = %ComputeLoop2541
  %.lcssa3967 = phi i32 [ %3287, %ComputeLoop2541 ]
  %.lcssa3966 = phi i32 [ %3288, %ComputeLoop2541 ]
  %div8.i.i.i.i.i957 = lshr i32 %spec.store.select.i.i.i956, 5
  %idxprom.i.i.i.i.i958 = zext nneg i32 %div8.i.i.i.i.i957 to i64
  %3293 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i.i.i958
  %3294 = trunc i64 %1332 to i32
  %3295 = lshr i64 %1332, 32
  %3296 = trunc i64 %3295 to i32
  %3297 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %3294, i32 0)
  %3298 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %3296, i32 %3297)
  %3299 = icmp eq i32 %3298, 0
  %3300 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3299)
  %3301 = extractvalue { i1, i64 } %3300, 0
  %3302 = extractvalue { i1, i64 } %3300, 1
  br i1 %3301, label %1333, label %1335

Flow3542:                                         ; preds = %1353, %if.then.i.i.i.i967
  %3303 = phi i1 [ false, %1353 ], [ true, %if.then.i.i.i.i967 ]
  br i1 %3303, label %if.end14.i.i.i1038, label %Flow3543, !amdgpu.uniform !42

ComputeLoop2546:                                  ; preds = %if.then12.i.i.i972, %ComputeLoop2546
  %Accumulator2548 = phi i32 [ %3307, %ComputeLoop2546 ], [ -1, %if.then12.i.i.i972 ]
  %ActiveBits2549 = phi i64 [ %3310, %ComputeLoop2546 ], [ %1350, %if.then12.i.i.i972 ]
  %3304 = call i64 @llvm.cttz.i64(i64 %ActiveBits2549, i1 true)
  %3305 = trunc i64 %3304 to i32
  %3306 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i.i.i978, i32 %3305)
  %3307 = and i32 %Accumulator2548, %3306
  %3308 = shl i64 1, %3304
  %3309 = xor i64 %3308, -1
  %3310 = and i64 %ActiveBits2549, %3309
  %3311 = icmp eq i64 %3310, 0
  br i1 %3311, label %ComputeEnd2547, label %ComputeLoop2546, !amdgpu.uniform !42

ComputeEnd2547:                                   ; preds = %ComputeLoop2546
  %.lcssa3968 = phi i32 [ %3307, %ComputeLoop2546 ]
  %3312 = trunc i64 %1350 to i32
  %3313 = lshr i64 %1350, 32
  %3314 = trunc i64 %3313 to i32
  %3315 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %3312, i32 0)
  %3316 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %3314, i32 %3315)
  %3317 = icmp eq i32 %3316, 0
  %3318 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3317)
  %3319 = extractvalue { i1, i64 } %3318, 0
  %3320 = extractvalue { i1, i64 } %3318, 1
  br i1 %3319, label %1351, label %1353

ComputeLoop2550:                                  ; preds = %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsINS_3cpp6stringEEERS1_T_.exit1046, %ComputeLoop2550
  %Accumulator2552 = phi i32 [ %3324, %ComputeLoop2550 ], [ -1, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsINS_3cpp6stringEEERS1_T_.exit1046 ]
  %ActiveBits2553 = phi i64 [ %3327, %ComputeLoop2550 ], [ %1440, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsINS_3cpp6stringEEERS1_T_.exit1046 ]
  %3321 = call i64 @llvm.cttz.i64(i64 %ActiveBits2553, i1 true)
  %3322 = trunc i64 %3321 to i32
  %3323 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i21.i.i1037, i32 %3322)
  %3324 = and i32 %Accumulator2552, %3323
  %3325 = shl i64 1, %3321
  %3326 = xor i64 %3325, -1
  %3327 = and i64 %ActiveBits2553, %3326
  %3328 = icmp eq i64 %3327, 0
  br i1 %3328, label %ComputeEnd2551, label %ComputeLoop2550, !amdgpu.uniform !42

ComputeEnd2551:                                   ; preds = %ComputeLoop2550
  %.lcssa3958 = phi i32 [ %3324, %ComputeLoop2550 ]
  %3329 = trunc i64 %1440 to i32
  %3330 = lshr i64 %1440, 32
  %3331 = trunc i64 %3330 to i32
  %3332 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %3329, i32 0)
  %3333 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %3331, i32 %3332)
  %3334 = icmp eq i32 %3333, 0
  %3335 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3334)
  %3336 = extractvalue { i1, i64 } %3335, 0
  %3337 = extractvalue { i1, i64 } %3335, 1
  br i1 %3336, label %1441, label %1444

ComputeLoop2554:                                  ; preds = %for.cond.i.i.i1050, %ComputeLoop2554
  %Accumulator2556 = phi i32 [ %3342, %ComputeLoop2554 ], [ 0, %for.cond.i.i.i1050 ]
  %OldValuePhi2557 = phi i32 [ %3341, %ComputeLoop2554 ], [ poison, %for.cond.i.i.i1050 ]
  %ActiveBits2558 = phi i64 [ %3345, %ComputeLoop2554 ], [ %1453, %for.cond.i.i.i1050 ]
  %3338 = call i64 @llvm.cttz.i64(i64 %ActiveBits2558, i1 true)
  %3339 = trunc i64 %3338 to i32
  %3340 = call i32 @llvm.amdgcn.readlane.i32(i32 %shl.i.i.i.i.i1062, i32 %3339)
  %3341 = call i32 @llvm.amdgcn.writelane.i32(i32 %Accumulator2556, i32 %3339, i32 %OldValuePhi2557)
  %3342 = or i32 %Accumulator2556, %3340
  %3343 = shl i64 1, %3338
  %3344 = xor i64 %3343, -1
  %3345 = and i64 %ActiveBits2558, %3344
  %3346 = icmp eq i64 %3345, 0
  br i1 %3346, label %ComputeEnd2555, label %ComputeLoop2554, !amdgpu.uniform !42

ComputeEnd2555:                                   ; preds = %ComputeLoop2554
  %.lcssa3953 = phi i32 [ %3341, %ComputeLoop2554 ]
  %.lcssa3952 = phi i32 [ %3342, %ComputeLoop2554 ]
  %div8.i.i.i.i.i1058 = lshr i32 %spec.store.select.i.i.i1057, 5
  %idxprom.i.i.i.i.i1059 = zext nneg i32 %div8.i.i.i.i.i1058 to i64
  %3347 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i.i.i1059
  %3348 = trunc i64 %1453 to i32
  %3349 = lshr i64 %1453, 32
  %3350 = trunc i64 %3349 to i32
  %3351 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %3348, i32 0)
  %3352 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %3350, i32 %3351)
  %3353 = icmp eq i32 %3352, 0
  %3354 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3353)
  %3355 = extractvalue { i1, i64 } %3354, 0
  %3356 = extractvalue { i1, i64 } %3354, 1
  br i1 %3355, label %1454, label %1456

Flow3518:                                         ; preds = %1474, %if.then.i.i.i.i1068
  %3357 = phi i1 [ false, %1474 ], [ true, %if.then.i.i.i.i1068 ]
  br i1 %3357, label %if.end14.i.i.i1137, label %Flow3519, !amdgpu.uniform !42

ComputeLoop2559:                                  ; preds = %if.then12.i.i.i1073, %ComputeLoop2559
  %Accumulator2561 = phi i32 [ %3361, %ComputeLoop2559 ], [ -1, %if.then12.i.i.i1073 ]
  %ActiveBits2562 = phi i64 [ %3364, %ComputeLoop2559 ], [ %1471, %if.then12.i.i.i1073 ]
  %3358 = call i64 @llvm.cttz.i64(i64 %ActiveBits2562, i1 true)
  %3359 = trunc i64 %3358 to i32
  %3360 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i.i.i1079, i32 %3359)
  %3361 = and i32 %Accumulator2561, %3360
  %3362 = shl i64 1, %3358
  %3363 = xor i64 %3362, -1
  %3364 = and i64 %ActiveBits2562, %3363
  %3365 = icmp eq i64 %3364, 0
  br i1 %3365, label %ComputeEnd2560, label %ComputeLoop2559, !amdgpu.uniform !42

ComputeEnd2560:                                   ; preds = %ComputeLoop2559
  %.lcssa3954 = phi i32 [ %3361, %ComputeLoop2559 ]
  %3366 = trunc i64 %1471 to i32
  %3367 = lshr i64 %1471, 32
  %3368 = trunc i64 %3367 to i32
  %3369 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %3366, i32 0)
  %3370 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %3368, i32 %3369)
  %3371 = icmp eq i32 %3370, 0
  %3372 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3371)
  %3373 = extractvalue { i1, i64 } %3372, 0
  %3374 = extractvalue { i1, i64 } %3372, 1
  br i1 %3373, label %1472, label %1474

ComputeLoop2563:                                  ; preds = %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIcEERS1_T_.exit1145, %ComputeLoop2563
  %Accumulator2565 = phi i32 [ %3378, %ComputeLoop2563 ], [ -1, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIcEERS1_T_.exit1145 ]
  %ActiveBits2566 = phi i64 [ %3381, %ComputeLoop2563 ], [ %1526, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIcEERS1_T_.exit1145 ]
  %3375 = call i64 @llvm.cttz.i64(i64 %ActiveBits2566, i1 true)
  %3376 = trunc i64 %3375 to i32
  %3377 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i21.i.i1136, i32 %3376)
  %3378 = and i32 %Accumulator2565, %3377
  %3379 = shl i64 1, %3375
  %3380 = xor i64 %3379, -1
  %3381 = and i64 %ActiveBits2566, %3380
  %3382 = icmp eq i64 %3381, 0
  br i1 %3382, label %ComputeEnd2564, label %ComputeLoop2563, !amdgpu.uniform !42

ComputeEnd2564:                                   ; preds = %ComputeLoop2563
  %.lcssa3948 = phi i32 [ %3378, %ComputeLoop2563 ]
  %3383 = trunc i64 %1526 to i32
  %3384 = lshr i64 %1526, 32
  %3385 = trunc i64 %3384 to i32
  %3386 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %3383, i32 0)
  %3387 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %3385, i32 %3386)
  %3388 = icmp eq i32 %3387, 0
  %3389 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3388)
  %3390 = extractvalue { i1, i64 } %3389, 0
  %3391 = extractvalue { i1, i64 } %3389, 1
  br i1 %3390, label %1527, label %1530

ComputeLoop2567:                                  ; preds = %for.cond.i.i1.i1154, %ComputeLoop2567
  %Accumulator2569 = phi i32 [ %3396, %ComputeLoop2567 ], [ 0, %for.cond.i.i1.i1154 ]
  %OldValuePhi2570 = phi i32 [ %3395, %ComputeLoop2567 ], [ poison, %for.cond.i.i1.i1154 ]
  %ActiveBits2571 = phi i64 [ %3399, %ComputeLoop2567 ], [ %1539, %for.cond.i.i1.i1154 ]
  %3392 = call i64 @llvm.cttz.i64(i64 %ActiveBits2571, i1 true)
  %3393 = trunc i64 %3392 to i32
  %3394 = call i32 @llvm.amdgcn.readlane.i32(i32 %shl.i.i.i.i.i1166, i32 %3393)
  %3395 = call i32 @llvm.amdgcn.writelane.i32(i32 %Accumulator2569, i32 %3393, i32 %OldValuePhi2570)
  %3396 = or i32 %Accumulator2569, %3394
  %3397 = shl i64 1, %3392
  %3398 = xor i64 %3397, -1
  %3399 = and i64 %ActiveBits2571, %3398
  %3400 = icmp eq i64 %3399, 0
  br i1 %3400, label %ComputeEnd2568, label %ComputeLoop2567, !amdgpu.uniform !42

ComputeEnd2568:                                   ; preds = %ComputeLoop2567
  %.lcssa3943 = phi i32 [ %3395, %ComputeLoop2567 ]
  %.lcssa3942 = phi i32 [ %3396, %ComputeLoop2567 ]
  %div8.i.i.i.i.i1162 = lshr i32 %spec.store.select.i.i.i1161, 5
  %idxprom.i.i.i.i.i1163 = zext nneg i32 %div8.i.i.i.i.i1162 to i64
  %3401 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i.i.i1163
  %3402 = trunc i64 %1539 to i32
  %3403 = lshr i64 %1539, 32
  %3404 = trunc i64 %3403 to i32
  %3405 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %3402, i32 0)
  %3406 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %3404, i32 %3405)
  %3407 = icmp eq i32 %3406, 0
  %3408 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3407)
  %3409 = extractvalue { i1, i64 } %3408, 0
  %3410 = extractvalue { i1, i64 } %3408, 1
  br i1 %3409, label %1540, label %1542

Flow3507:                                         ; preds = %1560, %if.then.i.i.i.i1172
  %3411 = phi i1 [ false, %1560 ], [ true, %if.then.i.i.i.i1172 ]
  br i1 %3411, label %if.end14.i.i.i1243, label %Flow3508, !amdgpu.uniform !42

ComputeLoop2572:                                  ; preds = %if.then12.i.i.i1177, %ComputeLoop2572
  %Accumulator2574 = phi i32 [ %3415, %ComputeLoop2572 ], [ -1, %if.then12.i.i.i1177 ]
  %ActiveBits2575 = phi i64 [ %3418, %ComputeLoop2572 ], [ %1557, %if.then12.i.i.i1177 ]
  %3412 = call i64 @llvm.cttz.i64(i64 %ActiveBits2575, i1 true)
  %3413 = trunc i64 %3412 to i32
  %3414 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i.i.i1183, i32 %3413)
  %3415 = and i32 %Accumulator2574, %3414
  %3416 = shl i64 1, %3412
  %3417 = xor i64 %3416, -1
  %3418 = and i64 %ActiveBits2575, %3417
  %3419 = icmp eq i64 %3418, 0
  br i1 %3419, label %ComputeEnd2573, label %ComputeLoop2572, !amdgpu.uniform !42

ComputeEnd2573:                                   ; preds = %ComputeLoop2572
  %.lcssa3944 = phi i32 [ %3415, %ComputeLoop2572 ]
  %3420 = trunc i64 %1557 to i32
  %3421 = lshr i64 %1557, 32
  %3422 = trunc i64 %3421 to i32
  %3423 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %3420, i32 0)
  %3424 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %3422, i32 %3423)
  %3425 = icmp eq i32 %3424, 0
  %3426 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3425)
  %3427 = extractvalue { i1, i64 } %3426, 0
  %3428 = extractvalue { i1, i64 } %3426, 1
  br i1 %3427, label %1558, label %1560

ComputeLoop2576:                                  ; preds = %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit1251, %ComputeLoop2576
  %Accumulator2578 = phi i32 [ %3432, %ComputeLoop2576 ], [ -1, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit1251 ]
  %ActiveBits2579 = phi i64 [ %3435, %ComputeLoop2576 ], [ %1637, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit1251 ]
  %3429 = call i64 @llvm.cttz.i64(i64 %ActiveBits2579, i1 true)
  %3430 = trunc i64 %3429 to i32
  %3431 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i21.i.i1242, i32 %3430)
  %3432 = and i32 %Accumulator2578, %3431
  %3433 = shl i64 1, %3429
  %3434 = xor i64 %3433, -1
  %3435 = and i64 %ActiveBits2579, %3434
  %3436 = icmp eq i64 %3435, 0
  br i1 %3436, label %ComputeEnd2577, label %ComputeLoop2576, !amdgpu.uniform !42

ComputeEnd2577:                                   ; preds = %ComputeLoop2576
  %.lcssa3938 = phi i32 [ %3432, %ComputeLoop2576 ]
  %3437 = trunc i64 %1637 to i32
  %3438 = lshr i64 %1637, 32
  %3439 = trunc i64 %3438 to i32
  %3440 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %3437, i32 0)
  %3441 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %3439, i32 %3440)
  %3442 = icmp eq i32 %3441, 0
  %3443 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3442)
  %3444 = extractvalue { i1, i64 } %3443, 0
  %3445 = extractvalue { i1, i64 } %3443, 1
  br i1 %3444, label %1638, label %1641

ComputeLoop2580:                                  ; preds = %for.cond.i.i.i1253, %ComputeLoop2580
  %Accumulator2582 = phi i32 [ %3450, %ComputeLoop2580 ], [ 0, %for.cond.i.i.i1253 ]
  %OldValuePhi2583 = phi i32 [ %3449, %ComputeLoop2580 ], [ poison, %for.cond.i.i.i1253 ]
  %ActiveBits2584 = phi i64 [ %3453, %ComputeLoop2580 ], [ %1650, %for.cond.i.i.i1253 ]
  %3446 = call i64 @llvm.cttz.i64(i64 %ActiveBits2584, i1 true)
  %3447 = trunc i64 %3446 to i32
  %3448 = call i32 @llvm.amdgcn.readlane.i32(i32 %shl.i.i.i.i.i1265, i32 %3447)
  %3449 = call i32 @llvm.amdgcn.writelane.i32(i32 %Accumulator2582, i32 %3447, i32 %OldValuePhi2583)
  %3450 = or i32 %Accumulator2582, %3448
  %3451 = shl i64 1, %3446
  %3452 = xor i64 %3451, -1
  %3453 = and i64 %ActiveBits2584, %3452
  %3454 = icmp eq i64 %3453, 0
  br i1 %3454, label %ComputeEnd2581, label %ComputeLoop2580, !amdgpu.uniform !42

ComputeEnd2581:                                   ; preds = %ComputeLoop2580
  %.lcssa3933 = phi i32 [ %3449, %ComputeLoop2580 ]
  %.lcssa3932 = phi i32 [ %3450, %ComputeLoop2580 ]
  %div8.i.i.i.i.i1261 = lshr i32 %spec.store.select.i.i.i1260, 5
  %idxprom.i.i.i.i.i1262 = zext nneg i32 %div8.i.i.i.i.i1261 to i64
  %3455 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i.i.i1262
  %3456 = trunc i64 %1650 to i32
  %3457 = lshr i64 %1650, 32
  %3458 = trunc i64 %3457 to i32
  %3459 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %3456, i32 0)
  %3460 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %3458, i32 %3459)
  %3461 = icmp eq i32 %3460, 0
  %3462 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3461)
  %3463 = extractvalue { i1, i64 } %3462, 0
  %3464 = extractvalue { i1, i64 } %3462, 1
  br i1 %3463, label %1651, label %1653

Flow3481:                                         ; preds = %1671, %if.then.i.i.i.i1271
  %3465 = phi i1 [ false, %1671 ], [ true, %if.then.i.i.i.i1271 ]
  br i1 %3465, label %if.end14.i.i.i1342, label %Flow3482, !amdgpu.uniform !42

ComputeLoop2585:                                  ; preds = %if.then12.i.i.i1276, %ComputeLoop2585
  %Accumulator2587 = phi i32 [ %3469, %ComputeLoop2585 ], [ -1, %if.then12.i.i.i1276 ]
  %ActiveBits2588 = phi i64 [ %3472, %ComputeLoop2585 ], [ %1668, %if.then12.i.i.i1276 ]
  %3466 = call i64 @llvm.cttz.i64(i64 %ActiveBits2588, i1 true)
  %3467 = trunc i64 %3466 to i32
  %3468 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i.i.i1282, i32 %3467)
  %3469 = and i32 %Accumulator2587, %3468
  %3470 = shl i64 1, %3466
  %3471 = xor i64 %3470, -1
  %3472 = and i64 %ActiveBits2588, %3471
  %3473 = icmp eq i64 %3472, 0
  br i1 %3473, label %ComputeEnd2586, label %ComputeLoop2585, !amdgpu.uniform !42

ComputeEnd2586:                                   ; preds = %ComputeLoop2585
  %.lcssa3934 = phi i32 [ %3469, %ComputeLoop2585 ]
  %3474 = trunc i64 %1668 to i32
  %3475 = lshr i64 %1668, 32
  %3476 = trunc i64 %3475 to i32
  %3477 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %3474, i32 0)
  %3478 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %3476, i32 %3477)
  %3479 = icmp eq i32 %3478, 0
  %3480 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3479)
  %3481 = extractvalue { i1, i64 } %3480, 0
  %3482 = extractvalue { i1, i64 } %3480, 1
  br i1 %3481, label %1669, label %1671

ComputeLoop2589:                                  ; preds = %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsINS_3cpp11string_viewEEERS1_T_.exit, %ComputeLoop2589
  %Accumulator2591 = phi i32 [ %3486, %ComputeLoop2589 ], [ -1, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsINS_3cpp11string_viewEEERS1_T_.exit ]
  %ActiveBits2592 = phi i64 [ %3489, %ComputeLoop2589 ], [ %1748, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsINS_3cpp11string_viewEEERS1_T_.exit ]
  %3483 = call i64 @llvm.cttz.i64(i64 %ActiveBits2592, i1 true)
  %3484 = trunc i64 %3483 to i32
  %3485 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i21.i.i1341, i32 %3484)
  %3486 = and i32 %Accumulator2591, %3485
  %3487 = shl i64 1, %3483
  %3488 = xor i64 %3487, -1
  %3489 = and i64 %ActiveBits2592, %3488
  %3490 = icmp eq i64 %3489, 0
  br i1 %3490, label %ComputeEnd2590, label %ComputeLoop2589, !amdgpu.uniform !42

ComputeEnd2590:                                   ; preds = %ComputeLoop2589
  %.lcssa3928 = phi i32 [ %3486, %ComputeLoop2589 ]
  %3491 = trunc i64 %1748 to i32
  %3492 = lshr i64 %1748, 32
  %3493 = trunc i64 %3492 to i32
  %3494 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %3491, i32 0)
  %3495 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %3493, i32 %3494)
  %3496 = icmp eq i32 %3495, 0
  %3497 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3496)
  %3498 = extractvalue { i1, i64 } %3497, 0
  %3499 = extractvalue { i1, i64 } %3497, 1
  br i1 %3498, label %1749, label %1752

ComputeLoop2593:                                  ; preds = %for.cond.i.i1.i1358, %ComputeLoop2593
  %Accumulator2595 = phi i32 [ %3504, %ComputeLoop2593 ], [ 0, %for.cond.i.i1.i1358 ]
  %OldValuePhi2596 = phi i32 [ %3503, %ComputeLoop2593 ], [ poison, %for.cond.i.i1.i1358 ]
  %ActiveBits2597 = phi i64 [ %3507, %ComputeLoop2593 ], [ %1761, %for.cond.i.i1.i1358 ]
  %3500 = call i64 @llvm.cttz.i64(i64 %ActiveBits2597, i1 true)
  %3501 = trunc i64 %3500 to i32
  %3502 = call i32 @llvm.amdgcn.readlane.i32(i32 %shl.i.i.i.i.i1370, i32 %3501)
  %3503 = call i32 @llvm.amdgcn.writelane.i32(i32 %Accumulator2595, i32 %3501, i32 %OldValuePhi2596)
  %3504 = or i32 %Accumulator2595, %3502
  %3505 = shl i64 1, %3500
  %3506 = xor i64 %3505, -1
  %3507 = and i64 %ActiveBits2597, %3506
  %3508 = icmp eq i64 %3507, 0
  br i1 %3508, label %ComputeEnd2594, label %ComputeLoop2593, !amdgpu.uniform !42

ComputeEnd2594:                                   ; preds = %ComputeLoop2593
  %.lcssa3923 = phi i32 [ %3503, %ComputeLoop2593 ]
  %.lcssa3922 = phi i32 [ %3504, %ComputeLoop2593 ]
  %div8.i.i.i.i.i1366 = lshr i32 %spec.store.select.i.i.i1365, 5
  %idxprom.i.i.i.i.i1367 = zext nneg i32 %div8.i.i.i.i.i1366 to i64
  %3509 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i.i.i1367
  %3510 = trunc i64 %1761 to i32
  %3511 = lshr i64 %1761, 32
  %3512 = trunc i64 %3511 to i32
  %3513 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %3510, i32 0)
  %3514 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %3512, i32 %3513)
  %3515 = icmp eq i32 %3514, 0
  %3516 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3515)
  %3517 = extractvalue { i1, i64 } %3516, 0
  %3518 = extractvalue { i1, i64 } %3516, 1
  br i1 %3517, label %1762, label %1764

Flow3455:                                         ; preds = %1782, %if.then.i.i.i.i1376
  %3519 = phi i1 [ false, %1782 ], [ true, %if.then.i.i.i.i1376 ]
  br i1 %3519, label %if.end14.i.i.i1447, label %Flow3456, !amdgpu.uniform !42

ComputeLoop2598:                                  ; preds = %if.then12.i.i.i1381, %ComputeLoop2598
  %Accumulator2600 = phi i32 [ %3523, %ComputeLoop2598 ], [ -1, %if.then12.i.i.i1381 ]
  %ActiveBits2601 = phi i64 [ %3526, %ComputeLoop2598 ], [ %1779, %if.then12.i.i.i1381 ]
  %3520 = call i64 @llvm.cttz.i64(i64 %ActiveBits2601, i1 true)
  %3521 = trunc i64 %3520 to i32
  %3522 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i.i.i1387, i32 %3521)
  %3523 = and i32 %Accumulator2600, %3522
  %3524 = shl i64 1, %3520
  %3525 = xor i64 %3524, -1
  %3526 = and i64 %ActiveBits2601, %3525
  %3527 = icmp eq i64 %3526, 0
  br i1 %3527, label %ComputeEnd2599, label %ComputeLoop2598, !amdgpu.uniform !42

ComputeEnd2599:                                   ; preds = %ComputeLoop2598
  %.lcssa3924 = phi i32 [ %3523, %ComputeLoop2598 ]
  %3528 = trunc i64 %1779 to i32
  %3529 = lshr i64 %1779, 32
  %3530 = trunc i64 %3529 to i32
  %3531 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %3528, i32 0)
  %3532 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %3530, i32 %3531)
  %3533 = icmp eq i32 %3532, 0
  %3534 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3533)
  %3535 = extractvalue { i1, i64 } %3534, 0
  %3536 = extractvalue { i1, i64 } %3534, 1
  br i1 %3535, label %1780, label %1782

ComputeLoop2602:                                  ; preds = %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit1455, %ComputeLoop2602
  %Accumulator2604 = phi i32 [ %3540, %ComputeLoop2602 ], [ -1, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit1455 ]
  %ActiveBits2605 = phi i64 [ %3543, %ComputeLoop2602 ], [ %1859, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit1455 ]
  %3537 = call i64 @llvm.cttz.i64(i64 %ActiveBits2605, i1 true)
  %3538 = trunc i64 %3537 to i32
  %3539 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i21.i.i1446, i32 %3538)
  %3540 = and i32 %Accumulator2604, %3539
  %3541 = shl i64 1, %3537
  %3542 = xor i64 %3541, -1
  %3543 = and i64 %ActiveBits2605, %3542
  %3544 = icmp eq i64 %3543, 0
  br i1 %3544, label %ComputeEnd2603, label %ComputeLoop2602, !amdgpu.uniform !42

ComputeEnd2603:                                   ; preds = %ComputeLoop2602
  %.lcssa3918 = phi i32 [ %3540, %ComputeLoop2602 ]
  %3545 = trunc i64 %1859 to i32
  %3546 = lshr i64 %1859, 32
  %3547 = trunc i64 %3546 to i32
  %3548 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %3545, i32 0)
  %3549 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %3547, i32 %3548)
  %3550 = icmp eq i32 %3549, 0
  %3551 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3550)
  %3552 = extractvalue { i1, i64 } %3551, 0
  %3553 = extractvalue { i1, i64 } %3551, 1
  br i1 %3552, label %1860, label %1863

ComputeLoop2606:                                  ; preds = %for.cond.i.i1.i1465, %ComputeLoop2606
  %Accumulator2608 = phi i32 [ %3558, %ComputeLoop2606 ], [ 0, %for.cond.i.i1.i1465 ]
  %OldValuePhi2609 = phi i32 [ %3557, %ComputeLoop2606 ], [ poison, %for.cond.i.i1.i1465 ]
  %ActiveBits2610 = phi i64 [ %3561, %ComputeLoop2606 ], [ %1876, %for.cond.i.i1.i1465 ]
  %3554 = call i64 @llvm.cttz.i64(i64 %ActiveBits2610, i1 true)
  %3555 = trunc i64 %3554 to i32
  %3556 = call i32 @llvm.amdgcn.readlane.i32(i32 %shl.i.i.i.i.i1477, i32 %3555)
  %3557 = call i32 @llvm.amdgcn.writelane.i32(i32 %Accumulator2608, i32 %3555, i32 %OldValuePhi2609)
  %3558 = or i32 %Accumulator2608, %3556
  %3559 = shl i64 1, %3554
  %3560 = xor i64 %3559, -1
  %3561 = and i64 %ActiveBits2610, %3560
  %3562 = icmp eq i64 %3561, 0
  br i1 %3562, label %ComputeEnd2607, label %ComputeLoop2606, !amdgpu.uniform !42

ComputeEnd2607:                                   ; preds = %ComputeLoop2606
  %.lcssa3912 = phi i32 [ %3557, %ComputeLoop2606 ]
  %.lcssa3911 = phi i32 [ %3558, %ComputeLoop2606 ]
  %div8.i.i.i.i.i1473 = lshr i32 %spec.store.select.i.i.i1472, 5
  %idxprom.i.i.i.i.i1474 = zext nneg i32 %div8.i.i.i.i.i1473 to i64
  %3563 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i.i.i1474
  %3564 = trunc i64 %1876 to i32
  %3565 = lshr i64 %1876, 32
  %3566 = trunc i64 %3565 to i32
  %3567 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %3564, i32 0)
  %3568 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %3566, i32 %3567)
  %3569 = icmp eq i32 %3568, 0
  %3570 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3569)
  %3571 = extractvalue { i1, i64 } %3570, 0
  %3572 = extractvalue { i1, i64 } %3570, 1
  br i1 %3571, label %1877, label %1879

Flow3429:                                         ; preds = %1897, %if.then.i.i.i.i1483
  %3573 = phi i1 [ false, %1897 ], [ true, %if.then.i.i.i.i1483 ]
  br i1 %3573, label %if.end14.i.i.i1554, label %Flow3430, !amdgpu.uniform !42

ComputeLoop2611:                                  ; preds = %if.then12.i.i.i1488, %ComputeLoop2611
  %Accumulator2613 = phi i32 [ %3577, %ComputeLoop2611 ], [ -1, %if.then12.i.i.i1488 ]
  %ActiveBits2614 = phi i64 [ %3580, %ComputeLoop2611 ], [ %1894, %if.then12.i.i.i1488 ]
  %3574 = call i64 @llvm.cttz.i64(i64 %ActiveBits2614, i1 true)
  %3575 = trunc i64 %3574 to i32
  %3576 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i.i.i1494, i32 %3575)
  %3577 = and i32 %Accumulator2613, %3576
  %3578 = shl i64 1, %3574
  %3579 = xor i64 %3578, -1
  %3580 = and i64 %ActiveBits2614, %3579
  %3581 = icmp eq i64 %3580, 0
  br i1 %3581, label %ComputeEnd2612, label %ComputeLoop2611, !amdgpu.uniform !42

ComputeEnd2612:                                   ; preds = %ComputeLoop2611
  %.lcssa3913 = phi i32 [ %3577, %ComputeLoop2611 ]
  %3582 = trunc i64 %1894 to i32
  %3583 = lshr i64 %1894, 32
  %3584 = trunc i64 %3583 to i32
  %3585 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %3582, i32 0)
  %3586 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %3584, i32 %3585)
  %3587 = icmp eq i32 %3586, 0
  %3588 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3587)
  %3589 = extractvalue { i1, i64 } %3588, 0
  %3590 = extractvalue { i1, i64 } %3588, 1
  br i1 %3589, label %1895, label %1897

ComputeLoop2615:                                  ; preds = %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit1562, %ComputeLoop2615
  %Accumulator2617 = phi i32 [ %3594, %ComputeLoop2615 ], [ -1, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit1562 ]
  %ActiveBits2618 = phi i64 [ %3597, %ComputeLoop2615 ], [ %1992, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit1562 ]
  %3591 = call i64 @llvm.cttz.i64(i64 %ActiveBits2618, i1 true)
  %3592 = trunc i64 %3591 to i32
  %3593 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i21.i.i1553, i32 %3592)
  %3594 = and i32 %Accumulator2617, %3593
  %3595 = shl i64 1, %3591
  %3596 = xor i64 %3595, -1
  %3597 = and i64 %ActiveBits2618, %3596
  %3598 = icmp eq i64 %3597, 0
  br i1 %3598, label %ComputeEnd2616, label %ComputeLoop2615, !amdgpu.uniform !42

ComputeEnd2616:                                   ; preds = %ComputeLoop2615
  %.lcssa3903 = phi i32 [ %3594, %ComputeLoop2615 ]
  %3599 = trunc i64 %1992 to i32
  %3600 = lshr i64 %1992, 32
  %3601 = trunc i64 %3600 to i32
  %3602 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %3599, i32 0)
  %3603 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %3601, i32 %3602)
  %3604 = icmp eq i32 %3603, 0
  %3605 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3604)
  %3606 = extractvalue { i1, i64 } %3605, 0
  %3607 = extractvalue { i1, i64 } %3605, 1
  br i1 %3606, label %1993, label %1996

ComputeLoop2619:                                  ; preds = %for.cond.i.i.i1566, %ComputeLoop2619
  %Accumulator2621 = phi i32 [ %3612, %ComputeLoop2619 ], [ 0, %for.cond.i.i.i1566 ]
  %OldValuePhi2622 = phi i32 [ %3611, %ComputeLoop2619 ], [ poison, %for.cond.i.i.i1566 ]
  %ActiveBits2623 = phi i64 [ %3615, %ComputeLoop2619 ], [ %2005, %for.cond.i.i.i1566 ]
  %3608 = call i64 @llvm.cttz.i64(i64 %ActiveBits2623, i1 true)
  %3609 = trunc i64 %3608 to i32
  %3610 = call i32 @llvm.amdgcn.readlane.i32(i32 %shl.i.i.i.i.i1578, i32 %3609)
  %3611 = call i32 @llvm.amdgcn.writelane.i32(i32 %Accumulator2621, i32 %3609, i32 %OldValuePhi2622)
  %3612 = or i32 %Accumulator2621, %3610
  %3613 = shl i64 1, %3608
  %3614 = xor i64 %3613, -1
  %3615 = and i64 %ActiveBits2623, %3614
  %3616 = icmp eq i64 %3615, 0
  br i1 %3616, label %ComputeEnd2620, label %ComputeLoop2619, !amdgpu.uniform !42

ComputeEnd2620:                                   ; preds = %ComputeLoop2619
  %.lcssa3898 = phi i32 [ %3611, %ComputeLoop2619 ]
  %.lcssa3897 = phi i32 [ %3612, %ComputeLoop2619 ]
  %div8.i.i.i.i.i1574 = lshr i32 %spec.store.select.i.i.i1573, 5
  %idxprom.i.i.i.i.i1575 = zext nneg i32 %div8.i.i.i.i.i1574 to i64
  %3617 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i.i.i1575
  %3618 = trunc i64 %2005 to i32
  %3619 = lshr i64 %2005, 32
  %3620 = trunc i64 %3619 to i32
  %3621 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %3618, i32 0)
  %3622 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %3620, i32 %3621)
  %3623 = icmp eq i32 %3622, 0
  %3624 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3623)
  %3625 = extractvalue { i1, i64 } %3624, 0
  %3626 = extractvalue { i1, i64 } %3624, 1
  br i1 %3625, label %2006, label %2008

Flow3403:                                         ; preds = %2026, %if.then.i.i.i.i1584
  %3627 = phi i1 [ false, %2026 ], [ true, %if.then.i.i.i.i1584 ]
  br i1 %3627, label %if.end14.i.i.i1653, label %Flow3404, !amdgpu.uniform !42

ComputeLoop2624:                                  ; preds = %if.then12.i.i.i1589, %ComputeLoop2624
  %Accumulator2626 = phi i32 [ %3631, %ComputeLoop2624 ], [ -1, %if.then12.i.i.i1589 ]
  %ActiveBits2627 = phi i64 [ %3634, %ComputeLoop2624 ], [ %2023, %if.then12.i.i.i1589 ]
  %3628 = call i64 @llvm.cttz.i64(i64 %ActiveBits2627, i1 true)
  %3629 = trunc i64 %3628 to i32
  %3630 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i.i.i1595, i32 %3629)
  %3631 = and i32 %Accumulator2626, %3630
  %3632 = shl i64 1, %3628
  %3633 = xor i64 %3632, -1
  %3634 = and i64 %ActiveBits2627, %3633
  %3635 = icmp eq i64 %3634, 0
  br i1 %3635, label %ComputeEnd2625, label %ComputeLoop2624, !amdgpu.uniform !42

ComputeEnd2625:                                   ; preds = %ComputeLoop2624
  %.lcssa3899 = phi i32 [ %3631, %ComputeLoop2624 ]
  %3636 = trunc i64 %2023 to i32
  %3637 = lshr i64 %2023, 32
  %3638 = trunc i64 %3637 to i32
  %3639 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %3636, i32 0)
  %3640 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %3638, i32 %3639)
  %3641 = icmp eq i32 %3640, 0
  %3642 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3641)
  %3643 = extractvalue { i1, i64 } %3642, 0
  %3644 = extractvalue { i1, i64 } %3642, 1
  br i1 %3643, label %2024, label %2026

ComputeLoop2628:                                  ; preds = %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIcEERS1_T_.exit1661, %ComputeLoop2628
  %Accumulator2630 = phi i32 [ %3648, %ComputeLoop2628 ], [ -1, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIcEERS1_T_.exit1661 ]
  %ActiveBits2631 = phi i64 [ %3651, %ComputeLoop2628 ], [ %2078, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIcEERS1_T_.exit1661 ]
  %3645 = call i64 @llvm.cttz.i64(i64 %ActiveBits2631, i1 true)
  %3646 = trunc i64 %3645 to i32
  %3647 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i21.i.i1652, i32 %3646)
  %3648 = and i32 %Accumulator2630, %3647
  %3649 = shl i64 1, %3645
  %3650 = xor i64 %3649, -1
  %3651 = and i64 %ActiveBits2631, %3650
  %3652 = icmp eq i64 %3651, 0
  br i1 %3652, label %ComputeEnd2629, label %ComputeLoop2628, !amdgpu.uniform !42

ComputeEnd2629:                                   ; preds = %ComputeLoop2628
  %.lcssa3893 = phi i32 [ %3648, %ComputeLoop2628 ]
  %3653 = trunc i64 %2078 to i32
  %3654 = lshr i64 %2078, 32
  %3655 = trunc i64 %3654 to i32
  %3656 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %3653, i32 0)
  %3657 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %3655, i32 %3656)
  %3658 = icmp eq i32 %3657, 0
  %3659 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3658)
  %3660 = extractvalue { i1, i64 } %3659, 0
  %3661 = extractvalue { i1, i64 } %3659, 1
  br i1 %3660, label %2079, label %2082

ComputeLoop2632:                                  ; preds = %for.cond.i.i.i1670, %ComputeLoop2632
  %Accumulator2634 = phi i32 [ %3666, %ComputeLoop2632 ], [ 0, %for.cond.i.i.i1670 ]
  %OldValuePhi2635 = phi i32 [ %3665, %ComputeLoop2632 ], [ poison, %for.cond.i.i.i1670 ]
  %ActiveBits2636 = phi i64 [ %3669, %ComputeLoop2632 ], [ %2100, %for.cond.i.i.i1670 ]
  %3662 = call i64 @llvm.cttz.i64(i64 %ActiveBits2636, i1 true)
  %3663 = trunc i64 %3662 to i32
  %3664 = call i32 @llvm.amdgcn.readlane.i32(i32 %shl.i.i.i.i.i1682, i32 %3663)
  %3665 = call i32 @llvm.amdgcn.writelane.i32(i32 %Accumulator2634, i32 %3663, i32 %OldValuePhi2635)
  %3666 = or i32 %Accumulator2634, %3664
  %3667 = shl i64 1, %3662
  %3668 = xor i64 %3667, -1
  %3669 = and i64 %ActiveBits2636, %3668
  %3670 = icmp eq i64 %3669, 0
  br i1 %3670, label %ComputeEnd2633, label %ComputeLoop2632, !amdgpu.uniform !42

ComputeEnd2633:                                   ; preds = %ComputeLoop2632
  %.lcssa3888 = phi i32 [ %3665, %ComputeLoop2632 ]
  %.lcssa3887 = phi i32 [ %3666, %ComputeLoop2632 ]
  %div8.i.i.i.i.i1678 = lshr i32 %spec.store.select.i.i.i1677, 5
  %idxprom.i.i.i.i.i1679 = zext nneg i32 %div8.i.i.i.i.i1678 to i64
  %3671 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i.i.i1679
  %3672 = trunc i64 %2100 to i32
  %3673 = lshr i64 %2100, 32
  %3674 = trunc i64 %3673 to i32
  %3675 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %3672, i32 0)
  %3676 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %3674, i32 %3675)
  %3677 = icmp eq i32 %3676, 0
  %3678 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3677)
  %3679 = extractvalue { i1, i64 } %3678, 0
  %3680 = extractvalue { i1, i64 } %3678, 1
  br i1 %3679, label %2101, label %2103

Flow3386:                                         ; preds = %2121, %if.then.i.i.i.i1688
  %3681 = phi i1 [ false, %2121 ], [ true, %if.then.i.i.i.i1688 ]
  br i1 %3681, label %if.end14.i.i.i1759, label %Flow3387, !amdgpu.uniform !42

ComputeLoop2637:                                  ; preds = %if.then12.i.i.i1693, %ComputeLoop2637
  %Accumulator2639 = phi i32 [ %3685, %ComputeLoop2637 ], [ -1, %if.then12.i.i.i1693 ]
  %ActiveBits2640 = phi i64 [ %3688, %ComputeLoop2637 ], [ %2118, %if.then12.i.i.i1693 ]
  %3682 = call i64 @llvm.cttz.i64(i64 %ActiveBits2640, i1 true)
  %3683 = trunc i64 %3682 to i32
  %3684 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i.i.i1699, i32 %3683)
  %3685 = and i32 %Accumulator2639, %3684
  %3686 = shl i64 1, %3682
  %3687 = xor i64 %3686, -1
  %3688 = and i64 %ActiveBits2640, %3687
  %3689 = icmp eq i64 %3688, 0
  br i1 %3689, label %ComputeEnd2638, label %ComputeLoop2637, !amdgpu.uniform !42

ComputeEnd2638:                                   ; preds = %ComputeLoop2637
  %.lcssa3889 = phi i32 [ %3685, %ComputeLoop2637 ]
  %3690 = trunc i64 %2118 to i32
  %3691 = lshr i64 %2118, 32
  %3692 = trunc i64 %3691 to i32
  %3693 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %3690, i32 0)
  %3694 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %3692, i32 %3693)
  %3695 = icmp eq i32 %3694, 0
  %3696 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3695)
  %3697 = extractvalue { i1, i64 } %3696, 0
  %3698 = extractvalue { i1, i64 } %3696, 1
  br i1 %3697, label %2119, label %2121

ComputeLoop2641:                                  ; preds = %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsINS_3cpp6stringEEERS1_T_.exit1767, %ComputeLoop2641
  %Accumulator2643 = phi i32 [ %3702, %ComputeLoop2641 ], [ -1, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsINS_3cpp6stringEEERS1_T_.exit1767 ]
  %ActiveBits2644 = phi i64 [ %3705, %ComputeLoop2641 ], [ %2195, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsINS_3cpp6stringEEERS1_T_.exit1767 ]
  %3699 = call i64 @llvm.cttz.i64(i64 %ActiveBits2644, i1 true)
  %3700 = trunc i64 %3699 to i32
  %3701 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i21.i.i1758, i32 %3700)
  %3702 = and i32 %Accumulator2643, %3701
  %3703 = shl i64 1, %3699
  %3704 = xor i64 %3703, -1
  %3705 = and i64 %ActiveBits2644, %3704
  %3706 = icmp eq i64 %3705, 0
  br i1 %3706, label %ComputeEnd2642, label %ComputeLoop2641, !amdgpu.uniform !42

ComputeEnd2642:                                   ; preds = %ComputeLoop2641
  %.lcssa3883 = phi i32 [ %3702, %ComputeLoop2641 ]
  %3707 = trunc i64 %2195 to i32
  %3708 = lshr i64 %2195, 32
  %3709 = trunc i64 %3708 to i32
  %3710 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %3707, i32 0)
  %3711 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %3709, i32 %3710)
  %3712 = icmp eq i32 %3711, 0
  %3713 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3712)
  %3714 = extractvalue { i1, i64 } %3713, 0
  %3715 = extractvalue { i1, i64 } %3713, 1
  br i1 %3714, label %2196, label %2199

ComputeLoop2645:                                  ; preds = %for.cond.i.i1.i1776, %ComputeLoop2645
  %Accumulator2647 = phi i32 [ %3720, %ComputeLoop2645 ], [ 0, %for.cond.i.i1.i1776 ]
  %OldValuePhi2648 = phi i32 [ %3719, %ComputeLoop2645 ], [ poison, %for.cond.i.i1.i1776 ]
  %ActiveBits2649 = phi i64 [ %3723, %ComputeLoop2645 ], [ %2208, %for.cond.i.i1.i1776 ]
  %3716 = call i64 @llvm.cttz.i64(i64 %ActiveBits2649, i1 true)
  %3717 = trunc i64 %3716 to i32
  %3718 = call i32 @llvm.amdgcn.readlane.i32(i32 %shl.i.i.i.i.i1788, i32 %3717)
  %3719 = call i32 @llvm.amdgcn.writelane.i32(i32 %Accumulator2647, i32 %3717, i32 %OldValuePhi2648)
  %3720 = or i32 %Accumulator2647, %3718
  %3721 = shl i64 1, %3716
  %3722 = xor i64 %3721, -1
  %3723 = and i64 %ActiveBits2649, %3722
  %3724 = icmp eq i64 %3723, 0
  br i1 %3724, label %ComputeEnd2646, label %ComputeLoop2645, !amdgpu.uniform !42

ComputeEnd2646:                                   ; preds = %ComputeLoop2645
  %.lcssa3878 = phi i32 [ %3719, %ComputeLoop2645 ]
  %.lcssa3877 = phi i32 [ %3720, %ComputeLoop2645 ]
  %div8.i.i.i.i.i1784 = lshr i32 %spec.store.select.i.i.i1783, 5
  %idxprom.i.i.i.i.i1785 = zext nneg i32 %div8.i.i.i.i.i1784 to i64
  %3725 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i.i.i1785
  %3726 = trunc i64 %2208 to i32
  %3727 = lshr i64 %2208, 32
  %3728 = trunc i64 %3727 to i32
  %3729 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %3726, i32 0)
  %3730 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %3728, i32 %3729)
  %3731 = icmp eq i32 %3730, 0
  %3732 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3731)
  %3733 = extractvalue { i1, i64 } %3732, 0
  %3734 = extractvalue { i1, i64 } %3732, 1
  br i1 %3733, label %2209, label %2211

Flow3360:                                         ; preds = %2229, %if.then.i.i.i.i1794
  %3735 = phi i1 [ false, %2229 ], [ true, %if.then.i.i.i.i1794 ]
  br i1 %3735, label %if.end14.i.i.i1865, label %Flow3361, !amdgpu.uniform !42

ComputeLoop2650:                                  ; preds = %if.then12.i.i.i1799, %ComputeLoop2650
  %Accumulator2652 = phi i32 [ %3739, %ComputeLoop2650 ], [ -1, %if.then12.i.i.i1799 ]
  %ActiveBits2653 = phi i64 [ %3742, %ComputeLoop2650 ], [ %2226, %if.then12.i.i.i1799 ]
  %3736 = call i64 @llvm.cttz.i64(i64 %ActiveBits2653, i1 true)
  %3737 = trunc i64 %3736 to i32
  %3738 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i.i.i1805, i32 %3737)
  %3739 = and i32 %Accumulator2652, %3738
  %3740 = shl i64 1, %3736
  %3741 = xor i64 %3740, -1
  %3742 = and i64 %ActiveBits2653, %3741
  %3743 = icmp eq i64 %3742, 0
  br i1 %3743, label %ComputeEnd2651, label %ComputeLoop2650, !amdgpu.uniform !42

ComputeEnd2651:                                   ; preds = %ComputeLoop2650
  %.lcssa3879 = phi i32 [ %3739, %ComputeLoop2650 ]
  %3744 = trunc i64 %2226 to i32
  %3745 = lshr i64 %2226, 32
  %3746 = trunc i64 %3745 to i32
  %3747 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %3744, i32 0)
  %3748 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %3746, i32 %3747)
  %3749 = icmp eq i32 %3748, 0
  %3750 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3749)
  %3751 = extractvalue { i1, i64 } %3750, 0
  %3752 = extractvalue { i1, i64 } %3750, 1
  br i1 %3751, label %2227, label %2229

ComputeLoop2654:                                  ; preds = %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit1873, %ComputeLoop2654
  %Accumulator2656 = phi i32 [ %3756, %ComputeLoop2654 ], [ -1, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit1873 ]
  %ActiveBits2657 = phi i64 [ %3759, %ComputeLoop2654 ], [ %2303, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit1873 ]
  %3753 = call i64 @llvm.cttz.i64(i64 %ActiveBits2657, i1 true)
  %3754 = trunc i64 %3753 to i32
  %3755 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i21.i.i1864, i32 %3754)
  %3756 = and i32 %Accumulator2656, %3755
  %3757 = shl i64 1, %3753
  %3758 = xor i64 %3757, -1
  %3759 = and i64 %ActiveBits2657, %3758
  %3760 = icmp eq i64 %3759, 0
  br i1 %3760, label %ComputeEnd2655, label %ComputeLoop2654, !amdgpu.uniform !42

ComputeEnd2655:                                   ; preds = %ComputeLoop2654
  %.lcssa3873 = phi i32 [ %3756, %ComputeLoop2654 ]
  %3761 = trunc i64 %2303 to i32
  %3762 = lshr i64 %2303, 32
  %3763 = trunc i64 %3762 to i32
  %3764 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %3761, i32 0)
  %3765 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %3763, i32 %3764)
  %3766 = icmp eq i32 %3765, 0
  %3767 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3766)
  %3768 = extractvalue { i1, i64 } %3767, 0
  %3769 = extractvalue { i1, i64 } %3767, 1
  br i1 %3768, label %2304, label %2307

ComputeLoop2658:                                  ; preds = %for.cond.i.i.i1880, %ComputeLoop2658
  %Accumulator2660 = phi i32 [ %3774, %ComputeLoop2658 ], [ 0, %for.cond.i.i.i1880 ]
  %OldValuePhi2661 = phi i32 [ %3773, %ComputeLoop2658 ], [ poison, %for.cond.i.i.i1880 ]
  %ActiveBits2662 = phi i64 [ %3777, %ComputeLoop2658 ], [ %2348, %for.cond.i.i.i1880 ]
  %3770 = call i64 @llvm.cttz.i64(i64 %ActiveBits2662, i1 true)
  %3771 = trunc i64 %3770 to i32
  %3772 = call i32 @llvm.amdgcn.readlane.i32(i32 %shl.i.i.i.i.i1892, i32 %3771)
  %3773 = call i32 @llvm.amdgcn.writelane.i32(i32 %Accumulator2660, i32 %3771, i32 %OldValuePhi2661)
  %3774 = or i32 %Accumulator2660, %3772
  %3775 = shl i64 1, %3770
  %3776 = xor i64 %3775, -1
  %3777 = and i64 %ActiveBits2662, %3776
  %3778 = icmp eq i64 %3777, 0
  br i1 %3778, label %ComputeEnd2659, label %ComputeLoop2658, !amdgpu.uniform !42

ComputeEnd2659:                                   ; preds = %ComputeLoop2658
  %.lcssa3866 = phi i32 [ %3773, %ComputeLoop2658 ]
  %.lcssa3865 = phi i32 [ %3774, %ComputeLoop2658 ]
  %div8.i.i.i.i.i1888 = lshr i32 %spec.store.select.i.i.i1887, 5
  %idxprom.i.i.i.i.i1889 = zext nneg i32 %div8.i.i.i.i.i1888 to i64
  %3779 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i.i.i1889
  %3780 = trunc i64 %2348 to i32
  %3781 = lshr i64 %2348, 32
  %3782 = trunc i64 %3781 to i32
  %3783 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %3780, i32 0)
  %3784 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %3782, i32 %3783)
  %3785 = icmp eq i32 %3784, 0
  %3786 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3785)
  %3787 = extractvalue { i1, i64 } %3786, 0
  %3788 = extractvalue { i1, i64 } %3786, 1
  br i1 %3787, label %2349, label %2351

Flow3336:                                         ; preds = %2369, %if.then.i.i.i.i1898
  %3789 = phi i1 [ false, %2369 ], [ true, %if.then.i.i.i.i1898 ]
  br i1 %3789, label %if.end14.i.i.i1969, label %Flow3337, !amdgpu.uniform !42

ComputeLoop2663:                                  ; preds = %if.then12.i.i.i1903, %ComputeLoop2663
  %Accumulator2665 = phi i32 [ %3793, %ComputeLoop2663 ], [ -1, %if.then12.i.i.i1903 ]
  %ActiveBits2666 = phi i64 [ %3796, %ComputeLoop2663 ], [ %2366, %if.then12.i.i.i1903 ]
  %3790 = call i64 @llvm.cttz.i64(i64 %ActiveBits2666, i1 true)
  %3791 = trunc i64 %3790 to i32
  %3792 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i.i.i1909, i32 %3791)
  %3793 = and i32 %Accumulator2665, %3792
  %3794 = shl i64 1, %3790
  %3795 = xor i64 %3794, -1
  %3796 = and i64 %ActiveBits2666, %3795
  %3797 = icmp eq i64 %3796, 0
  br i1 %3797, label %ComputeEnd2664, label %ComputeLoop2663, !amdgpu.uniform !42

ComputeEnd2664:                                   ; preds = %ComputeLoop2663
  %.lcssa3867 = phi i32 [ %3793, %ComputeLoop2663 ]
  %3798 = trunc i64 %2366 to i32
  %3799 = lshr i64 %2366, 32
  %3800 = trunc i64 %3799 to i32
  %3801 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %3798, i32 0)
  %3802 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %3800, i32 %3801)
  %3803 = icmp eq i32 %3802, 0
  %3804 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3803)
  %3805 = extractvalue { i1, i64 } %3804, 0
  %3806 = extractvalue { i1, i64 } %3804, 1
  br i1 %3805, label %2367, label %2369

ComputeLoop2667:                                  ; preds = %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsINS_3cpp6stringEEERS1_T_.exit1977, %ComputeLoop2667
  %Accumulator2669 = phi i32 [ %3810, %ComputeLoop2667 ], [ -1, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsINS_3cpp6stringEEERS1_T_.exit1977 ]
  %ActiveBits2670 = phi i64 [ %3813, %ComputeLoop2667 ], [ %2456, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsINS_3cpp6stringEEERS1_T_.exit1977 ]
  %3807 = call i64 @llvm.cttz.i64(i64 %ActiveBits2670, i1 true)
  %3808 = trunc i64 %3807 to i32
  %3809 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i21.i.i1968, i32 %3808)
  %3810 = and i32 %Accumulator2669, %3809
  %3811 = shl i64 1, %3807
  %3812 = xor i64 %3811, -1
  %3813 = and i64 %ActiveBits2670, %3812
  %3814 = icmp eq i64 %3813, 0
  br i1 %3814, label %ComputeEnd2668, label %ComputeLoop2667, !amdgpu.uniform !42

ComputeEnd2668:                                   ; preds = %ComputeLoop2667
  %.lcssa3857 = phi i32 [ %3810, %ComputeLoop2667 ]
  %3815 = trunc i64 %2456 to i32
  %3816 = lshr i64 %2456, 32
  %3817 = trunc i64 %3816 to i32
  %3818 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %3815, i32 0)
  %3819 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %3817, i32 %3818)
  %3820 = icmp eq i32 %3819, 0
  %3821 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3820)
  %3822 = extractvalue { i1, i64 } %3821, 0
  %3823 = extractvalue { i1, i64 } %3821, 1
  br i1 %3822, label %2457, label %2460

ComputeLoop2671:                                  ; preds = %for.cond.i.i.i1981, %ComputeLoop2671
  %Accumulator2673 = phi i32 [ %3828, %ComputeLoop2671 ], [ 0, %for.cond.i.i.i1981 ]
  %OldValuePhi2674 = phi i32 [ %3827, %ComputeLoop2671 ], [ poison, %for.cond.i.i.i1981 ]
  %ActiveBits2675 = phi i64 [ %3831, %ComputeLoop2671 ], [ %2469, %for.cond.i.i.i1981 ]
  %3824 = call i64 @llvm.cttz.i64(i64 %ActiveBits2675, i1 true)
  %3825 = trunc i64 %3824 to i32
  %3826 = call i32 @llvm.amdgcn.readlane.i32(i32 %shl.i.i.i.i.i1993, i32 %3825)
  %3827 = call i32 @llvm.amdgcn.writelane.i32(i32 %Accumulator2673, i32 %3825, i32 %OldValuePhi2674)
  %3828 = or i32 %Accumulator2673, %3826
  %3829 = shl i64 1, %3824
  %3830 = xor i64 %3829, -1
  %3831 = and i64 %ActiveBits2675, %3830
  %3832 = icmp eq i64 %3831, 0
  br i1 %3832, label %ComputeEnd2672, label %ComputeLoop2671, !amdgpu.uniform !42

ComputeEnd2672:                                   ; preds = %ComputeLoop2671
  %.lcssa3852 = phi i32 [ %3827, %ComputeLoop2671 ]
  %.lcssa3851 = phi i32 [ %3828, %ComputeLoop2671 ]
  %div8.i.i.i.i.i1989 = lshr i32 %spec.store.select.i.i.i1988, 5
  %idxprom.i.i.i.i.i1990 = zext nneg i32 %div8.i.i.i.i.i1989 to i64
  %3833 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i.i.i1990
  %3834 = trunc i64 %2469 to i32
  %3835 = lshr i64 %2469, 32
  %3836 = trunc i64 %3835 to i32
  %3837 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %3834, i32 0)
  %3838 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %3836, i32 %3837)
  %3839 = icmp eq i32 %3838, 0
  %3840 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3839)
  %3841 = extractvalue { i1, i64 } %3840, 0
  %3842 = extractvalue { i1, i64 } %3840, 1
  br i1 %3841, label %2470, label %2472

Flow3318:                                         ; preds = %2490, %if.then.i.i.i.i1999
  %3843 = phi i1 [ false, %2490 ], [ true, %if.then.i.i.i.i1999 ]
  br i1 %3843, label %if.end14.i.i.i2068, label %Flow3319, !amdgpu.uniform !42

ComputeLoop2676:                                  ; preds = %if.then12.i.i.i2004, %ComputeLoop2676
  %Accumulator2678 = phi i32 [ %3847, %ComputeLoop2676 ], [ -1, %if.then12.i.i.i2004 ]
  %ActiveBits2679 = phi i64 [ %3850, %ComputeLoop2676 ], [ %2487, %if.then12.i.i.i2004 ]
  %3844 = call i64 @llvm.cttz.i64(i64 %ActiveBits2679, i1 true)
  %3845 = trunc i64 %3844 to i32
  %3846 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i.i.i2010, i32 %3845)
  %3847 = and i32 %Accumulator2678, %3846
  %3848 = shl i64 1, %3844
  %3849 = xor i64 %3848, -1
  %3850 = and i64 %ActiveBits2679, %3849
  %3851 = icmp eq i64 %3850, 0
  br i1 %3851, label %ComputeEnd2677, label %ComputeLoop2676, !amdgpu.uniform !42

ComputeEnd2677:                                   ; preds = %ComputeLoop2676
  %.lcssa3853 = phi i32 [ %3847, %ComputeLoop2676 ]
  %3852 = trunc i64 %2487 to i32
  %3853 = lshr i64 %2487, 32
  %3854 = trunc i64 %3853 to i32
  %3855 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %3852, i32 0)
  %3856 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %3854, i32 %3855)
  %3857 = icmp eq i32 %3856, 0
  %3858 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3857)
  %3859 = extractvalue { i1, i64 } %3858, 0
  %3860 = extractvalue { i1, i64 } %3858, 1
  br i1 %3859, label %2488, label %2490

ComputeLoop2680:                                  ; preds = %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIcEERS1_T_.exit2076, %ComputeLoop2680
  %Accumulator2682 = phi i32 [ %3864, %ComputeLoop2680 ], [ -1, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIcEERS1_T_.exit2076 ]
  %ActiveBits2683 = phi i64 [ %3867, %ComputeLoop2680 ], [ %2542, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIcEERS1_T_.exit2076 ]
  %3861 = call i64 @llvm.cttz.i64(i64 %ActiveBits2683, i1 true)
  %3862 = trunc i64 %3861 to i32
  %3863 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i21.i.i2067, i32 %3862)
  %3864 = and i32 %Accumulator2682, %3863
  %3865 = shl i64 1, %3861
  %3866 = xor i64 %3865, -1
  %3867 = and i64 %ActiveBits2683, %3866
  %3868 = icmp eq i64 %3867, 0
  br i1 %3868, label %ComputeEnd2681, label %ComputeLoop2680, !amdgpu.uniform !42

ComputeEnd2681:                                   ; preds = %ComputeLoop2680
  %.lcssa = phi i32 [ %3864, %ComputeLoop2680 ]
  %3869 = trunc i64 %2542 to i32
  %3870 = lshr i64 %2542, 32
  %3871 = trunc i64 %3870 to i32
  %3872 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %3869, i32 0)
  %3873 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %3871, i32 %3872)
  %3874 = icmp eq i32 %3873, 0
  %3875 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3874)
  %3876 = extractvalue { i1, i64 } %3875, 0
  %3877 = extractvalue { i1, i64 } %3875, 1
  br i1 %3876, label %2543, label %2546
}

; Function Attrs: convergent mustprogress nounwind
define internal fastcc noundef range(i32 0, 2) i32 @_ZN22__llvm_libc_22_0_0_git7testing4Test8runTestsERKNS0_11TestOptionsE(ptr noundef nonnull readonly align 8 captures(none) dereferenceable(10) %Options) unnamed_addr #13 align 2 {
entry:
  %buffer.i.i.i2988 = alloca %"class.__llvm_libc_22_0_0_git::IntegerToString.21", align 8, addrspace(5)
  %buffer.i.i.i4144 = alloca %"class.__llvm_libc_22_0_0_git::IntegerToString", align 8, addrspace(5)
  %buffer.i.i.i3760 = alloca %"class.__llvm_libc_22_0_0_git::IntegerToString", align 8, addrspace(5)
  %buffer.i.i.i3369 = alloca %"class.__llvm_libc_22_0_0_git::IntegerToString", align 8, addrspace(5)
  %buffer.i.i.i2713 = alloca %"class.__llvm_libc_22_0_0_git::IntegerToString.21", align 8, addrspace(5)
  %buffer.i.i.i2448 = alloca %"class.__llvm_libc_22_0_0_git::IntegerToString.21", align 8, addrspace(5)
  %buffer.i.i.i = alloca %"class.__llvm_libc_22_0_0_git::IntegerToString", align 8, addrspace(5)
  %Ctx = alloca %"struct.__llvm_libc_22_0_0_git::testing::TestLogger", align 1, addrspace(5), !amdgpu.uniform !42
  %PrintColor = getelementptr inbounds nuw i8, ptr %Options, i64 8
  %0 = load i8, ptr %PrintColor, align 8, !tbaa !149, !range !152, !noundef !42
  %T.04.i = load ptr, ptr addrspace(1) @_ZN22__llvm_libc_22_0_0_git7testing4Test5StartE, align 8, !tbaa !63
  %tobool.not5.i = icmp ne ptr %T.04.i, null
  br i1 %tobool.not5.i, label %for.inc.i.preheader, label %Flow8611, !amdgpu.uniform !42

for.inc.i.preheader:                              ; preds = %entry
  br label %for.inc.i, !amdgpu.uniform !42

Flow8611:                                         ; preds = %Flow8610, %entry
  %1 = phi i32 [ %inc.i.lcssa, %Flow8610 ], [ 0, %entry ]
  br label %_ZN22__llvm_libc_22_0_0_git7testing4Test11getNumTestsEv.exit, !amdgpu.uniform !42

for.inc.i:                                        ; preds = %for.inc.i.preheader, %for.inc.i
  %phi.broken = phi i64 [ 0, %for.inc.i.preheader ], [ %2, %for.inc.i ]
  %T.07.i = phi ptr [ %T.0.i, %for.inc.i ], [ %T.04.i, %for.inc.i.preheader ]
  %N.06.i = phi i32 [ %inc.i, %for.inc.i ], [ 0, %for.inc.i.preheader ]
  %Next.i = getelementptr inbounds nuw i8, ptr %T.07.i, i64 8
  %inc.i = add nuw i32 %N.06.i, 1
  %T.0.i = load ptr, ptr %Next.i, align 8, !tbaa !63
  %tobool.not.i = icmp eq ptr %T.0.i, null
  %2 = call i64 @llvm.amdgcn.if.break.i64(i1 %tobool.not.i, i64 %phi.broken)
  %3 = call i1 @llvm.amdgcn.loop.i64(i64 %2)
  br i1 %3, label %Flow8610, label %for.inc.i

Flow8610:                                         ; preds = %for.inc.i
  %inc.i.lcssa = phi i32 [ %inc.i, %for.inc.i ]
  %.lcssa9365 = phi i64 [ %2, %for.inc.i ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa9365)
  br label %Flow8611, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git7testing4Test11getNumTestsEv.exit: ; preds = %Flow8611
  %loadedv = trunc nuw i8 %0 to i1
  %cond = select i1 %loadedv, ptr addrspace(4) @.str.8.26, ptr addrspace(4) @.str.9.27
  %cond6 = select i1 %loadedv, ptr addrspace(4) @.str.11.29, ptr addrspace(4) @.str.9.27
  %tobool.not = icmp ne i32 %1, 0
  %4 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %tobool.not)
  %5 = extractvalue { i1, i64 } %4, 0
  %6 = extractvalue { i1, i64 } %4, 1
  br i1 %5, label %for.cond.i.i.i.preheader, label %Flow8609

for.cond.i.i.i.preheader:                         ; preds = %_ZN22__llvm_libc_22_0_0_git7testing4Test11getNumTestsEv.exit
  %scevgep7277 = getelementptr i8, ptr addrspace(4) %cond, i64 -1
  br label %for.cond.i.i.i, !amdgpu.uniform !42

Flow8609:                                         ; preds = %1099, %_ZN22__llvm_libc_22_0_0_git7testing4Test11getNumTestsEv.exit
  %7 = phi ptr [ %T.0201.pre, %1099 ], [ %T.04.i, %_ZN22__llvm_libc_22_0_0_git7testing4Test11getNumTestsEv.exit ]
  call void @llvm.amdgcn.end.cf.i64(i64 %6)
  br label %if.end16, !amdgpu.uniform !42

for.cond.i.i.i:                                   ; preds = %for.cond.i.i.i.preheader, %for.cond.i.i.i
  %phi.broken8612 = phi i64 [ 0, %for.cond.i.i.i.preheader ], [ %9, %for.cond.i.i.i ]
  %lsr.iv7278 = phi ptr addrspace(4) [ %scevgep7280, %for.cond.i.i.i ], [ %scevgep7277, %for.cond.i.i.i.preheader ]
  %scevgep7280 = getelementptr i8, ptr addrspace(4) %lsr.iv7278, i64 1
  %8 = load i8, ptr addrspace(4) %scevgep7280, align 1, !tbaa !57
  %cmp.i.i.i = icmp eq i8 %8, 0
  %9 = call i64 @llvm.amdgcn.if.break.i64(i1 %cmp.i.i.i, i64 %phi.broken8612)
  %10 = call i1 @llvm.amdgcn.loop.i64(i64 %9)
  br i1 %10, label %_ZN22__llvm_libc_22_0_0_git3cpp11string_viewC2EPKc.exit.i, label %for.cond.i.i.i

_ZN22__llvm_libc_22_0_0_git3cpp11string_viewC2EPKc.exit.i: ; preds = %for.cond.i.i.i
  %scevgep7280.lcssa = phi ptr addrspace(4) [ %scevgep7280, %for.cond.i.i.i ]
  %.lcssa9364 = phi i64 [ %9, %for.cond.i.i.i ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa9364)
  %11 = tail call i32 @llvm.amdgcn.mbcnt.lo(i32 -1, i32 0)
  %12 = tail call noundef i32 @llvm.amdgcn.mbcnt.hi(i32 -1, i32 %11)
  %sh_prom.i.i.i.i = zext i32 %12 to i64
  br label %for.cond.i.i1.i, !amdgpu.uniform !42

for.cond.i.i1.i:                                  ; preds = %_ZN22__llvm_libc_22_0_0_git3cpp11string_viewC2EPKc.exit.i, %Flow8608
  %13 = phi i32 [ %51, %Flow8608 ], [ poison, %_ZN22__llvm_libc_22_0_0_git3cpp11string_viewC2EPKc.exit.i ]
  %14 = phi i64 [ %52, %Flow8608 ], [ poison, %_ZN22__llvm_libc_22_0_0_git3cpp11string_viewC2EPKc.exit.i ]
  %15 = phi i32 [ %53, %Flow8608 ], [ poison, %_ZN22__llvm_libc_22_0_0_git3cpp11string_viewC2EPKc.exit.i ]
  %retval.sroa.6.0.i.i.i = phi i32 [ %56, %Flow8608 ], [ undef, %_ZN22__llvm_libc_22_0_0_git3cpp11string_viewC2EPKc.exit.i ]
  %retval.sroa.2.0.i.i.i = phi i64 [ %55, %Flow8608 ], [ undef, %_ZN22__llvm_libc_22_0_0_git3cpp11string_viewC2EPKc.exit.i ]
  %retval.sroa.8.0.i.i.i = phi i32 [ %54, %Flow8608 ], [ undef, %_ZN22__llvm_libc_22_0_0_git3cpp11string_viewC2EPKc.exit.i ]
  %index.0.i.i.i = phi i32 [ %58, %Flow8608 ], [ 0, %_ZN22__llvm_libc_22_0_0_git3cpp11string_viewC2EPKc.exit.i ]
  %16 = load i32, ptr addrspace(1) @__llvm_rpc_client, align 8, !tbaa !68
  %cmp.not.i.i.i = icmp ult i32 %index.0.i.i.i, %16
  %spec.store.select.i.i.i = select i1 %cmp.not.i.i.i, i32 %index.0.i.i.i, i32 0
  %17 = tail call noundef i64 @llvm.amdgcn.ballot.i64(i1 true)
  %18 = and i32 %spec.store.select.i.i.i, 31
  %19 = lshr i64 %17, %sh_prom.i.i.i.i
  %20 = trunc i64 %19 to i32
  %conv4.i.i.i.i.i = and i32 %20, 1
  %shl.i.i.i.i.i = shl nuw i32 %conv4.i.i.i.i.i, %18
  %21 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop, !amdgpu.uniform !42

22:                                               ; preds = %ComputeEnd
  %sunkaddr = getelementptr inbounds i8, ptr addrspace(1) %6698, i64 40
  %23 = atomicrmw or ptr addrspace(1) %sunkaddr, i32 %.lcssa9358 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %24, !amdgpu.uniform !42

24:                                               ; preds = %22, %ComputeEnd
  %25 = phi i32 [ %23, %22 ], [ poison, %ComputeEnd ]
  call void @llvm.amdgcn.end.cf.i64(i64 %6707)
  %26 = call i32 @llvm.amdgcn.readfirstlane.i32(i32 %25)
  %27 = or i32 %26, %.lcssa9359
  %shl5.i.i.i.i.i = shl nuw i32 1, %18
  %and.i.i.i.i.i = and i32 %shl5.i.i.i.i.i, %27
  %tobool3.i.i.i.i = icmp ne i32 %and.i.i.i.i.i, 0
  %28 = tail call i64 @llvm.amdgcn.ballot.i64(i1 %tobool3.i.i.i.i)
  %and.i.i.i.i.i.i = and i64 %28, %17
  %cmp.i.not.i.i.i = icmp ne i64 %17, %and.i.i.i.i.i.i
  br i1 %cmp.i.not.i.i.i, label %if.then.i.i.i.i2, label %Flow8607, !amdgpu.uniform !42

if.then.i.i.i.i2:                                 ; preds = %24
  fence syncscope("agent") acquire
  %29 = load <2 x i64>, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !75
  %30 = extractelement <2 x i64> %29, i32 0
  %31 = inttoptr i64 %30 to ptr
  %32 = extractelement <2 x i64> %29, i32 1
  %33 = inttoptr i64 %32 to ptr
  %idxprom.i.i.i.i = zext i32 %spec.store.select.i.i.i to i64
  %arrayidx.i.i.i.i3 = getelementptr inbounds nuw i32, ptr %31, i64 %idxprom.i.i.i.i, !amdgpu.uniform !42
  %34 = load atomic i32, ptr %arrayidx.i.i.i.i3 monotonic, align 4
  %35 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %34)
  %arrayidx.i45.i.i.i = getelementptr inbounds nuw i32, ptr %33, i64 %idxprom.i.i.i.i, !amdgpu.uniform !42
  %36 = load atomic i32, ptr %arrayidx.i45.i.i.i monotonic, align 4
  %37 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %36)
  %cmp.i46.not.i.i.i = icmp ne i32 %35, %37
  br i1 %cmp.i46.not.i.i.i, label %if.then12.i.i.i, label %Flow8605, !amdgpu.uniform !42

if.then12.i.i.i:                                  ; preds = %if.then.i.i.i.i2
  fence syncscope("agent") release
  %38 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %17, i1 true)
  %iszero.i.i.i.i.i.i.i = icmp ne i64 %17, 0
  %cmp2.i.i.i.i.i.i = icmp eq i64 %38, %sh_prom.i.i.i.i
  %cmp.i.i.i.i.i.i = select i1 %iszero.i.i.i.i.i.i.i, i1 %cmp2.i.i.i.i.i.i, i1 false
  %conv4.i.i51.i.i.i = zext i1 %cmp.i.i.i.i.i.i to i32
  %shl.i.i52.i.i.i = shl nuw i32 %conv4.i.i51.i.i.i, %18
  %xor.i.i.i.i.i = xor i32 %shl.i.i52.i.i.i, -1
  %39 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop5468, !amdgpu.uniform !42

40:                                               ; preds = %ComputeEnd5469
  %sunkaddr7363 = getelementptr inbounds i8, ptr addrspace(1) %6698, i64 40
  %41 = atomicrmw and ptr addrspace(1) %sunkaddr7363, i32 %.lcssa9360 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %42, !amdgpu.uniform !42

42:                                               ; preds = %40, %ComputeEnd5469
  call void @llvm.amdgcn.end.cf.i64(i64 %6725)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %Flow8605, !amdgpu.uniform !42

if.end14.i.i.i:                                   ; preds = %Flow8605
  %43 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %17, i1 true)
  %iszero.i.i.i.i.i.i = icmp ne i64 %17, 0
  %cmp2.i.i.i.i.i = icmp eq i64 %43, %sh_prom.i.i.i.i
  %cmp.i.i.i.i.i = select i1 %iszero.i.i.i.i.i.i, i1 %cmp2.i.i.i.i.i, i1 false
  %44 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.i.i.i.i.i)
  %45 = extractvalue { i1, i64 } %44, 0
  %46 = extractvalue { i1, i64 } %44, 1
  br i1 %45, label %if.then16.i.i.i, label %if.end22.i.i.i

if.then16.i.i.i:                                  ; preds = %if.end14.i.i.i
  %47 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !76
  %opcode.i.i.i = getelementptr inbounds nuw %"struct.rpc::Header", ptr %47, i64 %idxprom.i.i.i.i, i32 1
  store i32 1660944387, ptr %opcode.i.i.i, align 8, !tbaa !77
  %arrayidx21.i.i.i = getelementptr inbounds nuw %"struct.rpc::Header", ptr %47, i64 %idxprom.i.i.i.i
  store i64 %17, ptr %arrayidx21.i.i.i, align 8, !tbaa !80
  br label %if.end22.i.i.i, !amdgpu.uniform !42

Flow8606:                                         ; preds = %if.end22.i.i.i, %Flow8605
  %48 = phi i32 [ %35, %if.end22.i.i.i ], [ %retval.sroa.8.0.i.i.i, %Flow8605 ]
  %49 = phi i64 [ %17, %if.end22.i.i.i ], [ %retval.sroa.2.0.i.i.i, %Flow8605 ]
  %50 = phi i32 [ %spec.store.select.i.i.i, %if.end22.i.i.i ], [ %retval.sroa.6.0.i.i.i, %Flow8605 ]
  br label %cleanup26.i.i.i, !amdgpu.uniform !42

if.end22.i.i.i:                                   ; preds = %if.then16.i.i.i, %if.end14.i.i.i
  call void @llvm.amdgcn.end.cf.i64(i64 %46)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %Flow8606, !amdgpu.uniform !42

Flow8607:                                         ; preds = %cleanup26.i.i.i, %24
  %51 = phi i32 [ %50, %cleanup26.i.i.i ], [ %13, %24 ]
  %52 = phi i64 [ %49, %cleanup26.i.i.i ], [ %14, %24 ]
  %53 = phi i32 [ %48, %cleanup26.i.i.i ], [ %15, %24 ]
  %54 = phi i32 [ %48, %cleanup26.i.i.i ], [ %retval.sroa.8.0.i.i.i, %24 ]
  %55 = phi i64 [ %49, %cleanup26.i.i.i ], [ %retval.sroa.2.0.i.i.i, %24 ]
  %56 = phi i32 [ %50, %cleanup26.i.i.i ], [ %retval.sroa.6.0.i.i.i, %24 ]
  %57 = phi i1 [ %cmp.i46.not.i.i.i, %cleanup26.i.i.i ], [ true, %24 ]
  br i1 %57, label %for.inc.i.i.i, label %Flow8608, !amdgpu.uniform !42

cleanup26.i.i.i:                                  ; preds = %Flow8606
  br label %Flow8607, !amdgpu.uniform !42

for.inc.i.i.i:                                    ; preds = %Flow8607
  %inc.i.i.i = add i32 %spec.store.select.i.i.i, 1
  br label %Flow8608, !amdgpu.uniform !42

Flow8608:                                         ; preds = %for.inc.i.i.i, %Flow8607
  %58 = phi i32 [ %inc.i.i.i, %for.inc.i.i.i ], [ poison, %Flow8607 ]
  %59 = phi i1 [ false, %for.inc.i.i.i ], [ true, %Flow8607 ]
  br i1 %59, label %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i, label %for.cond.i.i1.i, !amdgpu.uniform !42

_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i: ; preds = %Flow8608
  %.lcssa9363 = phi i32 [ %51, %Flow8608 ]
  %.lcssa9362 = phi i64 [ %52, %Flow8608 ]
  %.lcssa9361 = phi i32 [ %53, %Flow8608 ]
  %60 = addrspacecast ptr addrspace(4) %scevgep7280.lcssa to ptr
  %61 = addrspacecast ptr addrspace(4) %cond to ptr
  %sub.ptr.lhs.cast.i.i.i = ptrtoint ptr %60 to i64
  %sub.ptr.rhs.cast.i.i.i = ptrtoint ptr %61 to i64
  %sub.ptr.sub.i.i.i = sub i64 %sub.ptr.lhs.cast.i.i.i, %sub.ptr.rhs.cast.i.i.i
  fence acquire
  %62 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %mul.i.i.i.i.i.i = shl i32 %.lcssa9363, 6
  %idxprom.i19.i.i.i.i.i = zext i32 %mul.i.i.i.i.i.i to i64
  %arrayidx.i20.i.i.i.i.i = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %62, i64 %idxprom.i19.i.i.i.i.i
  %arrayidx.i22.i.i.i.i.i = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %arrayidx.i20.i.i.i.i.i, i64 %sh_prom.i.i.i.i
  store i64 %sub.ptr.sub.i.i.i, ptr %arrayidx.i22.i.i.i.i.i, align 8, !tbaa !82
  %spec.select.i.i.i.i.i.i.i = tail call i64 @llvm.umin.i64(i64 %sub.ptr.sub.i.i.i, i64 56)
  %63 = and i64 %spec.select.i.i.i.i.i.i.i, 15
  %64 = sub i64 %spec.select.i.i.i.i.i.i.i, %63
  %65 = icmp ne i64 %64, 0
  %66 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %65)
  %67 = extractvalue { i1, i64 } %66, 0
  %68 = extractvalue { i1, i64 } %66, 1
  br i1 %67, label %loop-memcpy-expansion5145.preheader, label %Flow8604

loop-memcpy-expansion5145.preheader:              ; preds = %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i
  %69 = shl nuw nsw i64 %sh_prom.i.i.i.i, 6
  %70 = add nuw nsw i64 %69, 8
  %scevgep7274 = getelementptr i8, ptr %62, i64 %70
  %71 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i, 6
  %scevgep7275 = getelementptr i8, ptr %scevgep7274, i64 %71
  br label %loop-memcpy-expansion5145, !amdgpu.uniform !42

Flow8604:                                         ; preds = %Flow8603, %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i
  call void @llvm.amdgcn.end.cf.i64(i64 %68)
  br label %loop-memcpy-residual-header5148, !amdgpu.uniform !42

loop-memcpy-expansion5145:                        ; preds = %loop-memcpy-expansion5145.preheader, %loop-memcpy-expansion5145
  %phi.broken8613 = phi i64 [ 0, %loop-memcpy-expansion5145.preheader ], [ %76, %loop-memcpy-expansion5145 ]
  %loop-index5146 = phi i64 [ %74, %loop-memcpy-expansion5145 ], [ 0, %loop-memcpy-expansion5145.preheader ]
  %72 = getelementptr i8, ptr addrspace(4) %cond, i64 %loop-index5146
  %73 = load <4 x i32>, ptr addrspace(4) %72, align 1
  %scevgep7276 = getelementptr i8, ptr %scevgep7275, i64 %loop-index5146
  store <4 x i32> %73, ptr %scevgep7276, align 1
  %74 = add i64 %loop-index5146, 16
  %75 = icmp uge i64 %74, %64
  %76 = call i64 @llvm.amdgcn.if.break.i64(i1 %75, i64 %phi.broken8613)
  %77 = call i1 @llvm.amdgcn.loop.i64(i64 %76)
  br i1 %77, label %Flow8603, label %loop-memcpy-expansion5145

Flow8602:                                         ; preds = %Flow8600, %Flow8601
  %78 = phi i64 [ %6424, %Flow8600 ], [ %6419, %Flow8601 ]
  %79 = phi i64 [ %6429, %Flow8600 ], [ %6420, %Flow8601 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %6423)
  br label %post-loop-memcpy-expansion5144, !amdgpu.uniform !42

loop-memcpy-residual5147:                         ; preds = %loop-memcpy-residual5147.preheader, %loop-memcpy-residual5147
  %phi.broken8614 = phi i64 [ 0, %loop-memcpy-residual5147.preheader ], [ %83, %loop-memcpy-residual5147 ]
  %residual-loop-index5149 = phi i64 [ %81, %loop-memcpy-residual5147 ], [ 0, %loop-memcpy-residual5147.preheader ]
  %scevgep7273 = getelementptr i8, ptr addrspace(4) %scevgep7272, i64 %residual-loop-index5149
  %80 = load i8, ptr addrspace(4) %scevgep7273, align 1
  %scevgep7271 = getelementptr i8, ptr %scevgep7270, i64 %residual-loop-index5149
  store i8 %80, ptr %scevgep7271, align 1
  %81 = add i64 %residual-loop-index5149, 1
  %82 = icmp uge i64 %81, %63
  %83 = call i64 @llvm.amdgcn.if.break.i64(i1 %82, i64 %phi.broken8614)
  %84 = call i1 @llvm.amdgcn.loop.i64(i64 %83)
  br i1 %84, label %Flow8600, label %loop-memcpy-residual5147

Flow8600:                                         ; preds = %loop-memcpy-residual5147
  %.lcssa9356 = phi i64 [ %83, %loop-memcpy-residual5147 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa9356)
  br label %Flow8602, !amdgpu.uniform !42

post-loop-memcpy-expansion5144:                   ; preds = %Flow8602
  %85 = addrspacecast ptr addrspace(4) %cond to ptr
  %86 = addrspacecast ptr addrspace(4) %scevgep7280.lcssa to ptr
  %87 = ptrtoint ptr %85 to i64
  %88 = ptrtoint ptr %86 to i64
  %tobool.not.i.i.i.i.i.i = icmp eq i32 %.lcssa9361, 0
  %conv.i.i.i.i.i.i = zext i1 %tobool.not.i.i.i.i.i.i to i32
  fence release
  %89 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %idxprom.i23.i.i.i.i.i = zext i32 %.lcssa9363 to i64
  %arrayidx.i24.i.i.i.i.i = getelementptr inbounds nuw i32, ptr %89, i64 %idxprom.i23.i.i.i.i.i
  store atomic i32 %conv.i.i.i.i.i.i, ptr %arrayidx.i24.i.i.i.i.i monotonic, align 4
  %90 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !84
  %arrayidx.i.i5.i.i = getelementptr inbounds nuw %"struct.rpc::Header", ptr %90, i64 %idxprom.i23.i.i.i.i.i, !amdgpu.uniform !42
  %91 = load i64, ptr %arrayidx.i.i5.i.i, align 8, !tbaa !80
  %scevgep7256 = getelementptr i8, ptr addrspace(4) %cond, i64 56
  %92 = add i64 %88, -56
  %93 = sub i64 %92, %87
  br label %while.cond.i.i.i.i, !amdgpu.uniform !42

while.cond.i.i.i.i:                               ; preds = %post-loop-memcpy-expansion5144, %Flow8599
  %phi.broken8617 = phi i64 [ 0, %post-loop-memcpy-expansion5144 ], [ %110, %Flow8599 ]
  %lsr.iv7262 = phi i64 [ %108, %Flow8599 ], [ %93, %post-loop-memcpy-expansion5144 ]
  %lsr.iv7257 = phi ptr addrspace(4) [ %107, %Flow8599 ], [ %scevgep7256, %post-loop-memcpy-expansion5144 ]
  %port.sroa.37.0.i.i = phi i32 [ %106, %Flow8599 ], [ %conv.i.i.i.i.i.i, %post-loop-memcpy-expansion5144 ]
  %idx.0.i.i.i.i = phi i64 [ %105, %Flow8599 ], [ 56, %post-loop-memcpy-expansion5144 ]
  %cmp.i.i.i.i4 = icmp ugt i64 %sub.ptr.sub.i.i.i, %idx.0.i.i.i.i
  %94 = tail call i64 @llvm.amdgcn.ballot.i64(i1 %cmp.i.i.i.i4)
  %and.i.i.i.i6.i.i = and i64 %94, %91
  %tobool.not.i.i.i.i = icmp ne i64 %and.i.i.i.i6.i.i, 0
  %95 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i22.i.i.i = getelementptr inbounds nuw i32, ptr %95, i64 %idxprom.i23.i.i.i.i.i, !amdgpu.uniform !42
  %96 = load atomic i32, ptr %arrayidx.i22.i.i.i monotonic, align 4
  %97 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %tobool.not.i.i.i.i)
  %98 = extractvalue { i1, i64 } %97, 0
  %99 = extractvalue { i1, i64 } %97, 1
  br i1 %98, label %cond.false.i8.i.i.i.i, label %Flow8599

cond.false.i8.i.i.i.i:                            ; preds = %while.cond.i.i.i.i
  %100 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %96)
  %101 = icmp ne i32 %100, %port.sroa.37.0.i.i
  br i1 %101, label %while.body.i.i21.i.i.i.i.preheader, label %Flow8598, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i.preheader:               ; preds = %cond.false.i8.i.i.i.i
  br label %while.body.i.i21.i.i.i.i, !amdgpu.uniform !42

Flow8598:                                         ; preds = %Flow8597, %cond.false.i8.i.i.i.i
  br label %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i:                         ; preds = %while.body.i.i21.i.i.i.i.preheader, %while.body.i.i21.i.i.i.i
  tail call void @llvm.amdgcn.s.sleep(i32 2)
  %102 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i22.i.i.i.i = getelementptr inbounds nuw i32, ptr %102, i64 %idxprom.i23.i.i.i.i.i, !amdgpu.uniform !42
  %103 = load atomic i32, ptr %arrayidx.i.i.i22.i.i.i.i monotonic, align 4
  %104 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %103)
  %cmp.i.not.i.i23.i.i.i.i = icmp eq i32 %104, %port.sroa.37.0.i.i
  br i1 %cmp.i.not.i.i23.i.i.i.i, label %Flow8597, label %while.body.i.i21.i.i.i.i, !amdgpu.uniform !42

Flow8597:                                         ; preds = %while.body.i.i21.i.i.i.i
  br label %Flow8598, !amdgpu.uniform !42

Flow8599:                                         ; preds = %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i, %while.cond.i.i.i.i
  %105 = phi i64 [ %add.i.i.i.i, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i ], [ poison, %while.cond.i.i.i.i ]
  %106 = phi i32 [ %conv.i.i26.i.i.i.i, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i ], [ poison, %while.cond.i.i.i.i ]
  %107 = phi ptr addrspace(4) [ %scevgep7258, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i ], [ poison, %while.cond.i.i.i.i ]
  %108 = phi i64 [ %lsr.iv.next7263, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i ], [ poison, %while.cond.i.i.i.i ]
  %109 = phi i1 [ false, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i ], [ true, %while.cond.i.i.i.i ]
  call void @llvm.amdgcn.end.cf.i64(i64 %99)
  %110 = call i64 @llvm.amdgcn.if.break.i64(i1 %109, i64 %phi.broken8617)
  %111 = call i1 @llvm.amdgcn.loop.i64(i64 %110)
  br i1 %111, label %cond.false.i.i.i, label %while.cond.i.i.i.i

_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i: ; preds = %Flow8598
  fence acquire
  %112 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.i.i.i.i4)
  %113 = extractvalue { i1, i64 } %112, 0
  %114 = extractvalue { i1, i64 } %112, 1
  br i1 %113, label %if.then.i.i.i.i.i.i.i, label %Flow8596

if.then.i.i.i.i.i.i.i:                            ; preds = %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i
  %sub.i.i.i.i.i.i.i = sub nuw i64 %sub.ptr.sub.i.i.i, %idx.0.i.i.i.i
  %spec.select.i.i.i36.i.i.i.i = tail call i64 @llvm.umin.i64(i64 %sub.i.i.i.i.i.i.i, i64 64)
  %115 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %116 = and i64 %spec.select.i.i.i36.i.i.i.i, 15
  %117 = sub i64 %spec.select.i.i.i36.i.i.i.i, %116
  %118 = icmp ne i64 %117, 0
  %119 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %118)
  %120 = extractvalue { i1, i64 } %119, 0
  %121 = extractvalue { i1, i64 } %119, 1
  br i1 %120, label %loop-memcpy-expansion5163.preheader, label %Flow8595

loop-memcpy-expansion5163.preheader:              ; preds = %if.then.i.i.i.i.i.i.i
  %arrayidx.i20.i35.i.i.i.i = getelementptr %"struct.rpc::Buffer", ptr %115, i64 %idxprom.i19.i.i.i.i.i
  %arrayidx.i22.i38.i.i.i.i = getelementptr %"struct.rpc::Buffer", ptr %arrayidx.i20.i35.i.i.i.i, i64 %sh_prom.i.i.i.i
  br label %loop-memcpy-expansion5163, !amdgpu.uniform !42

Flow8595:                                         ; preds = %Flow8594, %if.then.i.i.i.i.i.i.i
  call void @llvm.amdgcn.end.cf.i64(i64 %121)
  br label %loop-memcpy-residual-header5166, !amdgpu.uniform !42

loop-memcpy-expansion5163:                        ; preds = %loop-memcpy-expansion5163.preheader, %loop-memcpy-expansion5163
  %phi.broken8615 = phi i64 [ 0, %loop-memcpy-expansion5163.preheader ], [ %126, %loop-memcpy-expansion5163 ]
  %loop-index5164 = phi i64 [ %124, %loop-memcpy-expansion5163 ], [ 0, %loop-memcpy-expansion5163.preheader ]
  %scevgep7259 = getelementptr i8, ptr addrspace(4) %lsr.iv7257, i64 %loop-index5164
  %122 = load <4 x i32>, ptr addrspace(4) %scevgep7259, align 1
  %123 = getelementptr i8, ptr %arrayidx.i22.i38.i.i.i.i, i64 %loop-index5164
  store <4 x i32> %122, ptr %123, align 1
  %124 = add i64 %loop-index5164, 16
  %125 = icmp uge i64 %124, %117
  %126 = call i64 @llvm.amdgcn.if.break.i64(i1 %125, i64 %phi.broken8615)
  %127 = call i1 @llvm.amdgcn.loop.i64(i64 %126)
  br i1 %127, label %Flow8594, label %loop-memcpy-expansion5163

Flow8593:                                         ; preds = %Flow8592, %loop-memcpy-residual-header5166
  call void @llvm.amdgcn.end.cf.i64(i64 %6455)
  br label %Flow8596, !amdgpu.uniform !42

loop-memcpy-residual5165:                         ; preds = %loop-memcpy-residual5165.preheader, %loop-memcpy-residual5165
  %phi.broken8616 = phi i64 [ 0, %loop-memcpy-residual5165.preheader ], [ %131, %loop-memcpy-residual5165 ]
  %residual-loop-index5167 = phi i64 [ %129, %loop-memcpy-residual5165 ], [ 0, %loop-memcpy-residual5165.preheader ]
  %scevgep7268 = getelementptr i8, ptr addrspace(4) %scevgep7267, i64 %residual-loop-index5167
  %128 = load i8, ptr addrspace(4) %scevgep7268, align 1
  %scevgep7266 = getelementptr i8, ptr %scevgep7265, i64 %residual-loop-index5167
  store i8 %128, ptr %scevgep7266, align 1
  %129 = add i64 %residual-loop-index5167, 1
  %130 = icmp uge i64 %129, %116
  %131 = call i64 @llvm.amdgcn.if.break.i64(i1 %130, i64 %phi.broken8616)
  %132 = call i1 @llvm.amdgcn.loop.i64(i64 %131)
  br i1 %132, label %Flow8592, label %loop-memcpy-residual5165

Flow8592:                                         ; preds = %loop-memcpy-residual5165
  %.lcssa9353 = phi i64 [ %131, %loop-memcpy-residual5165 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa9353)
  br label %Flow8593, !amdgpu.uniform !42

_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i: ; preds = %Flow8596
  %conv.i.i26.i.i.i.i = xor i32 %port.sroa.37.0.i.i, 1
  fence release
  %133 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %arrayidx.i24.i29.i.i.i.i = getelementptr inbounds nuw i32, ptr %133, i64 %idxprom.i23.i.i.i.i.i
  store atomic i32 %conv.i.i26.i.i.i.i, ptr %arrayidx.i24.i29.i.i.i.i monotonic, align 4
  %add.i.i.i.i = add i64 %idx.0.i.i.i.i, 64
  %scevgep7258 = getelementptr i8, ptr addrspace(4) %lsr.iv7257, i64 64
  %lsr.iv.next7263 = add i64 %lsr.iv7262, -64
  br label %Flow8599, !amdgpu.uniform !42

cond.false.i.i.i:                                 ; preds = %Flow8599
  %.lcssa9355 = phi i64 [ %110, %Flow8599 ]
  %port.sroa.37.0.i.i.lcssa = phi i32 [ %port.sroa.37.0.i.i, %Flow8599 ]
  %.lcssa9354 = phi i32 [ %96, %Flow8599 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa9355)
  %134 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %.lcssa9354)
  %135 = icmp ne i32 %134, %port.sroa.37.0.i.i.lcssa
  %136 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %135)
  %137 = extractvalue { i1, i64 } %136, 0
  %138 = extractvalue { i1, i64 } %136, 1
  br i1 %137, label %while.body.i.i.i.i.preheader, label %Flow8591

while.body.i.i.i.i.preheader:                     ; preds = %cond.false.i.i.i
  br label %while.body.i.i.i.i, !amdgpu.uniform !42

Flow8591:                                         ; preds = %Flow8590, %cond.false.i.i.i
  call void @llvm.amdgcn.end.cf.i64(i64 %138)
  br label %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit, !amdgpu.uniform !42

while.body.i.i.i.i:                               ; preds = %while.body.i.i.i.i.preheader, %while.body.i.i.i.i
  %phi.broken8618 = phi i64 [ 0, %while.body.i.i.i.i.preheader ], [ %142, %while.body.i.i.i.i ]
  tail call void @llvm.amdgcn.s.sleep(i32 2)
  %139 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i8.i.i = getelementptr inbounds nuw i32, ptr %139, i64 %idxprom.i23.i.i.i.i.i, !amdgpu.uniform !42
  %140 = load atomic i32, ptr %arrayidx.i.i.i8.i.i monotonic, align 4
  %141 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %140)
  %cmp.i.not.i.i.i.i = icmp eq i32 %141, %port.sroa.37.0.i.i.lcssa
  %142 = call i64 @llvm.amdgcn.if.break.i64(i1 %cmp.i.not.i.i.i.i, i64 %phi.broken8618)
  %143 = call i1 @llvm.amdgcn.loop.i64(i64 %142)
  br i1 %143, label %Flow8590, label %while.body.i.i.i.i

Flow8590:                                         ; preds = %while.body.i.i.i.i
  %.lcssa9351 = phi i64 [ %142, %while.body.i.i.i.i ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa9351)
  br label %Flow8591, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit: ; preds = %Flow8591
  fence acquire
  tail call void @llvm.amdgcn.wave.barrier()
  fence syncscope("agent") release
  %144 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %.lcssa9362, i1 true)
  %iszero.i.i.i.i.i13.i.i = icmp ne i64 %.lcssa9362, 0
  %cmp2.i.i.i.i14.i.i = icmp eq i64 %144, %sh_prom.i.i.i.i
  %cmp.i.i.i.i15.i.i = select i1 %iszero.i.i.i.i.i13.i.i, i1 %cmp2.i.i.i.i14.i.i, i1 false
  %145 = and i32 %.lcssa9363, 31
  %conv4.i.i.i19.i.i = zext i1 %cmp.i.i.i.i15.i.i to i32
  %shl.i.i.i20.i.i = shl nuw i32 %conv4.i.i.i19.i.i, %145
  %xor.i.i.i21.i.i = xor i32 %shl.i.i.i20.i.i, -1
  %146 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop5472, !amdgpu.uniform !42

147:                                              ; preds = %ComputeEnd5473
  %div8.i.i.i16.i.i = lshr i32 %.lcssa9363, 5
  %idxprom.i.i.i17.i.i = zext nneg i32 %div8.i.i.i16.i.i to i64
  %148 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i17.i.i
  %sunkaddr7364 = getelementptr inbounds i8, ptr addrspace(1) %148, i64 40
  %149 = atomicrmw and ptr addrspace(1) %sunkaddr7364, i32 %.lcssa9350 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %150, !amdgpu.uniform !42

150:                                              ; preds = %147, %ComputeEnd5473
  call void @llvm.amdgcn.end.cf.i64(i64 %6742)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %for.cond.i.i1.i13, !amdgpu.uniform !42

for.cond.i.i1.i13:                                ; preds = %150, %Flow8589
  %151 = phi i32 [ %189, %Flow8589 ], [ poison, %150 ]
  %152 = phi i64 [ %190, %Flow8589 ], [ poison, %150 ]
  %153 = phi i32 [ %191, %Flow8589 ], [ poison, %150 ]
  %retval.sroa.6.0.i.i.i14 = phi i32 [ %194, %Flow8589 ], [ undef, %150 ]
  %retval.sroa.2.0.i.i.i16 = phi i64 [ %193, %Flow8589 ], [ undef, %150 ]
  %retval.sroa.8.0.i.i.i17 = phi i32 [ %192, %Flow8589 ], [ undef, %150 ]
  %index.0.i.i.i18 = phi i32 [ %196, %Flow8589 ], [ 0, %150 ]
  %154 = load i32, ptr addrspace(1) @__llvm_rpc_client, align 8, !tbaa !68
  %cmp.not.i.i.i19 = icmp ult i32 %index.0.i.i.i18, %154
  %spec.store.select.i.i.i20 = select i1 %cmp.not.i.i.i19, i32 %index.0.i.i.i18, i32 0
  %155 = tail call noundef i64 @llvm.amdgcn.ballot.i64(i1 true)
  %156 = and i32 %spec.store.select.i.i.i20, 31
  %157 = lshr i64 %155, %sh_prom.i.i.i.i
  %158 = trunc i64 %157 to i32
  %conv4.i.i.i.i.i24 = and i32 %158, 1
  %shl.i.i.i.i.i25 = shl nuw i32 %conv4.i.i.i.i.i24, %156
  %159 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop5476, !amdgpu.uniform !42

160:                                              ; preds = %ComputeEnd5477
  %sunkaddr7365 = getelementptr inbounds i8, ptr addrspace(1) %6752, i64 40
  %161 = atomicrmw or ptr addrspace(1) %sunkaddr7365, i32 %.lcssa9344 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %162, !amdgpu.uniform !42

162:                                              ; preds = %160, %ComputeEnd5477
  %163 = phi i32 [ %161, %160 ], [ poison, %ComputeEnd5477 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %6761)
  %164 = call i32 @llvm.amdgcn.readfirstlane.i32(i32 %163)
  %165 = or i32 %164, %.lcssa9345
  %shl5.i.i.i.i.i26 = shl nuw i32 1, %156
  %and.i.i.i.i.i27 = and i32 %shl5.i.i.i.i.i26, %165
  %tobool3.i.i.i.i28 = icmp ne i32 %and.i.i.i.i.i27, 0
  %166 = tail call i64 @llvm.amdgcn.ballot.i64(i1 %tobool3.i.i.i.i28)
  %and.i.i.i.i.i.i29 = and i64 %166, %155
  %cmp.i.not.i.i.i30 = icmp ne i64 %155, %and.i.i.i.i.i.i29
  br i1 %cmp.i.not.i.i.i30, label %if.then.i.i.i.i31, label %Flow8588, !amdgpu.uniform !42

if.then.i.i.i.i31:                                ; preds = %162
  fence syncscope("agent") acquire
  %167 = load <2 x i64>, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !75
  %168 = extractelement <2 x i64> %167, i32 0
  %169 = inttoptr i64 %168 to ptr
  %170 = extractelement <2 x i64> %167, i32 1
  %171 = inttoptr i64 %170 to ptr
  %idxprom.i.i.i.i32 = zext i32 %spec.store.select.i.i.i20 to i64
  %arrayidx.i.i.i.i33 = getelementptr inbounds nuw i32, ptr %169, i64 %idxprom.i.i.i.i32, !amdgpu.uniform !42
  %172 = load atomic i32, ptr %arrayidx.i.i.i.i33 monotonic, align 4
  %173 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %172)
  %arrayidx.i45.i.i.i34 = getelementptr inbounds nuw i32, ptr %171, i64 %idxprom.i.i.i.i32, !amdgpu.uniform !42
  %174 = load atomic i32, ptr %arrayidx.i45.i.i.i34 monotonic, align 4
  %175 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %174)
  %cmp.i46.not.i.i.i35 = icmp ne i32 %173, %175
  br i1 %cmp.i46.not.i.i.i35, label %if.then12.i.i.i36, label %Flow8586, !amdgpu.uniform !42

if.then12.i.i.i36:                                ; preds = %if.then.i.i.i.i31
  fence syncscope("agent") release
  %176 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %155, i1 true)
  %iszero.i.i.i.i.i.i.i37 = icmp ne i64 %155, 0
  %cmp2.i.i.i.i.i.i38 = icmp eq i64 %176, %sh_prom.i.i.i.i
  %cmp.i.i.i.i.i.i39 = select i1 %iszero.i.i.i.i.i.i.i37, i1 %cmp2.i.i.i.i.i.i38, i1 false
  %conv4.i.i51.i.i.i40 = zext i1 %cmp.i.i.i.i.i.i39 to i32
  %shl.i.i52.i.i.i41 = shl nuw i32 %conv4.i.i51.i.i.i40, %156
  %xor.i.i.i.i.i42 = xor i32 %shl.i.i52.i.i.i41, -1
  %177 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop5481, !amdgpu.uniform !42

178:                                              ; preds = %ComputeEnd5482
  %sunkaddr7366 = getelementptr inbounds i8, ptr addrspace(1) %6752, i64 40
  %179 = atomicrmw and ptr addrspace(1) %sunkaddr7366, i32 %.lcssa9346 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %180, !amdgpu.uniform !42

180:                                              ; preds = %178, %ComputeEnd5482
  call void @llvm.amdgcn.end.cf.i64(i64 %6779)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %Flow8586, !amdgpu.uniform !42

if.end14.i.i.i101:                                ; preds = %Flow8586
  %181 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %155, i1 true)
  %iszero.i.i.i.i.i.i102 = icmp ne i64 %155, 0
  %cmp2.i.i.i.i.i103 = icmp eq i64 %181, %sh_prom.i.i.i.i
  %cmp.i.i.i.i.i104 = select i1 %iszero.i.i.i.i.i.i102, i1 %cmp2.i.i.i.i.i103, i1 false
  %182 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.i.i.i.i.i104)
  %183 = extractvalue { i1, i64 } %182, 0
  %184 = extractvalue { i1, i64 } %182, 1
  br i1 %183, label %if.then16.i.i.i106, label %if.end22.i.i.i105

if.then16.i.i.i106:                               ; preds = %if.end14.i.i.i101
  %185 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !76
  %opcode.i.i.i107 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %185, i64 %idxprom.i.i.i.i32, i32 1
  store i32 1660944387, ptr %opcode.i.i.i107, align 8, !tbaa !77
  %arrayidx21.i.i.i108 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %185, i64 %idxprom.i.i.i.i32
  store i64 %155, ptr %arrayidx21.i.i.i108, align 8, !tbaa !80
  br label %if.end22.i.i.i105, !amdgpu.uniform !42

Flow8587:                                         ; preds = %if.end22.i.i.i105, %Flow8586
  %186 = phi i32 [ %173, %if.end22.i.i.i105 ], [ %retval.sroa.8.0.i.i.i17, %Flow8586 ]
  %187 = phi i64 [ %155, %if.end22.i.i.i105 ], [ %retval.sroa.2.0.i.i.i16, %Flow8586 ]
  %188 = phi i32 [ %spec.store.select.i.i.i20, %if.end22.i.i.i105 ], [ %retval.sroa.6.0.i.i.i14, %Flow8586 ]
  br label %cleanup26.i.i.i43, !amdgpu.uniform !42

if.end22.i.i.i105:                                ; preds = %if.then16.i.i.i106, %if.end14.i.i.i101
  call void @llvm.amdgcn.end.cf.i64(i64 %184)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %Flow8587, !amdgpu.uniform !42

Flow8588:                                         ; preds = %cleanup26.i.i.i43, %162
  %189 = phi i32 [ %188, %cleanup26.i.i.i43 ], [ %151, %162 ]
  %190 = phi i64 [ %187, %cleanup26.i.i.i43 ], [ %152, %162 ]
  %191 = phi i32 [ %186, %cleanup26.i.i.i43 ], [ %153, %162 ]
  %192 = phi i32 [ %186, %cleanup26.i.i.i43 ], [ %retval.sroa.8.0.i.i.i17, %162 ]
  %193 = phi i64 [ %187, %cleanup26.i.i.i43 ], [ %retval.sroa.2.0.i.i.i16, %162 ]
  %194 = phi i32 [ %188, %cleanup26.i.i.i43 ], [ %retval.sroa.6.0.i.i.i14, %162 ]
  %195 = phi i1 [ %cmp.i46.not.i.i.i35, %cleanup26.i.i.i43 ], [ true, %162 ]
  br i1 %195, label %for.inc.i.i.i48, label %Flow8589, !amdgpu.uniform !42

cleanup26.i.i.i43:                                ; preds = %Flow8587
  br label %Flow8588, !amdgpu.uniform !42

for.inc.i.i.i48:                                  ; preds = %Flow8588
  %inc.i.i.i53 = add i32 %spec.store.select.i.i.i20, 1
  br label %Flow8589, !amdgpu.uniform !42

Flow8589:                                         ; preds = %for.inc.i.i.i48, %Flow8588
  %196 = phi i32 [ %inc.i.i.i53, %for.inc.i.i.i48 ], [ poison, %Flow8588 ]
  %197 = phi i1 [ false, %for.inc.i.i.i48 ], [ true, %Flow8588 ]
  br i1 %197, label %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i54, label %for.cond.i.i1.i13, !amdgpu.uniform !42

_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i54: ; preds = %Flow8589
  %.lcssa9349 = phi i32 [ %189, %Flow8589 ]
  %.lcssa9348 = phi i64 [ %190, %Flow8589 ]
  %.lcssa9347 = phi i32 [ %191, %Flow8589 ]
  fence acquire
  %198 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %mul.i.i.i.i.i.i55 = shl i32 %.lcssa9349, 6
  %idxprom.i19.i.i.i.i.i56 = zext i32 %mul.i.i.i.i.i.i55 to i64
  %arrayidx.i20.i.i.i.i.i57 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %198, i64 %idxprom.i19.i.i.i.i.i56
  %arrayidx.i22.i.i.i.i.i58 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %arrayidx.i20.i.i.i.i.i57, i64 %sh_prom.i.i.i.i
  store i64 sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.12.30 to ptr), i64 13) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.12.30 to ptr) to i64)), ptr %arrayidx.i22.i.i.i.i.i58, align 8, !tbaa !82
  %spec.select.i.i.i.i.i.i.i59 = tail call i64 @llvm.umin.i64(i64 sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.12.30 to ptr), i64 13) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.12.30 to ptr) to i64)), i64 56)
  %199 = and i64 %spec.select.i.i.i.i.i.i.i59, 15
  %200 = sub i64 %spec.select.i.i.i.i.i.i.i59, %199
  %201 = icmp ne i64 %200, 0
  br i1 %201, label %loop-memcpy-expansion5169.preheader, label %Flow8585, !amdgpu.uniform !42

loop-memcpy-expansion5169.preheader:              ; preds = %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i54
  %202 = add nuw nsw i64 %78, 8
  %scevgep7253 = getelementptr i8, ptr %198, i64 %202
  %203 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i56, 6
  %scevgep7254 = getelementptr i8, ptr %scevgep7253, i64 %203
  br label %loop-memcpy-expansion5169, !amdgpu.uniform !42

Flow8585:                                         ; preds = %Flow8584, %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i54
  br label %loop-memcpy-residual-header5172, !amdgpu.uniform !42

loop-memcpy-expansion5169:                        ; preds = %loop-memcpy-expansion5169.preheader, %loop-memcpy-expansion5169
  %loop-index5170 = phi i64 [ %206, %loop-memcpy-expansion5169 ], [ 0, %loop-memcpy-expansion5169.preheader ]
  %204 = getelementptr i8, ptr addrspace(4) @.str.12.30, i64 %loop-index5170, !amdgpu.uniform !42
  %205 = load <4 x i32>, ptr addrspace(4) %204, align 1
  %scevgep7255 = getelementptr i8, ptr %scevgep7254, i64 %loop-index5170
  store <4 x i32> %205, ptr %scevgep7255, align 1
  %206 = add i64 %loop-index5170, 16
  %207 = icmp uge i64 %206, %200
  br i1 %207, label %Flow8584, label %loop-memcpy-expansion5169, !amdgpu.uniform !42

Flow8583:                                         ; preds = %Flow8581, %Flow8582
  %208 = phi i64 [ %6462, %Flow8581 ], [ %6459, %Flow8582 ]
  br label %post-loop-memcpy-expansion5168, !amdgpu.uniform !42

loop-memcpy-residual5171:                         ; preds = %loop-memcpy-residual5171.preheader, %loop-memcpy-residual5171
  %residual-loop-index5173 = phi i64 [ %210, %loop-memcpy-residual5171 ], [ 0, %loop-memcpy-residual5171.preheader ]
  %scevgep7252 = getelementptr i8, ptr addrspace(4) @.str.12.30, i64 %residual-loop-index5173, !amdgpu.uniform !42
  %209 = load i8, ptr addrspace(4) %scevgep7252, align 1
  %scevgep7251 = getelementptr i8, ptr %scevgep7250, i64 %residual-loop-index5173
  store i8 %209, ptr %scevgep7251, align 1
  %210 = add i64 %residual-loop-index5173, 1
  %211 = icmp uge i64 %210, %199
  br i1 %211, label %Flow8581, label %loop-memcpy-residual5171, !amdgpu.uniform !42

Flow8581:                                         ; preds = %loop-memcpy-residual5171
  br label %Flow8583, !amdgpu.uniform !42

post-loop-memcpy-expansion5168:                   ; preds = %Flow8583
  %tobool.not.i.i.i.i.i.i61 = icmp eq i32 %.lcssa9347, 0
  %conv.i.i.i.i.i.i62 = zext i1 %tobool.not.i.i.i.i.i.i61 to i32
  fence release
  %212 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %idxprom.i23.i.i.i.i.i63 = zext i32 %.lcssa9349 to i64
  %arrayidx.i24.i.i.i.i.i64 = getelementptr inbounds nuw i32, ptr %212, i64 %idxprom.i23.i.i.i.i.i63
  store atomic i32 %conv.i.i.i.i.i.i62, ptr %arrayidx.i24.i.i.i.i.i64 monotonic, align 4
  %213 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !84
  %arrayidx.i.i5.i.i65 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %213, i64 %idxprom.i23.i.i.i.i.i63, !amdgpu.uniform !42
  %214 = load i64, ptr %arrayidx.i.i5.i.i65, align 8, !tbaa !80
  br label %while.cond.i.i.i.i66, !amdgpu.uniform !42

while.cond.i.i.i.i66:                             ; preds = %post-loop-memcpy-expansion5168, %Flow8580
  %phi.broken8619 = phi i64 [ 0, %post-loop-memcpy-expansion5168 ], [ %231, %Flow8580 ]
  %lsr.iv7242 = phi i64 [ %229, %Flow8580 ], [ -43, %post-loop-memcpy-expansion5168 ]
  %lsr.iv7237 = phi ptr addrspace(4) [ %228, %Flow8580 ], [ getelementptr (i8, ptr addrspace(4) @.str.12.30, i64 56), %post-loop-memcpy-expansion5168 ]
  %port.sroa.37.0.i.i67 = phi i32 [ %227, %Flow8580 ], [ %conv.i.i.i.i.i.i62, %post-loop-memcpy-expansion5168 ]
  %idx.0.i.i.i.i68 = phi i64 [ %226, %Flow8580 ], [ 56, %post-loop-memcpy-expansion5168 ]
  %cmp.i.i.i.i69 = icmp ult i64 %idx.0.i.i.i.i68, sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.12.30 to ptr), i64 13) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.12.30 to ptr) to i64))
  %215 = tail call i64 @llvm.amdgcn.ballot.i64(i1 %cmp.i.i.i.i69)
  %and.i.i.i.i6.i.i70 = and i64 %215, %214
  %tobool.not.i.i.i.i71 = icmp ne i64 %and.i.i.i.i6.i.i70, 0
  %216 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i22.i.i.i72 = getelementptr inbounds nuw i32, ptr %216, i64 %idxprom.i23.i.i.i.i.i63, !amdgpu.uniform !42
  %217 = load atomic i32, ptr %arrayidx.i22.i.i.i72 monotonic, align 4
  %218 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %tobool.not.i.i.i.i71)
  %219 = extractvalue { i1, i64 } %218, 0
  %220 = extractvalue { i1, i64 } %218, 1
  br i1 %219, label %cond.false.i8.i.i.i.i73, label %Flow8580

cond.false.i8.i.i.i.i73:                          ; preds = %while.cond.i.i.i.i66
  %221 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %217)
  %222 = icmp ne i32 %221, %port.sroa.37.0.i.i67
  br i1 %222, label %while.body.i.i21.i.i.i.i74.preheader, label %Flow8579, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i74.preheader:             ; preds = %cond.false.i8.i.i.i.i73
  br label %while.body.i.i21.i.i.i.i74, !amdgpu.uniform !42

Flow8579:                                         ; preds = %Flow8578, %cond.false.i8.i.i.i.i73
  br label %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i77, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i74:                       ; preds = %while.body.i.i21.i.i.i.i74.preheader, %while.body.i.i21.i.i.i.i74
  tail call void @llvm.amdgcn.s.sleep(i32 2)
  %223 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i22.i.i.i.i75 = getelementptr inbounds nuw i32, ptr %223, i64 %idxprom.i23.i.i.i.i.i63, !amdgpu.uniform !42
  %224 = load atomic i32, ptr %arrayidx.i.i.i22.i.i.i.i75 monotonic, align 4
  %225 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %224)
  %cmp.i.not.i.i23.i.i.i.i76 = icmp eq i32 %225, %port.sroa.37.0.i.i67
  br i1 %cmp.i.not.i.i23.i.i.i.i76, label %Flow8578, label %while.body.i.i21.i.i.i.i74, !amdgpu.uniform !42

Flow8578:                                         ; preds = %while.body.i.i21.i.i.i.i74
  br label %Flow8579, !amdgpu.uniform !42

Flow8580:                                         ; preds = %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i78, %while.cond.i.i.i.i66
  %226 = phi i64 [ %add.i.i.i.i81, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i78 ], [ poison, %while.cond.i.i.i.i66 ]
  %227 = phi i32 [ %conv.i.i26.i.i.i.i79, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i78 ], [ poison, %while.cond.i.i.i.i66 ]
  %228 = phi ptr addrspace(4) [ %scevgep7238, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i78 ], [ poison, %while.cond.i.i.i.i66 ]
  %229 = phi i64 [ %lsr.iv.next7243, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i78 ], [ poison, %while.cond.i.i.i.i66 ]
  %230 = phi i1 [ false, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i78 ], [ true, %while.cond.i.i.i.i66 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %220)
  %231 = call i64 @llvm.amdgcn.if.break.i64(i1 %230, i64 %phi.broken8619)
  %232 = call i1 @llvm.amdgcn.loop.i64(i64 %231)
  br i1 %232, label %cond.false.i.i.i88, label %while.cond.i.i.i.i66

_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i77: ; preds = %Flow8579
  fence acquire
  br i1 %cmp.i.i.i.i69, label %if.then.i.i.i.i.i.i.i82, label %Flow8577, !amdgpu.uniform !42

if.then.i.i.i.i.i.i.i82:                          ; preds = %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i77
  %sub.i.i.i.i.i.i.i84 = sub nuw i64 sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.12.30 to ptr), i64 13) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.12.30 to ptr) to i64)), %idx.0.i.i.i.i68
  %spec.select.i.i.i36.i.i.i.i85 = tail call i64 @llvm.umin.i64(i64 %sub.i.i.i.i.i.i.i84, i64 64)
  %233 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %234 = and i64 %spec.select.i.i.i36.i.i.i.i85, 15
  %235 = sub i64 %spec.select.i.i.i36.i.i.i.i85, %234
  %236 = icmp ne i64 %235, 0
  br i1 %236, label %loop-memcpy-expansion5175.preheader, label %Flow8576, !amdgpu.uniform !42

loop-memcpy-expansion5175.preheader:              ; preds = %if.then.i.i.i.i.i.i.i82
  %arrayidx.i20.i35.i.i.i.i86 = getelementptr %"struct.rpc::Buffer", ptr %233, i64 %idxprom.i19.i.i.i.i.i56
  %arrayidx.i22.i38.i.i.i.i87 = getelementptr %"struct.rpc::Buffer", ptr %arrayidx.i20.i35.i.i.i.i86, i64 %sh_prom.i.i.i.i
  br label %loop-memcpy-expansion5175, !amdgpu.uniform !42

Flow8576:                                         ; preds = %Flow8575, %if.then.i.i.i.i.i.i.i82
  br label %loop-memcpy-residual-header5178, !amdgpu.uniform !42

loop-memcpy-expansion5175:                        ; preds = %loop-memcpy-expansion5175.preheader, %loop-memcpy-expansion5175
  %loop-index5176 = phi i64 [ %239, %loop-memcpy-expansion5175 ], [ 0, %loop-memcpy-expansion5175.preheader ]
  %scevgep7239 = getelementptr i8, ptr addrspace(4) %lsr.iv7237, i64 %loop-index5176, !amdgpu.uniform !42
  %237 = load <4 x i32>, ptr addrspace(4) %scevgep7239, align 1
  %238 = getelementptr i8, ptr %arrayidx.i22.i38.i.i.i.i87, i64 %loop-index5176
  store <4 x i32> %237, ptr %238, align 1
  %239 = add i64 %loop-index5176, 16
  %240 = icmp uge i64 %239, %235
  br i1 %240, label %Flow8575, label %loop-memcpy-expansion5175, !amdgpu.uniform !42

Flow8574:                                         ; preds = %Flow8573, %loop-memcpy-residual-header5178
  br label %Flow8577, !amdgpu.uniform !42

loop-memcpy-residual5177:                         ; preds = %loop-memcpy-residual5177.preheader, %loop-memcpy-residual5177
  %residual-loop-index5179 = phi i64 [ %242, %loop-memcpy-residual5177 ], [ 0, %loop-memcpy-residual5177.preheader ]
  %scevgep7248 = getelementptr i8, ptr addrspace(4) %scevgep7247, i64 %residual-loop-index5179, !amdgpu.uniform !42
  %241 = load i8, ptr addrspace(4) %scevgep7248, align 1
  %scevgep7246 = getelementptr i8, ptr %scevgep7245, i64 %residual-loop-index5179
  store i8 %241, ptr %scevgep7246, align 1
  %242 = add i64 %residual-loop-index5179, 1
  %243 = icmp uge i64 %242, %234
  br i1 %243, label %Flow8573, label %loop-memcpy-residual5177, !amdgpu.uniform !42

Flow8573:                                         ; preds = %loop-memcpy-residual5177
  br label %Flow8574, !amdgpu.uniform !42

_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i78: ; preds = %Flow8577
  %conv.i.i26.i.i.i.i79 = xor i32 %port.sroa.37.0.i.i67, 1
  fence release
  %244 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %arrayidx.i24.i29.i.i.i.i80 = getelementptr inbounds nuw i32, ptr %244, i64 %idxprom.i23.i.i.i.i.i63
  store atomic i32 %conv.i.i26.i.i.i.i79, ptr %arrayidx.i24.i29.i.i.i.i80 monotonic, align 4
  %add.i.i.i.i81 = add i64 %idx.0.i.i.i.i68, 64
  %scevgep7238 = getelementptr i8, ptr addrspace(4) %lsr.iv7237, i64 64
  %lsr.iv.next7243 = add i64 %lsr.iv7242, -64
  br label %Flow8580, !amdgpu.uniform !42

cond.false.i.i.i88:                               ; preds = %Flow8580
  %.lcssa9343 = phi i64 [ %231, %Flow8580 ]
  %port.sroa.37.0.i.i67.lcssa = phi i32 [ %port.sroa.37.0.i.i67, %Flow8580 ]
  %.lcssa9342 = phi i32 [ %217, %Flow8580 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa9343)
  %245 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %.lcssa9342)
  %246 = icmp ne i32 %245, %port.sroa.37.0.i.i67.lcssa
  %247 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %246)
  %248 = extractvalue { i1, i64 } %247, 0
  %249 = extractvalue { i1, i64 } %247, 1
  br i1 %248, label %while.body.i.i.i.i89.preheader, label %Flow8572

while.body.i.i.i.i89.preheader:                   ; preds = %cond.false.i.i.i88
  br label %while.body.i.i.i.i89, !amdgpu.uniform !42

Flow8572:                                         ; preds = %Flow8571, %cond.false.i.i.i88
  call void @llvm.amdgcn.end.cf.i64(i64 %249)
  br label %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit109, !amdgpu.uniform !42

while.body.i.i.i.i89:                             ; preds = %while.body.i.i.i.i89.preheader, %while.body.i.i.i.i89
  %phi.broken8620 = phi i64 [ 0, %while.body.i.i.i.i89.preheader ], [ %253, %while.body.i.i.i.i89 ]
  tail call void @llvm.amdgcn.s.sleep(i32 2)
  %250 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i8.i.i90 = getelementptr inbounds nuw i32, ptr %250, i64 %idxprom.i23.i.i.i.i.i63, !amdgpu.uniform !42
  %251 = load atomic i32, ptr %arrayidx.i.i.i8.i.i90 monotonic, align 4
  %252 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %251)
  %cmp.i.not.i.i.i.i91 = icmp eq i32 %252, %port.sroa.37.0.i.i67.lcssa
  %253 = call i64 @llvm.amdgcn.if.break.i64(i1 %cmp.i.not.i.i.i.i91, i64 %phi.broken8620)
  %254 = call i1 @llvm.amdgcn.loop.i64(i64 %253)
  br i1 %254, label %Flow8571, label %while.body.i.i.i.i89

Flow8571:                                         ; preds = %while.body.i.i.i.i89
  %.lcssa9341 = phi i64 [ %253, %while.body.i.i.i.i89 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa9341)
  br label %Flow8572, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit109: ; preds = %Flow8572
  fence acquire
  tail call void @llvm.amdgcn.wave.barrier()
  fence syncscope("agent") release
  %255 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %.lcssa9348, i1 true)
  %iszero.i.i.i.i.i13.i.i92 = icmp ne i64 %.lcssa9348, 0
  %cmp2.i.i.i.i14.i.i93 = icmp eq i64 %255, %sh_prom.i.i.i.i
  %cmp.i.i.i.i15.i.i94 = select i1 %iszero.i.i.i.i.i13.i.i92, i1 %cmp2.i.i.i.i14.i.i93, i1 false
  %256 = and i32 %.lcssa9349, 31
  %conv4.i.i.i19.i.i98 = zext i1 %cmp.i.i.i.i15.i.i94 to i32
  %shl.i.i.i20.i.i99 = shl nuw i32 %conv4.i.i.i19.i.i98, %256
  %xor.i.i.i21.i.i100 = xor i32 %shl.i.i.i20.i.i99, -1
  %257 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop5485, !amdgpu.uniform !42

258:                                              ; preds = %ComputeEnd5486
  %div8.i.i.i16.i.i95 = lshr i32 %.lcssa9349, 5
  %idxprom.i.i.i17.i.i96 = zext nneg i32 %div8.i.i.i16.i.i95 to i64
  %259 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i17.i.i96
  %sunkaddr7367 = getelementptr inbounds i8, ptr addrspace(1) %259, i64 40
  %260 = atomicrmw and ptr addrspace(1) %sunkaddr7367, i32 %.lcssa9340 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %261, !amdgpu.uniform !42

261:                                              ; preds = %258, %ComputeEnd5486
  call void @llvm.amdgcn.end.cf.i64(i64 %6796)
  tail call void @llvm.amdgcn.wave.barrier()
  %scevgep7233 = getelementptr i8, ptr addrspace(4) %cond6, i64 -1
  br label %for.cond.i.i.i110, !amdgpu.uniform !42

for.cond.i.i.i110:                                ; preds = %261, %for.cond.i.i.i110
  %phi.broken8621 = phi i64 [ 0, %261 ], [ %263, %for.cond.i.i.i110 ]
  %lsr.iv7234 = phi ptr addrspace(4) [ %scevgep7236, %for.cond.i.i.i110 ], [ %scevgep7233, %261 ]
  %scevgep7236 = getelementptr i8, ptr addrspace(4) %lsr.iv7234, i64 1
  %262 = load i8, ptr addrspace(4) %scevgep7236, align 1, !tbaa !57
  %cmp.i.i.i112 = icmp eq i8 %262, 0
  %263 = call i64 @llvm.amdgcn.if.break.i64(i1 %cmp.i.i.i112, i64 %phi.broken8621)
  %264 = call i1 @llvm.amdgcn.loop.i64(i64 %263)
  br i1 %264, label %for.cond.i.i1.i119.preheader, label %for.cond.i.i.i110

for.cond.i.i1.i119.preheader:                     ; preds = %for.cond.i.i.i110
  %scevgep7236.lcssa = phi ptr addrspace(4) [ %scevgep7236, %for.cond.i.i.i110 ]
  %.lcssa9339 = phi i64 [ %263, %for.cond.i.i.i110 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa9339)
  br label %for.cond.i.i1.i119, !amdgpu.uniform !42

for.cond.i.i1.i119:                               ; preds = %for.cond.i.i1.i119.preheader, %Flow8570
  %265 = phi i32 [ %303, %Flow8570 ], [ poison, %for.cond.i.i1.i119.preheader ]
  %266 = phi i64 [ %304, %Flow8570 ], [ poison, %for.cond.i.i1.i119.preheader ]
  %267 = phi i32 [ %305, %Flow8570 ], [ poison, %for.cond.i.i1.i119.preheader ]
  %retval.sroa.6.0.i.i.i120 = phi i32 [ %308, %Flow8570 ], [ undef, %for.cond.i.i1.i119.preheader ]
  %retval.sroa.2.0.i.i.i122 = phi i64 [ %307, %Flow8570 ], [ undef, %for.cond.i.i1.i119.preheader ]
  %retval.sroa.8.0.i.i.i123 = phi i32 [ %306, %Flow8570 ], [ undef, %for.cond.i.i1.i119.preheader ]
  %index.0.i.i.i124 = phi i32 [ %310, %Flow8570 ], [ 0, %for.cond.i.i1.i119.preheader ]
  %268 = load i32, ptr addrspace(1) @__llvm_rpc_client, align 8, !tbaa !68
  %cmp.not.i.i.i125 = icmp ult i32 %index.0.i.i.i124, %268
  %spec.store.select.i.i.i126 = select i1 %cmp.not.i.i.i125, i32 %index.0.i.i.i124, i32 0
  %269 = tail call noundef i64 @llvm.amdgcn.ballot.i64(i1 true)
  %270 = and i32 %spec.store.select.i.i.i126, 31
  %271 = lshr i64 %269, %sh_prom.i.i.i.i
  %272 = trunc i64 %271 to i32
  %conv4.i.i.i.i.i130 = and i32 %272, 1
  %shl.i.i.i.i.i131 = shl nuw i32 %conv4.i.i.i.i.i130, %270
  %273 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop5489, !amdgpu.uniform !42

274:                                              ; preds = %ComputeEnd5490
  %sunkaddr7368 = getelementptr inbounds i8, ptr addrspace(1) %6806, i64 40
  %275 = atomicrmw or ptr addrspace(1) %sunkaddr7368, i32 %.lcssa9333 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %276, !amdgpu.uniform !42

276:                                              ; preds = %274, %ComputeEnd5490
  %277 = phi i32 [ %275, %274 ], [ poison, %ComputeEnd5490 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %6815)
  %278 = call i32 @llvm.amdgcn.readfirstlane.i32(i32 %277)
  %279 = or i32 %278, %.lcssa9334
  %shl5.i.i.i.i.i132 = shl nuw i32 1, %270
  %and.i.i.i.i.i133 = and i32 %shl5.i.i.i.i.i132, %279
  %tobool3.i.i.i.i134 = icmp ne i32 %and.i.i.i.i.i133, 0
  %280 = tail call i64 @llvm.amdgcn.ballot.i64(i1 %tobool3.i.i.i.i134)
  %and.i.i.i.i.i.i135 = and i64 %280, %269
  %cmp.i.not.i.i.i136 = icmp ne i64 %269, %and.i.i.i.i.i.i135
  br i1 %cmp.i.not.i.i.i136, label %if.then.i.i.i.i137, label %Flow8569, !amdgpu.uniform !42

if.then.i.i.i.i137:                               ; preds = %276
  fence syncscope("agent") acquire
  %281 = load <2 x i64>, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !75
  %282 = extractelement <2 x i64> %281, i32 0
  %283 = inttoptr i64 %282 to ptr
  %284 = extractelement <2 x i64> %281, i32 1
  %285 = inttoptr i64 %284 to ptr
  %idxprom.i.i.i.i138 = zext i32 %spec.store.select.i.i.i126 to i64
  %arrayidx.i.i.i.i139 = getelementptr inbounds nuw i32, ptr %283, i64 %idxprom.i.i.i.i138, !amdgpu.uniform !42
  %286 = load atomic i32, ptr %arrayidx.i.i.i.i139 monotonic, align 4
  %287 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %286)
  %arrayidx.i45.i.i.i140 = getelementptr inbounds nuw i32, ptr %285, i64 %idxprom.i.i.i.i138, !amdgpu.uniform !42
  %288 = load atomic i32, ptr %arrayidx.i45.i.i.i140 monotonic, align 4
  %289 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %288)
  %cmp.i46.not.i.i.i141 = icmp ne i32 %287, %289
  br i1 %cmp.i46.not.i.i.i141, label %if.then12.i.i.i142, label %Flow8567, !amdgpu.uniform !42

if.then12.i.i.i142:                               ; preds = %if.then.i.i.i.i137
  fence syncscope("agent") release
  %290 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %269, i1 true)
  %iszero.i.i.i.i.i.i.i143 = icmp ne i64 %269, 0
  %cmp2.i.i.i.i.i.i144 = icmp eq i64 %290, %sh_prom.i.i.i.i
  %cmp.i.i.i.i.i.i145 = select i1 %iszero.i.i.i.i.i.i.i143, i1 %cmp2.i.i.i.i.i.i144, i1 false
  %conv4.i.i51.i.i.i146 = zext i1 %cmp.i.i.i.i.i.i145 to i32
  %shl.i.i52.i.i.i147 = shl nuw i32 %conv4.i.i51.i.i.i146, %270
  %xor.i.i.i.i.i148 = xor i32 %shl.i.i52.i.i.i147, -1
  %291 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop5494, !amdgpu.uniform !42

292:                                              ; preds = %ComputeEnd5495
  %sunkaddr7369 = getelementptr inbounds i8, ptr addrspace(1) %6806, i64 40
  %293 = atomicrmw and ptr addrspace(1) %sunkaddr7369, i32 %.lcssa9335 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %294, !amdgpu.uniform !42

294:                                              ; preds = %292, %ComputeEnd5495
  call void @llvm.amdgcn.end.cf.i64(i64 %6833)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %Flow8567, !amdgpu.uniform !42

if.end14.i.i.i207:                                ; preds = %Flow8567
  %295 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %269, i1 true)
  %iszero.i.i.i.i.i.i208 = icmp ne i64 %269, 0
  %cmp2.i.i.i.i.i209 = icmp eq i64 %295, %sh_prom.i.i.i.i
  %cmp.i.i.i.i.i210 = select i1 %iszero.i.i.i.i.i.i208, i1 %cmp2.i.i.i.i.i209, i1 false
  %296 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.i.i.i.i.i210)
  %297 = extractvalue { i1, i64 } %296, 0
  %298 = extractvalue { i1, i64 } %296, 1
  br i1 %297, label %if.then16.i.i.i212, label %if.end22.i.i.i211

if.then16.i.i.i212:                               ; preds = %if.end14.i.i.i207
  %299 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !76
  %opcode.i.i.i213 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %299, i64 %idxprom.i.i.i.i138, i32 1
  store i32 1660944387, ptr %opcode.i.i.i213, align 8, !tbaa !77
  %arrayidx21.i.i.i214 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %299, i64 %idxprom.i.i.i.i138
  store i64 %269, ptr %arrayidx21.i.i.i214, align 8, !tbaa !80
  br label %if.end22.i.i.i211, !amdgpu.uniform !42

Flow8568:                                         ; preds = %if.end22.i.i.i211, %Flow8567
  %300 = phi i32 [ %287, %if.end22.i.i.i211 ], [ %retval.sroa.8.0.i.i.i123, %Flow8567 ]
  %301 = phi i64 [ %269, %if.end22.i.i.i211 ], [ %retval.sroa.2.0.i.i.i122, %Flow8567 ]
  %302 = phi i32 [ %spec.store.select.i.i.i126, %if.end22.i.i.i211 ], [ %retval.sroa.6.0.i.i.i120, %Flow8567 ]
  br label %cleanup26.i.i.i149, !amdgpu.uniform !42

if.end22.i.i.i211:                                ; preds = %if.then16.i.i.i212, %if.end14.i.i.i207
  call void @llvm.amdgcn.end.cf.i64(i64 %298)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %Flow8568, !amdgpu.uniform !42

Flow8569:                                         ; preds = %cleanup26.i.i.i149, %276
  %303 = phi i32 [ %302, %cleanup26.i.i.i149 ], [ %265, %276 ]
  %304 = phi i64 [ %301, %cleanup26.i.i.i149 ], [ %266, %276 ]
  %305 = phi i32 [ %300, %cleanup26.i.i.i149 ], [ %267, %276 ]
  %306 = phi i32 [ %300, %cleanup26.i.i.i149 ], [ %retval.sroa.8.0.i.i.i123, %276 ]
  %307 = phi i64 [ %301, %cleanup26.i.i.i149 ], [ %retval.sroa.2.0.i.i.i122, %276 ]
  %308 = phi i32 [ %302, %cleanup26.i.i.i149 ], [ %retval.sroa.6.0.i.i.i120, %276 ]
  %309 = phi i1 [ %cmp.i46.not.i.i.i141, %cleanup26.i.i.i149 ], [ true, %276 ]
  br i1 %309, label %for.inc.i.i.i154, label %Flow8570, !amdgpu.uniform !42

cleanup26.i.i.i149:                               ; preds = %Flow8568
  br label %Flow8569, !amdgpu.uniform !42

for.inc.i.i.i154:                                 ; preds = %Flow8569
  %inc.i.i.i159 = add i32 %spec.store.select.i.i.i126, 1
  br label %Flow8570, !amdgpu.uniform !42

Flow8570:                                         ; preds = %for.inc.i.i.i154, %Flow8569
  %310 = phi i32 [ %inc.i.i.i159, %for.inc.i.i.i154 ], [ poison, %Flow8569 ]
  %311 = phi i1 [ false, %for.inc.i.i.i154 ], [ true, %Flow8569 ]
  br i1 %311, label %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i160, label %for.cond.i.i1.i119, !amdgpu.uniform !42

_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i160: ; preds = %Flow8570
  %.lcssa9338 = phi i32 [ %303, %Flow8570 ]
  %.lcssa9337 = phi i64 [ %304, %Flow8570 ]
  %.lcssa9336 = phi i32 [ %305, %Flow8570 ]
  %312 = addrspacecast ptr addrspace(4) %scevgep7236.lcssa to ptr
  %313 = addrspacecast ptr addrspace(4) %cond6 to ptr
  %sub.ptr.lhs.cast.i.i.i115 = ptrtoint ptr %312 to i64
  %sub.ptr.rhs.cast.i.i.i116 = ptrtoint ptr %313 to i64
  %sub.ptr.sub.i.i.i117 = sub i64 %sub.ptr.lhs.cast.i.i.i115, %sub.ptr.rhs.cast.i.i.i116
  fence acquire
  %314 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %mul.i.i.i.i.i.i161 = shl i32 %.lcssa9338, 6
  %idxprom.i19.i.i.i.i.i162 = zext i32 %mul.i.i.i.i.i.i161 to i64
  %arrayidx.i20.i.i.i.i.i163 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %314, i64 %idxprom.i19.i.i.i.i.i162
  %arrayidx.i22.i.i.i.i.i164 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %arrayidx.i20.i.i.i.i.i163, i64 %sh_prom.i.i.i.i
  store i64 %sub.ptr.sub.i.i.i117, ptr %arrayidx.i22.i.i.i.i.i164, align 8, !tbaa !82
  %spec.select.i.i.i.i.i.i.i165 = tail call i64 @llvm.umin.i64(i64 %sub.ptr.sub.i.i.i117, i64 56)
  %315 = and i64 %spec.select.i.i.i.i.i.i.i165, 15
  %316 = sub i64 %spec.select.i.i.i.i.i.i.i165, %315
  %317 = icmp ne i64 %316, 0
  %318 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %317)
  %319 = extractvalue { i1, i64 } %318, 0
  %320 = extractvalue { i1, i64 } %318, 1
  br i1 %319, label %loop-memcpy-expansion5181.preheader, label %Flow8566

loop-memcpy-expansion5181.preheader:              ; preds = %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i160
  %321 = add nuw nsw i64 %78, 8
  %scevgep7230 = getelementptr i8, ptr %314, i64 %321
  %322 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i162, 6
  %scevgep7231 = getelementptr i8, ptr %scevgep7230, i64 %322
  br label %loop-memcpy-expansion5181, !amdgpu.uniform !42

Flow8566:                                         ; preds = %Flow8565, %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i160
  call void @llvm.amdgcn.end.cf.i64(i64 %320)
  br label %loop-memcpy-residual-header5184, !amdgpu.uniform !42

loop-memcpy-expansion5181:                        ; preds = %loop-memcpy-expansion5181.preheader, %loop-memcpy-expansion5181
  %phi.broken8622 = phi i64 [ 0, %loop-memcpy-expansion5181.preheader ], [ %327, %loop-memcpy-expansion5181 ]
  %loop-index5182 = phi i64 [ %325, %loop-memcpy-expansion5181 ], [ 0, %loop-memcpy-expansion5181.preheader ]
  %323 = getelementptr i8, ptr addrspace(4) %cond6, i64 %loop-index5182
  %324 = load <4 x i32>, ptr addrspace(4) %323, align 1
  %scevgep7232 = getelementptr i8, ptr %scevgep7231, i64 %loop-index5182
  store <4 x i32> %324, ptr %scevgep7232, align 1
  %325 = add i64 %loop-index5182, 16
  %326 = icmp uge i64 %325, %316
  %327 = call i64 @llvm.amdgcn.if.break.i64(i1 %326, i64 %phi.broken8622)
  %328 = call i1 @llvm.amdgcn.loop.i64(i64 %327)
  br i1 %328, label %Flow8565, label %loop-memcpy-expansion5181

Flow8564:                                         ; preds = %Flow8562, %Flow8563
  %329 = phi i64 [ %6478, %Flow8562 ], [ %6470, %Flow8563 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %6473)
  br label %post-loop-memcpy-expansion5180, !amdgpu.uniform !42

loop-memcpy-residual5183:                         ; preds = %loop-memcpy-residual5183.preheader, %loop-memcpy-residual5183
  %phi.broken8623 = phi i64 [ 0, %loop-memcpy-residual5183.preheader ], [ %333, %loop-memcpy-residual5183 ]
  %residual-loop-index5185 = phi i64 [ %331, %loop-memcpy-residual5183 ], [ 0, %loop-memcpy-residual5183.preheader ]
  %scevgep7229 = getelementptr i8, ptr addrspace(4) %scevgep7228, i64 %residual-loop-index5185
  %330 = load i8, ptr addrspace(4) %scevgep7229, align 1
  %scevgep7227 = getelementptr i8, ptr %scevgep7226, i64 %residual-loop-index5185
  store i8 %330, ptr %scevgep7227, align 1
  %331 = add i64 %residual-loop-index5185, 1
  %332 = icmp uge i64 %331, %315
  %333 = call i64 @llvm.amdgcn.if.break.i64(i1 %332, i64 %phi.broken8623)
  %334 = call i1 @llvm.amdgcn.loop.i64(i64 %333)
  br i1 %334, label %Flow8562, label %loop-memcpy-residual5183

Flow8562:                                         ; preds = %loop-memcpy-residual5183
  %.lcssa9331 = phi i64 [ %333, %loop-memcpy-residual5183 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa9331)
  br label %Flow8564, !amdgpu.uniform !42

post-loop-memcpy-expansion5180:                   ; preds = %Flow8564
  %335 = addrspacecast ptr addrspace(4) %cond6 to ptr
  %336 = addrspacecast ptr addrspace(4) %scevgep7236.lcssa to ptr
  %337 = ptrtoint ptr %335 to i64
  %338 = ptrtoint ptr %336 to i64
  %tobool.not.i.i.i.i.i.i167 = icmp eq i32 %.lcssa9336, 0
  %conv.i.i.i.i.i.i168 = zext i1 %tobool.not.i.i.i.i.i.i167 to i32
  fence release
  %339 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %idxprom.i23.i.i.i.i.i169 = zext i32 %.lcssa9338 to i64
  %arrayidx.i24.i.i.i.i.i170 = getelementptr inbounds nuw i32, ptr %339, i64 %idxprom.i23.i.i.i.i.i169
  store atomic i32 %conv.i.i.i.i.i.i168, ptr %arrayidx.i24.i.i.i.i.i170 monotonic, align 4
  %340 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !84
  %arrayidx.i.i5.i.i171 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %340, i64 %idxprom.i23.i.i.i.i.i169, !amdgpu.uniform !42
  %341 = load i64, ptr %arrayidx.i.i5.i.i171, align 8, !tbaa !80
  %scevgep7212 = getelementptr i8, ptr addrspace(4) %cond6, i64 56
  %342 = add i64 %338, -56
  %343 = sub i64 %342, %337
  br label %while.cond.i.i.i.i172, !amdgpu.uniform !42

while.cond.i.i.i.i172:                            ; preds = %post-loop-memcpy-expansion5180, %Flow8561
  %phi.broken8626 = phi i64 [ 0, %post-loop-memcpy-expansion5180 ], [ %360, %Flow8561 ]
  %lsr.iv7218 = phi i64 [ %358, %Flow8561 ], [ %343, %post-loop-memcpy-expansion5180 ]
  %lsr.iv7213 = phi ptr addrspace(4) [ %357, %Flow8561 ], [ %scevgep7212, %post-loop-memcpy-expansion5180 ]
  %port.sroa.37.0.i.i173 = phi i32 [ %356, %Flow8561 ], [ %conv.i.i.i.i.i.i168, %post-loop-memcpy-expansion5180 ]
  %idx.0.i.i.i.i174 = phi i64 [ %355, %Flow8561 ], [ 56, %post-loop-memcpy-expansion5180 ]
  %cmp.i.i.i.i175 = icmp ugt i64 %sub.ptr.sub.i.i.i117, %idx.0.i.i.i.i174
  %344 = tail call i64 @llvm.amdgcn.ballot.i64(i1 %cmp.i.i.i.i175)
  %and.i.i.i.i6.i.i176 = and i64 %344, %341
  %tobool.not.i.i.i.i177 = icmp ne i64 %and.i.i.i.i6.i.i176, 0
  %345 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i22.i.i.i178 = getelementptr inbounds nuw i32, ptr %345, i64 %idxprom.i23.i.i.i.i.i169, !amdgpu.uniform !42
  %346 = load atomic i32, ptr %arrayidx.i22.i.i.i178 monotonic, align 4
  %347 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %tobool.not.i.i.i.i177)
  %348 = extractvalue { i1, i64 } %347, 0
  %349 = extractvalue { i1, i64 } %347, 1
  br i1 %348, label %cond.false.i8.i.i.i.i179, label %Flow8561

cond.false.i8.i.i.i.i179:                         ; preds = %while.cond.i.i.i.i172
  %350 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %346)
  %351 = icmp ne i32 %350, %port.sroa.37.0.i.i173
  br i1 %351, label %while.body.i.i21.i.i.i.i180.preheader, label %Flow8560, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i180.preheader:            ; preds = %cond.false.i8.i.i.i.i179
  br label %while.body.i.i21.i.i.i.i180, !amdgpu.uniform !42

Flow8560:                                         ; preds = %Flow8559, %cond.false.i8.i.i.i.i179
  br label %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i183, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i180:                      ; preds = %while.body.i.i21.i.i.i.i180.preheader, %while.body.i.i21.i.i.i.i180
  tail call void @llvm.amdgcn.s.sleep(i32 2)
  %352 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i22.i.i.i.i181 = getelementptr inbounds nuw i32, ptr %352, i64 %idxprom.i23.i.i.i.i.i169, !amdgpu.uniform !42
  %353 = load atomic i32, ptr %arrayidx.i.i.i22.i.i.i.i181 monotonic, align 4
  %354 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %353)
  %cmp.i.not.i.i23.i.i.i.i182 = icmp eq i32 %354, %port.sroa.37.0.i.i173
  br i1 %cmp.i.not.i.i23.i.i.i.i182, label %Flow8559, label %while.body.i.i21.i.i.i.i180, !amdgpu.uniform !42

Flow8559:                                         ; preds = %while.body.i.i21.i.i.i.i180
  br label %Flow8560, !amdgpu.uniform !42

Flow8561:                                         ; preds = %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i184, %while.cond.i.i.i.i172
  %355 = phi i64 [ %add.i.i.i.i187, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i184 ], [ poison, %while.cond.i.i.i.i172 ]
  %356 = phi i32 [ %conv.i.i26.i.i.i.i185, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i184 ], [ poison, %while.cond.i.i.i.i172 ]
  %357 = phi ptr addrspace(4) [ %scevgep7214, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i184 ], [ poison, %while.cond.i.i.i.i172 ]
  %358 = phi i64 [ %lsr.iv.next7219, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i184 ], [ poison, %while.cond.i.i.i.i172 ]
  %359 = phi i1 [ false, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i184 ], [ true, %while.cond.i.i.i.i172 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %349)
  %360 = call i64 @llvm.amdgcn.if.break.i64(i1 %359, i64 %phi.broken8626)
  %361 = call i1 @llvm.amdgcn.loop.i64(i64 %360)
  br i1 %361, label %cond.false.i.i.i194, label %while.cond.i.i.i.i172

_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i183: ; preds = %Flow8560
  fence acquire
  %362 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.i.i.i.i175)
  %363 = extractvalue { i1, i64 } %362, 0
  %364 = extractvalue { i1, i64 } %362, 1
  br i1 %363, label %if.then.i.i.i.i.i.i.i188, label %Flow8558

if.then.i.i.i.i.i.i.i188:                         ; preds = %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i183
  %sub.i.i.i.i.i.i.i190 = sub nuw i64 %sub.ptr.sub.i.i.i117, %idx.0.i.i.i.i174
  %spec.select.i.i.i36.i.i.i.i191 = tail call i64 @llvm.umin.i64(i64 %sub.i.i.i.i.i.i.i190, i64 64)
  %365 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %366 = and i64 %spec.select.i.i.i36.i.i.i.i191, 15
  %367 = sub i64 %spec.select.i.i.i36.i.i.i.i191, %366
  %368 = icmp ne i64 %367, 0
  %369 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %368)
  %370 = extractvalue { i1, i64 } %369, 0
  %371 = extractvalue { i1, i64 } %369, 1
  br i1 %370, label %loop-memcpy-expansion5205.preheader, label %Flow8557

loop-memcpy-expansion5205.preheader:              ; preds = %if.then.i.i.i.i.i.i.i188
  %arrayidx.i20.i35.i.i.i.i192 = getelementptr %"struct.rpc::Buffer", ptr %365, i64 %idxprom.i19.i.i.i.i.i162
  %arrayidx.i22.i38.i.i.i.i193 = getelementptr %"struct.rpc::Buffer", ptr %arrayidx.i20.i35.i.i.i.i192, i64 %sh_prom.i.i.i.i
  br label %loop-memcpy-expansion5205, !amdgpu.uniform !42

Flow8557:                                         ; preds = %Flow8556, %if.then.i.i.i.i.i.i.i188
  call void @llvm.amdgcn.end.cf.i64(i64 %371)
  br label %loop-memcpy-residual-header5208, !amdgpu.uniform !42

loop-memcpy-expansion5205:                        ; preds = %loop-memcpy-expansion5205.preheader, %loop-memcpy-expansion5205
  %phi.broken8624 = phi i64 [ 0, %loop-memcpy-expansion5205.preheader ], [ %376, %loop-memcpy-expansion5205 ]
  %loop-index5206 = phi i64 [ %374, %loop-memcpy-expansion5205 ], [ 0, %loop-memcpy-expansion5205.preheader ]
  %scevgep7215 = getelementptr i8, ptr addrspace(4) %lsr.iv7213, i64 %loop-index5206
  %372 = load <4 x i32>, ptr addrspace(4) %scevgep7215, align 1
  %373 = getelementptr i8, ptr %arrayidx.i22.i38.i.i.i.i193, i64 %loop-index5206
  store <4 x i32> %372, ptr %373, align 1
  %374 = add i64 %loop-index5206, 16
  %375 = icmp uge i64 %374, %367
  %376 = call i64 @llvm.amdgcn.if.break.i64(i1 %375, i64 %phi.broken8624)
  %377 = call i1 @llvm.amdgcn.loop.i64(i64 %376)
  br i1 %377, label %Flow8556, label %loop-memcpy-expansion5205

Flow8555:                                         ; preds = %Flow8554, %loop-memcpy-residual-header5208
  call void @llvm.amdgcn.end.cf.i64(i64 %6512)
  br label %Flow8558, !amdgpu.uniform !42

loop-memcpy-residual5207:                         ; preds = %loop-memcpy-residual5207.preheader, %loop-memcpy-residual5207
  %phi.broken8625 = phi i64 [ 0, %loop-memcpy-residual5207.preheader ], [ %381, %loop-memcpy-residual5207 ]
  %residual-loop-index5209 = phi i64 [ %379, %loop-memcpy-residual5207 ], [ 0, %loop-memcpy-residual5207.preheader ]
  %scevgep7224 = getelementptr i8, ptr addrspace(4) %scevgep7223, i64 %residual-loop-index5209
  %378 = load i8, ptr addrspace(4) %scevgep7224, align 1
  %scevgep7222 = getelementptr i8, ptr %scevgep7221, i64 %residual-loop-index5209
  store i8 %378, ptr %scevgep7222, align 1
  %379 = add i64 %residual-loop-index5209, 1
  %380 = icmp uge i64 %379, %366
  %381 = call i64 @llvm.amdgcn.if.break.i64(i1 %380, i64 %phi.broken8625)
  %382 = call i1 @llvm.amdgcn.loop.i64(i64 %381)
  br i1 %382, label %Flow8554, label %loop-memcpy-residual5207

Flow8554:                                         ; preds = %loop-memcpy-residual5207
  %.lcssa9328 = phi i64 [ %381, %loop-memcpy-residual5207 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa9328)
  br label %Flow8555, !amdgpu.uniform !42

_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i184: ; preds = %Flow8558
  %conv.i.i26.i.i.i.i185 = xor i32 %port.sroa.37.0.i.i173, 1
  fence release
  %383 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %arrayidx.i24.i29.i.i.i.i186 = getelementptr inbounds nuw i32, ptr %383, i64 %idxprom.i23.i.i.i.i.i169
  store atomic i32 %conv.i.i26.i.i.i.i185, ptr %arrayidx.i24.i29.i.i.i.i186 monotonic, align 4
  %add.i.i.i.i187 = add i64 %idx.0.i.i.i.i174, 64
  %scevgep7214 = getelementptr i8, ptr addrspace(4) %lsr.iv7213, i64 64
  %lsr.iv.next7219 = add i64 %lsr.iv7218, -64
  br label %Flow8561, !amdgpu.uniform !42

cond.false.i.i.i194:                              ; preds = %Flow8561
  %.lcssa9330 = phi i64 [ %360, %Flow8561 ]
  %port.sroa.37.0.i.i173.lcssa = phi i32 [ %port.sroa.37.0.i.i173, %Flow8561 ]
  %.lcssa9329 = phi i32 [ %346, %Flow8561 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa9330)
  %384 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %.lcssa9329)
  %385 = icmp ne i32 %384, %port.sroa.37.0.i.i173.lcssa
  %386 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %385)
  %387 = extractvalue { i1, i64 } %386, 0
  %388 = extractvalue { i1, i64 } %386, 1
  br i1 %387, label %while.body.i.i.i.i195.preheader, label %Flow8553

while.body.i.i.i.i195.preheader:                  ; preds = %cond.false.i.i.i194
  br label %while.body.i.i.i.i195, !amdgpu.uniform !42

Flow8553:                                         ; preds = %Flow8552, %cond.false.i.i.i194
  call void @llvm.amdgcn.end.cf.i64(i64 %388)
  br label %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit215, !amdgpu.uniform !42

while.body.i.i.i.i195:                            ; preds = %while.body.i.i.i.i195.preheader, %while.body.i.i.i.i195
  %phi.broken8627 = phi i64 [ 0, %while.body.i.i.i.i195.preheader ], [ %392, %while.body.i.i.i.i195 ]
  tail call void @llvm.amdgcn.s.sleep(i32 2)
  %389 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i8.i.i196 = getelementptr inbounds nuw i32, ptr %389, i64 %idxprom.i23.i.i.i.i.i169, !amdgpu.uniform !42
  %390 = load atomic i32, ptr %arrayidx.i.i.i8.i.i196 monotonic, align 4
  %391 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %390)
  %cmp.i.not.i.i.i.i197 = icmp eq i32 %391, %port.sroa.37.0.i.i173.lcssa
  %392 = call i64 @llvm.amdgcn.if.break.i64(i1 %cmp.i.not.i.i.i.i197, i64 %phi.broken8627)
  %393 = call i1 @llvm.amdgcn.loop.i64(i64 %392)
  br i1 %393, label %Flow8552, label %while.body.i.i.i.i195

Flow8552:                                         ; preds = %while.body.i.i.i.i195
  %.lcssa9326 = phi i64 [ %392, %while.body.i.i.i.i195 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa9326)
  br label %Flow8553, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit215: ; preds = %Flow8553
  fence acquire
  tail call void @llvm.amdgcn.wave.barrier()
  fence syncscope("agent") release
  %394 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %.lcssa9337, i1 true)
  %iszero.i.i.i.i.i13.i.i198 = icmp ne i64 %.lcssa9337, 0
  %cmp2.i.i.i.i14.i.i199 = icmp eq i64 %394, %sh_prom.i.i.i.i
  %cmp.i.i.i.i15.i.i200 = select i1 %iszero.i.i.i.i.i13.i.i198, i1 %cmp2.i.i.i.i14.i.i199, i1 false
  %395 = and i32 %.lcssa9338, 31
  %conv4.i.i.i19.i.i204 = zext i1 %cmp.i.i.i.i15.i.i200 to i32
  %shl.i.i.i20.i.i205 = shl nuw i32 %conv4.i.i.i19.i.i204, %395
  %xor.i.i.i21.i.i206 = xor i32 %shl.i.i.i20.i.i205, -1
  %396 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop5498, !amdgpu.uniform !42

397:                                              ; preds = %ComputeEnd5499
  %div8.i.i.i16.i.i201 = lshr i32 %.lcssa9338, 5
  %idxprom.i.i.i17.i.i202 = zext nneg i32 %div8.i.i.i16.i.i201 to i64
  %398 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i17.i.i202
  %sunkaddr7370 = getelementptr inbounds i8, ptr addrspace(1) %398, i64 40
  %399 = atomicrmw and ptr addrspace(1) %sunkaddr7370, i32 %.lcssa9325 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %400, !amdgpu.uniform !42

400:                                              ; preds = %397, %ComputeEnd5499
  call void @llvm.amdgcn.end.cf.i64(i64 %6850)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %for.cond.i.i1.i224, !amdgpu.uniform !42

for.cond.i.i1.i224:                               ; preds = %400, %Flow8551
  %401 = phi i32 [ %439, %Flow8551 ], [ poison, %400 ]
  %402 = phi i64 [ %440, %Flow8551 ], [ poison, %400 ]
  %403 = phi i32 [ %441, %Flow8551 ], [ poison, %400 ]
  %retval.sroa.6.0.i.i.i225 = phi i32 [ %444, %Flow8551 ], [ undef, %400 ]
  %retval.sroa.2.0.i.i.i227 = phi i64 [ %443, %Flow8551 ], [ undef, %400 ]
  %retval.sroa.8.0.i.i.i228 = phi i32 [ %442, %Flow8551 ], [ undef, %400 ]
  %index.0.i.i.i229 = phi i32 [ %446, %Flow8551 ], [ 0, %400 ]
  %404 = load i32, ptr addrspace(1) @__llvm_rpc_client, align 8, !tbaa !68
  %cmp.not.i.i.i230 = icmp ult i32 %index.0.i.i.i229, %404
  %spec.store.select.i.i.i231 = select i1 %cmp.not.i.i.i230, i32 %index.0.i.i.i229, i32 0
  %405 = tail call noundef i64 @llvm.amdgcn.ballot.i64(i1 true)
  %406 = and i32 %spec.store.select.i.i.i231, 31
  %407 = lshr i64 %405, %sh_prom.i.i.i.i
  %408 = trunc i64 %407 to i32
  %conv4.i.i.i.i.i235 = and i32 %408, 1
  %shl.i.i.i.i.i236 = shl nuw i32 %conv4.i.i.i.i.i235, %406
  %409 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop5502, !amdgpu.uniform !42

410:                                              ; preds = %ComputeEnd5503
  %sunkaddr7371 = getelementptr inbounds i8, ptr addrspace(1) %6860, i64 40
  %411 = atomicrmw or ptr addrspace(1) %sunkaddr7371, i32 %.lcssa9319 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %412, !amdgpu.uniform !42

412:                                              ; preds = %410, %ComputeEnd5503
  %413 = phi i32 [ %411, %410 ], [ poison, %ComputeEnd5503 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %6869)
  %414 = call i32 @llvm.amdgcn.readfirstlane.i32(i32 %413)
  %415 = or i32 %414, %.lcssa9320
  %shl5.i.i.i.i.i237 = shl nuw i32 1, %406
  %and.i.i.i.i.i238 = and i32 %shl5.i.i.i.i.i237, %415
  %tobool3.i.i.i.i239 = icmp ne i32 %and.i.i.i.i.i238, 0
  %416 = tail call i64 @llvm.amdgcn.ballot.i64(i1 %tobool3.i.i.i.i239)
  %and.i.i.i.i.i.i240 = and i64 %416, %405
  %cmp.i.not.i.i.i241 = icmp ne i64 %405, %and.i.i.i.i.i.i240
  br i1 %cmp.i.not.i.i.i241, label %if.then.i.i.i.i242, label %Flow8550, !amdgpu.uniform !42

if.then.i.i.i.i242:                               ; preds = %412
  fence syncscope("agent") acquire
  %417 = load <2 x i64>, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !75
  %418 = extractelement <2 x i64> %417, i32 0
  %419 = inttoptr i64 %418 to ptr
  %420 = extractelement <2 x i64> %417, i32 1
  %421 = inttoptr i64 %420 to ptr
  %idxprom.i.i.i.i243 = zext i32 %spec.store.select.i.i.i231 to i64
  %arrayidx.i.i.i.i244 = getelementptr inbounds nuw i32, ptr %419, i64 %idxprom.i.i.i.i243, !amdgpu.uniform !42
  %422 = load atomic i32, ptr %arrayidx.i.i.i.i244 monotonic, align 4
  %423 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %422)
  %arrayidx.i45.i.i.i245 = getelementptr inbounds nuw i32, ptr %421, i64 %idxprom.i.i.i.i243, !amdgpu.uniform !42
  %424 = load atomic i32, ptr %arrayidx.i45.i.i.i245 monotonic, align 4
  %425 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %424)
  %cmp.i46.not.i.i.i246 = icmp ne i32 %423, %425
  br i1 %cmp.i46.not.i.i.i246, label %if.then12.i.i.i247, label %Flow8548, !amdgpu.uniform !42

if.then12.i.i.i247:                               ; preds = %if.then.i.i.i.i242
  fence syncscope("agent") release
  %426 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %405, i1 true)
  %iszero.i.i.i.i.i.i.i248 = icmp ne i64 %405, 0
  %cmp2.i.i.i.i.i.i249 = icmp eq i64 %426, %sh_prom.i.i.i.i
  %cmp.i.i.i.i.i.i250 = select i1 %iszero.i.i.i.i.i.i.i248, i1 %cmp2.i.i.i.i.i.i249, i1 false
  %conv4.i.i51.i.i.i251 = zext i1 %cmp.i.i.i.i.i.i250 to i32
  %shl.i.i52.i.i.i252 = shl nuw i32 %conv4.i.i51.i.i.i251, %406
  %xor.i.i.i.i.i253 = xor i32 %shl.i.i52.i.i.i252, -1
  %427 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop5507, !amdgpu.uniform !42

428:                                              ; preds = %ComputeEnd5508
  %sunkaddr7372 = getelementptr inbounds i8, ptr addrspace(1) %6860, i64 40
  %429 = atomicrmw and ptr addrspace(1) %sunkaddr7372, i32 %.lcssa9321 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %430, !amdgpu.uniform !42

430:                                              ; preds = %428, %ComputeEnd5508
  call void @llvm.amdgcn.end.cf.i64(i64 %6887)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %Flow8548, !amdgpu.uniform !42

if.end14.i.i.i312:                                ; preds = %Flow8548
  %431 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %405, i1 true)
  %iszero.i.i.i.i.i.i313 = icmp ne i64 %405, 0
  %cmp2.i.i.i.i.i314 = icmp eq i64 %431, %sh_prom.i.i.i.i
  %cmp.i.i.i.i.i315 = select i1 %iszero.i.i.i.i.i.i313, i1 %cmp2.i.i.i.i.i314, i1 false
  %432 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.i.i.i.i.i315)
  %433 = extractvalue { i1, i64 } %432, 0
  %434 = extractvalue { i1, i64 } %432, 1
  br i1 %433, label %if.then16.i.i.i317, label %if.end22.i.i.i316

if.then16.i.i.i317:                               ; preds = %if.end14.i.i.i312
  %435 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !76
  %opcode.i.i.i318 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %435, i64 %idxprom.i.i.i.i243, i32 1
  store i32 1660944387, ptr %opcode.i.i.i318, align 8, !tbaa !77
  %arrayidx21.i.i.i319 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %435, i64 %idxprom.i.i.i.i243
  store i64 %405, ptr %arrayidx21.i.i.i319, align 8, !tbaa !80
  br label %if.end22.i.i.i316, !amdgpu.uniform !42

Flow8549:                                         ; preds = %if.end22.i.i.i316, %Flow8548
  %436 = phi i32 [ %423, %if.end22.i.i.i316 ], [ %retval.sroa.8.0.i.i.i228, %Flow8548 ]
  %437 = phi i64 [ %405, %if.end22.i.i.i316 ], [ %retval.sroa.2.0.i.i.i227, %Flow8548 ]
  %438 = phi i32 [ %spec.store.select.i.i.i231, %if.end22.i.i.i316 ], [ %retval.sroa.6.0.i.i.i225, %Flow8548 ]
  br label %cleanup26.i.i.i254, !amdgpu.uniform !42

if.end22.i.i.i316:                                ; preds = %if.then16.i.i.i317, %if.end14.i.i.i312
  call void @llvm.amdgcn.end.cf.i64(i64 %434)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %Flow8549, !amdgpu.uniform !42

Flow8550:                                         ; preds = %cleanup26.i.i.i254, %412
  %439 = phi i32 [ %438, %cleanup26.i.i.i254 ], [ %401, %412 ]
  %440 = phi i64 [ %437, %cleanup26.i.i.i254 ], [ %402, %412 ]
  %441 = phi i32 [ %436, %cleanup26.i.i.i254 ], [ %403, %412 ]
  %442 = phi i32 [ %436, %cleanup26.i.i.i254 ], [ %retval.sroa.8.0.i.i.i228, %412 ]
  %443 = phi i64 [ %437, %cleanup26.i.i.i254 ], [ %retval.sroa.2.0.i.i.i227, %412 ]
  %444 = phi i32 [ %438, %cleanup26.i.i.i254 ], [ %retval.sroa.6.0.i.i.i225, %412 ]
  %445 = phi i1 [ %cmp.i46.not.i.i.i246, %cleanup26.i.i.i254 ], [ true, %412 ]
  br i1 %445, label %for.inc.i.i.i259, label %Flow8551, !amdgpu.uniform !42

cleanup26.i.i.i254:                               ; preds = %Flow8549
  br label %Flow8550, !amdgpu.uniform !42

for.inc.i.i.i259:                                 ; preds = %Flow8550
  %inc.i.i.i264 = add i32 %spec.store.select.i.i.i231, 1
  br label %Flow8551, !amdgpu.uniform !42

Flow8551:                                         ; preds = %for.inc.i.i.i259, %Flow8550
  %446 = phi i32 [ %inc.i.i.i264, %for.inc.i.i.i259 ], [ poison, %Flow8550 ]
  %447 = phi i1 [ false, %for.inc.i.i.i259 ], [ true, %Flow8550 ]
  br i1 %447, label %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i265, label %for.cond.i.i1.i224, !amdgpu.uniform !42

_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i265: ; preds = %Flow8551
  %.lcssa9324 = phi i32 [ %439, %Flow8551 ]
  %.lcssa9323 = phi i64 [ %440, %Flow8551 ]
  %.lcssa9322 = phi i32 [ %441, %Flow8551 ]
  fence acquire
  %448 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %mul.i.i.i.i.i.i266 = shl i32 %.lcssa9324, 6
  %idxprom.i19.i.i.i.i.i267 = zext i32 %mul.i.i.i.i.i.i266 to i64
  %arrayidx.i20.i.i.i.i.i268 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %448, i64 %idxprom.i19.i.i.i.i.i267
  %arrayidx.i22.i.i.i.i.i269 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %arrayidx.i20.i.i.i.i.i268, i64 %sh_prom.i.i.i.i
  store i64 sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.13.31 to ptr), i64 8) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.13.31 to ptr) to i64)), ptr %arrayidx.i22.i.i.i.i.i269, align 8, !tbaa !82
  %spec.select.i.i.i.i.i.i.i270 = tail call i64 @llvm.umin.i64(i64 sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.13.31 to ptr), i64 8) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.13.31 to ptr) to i64)), i64 56)
  %449 = and i64 %spec.select.i.i.i.i.i.i.i270, 15
  %450 = sub i64 %spec.select.i.i.i.i.i.i.i270, %449
  %451 = icmp ne i64 %450, 0
  br i1 %451, label %loop-memcpy-expansion5211.preheader, label %Flow8547, !amdgpu.uniform !42

loop-memcpy-expansion5211.preheader:              ; preds = %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i265
  %452 = add nuw nsw i64 %78, 8
  %scevgep7209 = getelementptr i8, ptr %448, i64 %452
  %453 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i267, 6
  %scevgep7210 = getelementptr i8, ptr %scevgep7209, i64 %453
  br label %loop-memcpy-expansion5211, !amdgpu.uniform !42

Flow8547:                                         ; preds = %Flow8546, %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i265
  br label %loop-memcpy-residual-header5214, !amdgpu.uniform !42

loop-memcpy-expansion5211:                        ; preds = %loop-memcpy-expansion5211.preheader, %loop-memcpy-expansion5211
  %loop-index5212 = phi i64 [ %456, %loop-memcpy-expansion5211 ], [ 0, %loop-memcpy-expansion5211.preheader ]
  %454 = getelementptr i8, ptr addrspace(4) @.str.13.31, i64 %loop-index5212, !amdgpu.uniform !42
  %455 = load <4 x i32>, ptr addrspace(4) %454, align 1
  %scevgep7211 = getelementptr i8, ptr %scevgep7210, i64 %loop-index5212
  store <4 x i32> %455, ptr %scevgep7211, align 1
  %456 = add i64 %loop-index5212, 16
  %457 = icmp uge i64 %456, %450
  br i1 %457, label %Flow8546, label %loop-memcpy-expansion5211, !amdgpu.uniform !42

Flow8545:                                         ; preds = %Flow8543, %Flow8544
  %458 = phi i64 [ %6519, %Flow8543 ], [ %6516, %Flow8544 ]
  br label %post-loop-memcpy-expansion5210, !amdgpu.uniform !42

loop-memcpy-residual5213:                         ; preds = %loop-memcpy-residual5213.preheader, %loop-memcpy-residual5213
  %residual-loop-index5215 = phi i64 [ %460, %loop-memcpy-residual5213 ], [ 0, %loop-memcpy-residual5213.preheader ]
  %scevgep7208 = getelementptr i8, ptr addrspace(4) @.str.13.31, i64 %residual-loop-index5215, !amdgpu.uniform !42
  %459 = load i8, ptr addrspace(4) %scevgep7208, align 1
  %scevgep7207 = getelementptr i8, ptr %scevgep7206, i64 %residual-loop-index5215
  store i8 %459, ptr %scevgep7207, align 1
  %460 = add i64 %residual-loop-index5215, 1
  %461 = icmp uge i64 %460, %449
  br i1 %461, label %Flow8543, label %loop-memcpy-residual5213, !amdgpu.uniform !42

Flow8543:                                         ; preds = %loop-memcpy-residual5213
  br label %Flow8545, !amdgpu.uniform !42

post-loop-memcpy-expansion5210:                   ; preds = %Flow8545
  %tobool.not.i.i.i.i.i.i272 = icmp eq i32 %.lcssa9322, 0
  %conv.i.i.i.i.i.i273 = zext i1 %tobool.not.i.i.i.i.i.i272 to i32
  fence release
  %462 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %idxprom.i23.i.i.i.i.i274 = zext i32 %.lcssa9324 to i64
  %arrayidx.i24.i.i.i.i.i275 = getelementptr inbounds nuw i32, ptr %462, i64 %idxprom.i23.i.i.i.i.i274
  store atomic i32 %conv.i.i.i.i.i.i273, ptr %arrayidx.i24.i.i.i.i.i275 monotonic, align 4
  %463 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !84
  %arrayidx.i.i5.i.i276 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %463, i64 %idxprom.i23.i.i.i.i.i274, !amdgpu.uniform !42
  %464 = load i64, ptr %arrayidx.i.i5.i.i276, align 8, !tbaa !80
  br label %while.cond.i.i.i.i277, !amdgpu.uniform !42

while.cond.i.i.i.i277:                            ; preds = %post-loop-memcpy-expansion5210, %Flow8542
  %phi.broken8628 = phi i64 [ 0, %post-loop-memcpy-expansion5210 ], [ %481, %Flow8542 ]
  %lsr.iv7198 = phi i64 [ %479, %Flow8542 ], [ -48, %post-loop-memcpy-expansion5210 ]
  %lsr.iv7193 = phi ptr addrspace(4) [ %478, %Flow8542 ], [ getelementptr (i8, ptr addrspace(4) @.str.13.31, i64 56), %post-loop-memcpy-expansion5210 ]
  %port.sroa.37.0.i.i278 = phi i32 [ %477, %Flow8542 ], [ %conv.i.i.i.i.i.i273, %post-loop-memcpy-expansion5210 ]
  %idx.0.i.i.i.i279 = phi i64 [ %476, %Flow8542 ], [ 56, %post-loop-memcpy-expansion5210 ]
  %cmp.i.i.i.i280 = icmp ult i64 %idx.0.i.i.i.i279, sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.13.31 to ptr), i64 8) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.13.31 to ptr) to i64))
  %465 = tail call i64 @llvm.amdgcn.ballot.i64(i1 %cmp.i.i.i.i280)
  %and.i.i.i.i6.i.i281 = and i64 %465, %464
  %tobool.not.i.i.i.i282 = icmp ne i64 %and.i.i.i.i6.i.i281, 0
  %466 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i22.i.i.i283 = getelementptr inbounds nuw i32, ptr %466, i64 %idxprom.i23.i.i.i.i.i274, !amdgpu.uniform !42
  %467 = load atomic i32, ptr %arrayidx.i22.i.i.i283 monotonic, align 4
  %468 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %tobool.not.i.i.i.i282)
  %469 = extractvalue { i1, i64 } %468, 0
  %470 = extractvalue { i1, i64 } %468, 1
  br i1 %469, label %cond.false.i8.i.i.i.i284, label %Flow8542

cond.false.i8.i.i.i.i284:                         ; preds = %while.cond.i.i.i.i277
  %471 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %467)
  %472 = icmp ne i32 %471, %port.sroa.37.0.i.i278
  br i1 %472, label %while.body.i.i21.i.i.i.i285.preheader, label %Flow8541, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i285.preheader:            ; preds = %cond.false.i8.i.i.i.i284
  br label %while.body.i.i21.i.i.i.i285, !amdgpu.uniform !42

Flow8541:                                         ; preds = %Flow8540, %cond.false.i8.i.i.i.i284
  br label %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i288, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i285:                      ; preds = %while.body.i.i21.i.i.i.i285.preheader, %while.body.i.i21.i.i.i.i285
  tail call void @llvm.amdgcn.s.sleep(i32 2)
  %473 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i22.i.i.i.i286 = getelementptr inbounds nuw i32, ptr %473, i64 %idxprom.i23.i.i.i.i.i274, !amdgpu.uniform !42
  %474 = load atomic i32, ptr %arrayidx.i.i.i22.i.i.i.i286 monotonic, align 4
  %475 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %474)
  %cmp.i.not.i.i23.i.i.i.i287 = icmp eq i32 %475, %port.sroa.37.0.i.i278
  br i1 %cmp.i.not.i.i23.i.i.i.i287, label %Flow8540, label %while.body.i.i21.i.i.i.i285, !amdgpu.uniform !42

Flow8540:                                         ; preds = %while.body.i.i21.i.i.i.i285
  br label %Flow8541, !amdgpu.uniform !42

Flow8542:                                         ; preds = %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i289, %while.cond.i.i.i.i277
  %476 = phi i64 [ %add.i.i.i.i292, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i289 ], [ poison, %while.cond.i.i.i.i277 ]
  %477 = phi i32 [ %conv.i.i26.i.i.i.i290, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i289 ], [ poison, %while.cond.i.i.i.i277 ]
  %478 = phi ptr addrspace(4) [ %scevgep7194, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i289 ], [ poison, %while.cond.i.i.i.i277 ]
  %479 = phi i64 [ %lsr.iv.next7199, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i289 ], [ poison, %while.cond.i.i.i.i277 ]
  %480 = phi i1 [ false, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i289 ], [ true, %while.cond.i.i.i.i277 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %470)
  %481 = call i64 @llvm.amdgcn.if.break.i64(i1 %480, i64 %phi.broken8628)
  %482 = call i1 @llvm.amdgcn.loop.i64(i64 %481)
  br i1 %482, label %cond.false.i.i.i299, label %while.cond.i.i.i.i277

_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i288: ; preds = %Flow8541
  fence acquire
  br i1 %cmp.i.i.i.i280, label %if.then.i.i.i.i.i.i.i293, label %Flow8539, !amdgpu.uniform !42

if.then.i.i.i.i.i.i.i293:                         ; preds = %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i288
  %sub.i.i.i.i.i.i.i295 = sub nuw i64 sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.13.31 to ptr), i64 8) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.13.31 to ptr) to i64)), %idx.0.i.i.i.i279
  %spec.select.i.i.i36.i.i.i.i296 = tail call i64 @llvm.umin.i64(i64 %sub.i.i.i.i.i.i.i295, i64 64)
  %483 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %484 = and i64 %spec.select.i.i.i36.i.i.i.i296, 15
  %485 = sub i64 %spec.select.i.i.i36.i.i.i.i296, %484
  %486 = icmp ne i64 %485, 0
  br i1 %486, label %loop-memcpy-expansion5217.preheader, label %Flow8538, !amdgpu.uniform !42

loop-memcpy-expansion5217.preheader:              ; preds = %if.then.i.i.i.i.i.i.i293
  %arrayidx.i20.i35.i.i.i.i297 = getelementptr %"struct.rpc::Buffer", ptr %483, i64 %idxprom.i19.i.i.i.i.i267
  %arrayidx.i22.i38.i.i.i.i298 = getelementptr %"struct.rpc::Buffer", ptr %arrayidx.i20.i35.i.i.i.i297, i64 %sh_prom.i.i.i.i
  br label %loop-memcpy-expansion5217, !amdgpu.uniform !42

Flow8538:                                         ; preds = %Flow8537, %if.then.i.i.i.i.i.i.i293
  br label %loop-memcpy-residual-header5220, !amdgpu.uniform !42

loop-memcpy-expansion5217:                        ; preds = %loop-memcpy-expansion5217.preheader, %loop-memcpy-expansion5217
  %loop-index5218 = phi i64 [ %489, %loop-memcpy-expansion5217 ], [ 0, %loop-memcpy-expansion5217.preheader ]
  %scevgep7195 = getelementptr i8, ptr addrspace(4) %lsr.iv7193, i64 %loop-index5218, !amdgpu.uniform !42
  %487 = load <4 x i32>, ptr addrspace(4) %scevgep7195, align 1
  %488 = getelementptr i8, ptr %arrayidx.i22.i38.i.i.i.i298, i64 %loop-index5218
  store <4 x i32> %487, ptr %488, align 1
  %489 = add i64 %loop-index5218, 16
  %490 = icmp uge i64 %489, %485
  br i1 %490, label %Flow8537, label %loop-memcpy-expansion5217, !amdgpu.uniform !42

Flow8536:                                         ; preds = %Flow8535, %loop-memcpy-residual-header5220
  br label %Flow8539, !amdgpu.uniform !42

loop-memcpy-residual5219:                         ; preds = %loop-memcpy-residual5219.preheader, %loop-memcpy-residual5219
  %residual-loop-index5221 = phi i64 [ %492, %loop-memcpy-residual5219 ], [ 0, %loop-memcpy-residual5219.preheader ]
  %scevgep7204 = getelementptr i8, ptr addrspace(4) %scevgep7203, i64 %residual-loop-index5221, !amdgpu.uniform !42
  %491 = load i8, ptr addrspace(4) %scevgep7204, align 1
  %scevgep7202 = getelementptr i8, ptr %scevgep7201, i64 %residual-loop-index5221
  store i8 %491, ptr %scevgep7202, align 1
  %492 = add i64 %residual-loop-index5221, 1
  %493 = icmp uge i64 %492, %484
  br i1 %493, label %Flow8535, label %loop-memcpy-residual5219, !amdgpu.uniform !42

Flow8535:                                         ; preds = %loop-memcpy-residual5219
  br label %Flow8536, !amdgpu.uniform !42

_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i289: ; preds = %Flow8539
  %conv.i.i26.i.i.i.i290 = xor i32 %port.sroa.37.0.i.i278, 1
  fence release
  %494 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %arrayidx.i24.i29.i.i.i.i291 = getelementptr inbounds nuw i32, ptr %494, i64 %idxprom.i23.i.i.i.i.i274
  store atomic i32 %conv.i.i26.i.i.i.i290, ptr %arrayidx.i24.i29.i.i.i.i291 monotonic, align 4
  %add.i.i.i.i292 = add i64 %idx.0.i.i.i.i279, 64
  %scevgep7194 = getelementptr i8, ptr addrspace(4) %lsr.iv7193, i64 64
  %lsr.iv.next7199 = add i64 %lsr.iv7198, -64
  br label %Flow8542, !amdgpu.uniform !42

cond.false.i.i.i299:                              ; preds = %Flow8542
  %.lcssa9318 = phi i64 [ %481, %Flow8542 ]
  %port.sroa.37.0.i.i278.lcssa = phi i32 [ %port.sroa.37.0.i.i278, %Flow8542 ]
  %.lcssa9317 = phi i32 [ %467, %Flow8542 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa9318)
  %495 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %.lcssa9317)
  %496 = icmp ne i32 %495, %port.sroa.37.0.i.i278.lcssa
  %497 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %496)
  %498 = extractvalue { i1, i64 } %497, 0
  %499 = extractvalue { i1, i64 } %497, 1
  br i1 %498, label %while.body.i.i.i.i300.preheader, label %Flow8534

while.body.i.i.i.i300.preheader:                  ; preds = %cond.false.i.i.i299
  br label %while.body.i.i.i.i300, !amdgpu.uniform !42

Flow8534:                                         ; preds = %Flow8533, %cond.false.i.i.i299
  call void @llvm.amdgcn.end.cf.i64(i64 %499)
  br label %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit320, !amdgpu.uniform !42

while.body.i.i.i.i300:                            ; preds = %while.body.i.i.i.i300.preheader, %while.body.i.i.i.i300
  %phi.broken8629 = phi i64 [ 0, %while.body.i.i.i.i300.preheader ], [ %503, %while.body.i.i.i.i300 ]
  tail call void @llvm.amdgcn.s.sleep(i32 2)
  %500 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i8.i.i301 = getelementptr inbounds nuw i32, ptr %500, i64 %idxprom.i23.i.i.i.i.i274, !amdgpu.uniform !42
  %501 = load atomic i32, ptr %arrayidx.i.i.i8.i.i301 monotonic, align 4
  %502 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %501)
  %cmp.i.not.i.i.i.i302 = icmp eq i32 %502, %port.sroa.37.0.i.i278.lcssa
  %503 = call i64 @llvm.amdgcn.if.break.i64(i1 %cmp.i.not.i.i.i.i302, i64 %phi.broken8629)
  %504 = call i1 @llvm.amdgcn.loop.i64(i64 %503)
  br i1 %504, label %Flow8533, label %while.body.i.i.i.i300

Flow8533:                                         ; preds = %while.body.i.i.i.i300
  %.lcssa9316 = phi i64 [ %503, %while.body.i.i.i.i300 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa9316)
  br label %Flow8534, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit320: ; preds = %Flow8534
  fence acquire
  tail call void @llvm.amdgcn.wave.barrier()
  fence syncscope("agent") release
  %505 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %.lcssa9323, i1 true)
  %iszero.i.i.i.i.i13.i.i303 = icmp ne i64 %.lcssa9323, 0
  %cmp2.i.i.i.i14.i.i304 = icmp eq i64 %505, %sh_prom.i.i.i.i
  %cmp.i.i.i.i15.i.i305 = select i1 %iszero.i.i.i.i.i13.i.i303, i1 %cmp2.i.i.i.i14.i.i304, i1 false
  %506 = and i32 %.lcssa9324, 31
  %conv4.i.i.i19.i.i309 = zext i1 %cmp.i.i.i.i15.i.i305 to i32
  %shl.i.i.i20.i.i310 = shl nuw i32 %conv4.i.i.i19.i.i309, %506
  %xor.i.i.i21.i.i311 = xor i32 %shl.i.i.i20.i.i310, -1
  %507 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop5511, !amdgpu.uniform !42

508:                                              ; preds = %ComputeEnd5512
  %div8.i.i.i16.i.i306 = lshr i32 %.lcssa9324, 5
  %idxprom.i.i.i17.i.i307 = zext nneg i32 %div8.i.i.i16.i.i306 to i64
  %509 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i17.i.i307
  %sunkaddr7373 = getelementptr inbounds i8, ptr addrspace(1) %509, i64 40
  %510 = atomicrmw and ptr addrspace(1) %sunkaddr7373, i32 %.lcssa9315 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %511, !amdgpu.uniform !42

511:                                              ; preds = %508, %ComputeEnd5512
  call void @llvm.amdgcn.end.cf.i64(i64 %6904)
  tail call void @llvm.amdgcn.wave.barrier()
  call void @llvm.lifetime.start.p5(ptr addrspace(5) %buffer.i.i.i2988) #34, !noalias !153
  %retval.0.i.i.i.i.i.i = tail call noundef i32 @llvm.abs.i32(i32 %1, i1 false)
  br label %while.body.i.i.i.i.i.i, !amdgpu.uniform !42

while.body.i.i.i.i.i.i:                           ; preds = %511, %_ZN22__llvm_libc_22_0_0_git7details22StringBufferWriterImplILb0EE4pushEc.exit.i.i.i.i.i.i
  %phi.broken8630 = phi i64 [ 0, %511 ], [ %579, %_ZN22__llvm_libc_22_0_0_git7details22StringBufferWriterImplILb0EE4pushEc.exit.i.i.i.i.i.i ]
  %writer.sroa.10.0.i.i.i.i = phi i64 [ %writer.sroa.10.1.i.i.i.i, %_ZN22__llvm_libc_22_0_0_git7details22StringBufferWriterImplILb0EE4pushEc.exit.i.i.i.i.i.i ], [ 0, %511 ]
  %value.addr.09.i.i.i.i.i.i = phi i32 [ %div.i.i.i.i.i.i.i, %_ZN22__llvm_libc_22_0_0_git7details22StringBufferWriterImplILb0EE4pushEc.exit.i.i.i.i.i.i ], [ %retval.0.i.i.i.i.i.i, %511 ]
  br label %NodeBlock7523, !amdgpu.uniform !42

NodeBlock7523:                                    ; preds = %while.body.i.i.i.i.i.i
  %div.i.i.i.i.i.i.i = udiv i32 %value.addr.09.i.i.i.i.i.i, 10
  %.neg.i = mul i32 %div.i.i.i.i.i.i.i, 246
  %rem.i.i.i.i.i.i.decomposed.i = add i32 %.neg.i, %value.addr.09.i.i.i.i.i.i
  %conv.i.i.i.i.i.i.i = trunc i32 %rem.i.i.i.i.i.i.decomposed.i to i8
  %512 = zext i8 %conv.i.i.i.i.i.i.i to i16
  %Pivot7524 = icmp sge i16 %512, 5
  %513 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %Pivot7524)
  %514 = extractvalue { i1, i64 } %513, 0
  %515 = extractvalue { i1, i64 } %513, 1
  br i1 %514, label %NodeBlock7521, label %Flow8532

NodeBlock7521:                                    ; preds = %NodeBlock7523
  %Pivot7522 = icmp sge i16 %512, 7
  %516 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %Pivot7522)
  %517 = extractvalue { i1, i64 } %516, 0
  %518 = extractvalue { i1, i64 } %516, 1
  br i1 %517, label %NodeBlock7519, label %Flow8523

NodeBlock7519:                                    ; preds = %NodeBlock7521
  %Pivot7520 = icmp sge i16 %512, 8
  %519 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %Pivot7520)
  %520 = extractvalue { i1, i64 } %519, 0
  %521 = extractvalue { i1, i64 } %519, 1
  br i1 %520, label %NodeBlock7517, label %Flow8519

NodeBlock7517:                                    ; preds = %NodeBlock7519
  %Pivot7518 = icmp sge i16 %512, 9
  %522 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %Pivot7518)
  %523 = extractvalue { i1, i64 } %522, 0
  %524 = extractvalue { i1, i64 } %522, 1
  br i1 %523, label %sw.bb9.i.i.i.i.i.i.i.i, label %Flow8517

Flow8523:                                         ; preds = %Flow8520, %NodeBlock7521
  %525 = phi i8 [ %569, %Flow8520 ], [ poison, %NodeBlock7521 ]
  %526 = call { i1, i64 } @llvm.amdgcn.else.i64.i64(i64 %518)
  %527 = extractvalue { i1, i64 } %526, 0
  %528 = extractvalue { i1, i64 } %526, 1
  br i1 %527, label %NodeBlock7515, label %Flow8524

NodeBlock7515:                                    ; preds = %Flow8523
  %Pivot7516 = icmp sge i16 %512, 6
  %529 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %Pivot7516)
  %530 = extractvalue { i1, i64 } %529, 0
  %531 = extractvalue { i1, i64 } %529, 1
  br i1 %530, label %sw.bb6.i.i.i.i.i.i.i.i, label %Flow8521

Flow8532:                                         ; preds = %Flow8524, %NodeBlock7523
  %532 = phi i8 [ %571, %Flow8524 ], [ poison, %NodeBlock7523 ]
  %533 = call { i1, i64 } @llvm.amdgcn.else.i64.i64(i64 %515)
  %534 = extractvalue { i1, i64 } %533, 0
  %535 = extractvalue { i1, i64 } %533, 1
  br i1 %534, label %NodeBlock7513, label %if.then.i.i.i.i.i.i.i321

NodeBlock7513:                                    ; preds = %Flow8532
  %Pivot7514 = icmp sge i16 %512, 3
  %536 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %Pivot7514)
  %537 = extractvalue { i1, i64 } %536, 0
  %538 = extractvalue { i1, i64 } %536, 1
  br i1 %537, label %NodeBlock7511, label %Flow8530

NodeBlock7511:                                    ; preds = %NodeBlock7513
  %Pivot7512 = icmp sge i16 %512, 4
  %539 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %Pivot7512)
  %540 = extractvalue { i1, i64 } %539, 0
  %541 = extractvalue { i1, i64 } %539, 1
  br i1 %540, label %sw.bb4.i.i.i.i.i.i.i.i, label %Flow8525

Flow8530:                                         ; preds = %Flow8526, %NodeBlock7513
  %542 = phi i8 [ %572, %Flow8526 ], [ poison, %NodeBlock7513 ]
  %543 = call { i1, i64 } @llvm.amdgcn.else.i64.i64(i64 %538)
  %544 = extractvalue { i1, i64 } %543, 0
  %545 = extractvalue { i1, i64 } %543, 1
  br i1 %544, label %NodeBlock, label %Flow8531

NodeBlock:                                        ; preds = %Flow8530
  %Pivot = icmp sge i16 %512, 2
  %546 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %Pivot)
  %547 = extractvalue { i1, i64 } %546, 0
  %548 = extractvalue { i1, i64 } %546, 1
  br i1 %547, label %sw.bb2.i.i.i.i.i.i.i.i, label %Flow8528

Flow8528:                                         ; preds = %sw.bb2.i.i.i.i.i.i.i.i, %NodeBlock
  %549 = call { i1, i64 } @llvm.amdgcn.else.i64.i64(i64 %548)
  %550 = extractvalue { i1, i64 } %549, 0
  %551 = extractvalue { i1, i64 } %549, 1
  br i1 %550, label %LeafBlock, label %Flow8529

LeafBlock:                                        ; preds = %Flow8528
  %SwitchLeaf = icmp eq i16 %512, 1
  %552 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %SwitchLeaf)
  %553 = extractvalue { i1, i64 } %552, 0
  %554 = extractvalue { i1, i64 } %552, 1
  br i1 %553, label %sw.bb1.i.i.i.i.i.i.i.i, label %Flow8527

sw.bb1.i.i.i.i.i.i.i.i:                           ; preds = %LeafBlock
  br label %Flow8527, !amdgpu.uniform !42

sw.bb2.i.i.i.i.i.i.i.i:                           ; preds = %NodeBlock
  br label %Flow8528, !amdgpu.uniform !42

Flow8525:                                         ; preds = %sw.bb4.i.i.i.i.i.i.i.i, %NodeBlock7511
  %555 = call { i1, i64 } @llvm.amdgcn.else.i64.i64(i64 %541)
  %556 = extractvalue { i1, i64 } %555, 0
  %557 = extractvalue { i1, i64 } %555, 1
  br i1 %556, label %sw.bb3.i.i.i.i.i.i.i.i, label %Flow8526

sw.bb3.i.i.i.i.i.i.i.i:                           ; preds = %Flow8525
  br label %Flow8526, !amdgpu.uniform !42

sw.bb4.i.i.i.i.i.i.i.i:                           ; preds = %NodeBlock7511
  br label %Flow8525, !amdgpu.uniform !42

Flow8521:                                         ; preds = %sw.bb6.i.i.i.i.i.i.i.i, %NodeBlock7515
  %558 = call { i1, i64 } @llvm.amdgcn.else.i64.i64(i64 %531)
  %559 = extractvalue { i1, i64 } %558, 0
  %560 = extractvalue { i1, i64 } %558, 1
  br i1 %559, label %sw.bb5.i.i.i.i.i.i.i.i, label %Flow8522

sw.bb5.i.i.i.i.i.i.i.i:                           ; preds = %Flow8521
  br label %Flow8522, !amdgpu.uniform !42

sw.bb6.i.i.i.i.i.i.i.i:                           ; preds = %NodeBlock7515
  br label %Flow8521, !amdgpu.uniform !42

Flow8519:                                         ; preds = %Flow8518, %NodeBlock7519
  %561 = phi i8 [ %568, %Flow8518 ], [ poison, %NodeBlock7519 ]
  %562 = call { i1, i64 } @llvm.amdgcn.else.i64.i64(i64 %521)
  %563 = extractvalue { i1, i64 } %562, 0
  %564 = extractvalue { i1, i64 } %562, 1
  br i1 %563, label %sw.bb7.i.i.i.i.i.i.i.i, label %Flow8520

sw.bb7.i.i.i.i.i.i.i.i:                           ; preds = %Flow8519
  br label %Flow8520, !amdgpu.uniform !42

Flow8517:                                         ; preds = %sw.bb9.i.i.i.i.i.i.i.i, %NodeBlock7517
  %565 = call { i1, i64 } @llvm.amdgcn.else.i64.i64(i64 %524)
  %566 = extractvalue { i1, i64 } %565, 0
  %567 = extractvalue { i1, i64 } %565, 1
  br i1 %566, label %sw.bb8.i.i.i.i.i.i.i.i, label %Flow8518

sw.bb8.i.i.i.i.i.i.i.i:                           ; preds = %Flow8517
  br label %Flow8518, !amdgpu.uniform !42

sw.bb9.i.i.i.i.i.i.i.i:                           ; preds = %NodeBlock7517
  br label %Flow8517, !amdgpu.uniform !42

Flow8518:                                         ; preds = %sw.bb8.i.i.i.i.i.i.i.i, %Flow8517
  %568 = phi i8 [ 56, %sw.bb8.i.i.i.i.i.i.i.i ], [ 57, %Flow8517 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %567)
  br label %Flow8519, !amdgpu.uniform !42

Flow8520:                                         ; preds = %sw.bb7.i.i.i.i.i.i.i.i, %Flow8519
  %569 = phi i8 [ 55, %sw.bb7.i.i.i.i.i.i.i.i ], [ %561, %Flow8519 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %564)
  br label %Flow8523, !amdgpu.uniform !42

Flow8522:                                         ; preds = %sw.bb5.i.i.i.i.i.i.i.i, %Flow8521
  %570 = phi i8 [ 53, %sw.bb5.i.i.i.i.i.i.i.i ], [ 54, %Flow8521 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %560)
  br label %Flow8524, !amdgpu.uniform !42

Flow8524:                                         ; preds = %Flow8522, %Flow8523
  %571 = phi i8 [ %570, %Flow8522 ], [ %525, %Flow8523 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %528)
  br label %Flow8532, !amdgpu.uniform !42

Flow8526:                                         ; preds = %sw.bb3.i.i.i.i.i.i.i.i, %Flow8525
  %572 = phi i8 [ 51, %sw.bb3.i.i.i.i.i.i.i.i ], [ 52, %Flow8525 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %557)
  br label %Flow8530, !amdgpu.uniform !42

Flow8527:                                         ; preds = %sw.bb1.i.i.i.i.i.i.i.i, %LeafBlock
  %573 = phi i8 [ 49, %sw.bb1.i.i.i.i.i.i.i.i ], [ 48, %LeafBlock ]
  call void @llvm.amdgcn.end.cf.i64(i64 %554)
  br label %Flow8529, !amdgpu.uniform !42

Flow8529:                                         ; preds = %Flow8527, %Flow8528
  %574 = phi i8 [ %573, %Flow8527 ], [ 50, %Flow8528 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %551)
  br label %Flow8531, !amdgpu.uniform !42

Flow8531:                                         ; preds = %Flow8529, %Flow8530
  %575 = phi i8 [ %574, %Flow8529 ], [ %542, %Flow8530 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %545)
  br label %if.then.i.i.i.i.i.i.i321, !amdgpu.uniform !42

if.then.i.i.i.i.i.i.i321:                         ; preds = %Flow8532, %Flow8531
  %retval.0.i.i.i.i.i.i.i.i = phi i8 [ %532, %Flow8532 ], [ %575, %Flow8531 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %535)
  %cmp.i.i.i.i.i.i.i.i = icmp eq i64 %writer.sroa.10.0.i.i.i.i, 11
  %cmp.i.i.i.i.i.i.i.i.inv = xor i1 %cmp.i.i.i.i.i.i.i.i, true
  br i1 %cmp.i.i.i.i.i.i.i.i.inv, label %if.then3.i.i.i.i.i.i.i, label %_ZN22__llvm_libc_22_0_0_git7details22StringBufferWriterImplILb0EE4pushEc.exit.i.i.i.i.i.i, !amdgpu.uniform !42

if.then3.i.i.i.i.i.i.i:                           ; preds = %if.then.i.i.i.i.i.i.i321
  %576 = trunc i64 %writer.sroa.10.0.i.i.i.i to i32
  %577 = xor i32 %576, -1
  %578 = getelementptr i8, ptr addrspace(5) %buffer.i.i.i2988, i32 %577
  %arrayidx.i.i.i.i.i.i.i.i6017 = getelementptr i8, ptr addrspace(5) %578, i32 11
  %inc.i.i.i.i.i.i.i = add i64 %writer.sroa.10.0.i.i.i.i, 1
  store i8 %retval.0.i.i.i.i.i.i.i.i, ptr addrspace(5) %arrayidx.i.i.i.i.i.i.i.i6017, align 1, !tbaa !57, !noalias !153
  br label %_ZN22__llvm_libc_22_0_0_git7details22StringBufferWriterImplILb0EE4pushEc.exit.i.i.i.i.i.i, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git7details22StringBufferWriterImplILb0EE4pushEc.exit.i.i.i.i.i.i: ; preds = %if.then3.i.i.i.i.i.i.i, %if.then.i.i.i.i.i.i.i321
  %writer.sroa.10.1.i.i.i.i = phi i64 [ %inc.i.i.i.i.i.i.i, %if.then3.i.i.i.i.i.i.i ], [ 11, %if.then.i.i.i.i.i.i.i321 ]
  %cmp.i.i.i.i.i.i322 = icmp ult i32 %value.addr.09.i.i.i.i.i.i, 10
  %.not.i.i.i.i.i.i = or i1 %cmp.i.i.i.i.i.i.i.i, %cmp.i.i.i.i.i.i322
  %579 = call i64 @llvm.amdgcn.if.break.i64(i1 %.not.i.i.i.i.i.i, i64 %phi.broken8630)
  %580 = call i1 @llvm.amdgcn.loop.i64(i64 %579)
  br i1 %580, label %_ZN22__llvm_libc_22_0_0_git15IntegerToStringIiNS_7details3FmtILh10ELb0ELb0ELb0ELm1EEEE13IntegerWriter25write_unsigned_number_decEjRNS1_22StringBufferWriterImplILb0EEE.exit.i.i.i.i.i, label %while.body.i.i.i.i.i.i

_ZN22__llvm_libc_22_0_0_git15IntegerToStringIiNS_7details3FmtILh10ELb0ELb0ELb0ELm1EEEE13IntegerWriter25write_unsigned_number_decEjRNS1_22StringBufferWriterImplILb0EEE.exit.i.i.i.i.i: ; preds = %_ZN22__llvm_libc_22_0_0_git7details22StringBufferWriterImplILb0EE4pushEc.exit.i.i.i.i.i.i
  %writer.sroa.10.1.i.i.i.i.lcssa = phi i64 [ %writer.sroa.10.1.i.i.i.i, %_ZN22__llvm_libc_22_0_0_git7details22StringBufferWriterImplILb0EE4pushEc.exit.i.i.i.i.i.i ]
  %.lcssa9314 = phi i64 [ %579, %_ZN22__llvm_libc_22_0_0_git7details22StringBufferWriterImplILb0EE4pushEc.exit.i.i.i.i.i.i ]
  %cmp.i.i.i.i.i.i.i.i.lcssa = phi i1 [ %cmp.i.i.i.i.i.i.i.i, %_ZN22__llvm_libc_22_0_0_git7details22StringBufferWriterImplILb0EE4pushEc.exit.i.i.i.i.i.i ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa9314)
  %cmp.i19.i.i.i.i = icmp ne i64 %writer.sroa.10.1.i.i.i.i.lcssa, 0
  %or.cond.not.i.i.i.i = select i1 %cmp.i.i.i.i.i.i.i.i.lcssa, i1 true, i1 %cmp.i19.i.i.i.i
  %or.cond.not.i.i.i.i.inv = xor i1 %or.cond.not.i.i.i.i, true
  %581 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %or.cond.not.i.i.i.i.inv)
  %582 = extractvalue { i1, i64 } %581, 0
  %583 = extractvalue { i1, i64 } %581, 1
  br i1 %582, label %if.then3.i.i.lr.ph.i.i.i.i, label %while.end.i.i.i.i.i

if.then3.i.i.lr.ph.i.i.i.i:                       ; preds = %_ZN22__llvm_libc_22_0_0_git15IntegerToStringIiNS_7details3FmtILh10ELb0ELb0ELb0ELm1EEEE13IntegerWriter25write_unsigned_number_decEjRNS1_22StringBufferWriterImplILb0EEE.exit.i.i.i.i.i
  %arrayidx.i.i.i.i.i.i.i = getelementptr inbounds nuw i8, ptr addrspace(5) %buffer.i.i.i2988, i32 10
  store i8 48, ptr addrspace(5) %arrayidx.i.i.i.i.i.i.i, align 2, !tbaa !57, !noalias !153
  br label %while.end.i.i.i.i.i, !amdgpu.uniform !42

while.end.i.i.i.i.i:                              ; preds = %if.then3.i.i.lr.ph.i.i.i.i, %_ZN22__llvm_libc_22_0_0_git15IntegerToStringIiNS_7details3FmtILh10ELb0ELb0ELb0ELm1EEEE13IntegerWriter25write_unsigned_number_decEjRNS1_22StringBufferWriterImplILb0EEE.exit.i.i.i.i.i
  %writer.sroa.10.4.i.i.i.i = phi i64 [ 1, %if.then3.i.i.lr.ph.i.i.i.i ], [ %writer.sroa.10.1.i.i.i.i.lcssa, %_ZN22__llvm_libc_22_0_0_git15IntegerToStringIiNS_7details3FmtILh10ELb0ELb0ELb0ELm1EEEE13IntegerWriter25write_unsigned_number_decEjRNS1_22StringBufferWriterImplILb0EEE.exit.i.i.i.i.i ]
  call void @llvm.amdgcn.end.cf.i64(i64 %583)
  %cmp4.i.i.i.i.i = icmp sgt i32 %1, -1
  %brmerge.i.i.i.i.i = or i1 %cmp4.i.i.i.i.i, %cmp.i.i.i.i.i.i.i.i.lcssa
  %brmerge.i.i.i.i.i.inv = xor i1 %brmerge.i.i.i.i.i, true
  %584 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %brmerge.i.i.i.i.i.inv)
  %585 = extractvalue { i1, i64 } %584, 0
  %586 = extractvalue { i1, i64 } %584, 1
  br i1 %585, label %if.then.i14.i.i.i.i.i, label %_ZN22__llvm_libc_22_0_0_git15IntegerToStringIiNS_7details3FmtILh10ELb0ELb0ELb0ELm1EEEEC2Ei.exit.i.i.i

if.then.i14.i.i.i.i.i:                            ; preds = %while.end.i.i.i.i.i
  %cmp.i.i17.i.i.i.i.i = icmp ne i64 %writer.sroa.10.4.i.i.i.i, 11
  %587 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.i.i17.i.i.i.i.i)
  %588 = extractvalue { i1, i64 } %587, 0
  %589 = extractvalue { i1, i64 } %587, 1
  br i1 %588, label %if.then3.i18.i.i.i.i.i, label %Flow8516

if.then3.i18.i.i.i.i.i:                           ; preds = %if.then.i14.i.i.i.i.i
  %590 = trunc i64 %writer.sroa.10.4.i.i.i.i to i32
  %591 = xor i32 %590, -1
  %592 = getelementptr i8, ptr addrspace(5) %buffer.i.i.i2988, i32 %591
  %arrayidx.i.i19.i.i.i.i.i6018 = getelementptr i8, ptr addrspace(5) %592, i32 11
  %inc.i20.i.i.i.i.i = add i64 %writer.sroa.10.4.i.i.i.i, 1
  store i8 45, ptr addrspace(5) %arrayidx.i.i19.i.i.i.i.i6018, align 1, !tbaa !57, !noalias !153
  br label %Flow8516, !amdgpu.uniform !42

Flow8516:                                         ; preds = %if.then3.i18.i.i.i.i.i, %if.then.i14.i.i.i.i.i
  %593 = phi i64 [ %inc.i20.i.i.i.i.i, %if.then3.i18.i.i.i.i.i ], [ 11, %if.then.i14.i.i.i.i.i ]
  call void @llvm.amdgcn.end.cf.i64(i64 %589)
  br label %_ZN22__llvm_libc_22_0_0_git15IntegerToStringIiNS_7details3FmtILh10ELb0ELb0ELb0ELm1EEEEC2Ei.exit.i.i.i, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git15IntegerToStringIiNS_7details3FmtILh10ELb0ELb0ELb0ELm1EEEEC2Ei.exit.i.i.i: ; preds = %while.end.i.i.i.i.i, %Flow8516
  %writer.sroa.10.5.i.i.i.i = phi i64 [ %writer.sroa.10.4.i.i.i.i, %while.end.i.i.i.i.i ], [ %593, %Flow8516 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %586)
  %written.i.i.i.i = getelementptr inbounds nuw i8, ptr addrspace(5) %buffer.i.i.i2988, i32 16
  store i64 %writer.sroa.10.5.i.i.i.i, ptr addrspace(5) %written.i.i.i.i, align 8, !tbaa !91, !noalias !153
  %cmp.i.not.i.i.i.i.i = icmp ne i64 %writer.sroa.10.5.i.i.i.i, 0
  %594 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.i.not.i.i.i.i.i)
  %595 = extractvalue { i1, i64 } %594, 0
  %596 = extractvalue { i1, i64 } %594, 1
  br i1 %595, label %if.then.i.i.i.i.i.i, label %Flow8515

if.then.i.i.i.i.i.i:                              ; preds = %_ZN22__llvm_libc_22_0_0_git15IntegerToStringIiNS_7details3FmtILh10ELb0ELb0ELb0ELm1EEEEC2Ei.exit.i.i.i
  %597 = call { i64, i1 } @llvm.uadd.with.overflow.i64(i64 %writer.sroa.10.5.i.i.i.i, i64 1)
  %ov = extractvalue { i64, i1 } %597, 1
  %ov.inv = xor i1 %ov, true
  %598 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %ov.inv)
  %599 = extractvalue { i1, i64 } %598, 0
  %600 = extractvalue { i1, i64 } %598, 1
  br i1 %599, label %if.end.i.i.i.i.i.i.i, label %Flow8514

if.end.i.i.i.i.i.i.i:                             ; preds = %if.then.i.i.i.i.i.i
  %math = extractvalue { i64, i1 } %597, 0
  %cmp2.i.i.i.i.i.i.i = icmp ult i64 %math, 1676976733973595601
  %601 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp2.i.i.i.i.i.i.i)
  %602 = extractvalue { i1, i64 } %601, 0
  %603 = extractvalue { i1, i64 } %601, 1
  br i1 %602, label %if.then3.i.i.i.i4.i.i.i, label %if.end4.i.i.i.i.i.i.i

if.then3.i.i.i.i4.i.i.i:                          ; preds = %if.end.i.i.i.i.i.i.i
  %mul.i.i.i.i.i.i.i = mul nuw i64 %math, 11
  %div18.i.i.i.i.i.i.i = lshr i64 %mul.i.i.i.i.i.i.i, 3
  br label %if.end4.i.i.i.i.i.i.i, !amdgpu.uniform !42

Flow8514:                                         ; preds = %if.end4.i.i.i.i.i.i.i, %if.then.i.i.i.i.i.i
  %604 = phi ptr [ %cond.i.i.i, %if.end4.i.i.i.i.i.i.i ], [ addrspacecast (ptr addrspace(4) @_ZN22__llvm_libc_22_0_0_git3cpp6string14NULL_CHARACTERE to ptr), %if.then.i.i.i.i.i.i ]
  call void @llvm.amdgcn.end.cf.i64(i64 %600)
  br label %_ZN22__llvm_libc_22_0_0_git3cpp6string7reserveEm.exit.i.i.i.i.i.i, !amdgpu.uniform !42

if.end4.i.i.i.i.i.i.i:                            ; preds = %if.then3.i.i.i.i4.i.i.i, %if.end.i.i.i.i.i.i.i
  %new_capacity.addr.0.i.i.i.i.i.i.i = phi i64 [ %div18.i.i.i.i.i.i.i, %if.then3.i.i.i.i4.i.i.i ], [ %math, %if.end.i.i.i.i.i.i.i ]
  call void @llvm.amdgcn.end.cf.i64(i64 %603)
  %sub.i.i.i = add i64 %new_capacity.addr.0.i.i.i.i.i.i.i, 7
  %div2.i.i.i = and i64 %sub.i.i.i, -8
  %605 = load ptr, ptr addrspace(1) @_ZN12_GLOBAL__N_13ptrE, align 8, !tbaa !55
  %add.ptr.i.i.i = getelementptr inbounds nuw i8, ptr %605, i64 %div2.i.i.i
  store ptr %add.ptr.i.i.i, ptr addrspace(1) @_ZN12_GLOBAL__N_13ptrE, align 8, !tbaa !55
  %sub.ptr.lhs.cast.i.i.i323 = ptrtoint ptr %add.ptr.i.i.i to i64
  %sub.ptr.sub.i.i.i324 = sub i64 %sub.ptr.lhs.cast.i.i.i323, ptrtoint (ptr addrspacecast (ptr addrspace(1) @_ZN12_GLOBAL__N_16memoryE to ptr) to i64)
  %cmp.i.i.i325 = icmp ugt i64 %sub.ptr.sub.i.i.i324, 65335
  %cond.i.i.i = select i1 %cmp.i.i.i325, ptr null, ptr %605
  br label %Flow8514, !amdgpu.uniform !42

Flow8515:                                         ; preds = %Flow8513, %_ZN22__llvm_libc_22_0_0_git15IntegerToStringIiNS_7details3FmtILh10ELb0ELb0ELb0ELm1EEEEC2Ei.exit.i.i.i
  %606 = phi ptr [ %604, %Flow8513 ], [ addrspacecast (ptr addrspace(4) @_ZN22__llvm_libc_22_0_0_git3cpp6string14NULL_CHARACTERE to ptr), %_ZN22__llvm_libc_22_0_0_git15IntegerToStringIiNS_7details3FmtILh10ELb0ELb0ELb0ELm1EEEEC2Ei.exit.i.i.i ]
  call void @llvm.amdgcn.end.cf.i64(i64 %596)
  br label %if.end.i.i.i.i.i.i, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git3cpp6string7reserveEm.exit.i.i.i.i.i.i: ; preds = %Flow8514
  br label %loadstoreloop4959, !amdgpu.uniform !42

loadstoreloop4959:                                ; preds = %_ZN22__llvm_libc_22_0_0_git3cpp6string7reserveEm.exit.i.i.i.i.i.i, %loadstoreloop4959
  %phi.broken8631 = phi i64 [ 0, %_ZN22__llvm_libc_22_0_0_git3cpp6string7reserveEm.exit.i.i.i.i.i.i ], [ %611, %loadstoreloop4959 ]
  %607 = phi i64 [ %609, %loadstoreloop4959 ], [ 0, %_ZN22__llvm_libc_22_0_0_git3cpp6string7reserveEm.exit.i.i.i.i.i.i ]
  %608 = getelementptr i8, ptr %604, i64 %607
  store i8 0, ptr %608, align 1
  %609 = add i64 %607, 1
  %610 = icmp uge i64 %609, %writer.sroa.10.5.i.i.i.i
  %611 = call i64 @llvm.amdgcn.if.break.i64(i1 %610, i64 %phi.broken8631)
  %612 = call i1 @llvm.amdgcn.loop.i64(i64 %611)
  br i1 %612, label %Flow8513, label %loadstoreloop4959

Flow8513:                                         ; preds = %loadstoreloop4959
  %.lcssa9313 = phi i64 [ %611, %loadstoreloop4959 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa9313)
  br label %Flow8515, !amdgpu.uniform !42

if.end.i.i.i.i.i.i:                               ; preds = %Flow8515
  %cmp.not.i6.i.i.i.i.i.i = icmp ne ptr %606, addrspacecast (ptr addrspace(4) @_ZN22__llvm_libc_22_0_0_git3cpp6string14NULL_CHARACTERE to ptr)
  %613 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.not.i6.i.i.i.i.i.i)
  %614 = extractvalue { i1, i64 } %613, 0
  %615 = extractvalue { i1, i64 } %613, 1
  br i1 %614, label %if.then.i.i.i.i2.i.i.i, label %_ZN22__llvm_libc_22_0_0_git3cpp9to_stringEi.exit.i

if.then.i.i.i.i2.i.i.i:                           ; preds = %if.end.i.i.i.i.i.i
  %arrayidx.i.i.i.i3.i.i.i = getelementptr inbounds nuw i8, ptr %606, i64 %writer.sroa.10.5.i.i.i.i
  store i8 0, ptr %arrayidx.i.i.i.i3.i.i.i, align 1, !tbaa !57, !noalias !153
  br label %_ZN22__llvm_libc_22_0_0_git3cpp9to_stringEi.exit.i, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git3cpp9to_stringEi.exit.i: ; preds = %if.then.i.i.i.i2.i.i.i, %if.end.i.i.i.i.i.i
  call void @llvm.amdgcn.end.cf.i64(i64 %615)
  %616 = trunc i64 %writer.sroa.10.5.i.i.i.i to i32
  %617 = and i64 %writer.sroa.10.5.i.i.i.i, 15
  %618 = sub i64 %writer.sroa.10.5.i.i.i.i, %617
  %619 = icmp ne i64 %618, 0
  %620 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %619)
  %621 = extractvalue { i1, i64 } %620, 0
  %622 = extractvalue { i1, i64 } %620, 1
  br i1 %621, label %loop-memcpy-expansion5139.preheader, label %Flow8512

loop-memcpy-expansion5139.preheader:              ; preds = %_ZN22__llvm_libc_22_0_0_git3cpp9to_stringEi.exit.i
  %623 = sub i32 11, %616
  %scevgep7190 = getelementptr i8, ptr addrspace(5) %buffer.i.i.i2988, i32 %623
  br label %loop-memcpy-expansion5139, !amdgpu.uniform !42

Flow8512:                                         ; preds = %Flow8511, %_ZN22__llvm_libc_22_0_0_git3cpp9to_stringEi.exit.i
  call void @llvm.amdgcn.end.cf.i64(i64 %622)
  br label %loop-memcpy-residual-header5142, !amdgpu.uniform !42

loop-memcpy-expansion5139:                        ; preds = %loop-memcpy-expansion5139.preheader, %loop-memcpy-expansion5139
  %phi.broken8632 = phi i64 [ 0, %loop-memcpy-expansion5139.preheader ], [ %628, %loop-memcpy-expansion5139 ]
  %lsr.iv7191 = phi ptr addrspace(5) [ %scevgep7192, %loop-memcpy-expansion5139 ], [ %scevgep7190, %loop-memcpy-expansion5139.preheader ]
  %loop-index5140 = phi i64 [ %626, %loop-memcpy-expansion5139 ], [ 0, %loop-memcpy-expansion5139.preheader ]
  %624 = load <4 x i32>, ptr addrspace(5) %lsr.iv7191, align 1
  %625 = getelementptr i8, ptr %606, i64 %loop-index5140
  store <4 x i32> %624, ptr %625, align 1
  %626 = add i64 %loop-index5140, 16
  %scevgep7192 = getelementptr i8, ptr addrspace(5) %lsr.iv7191, i32 16
  %627 = icmp uge i64 %626, %618
  %628 = call i64 @llvm.amdgcn.if.break.i64(i1 %627, i64 %phi.broken8632)
  %629 = call i1 @llvm.amdgcn.loop.i64(i64 %628)
  br i1 %629, label %Flow8511, label %loop-memcpy-expansion5139

Flow8510:                                         ; preds = %Flow8509, %loop-memcpy-residual-header5142
  call void @llvm.amdgcn.end.cf.i64(i64 %6408)
  br label %post-loop-memcpy-expansion5138, !amdgpu.uniform !42

loop-memcpy-residual5141:                         ; preds = %loop-memcpy-residual5141.preheader, %loop-memcpy-residual5141
  %phi.broken8633 = phi i64 [ 0, %loop-memcpy-residual5141.preheader ], [ %633, %loop-memcpy-residual5141 ]
  %lsr.iv7186 = phi ptr addrspace(5) [ %scevgep7187, %loop-memcpy-residual5141 ], [ %scevgep7185, %loop-memcpy-residual5141.preheader ]
  %residual-loop-index5143 = phi i64 [ %631, %loop-memcpy-residual5141 ], [ 0, %loop-memcpy-residual5141.preheader ]
  %630 = load i8, ptr addrspace(5) %lsr.iv7186, align 1
  %scevgep7189 = getelementptr i8, ptr %scevgep7188, i64 %residual-loop-index5143
  store i8 %630, ptr %scevgep7189, align 1
  %631 = add i64 %residual-loop-index5143, 1
  %scevgep7187 = getelementptr i8, ptr addrspace(5) %lsr.iv7186, i32 1
  %632 = icmp uge i64 %631, %617
  %633 = call i64 @llvm.amdgcn.if.break.i64(i1 %632, i64 %phi.broken8633)
  %634 = call i1 @llvm.amdgcn.loop.i64(i64 %633)
  br i1 %634, label %Flow8509, label %loop-memcpy-residual5141

Flow8509:                                         ; preds = %loop-memcpy-residual5141
  %.lcssa9311 = phi i64 [ %633, %loop-memcpy-residual5141 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa9311)
  br label %Flow8510, !amdgpu.uniform !42

post-loop-memcpy-expansion5138:                   ; preds = %Flow8510
  call void @llvm.lifetime.end.p5(ptr addrspace(5) %buffer.i.i.i2988) #34, !noalias !153
  br label %for.cond.i.i.i327, !amdgpu.uniform !42

for.cond.i.i.i327:                                ; preds = %post-loop-memcpy-expansion5138, %Flow8508
  %635 = phi i32 [ %673, %Flow8508 ], [ poison, %post-loop-memcpy-expansion5138 ]
  %636 = phi i64 [ %674, %Flow8508 ], [ poison, %post-loop-memcpy-expansion5138 ]
  %637 = phi i32 [ %675, %Flow8508 ], [ poison, %post-loop-memcpy-expansion5138 ]
  %retval.sroa.6.0.i.i.i328 = phi i32 [ %678, %Flow8508 ], [ undef, %post-loop-memcpy-expansion5138 ]
  %retval.sroa.2.0.i.i.i330 = phi i64 [ %677, %Flow8508 ], [ undef, %post-loop-memcpy-expansion5138 ]
  %retval.sroa.8.0.i.i.i331 = phi i32 [ %676, %Flow8508 ], [ undef, %post-loop-memcpy-expansion5138 ]
  %index.0.i.i.i332 = phi i32 [ %680, %Flow8508 ], [ 0, %post-loop-memcpy-expansion5138 ]
  %638 = load i32, ptr addrspace(1) @__llvm_rpc_client, align 8, !tbaa !68
  %cmp.not.i.i.i333 = icmp ult i32 %index.0.i.i.i332, %638
  %spec.store.select.i.i.i334 = select i1 %cmp.not.i.i.i333, i32 %index.0.i.i.i332, i32 0
  %639 = tail call noundef i64 @llvm.amdgcn.ballot.i64(i1 true)
  %640 = and i32 %spec.store.select.i.i.i334, 31
  %641 = lshr i64 %639, %sh_prom.i.i.i.i
  %642 = trunc i64 %641 to i32
  %conv4.i.i.i.i.i338 = and i32 %642, 1
  %shl.i.i.i.i.i339 = shl nuw i32 %conv4.i.i.i.i.i338, %640
  %643 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop5515, !amdgpu.uniform !42

644:                                              ; preds = %ComputeEnd5516
  %sunkaddr7374 = getelementptr inbounds i8, ptr addrspace(1) %6914, i64 40
  %645 = atomicrmw or ptr addrspace(1) %sunkaddr7374, i32 %.lcssa9305 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %646, !amdgpu.uniform !42

646:                                              ; preds = %644, %ComputeEnd5516
  %647 = phi i32 [ %645, %644 ], [ poison, %ComputeEnd5516 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %6923)
  %648 = call i32 @llvm.amdgcn.readfirstlane.i32(i32 %647)
  %649 = or i32 %648, %.lcssa9306
  %shl5.i.i.i.i.i340 = shl nuw i32 1, %640
  %and.i.i.i.i.i341 = and i32 %shl5.i.i.i.i.i340, %649
  %tobool3.i.i.i.i342 = icmp ne i32 %and.i.i.i.i.i341, 0
  %650 = tail call i64 @llvm.amdgcn.ballot.i64(i1 %tobool3.i.i.i.i342)
  %and.i.i.i.i.i.i343 = and i64 %650, %639
  %cmp.i.not.i.i.i344 = icmp ne i64 %639, %and.i.i.i.i.i.i343
  br i1 %cmp.i.not.i.i.i344, label %if.then.i.i.i.i345, label %Flow8507, !amdgpu.uniform !42

if.then.i.i.i.i345:                               ; preds = %646
  fence syncscope("agent") acquire
  %651 = load <2 x i64>, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !75
  %652 = extractelement <2 x i64> %651, i32 0
  %653 = inttoptr i64 %652 to ptr
  %654 = extractelement <2 x i64> %651, i32 1
  %655 = inttoptr i64 %654 to ptr
  %idxprom.i.i.i.i346 = zext i32 %spec.store.select.i.i.i334 to i64
  %arrayidx.i.i.i.i347 = getelementptr inbounds nuw i32, ptr %653, i64 %idxprom.i.i.i.i346, !amdgpu.uniform !42
  %656 = load atomic i32, ptr %arrayidx.i.i.i.i347 monotonic, align 4
  %657 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %656)
  %arrayidx.i45.i.i.i348 = getelementptr inbounds nuw i32, ptr %655, i64 %idxprom.i.i.i.i346, !amdgpu.uniform !42
  %658 = load atomic i32, ptr %arrayidx.i45.i.i.i348 monotonic, align 4
  %659 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %658)
  %cmp.i46.not.i.i.i349 = icmp ne i32 %657, %659
  br i1 %cmp.i46.not.i.i.i349, label %if.then12.i.i.i350, label %Flow8505, !amdgpu.uniform !42

if.then12.i.i.i350:                               ; preds = %if.then.i.i.i.i345
  fence syncscope("agent") release
  %660 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %639, i1 true)
  %iszero.i.i.i.i.i.i.i351 = icmp ne i64 %639, 0
  %cmp2.i.i.i.i.i.i352 = icmp eq i64 %660, %sh_prom.i.i.i.i
  %cmp.i.i.i.i.i2.i = select i1 %iszero.i.i.i.i.i.i.i351, i1 %cmp2.i.i.i.i.i.i352, i1 false
  %conv4.i.i51.i.i.i353 = zext i1 %cmp.i.i.i.i.i2.i to i32
  %shl.i.i52.i.i.i354 = shl nuw i32 %conv4.i.i51.i.i.i353, %640
  %xor.i.i.i.i.i355 = xor i32 %shl.i.i52.i.i.i354, -1
  %661 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop5520, !amdgpu.uniform !42

662:                                              ; preds = %ComputeEnd5521
  %sunkaddr7375 = getelementptr inbounds i8, ptr addrspace(1) %6914, i64 40
  %663 = atomicrmw and ptr addrspace(1) %sunkaddr7375, i32 %.lcssa9307 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %664, !amdgpu.uniform !42

664:                                              ; preds = %662, %ComputeEnd5521
  call void @llvm.amdgcn.end.cf.i64(i64 %6941)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %Flow8505, !amdgpu.uniform !42

if.end14.i.i.i413:                                ; preds = %Flow8505
  %665 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %639, i1 true)
  %iszero.i.i.i.i.i.i414 = icmp ne i64 %639, 0
  %cmp2.i.i.i.i.i415 = icmp eq i64 %665, %sh_prom.i.i.i.i
  %cmp.i.i.i.i.i416 = select i1 %iszero.i.i.i.i.i.i414, i1 %cmp2.i.i.i.i.i415, i1 false
  %666 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.i.i.i.i.i416)
  %667 = extractvalue { i1, i64 } %666, 0
  %668 = extractvalue { i1, i64 } %666, 1
  br i1 %667, label %if.then16.i.i.i418, label %if.end22.i.i.i417

if.then16.i.i.i418:                               ; preds = %if.end14.i.i.i413
  %669 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !76
  %opcode.i.i.i419 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %669, i64 %idxprom.i.i.i.i346, i32 1
  store i32 1660944387, ptr %opcode.i.i.i419, align 8, !tbaa !77
  %arrayidx21.i.i.i420 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %669, i64 %idxprom.i.i.i.i346
  store i64 %639, ptr %arrayidx21.i.i.i420, align 8, !tbaa !80
  br label %if.end22.i.i.i417, !amdgpu.uniform !42

Flow8506:                                         ; preds = %if.end22.i.i.i417, %Flow8505
  %670 = phi i32 [ %657, %if.end22.i.i.i417 ], [ %retval.sroa.8.0.i.i.i331, %Flow8505 ]
  %671 = phi i64 [ %639, %if.end22.i.i.i417 ], [ %retval.sroa.2.0.i.i.i330, %Flow8505 ]
  %672 = phi i32 [ %spec.store.select.i.i.i334, %if.end22.i.i.i417 ], [ %retval.sroa.6.0.i.i.i328, %Flow8505 ]
  br label %cleanup26.i.i.i356, !amdgpu.uniform !42

if.end22.i.i.i417:                                ; preds = %if.then16.i.i.i418, %if.end14.i.i.i413
  call void @llvm.amdgcn.end.cf.i64(i64 %668)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %Flow8506, !amdgpu.uniform !42

Flow8507:                                         ; preds = %cleanup26.i.i.i356, %646
  %673 = phi i32 [ %672, %cleanup26.i.i.i356 ], [ %635, %646 ]
  %674 = phi i64 [ %671, %cleanup26.i.i.i356 ], [ %636, %646 ]
  %675 = phi i32 [ %670, %cleanup26.i.i.i356 ], [ %637, %646 ]
  %676 = phi i32 [ %670, %cleanup26.i.i.i356 ], [ %retval.sroa.8.0.i.i.i331, %646 ]
  %677 = phi i64 [ %671, %cleanup26.i.i.i356 ], [ %retval.sroa.2.0.i.i.i330, %646 ]
  %678 = phi i32 [ %672, %cleanup26.i.i.i356 ], [ %retval.sroa.6.0.i.i.i328, %646 ]
  %679 = phi i1 [ %cmp.i46.not.i.i.i349, %cleanup26.i.i.i356 ], [ true, %646 ]
  br i1 %679, label %for.inc.i.i.i361, label %Flow8508, !amdgpu.uniform !42

cleanup26.i.i.i356:                               ; preds = %Flow8506
  br label %Flow8507, !amdgpu.uniform !42

for.inc.i.i.i361:                                 ; preds = %Flow8507
  %inc.i.i.i366 = add i32 %spec.store.select.i.i.i334, 1
  br label %Flow8508, !amdgpu.uniform !42

Flow8508:                                         ; preds = %for.inc.i.i.i361, %Flow8507
  %680 = phi i32 [ %inc.i.i.i366, %for.inc.i.i.i361 ], [ poison, %Flow8507 ]
  %681 = phi i1 [ false, %for.inc.i.i.i361 ], [ true, %Flow8507 ]
  br i1 %681, label %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i367, label %for.cond.i.i.i327, !amdgpu.uniform !42

_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i367: ; preds = %Flow8508
  %.lcssa9310 = phi i32 [ %673, %Flow8508 ]
  %.lcssa9309 = phi i64 [ %674, %Flow8508 ]
  %.lcssa9308 = phi i32 [ %675, %Flow8508 ]
  fence acquire
  %682 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %mul.i.i.i.i.i.i368 = shl i32 %.lcssa9310, 6
  %idxprom.i19.i.i.i.i.i369 = zext i32 %mul.i.i.i.i.i.i368 to i64
  %arrayidx.i20.i.i.i.i.i370 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %682, i64 %idxprom.i19.i.i.i.i.i369
  %arrayidx.i22.i.i.i.i.i371 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %arrayidx.i20.i.i.i.i.i370, i64 %sh_prom.i.i.i.i
  store i64 %writer.sroa.10.5.i.i.i.i, ptr %arrayidx.i22.i.i.i.i.i371, align 8, !tbaa !82
  %spec.select.i.i.i.i.i.i.i372 = tail call i64 @llvm.umin.i64(i64 %writer.sroa.10.5.i.i.i.i, i64 56)
  %683 = and i64 %spec.select.i.i.i.i.i.i.i372, 15
  %684 = sub i64 %spec.select.i.i.i.i.i.i.i372, %683
  %685 = icmp ne i64 %684, 0
  %686 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %685)
  %687 = extractvalue { i1, i64 } %686, 0
  %688 = extractvalue { i1, i64 } %686, 1
  br i1 %687, label %loop-memcpy-expansion5085.preheader, label %Flow8504

loop-memcpy-expansion5085.preheader:              ; preds = %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i367
  %689 = add nuw nsw i64 %78, 8
  %scevgep7182 = getelementptr i8, ptr %682, i64 %689
  %690 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i369, 6
  %scevgep7183 = getelementptr i8, ptr %scevgep7182, i64 %690
  br label %loop-memcpy-expansion5085, !amdgpu.uniform !42

Flow8504:                                         ; preds = %Flow8503, %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i367
  call void @llvm.amdgcn.end.cf.i64(i64 %688)
  br label %loop-memcpy-residual-header5088, !amdgpu.uniform !42

loop-memcpy-expansion5085:                        ; preds = %loop-memcpy-expansion5085.preheader, %loop-memcpy-expansion5085
  %phi.broken8634 = phi i64 [ 0, %loop-memcpy-expansion5085.preheader ], [ %695, %loop-memcpy-expansion5085 ]
  %loop-index5086 = phi i64 [ %693, %loop-memcpy-expansion5085 ], [ 0, %loop-memcpy-expansion5085.preheader ]
  %691 = getelementptr i8, ptr %606, i64 %loop-index5086
  %692 = load <4 x i32>, ptr %691, align 1
  %scevgep7184 = getelementptr i8, ptr %scevgep7183, i64 %loop-index5086
  store <4 x i32> %692, ptr %scevgep7184, align 1
  %693 = add i64 %loop-index5086, 16
  %694 = icmp uge i64 %693, %684
  %695 = call i64 @llvm.amdgcn.if.break.i64(i1 %694, i64 %phi.broken8634)
  %696 = call i1 @llvm.amdgcn.loop.i64(i64 %695)
  br i1 %696, label %Flow8503, label %loop-memcpy-expansion5085

Flow8502:                                         ; preds = %Flow8500, %Flow8501
  %697 = phi i64 [ %6351, %Flow8500 ], [ %6343, %Flow8501 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %6346)
  br label %post-loop-memcpy-expansion5084, !amdgpu.uniform !42

loop-memcpy-residual5087:                         ; preds = %loop-memcpy-residual5087.preheader, %loop-memcpy-residual5087
  %phi.broken8635 = phi i64 [ 0, %loop-memcpy-residual5087.preheader ], [ %701, %loop-memcpy-residual5087 ]
  %residual-loop-index5089 = phi i64 [ %699, %loop-memcpy-residual5087 ], [ 0, %loop-memcpy-residual5087.preheader ]
  %scevgep7181 = getelementptr i8, ptr %scevgep7180, i64 %residual-loop-index5089
  %698 = load i8, ptr %scevgep7181, align 1
  %scevgep7179 = getelementptr i8, ptr %scevgep7178, i64 %residual-loop-index5089
  store i8 %698, ptr %scevgep7179, align 1
  %699 = add i64 %residual-loop-index5089, 1
  %700 = icmp uge i64 %699, %683
  %701 = call i64 @llvm.amdgcn.if.break.i64(i1 %700, i64 %phi.broken8635)
  %702 = call i1 @llvm.amdgcn.loop.i64(i64 %701)
  br i1 %702, label %Flow8500, label %loop-memcpy-residual5087

Flow8500:                                         ; preds = %loop-memcpy-residual5087
  %.lcssa9303 = phi i64 [ %701, %loop-memcpy-residual5087 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa9303)
  br label %Flow8502, !amdgpu.uniform !42

post-loop-memcpy-expansion5084:                   ; preds = %Flow8502
  %tobool.not.i.i.i.i.i.i374 = icmp eq i32 %.lcssa9308, 0
  %conv.i.i.i.i.i.i375 = zext i1 %tobool.not.i.i.i.i.i.i374 to i32
  fence release
  %703 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %idxprom.i23.i.i.i.i.i376 = zext i32 %.lcssa9310 to i64
  %arrayidx.i24.i.i.i.i.i377 = getelementptr inbounds nuw i32, ptr %703, i64 %idxprom.i23.i.i.i.i.i376
  store atomic i32 %conv.i.i.i.i.i.i375, ptr %arrayidx.i24.i.i.i.i.i377 monotonic, align 4
  %704 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !84
  %arrayidx.i.i5.i.i378 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %704, i64 %idxprom.i23.i.i.i.i.i376, !amdgpu.uniform !42
  %705 = load i64, ptr %arrayidx.i.i5.i.i378, align 8, !tbaa !80
  %scevgep7164 = getelementptr i8, ptr %606, i64 56
  %706 = add i64 %writer.sroa.10.5.i.i.i.i, -56
  br label %while.cond.i.i.i.i379, !amdgpu.uniform !42

while.cond.i.i.i.i379:                            ; preds = %post-loop-memcpy-expansion5084, %Flow8499
  %phi.broken8638 = phi i64 [ 0, %post-loop-memcpy-expansion5084 ], [ %723, %Flow8499 ]
  %lsr.iv7170 = phi i64 [ %721, %Flow8499 ], [ %706, %post-loop-memcpy-expansion5084 ]
  %lsr.iv7165 = phi ptr [ %720, %Flow8499 ], [ %scevgep7164, %post-loop-memcpy-expansion5084 ]
  %port.sroa.37.0.i.i380 = phi i32 [ %719, %Flow8499 ], [ %conv.i.i.i.i.i.i375, %post-loop-memcpy-expansion5084 ]
  %idx.0.i.i.i.i381 = phi i64 [ %718, %Flow8499 ], [ 56, %post-loop-memcpy-expansion5084 ]
  %cmp.i.i.i.i382 = icmp ugt i64 %writer.sroa.10.5.i.i.i.i, %idx.0.i.i.i.i381
  %707 = tail call i64 @llvm.amdgcn.ballot.i64(i1 %cmp.i.i.i.i382)
  %and.i.i.i.i6.i.i383 = and i64 %707, %705
  %tobool.not.i.i.i.i384 = icmp ne i64 %and.i.i.i.i6.i.i383, 0
  %708 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i22.i.i.i385 = getelementptr inbounds nuw i32, ptr %708, i64 %idxprom.i23.i.i.i.i.i376, !amdgpu.uniform !42
  %709 = load atomic i32, ptr %arrayidx.i22.i.i.i385 monotonic, align 4
  %710 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %tobool.not.i.i.i.i384)
  %711 = extractvalue { i1, i64 } %710, 0
  %712 = extractvalue { i1, i64 } %710, 1
  br i1 %711, label %cond.false.i8.i.i.i.i386, label %Flow8499

cond.false.i8.i.i.i.i386:                         ; preds = %while.cond.i.i.i.i379
  %713 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %709)
  %714 = icmp ne i32 %713, %port.sroa.37.0.i.i380
  br i1 %714, label %while.body.i.i21.i.i.i.i387.preheader, label %Flow8498, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i387.preheader:            ; preds = %cond.false.i8.i.i.i.i386
  br label %while.body.i.i21.i.i.i.i387, !amdgpu.uniform !42

Flow8498:                                         ; preds = %Flow8497, %cond.false.i8.i.i.i.i386
  br label %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i390, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i387:                      ; preds = %while.body.i.i21.i.i.i.i387.preheader, %while.body.i.i21.i.i.i.i387
  tail call void @llvm.amdgcn.s.sleep(i32 2)
  %715 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i22.i.i.i.i388 = getelementptr inbounds nuw i32, ptr %715, i64 %idxprom.i23.i.i.i.i.i376, !amdgpu.uniform !42
  %716 = load atomic i32, ptr %arrayidx.i.i.i22.i.i.i.i388 monotonic, align 4
  %717 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %716)
  %cmp.i.not.i.i23.i.i.i.i389 = icmp eq i32 %717, %port.sroa.37.0.i.i380
  br i1 %cmp.i.not.i.i23.i.i.i.i389, label %Flow8497, label %while.body.i.i21.i.i.i.i387, !amdgpu.uniform !42

Flow8497:                                         ; preds = %while.body.i.i21.i.i.i.i387
  br label %Flow8498, !amdgpu.uniform !42

Flow8499:                                         ; preds = %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i391, %while.cond.i.i.i.i379
  %718 = phi i64 [ %add.i.i.i.i394, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i391 ], [ poison, %while.cond.i.i.i.i379 ]
  %719 = phi i32 [ %conv.i.i26.i.i.i.i392, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i391 ], [ poison, %while.cond.i.i.i.i379 ]
  %720 = phi ptr [ %scevgep7166, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i391 ], [ poison, %while.cond.i.i.i.i379 ]
  %721 = phi i64 [ %lsr.iv.next7171, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i391 ], [ poison, %while.cond.i.i.i.i379 ]
  %722 = phi i1 [ false, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i391 ], [ true, %while.cond.i.i.i.i379 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %712)
  %723 = call i64 @llvm.amdgcn.if.break.i64(i1 %722, i64 %phi.broken8638)
  %724 = call i1 @llvm.amdgcn.loop.i64(i64 %723)
  br i1 %724, label %cond.false.i.i.i400, label %while.cond.i.i.i.i379

_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i390: ; preds = %Flow8498
  fence acquire
  %725 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.i.i.i.i382)
  %726 = extractvalue { i1, i64 } %725, 0
  %727 = extractvalue { i1, i64 } %725, 1
  br i1 %726, label %if.then.i.i.i.i.i.i3.i, label %Flow8496

if.then.i.i.i.i.i.i3.i:                           ; preds = %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i390
  %sub.i.i.i.i.i.i.i396 = sub nuw i64 %writer.sroa.10.5.i.i.i.i, %idx.0.i.i.i.i381
  %spec.select.i.i.i36.i.i.i.i397 = tail call i64 @llvm.umin.i64(i64 %sub.i.i.i.i.i.i.i396, i64 64)
  %728 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %729 = and i64 %spec.select.i.i.i36.i.i.i.i397, 15
  %730 = sub i64 %spec.select.i.i.i36.i.i.i.i397, %729
  %731 = icmp ne i64 %730, 0
  %732 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %731)
  %733 = extractvalue { i1, i64 } %732, 0
  %734 = extractvalue { i1, i64 } %732, 1
  br i1 %733, label %loop-memcpy-expansion5079.preheader, label %Flow8495

loop-memcpy-expansion5079.preheader:              ; preds = %if.then.i.i.i.i.i.i3.i
  %arrayidx.i20.i35.i.i.i.i398 = getelementptr %"struct.rpc::Buffer", ptr %728, i64 %idxprom.i19.i.i.i.i.i369
  %arrayidx.i22.i38.i.i.i.i399 = getelementptr %"struct.rpc::Buffer", ptr %arrayidx.i20.i35.i.i.i.i398, i64 %sh_prom.i.i.i.i
  br label %loop-memcpy-expansion5079, !amdgpu.uniform !42

Flow8495:                                         ; preds = %Flow8494, %if.then.i.i.i.i.i.i3.i
  call void @llvm.amdgcn.end.cf.i64(i64 %734)
  br label %loop-memcpy-residual-header5082, !amdgpu.uniform !42

loop-memcpy-expansion5079:                        ; preds = %loop-memcpy-expansion5079.preheader, %loop-memcpy-expansion5079
  %phi.broken8636 = phi i64 [ 0, %loop-memcpy-expansion5079.preheader ], [ %739, %loop-memcpy-expansion5079 ]
  %loop-index5080 = phi i64 [ %737, %loop-memcpy-expansion5079 ], [ 0, %loop-memcpy-expansion5079.preheader ]
  %scevgep7167 = getelementptr i8, ptr %lsr.iv7165, i64 %loop-index5080
  %735 = load <4 x i32>, ptr %scevgep7167, align 1
  %736 = getelementptr i8, ptr %arrayidx.i22.i38.i.i.i.i399, i64 %loop-index5080
  store <4 x i32> %735, ptr %736, align 1
  %737 = add i64 %loop-index5080, 16
  %738 = icmp uge i64 %737, %730
  %739 = call i64 @llvm.amdgcn.if.break.i64(i1 %738, i64 %phi.broken8636)
  %740 = call i1 @llvm.amdgcn.loop.i64(i64 %739)
  br i1 %740, label %Flow8494, label %loop-memcpy-expansion5079

Flow8493:                                         ; preds = %Flow8492, %loop-memcpy-residual-header5082
  call void @llvm.amdgcn.end.cf.i64(i64 %6336)
  br label %Flow8496, !amdgpu.uniform !42

loop-memcpy-residual5081:                         ; preds = %loop-memcpy-residual5081.preheader, %loop-memcpy-residual5081
  %phi.broken8637 = phi i64 [ 0, %loop-memcpy-residual5081.preheader ], [ %744, %loop-memcpy-residual5081 ]
  %residual-loop-index5083 = phi i64 [ %742, %loop-memcpy-residual5081 ], [ 0, %loop-memcpy-residual5081.preheader ]
  %scevgep7176 = getelementptr i8, ptr %scevgep7175, i64 %residual-loop-index5083
  %741 = load i8, ptr %scevgep7176, align 1
  %scevgep7174 = getelementptr i8, ptr %scevgep7173, i64 %residual-loop-index5083
  store i8 %741, ptr %scevgep7174, align 1
  %742 = add i64 %residual-loop-index5083, 1
  %743 = icmp uge i64 %742, %729
  %744 = call i64 @llvm.amdgcn.if.break.i64(i1 %743, i64 %phi.broken8637)
  %745 = call i1 @llvm.amdgcn.loop.i64(i64 %744)
  br i1 %745, label %Flow8492, label %loop-memcpy-residual5081

Flow8492:                                         ; preds = %loop-memcpy-residual5081
  %.lcssa9300 = phi i64 [ %744, %loop-memcpy-residual5081 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa9300)
  br label %Flow8493, !amdgpu.uniform !42

_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i391: ; preds = %Flow8496
  %conv.i.i26.i.i.i.i392 = xor i32 %port.sroa.37.0.i.i380, 1
  fence release
  %746 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %arrayidx.i24.i29.i.i.i.i393 = getelementptr inbounds nuw i32, ptr %746, i64 %idxprom.i23.i.i.i.i.i376
  store atomic i32 %conv.i.i26.i.i.i.i392, ptr %arrayidx.i24.i29.i.i.i.i393 monotonic, align 4
  %add.i.i.i.i394 = add i64 %idx.0.i.i.i.i381, 64
  %scevgep7166 = getelementptr i8, ptr %lsr.iv7165, i64 64
  %lsr.iv.next7171 = add i64 %lsr.iv7170, -64
  br label %Flow8499, !amdgpu.uniform !42

cond.false.i.i.i400:                              ; preds = %Flow8499
  %.lcssa9302 = phi i64 [ %723, %Flow8499 ]
  %port.sroa.37.0.i.i380.lcssa = phi i32 [ %port.sroa.37.0.i.i380, %Flow8499 ]
  %.lcssa9301 = phi i32 [ %709, %Flow8499 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa9302)
  %747 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %.lcssa9301)
  %748 = icmp ne i32 %747, %port.sroa.37.0.i.i380.lcssa
  %749 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %748)
  %750 = extractvalue { i1, i64 } %749, 0
  %751 = extractvalue { i1, i64 } %749, 1
  br i1 %750, label %while.body.i.i.i.i401.preheader, label %Flow8491

while.body.i.i.i.i401.preheader:                  ; preds = %cond.false.i.i.i400
  br label %while.body.i.i.i.i401, !amdgpu.uniform !42

Flow8491:                                         ; preds = %Flow8490, %cond.false.i.i.i400
  call void @llvm.amdgcn.end.cf.i64(i64 %751)
  br label %_ZN22__llvm_libc_22_0_0_git15write_to_stderrENS_3cpp11string_viewE.exit.i, !amdgpu.uniform !42

while.body.i.i.i.i401:                            ; preds = %while.body.i.i.i.i401.preheader, %while.body.i.i.i.i401
  %phi.broken8639 = phi i64 [ 0, %while.body.i.i.i.i401.preheader ], [ %755, %while.body.i.i.i.i401 ]
  tail call void @llvm.amdgcn.s.sleep(i32 2)
  %752 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i8.i.i402 = getelementptr inbounds nuw i32, ptr %752, i64 %idxprom.i23.i.i.i.i.i376, !amdgpu.uniform !42
  %753 = load atomic i32, ptr %arrayidx.i.i.i8.i.i402 monotonic, align 4
  %754 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %753)
  %cmp.i.not.i.i.i.i403 = icmp eq i32 %754, %port.sroa.37.0.i.i380.lcssa
  %755 = call i64 @llvm.amdgcn.if.break.i64(i1 %cmp.i.not.i.i.i.i403, i64 %phi.broken8639)
  %756 = call i1 @llvm.amdgcn.loop.i64(i64 %755)
  br i1 %756, label %Flow8490, label %while.body.i.i.i.i401

Flow8490:                                         ; preds = %while.body.i.i.i.i401
  %.lcssa9298 = phi i64 [ %755, %while.body.i.i.i.i401 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa9298)
  br label %Flow8491, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git15write_to_stderrENS_3cpp11string_viewE.exit.i: ; preds = %Flow8491
  fence acquire
  tail call void @llvm.amdgcn.wave.barrier()
  fence syncscope("agent") release
  %757 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %.lcssa9309, i1 true)
  %iszero.i.i.i.i.i13.i.i404 = icmp ne i64 %.lcssa9309, 0
  %cmp2.i.i.i.i14.i.i405 = icmp eq i64 %757, %sh_prom.i.i.i.i
  %cmp.i.i.i.i15.i.i406 = select i1 %iszero.i.i.i.i.i13.i.i404, i1 %cmp2.i.i.i.i14.i.i405, i1 false
  %758 = and i32 %.lcssa9310, 31
  %conv4.i.i.i19.i.i410 = zext i1 %cmp.i.i.i.i15.i.i406 to i32
  %shl.i.i.i20.i.i411 = shl nuw i32 %conv4.i.i.i19.i.i410, %758
  %xor.i.i.i21.i.i412 = xor i32 %shl.i.i.i20.i.i411, -1
  %759 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop5524, !amdgpu.uniform !42

760:                                              ; preds = %ComputeEnd5525
  %div8.i.i.i16.i.i407 = lshr i32 %.lcssa9310, 5
  %idxprom.i.i.i17.i.i408 = zext nneg i32 %div8.i.i.i16.i.i407 to i64
  %761 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i17.i.i408
  %sunkaddr7376 = getelementptr inbounds i8, ptr addrspace(1) %761, i64 40
  %762 = atomicrmw and ptr addrspace(1) %sunkaddr7376, i32 %.lcssa9297 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %763, !amdgpu.uniform !42

763:                                              ; preds = %760, %ComputeEnd5525
  call void @llvm.amdgcn.end.cf.i64(i64 %6958)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %for.cond.i.i1.i429, !amdgpu.uniform !42

for.cond.i.i1.i429:                               ; preds = %763, %Flow8489
  %764 = phi i32 [ %802, %Flow8489 ], [ poison, %763 ]
  %765 = phi i64 [ %803, %Flow8489 ], [ poison, %763 ]
  %766 = phi i32 [ %804, %Flow8489 ], [ poison, %763 ]
  %retval.sroa.6.0.i.i.i430 = phi i32 [ %807, %Flow8489 ], [ undef, %763 ]
  %retval.sroa.2.0.i.i.i432 = phi i64 [ %806, %Flow8489 ], [ undef, %763 ]
  %retval.sroa.8.0.i.i.i433 = phi i32 [ %805, %Flow8489 ], [ undef, %763 ]
  %index.0.i.i.i434 = phi i32 [ %809, %Flow8489 ], [ 0, %763 ]
  %767 = load i32, ptr addrspace(1) @__llvm_rpc_client, align 8, !tbaa !68
  %cmp.not.i.i.i435 = icmp ult i32 %index.0.i.i.i434, %767
  %spec.store.select.i.i.i436 = select i1 %cmp.not.i.i.i435, i32 %index.0.i.i.i434, i32 0
  %768 = tail call noundef i64 @llvm.amdgcn.ballot.i64(i1 true)
  %769 = and i32 %spec.store.select.i.i.i436, 31
  %770 = lshr i64 %768, %sh_prom.i.i.i.i
  %771 = trunc i64 %770 to i32
  %conv4.i.i.i.i.i440 = and i32 %771, 1
  %shl.i.i.i.i.i441 = shl nuw i32 %conv4.i.i.i.i.i440, %769
  %772 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop5528, !amdgpu.uniform !42

773:                                              ; preds = %ComputeEnd5529
  %sunkaddr7377 = getelementptr inbounds i8, ptr addrspace(1) %6968, i64 40
  %774 = atomicrmw or ptr addrspace(1) %sunkaddr7377, i32 %.lcssa9291 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %775, !amdgpu.uniform !42

775:                                              ; preds = %773, %ComputeEnd5529
  %776 = phi i32 [ %774, %773 ], [ poison, %ComputeEnd5529 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %6977)
  %777 = call i32 @llvm.amdgcn.readfirstlane.i32(i32 %776)
  %778 = or i32 %777, %.lcssa9292
  %shl5.i.i.i.i.i442 = shl nuw i32 1, %769
  %and.i.i.i.i.i443 = and i32 %shl5.i.i.i.i.i442, %778
  %tobool3.i.i.i.i444 = icmp ne i32 %and.i.i.i.i.i443, 0
  %779 = tail call i64 @llvm.amdgcn.ballot.i64(i1 %tobool3.i.i.i.i444)
  %and.i.i.i.i.i.i445 = and i64 %779, %768
  %cmp.i.not.i.i.i446 = icmp ne i64 %768, %and.i.i.i.i.i.i445
  br i1 %cmp.i.not.i.i.i446, label %if.then.i.i.i.i447, label %Flow8488, !amdgpu.uniform !42

if.then.i.i.i.i447:                               ; preds = %775
  fence syncscope("agent") acquire
  %780 = load <2 x i64>, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !75
  %781 = extractelement <2 x i64> %780, i32 0
  %782 = inttoptr i64 %781 to ptr
  %783 = extractelement <2 x i64> %780, i32 1
  %784 = inttoptr i64 %783 to ptr
  %idxprom.i.i.i.i448 = zext i32 %spec.store.select.i.i.i436 to i64
  %arrayidx.i.i.i.i449 = getelementptr inbounds nuw i32, ptr %782, i64 %idxprom.i.i.i.i448, !amdgpu.uniform !42
  %785 = load atomic i32, ptr %arrayidx.i.i.i.i449 monotonic, align 4
  %786 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %785)
  %arrayidx.i45.i.i.i450 = getelementptr inbounds nuw i32, ptr %784, i64 %idxprom.i.i.i.i448, !amdgpu.uniform !42
  %787 = load atomic i32, ptr %arrayidx.i45.i.i.i450 monotonic, align 4
  %788 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %787)
  %cmp.i46.not.i.i.i451 = icmp ne i32 %786, %788
  br i1 %cmp.i46.not.i.i.i451, label %if.then12.i.i.i452, label %Flow8486, !amdgpu.uniform !42

if.then12.i.i.i452:                               ; preds = %if.then.i.i.i.i447
  fence syncscope("agent") release
  %789 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %768, i1 true)
  %iszero.i.i.i.i.i.i.i453 = icmp ne i64 %768, 0
  %cmp2.i.i.i.i.i.i454 = icmp eq i64 %789, %sh_prom.i.i.i.i
  %cmp.i.i.i.i.i.i455 = select i1 %iszero.i.i.i.i.i.i.i453, i1 %cmp2.i.i.i.i.i.i454, i1 false
  %conv4.i.i51.i.i.i456 = zext i1 %cmp.i.i.i.i.i.i455 to i32
  %shl.i.i52.i.i.i457 = shl nuw i32 %conv4.i.i51.i.i.i456, %769
  %xor.i.i.i.i.i458 = xor i32 %shl.i.i52.i.i.i457, -1
  %790 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop5533, !amdgpu.uniform !42

791:                                              ; preds = %ComputeEnd5534
  %sunkaddr7378 = getelementptr inbounds i8, ptr addrspace(1) %6968, i64 40
  %792 = atomicrmw and ptr addrspace(1) %sunkaddr7378, i32 %.lcssa9293 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %793, !amdgpu.uniform !42

793:                                              ; preds = %791, %ComputeEnd5534
  call void @llvm.amdgcn.end.cf.i64(i64 %6995)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %Flow8486, !amdgpu.uniform !42

if.end14.i.i.i518:                                ; preds = %Flow8486
  %794 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %768, i1 true)
  %iszero.i.i.i.i.i.i519 = icmp ne i64 %768, 0
  %cmp2.i.i.i.i.i520 = icmp eq i64 %794, %sh_prom.i.i.i.i
  %cmp.i.i.i.i.i521 = select i1 %iszero.i.i.i.i.i.i519, i1 %cmp2.i.i.i.i.i520, i1 false
  %795 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.i.i.i.i.i521)
  %796 = extractvalue { i1, i64 } %795, 0
  %797 = extractvalue { i1, i64 } %795, 1
  br i1 %796, label %if.then16.i.i.i523, label %if.end22.i.i.i522

if.then16.i.i.i523:                               ; preds = %if.end14.i.i.i518
  %798 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !76
  %opcode.i.i.i524 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %798, i64 %idxprom.i.i.i.i448, i32 1
  store i32 1660944387, ptr %opcode.i.i.i524, align 8, !tbaa !77
  %arrayidx21.i.i.i525 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %798, i64 %idxprom.i.i.i.i448
  store i64 %768, ptr %arrayidx21.i.i.i525, align 8, !tbaa !80
  br label %if.end22.i.i.i522, !amdgpu.uniform !42

Flow8487:                                         ; preds = %if.end22.i.i.i522, %Flow8486
  %799 = phi i32 [ %786, %if.end22.i.i.i522 ], [ %retval.sroa.8.0.i.i.i433, %Flow8486 ]
  %800 = phi i64 [ %768, %if.end22.i.i.i522 ], [ %retval.sroa.2.0.i.i.i432, %Flow8486 ]
  %801 = phi i32 [ %spec.store.select.i.i.i436, %if.end22.i.i.i522 ], [ %retval.sroa.6.0.i.i.i430, %Flow8486 ]
  br label %cleanup26.i.i.i459, !amdgpu.uniform !42

if.end22.i.i.i522:                                ; preds = %if.then16.i.i.i523, %if.end14.i.i.i518
  call void @llvm.amdgcn.end.cf.i64(i64 %797)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %Flow8487, !amdgpu.uniform !42

Flow8488:                                         ; preds = %cleanup26.i.i.i459, %775
  %802 = phi i32 [ %801, %cleanup26.i.i.i459 ], [ %764, %775 ]
  %803 = phi i64 [ %800, %cleanup26.i.i.i459 ], [ %765, %775 ]
  %804 = phi i32 [ %799, %cleanup26.i.i.i459 ], [ %766, %775 ]
  %805 = phi i32 [ %799, %cleanup26.i.i.i459 ], [ %retval.sroa.8.0.i.i.i433, %775 ]
  %806 = phi i64 [ %800, %cleanup26.i.i.i459 ], [ %retval.sroa.2.0.i.i.i432, %775 ]
  %807 = phi i32 [ %801, %cleanup26.i.i.i459 ], [ %retval.sroa.6.0.i.i.i430, %775 ]
  %808 = phi i1 [ %cmp.i46.not.i.i.i451, %cleanup26.i.i.i459 ], [ true, %775 ]
  br i1 %808, label %for.inc.i.i.i464, label %Flow8489, !amdgpu.uniform !42

cleanup26.i.i.i459:                               ; preds = %Flow8487
  br label %Flow8488, !amdgpu.uniform !42

for.inc.i.i.i464:                                 ; preds = %Flow8488
  %inc.i.i.i469 = add i32 %spec.store.select.i.i.i436, 1
  br label %Flow8489, !amdgpu.uniform !42

Flow8489:                                         ; preds = %for.inc.i.i.i464, %Flow8488
  %809 = phi i32 [ %inc.i.i.i469, %for.inc.i.i.i464 ], [ poison, %Flow8488 ]
  %810 = phi i1 [ false, %for.inc.i.i.i464 ], [ true, %Flow8488 ]
  br i1 %810, label %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i470, label %for.cond.i.i1.i429, !amdgpu.uniform !42

_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i470: ; preds = %Flow8489
  %.lcssa9296 = phi i32 [ %802, %Flow8489 ]
  %.lcssa9295 = phi i64 [ %803, %Flow8489 ]
  %.lcssa9294 = phi i32 [ %804, %Flow8489 ]
  fence acquire
  %811 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %mul.i.i.i.i.i.i471 = shl i32 %.lcssa9296, 6
  %idxprom.i19.i.i.i.i.i472 = zext i32 %mul.i.i.i.i.i.i471 to i64
  %arrayidx.i20.i.i.i.i.i473 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %811, i64 %idxprom.i19.i.i.i.i.i472
  %arrayidx.i22.i.i.i.i.i474 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %arrayidx.i20.i.i.i.i.i473, i64 %sh_prom.i.i.i.i
  store i64 sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.14.32 to ptr), i64 5) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.14.32 to ptr) to i64)), ptr %arrayidx.i22.i.i.i.i.i474, align 8, !tbaa !82
  %spec.select.i.i.i.i.i.i.i475 = tail call i64 @llvm.umin.i64(i64 sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.14.32 to ptr), i64 5) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.14.32 to ptr) to i64)), i64 56)
  %812 = and i64 %spec.select.i.i.i.i.i.i.i475, 15
  %813 = sub i64 %spec.select.i.i.i.i.i.i.i475, %812
  %814 = icmp ne i64 %813, 0
  br i1 %814, label %loop-memcpy-expansion5223.preheader, label %Flow8485, !amdgpu.uniform !42

loop-memcpy-expansion5223.preheader:              ; preds = %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i470
  %815 = add nuw nsw i64 %78, 8
  %scevgep7161 = getelementptr i8, ptr %811, i64 %815
  %816 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i472, 6
  %scevgep7162 = getelementptr i8, ptr %scevgep7161, i64 %816
  br label %loop-memcpy-expansion5223, !amdgpu.uniform !42

Flow8485:                                         ; preds = %Flow8484, %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i470
  br label %loop-memcpy-residual-header5226, !amdgpu.uniform !42

loop-memcpy-expansion5223:                        ; preds = %loop-memcpy-expansion5223.preheader, %loop-memcpy-expansion5223
  %loop-index5224 = phi i64 [ %819, %loop-memcpy-expansion5223 ], [ 0, %loop-memcpy-expansion5223.preheader ]
  %817 = getelementptr i8, ptr addrspace(4) @.str.14.32, i64 %loop-index5224, !amdgpu.uniform !42
  %818 = load <4 x i32>, ptr addrspace(4) %817, align 1
  %scevgep7163 = getelementptr i8, ptr %scevgep7162, i64 %loop-index5224
  store <4 x i32> %818, ptr %scevgep7163, align 1
  %819 = add i64 %loop-index5224, 16
  %820 = icmp uge i64 %819, %813
  br i1 %820, label %Flow8484, label %loop-memcpy-expansion5223, !amdgpu.uniform !42

Flow8483:                                         ; preds = %Flow8481, %Flow8482
  %821 = phi i64 [ %6525, %Flow8481 ], [ %6522, %Flow8482 ]
  br label %post-loop-memcpy-expansion5222, !amdgpu.uniform !42

loop-memcpy-residual5225:                         ; preds = %loop-memcpy-residual5225.preheader, %loop-memcpy-residual5225
  %residual-loop-index5227 = phi i64 [ %823, %loop-memcpy-residual5225 ], [ 0, %loop-memcpy-residual5225.preheader ]
  %scevgep7160 = getelementptr i8, ptr addrspace(4) @.str.14.32, i64 %residual-loop-index5227, !amdgpu.uniform !42
  %822 = load i8, ptr addrspace(4) %scevgep7160, align 1
  %scevgep7159 = getelementptr i8, ptr %scevgep7158, i64 %residual-loop-index5227
  store i8 %822, ptr %scevgep7159, align 1
  %823 = add i64 %residual-loop-index5227, 1
  %824 = icmp uge i64 %823, %812
  br i1 %824, label %Flow8481, label %loop-memcpy-residual5225, !amdgpu.uniform !42

Flow8481:                                         ; preds = %loop-memcpy-residual5225
  br label %Flow8483, !amdgpu.uniform !42

post-loop-memcpy-expansion5222:                   ; preds = %Flow8483
  %tobool.not.i.i.i.i.i.i477 = icmp eq i32 %.lcssa9294, 0
  %conv.i.i.i.i.i.i478 = zext i1 %tobool.not.i.i.i.i.i.i477 to i32
  fence release
  %825 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %idxprom.i23.i.i.i.i.i479 = zext i32 %.lcssa9296 to i64
  %arrayidx.i24.i.i.i.i.i480 = getelementptr inbounds nuw i32, ptr %825, i64 %idxprom.i23.i.i.i.i.i479
  store atomic i32 %conv.i.i.i.i.i.i478, ptr %arrayidx.i24.i.i.i.i.i480 monotonic, align 4
  %826 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !84
  %arrayidx.i.i5.i.i481 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %826, i64 %idxprom.i23.i.i.i.i.i479, !amdgpu.uniform !42
  %827 = load i64, ptr %arrayidx.i.i5.i.i481, align 8, !tbaa !80
  br label %while.cond.i.i.i.i482, !amdgpu.uniform !42

while.cond.i.i.i.i482:                            ; preds = %post-loop-memcpy-expansion5222, %Flow8480
  %phi.broken8640 = phi i64 [ 0, %post-loop-memcpy-expansion5222 ], [ %844, %Flow8480 ]
  %lsr.iv7150 = phi i64 [ %842, %Flow8480 ], [ -51, %post-loop-memcpy-expansion5222 ]
  %lsr.iv7145 = phi ptr addrspace(4) [ %841, %Flow8480 ], [ getelementptr (i8, ptr addrspace(4) @.str.14.32, i64 56), %post-loop-memcpy-expansion5222 ]
  %port.sroa.37.0.i.i483 = phi i32 [ %840, %Flow8480 ], [ %conv.i.i.i.i.i.i478, %post-loop-memcpy-expansion5222 ]
  %idx.0.i.i.i.i484 = phi i64 [ %839, %Flow8480 ], [ 56, %post-loop-memcpy-expansion5222 ]
  %cmp.i.i.i.i485 = icmp ult i64 %idx.0.i.i.i.i484, sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.14.32 to ptr), i64 5) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.14.32 to ptr) to i64))
  %828 = tail call i64 @llvm.amdgcn.ballot.i64(i1 %cmp.i.i.i.i485)
  %and.i.i.i.i6.i.i486 = and i64 %828, %827
  %tobool.not.i.i.i.i487 = icmp ne i64 %and.i.i.i.i6.i.i486, 0
  %829 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i22.i.i.i488 = getelementptr inbounds nuw i32, ptr %829, i64 %idxprom.i23.i.i.i.i.i479, !amdgpu.uniform !42
  %830 = load atomic i32, ptr %arrayidx.i22.i.i.i488 monotonic, align 4
  %831 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %tobool.not.i.i.i.i487)
  %832 = extractvalue { i1, i64 } %831, 0
  %833 = extractvalue { i1, i64 } %831, 1
  br i1 %832, label %cond.false.i8.i.i.i.i489, label %Flow8480

cond.false.i8.i.i.i.i489:                         ; preds = %while.cond.i.i.i.i482
  %834 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %830)
  %835 = icmp ne i32 %834, %port.sroa.37.0.i.i483
  br i1 %835, label %while.body.i.i21.i.i.i.i490.preheader, label %Flow8479, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i490.preheader:            ; preds = %cond.false.i8.i.i.i.i489
  br label %while.body.i.i21.i.i.i.i490, !amdgpu.uniform !42

Flow8479:                                         ; preds = %Flow8478, %cond.false.i8.i.i.i.i489
  br label %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i493, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i490:                      ; preds = %while.body.i.i21.i.i.i.i490.preheader, %while.body.i.i21.i.i.i.i490
  tail call void @llvm.amdgcn.s.sleep(i32 2)
  %836 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i22.i.i.i.i491 = getelementptr inbounds nuw i32, ptr %836, i64 %idxprom.i23.i.i.i.i.i479, !amdgpu.uniform !42
  %837 = load atomic i32, ptr %arrayidx.i.i.i22.i.i.i.i491 monotonic, align 4
  %838 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %837)
  %cmp.i.not.i.i23.i.i.i.i492 = icmp eq i32 %838, %port.sroa.37.0.i.i483
  br i1 %cmp.i.not.i.i23.i.i.i.i492, label %Flow8478, label %while.body.i.i21.i.i.i.i490, !amdgpu.uniform !42

Flow8478:                                         ; preds = %while.body.i.i21.i.i.i.i490
  br label %Flow8479, !amdgpu.uniform !42

Flow8480:                                         ; preds = %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i494, %while.cond.i.i.i.i482
  %839 = phi i64 [ %add.i.i.i.i497, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i494 ], [ poison, %while.cond.i.i.i.i482 ]
  %840 = phi i32 [ %conv.i.i26.i.i.i.i495, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i494 ], [ poison, %while.cond.i.i.i.i482 ]
  %841 = phi ptr addrspace(4) [ %scevgep7146, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i494 ], [ poison, %while.cond.i.i.i.i482 ]
  %842 = phi i64 [ %lsr.iv.next7151, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i494 ], [ poison, %while.cond.i.i.i.i482 ]
  %843 = phi i1 [ false, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i494 ], [ true, %while.cond.i.i.i.i482 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %833)
  %844 = call i64 @llvm.amdgcn.if.break.i64(i1 %843, i64 %phi.broken8640)
  %845 = call i1 @llvm.amdgcn.loop.i64(i64 %844)
  br i1 %845, label %cond.false.i.i.i504, label %while.cond.i.i.i.i482

_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i493: ; preds = %Flow8479
  fence acquire
  br i1 %cmp.i.i.i.i485, label %if.then.i.i.i.i.i.i.i498, label %Flow8477, !amdgpu.uniform !42

if.then.i.i.i.i.i.i.i498:                         ; preds = %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i493
  %sub.i.i.i.i.i.i.i500 = sub nuw i64 sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.14.32 to ptr), i64 5) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.14.32 to ptr) to i64)), %idx.0.i.i.i.i484
  %spec.select.i.i.i36.i.i.i.i501 = tail call i64 @llvm.umin.i64(i64 %sub.i.i.i.i.i.i.i500, i64 64)
  %846 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %847 = and i64 %spec.select.i.i.i36.i.i.i.i501, 15
  %848 = sub i64 %spec.select.i.i.i36.i.i.i.i501, %847
  %849 = icmp ne i64 %848, 0
  br i1 %849, label %loop-memcpy-expansion5229.preheader, label %Flow8476, !amdgpu.uniform !42

loop-memcpy-expansion5229.preheader:              ; preds = %if.then.i.i.i.i.i.i.i498
  %arrayidx.i20.i35.i.i.i.i502 = getelementptr %"struct.rpc::Buffer", ptr %846, i64 %idxprom.i19.i.i.i.i.i472
  %arrayidx.i22.i38.i.i.i.i503 = getelementptr %"struct.rpc::Buffer", ptr %arrayidx.i20.i35.i.i.i.i502, i64 %sh_prom.i.i.i.i
  br label %loop-memcpy-expansion5229, !amdgpu.uniform !42

Flow8476:                                         ; preds = %Flow8475, %if.then.i.i.i.i.i.i.i498
  br label %loop-memcpy-residual-header5232, !amdgpu.uniform !42

loop-memcpy-expansion5229:                        ; preds = %loop-memcpy-expansion5229.preheader, %loop-memcpy-expansion5229
  %loop-index5230 = phi i64 [ %852, %loop-memcpy-expansion5229 ], [ 0, %loop-memcpy-expansion5229.preheader ]
  %scevgep7147 = getelementptr i8, ptr addrspace(4) %lsr.iv7145, i64 %loop-index5230, !amdgpu.uniform !42
  %850 = load <4 x i32>, ptr addrspace(4) %scevgep7147, align 1
  %851 = getelementptr i8, ptr %arrayidx.i22.i38.i.i.i.i503, i64 %loop-index5230
  store <4 x i32> %850, ptr %851, align 1
  %852 = add i64 %loop-index5230, 16
  %853 = icmp uge i64 %852, %848
  br i1 %853, label %Flow8475, label %loop-memcpy-expansion5229, !amdgpu.uniform !42

Flow8474:                                         ; preds = %Flow8473, %loop-memcpy-residual-header5232
  br label %Flow8477, !amdgpu.uniform !42

loop-memcpy-residual5231:                         ; preds = %loop-memcpy-residual5231.preheader, %loop-memcpy-residual5231
  %residual-loop-index5233 = phi i64 [ %855, %loop-memcpy-residual5231 ], [ 0, %loop-memcpy-residual5231.preheader ]
  %scevgep7156 = getelementptr i8, ptr addrspace(4) %scevgep7155, i64 %residual-loop-index5233, !amdgpu.uniform !42
  %854 = load i8, ptr addrspace(4) %scevgep7156, align 1
  %scevgep7154 = getelementptr i8, ptr %scevgep7153, i64 %residual-loop-index5233
  store i8 %854, ptr %scevgep7154, align 1
  %855 = add i64 %residual-loop-index5233, 1
  %856 = icmp uge i64 %855, %847
  br i1 %856, label %Flow8473, label %loop-memcpy-residual5231, !amdgpu.uniform !42

Flow8473:                                         ; preds = %loop-memcpy-residual5231
  br label %Flow8474, !amdgpu.uniform !42

_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i494: ; preds = %Flow8477
  %conv.i.i26.i.i.i.i495 = xor i32 %port.sroa.37.0.i.i483, 1
  fence release
  %857 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %arrayidx.i24.i29.i.i.i.i496 = getelementptr inbounds nuw i32, ptr %857, i64 %idxprom.i23.i.i.i.i.i479
  store atomic i32 %conv.i.i26.i.i.i.i495, ptr %arrayidx.i24.i29.i.i.i.i496 monotonic, align 4
  %add.i.i.i.i497 = add i64 %idx.0.i.i.i.i484, 64
  %scevgep7146 = getelementptr i8, ptr addrspace(4) %lsr.iv7145, i64 64
  %lsr.iv.next7151 = add i64 %lsr.iv7150, -64
  br label %Flow8480, !amdgpu.uniform !42

cond.false.i.i.i504:                              ; preds = %Flow8480
  %.lcssa9290 = phi i64 [ %844, %Flow8480 ]
  %port.sroa.37.0.i.i483.lcssa = phi i32 [ %port.sroa.37.0.i.i483, %Flow8480 ]
  %.lcssa9289 = phi i32 [ %830, %Flow8480 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa9290)
  %858 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %.lcssa9289)
  %859 = icmp ne i32 %858, %port.sroa.37.0.i.i483.lcssa
  %860 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %859)
  %861 = extractvalue { i1, i64 } %860, 0
  %862 = extractvalue { i1, i64 } %860, 1
  br i1 %861, label %while.body.i.i.i.i505.preheader, label %Flow8472

while.body.i.i.i.i505.preheader:                  ; preds = %cond.false.i.i.i504
  br label %while.body.i.i.i.i505, !amdgpu.uniform !42

Flow8472:                                         ; preds = %Flow8471, %cond.false.i.i.i504
  call void @llvm.amdgcn.end.cf.i64(i64 %862)
  br label %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit526, !amdgpu.uniform !42

while.body.i.i.i.i505:                            ; preds = %while.body.i.i.i.i505.preheader, %while.body.i.i.i.i505
  %phi.broken8641 = phi i64 [ 0, %while.body.i.i.i.i505.preheader ], [ %866, %while.body.i.i.i.i505 ]
  tail call void @llvm.amdgcn.s.sleep(i32 2)
  %863 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i8.i.i506 = getelementptr inbounds nuw i32, ptr %863, i64 %idxprom.i23.i.i.i.i.i479, !amdgpu.uniform !42
  %864 = load atomic i32, ptr %arrayidx.i.i.i8.i.i506 monotonic, align 4
  %865 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %864)
  %cmp.i.not.i.i.i.i507 = icmp eq i32 %865, %port.sroa.37.0.i.i483.lcssa
  %866 = call i64 @llvm.amdgcn.if.break.i64(i1 %cmp.i.not.i.i.i.i507, i64 %phi.broken8641)
  %867 = call i1 @llvm.amdgcn.loop.i64(i64 %866)
  br i1 %867, label %Flow8471, label %while.body.i.i.i.i505

Flow8471:                                         ; preds = %while.body.i.i.i.i505
  %.lcssa9288 = phi i64 [ %866, %while.body.i.i.i.i505 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa9288)
  br label %Flow8472, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit526: ; preds = %Flow8472
  fence acquire
  tail call void @llvm.amdgcn.wave.barrier()
  fence syncscope("agent") release
  %868 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %.lcssa9295, i1 true)
  %iszero.i.i.i.i.i13.i.i509 = icmp ne i64 %.lcssa9295, 0
  %cmp2.i.i.i.i14.i.i510 = icmp eq i64 %868, %sh_prom.i.i.i.i
  %cmp.i.i.i.i15.i.i511 = select i1 %iszero.i.i.i.i.i13.i.i509, i1 %cmp2.i.i.i.i14.i.i510, i1 false
  %869 = and i32 %.lcssa9296, 31
  %conv4.i.i.i19.i.i515 = zext i1 %cmp.i.i.i.i15.i.i511 to i32
  %shl.i.i.i20.i.i516 = shl nuw i32 %conv4.i.i.i19.i.i515, %869
  %xor.i.i.i21.i.i517 = xor i32 %shl.i.i.i20.i.i516, -1
  %870 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop5537, !amdgpu.uniform !42

871:                                              ; preds = %ComputeEnd5538
  %div8.i.i.i16.i.i512 = lshr i32 %.lcssa9296, 5
  %idxprom.i.i.i17.i.i513 = zext nneg i32 %div8.i.i.i16.i.i512 to i64
  %872 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i17.i.i513
  %sunkaddr7379 = getelementptr inbounds i8, ptr addrspace(1) %872, i64 40
  %873 = atomicrmw and ptr addrspace(1) %sunkaddr7379, i32 %.lcssa9287 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %874, !amdgpu.uniform !42

874:                                              ; preds = %871, %ComputeEnd5538
  call void @llvm.amdgcn.end.cf.i64(i64 %7012)
  tail call void @llvm.amdgcn.wave.barrier()
  %cmp = icmp sgt i32 %1, 1
  %875 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp)
  %876 = extractvalue { i1, i64 } %875, 0
  %877 = extractvalue { i1, i64 } %875, 1
  br i1 %876, label %for.cond.i.i1.i535.preheader, label %Flow8470

for.cond.i.i1.i535.preheader:                     ; preds = %874
  br label %for.cond.i.i1.i535, !amdgpu.uniform !42

Flow8470:                                         ; preds = %988, %874
  call void @llvm.amdgcn.end.cf.i64(i64 %877)
  br label %for.cond.i.i1.i641.preheader, !amdgpu.uniform !42

for.cond.i.i1.i535:                               ; preds = %for.cond.i.i1.i535.preheader, %Flow8469
  %878 = phi i32 [ %916, %Flow8469 ], [ poison, %for.cond.i.i1.i535.preheader ]
  %879 = phi i64 [ %917, %Flow8469 ], [ poison, %for.cond.i.i1.i535.preheader ]
  %880 = phi i32 [ %918, %Flow8469 ], [ poison, %for.cond.i.i1.i535.preheader ]
  %retval.sroa.6.0.i.i.i536 = phi i32 [ %921, %Flow8469 ], [ undef, %for.cond.i.i1.i535.preheader ]
  %retval.sroa.2.0.i.i.i538 = phi i64 [ %920, %Flow8469 ], [ undef, %for.cond.i.i1.i535.preheader ]
  %retval.sroa.8.0.i.i.i539 = phi i32 [ %919, %Flow8469 ], [ undef, %for.cond.i.i1.i535.preheader ]
  %index.0.i.i.i540 = phi i32 [ %923, %Flow8469 ], [ 0, %for.cond.i.i1.i535.preheader ]
  %881 = load i32, ptr addrspace(1) @__llvm_rpc_client, align 8, !tbaa !68
  %cmp.not.i.i.i541 = icmp ult i32 %index.0.i.i.i540, %881
  %spec.store.select.i.i.i542 = select i1 %cmp.not.i.i.i541, i32 %index.0.i.i.i540, i32 0
  %882 = tail call noundef i64 @llvm.amdgcn.ballot.i64(i1 true)
  %883 = and i32 %spec.store.select.i.i.i542, 31
  %884 = lshr i64 %882, %sh_prom.i.i.i.i
  %885 = trunc i64 %884 to i32
  %conv4.i.i.i.i.i546 = and i32 %885, 1
  %shl.i.i.i.i.i547 = shl nuw i32 %conv4.i.i.i.i.i546, %883
  %886 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop5541, !amdgpu.uniform !42

887:                                              ; preds = %ComputeEnd5542
  %sunkaddr7380 = getelementptr inbounds i8, ptr addrspace(1) %7022, i64 40
  %888 = atomicrmw or ptr addrspace(1) %sunkaddr7380, i32 %.lcssa9281 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %889, !amdgpu.uniform !42

889:                                              ; preds = %887, %ComputeEnd5542
  %890 = phi i32 [ %888, %887 ], [ poison, %ComputeEnd5542 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %7031)
  %891 = call i32 @llvm.amdgcn.readfirstlane.i32(i32 %890)
  %892 = or i32 %891, %.lcssa9282
  %shl5.i.i.i.i.i548 = shl nuw i32 1, %883
  %and.i.i.i.i.i549 = and i32 %shl5.i.i.i.i.i548, %892
  %tobool3.i.i.i.i550 = icmp ne i32 %and.i.i.i.i.i549, 0
  %893 = tail call i64 @llvm.amdgcn.ballot.i64(i1 %tobool3.i.i.i.i550)
  %and.i.i.i.i.i.i551 = and i64 %893, %882
  %cmp.i.not.i.i.i552 = icmp ne i64 %882, %and.i.i.i.i.i.i551
  br i1 %cmp.i.not.i.i.i552, label %if.then.i.i.i.i553, label %Flow8468, !amdgpu.uniform !42

if.then.i.i.i.i553:                               ; preds = %889
  fence syncscope("agent") acquire
  %894 = load <2 x i64>, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !75
  %895 = extractelement <2 x i64> %894, i32 0
  %896 = inttoptr i64 %895 to ptr
  %897 = extractelement <2 x i64> %894, i32 1
  %898 = inttoptr i64 %897 to ptr
  %idxprom.i.i.i.i554 = zext i32 %spec.store.select.i.i.i542 to i64
  %arrayidx.i.i.i.i555 = getelementptr inbounds nuw i32, ptr %896, i64 %idxprom.i.i.i.i554, !amdgpu.uniform !42
  %899 = load atomic i32, ptr %arrayidx.i.i.i.i555 monotonic, align 4
  %900 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %899)
  %arrayidx.i45.i.i.i556 = getelementptr inbounds nuw i32, ptr %898, i64 %idxprom.i.i.i.i554, !amdgpu.uniform !42
  %901 = load atomic i32, ptr %arrayidx.i45.i.i.i556 monotonic, align 4
  %902 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %901)
  %cmp.i46.not.i.i.i557 = icmp ne i32 %900, %902
  br i1 %cmp.i46.not.i.i.i557, label %if.then12.i.i.i558, label %Flow8466, !amdgpu.uniform !42

if.then12.i.i.i558:                               ; preds = %if.then.i.i.i.i553
  fence syncscope("agent") release
  %903 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %882, i1 true)
  %iszero.i.i.i.i.i.i.i559 = icmp ne i64 %882, 0
  %cmp2.i.i.i.i.i.i560 = icmp eq i64 %903, %sh_prom.i.i.i.i
  %cmp.i.i.i.i.i.i561 = select i1 %iszero.i.i.i.i.i.i.i559, i1 %cmp2.i.i.i.i.i.i560, i1 false
  %conv4.i.i51.i.i.i562 = zext i1 %cmp.i.i.i.i.i.i561 to i32
  %shl.i.i52.i.i.i563 = shl nuw i32 %conv4.i.i51.i.i.i562, %883
  %xor.i.i.i.i.i564 = xor i32 %shl.i.i52.i.i.i563, -1
  %904 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop5546, !amdgpu.uniform !42

905:                                              ; preds = %ComputeEnd5547
  %sunkaddr7381 = getelementptr inbounds i8, ptr addrspace(1) %7022, i64 40
  %906 = atomicrmw and ptr addrspace(1) %sunkaddr7381, i32 %.lcssa9283 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %907, !amdgpu.uniform !42

907:                                              ; preds = %905, %ComputeEnd5547
  call void @llvm.amdgcn.end.cf.i64(i64 %7049)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %Flow8466, !amdgpu.uniform !42

if.end14.i.i.i624:                                ; preds = %Flow8466
  %908 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %882, i1 true)
  %iszero.i.i.i.i.i.i625 = icmp ne i64 %882, 0
  %cmp2.i.i.i.i.i626 = icmp eq i64 %908, %sh_prom.i.i.i.i
  %cmp.i.i.i.i.i627 = select i1 %iszero.i.i.i.i.i.i625, i1 %cmp2.i.i.i.i.i626, i1 false
  %909 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.i.i.i.i.i627)
  %910 = extractvalue { i1, i64 } %909, 0
  %911 = extractvalue { i1, i64 } %909, 1
  br i1 %910, label %if.then16.i.i.i629, label %if.end22.i.i.i628

if.then16.i.i.i629:                               ; preds = %if.end14.i.i.i624
  %912 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !76
  %opcode.i.i.i630 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %912, i64 %idxprom.i.i.i.i554, i32 1
  store i32 1660944387, ptr %opcode.i.i.i630, align 8, !tbaa !77
  %arrayidx21.i.i.i631 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %912, i64 %idxprom.i.i.i.i554
  store i64 %882, ptr %arrayidx21.i.i.i631, align 8, !tbaa !80
  br label %if.end22.i.i.i628, !amdgpu.uniform !42

Flow8467:                                         ; preds = %if.end22.i.i.i628, %Flow8466
  %913 = phi i32 [ %900, %if.end22.i.i.i628 ], [ %retval.sroa.8.0.i.i.i539, %Flow8466 ]
  %914 = phi i64 [ %882, %if.end22.i.i.i628 ], [ %retval.sroa.2.0.i.i.i538, %Flow8466 ]
  %915 = phi i32 [ %spec.store.select.i.i.i542, %if.end22.i.i.i628 ], [ %retval.sroa.6.0.i.i.i536, %Flow8466 ]
  br label %cleanup26.i.i.i565, !amdgpu.uniform !42

if.end22.i.i.i628:                                ; preds = %if.then16.i.i.i629, %if.end14.i.i.i624
  call void @llvm.amdgcn.end.cf.i64(i64 %911)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %Flow8467, !amdgpu.uniform !42

Flow8468:                                         ; preds = %cleanup26.i.i.i565, %889
  %916 = phi i32 [ %915, %cleanup26.i.i.i565 ], [ %878, %889 ]
  %917 = phi i64 [ %914, %cleanup26.i.i.i565 ], [ %879, %889 ]
  %918 = phi i32 [ %913, %cleanup26.i.i.i565 ], [ %880, %889 ]
  %919 = phi i32 [ %913, %cleanup26.i.i.i565 ], [ %retval.sroa.8.0.i.i.i539, %889 ]
  %920 = phi i64 [ %914, %cleanup26.i.i.i565 ], [ %retval.sroa.2.0.i.i.i538, %889 ]
  %921 = phi i32 [ %915, %cleanup26.i.i.i565 ], [ %retval.sroa.6.0.i.i.i536, %889 ]
  %922 = phi i1 [ %cmp.i46.not.i.i.i557, %cleanup26.i.i.i565 ], [ true, %889 ]
  br i1 %922, label %for.inc.i.i.i570, label %Flow8469, !amdgpu.uniform !42

cleanup26.i.i.i565:                               ; preds = %Flow8467
  br label %Flow8468, !amdgpu.uniform !42

for.inc.i.i.i570:                                 ; preds = %Flow8468
  %inc.i.i.i575 = add i32 %spec.store.select.i.i.i542, 1
  br label %Flow8469, !amdgpu.uniform !42

Flow8469:                                         ; preds = %for.inc.i.i.i570, %Flow8468
  %923 = phi i32 [ %inc.i.i.i575, %for.inc.i.i.i570 ], [ poison, %Flow8468 ]
  %924 = phi i1 [ false, %for.inc.i.i.i570 ], [ true, %Flow8468 ]
  br i1 %924, label %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i576, label %for.cond.i.i1.i535, !amdgpu.uniform !42

_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i576: ; preds = %Flow8469
  %.lcssa9286 = phi i32 [ %916, %Flow8469 ]
  %.lcssa9285 = phi i64 [ %917, %Flow8469 ]
  %.lcssa9284 = phi i32 [ %918, %Flow8469 ]
  fence acquire
  %925 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %mul.i.i.i.i.i.i577 = shl i32 %.lcssa9286, 6
  %idxprom.i19.i.i.i.i.i578 = zext i32 %mul.i.i.i.i.i.i577 to i64
  %arrayidx.i20.i.i.i.i.i579 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %925, i64 %idxprom.i19.i.i.i.i.i578
  %arrayidx.i22.i.i.i.i.i580 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %arrayidx.i20.i.i.i.i.i579, i64 %sh_prom.i.i.i.i
  store i64 sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.15.33 to ptr), i64 1) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.15.33 to ptr) to i64)), ptr %arrayidx.i22.i.i.i.i.i580, align 8, !tbaa !82
  %spec.select.i.i.i.i.i.i.i581 = tail call i64 @llvm.umin.i64(i64 sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.15.33 to ptr), i64 1) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.15.33 to ptr) to i64)), i64 56)
  %926 = and i64 %spec.select.i.i.i.i.i.i.i581, 15
  %927 = sub i64 %spec.select.i.i.i.i.i.i.i581, %926
  %928 = icmp ne i64 %927, 0
  br i1 %928, label %loop-memcpy-expansion5235.preheader, label %Flow8465, !amdgpu.uniform !42

loop-memcpy-expansion5235.preheader:              ; preds = %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i576
  %929 = add nuw nsw i64 %78, 8
  %scevgep7142 = getelementptr i8, ptr %925, i64 %929
  %930 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i578, 6
  %scevgep7143 = getelementptr i8, ptr %scevgep7142, i64 %930
  br label %loop-memcpy-expansion5235, !amdgpu.uniform !42

Flow8465:                                         ; preds = %Flow8464, %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i576
  br label %loop-memcpy-residual-header5238, !amdgpu.uniform !42

loop-memcpy-expansion5235:                        ; preds = %loop-memcpy-expansion5235.preheader, %loop-memcpy-expansion5235
  %loop-index5236 = phi i64 [ %933, %loop-memcpy-expansion5235 ], [ 0, %loop-memcpy-expansion5235.preheader ]
  %931 = getelementptr i8, ptr addrspace(4) @.str.15.33, i64 %loop-index5236, !amdgpu.uniform !42
  %932 = load <4 x i32>, ptr addrspace(4) %931, align 1
  %scevgep7144 = getelementptr i8, ptr %scevgep7143, i64 %loop-index5236
  store <4 x i32> %932, ptr %scevgep7144, align 1
  %933 = add i64 %loop-index5236, 16
  %934 = icmp uge i64 %933, %927
  br i1 %934, label %Flow8464, label %loop-memcpy-expansion5235, !amdgpu.uniform !42

Flow8463:                                         ; preds = %Flow8461, %Flow8462
  %935 = phi i64 [ %6533, %Flow8461 ], [ %6530, %Flow8462 ]
  br label %post-loop-memcpy-expansion5234, !amdgpu.uniform !42

loop-memcpy-residual5237:                         ; preds = %loop-memcpy-residual5237.preheader, %loop-memcpy-residual5237
  %residual-loop-index5239 = phi i64 [ %937, %loop-memcpy-residual5237 ], [ 0, %loop-memcpy-residual5237.preheader ]
  %scevgep7141 = getelementptr i8, ptr addrspace(4) @.str.15.33, i64 %residual-loop-index5239, !amdgpu.uniform !42
  %936 = load i8, ptr addrspace(4) %scevgep7141, align 1
  %scevgep7140 = getelementptr i8, ptr %scevgep7139, i64 %residual-loop-index5239
  store i8 %936, ptr %scevgep7140, align 1
  %937 = add i64 %residual-loop-index5239, 1
  %938 = icmp uge i64 %937, %926
  br i1 %938, label %Flow8461, label %loop-memcpy-residual5237, !amdgpu.uniform !42

Flow8461:                                         ; preds = %loop-memcpy-residual5237
  br label %Flow8463, !amdgpu.uniform !42

post-loop-memcpy-expansion5234:                   ; preds = %Flow8463
  %tobool.not.i.i.i.i.i.i583 = icmp eq i32 %.lcssa9284, 0
  %conv.i.i.i.i.i.i584 = zext i1 %tobool.not.i.i.i.i.i.i583 to i32
  fence release
  %939 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %idxprom.i23.i.i.i.i.i585 = zext i32 %.lcssa9286 to i64
  %arrayidx.i24.i.i.i.i.i586 = getelementptr inbounds nuw i32, ptr %939, i64 %idxprom.i23.i.i.i.i.i585
  store atomic i32 %conv.i.i.i.i.i.i584, ptr %arrayidx.i24.i.i.i.i.i586 monotonic, align 4
  %940 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !84
  %arrayidx.i.i5.i.i587 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %940, i64 %idxprom.i23.i.i.i.i.i585, !amdgpu.uniform !42
  %941 = load i64, ptr %arrayidx.i.i5.i.i587, align 8, !tbaa !80
  br label %while.cond.i.i.i.i588, !amdgpu.uniform !42

while.cond.i.i.i.i588:                            ; preds = %post-loop-memcpy-expansion5234, %Flow8460
  %phi.broken8642 = phi i64 [ 0, %post-loop-memcpy-expansion5234 ], [ %958, %Flow8460 ]
  %lsr.iv7131 = phi i64 [ %956, %Flow8460 ], [ -55, %post-loop-memcpy-expansion5234 ]
  %lsr.iv7126 = phi ptr addrspace(4) [ %955, %Flow8460 ], [ getelementptr (i8, ptr addrspace(4) @.str.15.33, i64 56), %post-loop-memcpy-expansion5234 ]
  %port.sroa.37.0.i.i589 = phi i32 [ %954, %Flow8460 ], [ %conv.i.i.i.i.i.i584, %post-loop-memcpy-expansion5234 ]
  %idx.0.i.i.i.i590 = phi i64 [ %953, %Flow8460 ], [ 56, %post-loop-memcpy-expansion5234 ]
  %cmp.i.i.i.i591 = icmp ult i64 %idx.0.i.i.i.i590, sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.15.33 to ptr), i64 1) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.15.33 to ptr) to i64))
  %942 = tail call i64 @llvm.amdgcn.ballot.i64(i1 %cmp.i.i.i.i591)
  %and.i.i.i.i6.i.i592 = and i64 %942, %941
  %tobool.not.i.i.i.i593 = icmp ne i64 %and.i.i.i.i6.i.i592, 0
  %943 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i22.i.i.i594 = getelementptr inbounds nuw i32, ptr %943, i64 %idxprom.i23.i.i.i.i.i585, !amdgpu.uniform !42
  %944 = load atomic i32, ptr %arrayidx.i22.i.i.i594 monotonic, align 4
  %945 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %tobool.not.i.i.i.i593)
  %946 = extractvalue { i1, i64 } %945, 0
  %947 = extractvalue { i1, i64 } %945, 1
  br i1 %946, label %cond.false.i8.i.i.i.i595, label %Flow8460

cond.false.i8.i.i.i.i595:                         ; preds = %while.cond.i.i.i.i588
  %948 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %944)
  %949 = icmp ne i32 %948, %port.sroa.37.0.i.i589
  br i1 %949, label %while.body.i.i21.i.i.i.i596.preheader, label %Flow8459, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i596.preheader:            ; preds = %cond.false.i8.i.i.i.i595
  br label %while.body.i.i21.i.i.i.i596, !amdgpu.uniform !42

Flow8459:                                         ; preds = %Flow8458, %cond.false.i8.i.i.i.i595
  br label %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i599, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i596:                      ; preds = %while.body.i.i21.i.i.i.i596.preheader, %while.body.i.i21.i.i.i.i596
  tail call void @llvm.amdgcn.s.sleep(i32 2)
  %950 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i22.i.i.i.i597 = getelementptr inbounds nuw i32, ptr %950, i64 %idxprom.i23.i.i.i.i.i585, !amdgpu.uniform !42
  %951 = load atomic i32, ptr %arrayidx.i.i.i22.i.i.i.i597 monotonic, align 4
  %952 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %951)
  %cmp.i.not.i.i23.i.i.i.i598 = icmp eq i32 %952, %port.sroa.37.0.i.i589
  br i1 %cmp.i.not.i.i23.i.i.i.i598, label %Flow8458, label %while.body.i.i21.i.i.i.i596, !amdgpu.uniform !42

Flow8458:                                         ; preds = %while.body.i.i21.i.i.i.i596
  br label %Flow8459, !amdgpu.uniform !42

Flow8460:                                         ; preds = %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i600, %while.cond.i.i.i.i588
  %953 = phi i64 [ %add.i.i.i.i603, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i600 ], [ poison, %while.cond.i.i.i.i588 ]
  %954 = phi i32 [ %conv.i.i26.i.i.i.i601, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i600 ], [ poison, %while.cond.i.i.i.i588 ]
  %955 = phi ptr addrspace(4) [ %scevgep7127, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i600 ], [ poison, %while.cond.i.i.i.i588 ]
  %956 = phi i64 [ %lsr.iv.next7132, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i600 ], [ poison, %while.cond.i.i.i.i588 ]
  %957 = phi i1 [ false, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i600 ], [ true, %while.cond.i.i.i.i588 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %947)
  %958 = call i64 @llvm.amdgcn.if.break.i64(i1 %957, i64 %phi.broken8642)
  %959 = call i1 @llvm.amdgcn.loop.i64(i64 %958)
  br i1 %959, label %cond.false.i.i.i610, label %while.cond.i.i.i.i588

_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i599: ; preds = %Flow8459
  fence acquire
  br i1 %cmp.i.i.i.i591, label %if.then.i.i.i.i.i.i.i604, label %Flow8457, !amdgpu.uniform !42

if.then.i.i.i.i.i.i.i604:                         ; preds = %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i599
  %sub.i.i.i.i.i.i.i606 = sub nuw i64 sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.15.33 to ptr), i64 1) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.15.33 to ptr) to i64)), %idx.0.i.i.i.i590
  %spec.select.i.i.i36.i.i.i.i607 = tail call i64 @llvm.umin.i64(i64 %sub.i.i.i.i.i.i.i606, i64 64)
  %960 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %961 = and i64 %spec.select.i.i.i36.i.i.i.i607, 15
  %962 = sub i64 %spec.select.i.i.i36.i.i.i.i607, %961
  %963 = icmp ne i64 %962, 0
  br i1 %963, label %loop-memcpy-expansion5241.preheader, label %Flow8456, !amdgpu.uniform !42

loop-memcpy-expansion5241.preheader:              ; preds = %if.then.i.i.i.i.i.i.i604
  %arrayidx.i20.i35.i.i.i.i608 = getelementptr %"struct.rpc::Buffer", ptr %960, i64 %idxprom.i19.i.i.i.i.i578
  %arrayidx.i22.i38.i.i.i.i609 = getelementptr %"struct.rpc::Buffer", ptr %arrayidx.i20.i35.i.i.i.i608, i64 %sh_prom.i.i.i.i
  br label %loop-memcpy-expansion5241, !amdgpu.uniform !42

Flow8456:                                         ; preds = %Flow8455, %if.then.i.i.i.i.i.i.i604
  br label %loop-memcpy-residual-header5244, !amdgpu.uniform !42

loop-memcpy-expansion5241:                        ; preds = %loop-memcpy-expansion5241.preheader, %loop-memcpy-expansion5241
  %loop-index5242 = phi i64 [ %966, %loop-memcpy-expansion5241 ], [ 0, %loop-memcpy-expansion5241.preheader ]
  %scevgep7128 = getelementptr i8, ptr addrspace(4) %lsr.iv7126, i64 %loop-index5242, !amdgpu.uniform !42
  %964 = load <4 x i32>, ptr addrspace(4) %scevgep7128, align 1
  %965 = getelementptr i8, ptr %arrayidx.i22.i38.i.i.i.i609, i64 %loop-index5242
  store <4 x i32> %964, ptr %965, align 1
  %966 = add i64 %loop-index5242, 16
  %967 = icmp uge i64 %966, %962
  br i1 %967, label %Flow8455, label %loop-memcpy-expansion5241, !amdgpu.uniform !42

Flow8454:                                         ; preds = %Flow8453, %loop-memcpy-residual-header5244
  br label %Flow8457, !amdgpu.uniform !42

loop-memcpy-residual5243:                         ; preds = %loop-memcpy-residual5243.preheader, %loop-memcpy-residual5243
  %residual-loop-index5245 = phi i64 [ %969, %loop-memcpy-residual5243 ], [ 0, %loop-memcpy-residual5243.preheader ]
  %scevgep7137 = getelementptr i8, ptr addrspace(4) %scevgep7136, i64 %residual-loop-index5245, !amdgpu.uniform !42
  %968 = load i8, ptr addrspace(4) %scevgep7137, align 1
  %scevgep7135 = getelementptr i8, ptr %scevgep7134, i64 %residual-loop-index5245
  store i8 %968, ptr %scevgep7135, align 1
  %969 = add i64 %residual-loop-index5245, 1
  %970 = icmp uge i64 %969, %961
  br i1 %970, label %Flow8453, label %loop-memcpy-residual5243, !amdgpu.uniform !42

Flow8453:                                         ; preds = %loop-memcpy-residual5243
  br label %Flow8454, !amdgpu.uniform !42

_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i600: ; preds = %Flow8457
  %conv.i.i26.i.i.i.i601 = xor i32 %port.sroa.37.0.i.i589, 1
  fence release
  %971 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %arrayidx.i24.i29.i.i.i.i602 = getelementptr inbounds nuw i32, ptr %971, i64 %idxprom.i23.i.i.i.i.i585
  store atomic i32 %conv.i.i26.i.i.i.i601, ptr %arrayidx.i24.i29.i.i.i.i602 monotonic, align 4
  %add.i.i.i.i603 = add i64 %idx.0.i.i.i.i590, 64
  %scevgep7127 = getelementptr i8, ptr addrspace(4) %lsr.iv7126, i64 64
  %lsr.iv.next7132 = add i64 %lsr.iv7131, -64
  br label %Flow8460, !amdgpu.uniform !42

cond.false.i.i.i610:                              ; preds = %Flow8460
  %.lcssa9280 = phi i64 [ %958, %Flow8460 ]
  %port.sroa.37.0.i.i589.lcssa = phi i32 [ %port.sroa.37.0.i.i589, %Flow8460 ]
  %.lcssa9279 = phi i32 [ %944, %Flow8460 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa9280)
  %972 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %.lcssa9279)
  %973 = icmp ne i32 %972, %port.sroa.37.0.i.i589.lcssa
  %974 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %973)
  %975 = extractvalue { i1, i64 } %974, 0
  %976 = extractvalue { i1, i64 } %974, 1
  br i1 %975, label %while.body.i.i.i.i611.preheader, label %Flow8452

while.body.i.i.i.i611.preheader:                  ; preds = %cond.false.i.i.i610
  br label %while.body.i.i.i.i611, !amdgpu.uniform !42

Flow8452:                                         ; preds = %Flow8451, %cond.false.i.i.i610
  call void @llvm.amdgcn.end.cf.i64(i64 %976)
  br label %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit632, !amdgpu.uniform !42

while.body.i.i.i.i611:                            ; preds = %while.body.i.i.i.i611.preheader, %while.body.i.i.i.i611
  %phi.broken8643 = phi i64 [ 0, %while.body.i.i.i.i611.preheader ], [ %980, %while.body.i.i.i.i611 ]
  tail call void @llvm.amdgcn.s.sleep(i32 2)
  %977 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i8.i.i612 = getelementptr inbounds nuw i32, ptr %977, i64 %idxprom.i23.i.i.i.i.i585, !amdgpu.uniform !42
  %978 = load atomic i32, ptr %arrayidx.i.i.i8.i.i612 monotonic, align 4
  %979 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %978)
  %cmp.i.not.i.i.i.i613 = icmp eq i32 %979, %port.sroa.37.0.i.i589.lcssa
  %980 = call i64 @llvm.amdgcn.if.break.i64(i1 %cmp.i.not.i.i.i.i613, i64 %phi.broken8643)
  %981 = call i1 @llvm.amdgcn.loop.i64(i64 %980)
  br i1 %981, label %Flow8451, label %while.body.i.i.i.i611

Flow8451:                                         ; preds = %while.body.i.i.i.i611
  %.lcssa9278 = phi i64 [ %980, %while.body.i.i.i.i611 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa9278)
  br label %Flow8452, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit632: ; preds = %Flow8452
  fence acquire
  tail call void @llvm.amdgcn.wave.barrier()
  fence syncscope("agent") release
  %982 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %.lcssa9285, i1 true)
  %iszero.i.i.i.i.i13.i.i615 = icmp ne i64 %.lcssa9285, 0
  %cmp2.i.i.i.i14.i.i616 = icmp eq i64 %982, %sh_prom.i.i.i.i
  %cmp.i.i.i.i15.i.i617 = select i1 %iszero.i.i.i.i.i13.i.i615, i1 %cmp2.i.i.i.i14.i.i616, i1 false
  %983 = and i32 %.lcssa9286, 31
  %conv4.i.i.i19.i.i621 = zext i1 %cmp.i.i.i.i15.i.i617 to i32
  %shl.i.i.i20.i.i622 = shl nuw i32 %conv4.i.i.i19.i.i621, %983
  %xor.i.i.i21.i.i623 = xor i32 %shl.i.i.i20.i.i622, -1
  %984 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop5550, !amdgpu.uniform !42

985:                                              ; preds = %ComputeEnd5551
  %div8.i.i.i16.i.i618 = lshr i32 %.lcssa9286, 5
  %idxprom.i.i.i17.i.i619 = zext nneg i32 %div8.i.i.i16.i.i618 to i64
  %986 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i17.i.i619
  %sunkaddr7382 = getelementptr inbounds i8, ptr addrspace(1) %986, i64 40
  %987 = atomicrmw and ptr addrspace(1) %sunkaddr7382, i32 %.lcssa9277 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %988, !amdgpu.uniform !42

988:                                              ; preds = %985, %ComputeEnd5551
  call void @llvm.amdgcn.end.cf.i64(i64 %7066)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %Flow8470, !amdgpu.uniform !42

for.cond.i.i1.i641.preheader:                     ; preds = %Flow8470
  br label %for.cond.i.i1.i641, !amdgpu.uniform !42

for.cond.i.i1.i641:                               ; preds = %for.cond.i.i1.i641.preheader, %Flow8450
  %989 = phi i32 [ %1027, %Flow8450 ], [ poison, %for.cond.i.i1.i641.preheader ]
  %990 = phi i64 [ %1028, %Flow8450 ], [ poison, %for.cond.i.i1.i641.preheader ]
  %991 = phi i32 [ %1029, %Flow8450 ], [ poison, %for.cond.i.i1.i641.preheader ]
  %retval.sroa.6.0.i.i.i642 = phi i32 [ %1032, %Flow8450 ], [ undef, %for.cond.i.i1.i641.preheader ]
  %retval.sroa.2.0.i.i.i644 = phi i64 [ %1031, %Flow8450 ], [ undef, %for.cond.i.i1.i641.preheader ]
  %retval.sroa.8.0.i.i.i645 = phi i32 [ %1030, %Flow8450 ], [ undef, %for.cond.i.i1.i641.preheader ]
  %index.0.i.i.i646 = phi i32 [ %1034, %Flow8450 ], [ 0, %for.cond.i.i1.i641.preheader ]
  %992 = load i32, ptr addrspace(1) @__llvm_rpc_client, align 8, !tbaa !68
  %cmp.not.i.i.i647 = icmp ult i32 %index.0.i.i.i646, %992
  %spec.store.select.i.i.i648 = select i1 %cmp.not.i.i.i647, i32 %index.0.i.i.i646, i32 0
  %993 = tail call noundef i64 @llvm.amdgcn.ballot.i64(i1 true)
  %994 = and i32 %spec.store.select.i.i.i648, 31
  %995 = lshr i64 %993, %sh_prom.i.i.i.i
  %996 = trunc i64 %995 to i32
  %conv4.i.i.i.i.i652 = and i32 %996, 1
  %shl.i.i.i.i.i653 = shl nuw i32 %conv4.i.i.i.i.i652, %994
  %997 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop5554, !amdgpu.uniform !42

998:                                              ; preds = %ComputeEnd5555
  %sunkaddr7383 = getelementptr inbounds i8, ptr addrspace(1) %7076, i64 40
  %999 = atomicrmw or ptr addrspace(1) %sunkaddr7383, i32 %.lcssa9271 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %1000, !amdgpu.uniform !42

1000:                                             ; preds = %998, %ComputeEnd5555
  %1001 = phi i32 [ %999, %998 ], [ poison, %ComputeEnd5555 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %7085)
  %1002 = call i32 @llvm.amdgcn.readfirstlane.i32(i32 %1001)
  %1003 = or i32 %1002, %.lcssa9272
  %shl5.i.i.i.i.i654 = shl nuw i32 1, %994
  %and.i.i.i.i.i655 = and i32 %shl5.i.i.i.i.i654, %1003
  %tobool3.i.i.i.i656 = icmp ne i32 %and.i.i.i.i.i655, 0
  %1004 = tail call i64 @llvm.amdgcn.ballot.i64(i1 %tobool3.i.i.i.i656)
  %and.i.i.i.i.i.i657 = and i64 %1004, %993
  %cmp.i.not.i.i.i658 = icmp ne i64 %993, %and.i.i.i.i.i.i657
  br i1 %cmp.i.not.i.i.i658, label %if.then.i.i.i.i659, label %Flow8449, !amdgpu.uniform !42

if.then.i.i.i.i659:                               ; preds = %1000
  fence syncscope("agent") acquire
  %1005 = load <2 x i64>, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !75
  %1006 = extractelement <2 x i64> %1005, i32 0
  %1007 = inttoptr i64 %1006 to ptr
  %1008 = extractelement <2 x i64> %1005, i32 1
  %1009 = inttoptr i64 %1008 to ptr
  %idxprom.i.i.i.i660 = zext i32 %spec.store.select.i.i.i648 to i64
  %arrayidx.i.i.i.i661 = getelementptr inbounds nuw i32, ptr %1007, i64 %idxprom.i.i.i.i660, !amdgpu.uniform !42
  %1010 = load atomic i32, ptr %arrayidx.i.i.i.i661 monotonic, align 4
  %1011 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %1010)
  %arrayidx.i45.i.i.i662 = getelementptr inbounds nuw i32, ptr %1009, i64 %idxprom.i.i.i.i660, !amdgpu.uniform !42
  %1012 = load atomic i32, ptr %arrayidx.i45.i.i.i662 monotonic, align 4
  %1013 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %1012)
  %cmp.i46.not.i.i.i663 = icmp ne i32 %1011, %1013
  br i1 %cmp.i46.not.i.i.i663, label %if.then12.i.i.i664, label %Flow8447, !amdgpu.uniform !42

if.then12.i.i.i664:                               ; preds = %if.then.i.i.i.i659
  fence syncscope("agent") release
  %1014 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %993, i1 true)
  %iszero.i.i.i.i.i.i.i665 = icmp ne i64 %993, 0
  %cmp2.i.i.i.i.i.i666 = icmp eq i64 %1014, %sh_prom.i.i.i.i
  %cmp.i.i.i.i.i.i667 = select i1 %iszero.i.i.i.i.i.i.i665, i1 %cmp2.i.i.i.i.i.i666, i1 false
  %conv4.i.i51.i.i.i668 = zext i1 %cmp.i.i.i.i.i.i667 to i32
  %shl.i.i52.i.i.i669 = shl nuw i32 %conv4.i.i51.i.i.i668, %994
  %xor.i.i.i.i.i670 = xor i32 %shl.i.i52.i.i.i669, -1
  %1015 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop5559, !amdgpu.uniform !42

1016:                                             ; preds = %ComputeEnd5560
  %sunkaddr7384 = getelementptr inbounds i8, ptr addrspace(1) %7076, i64 40
  %1017 = atomicrmw and ptr addrspace(1) %sunkaddr7384, i32 %.lcssa9273 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %1018, !amdgpu.uniform !42

1018:                                             ; preds = %1016, %ComputeEnd5560
  call void @llvm.amdgcn.end.cf.i64(i64 %7103)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %Flow8447, !amdgpu.uniform !42

if.end14.i.i.i730:                                ; preds = %Flow8447
  %1019 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %993, i1 true)
  %iszero.i.i.i.i.i.i731 = icmp ne i64 %993, 0
  %cmp2.i.i.i.i.i732 = icmp eq i64 %1019, %sh_prom.i.i.i.i
  %cmp.i.i.i.i.i733 = select i1 %iszero.i.i.i.i.i.i731, i1 %cmp2.i.i.i.i.i732, i1 false
  %1020 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.i.i.i.i.i733)
  %1021 = extractvalue { i1, i64 } %1020, 0
  %1022 = extractvalue { i1, i64 } %1020, 1
  br i1 %1021, label %if.then16.i.i.i735, label %if.end22.i.i.i734

if.then16.i.i.i735:                               ; preds = %if.end14.i.i.i730
  %1023 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !76
  %opcode.i.i.i736 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %1023, i64 %idxprom.i.i.i.i660, i32 1
  store i32 1660944387, ptr %opcode.i.i.i736, align 8, !tbaa !77
  %arrayidx21.i.i.i737 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %1023, i64 %idxprom.i.i.i.i660
  store i64 %993, ptr %arrayidx21.i.i.i737, align 8, !tbaa !80
  br label %if.end22.i.i.i734, !amdgpu.uniform !42

Flow8448:                                         ; preds = %if.end22.i.i.i734, %Flow8447
  %1024 = phi i32 [ %1011, %if.end22.i.i.i734 ], [ %retval.sroa.8.0.i.i.i645, %Flow8447 ]
  %1025 = phi i64 [ %993, %if.end22.i.i.i734 ], [ %retval.sroa.2.0.i.i.i644, %Flow8447 ]
  %1026 = phi i32 [ %spec.store.select.i.i.i648, %if.end22.i.i.i734 ], [ %retval.sroa.6.0.i.i.i642, %Flow8447 ]
  br label %cleanup26.i.i.i671, !amdgpu.uniform !42

if.end22.i.i.i734:                                ; preds = %if.then16.i.i.i735, %if.end14.i.i.i730
  call void @llvm.amdgcn.end.cf.i64(i64 %1022)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %Flow8448, !amdgpu.uniform !42

Flow8449:                                         ; preds = %cleanup26.i.i.i671, %1000
  %1027 = phi i32 [ %1026, %cleanup26.i.i.i671 ], [ %989, %1000 ]
  %1028 = phi i64 [ %1025, %cleanup26.i.i.i671 ], [ %990, %1000 ]
  %1029 = phi i32 [ %1024, %cleanup26.i.i.i671 ], [ %991, %1000 ]
  %1030 = phi i32 [ %1024, %cleanup26.i.i.i671 ], [ %retval.sroa.8.0.i.i.i645, %1000 ]
  %1031 = phi i64 [ %1025, %cleanup26.i.i.i671 ], [ %retval.sroa.2.0.i.i.i644, %1000 ]
  %1032 = phi i32 [ %1026, %cleanup26.i.i.i671 ], [ %retval.sroa.6.0.i.i.i642, %1000 ]
  %1033 = phi i1 [ %cmp.i46.not.i.i.i663, %cleanup26.i.i.i671 ], [ true, %1000 ]
  br i1 %1033, label %for.inc.i.i.i676, label %Flow8450, !amdgpu.uniform !42

cleanup26.i.i.i671:                               ; preds = %Flow8448
  br label %Flow8449, !amdgpu.uniform !42

for.inc.i.i.i676:                                 ; preds = %Flow8449
  %inc.i.i.i681 = add i32 %spec.store.select.i.i.i648, 1
  br label %Flow8450, !amdgpu.uniform !42

Flow8450:                                         ; preds = %for.inc.i.i.i676, %Flow8449
  %1034 = phi i32 [ %inc.i.i.i681, %for.inc.i.i.i676 ], [ poison, %Flow8449 ]
  %1035 = phi i1 [ false, %for.inc.i.i.i676 ], [ true, %Flow8449 ]
  br i1 %1035, label %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i682, label %for.cond.i.i1.i641, !amdgpu.uniform !42

_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i682: ; preds = %Flow8450
  %.lcssa9276 = phi i32 [ %1027, %Flow8450 ]
  %.lcssa9275 = phi i64 [ %1028, %Flow8450 ]
  %.lcssa9274 = phi i32 [ %1029, %Flow8450 ]
  fence acquire
  %1036 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %mul.i.i.i.i.i.i683 = shl i32 %.lcssa9276, 6
  %idxprom.i19.i.i.i.i.i684 = zext i32 %mul.i.i.i.i.i.i683 to i64
  %arrayidx.i20.i.i.i.i.i685 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %1036, i64 %idxprom.i19.i.i.i.i.i684
  %arrayidx.i22.i.i.i.i.i686 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %arrayidx.i20.i.i.i.i.i685, i64 %sh_prom.i.i.i.i
  store i64 sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.16 to ptr), i64 20) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.16 to ptr) to i64)), ptr %arrayidx.i22.i.i.i.i.i686, align 8, !tbaa !82
  %spec.select.i.i.i.i.i.i.i687 = tail call i64 @llvm.umin.i64(i64 sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.16 to ptr), i64 20) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.16 to ptr) to i64)), i64 56)
  %1037 = and i64 %spec.select.i.i.i.i.i.i.i687, 15
  %1038 = sub i64 %spec.select.i.i.i.i.i.i.i687, %1037
  %1039 = icmp ne i64 %1038, 0
  br i1 %1039, label %loop-memcpy-expansion5247.preheader, label %Flow8446, !amdgpu.uniform !42

loop-memcpy-expansion5247.preheader:              ; preds = %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i682
  %1040 = add nuw nsw i64 %78, 8
  %scevgep7123 = getelementptr i8, ptr %1036, i64 %1040
  %1041 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i684, 6
  %scevgep7124 = getelementptr i8, ptr %scevgep7123, i64 %1041
  br label %loop-memcpy-expansion5247, !amdgpu.uniform !42

Flow8446:                                         ; preds = %Flow8445, %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i682
  br label %loop-memcpy-residual-header5250, !amdgpu.uniform !42

loop-memcpy-expansion5247:                        ; preds = %loop-memcpy-expansion5247.preheader, %loop-memcpy-expansion5247
  %loop-index5248 = phi i64 [ %1044, %loop-memcpy-expansion5247 ], [ 0, %loop-memcpy-expansion5247.preheader ]
  %1042 = getelementptr i8, ptr addrspace(4) @.str.16, i64 %loop-index5248, !amdgpu.uniform !42
  %1043 = load <4 x i32>, ptr addrspace(4) %1042, align 1
  %scevgep7125 = getelementptr i8, ptr %scevgep7124, i64 %loop-index5248
  store <4 x i32> %1043, ptr %scevgep7125, align 1
  %1044 = add i64 %loop-index5248, 16
  %1045 = icmp uge i64 %1044, %1038
  br i1 %1045, label %Flow8445, label %loop-memcpy-expansion5247, !amdgpu.uniform !42

Flow8444:                                         ; preds = %Flow8442, %Flow8443
  %1046 = phi i64 [ %6541, %Flow8442 ], [ %6538, %Flow8443 ]
  br label %post-loop-memcpy-expansion5246, !amdgpu.uniform !42

loop-memcpy-residual5249:                         ; preds = %loop-memcpy-residual5249.preheader, %loop-memcpy-residual5249
  %residual-loop-index5251 = phi i64 [ %1048, %loop-memcpy-residual5249 ], [ 0, %loop-memcpy-residual5249.preheader ]
  %scevgep7122 = getelementptr i8, ptr addrspace(4) getelementptr inbounds nuw (i8, ptr addrspace(4) @.str.16, i64 16), i64 %residual-loop-index5251, !amdgpu.uniform !42
  %1047 = load i8, ptr addrspace(4) %scevgep7122, align 1
  %scevgep7121 = getelementptr i8, ptr %scevgep7120, i64 %residual-loop-index5251
  store i8 %1047, ptr %scevgep7121, align 1
  %1048 = add i64 %residual-loop-index5251, 1
  %1049 = icmp uge i64 %1048, %1037
  br i1 %1049, label %Flow8442, label %loop-memcpy-residual5249, !amdgpu.uniform !42

Flow8442:                                         ; preds = %loop-memcpy-residual5249
  br label %Flow8444, !amdgpu.uniform !42

post-loop-memcpy-expansion5246:                   ; preds = %Flow8444
  %tobool.not.i.i.i.i.i.i689 = icmp eq i32 %.lcssa9274, 0
  %conv.i.i.i.i.i.i690 = zext i1 %tobool.not.i.i.i.i.i.i689 to i32
  fence release
  %1050 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %idxprom.i23.i.i.i.i.i691 = zext i32 %.lcssa9276 to i64
  %arrayidx.i24.i.i.i.i.i692 = getelementptr inbounds nuw i32, ptr %1050, i64 %idxprom.i23.i.i.i.i.i691
  store atomic i32 %conv.i.i.i.i.i.i690, ptr %arrayidx.i24.i.i.i.i.i692 monotonic, align 4
  %1051 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !84
  %arrayidx.i.i5.i.i693 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %1051, i64 %idxprom.i23.i.i.i.i.i691, !amdgpu.uniform !42
  %1052 = load i64, ptr %arrayidx.i.i5.i.i693, align 8, !tbaa !80
  br label %while.cond.i.i.i.i694, !amdgpu.uniform !42

while.cond.i.i.i.i694:                            ; preds = %post-loop-memcpy-expansion5246, %Flow8441
  %phi.broken8644 = phi i64 [ 0, %post-loop-memcpy-expansion5246 ], [ %1069, %Flow8441 ]
  %lsr.iv7112 = phi i64 [ %1067, %Flow8441 ], [ -36, %post-loop-memcpy-expansion5246 ]
  %lsr.iv7107 = phi ptr addrspace(4) [ %1066, %Flow8441 ], [ getelementptr (i8, ptr addrspace(4) @.str.16, i64 56), %post-loop-memcpy-expansion5246 ]
  %port.sroa.37.0.i.i695 = phi i32 [ %1065, %Flow8441 ], [ %conv.i.i.i.i.i.i690, %post-loop-memcpy-expansion5246 ]
  %idx.0.i.i.i.i696 = phi i64 [ %1064, %Flow8441 ], [ 56, %post-loop-memcpy-expansion5246 ]
  %cmp.i.i.i.i697 = icmp ult i64 %idx.0.i.i.i.i696, sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.16 to ptr), i64 20) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.16 to ptr) to i64))
  %1053 = tail call i64 @llvm.amdgcn.ballot.i64(i1 %cmp.i.i.i.i697)
  %and.i.i.i.i6.i.i698 = and i64 %1053, %1052
  %tobool.not.i.i.i.i699 = icmp ne i64 %and.i.i.i.i6.i.i698, 0
  %1054 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i22.i.i.i700 = getelementptr inbounds nuw i32, ptr %1054, i64 %idxprom.i23.i.i.i.i.i691, !amdgpu.uniform !42
  %1055 = load atomic i32, ptr %arrayidx.i22.i.i.i700 monotonic, align 4
  %1056 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %tobool.not.i.i.i.i699)
  %1057 = extractvalue { i1, i64 } %1056, 0
  %1058 = extractvalue { i1, i64 } %1056, 1
  br i1 %1057, label %cond.false.i8.i.i.i.i701, label %Flow8441

cond.false.i8.i.i.i.i701:                         ; preds = %while.cond.i.i.i.i694
  %1059 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %1055)
  %1060 = icmp ne i32 %1059, %port.sroa.37.0.i.i695
  br i1 %1060, label %while.body.i.i21.i.i.i.i702.preheader, label %Flow8440, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i702.preheader:            ; preds = %cond.false.i8.i.i.i.i701
  br label %while.body.i.i21.i.i.i.i702, !amdgpu.uniform !42

Flow8440:                                         ; preds = %Flow8439, %cond.false.i8.i.i.i.i701
  br label %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i705, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i702:                      ; preds = %while.body.i.i21.i.i.i.i702.preheader, %while.body.i.i21.i.i.i.i702
  tail call void @llvm.amdgcn.s.sleep(i32 2)
  %1061 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i22.i.i.i.i703 = getelementptr inbounds nuw i32, ptr %1061, i64 %idxprom.i23.i.i.i.i.i691, !amdgpu.uniform !42
  %1062 = load atomic i32, ptr %arrayidx.i.i.i22.i.i.i.i703 monotonic, align 4
  %1063 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %1062)
  %cmp.i.not.i.i23.i.i.i.i704 = icmp eq i32 %1063, %port.sroa.37.0.i.i695
  br i1 %cmp.i.not.i.i23.i.i.i.i704, label %Flow8439, label %while.body.i.i21.i.i.i.i702, !amdgpu.uniform !42

Flow8439:                                         ; preds = %while.body.i.i21.i.i.i.i702
  br label %Flow8440, !amdgpu.uniform !42

Flow8441:                                         ; preds = %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i706, %while.cond.i.i.i.i694
  %1064 = phi i64 [ %add.i.i.i.i709, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i706 ], [ poison, %while.cond.i.i.i.i694 ]
  %1065 = phi i32 [ %conv.i.i26.i.i.i.i707, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i706 ], [ poison, %while.cond.i.i.i.i694 ]
  %1066 = phi ptr addrspace(4) [ %scevgep7108, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i706 ], [ poison, %while.cond.i.i.i.i694 ]
  %1067 = phi i64 [ %lsr.iv.next7113, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i706 ], [ poison, %while.cond.i.i.i.i694 ]
  %1068 = phi i1 [ false, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i706 ], [ true, %while.cond.i.i.i.i694 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %1058)
  %1069 = call i64 @llvm.amdgcn.if.break.i64(i1 %1068, i64 %phi.broken8644)
  %1070 = call i1 @llvm.amdgcn.loop.i64(i64 %1069)
  br i1 %1070, label %cond.false.i.i.i716, label %while.cond.i.i.i.i694

_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i705: ; preds = %Flow8440
  fence acquire
  br i1 %cmp.i.i.i.i697, label %if.then.i.i.i.i.i.i.i710, label %Flow8438, !amdgpu.uniform !42

if.then.i.i.i.i.i.i.i710:                         ; preds = %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i705
  %sub.i.i.i.i.i.i.i712 = sub nuw i64 sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.16 to ptr), i64 20) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.16 to ptr) to i64)), %idx.0.i.i.i.i696
  %spec.select.i.i.i36.i.i.i.i713 = tail call i64 @llvm.umin.i64(i64 %sub.i.i.i.i.i.i.i712, i64 64)
  %1071 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %1072 = and i64 %spec.select.i.i.i36.i.i.i.i713, 15
  %1073 = sub i64 %spec.select.i.i.i36.i.i.i.i713, %1072
  %1074 = icmp ne i64 %1073, 0
  br i1 %1074, label %loop-memcpy-expansion5253.preheader, label %Flow8437, !amdgpu.uniform !42

loop-memcpy-expansion5253.preheader:              ; preds = %if.then.i.i.i.i.i.i.i710
  %arrayidx.i20.i35.i.i.i.i714 = getelementptr %"struct.rpc::Buffer", ptr %1071, i64 %idxprom.i19.i.i.i.i.i684
  %arrayidx.i22.i38.i.i.i.i715 = getelementptr %"struct.rpc::Buffer", ptr %arrayidx.i20.i35.i.i.i.i714, i64 %sh_prom.i.i.i.i
  br label %loop-memcpy-expansion5253, !amdgpu.uniform !42

Flow8437:                                         ; preds = %Flow8436, %if.then.i.i.i.i.i.i.i710
  br label %loop-memcpy-residual-header5256, !amdgpu.uniform !42

loop-memcpy-expansion5253:                        ; preds = %loop-memcpy-expansion5253.preheader, %loop-memcpy-expansion5253
  %loop-index5254 = phi i64 [ %1077, %loop-memcpy-expansion5253 ], [ 0, %loop-memcpy-expansion5253.preheader ]
  %scevgep7109 = getelementptr i8, ptr addrspace(4) %lsr.iv7107, i64 %loop-index5254, !amdgpu.uniform !42
  %1075 = load <4 x i32>, ptr addrspace(4) %scevgep7109, align 1
  %1076 = getelementptr i8, ptr %arrayidx.i22.i38.i.i.i.i715, i64 %loop-index5254
  store <4 x i32> %1075, ptr %1076, align 1
  %1077 = add i64 %loop-index5254, 16
  %1078 = icmp uge i64 %1077, %1073
  br i1 %1078, label %Flow8436, label %loop-memcpy-expansion5253, !amdgpu.uniform !42

Flow8435:                                         ; preds = %Flow8434, %loop-memcpy-residual-header5256
  br label %Flow8438, !amdgpu.uniform !42

loop-memcpy-residual5255:                         ; preds = %loop-memcpy-residual5255.preheader, %loop-memcpy-residual5255
  %residual-loop-index5257 = phi i64 [ %1080, %loop-memcpy-residual5255 ], [ 0, %loop-memcpy-residual5255.preheader ]
  %scevgep7118 = getelementptr i8, ptr addrspace(4) %scevgep7117, i64 %residual-loop-index5257, !amdgpu.uniform !42
  %1079 = load i8, ptr addrspace(4) %scevgep7118, align 1
  %scevgep7116 = getelementptr i8, ptr %scevgep7115, i64 %residual-loop-index5257
  store i8 %1079, ptr %scevgep7116, align 1
  %1080 = add i64 %residual-loop-index5257, 1
  %1081 = icmp uge i64 %1080, %1072
  br i1 %1081, label %Flow8434, label %loop-memcpy-residual5255, !amdgpu.uniform !42

Flow8434:                                         ; preds = %loop-memcpy-residual5255
  br label %Flow8435, !amdgpu.uniform !42

_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i706: ; preds = %Flow8438
  %conv.i.i26.i.i.i.i707 = xor i32 %port.sroa.37.0.i.i695, 1
  fence release
  %1082 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %arrayidx.i24.i29.i.i.i.i708 = getelementptr inbounds nuw i32, ptr %1082, i64 %idxprom.i23.i.i.i.i.i691
  store atomic i32 %conv.i.i26.i.i.i.i707, ptr %arrayidx.i24.i29.i.i.i.i708 monotonic, align 4
  %add.i.i.i.i709 = add i64 %idx.0.i.i.i.i696, 64
  %scevgep7108 = getelementptr i8, ptr addrspace(4) %lsr.iv7107, i64 64
  %lsr.iv.next7113 = add i64 %lsr.iv7112, -64
  br label %Flow8441, !amdgpu.uniform !42

cond.false.i.i.i716:                              ; preds = %Flow8441
  %.lcssa9270 = phi i64 [ %1069, %Flow8441 ]
  %port.sroa.37.0.i.i695.lcssa = phi i32 [ %port.sroa.37.0.i.i695, %Flow8441 ]
  %.lcssa9269 = phi i32 [ %1055, %Flow8441 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa9270)
  %1083 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %.lcssa9269)
  %1084 = icmp ne i32 %1083, %port.sroa.37.0.i.i695.lcssa
  %1085 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %1084)
  %1086 = extractvalue { i1, i64 } %1085, 0
  %1087 = extractvalue { i1, i64 } %1085, 1
  br i1 %1086, label %while.body.i.i.i.i717.preheader, label %Flow8433

while.body.i.i.i.i717.preheader:                  ; preds = %cond.false.i.i.i716
  br label %while.body.i.i.i.i717, !amdgpu.uniform !42

Flow8433:                                         ; preds = %Flow8432, %cond.false.i.i.i716
  call void @llvm.amdgcn.end.cf.i64(i64 %1087)
  br label %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit738, !amdgpu.uniform !42

while.body.i.i.i.i717:                            ; preds = %while.body.i.i.i.i717.preheader, %while.body.i.i.i.i717
  %phi.broken8645 = phi i64 [ 0, %while.body.i.i.i.i717.preheader ], [ %1091, %while.body.i.i.i.i717 ]
  tail call void @llvm.amdgcn.s.sleep(i32 2)
  %1088 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i8.i.i718 = getelementptr inbounds nuw i32, ptr %1088, i64 %idxprom.i23.i.i.i.i.i691, !amdgpu.uniform !42
  %1089 = load atomic i32, ptr %arrayidx.i.i.i8.i.i718 monotonic, align 4
  %1090 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %1089)
  %cmp.i.not.i.i.i.i719 = icmp eq i32 %1090, %port.sroa.37.0.i.i695.lcssa
  %1091 = call i64 @llvm.amdgcn.if.break.i64(i1 %cmp.i.not.i.i.i.i719, i64 %phi.broken8645)
  %1092 = call i1 @llvm.amdgcn.loop.i64(i64 %1091)
  br i1 %1092, label %Flow8432, label %while.body.i.i.i.i717

Flow8432:                                         ; preds = %while.body.i.i.i.i717
  %.lcssa9268 = phi i64 [ %1091, %while.body.i.i.i.i717 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa9268)
  br label %Flow8433, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit738: ; preds = %Flow8433
  fence acquire
  tail call void @llvm.amdgcn.wave.barrier()
  fence syncscope("agent") release
  %1093 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %.lcssa9275, i1 true)
  %iszero.i.i.i.i.i13.i.i721 = icmp ne i64 %.lcssa9275, 0
  %cmp2.i.i.i.i14.i.i722 = icmp eq i64 %1093, %sh_prom.i.i.i.i
  %cmp.i.i.i.i15.i.i723 = select i1 %iszero.i.i.i.i.i13.i.i721, i1 %cmp2.i.i.i.i14.i.i722, i1 false
  %1094 = and i32 %.lcssa9276, 31
  %conv4.i.i.i19.i.i727 = zext i1 %cmp.i.i.i.i15.i.i723 to i32
  %shl.i.i.i20.i.i728 = shl nuw i32 %conv4.i.i.i19.i.i727, %1094
  %xor.i.i.i21.i.i729 = xor i32 %shl.i.i.i20.i.i728, -1
  %1095 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop5563, !amdgpu.uniform !42

1096:                                             ; preds = %ComputeEnd5564
  %div8.i.i.i16.i.i724 = lshr i32 %.lcssa9276, 5
  %idxprom.i.i.i17.i.i725 = zext nneg i32 %div8.i.i.i16.i.i724 to i64
  %1097 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i17.i.i725
  %sunkaddr7385 = getelementptr inbounds i8, ptr addrspace(1) %1097, i64 40
  %1098 = atomicrmw and ptr addrspace(1) %sunkaddr7385, i32 %.lcssa9267 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %1099, !amdgpu.uniform !42

1099:                                             ; preds = %1096, %ComputeEnd5564
  call void @llvm.amdgcn.end.cf.i64(i64 %7120)
  tail call void @llvm.amdgcn.wave.barrier()
  %T.0201.pre = load ptr, ptr addrspace(1) @_ZN22__llvm_libc_22_0_0_git7testing4Test5StartE, align 8, !tbaa !63
  br label %Flow8609, !amdgpu.uniform !42

if.end16:                                         ; preds = %Flow8609
  %cmp17.not202 = icmp ne ptr %7, null
  %1100 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp17.not202)
  %1101 = extractvalue { i1, i64 } %1100, 0
  %1102 = extractvalue { i1, i64 } %1100, 1
  br i1 %1101, label %for.body.lr.ph, label %Flow8431

for.body.lr.ph:                                   ; preds = %if.end16
  %1103 = call ptr @llvm.amdgcn.addrspacecast.nonnull.p0.p5(ptr addrspace(5) %Ctx)
  %cond3 = select i1 %loadedv, ptr addrspace(4) @.str.10.28, ptr addrspace(4) @.str.9.27
  %1104 = addrspacecast ptr addrspace(4) %cond6 to ptr
  %1105 = addrspacecast ptr addrspace(4) %cond3 to ptr
  %1106 = addrspacecast ptr addrspace(4) %cond to ptr
  %sub.ptr.rhs.cast.i.i.i754 = ptrtoint ptr %1106 to i64
  %sub.ptr.rhs.cast.i.i.i967 = ptrtoint ptr %1104 to i64
  %1107 = load i64, ptr addrspace(4) @__llvm_libc_clock_freq, align 8
  %tobool.not.i1271 = icmp ne i64 %1107, 0
  %cmp.i = icmp sle i64 %1107, 1000000
  %div.i = udiv i64 %1107, 1000000
  %sub.ptr.rhs.cast.i.i.i1289 = ptrtoint ptr %1105 to i64
  %spec.select.i.i.i.i.i.i.i909 = call i64 @llvm.umin.i64(i64 sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.17.34 to ptr), i64 13) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.17.34 to ptr) to i64)), i64 56)
  %spec.select.i.i.i.i.i.i.i1970 = call i64 @llvm.umin.i64(i64 sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.19.36 to ptr), i64 13) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.19.36 to ptr) to i64)), i64 56)
  %spec.select.i.i.i.i.i.i.i2290 = call i64 @llvm.umin.i64(i64 sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.20.37 to ptr), i64 2) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.20.37 to ptr) to i64)), i64 56)
  %spec.select.i.i.i.i.i.i.i2936 = call i64 @llvm.umin.i64(i64 sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.23 to ptr), i64 5) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.23 to ptr) to i64)), i64 56)
  %spec.select.i.i.i.i.i.i.i3211 = call i64 @llvm.umin.i64(i64 sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.24.40 to ptr), i64 5) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.24.40 to ptr) to i64)), i64 56)
  %spec.select.i.i.i.i.i.i.i2661 = call i64 @llvm.umin.i64(i64 sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.22.39 to ptr), i64 5) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.22.39 to ptr) to i64)), i64 56)
  %spec.select.i.i.i.i.i.i.i2396 = call i64 @llvm.umin.i64(i64 sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.21.38 to ptr), i64 25) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.21.38 to ptr) to i64)), i64 56)
  %spec.select.i.i.i.i.i.i.i1444 = call i64 @llvm.umin.i64(i64 sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.18.35 to ptr), i64 13) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.18.35 to ptr) to i64)), i64 56)
  %1108 = and i64 %spec.select.i.i.i.i.i.i.i909, 15
  %1109 = sub i64 %spec.select.i.i.i.i.i.i.i909, %1108
  %1110 = icmp ne i64 %1109, 0
  %1111 = icmp eq i64 %1108, 0
  %1112 = and i64 %spec.select.i.i.i.i.i.i.i1970, 15
  %1113 = sub i64 %spec.select.i.i.i.i.i.i.i1970, %1112
  %1114 = icmp ne i64 %1113, 0
  %1115 = icmp eq i64 %1112, 0
  %1116 = and i64 %spec.select.i.i.i.i.i.i.i2290, 15
  %1117 = sub i64 %spec.select.i.i.i.i.i.i.i2290, %1116
  %1118 = icmp ne i64 %1117, 0
  %1119 = icmp eq i64 %1116, 0
  %1120 = and i64 %spec.select.i.i.i.i.i.i.i2936, 15
  %1121 = sub i64 %spec.select.i.i.i.i.i.i.i2936, %1120
  %1122 = icmp ne i64 %1121, 0
  %1123 = icmp eq i64 %1120, 0
  %1124 = and i64 %spec.select.i.i.i.i.i.i.i3211, 15
  %1125 = sub i64 %spec.select.i.i.i.i.i.i.i3211, %1124
  %1126 = icmp ne i64 %1125, 0
  %1127 = icmp eq i64 %1124, 0
  %1128 = and i64 %spec.select.i.i.i.i.i.i.i2661, 15
  %1129 = sub i64 %spec.select.i.i.i.i.i.i.i2661, %1128
  %1130 = icmp ne i64 %1129, 0
  %1131 = icmp eq i64 %1128, 0
  %1132 = and i64 %spec.select.i.i.i.i.i.i.i2396, 15
  %1133 = sub i64 %spec.select.i.i.i.i.i.i.i2396, %1132
  %1134 = icmp ne i64 %1133, 0
  %1135 = icmp eq i64 %1132, 0
  %1136 = and i64 %spec.select.i.i.i.i.i.i.i1444, 15
  %1137 = sub i64 %spec.select.i.i.i.i.i.i.i1444, %1136
  %1138 = icmp ne i64 %1137, 0
  %1139 = icmp eq i64 %1136, 0
  %scevgep6586 = getelementptr i8, ptr addrspace(4) %cond, i64 -1
  %scevgep6600 = getelementptr i8, ptr addrspace(4) %cond, i64 56
  %1140 = sub i64 -56, %sub.ptr.rhs.cast.i.i.i754
  %scevgep6634 = getelementptr i8, ptr addrspace(4) %cond6, i64 -1
  %scevgep6648 = getelementptr i8, ptr addrspace(4) %cond6, i64 56
  %1141 = sub i64 -56, %sub.ptr.rhs.cast.i.i.i967
  %scevgep6819 = getelementptr i8, ptr addrspace(5) %buffer.i.i.i2988, i32 21
  %scevgep6873 = getelementptr i8, ptr addrspace(5) %buffer.i.i.i2988, i32 21
  %scevgep6927 = getelementptr i8, ptr addrspace(5) %buffer.i.i.i2988, i32 21
  %scevgep7002 = getelementptr i8, ptr addrspace(4) %cond3, i64 -1
  %scevgep7016 = getelementptr i8, ptr addrspace(4) %cond3, i64 56
  %1142 = sub i64 -56, %sub.ptr.rhs.cast.i.i.i1289
  br label %for.body, !amdgpu.uniform !42

Flow8430:                                         ; preds = %cleanup
  %.lcssa9266 = phi i64 [ %4454, %cleanup ]
  %.lcssa9265 = phi i32 [ %1337, %cleanup ]
  %.lcssa9264 = phi i32 [ %1338, %cleanup ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa9266)
  br label %Flow8431, !amdgpu.uniform !42

for.cond.cleanup:                                 ; preds = %Flow8431
  %cmp89 = icmp sle i32 %1148, 0
  %1143 = call i32 @llvm.amdgcn.mbcnt.lo(i32 -1, i32 0)
  %1144 = call noundef i32 @llvm.amdgcn.mbcnt.hi(i32 -1, i32 %1143)
  %sh_prom.i.i.i.i3270 = zext i32 %1144 to i64
  %1145 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp89)
  %1146 = extractvalue { i1, i64 } %1145, 0
  %1147 = extractvalue { i1, i64 } %1145, 1
  br i1 %1146, label %for.cond.i.i1.i4430.preheader, label %Flow7916

for.cond.i.i1.i4430.preheader:                    ; preds = %for.cond.cleanup
  br label %for.cond.i.i1.i4430, !amdgpu.uniform !42

for.cond.i.i1.i3271.preheader:                    ; preds = %Flow7916
  br label %for.cond.i.i1.i3271, !amdgpu.uniform !42

Flow8431:                                         ; preds = %Flow8430, %if.end16
  %1148 = phi i32 [ %.lcssa9265, %Flow8430 ], [ %1, %if.end16 ]
  %1149 = phi i32 [ %.lcssa9264, %Flow8430 ], [ 0, %if.end16 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %1102)
  br label %for.cond.cleanup, !amdgpu.uniform !42

for.body:                                         ; preds = %for.body.lr.ph, %cleanup
  %phi.broken8766 = phi i64 [ 0, %for.body.lr.ph ], [ %4454, %cleanup ]
  %1150 = phi i32 [ %1404, %cleanup ], [ poison, %for.body.lr.ph ]
  %1151 = phi i64 [ %1403, %cleanup ], [ poison, %for.body.lr.ph ]
  %1152 = phi i32 [ %1402, %cleanup ], [ poison, %for.body.lr.ph ]
  %1153 = phi i32 [ %1401, %cleanup ], [ poison, %for.body.lr.ph ]
  %1154 = phi i64 [ %1400, %cleanup ], [ poison, %for.body.lr.ph ]
  %1155 = phi i32 [ %1399, %cleanup ], [ poison, %for.body.lr.ph ]
  %1156 = phi i32 [ %1398, %cleanup ], [ poison, %for.body.lr.ph ]
  %1157 = phi i64 [ %1397, %cleanup ], [ poison, %for.body.lr.ph ]
  %1158 = phi i32 [ %1396, %cleanup ], [ poison, %for.body.lr.ph ]
  %1159 = phi i32 [ %1395, %cleanup ], [ poison, %for.body.lr.ph ]
  %1160 = phi i64 [ %1394, %cleanup ], [ poison, %for.body.lr.ph ]
  %1161 = phi i32 [ %1393, %cleanup ], [ poison, %for.body.lr.ph ]
  %1162 = phi i32 [ %1392, %cleanup ], [ poison, %for.body.lr.ph ]
  %1163 = phi i64 [ %1391, %cleanup ], [ poison, %for.body.lr.ph ]
  %1164 = phi i32 [ %1390, %cleanup ], [ poison, %for.body.lr.ph ]
  %1165 = phi i32 [ %1389, %cleanup ], [ poison, %for.body.lr.ph ]
  %1166 = phi i64 [ %1388, %cleanup ], [ poison, %for.body.lr.ph ]
  %1167 = phi i32 [ %1387, %cleanup ], [ poison, %for.body.lr.ph ]
  %1168 = phi i32 [ %1386, %cleanup ], [ poison, %for.body.lr.ph ]
  %1169 = phi i64 [ %1385, %cleanup ], [ poison, %for.body.lr.ph ]
  %1170 = phi i32 [ %1384, %cleanup ], [ poison, %for.body.lr.ph ]
  %1171 = phi i32 [ %1383, %cleanup ], [ poison, %for.body.lr.ph ]
  %1172 = phi i64 [ %1382, %cleanup ], [ poison, %for.body.lr.ph ]
  %1173 = phi i32 [ %1381, %cleanup ], [ poison, %for.body.lr.ph ]
  %1174 = phi i32 [ %1380, %cleanup ], [ poison, %for.body.lr.ph ]
  %1175 = phi i64 [ %1379, %cleanup ], [ poison, %for.body.lr.ph ]
  %1176 = phi i32 [ %1378, %cleanup ], [ poison, %for.body.lr.ph ]
  %1177 = phi i32 [ %1377, %cleanup ], [ poison, %for.body.lr.ph ]
  %1178 = phi i64 [ %1376, %cleanup ], [ poison, %for.body.lr.ph ]
  %1179 = phi i32 [ %1375, %cleanup ], [ poison, %for.body.lr.ph ]
  %1180 = phi i32 [ %1374, %cleanup ], [ poison, %for.body.lr.ph ]
  %1181 = phi i64 [ %1373, %cleanup ], [ poison, %for.body.lr.ph ]
  %1182 = phi i32 [ %1372, %cleanup ], [ poison, %for.body.lr.ph ]
  %1183 = phi i32 [ %1371, %cleanup ], [ poison, %for.body.lr.ph ]
  %1184 = phi i64 [ %1370, %cleanup ], [ poison, %for.body.lr.ph ]
  %1185 = phi i32 [ %1369, %cleanup ], [ poison, %for.body.lr.ph ]
  %1186 = phi i32 [ %1368, %cleanup ], [ poison, %for.body.lr.ph ]
  %1187 = phi i64 [ %1367, %cleanup ], [ poison, %for.body.lr.ph ]
  %1188 = phi i32 [ %1366, %cleanup ], [ poison, %for.body.lr.ph ]
  %1189 = phi i32 [ %1365, %cleanup ], [ poison, %for.body.lr.ph ]
  %1190 = phi i64 [ %1364, %cleanup ], [ poison, %for.body.lr.ph ]
  %1191 = phi i32 [ %1363, %cleanup ], [ poison, %for.body.lr.ph ]
  %1192 = phi i32 [ %1362, %cleanup ], [ poison, %for.body.lr.ph ]
  %1193 = phi i64 [ %1361, %cleanup ], [ poison, %for.body.lr.ph ]
  %1194 = phi i32 [ %1360, %cleanup ], [ poison, %for.body.lr.ph ]
  %1195 = phi i32 [ %1359, %cleanup ], [ poison, %for.body.lr.ph ]
  %1196 = phi i64 [ %1358, %cleanup ], [ poison, %for.body.lr.ph ]
  %1197 = phi i32 [ %1357, %cleanup ], [ poison, %for.body.lr.ph ]
  %1198 = phi i32 [ %1356, %cleanup ], [ poison, %for.body.lr.ph ]
  %1199 = phi i64 [ %1355, %cleanup ], [ poison, %for.body.lr.ph ]
  %1200 = phi i32 [ %1354, %cleanup ], [ poison, %for.body.lr.ph ]
  %1201 = phi i32 [ %1353, %cleanup ], [ poison, %for.body.lr.ph ]
  %1202 = phi i64 [ %1352, %cleanup ], [ poison, %for.body.lr.ph ]
  %1203 = phi i32 [ %1351, %cleanup ], [ poison, %for.body.lr.ph ]
  %1204 = phi i32 [ %1350, %cleanup ], [ poison, %for.body.lr.ph ]
  %1205 = phi i64 [ %1349, %cleanup ], [ poison, %for.body.lr.ph ]
  %1206 = phi i32 [ %1348, %cleanup ], [ poison, %for.body.lr.ph ]
  %1207 = phi i32 [ %1347, %cleanup ], [ poison, %for.body.lr.ph ]
  %1208 = phi i64 [ %1346, %cleanup ], [ poison, %for.body.lr.ph ]
  %1209 = phi i32 [ %1345, %cleanup ], [ poison, %for.body.lr.ph ]
  %1210 = phi i32 [ %1344, %cleanup ], [ poison, %for.body.lr.ph ]
  %1211 = phi i64 [ %1343, %cleanup ], [ poison, %for.body.lr.ph ]
  %1212 = phi i32 [ %1342, %cleanup ], [ poison, %for.body.lr.ph ]
  %1213 = phi i32 [ %1341, %cleanup ], [ poison, %for.body.lr.ph ]
  %1214 = phi i64 [ %1340, %cleanup ], [ poison, %for.body.lr.ph ]
  %1215 = phi i32 [ %1339, %cleanup ], [ poison, %for.body.lr.ph ]
  %T.0207 = phi ptr [ %T.0, %cleanup ], [ %7, %for.body.lr.ph ]
  %TestCount.0206 = phi i32 [ %1337, %cleanup ], [ %1, %for.body.lr.ph ]
  %FailCount.0205 = phi i32 [ %1338, %cleanup ], [ 0, %for.body.lr.ph ]
  %vtable = load ptr addrspace(1), ptr %T.0207, align 8, !tbaa !158
  %vfn = getelementptr inbounds nuw i8, ptr addrspace(1) %vtable, i64 40
  %1216 = load ptr addrspace(1), ptr addrspace(1) %vfn, align 8
  %call18 = call noundef addrspace(1) ptr %1216(ptr noundef nonnull align 8 dereferenceable(24) %T.0207) #33
  %1217 = load ptr, ptr %Options, align 8, !tbaa !160
  %tobool19.not.not = icmp ne ptr %1217, null
  %1218 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %tobool19.not.not)
  %1219 = extractvalue { i1, i64 } %1218, 0
  %1220 = extractvalue { i1, i64 } %1218, 1
  br i1 %1219, label %land.rhs, label %Flow8427

Flow8428:                                         ; preds = %if.then35, %land.end
  %1221 = phi i32 [ %dec, %if.then35 ], [ poison, %land.end ]
  %1222 = phi i1 [ false, %if.then35 ], [ true, %land.end ]
  call void @llvm.amdgcn.end.cf.i64(i64 %1336)
  br label %Flow8427, !amdgpu.uniform !42

for.cond.i.i.i748.preheader:                      ; preds = %Flow8427
  br label %for.cond.i.i.i748, !amdgpu.uniform !42

land.rhs:                                         ; preds = %for.body
  %1223 = load i8, ptr %call18, align 1, !tbaa !57
  %tobool.not3.i.i = icmp ne i8 %1223, 0
  %1224 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %tobool.not3.i.i)
  %1225 = extractvalue { i1, i64 } %1224, 0
  %1226 = extractvalue { i1, i64 } %1224, 1
  br i1 %1225, label %for.inc.i.i.preheader, label %Flow7937

for.inc.i.i.preheader:                            ; preds = %land.rhs
  %scevgep6568 = getelementptr i8, ptr %call18, i64 1
  br label %for.inc.i.i, !amdgpu.uniform !42

Flow7937:                                         ; preds = %Flow7936, %land.rhs
  %1227 = phi i64 [ %inc.i.i.lcssa, %Flow7936 ], [ 0, %land.rhs ]
  call void @llvm.amdgcn.end.cf.i64(i64 %1226)
  br label %_ZN22__llvm_libc_22_0_0_git8internal13string_lengthIcEEmPKT_.exit.i, !amdgpu.uniform !42

for.inc.i.i:                                      ; preds = %for.inc.i.i.preheader, %for.inc.i.i
  %phi.broken8646 = phi i64 [ 0, %for.inc.i.i.preheader ], [ %1229, %for.inc.i.i ]
  %length.05.i.i = phi i64 [ %inc.i.i, %for.inc.i.i ], [ 0, %for.inc.i.i.preheader ]
  %scevgep6569 = getelementptr i8, ptr %scevgep6568, i64 %length.05.i.i
  %inc.i.i = add i64 %length.05.i.i, 1
  %1228 = load i8, ptr %scevgep6569, align 1, !tbaa !57
  %tobool.not.i.i = icmp eq i8 %1228, 0
  %1229 = call i64 @llvm.amdgcn.if.break.i64(i1 %tobool.not.i.i, i64 %phi.broken8646)
  %1230 = call i1 @llvm.amdgcn.loop.i64(i64 %1229)
  br i1 %1230, label %Flow7936, label %for.inc.i.i

Flow7936:                                         ; preds = %for.inc.i.i
  %inc.i.i.lcssa = phi i64 [ %inc.i.i, %for.inc.i.i ]
  %.lcssa8968 = phi i64 [ %1229, %for.inc.i.i ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa8968)
  br label %Flow7937, !amdgpu.uniform !42

Flow8427:                                         ; preds = %Flow8428, %for.body
  %1231 = phi i32 [ %1221, %Flow8428 ], [ poison, %for.body ]
  %1232 = phi i1 [ %1222, %Flow8428 ], [ true, %for.body ]
  call void @llvm.amdgcn.end.cf.i64(i64 %1220)
  %1233 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %1232)
  %1234 = extractvalue { i1, i64 } %1233, 0
  %1235 = extractvalue { i1, i64 } %1233, 1
  br i1 %1234, label %for.cond.i.i.i748.preheader, label %Flow8429

_ZN22__llvm_libc_22_0_0_git8internal13string_lengthIcEEmPKT_.exit.i: ; preds = %Flow7937
  %cmp.i.not.i.i = icmp eq i64 %1227, 0
  %cmp.i.not.i.i.inv = xor i1 %cmp.i.not.i.i, true
  %1236 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.i.not.i.i.inv)
  %1237 = extractvalue { i1, i64 } %1236, 0
  %1238 = extractvalue { i1, i64 } %1236, 1
  br i1 %1237, label %if.then.i.i.i, label %Flow7935

if.then.i.i.i:                                    ; preds = %_ZN22__llvm_libc_22_0_0_git8internal13string_lengthIcEEmPKT_.exit.i
  %1239 = call { i64, i1 } @llvm.uadd.with.overflow.i64(i64 %1227, i64 1)
  %ov7387 = extractvalue { i64, i1 } %1239, 1
  %ov7387.inv = xor i1 %ov7387, true
  %1240 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %ov7387.inv)
  %1241 = extractvalue { i1, i64 } %1240, 0
  %1242 = extractvalue { i1, i64 } %1240, 1
  br i1 %1241, label %if.end.i.i.i.i, label %Flow7934

if.end.i.i.i.i:                                   ; preds = %if.then.i.i.i
  %math7386 = extractvalue { i64, i1 } %1239, 0
  %cmp2.i.i.i.i = icmp ult i64 %math7386, 1676976733973595601
  %1243 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp2.i.i.i.i)
  %1244 = extractvalue { i1, i64 } %1243, 0
  %1245 = extractvalue { i1, i64 } %1243, 1
  br i1 %1244, label %if.then3.i.i.i.i, label %if.end4.i.i.i.i

if.then3.i.i.i.i:                                 ; preds = %if.end.i.i.i.i
  %mul.i.i.i.i = mul nuw i64 %math7386, 11
  %div18.i.i.i.i = lshr i64 %mul.i.i.i.i, 3
  br label %if.end4.i.i.i.i, !amdgpu.uniform !42

Flow7934:                                         ; preds = %if.end4.i.i.i.i, %if.then.i.i.i
  %1246 = phi ptr [ %cond.i.i, %if.end4.i.i.i.i ], [ addrspacecast (ptr addrspace(4) @_ZN22__llvm_libc_22_0_0_git3cpp6string14NULL_CHARACTERE to ptr), %if.then.i.i.i ]
  call void @llvm.amdgcn.end.cf.i64(i64 %1242)
  br label %_ZN22__llvm_libc_22_0_0_git3cpp6string7reserveEm.exit.i.i.i, !amdgpu.uniform !42

if.end4.i.i.i.i:                                  ; preds = %if.then3.i.i.i.i, %if.end.i.i.i.i
  %new_capacity.addr.0.i.i.i.i = phi i64 [ %div18.i.i.i.i, %if.then3.i.i.i.i ], [ %math7386, %if.end.i.i.i.i ]
  call void @llvm.amdgcn.end.cf.i64(i64 %1245)
  %sub.i.i = add i64 %new_capacity.addr.0.i.i.i.i, 7
  %div2.i.i = and i64 %sub.i.i, -8
  %1247 = load ptr, ptr addrspace(1) @_ZN12_GLOBAL__N_13ptrE, align 8, !tbaa !55
  %add.ptr.i.i = getelementptr inbounds nuw i8, ptr %1247, i64 %div2.i.i
  store ptr %add.ptr.i.i, ptr addrspace(1) @_ZN12_GLOBAL__N_13ptrE, align 8, !tbaa !55
  %sub.ptr.lhs.cast.i.i = ptrtoint ptr %add.ptr.i.i to i64
  %sub.ptr.sub.i.i = sub i64 %sub.ptr.lhs.cast.i.i, ptrtoint (ptr addrspacecast (ptr addrspace(1) @_ZN12_GLOBAL__N_16memoryE to ptr) to i64)
  %cmp.i.i = icmp ugt i64 %sub.ptr.sub.i.i, 65335
  %cond.i.i = select i1 %cmp.i.i, ptr null, ptr %1247
  br label %Flow7934, !amdgpu.uniform !42

Flow7935:                                         ; preds = %Flow7933, %_ZN22__llvm_libc_22_0_0_git8internal13string_lengthIcEEmPKT_.exit.i
  %1248 = phi ptr [ %1246, %Flow7933 ], [ addrspacecast (ptr addrspace(4) @_ZN22__llvm_libc_22_0_0_git3cpp6string14NULL_CHARACTERE to ptr), %_ZN22__llvm_libc_22_0_0_git8internal13string_lengthIcEEmPKT_.exit.i ]
  call void @llvm.amdgcn.end.cf.i64(i64 %1238)
  br label %if.end.i.i.i, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git3cpp6string7reserveEm.exit.i.i.i: ; preds = %Flow7934
  br label %loadstoreloop4963, !amdgpu.uniform !42

loadstoreloop4963:                                ; preds = %_ZN22__llvm_libc_22_0_0_git3cpp6string7reserveEm.exit.i.i.i, %loadstoreloop4963
  %phi.broken8647 = phi i64 [ 0, %_ZN22__llvm_libc_22_0_0_git3cpp6string7reserveEm.exit.i.i.i ], [ %1253, %loadstoreloop4963 ]
  %1249 = phi i64 [ %1251, %loadstoreloop4963 ], [ 0, %_ZN22__llvm_libc_22_0_0_git3cpp6string7reserveEm.exit.i.i.i ]
  %1250 = getelementptr i8, ptr %1246, i64 %1249
  store i8 0, ptr %1250, align 1
  %1251 = add i64 %1249, 1
  %1252 = icmp uge i64 %1251, %1227
  %1253 = call i64 @llvm.amdgcn.if.break.i64(i1 %1252, i64 %phi.broken8647)
  %1254 = call i1 @llvm.amdgcn.loop.i64(i64 %1253)
  br i1 %1254, label %Flow7933, label %loadstoreloop4963

Flow7933:                                         ; preds = %loadstoreloop4963
  %.lcssa8969 = phi i64 [ %1253, %loadstoreloop4963 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa8969)
  br label %Flow7935, !amdgpu.uniform !42

if.end.i.i.i:                                     ; preds = %Flow7935
  %cmp.not.i6.i.i.i = icmp ne ptr %1248, addrspacecast (ptr addrspace(4) @_ZN22__llvm_libc_22_0_0_git3cpp6string14NULL_CHARACTERE to ptr)
  %1255 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.not.i6.i.i.i)
  %1256 = extractvalue { i1, i64 } %1255, 0
  %1257 = extractvalue { i1, i64 } %1255, 1
  br i1 %1256, label %if.then.i.i.i.i, label %_ZN22__llvm_libc_22_0_0_git3cpp6stringC2EPKc.exit

if.then.i.i.i.i:                                  ; preds = %if.end.i.i.i
  %arrayidx.i.i.i.i = getelementptr inbounds nuw i8, ptr %1248, i64 %1227
  store i8 0, ptr %arrayidx.i.i.i.i, align 1, !tbaa !57
  br label %_ZN22__llvm_libc_22_0_0_git3cpp6stringC2EPKc.exit, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git3cpp6stringC2EPKc.exit: ; preds = %if.then.i.i.i.i, %if.end.i.i.i
  call void @llvm.amdgcn.end.cf.i64(i64 %1257)
  %1258 = and i64 %1227, 15
  %1259 = sub i64 %1227, %1258
  %1260 = icmp ne i64 %1259, 0
  %1261 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %1260)
  %1262 = extractvalue { i1, i64 } %1261, 0
  %1263 = extractvalue { i1, i64 } %1261, 1
  br i1 %1262, label %loop-memcpy-expansion5097.preheader, label %Flow7932

loop-memcpy-expansion5097.preheader:              ; preds = %_ZN22__llvm_libc_22_0_0_git3cpp6stringC2EPKc.exit
  br label %loop-memcpy-expansion5097, !amdgpu.uniform !42

Flow7932:                                         ; preds = %Flow7931, %_ZN22__llvm_libc_22_0_0_git3cpp6stringC2EPKc.exit
  call void @llvm.amdgcn.end.cf.i64(i64 %1263)
  br label %loop-memcpy-residual-header5100, !amdgpu.uniform !42

loop-memcpy-expansion5097:                        ; preds = %loop-memcpy-expansion5097.preheader, %loop-memcpy-expansion5097
  %phi.broken8648 = phi i64 [ 0, %loop-memcpy-expansion5097.preheader ], [ %1269, %loop-memcpy-expansion5097 ]
  %loop-index5098 = phi i64 [ %1267, %loop-memcpy-expansion5097 ], [ 0, %loop-memcpy-expansion5097.preheader ]
  %1264 = getelementptr i8, ptr %call18, i64 %loop-index5098
  %1265 = load <4 x i32>, ptr %1264, align 1
  %1266 = getelementptr i8, ptr %1248, i64 %loop-index5098
  store <4 x i32> %1265, ptr %1266, align 1
  %1267 = add i64 %loop-index5098, 16
  %1268 = icmp uge i64 %1267, %1259
  %1269 = call i64 @llvm.amdgcn.if.break.i64(i1 %1268, i64 %phi.broken8648)
  %1270 = call i1 @llvm.amdgcn.loop.i64(i64 %1269)
  br i1 %1270, label %Flow7931, label %loop-memcpy-expansion5097

Flow7930:                                         ; preds = %Flow7929, %loop-memcpy-residual-header5100
  call void @llvm.amdgcn.end.cf.i64(i64 %6359)
  br label %post-loop-memcpy-expansion5096, !amdgpu.uniform !42

loop-memcpy-residual5099:                         ; preds = %loop-memcpy-residual5099.preheader, %loop-memcpy-residual5099
  %phi.broken8649 = phi i64 [ 0, %loop-memcpy-residual5099.preheader ], [ %1274, %loop-memcpy-residual5099 ]
  %residual-loop-index5101 = phi i64 [ %1272, %loop-memcpy-residual5099 ], [ 0, %loop-memcpy-residual5099.preheader ]
  %scevgep6573 = getelementptr i8, ptr %scevgep6572, i64 %residual-loop-index5101
  %1271 = load i8, ptr %scevgep6573, align 1
  %scevgep6571 = getelementptr i8, ptr %scevgep6570, i64 %residual-loop-index5101
  store i8 %1271, ptr %scevgep6571, align 1
  %1272 = add i64 %residual-loop-index5101, 1
  %1273 = icmp uge i64 %1272, %1258
  %1274 = call i64 @llvm.amdgcn.if.break.i64(i1 %1273, i64 %phi.broken8649)
  %1275 = call i1 @llvm.amdgcn.loop.i64(i64 %1274)
  br i1 %1275, label %Flow7929, label %loop-memcpy-residual5099

Flow7929:                                         ; preds = %loop-memcpy-residual5099
  %.lcssa8971 = phi i64 [ %1274, %loop-memcpy-residual5099 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa8971)
  br label %Flow7930, !amdgpu.uniform !42

post-loop-memcpy-expansion5096:                   ; preds = %Flow7930
  %1276 = load ptr, ptr %Options, align 8, !tbaa !160
  %1277 = load i8, ptr %1276, align 1, !tbaa !57
  %tobool.not3.i.i158 = icmp ne i8 %1277, 0
  %1278 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %tobool.not3.i.i158)
  %1279 = extractvalue { i1, i64 } %1278, 0
  %1280 = extractvalue { i1, i64 } %1278, 1
  br i1 %1279, label %for.inc.i.i159.preheader, label %Flow7928

for.inc.i.i159.preheader:                         ; preds = %post-loop-memcpy-expansion5096
  %scevgep6574 = getelementptr i8, ptr %1276, i64 1
  br label %for.inc.i.i159, !amdgpu.uniform !42

Flow7928:                                         ; preds = %Flow7927, %post-loop-memcpy-expansion5096
  %1281 = phi i64 [ %inc.i.i163.lcssa, %Flow7927 ], [ 0, %post-loop-memcpy-expansion5096 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %1280)
  br label %_ZN22__llvm_libc_22_0_0_git8internal13string_lengthIcEEmPKT_.exit.i165, !amdgpu.uniform !42

for.inc.i.i159:                                   ; preds = %for.inc.i.i159.preheader, %for.inc.i.i159
  %phi.broken8650 = phi i64 [ 0, %for.inc.i.i159.preheader ], [ %1283, %for.inc.i.i159 ]
  %length.05.i.i160 = phi i64 [ %inc.i.i163, %for.inc.i.i159 ], [ 0, %for.inc.i.i159.preheader ]
  %scevgep6575 = getelementptr i8, ptr %scevgep6574, i64 %length.05.i.i160
  %inc.i.i163 = add i64 %length.05.i.i160, 1
  %1282 = load i8, ptr %scevgep6575, align 1, !tbaa !57
  %tobool.not.i.i164 = icmp eq i8 %1282, 0
  %1283 = call i64 @llvm.amdgcn.if.break.i64(i1 %tobool.not.i.i164, i64 %phi.broken8650)
  %1284 = call i1 @llvm.amdgcn.loop.i64(i64 %1283)
  br i1 %1284, label %Flow7927, label %for.inc.i.i159

Flow7927:                                         ; preds = %for.inc.i.i159
  %inc.i.i163.lcssa = phi i64 [ %inc.i.i163, %for.inc.i.i159 ]
  %.lcssa8972 = phi i64 [ %1283, %for.inc.i.i159 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa8972)
  br label %Flow7928, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git8internal13string_lengthIcEEmPKT_.exit.i165: ; preds = %Flow7928
  %cmp.i.not.i.i169 = icmp ne i64 %1281, 0
  %1285 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.i.not.i.i169)
  %1286 = extractvalue { i1, i64 } %1285, 0
  %1287 = extractvalue { i1, i64 } %1285, 1
  br i1 %1286, label %if.then.i.i.i170, label %Flow7926

if.then.i.i.i170:                                 ; preds = %_ZN22__llvm_libc_22_0_0_git8internal13string_lengthIcEEmPKT_.exit.i165
  %1288 = call { i64, i1 } @llvm.uadd.with.overflow.i64(i64 %1281, i64 1)
  %ov7389 = extractvalue { i64, i1 } %1288, 1
  %ov7389.inv = xor i1 %ov7389, true
  %1289 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %ov7389.inv)
  %1290 = extractvalue { i1, i64 } %1289, 0
  %1291 = extractvalue { i1, i64 } %1289, 1
  br i1 %1290, label %if.end.i.i.i.i173, label %Flow7925

if.end.i.i.i.i173:                                ; preds = %if.then.i.i.i170
  %math7388 = extractvalue { i64, i1 } %1288, 0
  %cmp2.i.i.i.i174 = icmp ult i64 %math7388, 1676976733973595601
  %1292 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp2.i.i.i.i174)
  %1293 = extractvalue { i1, i64 } %1292, 0
  %1294 = extractvalue { i1, i64 } %1292, 1
  br i1 %1293, label %if.then3.i.i.i.i189, label %if.end4.i.i.i.i175

if.then3.i.i.i.i189:                              ; preds = %if.end.i.i.i.i173
  %mul.i.i.i.i190 = mul nuw i64 %math7388, 11
  %div18.i.i.i.i191 = lshr i64 %mul.i.i.i.i190, 3
  br label %if.end4.i.i.i.i175, !amdgpu.uniform !42

Flow7925:                                         ; preds = %if.end4.i.i.i.i175, %if.then.i.i.i170
  %1295 = phi ptr [ %cond.i.i747, %if.end4.i.i.i.i175 ], [ addrspacecast (ptr addrspace(4) @_ZN22__llvm_libc_22_0_0_git3cpp6string14NULL_CHARACTERE to ptr), %if.then.i.i.i170 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %1291)
  br label %_ZN22__llvm_libc_22_0_0_git3cpp6string7reserveEm.exit.i.i.i180, !amdgpu.uniform !42

if.end4.i.i.i.i175:                               ; preds = %if.then3.i.i.i.i189, %if.end.i.i.i.i173
  %new_capacity.addr.0.i.i.i.i176 = phi i64 [ %div18.i.i.i.i191, %if.then3.i.i.i.i189 ], [ %math7388, %if.end.i.i.i.i173 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %1294)
  %sub.i.i740 = add i64 %new_capacity.addr.0.i.i.i.i176, 7
  %div2.i.i741 = and i64 %sub.i.i740, -8
  %1296 = load ptr, ptr addrspace(1) @_ZN12_GLOBAL__N_13ptrE, align 8, !tbaa !55
  %add.ptr.i.i742 = getelementptr inbounds nuw i8, ptr %1296, i64 %div2.i.i741
  store ptr %add.ptr.i.i742, ptr addrspace(1) @_ZN12_GLOBAL__N_13ptrE, align 8, !tbaa !55
  %sub.ptr.lhs.cast.i.i744 = ptrtoint ptr %add.ptr.i.i742 to i64
  %sub.ptr.sub.i.i745 = sub i64 %sub.ptr.lhs.cast.i.i744, ptrtoint (ptr addrspacecast (ptr addrspace(1) @_ZN12_GLOBAL__N_16memoryE to ptr) to i64)
  %cmp.i.i746 = icmp ugt i64 %sub.ptr.sub.i.i745, 65335
  %cond.i.i747 = select i1 %cmp.i.i746, ptr null, ptr %1296
  br label %Flow7925, !amdgpu.uniform !42

Flow7926:                                         ; preds = %Flow7924, %_ZN22__llvm_libc_22_0_0_git8internal13string_lengthIcEEmPKT_.exit.i165
  %1297 = phi ptr [ %1295, %Flow7924 ], [ addrspacecast (ptr addrspace(4) @_ZN22__llvm_libc_22_0_0_git3cpp6string14NULL_CHARACTERE to ptr), %_ZN22__llvm_libc_22_0_0_git8internal13string_lengthIcEEmPKT_.exit.i165 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %1287)
  br label %if.end.i.i.i184, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git3cpp6string7reserveEm.exit.i.i.i180: ; preds = %Flow7925
  br label %loadstoreloop4961, !amdgpu.uniform !42

loadstoreloop4961:                                ; preds = %_ZN22__llvm_libc_22_0_0_git3cpp6string7reserveEm.exit.i.i.i180, %loadstoreloop4961
  %phi.broken8651 = phi i64 [ 0, %_ZN22__llvm_libc_22_0_0_git3cpp6string7reserveEm.exit.i.i.i180 ], [ %1302, %loadstoreloop4961 ]
  %1298 = phi i64 [ %1300, %loadstoreloop4961 ], [ 0, %_ZN22__llvm_libc_22_0_0_git3cpp6string7reserveEm.exit.i.i.i180 ]
  %1299 = getelementptr i8, ptr %1295, i64 %1298
  store i8 0, ptr %1299, align 1
  %1300 = add i64 %1298, 1
  %1301 = icmp uge i64 %1300, %1281
  %1302 = call i64 @llvm.amdgcn.if.break.i64(i1 %1301, i64 %phi.broken8651)
  %1303 = call i1 @llvm.amdgcn.loop.i64(i64 %1302)
  br i1 %1303, label %Flow7924, label %loadstoreloop4961

Flow7924:                                         ; preds = %loadstoreloop4961
  %.lcssa8973 = phi i64 [ %1302, %loadstoreloop4961 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa8973)
  br label %Flow7926, !amdgpu.uniform !42

if.end.i.i.i184:                                  ; preds = %Flow7926
  %cmp.not.i6.i.i.i185 = icmp ne ptr %1297, addrspacecast (ptr addrspace(4) @_ZN22__llvm_libc_22_0_0_git3cpp6string14NULL_CHARACTERE to ptr)
  %1304 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.not.i6.i.i.i185)
  %1305 = extractvalue { i1, i64 } %1304, 0
  %1306 = extractvalue { i1, i64 } %1304, 1
  br i1 %1305, label %if.then.i.i.i.i186, label %_ZN22__llvm_libc_22_0_0_git3cpp6stringC2EPKc.exit192

if.then.i.i.i.i186:                               ; preds = %if.end.i.i.i184
  %arrayidx.i.i.i.i187 = getelementptr inbounds nuw i8, ptr %1297, i64 %1281
  store i8 0, ptr %arrayidx.i.i.i.i187, align 1, !tbaa !57
  br label %_ZN22__llvm_libc_22_0_0_git3cpp6stringC2EPKc.exit192, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git3cpp6stringC2EPKc.exit192: ; preds = %if.then.i.i.i.i186, %if.end.i.i.i184
  call void @llvm.amdgcn.end.cf.i64(i64 %1306)
  %1307 = and i64 %1281, 15
  %1308 = sub i64 %1281, %1307
  %1309 = icmp ne i64 %1308, 0
  %1310 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %1309)
  %1311 = extractvalue { i1, i64 } %1310, 0
  %1312 = extractvalue { i1, i64 } %1310, 1
  br i1 %1311, label %loop-memcpy-expansion5091.preheader, label %Flow7923

loop-memcpy-expansion5091.preheader:              ; preds = %_ZN22__llvm_libc_22_0_0_git3cpp6stringC2EPKc.exit192
  br label %loop-memcpy-expansion5091, !amdgpu.uniform !42

Flow7923:                                         ; preds = %Flow7922, %_ZN22__llvm_libc_22_0_0_git3cpp6stringC2EPKc.exit192
  call void @llvm.amdgcn.end.cf.i64(i64 %1312)
  br label %loop-memcpy-residual-header5094, !amdgpu.uniform !42

loop-memcpy-expansion5091:                        ; preds = %loop-memcpy-expansion5091.preheader, %loop-memcpy-expansion5091
  %phi.broken8652 = phi i64 [ 0, %loop-memcpy-expansion5091.preheader ], [ %1318, %loop-memcpy-expansion5091 ]
  %loop-index5092 = phi i64 [ %1316, %loop-memcpy-expansion5091 ], [ 0, %loop-memcpy-expansion5091.preheader ]
  %1313 = getelementptr i8, ptr %1276, i64 %loop-index5092
  %1314 = load <4 x i32>, ptr %1313, align 1
  %1315 = getelementptr i8, ptr %1297, i64 %loop-index5092
  store <4 x i32> %1314, ptr %1315, align 1
  %1316 = add i64 %loop-index5092, 16
  %1317 = icmp uge i64 %1316, %1308
  %1318 = call i64 @llvm.amdgcn.if.break.i64(i1 %1317, i64 %phi.broken8652)
  %1319 = call i1 @llvm.amdgcn.loop.i64(i64 %1318)
  br i1 %1319, label %Flow7922, label %loop-memcpy-expansion5091

Flow7921:                                         ; preds = %Flow7920, %loop-memcpy-residual-header5094
  call void @llvm.amdgcn.end.cf.i64(i64 %6355)
  br label %post-loop-memcpy-expansion5090, !amdgpu.uniform !42

loop-memcpy-residual5093:                         ; preds = %loop-memcpy-residual5093.preheader, %loop-memcpy-residual5093
  %phi.broken8653 = phi i64 [ 0, %loop-memcpy-residual5093.preheader ], [ %1323, %loop-memcpy-residual5093 ]
  %residual-loop-index5095 = phi i64 [ %1321, %loop-memcpy-residual5093 ], [ 0, %loop-memcpy-residual5093.preheader ]
  %scevgep6579 = getelementptr i8, ptr %scevgep6578, i64 %residual-loop-index5095
  %1320 = load i8, ptr %scevgep6579, align 1
  %scevgep6577 = getelementptr i8, ptr %scevgep6576, i64 %residual-loop-index5095
  store i8 %1320, ptr %scevgep6577, align 1
  %1321 = add i64 %residual-loop-index5095, 1
  %1322 = icmp uge i64 %1321, %1307
  %1323 = call i64 @llvm.amdgcn.if.break.i64(i1 %1322, i64 %phi.broken8653)
  %1324 = call i1 @llvm.amdgcn.loop.i64(i64 %1323)
  br i1 %1324, label %Flow7920, label %loop-memcpy-residual5093

Flow7920:                                         ; preds = %loop-memcpy-residual5093
  %.lcssa8975 = phi i64 [ %1323, %loop-memcpy-residual5093 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa8975)
  br label %Flow7921, !amdgpu.uniform !42

post-loop-memcpy-expansion5090:                   ; preds = %Flow7921
  %cmp.i.i.i.i = icmp ne i64 %1227, %1281
  %brmerge = or i1 %cmp.i.not.i.i, %cmp.i.i.i.i
  %brmerge.inv = xor i1 %brmerge, true
  %1325 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %brmerge.inv)
  %1326 = extractvalue { i1, i64 } %1325, 0
  %1327 = extractvalue { i1, i64 } %1325, 1
  br i1 %1326, label %for.body.i.i.i.i.i.preheader, label %Flow7919

for.body.i.i.i.i.i.preheader:                     ; preds = %post-loop-memcpy-expansion5090
  %1328 = add i64 %1227, -1
  br label %for.body.i.i.i.i.i, !amdgpu.uniform !42

Flow7919:                                         ; preds = %Flow7918, %post-loop-memcpy-expansion5090
  %1329 = phi i1 [ %tobool.not.i.i.i.i.i.lcssa, %Flow7918 ], [ %cmp.i.i.i.i, %post-loop-memcpy-expansion5090 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %1327)
  br label %land.end, !amdgpu.uniform !42

for.body.i.i.i.i.i:                               ; preds = %for.body.i.i.i.i.i.preheader, %for.body.i.i.i.i.i
  %phi.broken8654 = phi i64 [ 0, %for.body.i.i.i.i.i.preheader ], [ %1332, %for.body.i.i.i.i.i ]
  %lsr.iv6584 = phi ptr [ %scevgep6585, %for.body.i.i.i.i.i ], [ %1248, %for.body.i.i.i.i.i.preheader ]
  %lsr.iv6582 = phi ptr [ %scevgep6583, %for.body.i.i.i.i.i ], [ %1297, %for.body.i.i.i.i.i.preheader ]
  %lsr.iv6580 = phi i64 [ %lsr.iv.next6581, %for.body.i.i.i.i.i ], [ %1328, %for.body.i.i.i.i.i.preheader ]
  %1330 = load i8, ptr %lsr.iv6584, align 1, !tbaa !57
  %1331 = load i8, ptr %lsr.iv6582, align 1, !tbaa !57
  %tobool.not.i.i.i.i.i = icmp ne i8 %1330, %1331
  %exitcond.not.i.i.i.i.i = icmp eq i64 %lsr.iv6580, 0
  %or.cond.i.i.i.i.i = select i1 %tobool.not.i.i.i.i.i, i1 true, i1 %exitcond.not.i.i.i.i.i
  %lsr.iv.next6581 = add i64 %lsr.iv6580, -1
  %scevgep6583 = getelementptr i8, ptr %lsr.iv6582, i64 1
  %scevgep6585 = getelementptr i8, ptr %lsr.iv6584, i64 1
  %1332 = call i64 @llvm.amdgcn.if.break.i64(i1 %or.cond.i.i.i.i.i, i64 %phi.broken8654)
  %1333 = call i1 @llvm.amdgcn.loop.i64(i64 %1332)
  br i1 %1333, label %Flow7918, label %for.body.i.i.i.i.i

Flow7918:                                         ; preds = %for.body.i.i.i.i.i
  %tobool.not.i.i.i.i.i.lcssa = phi i1 [ %tobool.not.i.i.i.i.i, %for.body.i.i.i.i.i ]
  %.lcssa8976 = phi i64 [ %1332, %for.body.i.i.i.i.i ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa8976)
  br label %Flow7919, !amdgpu.uniform !42

land.end:                                         ; preds = %Flow7919
  %1334 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %1329)
  %1335 = extractvalue { i1, i64 } %1334, 0
  %1336 = extractvalue { i1, i64 } %1334, 1
  br i1 %1335, label %if.then35, label %Flow8428

if.then35:                                        ; preds = %land.end
  %dec = add nsw i32 %TestCount.0206, -1
  br label %Flow8428, !amdgpu.uniform !42

Flow8429:                                         ; preds = %sw.epilog, %Flow8427
  %1337 = phi i32 [ %TestCount.0206, %sw.epilog ], [ %1231, %Flow8427 ]
  %1338 = phi i32 [ %2089, %sw.epilog ], [ %FailCount.0205, %Flow8427 ]
  %1339 = phi i32 [ %2090, %sw.epilog ], [ %1215, %Flow8427 ]
  %1340 = phi i64 [ %2091, %sw.epilog ], [ %1214, %Flow8427 ]
  %1341 = phi i32 [ %2092, %sw.epilog ], [ %1213, %Flow8427 ]
  %1342 = phi i32 [ %2093, %sw.epilog ], [ %1212, %Flow8427 ]
  %1343 = phi i64 [ %2094, %sw.epilog ], [ %1211, %Flow8427 ]
  %1344 = phi i32 [ %2095, %sw.epilog ], [ %1210, %Flow8427 ]
  %1345 = phi i32 [ %2096, %sw.epilog ], [ %1209, %Flow8427 ]
  %1346 = phi i64 [ %2097, %sw.epilog ], [ %1208, %Flow8427 ]
  %1347 = phi i32 [ %2098, %sw.epilog ], [ %1207, %Flow8427 ]
  %1348 = phi i32 [ %2099, %sw.epilog ], [ %1206, %Flow8427 ]
  %1349 = phi i64 [ %2100, %sw.epilog ], [ %1205, %Flow8427 ]
  %1350 = phi i32 [ %2101, %sw.epilog ], [ %1204, %Flow8427 ]
  %1351 = phi i32 [ %2102, %sw.epilog ], [ %1203, %Flow8427 ]
  %1352 = phi i64 [ %2103, %sw.epilog ], [ %1202, %Flow8427 ]
  %1353 = phi i32 [ %2104, %sw.epilog ], [ %1201, %Flow8427 ]
  %1354 = phi i32 [ %2105, %sw.epilog ], [ %1200, %Flow8427 ]
  %1355 = phi i64 [ %2106, %sw.epilog ], [ %1199, %Flow8427 ]
  %1356 = phi i32 [ %2107, %sw.epilog ], [ %1198, %Flow8427 ]
  %1357 = phi i32 [ %2108, %sw.epilog ], [ %1197, %Flow8427 ]
  %1358 = phi i64 [ %2109, %sw.epilog ], [ %1196, %Flow8427 ]
  %1359 = phi i32 [ %2110, %sw.epilog ], [ %1195, %Flow8427 ]
  %1360 = phi i32 [ %2111, %sw.epilog ], [ %1194, %Flow8427 ]
  %1361 = phi i64 [ %2112, %sw.epilog ], [ %1193, %Flow8427 ]
  %1362 = phi i32 [ %2113, %sw.epilog ], [ %1192, %Flow8427 ]
  %1363 = phi i32 [ %2114, %sw.epilog ], [ %1191, %Flow8427 ]
  %1364 = phi i64 [ %2115, %sw.epilog ], [ %1190, %Flow8427 ]
  %1365 = phi i32 [ %2116, %sw.epilog ], [ %1189, %Flow8427 ]
  %1366 = phi i32 [ %2117, %sw.epilog ], [ %1188, %Flow8427 ]
  %1367 = phi i64 [ %2118, %sw.epilog ], [ %1187, %Flow8427 ]
  %1368 = phi i32 [ %2119, %sw.epilog ], [ %1186, %Flow8427 ]
  %1369 = phi i32 [ %2120, %sw.epilog ], [ %1185, %Flow8427 ]
  %1370 = phi i64 [ %2121, %sw.epilog ], [ %1184, %Flow8427 ]
  %1371 = phi i32 [ %2122, %sw.epilog ], [ %1183, %Flow8427 ]
  %1372 = phi i32 [ %2123, %sw.epilog ], [ %1182, %Flow8427 ]
  %1373 = phi i64 [ %2124, %sw.epilog ], [ %1181, %Flow8427 ]
  %1374 = phi i32 [ %2125, %sw.epilog ], [ %1180, %Flow8427 ]
  %1375 = phi i32 [ %2126, %sw.epilog ], [ %1179, %Flow8427 ]
  %1376 = phi i64 [ %2127, %sw.epilog ], [ %1178, %Flow8427 ]
  %1377 = phi i32 [ %2128, %sw.epilog ], [ %1177, %Flow8427 ]
  %1378 = phi i32 [ %2129, %sw.epilog ], [ %1176, %Flow8427 ]
  %1379 = phi i64 [ %2130, %sw.epilog ], [ %1175, %Flow8427 ]
  %1380 = phi i32 [ %2131, %sw.epilog ], [ %1174, %Flow8427 ]
  %1381 = phi i32 [ %2132, %sw.epilog ], [ %1173, %Flow8427 ]
  %1382 = phi i64 [ %2133, %sw.epilog ], [ %1172, %Flow8427 ]
  %1383 = phi i32 [ %2134, %sw.epilog ], [ %1171, %Flow8427 ]
  %1384 = phi i32 [ %2135, %sw.epilog ], [ %1170, %Flow8427 ]
  %1385 = phi i64 [ %2136, %sw.epilog ], [ %1169, %Flow8427 ]
  %1386 = phi i32 [ %2137, %sw.epilog ], [ %1168, %Flow8427 ]
  %1387 = phi i32 [ %2138, %sw.epilog ], [ %1167, %Flow8427 ]
  %1388 = phi i64 [ %2139, %sw.epilog ], [ %1166, %Flow8427 ]
  %1389 = phi i32 [ %2140, %sw.epilog ], [ %1165, %Flow8427 ]
  %1390 = phi i32 [ %.lcssa9039, %sw.epilog ], [ %1164, %Flow8427 ]
  %1391 = phi i64 [ %.lcssa9040, %sw.epilog ], [ %1163, %Flow8427 ]
  %1392 = phi i32 [ %.lcssa9041, %sw.epilog ], [ %1162, %Flow8427 ]
  %1393 = phi i32 [ %.lcssa9024, %sw.epilog ], [ %1161, %Flow8427 ]
  %1394 = phi i64 [ %.lcssa9025, %sw.epilog ], [ %1160, %Flow8427 ]
  %1395 = phi i32 [ %.lcssa9026, %sw.epilog ], [ %1159, %Flow8427 ]
  %1396 = phi i32 [ %.lcssa9008, %sw.epilog ], [ %1158, %Flow8427 ]
  %1397 = phi i64 [ %.lcssa9009, %sw.epilog ], [ %1157, %Flow8427 ]
  %1398 = phi i32 [ %.lcssa9010, %sw.epilog ], [ %1156, %Flow8427 ]
  %1399 = phi i32 [ %.lcssa8996, %sw.epilog ], [ %1155, %Flow8427 ]
  %1400 = phi i64 [ %.lcssa8997, %sw.epilog ], [ %1154, %Flow8427 ]
  %1401 = phi i32 [ %.lcssa8998, %sw.epilog ], [ %1153, %Flow8427 ]
  %1402 = phi i32 [ %.lcssa8981, %sw.epilog ], [ %1152, %Flow8427 ]
  %1403 = phi i64 [ %.lcssa8982, %sw.epilog ], [ %1151, %Flow8427 ]
  %1404 = phi i32 [ %.lcssa8983, %sw.epilog ], [ %1150, %Flow8427 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %1235)
  br label %cleanup, !amdgpu.uniform !42

for.cond.i.i.i748:                                ; preds = %for.cond.i.i.i748.preheader, %for.cond.i.i.i748
  %phi.broken8655 = phi i64 [ 0, %for.cond.i.i.i748.preheader ], [ %1406, %for.cond.i.i.i748 ]
  %lsr.iv6587 = phi ptr addrspace(4) [ %scevgep6589, %for.cond.i.i.i748 ], [ %scevgep6586, %for.cond.i.i.i748.preheader ]
  %scevgep6589 = getelementptr i8, ptr addrspace(4) %lsr.iv6587, i64 1
  %1405 = load i8, ptr addrspace(4) %scevgep6589, align 1, !tbaa !57
  %cmp.i.i.i750 = icmp eq i8 %1405, 0
  %1406 = call i64 @llvm.amdgcn.if.break.i64(i1 %cmp.i.i.i750, i64 %phi.broken8655)
  %1407 = call i1 @llvm.amdgcn.loop.i64(i64 %1406)
  br i1 %1407, label %_ZN22__llvm_libc_22_0_0_git3cpp11string_viewC2EPKc.exit.i752, label %for.cond.i.i.i748

_ZN22__llvm_libc_22_0_0_git3cpp11string_viewC2EPKc.exit.i752: ; preds = %for.cond.i.i.i748
  %scevgep6589.lcssa = phi ptr addrspace(4) [ %scevgep6589, %for.cond.i.i.i748 ]
  %.lcssa8977 = phi i64 [ %1406, %for.cond.i.i.i748 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa8977)
  %1408 = call i32 @llvm.amdgcn.mbcnt.lo(i32 -1, i32 0)
  %1409 = call noundef i32 @llvm.amdgcn.mbcnt.hi(i32 -1, i32 %1408)
  %sh_prom.i.i.i.i756 = zext i32 %1409 to i64
  br label %for.cond.i.i1.i757, !amdgpu.uniform !42

for.cond.i.i1.i757:                               ; preds = %_ZN22__llvm_libc_22_0_0_git3cpp11string_viewC2EPKc.exit.i752, %Flow8426
  %1410 = phi i32 [ %1447, %Flow8426 ], [ %1150, %_ZN22__llvm_libc_22_0_0_git3cpp11string_viewC2EPKc.exit.i752 ]
  %1411 = phi i64 [ %1448, %Flow8426 ], [ %1151, %_ZN22__llvm_libc_22_0_0_git3cpp11string_viewC2EPKc.exit.i752 ]
  %1412 = phi i32 [ %1449, %Flow8426 ], [ %1152, %_ZN22__llvm_libc_22_0_0_git3cpp11string_viewC2EPKc.exit.i752 ]
  %retval.sroa.6.0.i.i.i758 = phi i32 [ %1452, %Flow8426 ], [ undef, %_ZN22__llvm_libc_22_0_0_git3cpp11string_viewC2EPKc.exit.i752 ]
  %retval.sroa.2.0.i.i.i760 = phi i64 [ %1451, %Flow8426 ], [ undef, %_ZN22__llvm_libc_22_0_0_git3cpp11string_viewC2EPKc.exit.i752 ]
  %retval.sroa.8.0.i.i.i761 = phi i32 [ %1450, %Flow8426 ], [ undef, %_ZN22__llvm_libc_22_0_0_git3cpp11string_viewC2EPKc.exit.i752 ]
  %index.0.i.i.i762 = phi i32 [ %1454, %Flow8426 ], [ 0, %_ZN22__llvm_libc_22_0_0_git3cpp11string_viewC2EPKc.exit.i752 ]
  %1413 = load i32, ptr addrspace(1) @__llvm_rpc_client, align 8, !tbaa !68
  %cmp.not.i.i.i763 = icmp ult i32 %index.0.i.i.i762, %1413
  %spec.store.select.i.i.i764 = select i1 %cmp.not.i.i.i763, i32 %index.0.i.i.i762, i32 0
  %1414 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  %1415 = and i32 %spec.store.select.i.i.i764, 31
  %1416 = lshr i64 %1414, %sh_prom.i.i.i.i756
  %1417 = trunc i64 %1416 to i32
  %conv4.i.i.i.i.i768 = and i32 %1417, 1
  %shl.i.i.i.i.i769 = shl nuw i32 %conv4.i.i.i.i.i768, %1415
  br label %ComputeLoop5567, !amdgpu.uniform !42

1418:                                             ; preds = %ComputeEnd5568
  %sunkaddr7390 = getelementptr inbounds i8, ptr addrspace(1) %7130, i64 40
  %1419 = atomicrmw or ptr addrspace(1) %sunkaddr7390, i32 %.lcssa8978 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %1420, !amdgpu.uniform !42

1420:                                             ; preds = %1418, %ComputeEnd5568
  %1421 = phi i32 [ %1419, %1418 ], [ poison, %ComputeEnd5568 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %7139)
  %1422 = call i32 @llvm.amdgcn.readfirstlane.i32(i32 %1421)
  %1423 = or i32 %1422, %.lcssa8979
  %shl5.i.i.i.i.i770 = shl nuw i32 1, %1415
  %and.i.i.i.i.i771 = and i32 %shl5.i.i.i.i.i770, %1423
  %tobool3.i.i.i.i772 = icmp ne i32 %and.i.i.i.i.i771, 0
  %1424 = call i64 @llvm.amdgcn.ballot.i64(i1 %tobool3.i.i.i.i772)
  %and.i.i.i.i.i.i773 = and i64 %1424, %1414
  %cmp.i.not.i.i.i774 = icmp ne i64 %1414, %and.i.i.i.i.i.i773
  br i1 %cmp.i.not.i.i.i774, label %if.then.i.i.i.i775, label %Flow8425, !amdgpu.uniform !42

if.then.i.i.i.i775:                               ; preds = %1420
  fence syncscope("agent") acquire
  %1425 = load <2 x i64>, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !75
  %1426 = extractelement <2 x i64> %1425, i32 0
  %1427 = inttoptr i64 %1426 to ptr
  %1428 = extractelement <2 x i64> %1425, i32 1
  %1429 = inttoptr i64 %1428 to ptr
  %idxprom.i.i.i.i776 = zext i32 %spec.store.select.i.i.i764 to i64
  %arrayidx.i.i.i.i777 = getelementptr inbounds nuw i32, ptr %1427, i64 %idxprom.i.i.i.i776, !amdgpu.uniform !42
  %1430 = load atomic i32, ptr %arrayidx.i.i.i.i777 monotonic, align 4
  %1431 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %1430)
  %arrayidx.i45.i.i.i778 = getelementptr inbounds nuw i32, ptr %1429, i64 %idxprom.i.i.i.i776, !amdgpu.uniform !42
  %1432 = load atomic i32, ptr %arrayidx.i45.i.i.i778 monotonic, align 4
  %1433 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %1432)
  %cmp.i46.not.i.i.i779 = icmp ne i32 %1431, %1433
  br i1 %cmp.i46.not.i.i.i779, label %if.then12.i.i.i780, label %Flow8423, !amdgpu.uniform !42

if.then12.i.i.i780:                               ; preds = %if.then.i.i.i.i775
  fence syncscope("agent") release
  %1434 = call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %1414, i1 true)
  %iszero.i.i.i.i.i.i.i781 = icmp ne i64 %1414, 0
  %cmp2.i.i.i.i.i.i782 = icmp eq i64 %1434, %sh_prom.i.i.i.i756
  %cmp.i.i.i.i.i.i783 = select i1 %iszero.i.i.i.i.i.i.i781, i1 %cmp2.i.i.i.i.i.i782, i1 false
  %conv4.i.i51.i.i.i784 = zext i1 %cmp.i.i.i.i.i.i783 to i32
  %shl.i.i52.i.i.i785 = shl nuw i32 %conv4.i.i51.i.i.i784, %1415
  %xor.i.i.i.i.i786 = xor i32 %shl.i.i52.i.i.i785, -1
  %1435 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop5572, !amdgpu.uniform !42

1436:                                             ; preds = %ComputeEnd5573
  %sunkaddr7391 = getelementptr inbounds i8, ptr addrspace(1) %7130, i64 40
  %1437 = atomicrmw and ptr addrspace(1) %sunkaddr7391, i32 %.lcssa8980 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %1438, !amdgpu.uniform !42

1438:                                             ; preds = %1436, %ComputeEnd5573
  call void @llvm.amdgcn.end.cf.i64(i64 %7157)
  call void @llvm.amdgcn.wave.barrier()
  br label %Flow8423, !amdgpu.uniform !42

if.end14.i.i.i846:                                ; preds = %Flow8423
  %1439 = call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %1414, i1 true)
  %iszero.i.i.i.i.i.i847 = icmp ne i64 %1414, 0
  %cmp2.i.i.i.i.i848 = icmp eq i64 %1439, %sh_prom.i.i.i.i756
  %cmp.i.i.i.i.i849 = select i1 %iszero.i.i.i.i.i.i847, i1 %cmp2.i.i.i.i.i848, i1 false
  %1440 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.i.i.i.i.i849)
  %1441 = extractvalue { i1, i64 } %1440, 0
  %1442 = extractvalue { i1, i64 } %1440, 1
  br i1 %1441, label %if.then16.i.i.i851, label %if.end22.i.i.i850

if.then16.i.i.i851:                               ; preds = %if.end14.i.i.i846
  %1443 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !76
  %opcode.i.i.i852 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %1443, i64 %idxprom.i.i.i.i776, i32 1
  store i32 1660944387, ptr %opcode.i.i.i852, align 8, !tbaa !77
  %arrayidx21.i.i.i853 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %1443, i64 %idxprom.i.i.i.i776
  store i64 %1414, ptr %arrayidx21.i.i.i853, align 8, !tbaa !80
  br label %if.end22.i.i.i850, !amdgpu.uniform !42

Flow8424:                                         ; preds = %if.end22.i.i.i850, %Flow8423
  %1444 = phi i32 [ %1431, %if.end22.i.i.i850 ], [ %retval.sroa.8.0.i.i.i761, %Flow8423 ]
  %1445 = phi i64 [ %1414, %if.end22.i.i.i850 ], [ %retval.sroa.2.0.i.i.i760, %Flow8423 ]
  %1446 = phi i32 [ %spec.store.select.i.i.i764, %if.end22.i.i.i850 ], [ %retval.sroa.6.0.i.i.i758, %Flow8423 ]
  br label %cleanup26.i.i.i787, !amdgpu.uniform !42

if.end22.i.i.i850:                                ; preds = %if.then16.i.i.i851, %if.end14.i.i.i846
  call void @llvm.amdgcn.end.cf.i64(i64 %1442)
  call void @llvm.amdgcn.wave.barrier()
  br label %Flow8424, !amdgpu.uniform !42

Flow8425:                                         ; preds = %cleanup26.i.i.i787, %1420
  %1447 = phi i32 [ %1446, %cleanup26.i.i.i787 ], [ %1410, %1420 ]
  %1448 = phi i64 [ %1445, %cleanup26.i.i.i787 ], [ %1411, %1420 ]
  %1449 = phi i32 [ %1444, %cleanup26.i.i.i787 ], [ %1412, %1420 ]
  %1450 = phi i32 [ %1444, %cleanup26.i.i.i787 ], [ %retval.sroa.8.0.i.i.i761, %1420 ]
  %1451 = phi i64 [ %1445, %cleanup26.i.i.i787 ], [ %retval.sroa.2.0.i.i.i760, %1420 ]
  %1452 = phi i32 [ %1446, %cleanup26.i.i.i787 ], [ %retval.sroa.6.0.i.i.i758, %1420 ]
  %1453 = phi i1 [ %cmp.i46.not.i.i.i779, %cleanup26.i.i.i787 ], [ true, %1420 ]
  br i1 %1453, label %for.inc.i.i.i792, label %Flow8426, !amdgpu.uniform !42

cleanup26.i.i.i787:                               ; preds = %Flow8424
  br label %Flow8425, !amdgpu.uniform !42

for.inc.i.i.i792:                                 ; preds = %Flow8425
  %inc.i.i.i797 = add i32 %spec.store.select.i.i.i764, 1
  br label %Flow8426, !amdgpu.uniform !42

Flow8426:                                         ; preds = %for.inc.i.i.i792, %Flow8425
  %1454 = phi i32 [ %inc.i.i.i797, %for.inc.i.i.i792 ], [ poison, %Flow8425 ]
  %1455 = phi i1 [ false, %for.inc.i.i.i792 ], [ true, %Flow8425 ]
  br i1 %1455, label %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i798, label %for.cond.i.i1.i757, !amdgpu.uniform !42

_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i798: ; preds = %Flow8426
  %.lcssa8983 = phi i32 [ %1447, %Flow8426 ]
  %.lcssa8982 = phi i64 [ %1448, %Flow8426 ]
  %.lcssa8981 = phi i32 [ %1449, %Flow8426 ]
  %1456 = addrspacecast ptr addrspace(4) %scevgep6589.lcssa to ptr
  %1457 = addrspacecast ptr addrspace(4) %cond to ptr
  %1458 = ptrtoint ptr %1457 to i64
  %sub.ptr.lhs.cast.i.i.i753 = ptrtoint ptr %1456 to i64
  %sub.ptr.sub.i.i.i755 = sub i64 %sub.ptr.lhs.cast.i.i.i753, %1458
  fence acquire
  %1459 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %mul.i.i.i.i.i.i799 = shl i32 %.lcssa8983, 6
  %idxprom.i19.i.i.i.i.i800 = zext i32 %mul.i.i.i.i.i.i799 to i64
  %arrayidx.i20.i.i.i.i.i801 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %1459, i64 %idxprom.i19.i.i.i.i.i800
  %arrayidx.i22.i.i.i.i.i802 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %arrayidx.i20.i.i.i.i.i801, i64 %sh_prom.i.i.i.i756
  store i64 %sub.ptr.sub.i.i.i755, ptr %arrayidx.i22.i.i.i.i.i802, align 8, !tbaa !82
  %spec.select.i.i.i.i.i.i.i803 = call i64 @llvm.umin.i64(i64 %sub.ptr.sub.i.i.i755, i64 56)
  %1460 = and i64 %spec.select.i.i.i.i.i.i.i803, 15
  %1461 = sub i64 %spec.select.i.i.i.i.i.i.i803, %1460
  %1462 = icmp ne i64 %1461, 0
  %1463 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %1462)
  %1464 = extractvalue { i1, i64 } %1463, 0
  %1465 = extractvalue { i1, i64 } %1463, 1
  br i1 %1464, label %loop-memcpy-expansion5151.preheader, label %Flow8422

loop-memcpy-expansion5151.preheader:              ; preds = %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i798
  %scevgep6590 = getelementptr i8, ptr %1459, i64 8
  %1466 = shl nuw nsw i64 %sh_prom.i.i.i.i756, 6
  %scevgep6591 = getelementptr i8, ptr %scevgep6590, i64 %1466
  %1467 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i800, 6
  %scevgep6592 = getelementptr i8, ptr %scevgep6591, i64 %1467
  br label %loop-memcpy-expansion5151, !amdgpu.uniform !42

Flow8422:                                         ; preds = %Flow8421, %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i798
  call void @llvm.amdgcn.end.cf.i64(i64 %1465)
  br label %loop-memcpy-residual-header5154, !amdgpu.uniform !42

loop-memcpy-expansion5151:                        ; preds = %loop-memcpy-expansion5151.preheader, %loop-memcpy-expansion5151
  %phi.broken8656 = phi i64 [ 0, %loop-memcpy-expansion5151.preheader ], [ %1472, %loop-memcpy-expansion5151 ]
  %loop-index5152 = phi i64 [ %1470, %loop-memcpy-expansion5151 ], [ 0, %loop-memcpy-expansion5151.preheader ]
  %1468 = getelementptr i8, ptr addrspace(4) %cond, i64 %loop-index5152
  %1469 = load <4 x i32>, ptr addrspace(4) %1468, align 1
  %scevgep6593 = getelementptr i8, ptr %scevgep6592, i64 %loop-index5152
  store <4 x i32> %1469, ptr %scevgep6593, align 1
  %1470 = add i64 %loop-index5152, 16
  %1471 = icmp uge i64 %1470, %1461
  %1472 = call i64 @llvm.amdgcn.if.break.i64(i1 %1471, i64 %phi.broken8656)
  %1473 = call i1 @llvm.amdgcn.loop.i64(i64 %1472)
  br i1 %1473, label %Flow8421, label %loop-memcpy-expansion5151

Flow8420:                                         ; preds = %Flow8418, %Flow8419
  %1474 = phi i64 [ %6439, %Flow8418 ], [ %6434, %Flow8419 ]
  %1475 = phi i64 [ %6441, %Flow8418 ], [ %6435, %Flow8419 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %6438)
  br label %post-loop-memcpy-expansion5150, !amdgpu.uniform !42

loop-memcpy-residual5153:                         ; preds = %loop-memcpy-residual5153.preheader, %loop-memcpy-residual5153
  %phi.broken8657 = phi i64 [ 0, %loop-memcpy-residual5153.preheader ], [ %1479, %loop-memcpy-residual5153 ]
  %residual-loop-index5155 = phi i64 [ %1477, %loop-memcpy-residual5153 ], [ 0, %loop-memcpy-residual5153.preheader ]
  %scevgep6599 = getelementptr i8, ptr addrspace(4) %scevgep6598, i64 %residual-loop-index5155
  %1476 = load i8, ptr addrspace(4) %scevgep6599, align 1
  %scevgep6597 = getelementptr i8, ptr %scevgep6596, i64 %residual-loop-index5155
  store i8 %1476, ptr %scevgep6597, align 1
  %1477 = add i64 %residual-loop-index5155, 1
  %1478 = icmp uge i64 %1477, %1460
  %1479 = call i64 @llvm.amdgcn.if.break.i64(i1 %1478, i64 %phi.broken8657)
  %1480 = call i1 @llvm.amdgcn.loop.i64(i64 %1479)
  br i1 %1480, label %Flow8418, label %loop-memcpy-residual5153

Flow8418:                                         ; preds = %loop-memcpy-residual5153
  %.lcssa8985 = phi i64 [ %1479, %loop-memcpy-residual5153 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa8985)
  br label %Flow8420, !amdgpu.uniform !42

post-loop-memcpy-expansion5150:                   ; preds = %Flow8420
  %1481 = addrspacecast ptr addrspace(4) %scevgep6589.lcssa to ptr
  %1482 = ptrtoint ptr %1481 to i64
  %tobool.not.i.i.i.i.i.i805 = icmp eq i32 %.lcssa8981, 0
  %conv.i.i.i.i.i.i806 = zext i1 %tobool.not.i.i.i.i.i.i805 to i32
  fence release
  %1483 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %idxprom.i23.i.i.i.i.i807 = zext i32 %.lcssa8983 to i64
  %arrayidx.i24.i.i.i.i.i808 = getelementptr inbounds nuw i32, ptr %1483, i64 %idxprom.i23.i.i.i.i.i807
  store atomic i32 %conv.i.i.i.i.i.i806, ptr %arrayidx.i24.i.i.i.i.i808 monotonic, align 4
  %1484 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !84
  %arrayidx.i.i5.i.i809 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %1484, i64 %idxprom.i23.i.i.i.i.i807
  %1485 = load i64, ptr %arrayidx.i.i5.i.i809, align 8, !tbaa !80
  %1486 = add i64 %1140, %1482
  br label %while.cond.i.i.i.i810, !amdgpu.uniform !42

while.cond.i.i.i.i810:                            ; preds = %post-loop-memcpy-expansion5150, %Flow8417
  %phi.broken8661 = phi i64 [ 0, %post-loop-memcpy-expansion5150 ], [ %1508, %Flow8417 ]
  %lsr.iv6606 = phi i64 [ %1506, %Flow8417 ], [ %1486, %post-loop-memcpy-expansion5150 ]
  %lsr.iv6601 = phi ptr addrspace(4) [ %1505, %Flow8417 ], [ %scevgep6600, %post-loop-memcpy-expansion5150 ]
  %port.sroa.37.0.i.i811 = phi i32 [ %1504, %Flow8417 ], [ %conv.i.i.i.i.i.i806, %post-loop-memcpy-expansion5150 ]
  %idx.0.i.i.i.i812 = phi i64 [ %1503, %Flow8417 ], [ 56, %post-loop-memcpy-expansion5150 ]
  %cmp.i.i.i.i813 = icmp ugt i64 %sub.ptr.sub.i.i.i755, %idx.0.i.i.i.i812
  %1487 = call i64 @llvm.amdgcn.ballot.i64(i1 %cmp.i.i.i.i813)
  %and.i.i.i.i6.i.i814 = and i64 %1487, %1485
  %tobool.not.i.i.i.i815 = icmp ne i64 %and.i.i.i.i6.i.i814, 0
  %1488 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i22.i.i.i816 = getelementptr inbounds nuw i32, ptr %1488, i64 %idxprom.i23.i.i.i.i.i807
  %1489 = load atomic i32, ptr %arrayidx.i22.i.i.i816 monotonic, align 4
  %1490 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %tobool.not.i.i.i.i815)
  %1491 = extractvalue { i1, i64 } %1490, 0
  %1492 = extractvalue { i1, i64 } %1490, 1
  br i1 %1491, label %cond.false.i8.i.i.i.i817, label %Flow8417

cond.false.i8.i.i.i.i817:                         ; preds = %while.cond.i.i.i.i810
  %1493 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %1489)
  %1494 = icmp ne i32 %1493, %port.sroa.37.0.i.i811
  %1495 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %1494)
  %1496 = extractvalue { i1, i64 } %1495, 0
  %1497 = extractvalue { i1, i64 } %1495, 1
  br i1 %1496, label %while.body.i.i21.i.i.i.i818.preheader, label %Flow8416

while.body.i.i21.i.i.i.i818.preheader:            ; preds = %cond.false.i8.i.i.i.i817
  br label %while.body.i.i21.i.i.i.i818, !amdgpu.uniform !42

Flow8416:                                         ; preds = %Flow8415, %cond.false.i8.i.i.i.i817
  call void @llvm.amdgcn.end.cf.i64(i64 %1497)
  br label %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i821, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i818:                      ; preds = %while.body.i.i21.i.i.i.i818.preheader, %while.body.i.i21.i.i.i.i818
  %phi.broken8658 = phi i64 [ 0, %while.body.i.i21.i.i.i.i818.preheader ], [ %1501, %while.body.i.i21.i.i.i.i818 ]
  call void @llvm.amdgcn.s.sleep(i32 2)
  %1498 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i22.i.i.i.i819 = getelementptr inbounds nuw i32, ptr %1498, i64 %idxprom.i23.i.i.i.i.i807
  %1499 = load atomic i32, ptr %arrayidx.i.i.i22.i.i.i.i819 monotonic, align 4
  %1500 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %1499)
  %cmp.i.not.i.i23.i.i.i.i820 = icmp eq i32 %1500, %port.sroa.37.0.i.i811
  %1501 = call i64 @llvm.amdgcn.if.break.i64(i1 %cmp.i.not.i.i23.i.i.i.i820, i64 %phi.broken8658)
  %1502 = call i1 @llvm.amdgcn.loop.i64(i64 %1501)
  br i1 %1502, label %Flow8415, label %while.body.i.i21.i.i.i.i818

Flow8415:                                         ; preds = %while.body.i.i21.i.i.i.i818
  %.lcssa8986 = phi i64 [ %1501, %while.body.i.i21.i.i.i.i818 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa8986)
  br label %Flow8416, !amdgpu.uniform !42

Flow8417:                                         ; preds = %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i822, %while.cond.i.i.i.i810
  %1503 = phi i64 [ %add.i.i.i.i825, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i822 ], [ poison, %while.cond.i.i.i.i810 ]
  %1504 = phi i32 [ %conv.i.i26.i.i.i.i823, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i822 ], [ poison, %while.cond.i.i.i.i810 ]
  %1505 = phi ptr addrspace(4) [ %scevgep6602, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i822 ], [ poison, %while.cond.i.i.i.i810 ]
  %1506 = phi i64 [ %lsr.iv.next6607, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i822 ], [ poison, %while.cond.i.i.i.i810 ]
  %1507 = phi i1 [ false, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i822 ], [ true, %while.cond.i.i.i.i810 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %1492)
  %1508 = call i64 @llvm.amdgcn.if.break.i64(i1 %1507, i64 %phi.broken8661)
  %1509 = call i1 @llvm.amdgcn.loop.i64(i64 %1508)
  br i1 %1509, label %cond.false.i.i.i832, label %while.cond.i.i.i.i810

_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i821: ; preds = %Flow8416
  fence acquire
  %1510 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.i.i.i.i813)
  %1511 = extractvalue { i1, i64 } %1510, 0
  %1512 = extractvalue { i1, i64 } %1510, 1
  br i1 %1511, label %if.then.i.i.i.i.i.i.i826, label %Flow8414

if.then.i.i.i.i.i.i.i826:                         ; preds = %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i821
  %sub.i.i.i.i.i.i.i828 = sub nuw i64 %sub.ptr.sub.i.i.i755, %idx.0.i.i.i.i812
  %spec.select.i.i.i36.i.i.i.i829 = call i64 @llvm.umin.i64(i64 %sub.i.i.i.i.i.i.i828, i64 64)
  %1513 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %1514 = and i64 %spec.select.i.i.i36.i.i.i.i829, 15
  %1515 = sub i64 %spec.select.i.i.i36.i.i.i.i829, %1514
  %1516 = icmp ne i64 %1515, 0
  %1517 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %1516)
  %1518 = extractvalue { i1, i64 } %1517, 0
  %1519 = extractvalue { i1, i64 } %1517, 1
  br i1 %1518, label %loop-memcpy-expansion5259.preheader, label %Flow8413

loop-memcpy-expansion5259.preheader:              ; preds = %if.then.i.i.i.i.i.i.i826
  %arrayidx.i20.i35.i.i.i.i830 = getelementptr %"struct.rpc::Buffer", ptr %1513, i64 %idxprom.i19.i.i.i.i.i800
  %arrayidx.i22.i38.i.i.i.i831 = getelementptr %"struct.rpc::Buffer", ptr %arrayidx.i20.i35.i.i.i.i830, i64 %sh_prom.i.i.i.i756
  br label %loop-memcpy-expansion5259, !amdgpu.uniform !42

Flow8413:                                         ; preds = %Flow8412, %if.then.i.i.i.i.i.i.i826
  call void @llvm.amdgcn.end.cf.i64(i64 %1519)
  br label %loop-memcpy-residual-header5262, !amdgpu.uniform !42

loop-memcpy-expansion5259:                        ; preds = %loop-memcpy-expansion5259.preheader, %loop-memcpy-expansion5259
  %phi.broken8659 = phi i64 [ 0, %loop-memcpy-expansion5259.preheader ], [ %1524, %loop-memcpy-expansion5259 ]
  %loop-index5260 = phi i64 [ %1522, %loop-memcpy-expansion5259 ], [ 0, %loop-memcpy-expansion5259.preheader ]
  %scevgep6603 = getelementptr i8, ptr addrspace(4) %lsr.iv6601, i64 %loop-index5260
  %1520 = load <4 x i32>, ptr addrspace(4) %scevgep6603, align 1
  %1521 = getelementptr i8, ptr %arrayidx.i22.i38.i.i.i.i831, i64 %loop-index5260
  store <4 x i32> %1520, ptr %1521, align 1
  %1522 = add i64 %loop-index5260, 16
  %1523 = icmp uge i64 %1522, %1515
  %1524 = call i64 @llvm.amdgcn.if.break.i64(i1 %1523, i64 %phi.broken8659)
  %1525 = call i1 @llvm.amdgcn.loop.i64(i64 %1524)
  br i1 %1525, label %Flow8412, label %loop-memcpy-expansion5259

Flow8411:                                         ; preds = %Flow8410, %loop-memcpy-residual-header5262
  call void @llvm.amdgcn.end.cf.i64(i64 %6548)
  br label %Flow8414, !amdgpu.uniform !42

loop-memcpy-residual5261:                         ; preds = %loop-memcpy-residual5261.preheader, %loop-memcpy-residual5261
  %phi.broken8660 = phi i64 [ 0, %loop-memcpy-residual5261.preheader ], [ %1529, %loop-memcpy-residual5261 ]
  %residual-loop-index5263 = phi i64 [ %1527, %loop-memcpy-residual5261 ], [ 0, %loop-memcpy-residual5261.preheader ]
  %scevgep6612 = getelementptr i8, ptr addrspace(4) %scevgep6611, i64 %residual-loop-index5263
  %1526 = load i8, ptr addrspace(4) %scevgep6612, align 1
  %scevgep6610 = getelementptr i8, ptr %scevgep6609, i64 %residual-loop-index5263
  store i8 %1526, ptr %scevgep6610, align 1
  %1527 = add i64 %residual-loop-index5263, 1
  %1528 = icmp uge i64 %1527, %1514
  %1529 = call i64 @llvm.amdgcn.if.break.i64(i1 %1528, i64 %phi.broken8660)
  %1530 = call i1 @llvm.amdgcn.loop.i64(i64 %1529)
  br i1 %1530, label %Flow8410, label %loop-memcpy-residual5261

Flow8410:                                         ; preds = %loop-memcpy-residual5261
  %.lcssa8988 = phi i64 [ %1529, %loop-memcpy-residual5261 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa8988)
  br label %Flow8411, !amdgpu.uniform !42

_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i822: ; preds = %Flow8414
  %conv.i.i26.i.i.i.i823 = xor i32 %port.sroa.37.0.i.i811, 1
  fence release
  %1531 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %arrayidx.i24.i29.i.i.i.i824 = getelementptr inbounds nuw i32, ptr %1531, i64 %idxprom.i23.i.i.i.i.i807
  store atomic i32 %conv.i.i26.i.i.i.i823, ptr %arrayidx.i24.i29.i.i.i.i824 monotonic, align 4
  %add.i.i.i.i825 = add i64 %idx.0.i.i.i.i812, 64
  %scevgep6602 = getelementptr i8, ptr addrspace(4) %lsr.iv6601, i64 64
  %lsr.iv.next6607 = add i64 %lsr.iv6606, -64
  br label %Flow8417, !amdgpu.uniform !42

cond.false.i.i.i832:                              ; preds = %Flow8417
  %.lcssa8990 = phi i64 [ %1508, %Flow8417 ]
  %port.sroa.37.0.i.i811.lcssa = phi i32 [ %port.sroa.37.0.i.i811, %Flow8417 ]
  %.lcssa8989 = phi i32 [ %1489, %Flow8417 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa8990)
  %1532 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %.lcssa8989)
  %1533 = icmp ne i32 %1532, %port.sroa.37.0.i.i811.lcssa
  %1534 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %1533)
  %1535 = extractvalue { i1, i64 } %1534, 0
  %1536 = extractvalue { i1, i64 } %1534, 1
  br i1 %1535, label %while.body.i.i.i.i833.preheader, label %Flow8409

while.body.i.i.i.i833.preheader:                  ; preds = %cond.false.i.i.i832
  br label %while.body.i.i.i.i833, !amdgpu.uniform !42

Flow8409:                                         ; preds = %Flow8408, %cond.false.i.i.i832
  call void @llvm.amdgcn.end.cf.i64(i64 %1536)
  br label %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit854, !amdgpu.uniform !42

while.body.i.i.i.i833:                            ; preds = %while.body.i.i.i.i833.preheader, %while.body.i.i.i.i833
  %phi.broken8662 = phi i64 [ 0, %while.body.i.i.i.i833.preheader ], [ %1540, %while.body.i.i.i.i833 ]
  call void @llvm.amdgcn.s.sleep(i32 2)
  %1537 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i8.i.i834 = getelementptr inbounds nuw i32, ptr %1537, i64 %idxprom.i23.i.i.i.i.i807
  %1538 = load atomic i32, ptr %arrayidx.i.i.i8.i.i834 monotonic, align 4
  %1539 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %1538)
  %cmp.i.not.i.i.i.i835 = icmp eq i32 %1539, %port.sroa.37.0.i.i811.lcssa
  %1540 = call i64 @llvm.amdgcn.if.break.i64(i1 %cmp.i.not.i.i.i.i835, i64 %phi.broken8662)
  %1541 = call i1 @llvm.amdgcn.loop.i64(i64 %1540)
  br i1 %1541, label %Flow8408, label %while.body.i.i.i.i833

Flow8408:                                         ; preds = %while.body.i.i.i.i833
  %.lcssa8991 = phi i64 [ %1540, %while.body.i.i.i.i833 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa8991)
  br label %Flow8409, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit854: ; preds = %Flow8409
  fence acquire
  call void @llvm.amdgcn.wave.barrier()
  fence syncscope("agent") release
  %1542 = call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %.lcssa8982, i1 true)
  %iszero.i.i.i.i.i13.i.i837 = icmp ne i64 %.lcssa8982, 0
  %cmp2.i.i.i.i14.i.i838 = icmp eq i64 %1542, %sh_prom.i.i.i.i756
  %cmp.i.i.i.i15.i.i839 = select i1 %iszero.i.i.i.i.i13.i.i837, i1 %cmp2.i.i.i.i14.i.i838, i1 false
  %1543 = and i32 %.lcssa8983, 31
  %conv4.i.i.i19.i.i843 = zext i1 %cmp.i.i.i.i15.i.i839 to i32
  %shl.i.i.i20.i.i844 = shl nuw i32 %conv4.i.i.i19.i.i843, %1543
  %xor.i.i.i21.i.i845 = xor i32 %shl.i.i.i20.i.i844, -1
  %1544 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop5576, !amdgpu.uniform !42

1545:                                             ; preds = %ComputeEnd5577
  %div8.i.i.i16.i.i840 = lshr i32 %.lcssa8983, 5
  %idxprom.i.i.i17.i.i841 = zext nneg i32 %div8.i.i.i16.i.i840 to i64
  %1546 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i17.i.i841
  %sunkaddr7392 = getelementptr inbounds i8, ptr addrspace(1) %1546, i64 40
  %1547 = atomicrmw and ptr addrspace(1) %sunkaddr7392, i32 %.lcssa8992 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %1548, !amdgpu.uniform !42

1548:                                             ; preds = %1545, %ComputeEnd5577
  call void @llvm.amdgcn.end.cf.i64(i64 %7174)
  call void @llvm.amdgcn.wave.barrier()
  br label %for.cond.i.i1.i863, !amdgpu.uniform !42

for.cond.i.i1.i863:                               ; preds = %1548, %Flow8407
  %1549 = phi i32 [ %1586, %Flow8407 ], [ %1153, %1548 ]
  %1550 = phi i64 [ %1587, %Flow8407 ], [ %1154, %1548 ]
  %1551 = phi i32 [ %1588, %Flow8407 ], [ %1155, %1548 ]
  %retval.sroa.6.0.i.i.i864 = phi i32 [ %1591, %Flow8407 ], [ undef, %1548 ]
  %retval.sroa.2.0.i.i.i866 = phi i64 [ %1590, %Flow8407 ], [ undef, %1548 ]
  %retval.sroa.8.0.i.i.i867 = phi i32 [ %1589, %Flow8407 ], [ undef, %1548 ]
  %index.0.i.i.i868 = phi i32 [ %1593, %Flow8407 ], [ 0, %1548 ]
  %1552 = load i32, ptr addrspace(1) @__llvm_rpc_client, align 8, !tbaa !68
  %cmp.not.i.i.i869 = icmp ult i32 %index.0.i.i.i868, %1552
  %spec.store.select.i.i.i870 = select i1 %cmp.not.i.i.i869, i32 %index.0.i.i.i868, i32 0
  %1553 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  %1554 = and i32 %spec.store.select.i.i.i870, 31
  %1555 = lshr i64 %1553, %sh_prom.i.i.i.i756
  %1556 = trunc i64 %1555 to i32
  %conv4.i.i.i.i.i874 = and i32 %1556, 1
  %shl.i.i.i.i.i875 = shl nuw i32 %conv4.i.i.i.i.i874, %1554
  br label %ComputeLoop5580, !amdgpu.uniform !42

1557:                                             ; preds = %ComputeEnd5581
  %sunkaddr7393 = getelementptr inbounds i8, ptr addrspace(1) %7184, i64 40
  %1558 = atomicrmw or ptr addrspace(1) %sunkaddr7393, i32 %.lcssa8993 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %1559, !amdgpu.uniform !42

1559:                                             ; preds = %1557, %ComputeEnd5581
  %1560 = phi i32 [ %1558, %1557 ], [ poison, %ComputeEnd5581 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %7193)
  %1561 = call i32 @llvm.amdgcn.readfirstlane.i32(i32 %1560)
  %1562 = or i32 %1561, %.lcssa8994
  %shl5.i.i.i.i.i876 = shl nuw i32 1, %1554
  %and.i.i.i.i.i877 = and i32 %shl5.i.i.i.i.i876, %1562
  %tobool3.i.i.i.i878 = icmp ne i32 %and.i.i.i.i.i877, 0
  %1563 = call i64 @llvm.amdgcn.ballot.i64(i1 %tobool3.i.i.i.i878)
  %and.i.i.i.i.i.i879 = and i64 %1563, %1553
  %cmp.i.not.i.i.i880 = icmp ne i64 %1553, %and.i.i.i.i.i.i879
  br i1 %cmp.i.not.i.i.i880, label %if.then.i.i.i.i881, label %Flow8406, !amdgpu.uniform !42

if.then.i.i.i.i881:                               ; preds = %1559
  fence syncscope("agent") acquire
  %1564 = load <2 x i64>, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !75
  %1565 = extractelement <2 x i64> %1564, i32 0
  %1566 = inttoptr i64 %1565 to ptr
  %1567 = extractelement <2 x i64> %1564, i32 1
  %1568 = inttoptr i64 %1567 to ptr
  %idxprom.i.i.i.i882 = zext i32 %spec.store.select.i.i.i870 to i64
  %arrayidx.i.i.i.i883 = getelementptr inbounds nuw i32, ptr %1566, i64 %idxprom.i.i.i.i882, !amdgpu.uniform !42
  %1569 = load atomic i32, ptr %arrayidx.i.i.i.i883 monotonic, align 4
  %1570 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %1569)
  %arrayidx.i45.i.i.i884 = getelementptr inbounds nuw i32, ptr %1568, i64 %idxprom.i.i.i.i882, !amdgpu.uniform !42
  %1571 = load atomic i32, ptr %arrayidx.i45.i.i.i884 monotonic, align 4
  %1572 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %1571)
  %cmp.i46.not.i.i.i885 = icmp ne i32 %1570, %1572
  br i1 %cmp.i46.not.i.i.i885, label %if.then12.i.i.i886, label %Flow8404, !amdgpu.uniform !42

if.then12.i.i.i886:                               ; preds = %if.then.i.i.i.i881
  fence syncscope("agent") release
  %1573 = call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %1553, i1 true)
  %iszero.i.i.i.i.i.i.i887 = icmp ne i64 %1553, 0
  %cmp2.i.i.i.i.i.i888 = icmp eq i64 %1573, %sh_prom.i.i.i.i756
  %cmp.i.i.i.i.i.i889 = select i1 %iszero.i.i.i.i.i.i.i887, i1 %cmp2.i.i.i.i.i.i888, i1 false
  %conv4.i.i51.i.i.i890 = zext i1 %cmp.i.i.i.i.i.i889 to i32
  %shl.i.i52.i.i.i891 = shl nuw i32 %conv4.i.i51.i.i.i890, %1554
  %xor.i.i.i.i.i892 = xor i32 %shl.i.i52.i.i.i891, -1
  %1574 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop5585, !amdgpu.uniform !42

1575:                                             ; preds = %ComputeEnd5586
  %sunkaddr7394 = getelementptr inbounds i8, ptr addrspace(1) %7184, i64 40
  %1576 = atomicrmw and ptr addrspace(1) %sunkaddr7394, i32 %.lcssa8995 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %1577, !amdgpu.uniform !42

1577:                                             ; preds = %1575, %ComputeEnd5586
  call void @llvm.amdgcn.end.cf.i64(i64 %7211)
  call void @llvm.amdgcn.wave.barrier()
  br label %Flow8404, !amdgpu.uniform !42

if.end14.i.i.i952:                                ; preds = %Flow8404
  %1578 = call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %1553, i1 true)
  %iszero.i.i.i.i.i.i953 = icmp ne i64 %1553, 0
  %cmp2.i.i.i.i.i954 = icmp eq i64 %1578, %sh_prom.i.i.i.i756
  %cmp.i.i.i.i.i955 = select i1 %iszero.i.i.i.i.i.i953, i1 %cmp2.i.i.i.i.i954, i1 false
  %1579 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.i.i.i.i.i955)
  %1580 = extractvalue { i1, i64 } %1579, 0
  %1581 = extractvalue { i1, i64 } %1579, 1
  br i1 %1580, label %if.then16.i.i.i957, label %if.end22.i.i.i956

if.then16.i.i.i957:                               ; preds = %if.end14.i.i.i952
  %1582 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !76
  %opcode.i.i.i958 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %1582, i64 %idxprom.i.i.i.i882, i32 1
  store i32 1660944387, ptr %opcode.i.i.i958, align 8, !tbaa !77
  %arrayidx21.i.i.i959 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %1582, i64 %idxprom.i.i.i.i882
  store i64 %1553, ptr %arrayidx21.i.i.i959, align 8, !tbaa !80
  br label %if.end22.i.i.i956, !amdgpu.uniform !42

Flow8405:                                         ; preds = %if.end22.i.i.i956, %Flow8404
  %1583 = phi i32 [ %1570, %if.end22.i.i.i956 ], [ %retval.sroa.8.0.i.i.i867, %Flow8404 ]
  %1584 = phi i64 [ %1553, %if.end22.i.i.i956 ], [ %retval.sroa.2.0.i.i.i866, %Flow8404 ]
  %1585 = phi i32 [ %spec.store.select.i.i.i870, %if.end22.i.i.i956 ], [ %retval.sroa.6.0.i.i.i864, %Flow8404 ]
  br label %cleanup26.i.i.i893, !amdgpu.uniform !42

if.end22.i.i.i956:                                ; preds = %if.then16.i.i.i957, %if.end14.i.i.i952
  call void @llvm.amdgcn.end.cf.i64(i64 %1581)
  call void @llvm.amdgcn.wave.barrier()
  br label %Flow8405, !amdgpu.uniform !42

Flow8406:                                         ; preds = %cleanup26.i.i.i893, %1559
  %1586 = phi i32 [ %1585, %cleanup26.i.i.i893 ], [ %1549, %1559 ]
  %1587 = phi i64 [ %1584, %cleanup26.i.i.i893 ], [ %1550, %1559 ]
  %1588 = phi i32 [ %1583, %cleanup26.i.i.i893 ], [ %1551, %1559 ]
  %1589 = phi i32 [ %1583, %cleanup26.i.i.i893 ], [ %retval.sroa.8.0.i.i.i867, %1559 ]
  %1590 = phi i64 [ %1584, %cleanup26.i.i.i893 ], [ %retval.sroa.2.0.i.i.i866, %1559 ]
  %1591 = phi i32 [ %1585, %cleanup26.i.i.i893 ], [ %retval.sroa.6.0.i.i.i864, %1559 ]
  %1592 = phi i1 [ %cmp.i46.not.i.i.i885, %cleanup26.i.i.i893 ], [ true, %1559 ]
  br i1 %1592, label %for.inc.i.i.i898, label %Flow8407, !amdgpu.uniform !42

cleanup26.i.i.i893:                               ; preds = %Flow8405
  br label %Flow8406, !amdgpu.uniform !42

for.inc.i.i.i898:                                 ; preds = %Flow8406
  %inc.i.i.i903 = add i32 %spec.store.select.i.i.i870, 1
  br label %Flow8407, !amdgpu.uniform !42

Flow8407:                                         ; preds = %for.inc.i.i.i898, %Flow8406
  %1593 = phi i32 [ %inc.i.i.i903, %for.inc.i.i.i898 ], [ poison, %Flow8406 ]
  %1594 = phi i1 [ false, %for.inc.i.i.i898 ], [ true, %Flow8406 ]
  br i1 %1594, label %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i904, label %for.cond.i.i1.i863, !amdgpu.uniform !42

_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i904: ; preds = %Flow8407
  %.lcssa8998 = phi i32 [ %1586, %Flow8407 ]
  %.lcssa8997 = phi i64 [ %1587, %Flow8407 ]
  %.lcssa8996 = phi i32 [ %1588, %Flow8407 ]
  fence acquire
  %1595 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %mul.i.i.i.i.i.i905 = shl i32 %.lcssa8998, 6
  %idxprom.i19.i.i.i.i.i906 = zext i32 %mul.i.i.i.i.i.i905 to i64
  %arrayidx.i20.i.i.i.i.i907 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %1595, i64 %idxprom.i19.i.i.i.i.i906
  %arrayidx.i22.i.i.i.i.i908 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %arrayidx.i20.i.i.i.i.i907, i64 %sh_prom.i.i.i.i756
  store i64 sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.17.34 to ptr), i64 13) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.17.34 to ptr) to i64)), ptr %arrayidx.i22.i.i.i.i.i908, align 8, !tbaa !82
  br i1 %1110, label %loop-memcpy-expansion5265.preheader, label %Flow8403, !amdgpu.uniform !42

loop-memcpy-expansion5265.preheader:              ; preds = %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i904
  %scevgep6613 = getelementptr i8, ptr %1595, i64 8
  %scevgep6614 = getelementptr i8, ptr %scevgep6613, i64 %1474
  %1596 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i906, 6
  %scevgep6615 = getelementptr i8, ptr %scevgep6614, i64 %1596
  br label %loop-memcpy-expansion5265, !amdgpu.uniform !42

Flow8403:                                         ; preds = %Flow8402, %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i904
  br label %loop-memcpy-residual-header5268, !amdgpu.uniform !42

loop-memcpy-expansion5265:                        ; preds = %loop-memcpy-expansion5265.preheader, %loop-memcpy-expansion5265
  %loop-index5266 = phi i64 [ %1599, %loop-memcpy-expansion5265 ], [ 0, %loop-memcpy-expansion5265.preheader ]
  %1597 = getelementptr i8, ptr addrspace(4) @.str.17.34, i64 %loop-index5266, !amdgpu.uniform !42
  %1598 = load <4 x i32>, ptr addrspace(4) %1597, align 1
  %scevgep6616 = getelementptr i8, ptr %scevgep6615, i64 %loop-index5266
  store <4 x i32> %1598, ptr %scevgep6616, align 1
  %1599 = add i64 %loop-index5266, 16
  %1600 = icmp uge i64 %1599, %1109
  br i1 %1600, label %Flow8402, label %loop-memcpy-expansion5265, !amdgpu.uniform !42

Flow8401:                                         ; preds = %Flow8399, %Flow8400
  %1601 = phi i64 [ %6553, %Flow8399 ], [ %6551, %Flow8400 ]
  br label %post-loop-memcpy-expansion5264, !amdgpu.uniform !42

loop-memcpy-residual5267:                         ; preds = %loop-memcpy-residual5267.preheader, %loop-memcpy-residual5267
  %residual-loop-index5269 = phi i64 [ %1603, %loop-memcpy-residual5267 ], [ 0, %loop-memcpy-residual5267.preheader ]
  %scevgep6621 = getelementptr i8, ptr addrspace(4) @.str.17.34, i64 %residual-loop-index5269, !amdgpu.uniform !42
  %1602 = load i8, ptr addrspace(4) %scevgep6621, align 1
  %scevgep6620 = getelementptr i8, ptr %scevgep6619, i64 %residual-loop-index5269
  store i8 %1602, ptr %scevgep6620, align 1
  %1603 = add i64 %residual-loop-index5269, 1
  %1604 = icmp uge i64 %1603, %1108
  br i1 %1604, label %Flow8399, label %loop-memcpy-residual5267, !amdgpu.uniform !42

Flow8399:                                         ; preds = %loop-memcpy-residual5267
  br label %Flow8401, !amdgpu.uniform !42

post-loop-memcpy-expansion5264:                   ; preds = %Flow8401
  %tobool.not.i.i.i.i.i.i911 = icmp eq i32 %.lcssa8996, 0
  %conv.i.i.i.i.i.i912 = zext i1 %tobool.not.i.i.i.i.i.i911 to i32
  fence release
  %1605 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %idxprom.i23.i.i.i.i.i913 = zext i32 %.lcssa8998 to i64
  %arrayidx.i24.i.i.i.i.i914 = getelementptr inbounds nuw i32, ptr %1605, i64 %idxprom.i23.i.i.i.i.i913
  store atomic i32 %conv.i.i.i.i.i.i912, ptr %arrayidx.i24.i.i.i.i.i914 monotonic, align 4
  %1606 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !84
  %arrayidx.i.i5.i.i915 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %1606, i64 %idxprom.i23.i.i.i.i.i913
  %1607 = load i64, ptr %arrayidx.i.i5.i.i915, align 8, !tbaa !80
  br label %while.cond.i.i.i.i916, !amdgpu.uniform !42

while.cond.i.i.i.i916:                            ; preds = %post-loop-memcpy-expansion5264, %Flow8398
  %phi.broken8664 = phi i64 [ 0, %post-loop-memcpy-expansion5264 ], [ %1629, %Flow8398 ]
  %lsr.iv6627 = phi i64 [ %1627, %Flow8398 ], [ -43, %post-loop-memcpy-expansion5264 ]
  %lsr.iv6622 = phi ptr addrspace(4) [ %1626, %Flow8398 ], [ getelementptr (i8, ptr addrspace(4) @.str.17.34, i64 56), %post-loop-memcpy-expansion5264 ]
  %port.sroa.37.0.i.i917 = phi i32 [ %1625, %Flow8398 ], [ %conv.i.i.i.i.i.i912, %post-loop-memcpy-expansion5264 ]
  %idx.0.i.i.i.i918 = phi i64 [ %1624, %Flow8398 ], [ 56, %post-loop-memcpy-expansion5264 ]
  %cmp.i.i.i.i919 = icmp ult i64 %idx.0.i.i.i.i918, sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.17.34 to ptr), i64 13) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.17.34 to ptr) to i64))
  %1608 = call i64 @llvm.amdgcn.ballot.i64(i1 %cmp.i.i.i.i919)
  %and.i.i.i.i6.i.i920 = and i64 %1608, %1607
  %tobool.not.i.i.i.i921 = icmp ne i64 %and.i.i.i.i6.i.i920, 0
  %1609 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i22.i.i.i922 = getelementptr inbounds nuw i32, ptr %1609, i64 %idxprom.i23.i.i.i.i.i913
  %1610 = load atomic i32, ptr %arrayidx.i22.i.i.i922 monotonic, align 4
  %1611 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %tobool.not.i.i.i.i921)
  %1612 = extractvalue { i1, i64 } %1611, 0
  %1613 = extractvalue { i1, i64 } %1611, 1
  br i1 %1612, label %cond.false.i8.i.i.i.i923, label %Flow8398

cond.false.i8.i.i.i.i923:                         ; preds = %while.cond.i.i.i.i916
  %1614 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %1610)
  %1615 = icmp ne i32 %1614, %port.sroa.37.0.i.i917
  %1616 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %1615)
  %1617 = extractvalue { i1, i64 } %1616, 0
  %1618 = extractvalue { i1, i64 } %1616, 1
  br i1 %1617, label %while.body.i.i21.i.i.i.i924.preheader, label %Flow8397

while.body.i.i21.i.i.i.i924.preheader:            ; preds = %cond.false.i8.i.i.i.i923
  br label %while.body.i.i21.i.i.i.i924, !amdgpu.uniform !42

Flow8397:                                         ; preds = %Flow8396, %cond.false.i8.i.i.i.i923
  call void @llvm.amdgcn.end.cf.i64(i64 %1618)
  br label %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i927, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i924:                      ; preds = %while.body.i.i21.i.i.i.i924.preheader, %while.body.i.i21.i.i.i.i924
  %phi.broken8663 = phi i64 [ 0, %while.body.i.i21.i.i.i.i924.preheader ], [ %1622, %while.body.i.i21.i.i.i.i924 ]
  call void @llvm.amdgcn.s.sleep(i32 2)
  %1619 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i22.i.i.i.i925 = getelementptr inbounds nuw i32, ptr %1619, i64 %idxprom.i23.i.i.i.i.i913
  %1620 = load atomic i32, ptr %arrayidx.i.i.i22.i.i.i.i925 monotonic, align 4
  %1621 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %1620)
  %cmp.i.not.i.i23.i.i.i.i926 = icmp eq i32 %1621, %port.sroa.37.0.i.i917
  %1622 = call i64 @llvm.amdgcn.if.break.i64(i1 %cmp.i.not.i.i23.i.i.i.i926, i64 %phi.broken8663)
  %1623 = call i1 @llvm.amdgcn.loop.i64(i64 %1622)
  br i1 %1623, label %Flow8396, label %while.body.i.i21.i.i.i.i924

Flow8396:                                         ; preds = %while.body.i.i21.i.i.i.i924
  %.lcssa8999 = phi i64 [ %1622, %while.body.i.i21.i.i.i.i924 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa8999)
  br label %Flow8397, !amdgpu.uniform !42

Flow8398:                                         ; preds = %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i928, %while.cond.i.i.i.i916
  %1624 = phi i64 [ %add.i.i.i.i931, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i928 ], [ poison, %while.cond.i.i.i.i916 ]
  %1625 = phi i32 [ %conv.i.i26.i.i.i.i929, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i928 ], [ poison, %while.cond.i.i.i.i916 ]
  %1626 = phi ptr addrspace(4) [ %scevgep6623, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i928 ], [ poison, %while.cond.i.i.i.i916 ]
  %1627 = phi i64 [ %lsr.iv.next6628, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i928 ], [ poison, %while.cond.i.i.i.i916 ]
  %1628 = phi i1 [ false, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i928 ], [ true, %while.cond.i.i.i.i916 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %1613)
  %1629 = call i64 @llvm.amdgcn.if.break.i64(i1 %1628, i64 %phi.broken8664)
  %1630 = call i1 @llvm.amdgcn.loop.i64(i64 %1629)
  br i1 %1630, label %cond.false.i.i.i938, label %while.cond.i.i.i.i916

_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i927: ; preds = %Flow8397
  fence acquire
  br i1 %cmp.i.i.i.i919, label %if.then.i.i.i.i.i.i.i932, label %Flow8395, !amdgpu.uniform !42

if.then.i.i.i.i.i.i.i932:                         ; preds = %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i927
  %sub.i.i.i.i.i.i.i934 = sub nuw i64 sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.17.34 to ptr), i64 13) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.17.34 to ptr) to i64)), %idx.0.i.i.i.i918
  %spec.select.i.i.i36.i.i.i.i935 = call i64 @llvm.umin.i64(i64 %sub.i.i.i.i.i.i.i934, i64 64)
  %1631 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %1632 = and i64 %spec.select.i.i.i36.i.i.i.i935, 15
  %1633 = sub i64 %spec.select.i.i.i36.i.i.i.i935, %1632
  %1634 = icmp ne i64 %1633, 0
  br i1 %1634, label %loop-memcpy-expansion5271.preheader, label %Flow8394, !amdgpu.uniform !42

loop-memcpy-expansion5271.preheader:              ; preds = %if.then.i.i.i.i.i.i.i932
  %arrayidx.i20.i35.i.i.i.i936 = getelementptr %"struct.rpc::Buffer", ptr %1631, i64 %idxprom.i19.i.i.i.i.i906
  %arrayidx.i22.i38.i.i.i.i937 = getelementptr %"struct.rpc::Buffer", ptr %arrayidx.i20.i35.i.i.i.i936, i64 %sh_prom.i.i.i.i756
  br label %loop-memcpy-expansion5271, !amdgpu.uniform !42

Flow8394:                                         ; preds = %Flow8393, %if.then.i.i.i.i.i.i.i932
  br label %loop-memcpy-residual-header5274, !amdgpu.uniform !42

loop-memcpy-expansion5271:                        ; preds = %loop-memcpy-expansion5271.preheader, %loop-memcpy-expansion5271
  %loop-index5272 = phi i64 [ %1637, %loop-memcpy-expansion5271 ], [ 0, %loop-memcpy-expansion5271.preheader ]
  %scevgep6624 = getelementptr i8, ptr addrspace(4) %lsr.iv6622, i64 %loop-index5272, !amdgpu.uniform !42
  %1635 = load <4 x i32>, ptr addrspace(4) %scevgep6624, align 1
  %1636 = getelementptr i8, ptr %arrayidx.i22.i38.i.i.i.i937, i64 %loop-index5272
  store <4 x i32> %1635, ptr %1636, align 1
  %1637 = add i64 %loop-index5272, 16
  %1638 = icmp uge i64 %1637, %1633
  br i1 %1638, label %Flow8393, label %loop-memcpy-expansion5271, !amdgpu.uniform !42

Flow8392:                                         ; preds = %Flow8391, %loop-memcpy-residual-header5274
  br label %Flow8395, !amdgpu.uniform !42

loop-memcpy-residual5273:                         ; preds = %loop-memcpy-residual5273.preheader, %loop-memcpy-residual5273
  %residual-loop-index5275 = phi i64 [ %1640, %loop-memcpy-residual5273 ], [ 0, %loop-memcpy-residual5273.preheader ]
  %scevgep6633 = getelementptr i8, ptr addrspace(4) %scevgep6632, i64 %residual-loop-index5275, !amdgpu.uniform !42
  %1639 = load i8, ptr addrspace(4) %scevgep6633, align 1
  %scevgep6631 = getelementptr i8, ptr %scevgep6630, i64 %residual-loop-index5275
  store i8 %1639, ptr %scevgep6631, align 1
  %1640 = add i64 %residual-loop-index5275, 1
  %1641 = icmp uge i64 %1640, %1632
  br i1 %1641, label %Flow8391, label %loop-memcpy-residual5273, !amdgpu.uniform !42

Flow8391:                                         ; preds = %loop-memcpy-residual5273
  br label %Flow8392, !amdgpu.uniform !42

_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i928: ; preds = %Flow8395
  %conv.i.i26.i.i.i.i929 = xor i32 %port.sroa.37.0.i.i917, 1
  fence release
  %1642 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %arrayidx.i24.i29.i.i.i.i930 = getelementptr inbounds nuw i32, ptr %1642, i64 %idxprom.i23.i.i.i.i.i913
  store atomic i32 %conv.i.i26.i.i.i.i929, ptr %arrayidx.i24.i29.i.i.i.i930 monotonic, align 4
  %add.i.i.i.i931 = add i64 %idx.0.i.i.i.i918, 64
  %scevgep6623 = getelementptr i8, ptr addrspace(4) %lsr.iv6622, i64 64
  %lsr.iv.next6628 = add i64 %lsr.iv6627, -64
  br label %Flow8398, !amdgpu.uniform !42

cond.false.i.i.i938:                              ; preds = %Flow8398
  %.lcssa9001 = phi i64 [ %1629, %Flow8398 ]
  %port.sroa.37.0.i.i917.lcssa = phi i32 [ %port.sroa.37.0.i.i917, %Flow8398 ]
  %.lcssa9000 = phi i32 [ %1610, %Flow8398 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa9001)
  %1643 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %.lcssa9000)
  %1644 = icmp ne i32 %1643, %port.sroa.37.0.i.i917.lcssa
  %1645 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %1644)
  %1646 = extractvalue { i1, i64 } %1645, 0
  %1647 = extractvalue { i1, i64 } %1645, 1
  br i1 %1646, label %while.body.i.i.i.i939.preheader, label %Flow8390

while.body.i.i.i.i939.preheader:                  ; preds = %cond.false.i.i.i938
  br label %while.body.i.i.i.i939, !amdgpu.uniform !42

Flow8390:                                         ; preds = %Flow8389, %cond.false.i.i.i938
  call void @llvm.amdgcn.end.cf.i64(i64 %1647)
  br label %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit960, !amdgpu.uniform !42

while.body.i.i.i.i939:                            ; preds = %while.body.i.i.i.i939.preheader, %while.body.i.i.i.i939
  %phi.broken8665 = phi i64 [ 0, %while.body.i.i.i.i939.preheader ], [ %1651, %while.body.i.i.i.i939 ]
  call void @llvm.amdgcn.s.sleep(i32 2)
  %1648 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i8.i.i940 = getelementptr inbounds nuw i32, ptr %1648, i64 %idxprom.i23.i.i.i.i.i913
  %1649 = load atomic i32, ptr %arrayidx.i.i.i8.i.i940 monotonic, align 4
  %1650 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %1649)
  %cmp.i.not.i.i.i.i941 = icmp eq i32 %1650, %port.sroa.37.0.i.i917.lcssa
  %1651 = call i64 @llvm.amdgcn.if.break.i64(i1 %cmp.i.not.i.i.i.i941, i64 %phi.broken8665)
  %1652 = call i1 @llvm.amdgcn.loop.i64(i64 %1651)
  br i1 %1652, label %Flow8389, label %while.body.i.i.i.i939

Flow8389:                                         ; preds = %while.body.i.i.i.i939
  %.lcssa9002 = phi i64 [ %1651, %while.body.i.i.i.i939 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa9002)
  br label %Flow8390, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit960: ; preds = %Flow8390
  fence acquire
  call void @llvm.amdgcn.wave.barrier()
  fence syncscope("agent") release
  %1653 = call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %.lcssa8997, i1 true)
  %iszero.i.i.i.i.i13.i.i943 = icmp ne i64 %.lcssa8997, 0
  %cmp2.i.i.i.i14.i.i944 = icmp eq i64 %1653, %sh_prom.i.i.i.i756
  %cmp.i.i.i.i15.i.i945 = select i1 %iszero.i.i.i.i.i13.i.i943, i1 %cmp2.i.i.i.i14.i.i944, i1 false
  %1654 = and i32 %.lcssa8998, 31
  %conv4.i.i.i19.i.i949 = zext i1 %cmp.i.i.i.i15.i.i945 to i32
  %shl.i.i.i20.i.i950 = shl nuw i32 %conv4.i.i.i19.i.i949, %1654
  %xor.i.i.i21.i.i951 = xor i32 %shl.i.i.i20.i.i950, -1
  %1655 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop5589, !amdgpu.uniform !42

1656:                                             ; preds = %ComputeEnd5590
  %div8.i.i.i16.i.i946 = lshr i32 %.lcssa8998, 5
  %idxprom.i.i.i17.i.i947 = zext nneg i32 %div8.i.i.i16.i.i946 to i64
  %1657 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i17.i.i947
  %sunkaddr7395 = getelementptr inbounds i8, ptr addrspace(1) %1657, i64 40
  %1658 = atomicrmw and ptr addrspace(1) %sunkaddr7395, i32 %.lcssa9003 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %1659, !amdgpu.uniform !42

1659:                                             ; preds = %1656, %ComputeEnd5590
  call void @llvm.amdgcn.end.cf.i64(i64 %7228)
  call void @llvm.amdgcn.wave.barrier()
  br label %for.cond.i.i.i961, !amdgpu.uniform !42

for.cond.i.i.i961:                                ; preds = %1659, %for.cond.i.i.i961
  %phi.broken8666 = phi i64 [ 0, %1659 ], [ %1661, %for.cond.i.i.i961 ]
  %lsr.iv6635 = phi ptr addrspace(4) [ %scevgep6637, %for.cond.i.i.i961 ], [ %scevgep6634, %1659 ]
  %scevgep6637 = getelementptr i8, ptr addrspace(4) %lsr.iv6635, i64 1
  %1660 = load i8, ptr addrspace(4) %scevgep6637, align 1, !tbaa !57
  %cmp.i.i.i963 = icmp eq i8 %1660, 0
  %1661 = call i64 @llvm.amdgcn.if.break.i64(i1 %cmp.i.i.i963, i64 %phi.broken8666)
  %1662 = call i1 @llvm.amdgcn.loop.i64(i64 %1661)
  br i1 %1662, label %for.cond.i.i1.i970.preheader, label %for.cond.i.i.i961

for.cond.i.i1.i970.preheader:                     ; preds = %for.cond.i.i.i961
  %scevgep6637.lcssa = phi ptr addrspace(4) [ %scevgep6637, %for.cond.i.i.i961 ]
  %.lcssa9004 = phi i64 [ %1661, %for.cond.i.i.i961 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa9004)
  br label %for.cond.i.i1.i970, !amdgpu.uniform !42

for.cond.i.i1.i970:                               ; preds = %for.cond.i.i1.i970.preheader, %Flow8388
  %1663 = phi i32 [ %1700, %Flow8388 ], [ %1156, %for.cond.i.i1.i970.preheader ]
  %1664 = phi i64 [ %1701, %Flow8388 ], [ %1157, %for.cond.i.i1.i970.preheader ]
  %1665 = phi i32 [ %1702, %Flow8388 ], [ %1158, %for.cond.i.i1.i970.preheader ]
  %retval.sroa.6.0.i.i.i971 = phi i32 [ %1705, %Flow8388 ], [ undef, %for.cond.i.i1.i970.preheader ]
  %retval.sroa.2.0.i.i.i973 = phi i64 [ %1704, %Flow8388 ], [ undef, %for.cond.i.i1.i970.preheader ]
  %retval.sroa.8.0.i.i.i974 = phi i32 [ %1703, %Flow8388 ], [ undef, %for.cond.i.i1.i970.preheader ]
  %index.0.i.i.i975 = phi i32 [ %1707, %Flow8388 ], [ 0, %for.cond.i.i1.i970.preheader ]
  %1666 = load i32, ptr addrspace(1) @__llvm_rpc_client, align 8, !tbaa !68
  %cmp.not.i.i.i976 = icmp ult i32 %index.0.i.i.i975, %1666
  %spec.store.select.i.i.i977 = select i1 %cmp.not.i.i.i976, i32 %index.0.i.i.i975, i32 0
  %1667 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  %1668 = and i32 %spec.store.select.i.i.i977, 31
  %1669 = lshr i64 %1667, %sh_prom.i.i.i.i756
  %1670 = trunc i64 %1669 to i32
  %conv4.i.i.i.i.i981 = and i32 %1670, 1
  %shl.i.i.i.i.i982 = shl nuw i32 %conv4.i.i.i.i.i981, %1668
  br label %ComputeLoop5593, !amdgpu.uniform !42

1671:                                             ; preds = %ComputeEnd5594
  %sunkaddr7396 = getelementptr inbounds i8, ptr addrspace(1) %7238, i64 40
  %1672 = atomicrmw or ptr addrspace(1) %sunkaddr7396, i32 %.lcssa9005 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %1673, !amdgpu.uniform !42

1673:                                             ; preds = %1671, %ComputeEnd5594
  %1674 = phi i32 [ %1672, %1671 ], [ poison, %ComputeEnd5594 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %7247)
  %1675 = call i32 @llvm.amdgcn.readfirstlane.i32(i32 %1674)
  %1676 = or i32 %1675, %.lcssa9006
  %shl5.i.i.i.i.i983 = shl nuw i32 1, %1668
  %and.i.i.i.i.i984 = and i32 %shl5.i.i.i.i.i983, %1676
  %tobool3.i.i.i.i985 = icmp ne i32 %and.i.i.i.i.i984, 0
  %1677 = call i64 @llvm.amdgcn.ballot.i64(i1 %tobool3.i.i.i.i985)
  %and.i.i.i.i.i.i986 = and i64 %1677, %1667
  %cmp.i.not.i.i.i987 = icmp ne i64 %1667, %and.i.i.i.i.i.i986
  br i1 %cmp.i.not.i.i.i987, label %if.then.i.i.i.i988, label %Flow8387, !amdgpu.uniform !42

if.then.i.i.i.i988:                               ; preds = %1673
  fence syncscope("agent") acquire
  %1678 = load <2 x i64>, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !75
  %1679 = extractelement <2 x i64> %1678, i32 0
  %1680 = inttoptr i64 %1679 to ptr
  %1681 = extractelement <2 x i64> %1678, i32 1
  %1682 = inttoptr i64 %1681 to ptr
  %idxprom.i.i.i.i989 = zext i32 %spec.store.select.i.i.i977 to i64
  %arrayidx.i.i.i.i990 = getelementptr inbounds nuw i32, ptr %1680, i64 %idxprom.i.i.i.i989, !amdgpu.uniform !42
  %1683 = load atomic i32, ptr %arrayidx.i.i.i.i990 monotonic, align 4
  %1684 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %1683)
  %arrayidx.i45.i.i.i991 = getelementptr inbounds nuw i32, ptr %1682, i64 %idxprom.i.i.i.i989, !amdgpu.uniform !42
  %1685 = load atomic i32, ptr %arrayidx.i45.i.i.i991 monotonic, align 4
  %1686 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %1685)
  %cmp.i46.not.i.i.i992 = icmp ne i32 %1684, %1686
  br i1 %cmp.i46.not.i.i.i992, label %if.then12.i.i.i993, label %Flow8385, !amdgpu.uniform !42

if.then12.i.i.i993:                               ; preds = %if.then.i.i.i.i988
  fence syncscope("agent") release
  %1687 = call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %1667, i1 true)
  %iszero.i.i.i.i.i.i.i994 = icmp ne i64 %1667, 0
  %cmp2.i.i.i.i.i.i995 = icmp eq i64 %1687, %sh_prom.i.i.i.i756
  %cmp.i.i.i.i.i.i996 = select i1 %iszero.i.i.i.i.i.i.i994, i1 %cmp2.i.i.i.i.i.i995, i1 false
  %conv4.i.i51.i.i.i997 = zext i1 %cmp.i.i.i.i.i.i996 to i32
  %shl.i.i52.i.i.i998 = shl nuw i32 %conv4.i.i51.i.i.i997, %1668
  %xor.i.i.i.i.i999 = xor i32 %shl.i.i52.i.i.i998, -1
  %1688 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop5598, !amdgpu.uniform !42

1689:                                             ; preds = %ComputeEnd5599
  %sunkaddr7397 = getelementptr inbounds i8, ptr addrspace(1) %7238, i64 40
  %1690 = atomicrmw and ptr addrspace(1) %sunkaddr7397, i32 %.lcssa9007 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %1691, !amdgpu.uniform !42

1691:                                             ; preds = %1689, %ComputeEnd5599
  call void @llvm.amdgcn.end.cf.i64(i64 %7265)
  call void @llvm.amdgcn.wave.barrier()
  br label %Flow8385, !amdgpu.uniform !42

if.end14.i.i.i1059:                               ; preds = %Flow8385
  %1692 = call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %1667, i1 true)
  %iszero.i.i.i.i.i.i1060 = icmp ne i64 %1667, 0
  %cmp2.i.i.i.i.i1061 = icmp eq i64 %1692, %sh_prom.i.i.i.i756
  %cmp.i.i.i.i.i1062 = select i1 %iszero.i.i.i.i.i.i1060, i1 %cmp2.i.i.i.i.i1061, i1 false
  %1693 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.i.i.i.i.i1062)
  %1694 = extractvalue { i1, i64 } %1693, 0
  %1695 = extractvalue { i1, i64 } %1693, 1
  br i1 %1694, label %if.then16.i.i.i1064, label %if.end22.i.i.i1063

if.then16.i.i.i1064:                              ; preds = %if.end14.i.i.i1059
  %1696 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !76
  %opcode.i.i.i1065 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %1696, i64 %idxprom.i.i.i.i989, i32 1
  store i32 1660944387, ptr %opcode.i.i.i1065, align 8, !tbaa !77
  %arrayidx21.i.i.i1066 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %1696, i64 %idxprom.i.i.i.i989
  store i64 %1667, ptr %arrayidx21.i.i.i1066, align 8, !tbaa !80
  br label %if.end22.i.i.i1063, !amdgpu.uniform !42

Flow8386:                                         ; preds = %if.end22.i.i.i1063, %Flow8385
  %1697 = phi i32 [ %1684, %if.end22.i.i.i1063 ], [ %retval.sroa.8.0.i.i.i974, %Flow8385 ]
  %1698 = phi i64 [ %1667, %if.end22.i.i.i1063 ], [ %retval.sroa.2.0.i.i.i973, %Flow8385 ]
  %1699 = phi i32 [ %spec.store.select.i.i.i977, %if.end22.i.i.i1063 ], [ %retval.sroa.6.0.i.i.i971, %Flow8385 ]
  br label %cleanup26.i.i.i1000, !amdgpu.uniform !42

if.end22.i.i.i1063:                               ; preds = %if.then16.i.i.i1064, %if.end14.i.i.i1059
  call void @llvm.amdgcn.end.cf.i64(i64 %1695)
  call void @llvm.amdgcn.wave.barrier()
  br label %Flow8386, !amdgpu.uniform !42

Flow8387:                                         ; preds = %cleanup26.i.i.i1000, %1673
  %1700 = phi i32 [ %1699, %cleanup26.i.i.i1000 ], [ %1663, %1673 ]
  %1701 = phi i64 [ %1698, %cleanup26.i.i.i1000 ], [ %1664, %1673 ]
  %1702 = phi i32 [ %1697, %cleanup26.i.i.i1000 ], [ %1665, %1673 ]
  %1703 = phi i32 [ %1697, %cleanup26.i.i.i1000 ], [ %retval.sroa.8.0.i.i.i974, %1673 ]
  %1704 = phi i64 [ %1698, %cleanup26.i.i.i1000 ], [ %retval.sroa.2.0.i.i.i973, %1673 ]
  %1705 = phi i32 [ %1699, %cleanup26.i.i.i1000 ], [ %retval.sroa.6.0.i.i.i971, %1673 ]
  %1706 = phi i1 [ %cmp.i46.not.i.i.i992, %cleanup26.i.i.i1000 ], [ true, %1673 ]
  br i1 %1706, label %for.inc.i.i.i1005, label %Flow8388, !amdgpu.uniform !42

cleanup26.i.i.i1000:                              ; preds = %Flow8386
  br label %Flow8387, !amdgpu.uniform !42

for.inc.i.i.i1005:                                ; preds = %Flow8387
  %inc.i.i.i1010 = add i32 %spec.store.select.i.i.i977, 1
  br label %Flow8388, !amdgpu.uniform !42

Flow8388:                                         ; preds = %for.inc.i.i.i1005, %Flow8387
  %1707 = phi i32 [ %inc.i.i.i1010, %for.inc.i.i.i1005 ], [ poison, %Flow8387 ]
  %1708 = phi i1 [ false, %for.inc.i.i.i1005 ], [ true, %Flow8387 ]
  br i1 %1708, label %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1011, label %for.cond.i.i1.i970, !amdgpu.uniform !42

_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1011: ; preds = %Flow8388
  %.lcssa9010 = phi i32 [ %1700, %Flow8388 ]
  %.lcssa9009 = phi i64 [ %1701, %Flow8388 ]
  %.lcssa9008 = phi i32 [ %1702, %Flow8388 ]
  %1709 = addrspacecast ptr addrspace(4) %scevgep6637.lcssa to ptr
  %1710 = addrspacecast ptr addrspace(4) %cond6 to ptr
  %1711 = ptrtoint ptr %1710 to i64
  %sub.ptr.lhs.cast.i.i.i966 = ptrtoint ptr %1709 to i64
  %sub.ptr.sub.i.i.i968 = sub i64 %sub.ptr.lhs.cast.i.i.i966, %1711
  fence acquire
  %1712 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %mul.i.i.i.i.i.i1012 = shl i32 %.lcssa9010, 6
  %idxprom.i19.i.i.i.i.i1013 = zext i32 %mul.i.i.i.i.i.i1012 to i64
  %arrayidx.i20.i.i.i.i.i1014 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %1712, i64 %idxprom.i19.i.i.i.i.i1013
  %arrayidx.i22.i.i.i.i.i1015 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %arrayidx.i20.i.i.i.i.i1014, i64 %sh_prom.i.i.i.i756
  store i64 %sub.ptr.sub.i.i.i968, ptr %arrayidx.i22.i.i.i.i.i1015, align 8, !tbaa !82
  %spec.select.i.i.i.i.i.i.i1016 = call i64 @llvm.umin.i64(i64 %sub.ptr.sub.i.i.i968, i64 56)
  %1713 = and i64 %spec.select.i.i.i.i.i.i.i1016, 15
  %1714 = sub i64 %spec.select.i.i.i.i.i.i.i1016, %1713
  %1715 = icmp ne i64 %1714, 0
  %1716 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %1715)
  %1717 = extractvalue { i1, i64 } %1716, 0
  %1718 = extractvalue { i1, i64 } %1716, 1
  br i1 %1717, label %loop-memcpy-expansion5187.preheader, label %Flow8384

loop-memcpy-expansion5187.preheader:              ; preds = %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1011
  %scevgep6638 = getelementptr i8, ptr %1712, i64 8
  %scevgep6639 = getelementptr i8, ptr %scevgep6638, i64 %1474
  %1719 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i1013, 6
  %scevgep6640 = getelementptr i8, ptr %scevgep6639, i64 %1719
  br label %loop-memcpy-expansion5187, !amdgpu.uniform !42

Flow8384:                                         ; preds = %Flow8383, %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1011
  call void @llvm.amdgcn.end.cf.i64(i64 %1718)
  br label %loop-memcpy-residual-header5190, !amdgpu.uniform !42

loop-memcpy-expansion5187:                        ; preds = %loop-memcpy-expansion5187.preheader, %loop-memcpy-expansion5187
  %phi.broken8667 = phi i64 [ 0, %loop-memcpy-expansion5187.preheader ], [ %1724, %loop-memcpy-expansion5187 ]
  %loop-index5188 = phi i64 [ %1722, %loop-memcpy-expansion5187 ], [ 0, %loop-memcpy-expansion5187.preheader ]
  %1720 = getelementptr i8, ptr addrspace(4) %cond6, i64 %loop-index5188
  %1721 = load <4 x i32>, ptr addrspace(4) %1720, align 1
  %scevgep6641 = getelementptr i8, ptr %scevgep6640, i64 %loop-index5188
  store <4 x i32> %1721, ptr %scevgep6641, align 1
  %1722 = add i64 %loop-index5188, 16
  %1723 = icmp uge i64 %1722, %1714
  %1724 = call i64 @llvm.amdgcn.if.break.i64(i1 %1723, i64 %phi.broken8667)
  %1725 = call i1 @llvm.amdgcn.loop.i64(i64 %1724)
  br i1 %1725, label %Flow8383, label %loop-memcpy-expansion5187

Flow8382:                                         ; preds = %Flow8380, %Flow8381
  %1726 = phi i64 [ %6488, %Flow8380 ], [ %6483, %Flow8381 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %6486)
  br label %post-loop-memcpy-expansion5186, !amdgpu.uniform !42

loop-memcpy-residual5189:                         ; preds = %loop-memcpy-residual5189.preheader, %loop-memcpy-residual5189
  %phi.broken8668 = phi i64 [ 0, %loop-memcpy-residual5189.preheader ], [ %1730, %loop-memcpy-residual5189 ]
  %residual-loop-index5191 = phi i64 [ %1728, %loop-memcpy-residual5189 ], [ 0, %loop-memcpy-residual5189.preheader ]
  %scevgep6647 = getelementptr i8, ptr addrspace(4) %scevgep6646, i64 %residual-loop-index5191
  %1727 = load i8, ptr addrspace(4) %scevgep6647, align 1
  %scevgep6645 = getelementptr i8, ptr %scevgep6644, i64 %residual-loop-index5191
  store i8 %1727, ptr %scevgep6645, align 1
  %1728 = add i64 %residual-loop-index5191, 1
  %1729 = icmp uge i64 %1728, %1713
  %1730 = call i64 @llvm.amdgcn.if.break.i64(i1 %1729, i64 %phi.broken8668)
  %1731 = call i1 @llvm.amdgcn.loop.i64(i64 %1730)
  br i1 %1731, label %Flow8380, label %loop-memcpy-residual5189

Flow8380:                                         ; preds = %loop-memcpy-residual5189
  %.lcssa9012 = phi i64 [ %1730, %loop-memcpy-residual5189 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa9012)
  br label %Flow8382, !amdgpu.uniform !42

post-loop-memcpy-expansion5186:                   ; preds = %Flow8382
  %1732 = addrspacecast ptr addrspace(4) %scevgep6637.lcssa to ptr
  %1733 = ptrtoint ptr %1732 to i64
  %tobool.not.i.i.i.i.i.i1018 = icmp eq i32 %.lcssa9008, 0
  %conv.i.i.i.i.i.i1019 = zext i1 %tobool.not.i.i.i.i.i.i1018 to i32
  fence release
  %1734 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %idxprom.i23.i.i.i.i.i1020 = zext i32 %.lcssa9010 to i64
  %arrayidx.i24.i.i.i.i.i1021 = getelementptr inbounds nuw i32, ptr %1734, i64 %idxprom.i23.i.i.i.i.i1020
  store atomic i32 %conv.i.i.i.i.i.i1019, ptr %arrayidx.i24.i.i.i.i.i1021 monotonic, align 4
  %1735 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !84
  %arrayidx.i.i5.i.i1022 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %1735, i64 %idxprom.i23.i.i.i.i.i1020
  %1736 = load i64, ptr %arrayidx.i.i5.i.i1022, align 8, !tbaa !80
  %1737 = add i64 %1141, %1733
  br label %while.cond.i.i.i.i1023, !amdgpu.uniform !42

while.cond.i.i.i.i1023:                           ; preds = %post-loop-memcpy-expansion5186, %Flow8379
  %phi.broken8672 = phi i64 [ 0, %post-loop-memcpy-expansion5186 ], [ %1759, %Flow8379 ]
  %lsr.iv6654 = phi i64 [ %1757, %Flow8379 ], [ %1737, %post-loop-memcpy-expansion5186 ]
  %lsr.iv6649 = phi ptr addrspace(4) [ %1756, %Flow8379 ], [ %scevgep6648, %post-loop-memcpy-expansion5186 ]
  %port.sroa.37.0.i.i1024 = phi i32 [ %1755, %Flow8379 ], [ %conv.i.i.i.i.i.i1019, %post-loop-memcpy-expansion5186 ]
  %idx.0.i.i.i.i1025 = phi i64 [ %1754, %Flow8379 ], [ 56, %post-loop-memcpy-expansion5186 ]
  %cmp.i.i.i.i1026 = icmp ugt i64 %sub.ptr.sub.i.i.i968, %idx.0.i.i.i.i1025
  %1738 = call i64 @llvm.amdgcn.ballot.i64(i1 %cmp.i.i.i.i1026)
  %and.i.i.i.i6.i.i1027 = and i64 %1738, %1736
  %tobool.not.i.i.i.i1028 = icmp ne i64 %and.i.i.i.i6.i.i1027, 0
  %1739 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i22.i.i.i1029 = getelementptr inbounds nuw i32, ptr %1739, i64 %idxprom.i23.i.i.i.i.i1020
  %1740 = load atomic i32, ptr %arrayidx.i22.i.i.i1029 monotonic, align 4
  %1741 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %tobool.not.i.i.i.i1028)
  %1742 = extractvalue { i1, i64 } %1741, 0
  %1743 = extractvalue { i1, i64 } %1741, 1
  br i1 %1742, label %cond.false.i8.i.i.i.i1030, label %Flow8379

cond.false.i8.i.i.i.i1030:                        ; preds = %while.cond.i.i.i.i1023
  %1744 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %1740)
  %1745 = icmp ne i32 %1744, %port.sroa.37.0.i.i1024
  %1746 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %1745)
  %1747 = extractvalue { i1, i64 } %1746, 0
  %1748 = extractvalue { i1, i64 } %1746, 1
  br i1 %1747, label %while.body.i.i21.i.i.i.i1031.preheader, label %Flow8378

while.body.i.i21.i.i.i.i1031.preheader:           ; preds = %cond.false.i8.i.i.i.i1030
  br label %while.body.i.i21.i.i.i.i1031, !amdgpu.uniform !42

Flow8378:                                         ; preds = %Flow8377, %cond.false.i8.i.i.i.i1030
  call void @llvm.amdgcn.end.cf.i64(i64 %1748)
  br label %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i1034, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i1031:                     ; preds = %while.body.i.i21.i.i.i.i1031.preheader, %while.body.i.i21.i.i.i.i1031
  %phi.broken8669 = phi i64 [ 0, %while.body.i.i21.i.i.i.i1031.preheader ], [ %1752, %while.body.i.i21.i.i.i.i1031 ]
  call void @llvm.amdgcn.s.sleep(i32 2)
  %1749 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i22.i.i.i.i1032 = getelementptr inbounds nuw i32, ptr %1749, i64 %idxprom.i23.i.i.i.i.i1020
  %1750 = load atomic i32, ptr %arrayidx.i.i.i22.i.i.i.i1032 monotonic, align 4
  %1751 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %1750)
  %cmp.i.not.i.i23.i.i.i.i1033 = icmp eq i32 %1751, %port.sroa.37.0.i.i1024
  %1752 = call i64 @llvm.amdgcn.if.break.i64(i1 %cmp.i.not.i.i23.i.i.i.i1033, i64 %phi.broken8669)
  %1753 = call i1 @llvm.amdgcn.loop.i64(i64 %1752)
  br i1 %1753, label %Flow8377, label %while.body.i.i21.i.i.i.i1031

Flow8377:                                         ; preds = %while.body.i.i21.i.i.i.i1031
  %.lcssa9013 = phi i64 [ %1752, %while.body.i.i21.i.i.i.i1031 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa9013)
  br label %Flow8378, !amdgpu.uniform !42

Flow8379:                                         ; preds = %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1035, %while.cond.i.i.i.i1023
  %1754 = phi i64 [ %add.i.i.i.i1038, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1035 ], [ poison, %while.cond.i.i.i.i1023 ]
  %1755 = phi i32 [ %conv.i.i26.i.i.i.i1036, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1035 ], [ poison, %while.cond.i.i.i.i1023 ]
  %1756 = phi ptr addrspace(4) [ %scevgep6650, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1035 ], [ poison, %while.cond.i.i.i.i1023 ]
  %1757 = phi i64 [ %lsr.iv.next6655, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1035 ], [ poison, %while.cond.i.i.i.i1023 ]
  %1758 = phi i1 [ false, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1035 ], [ true, %while.cond.i.i.i.i1023 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %1743)
  %1759 = call i64 @llvm.amdgcn.if.break.i64(i1 %1758, i64 %phi.broken8672)
  %1760 = call i1 @llvm.amdgcn.loop.i64(i64 %1759)
  br i1 %1760, label %cond.false.i.i.i1045, label %while.cond.i.i.i.i1023

_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i1034: ; preds = %Flow8378
  fence acquire
  %1761 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.i.i.i.i1026)
  %1762 = extractvalue { i1, i64 } %1761, 0
  %1763 = extractvalue { i1, i64 } %1761, 1
  br i1 %1762, label %if.then.i.i.i.i.i.i.i1039, label %Flow8376

if.then.i.i.i.i.i.i.i1039:                        ; preds = %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i1034
  %sub.i.i.i.i.i.i.i1041 = sub nuw i64 %sub.ptr.sub.i.i.i968, %idx.0.i.i.i.i1025
  %spec.select.i.i.i36.i.i.i.i1042 = call i64 @llvm.umin.i64(i64 %sub.i.i.i.i.i.i.i1041, i64 64)
  %1764 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %1765 = and i64 %spec.select.i.i.i36.i.i.i.i1042, 15
  %1766 = sub i64 %spec.select.i.i.i36.i.i.i.i1042, %1765
  %1767 = icmp ne i64 %1766, 0
  %1768 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %1767)
  %1769 = extractvalue { i1, i64 } %1768, 0
  %1770 = extractvalue { i1, i64 } %1768, 1
  br i1 %1769, label %loop-memcpy-expansion5277.preheader, label %Flow8375

loop-memcpy-expansion5277.preheader:              ; preds = %if.then.i.i.i.i.i.i.i1039
  %arrayidx.i20.i35.i.i.i.i1043 = getelementptr %"struct.rpc::Buffer", ptr %1764, i64 %idxprom.i19.i.i.i.i.i1013
  %arrayidx.i22.i38.i.i.i.i1044 = getelementptr %"struct.rpc::Buffer", ptr %arrayidx.i20.i35.i.i.i.i1043, i64 %sh_prom.i.i.i.i756
  br label %loop-memcpy-expansion5277, !amdgpu.uniform !42

Flow8375:                                         ; preds = %Flow8374, %if.then.i.i.i.i.i.i.i1039
  call void @llvm.amdgcn.end.cf.i64(i64 %1770)
  br label %loop-memcpy-residual-header5280, !amdgpu.uniform !42

loop-memcpy-expansion5277:                        ; preds = %loop-memcpy-expansion5277.preheader, %loop-memcpy-expansion5277
  %phi.broken8670 = phi i64 [ 0, %loop-memcpy-expansion5277.preheader ], [ %1775, %loop-memcpy-expansion5277 ]
  %loop-index5278 = phi i64 [ %1773, %loop-memcpy-expansion5277 ], [ 0, %loop-memcpy-expansion5277.preheader ]
  %scevgep6651 = getelementptr i8, ptr addrspace(4) %lsr.iv6649, i64 %loop-index5278
  %1771 = load <4 x i32>, ptr addrspace(4) %scevgep6651, align 1
  %1772 = getelementptr i8, ptr %arrayidx.i22.i38.i.i.i.i1044, i64 %loop-index5278
  store <4 x i32> %1771, ptr %1772, align 1
  %1773 = add i64 %loop-index5278, 16
  %1774 = icmp uge i64 %1773, %1766
  %1775 = call i64 @llvm.amdgcn.if.break.i64(i1 %1774, i64 %phi.broken8670)
  %1776 = call i1 @llvm.amdgcn.loop.i64(i64 %1775)
  br i1 %1776, label %Flow8374, label %loop-memcpy-expansion5277

Flow8373:                                         ; preds = %Flow8372, %loop-memcpy-residual-header5280
  call void @llvm.amdgcn.end.cf.i64(i64 %6560)
  br label %Flow8376, !amdgpu.uniform !42

loop-memcpy-residual5279:                         ; preds = %loop-memcpy-residual5279.preheader, %loop-memcpy-residual5279
  %phi.broken8671 = phi i64 [ 0, %loop-memcpy-residual5279.preheader ], [ %1780, %loop-memcpy-residual5279 ]
  %residual-loop-index5281 = phi i64 [ %1778, %loop-memcpy-residual5279 ], [ 0, %loop-memcpy-residual5279.preheader ]
  %scevgep6660 = getelementptr i8, ptr addrspace(4) %scevgep6659, i64 %residual-loop-index5281
  %1777 = load i8, ptr addrspace(4) %scevgep6660, align 1
  %scevgep6658 = getelementptr i8, ptr %scevgep6657, i64 %residual-loop-index5281
  store i8 %1777, ptr %scevgep6658, align 1
  %1778 = add i64 %residual-loop-index5281, 1
  %1779 = icmp uge i64 %1778, %1765
  %1780 = call i64 @llvm.amdgcn.if.break.i64(i1 %1779, i64 %phi.broken8671)
  %1781 = call i1 @llvm.amdgcn.loop.i64(i64 %1780)
  br i1 %1781, label %Flow8372, label %loop-memcpy-residual5279

Flow8372:                                         ; preds = %loop-memcpy-residual5279
  %.lcssa9015 = phi i64 [ %1780, %loop-memcpy-residual5279 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa9015)
  br label %Flow8373, !amdgpu.uniform !42

_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1035: ; preds = %Flow8376
  %conv.i.i26.i.i.i.i1036 = xor i32 %port.sroa.37.0.i.i1024, 1
  fence release
  %1782 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %arrayidx.i24.i29.i.i.i.i1037 = getelementptr inbounds nuw i32, ptr %1782, i64 %idxprom.i23.i.i.i.i.i1020
  store atomic i32 %conv.i.i26.i.i.i.i1036, ptr %arrayidx.i24.i29.i.i.i.i1037 monotonic, align 4
  %add.i.i.i.i1038 = add i64 %idx.0.i.i.i.i1025, 64
  %scevgep6650 = getelementptr i8, ptr addrspace(4) %lsr.iv6649, i64 64
  %lsr.iv.next6655 = add i64 %lsr.iv6654, -64
  br label %Flow8379, !amdgpu.uniform !42

cond.false.i.i.i1045:                             ; preds = %Flow8379
  %.lcssa9017 = phi i64 [ %1759, %Flow8379 ]
  %port.sroa.37.0.i.i1024.lcssa = phi i32 [ %port.sroa.37.0.i.i1024, %Flow8379 ]
  %.lcssa9016 = phi i32 [ %1740, %Flow8379 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa9017)
  %1783 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %.lcssa9016)
  %1784 = icmp ne i32 %1783, %port.sroa.37.0.i.i1024.lcssa
  %1785 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %1784)
  %1786 = extractvalue { i1, i64 } %1785, 0
  %1787 = extractvalue { i1, i64 } %1785, 1
  br i1 %1786, label %while.body.i.i.i.i1046.preheader, label %Flow8371

while.body.i.i.i.i1046.preheader:                 ; preds = %cond.false.i.i.i1045
  br label %while.body.i.i.i.i1046, !amdgpu.uniform !42

Flow8371:                                         ; preds = %Flow8370, %cond.false.i.i.i1045
  call void @llvm.amdgcn.end.cf.i64(i64 %1787)
  br label %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit1067, !amdgpu.uniform !42

while.body.i.i.i.i1046:                           ; preds = %while.body.i.i.i.i1046.preheader, %while.body.i.i.i.i1046
  %phi.broken8673 = phi i64 [ 0, %while.body.i.i.i.i1046.preheader ], [ %1791, %while.body.i.i.i.i1046 ]
  call void @llvm.amdgcn.s.sleep(i32 2)
  %1788 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i8.i.i1047 = getelementptr inbounds nuw i32, ptr %1788, i64 %idxprom.i23.i.i.i.i.i1020
  %1789 = load atomic i32, ptr %arrayidx.i.i.i8.i.i1047 monotonic, align 4
  %1790 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %1789)
  %cmp.i.not.i.i.i.i1048 = icmp eq i32 %1790, %port.sroa.37.0.i.i1024.lcssa
  %1791 = call i64 @llvm.amdgcn.if.break.i64(i1 %cmp.i.not.i.i.i.i1048, i64 %phi.broken8673)
  %1792 = call i1 @llvm.amdgcn.loop.i64(i64 %1791)
  br i1 %1792, label %Flow8370, label %while.body.i.i.i.i1046

Flow8370:                                         ; preds = %while.body.i.i.i.i1046
  %.lcssa9018 = phi i64 [ %1791, %while.body.i.i.i.i1046 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa9018)
  br label %Flow8371, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit1067: ; preds = %Flow8371
  fence acquire
  call void @llvm.amdgcn.wave.barrier()
  fence syncscope("agent") release
  %1793 = call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %.lcssa9009, i1 true)
  %iszero.i.i.i.i.i13.i.i1050 = icmp ne i64 %.lcssa9009, 0
  %cmp2.i.i.i.i14.i.i1051 = icmp eq i64 %1793, %sh_prom.i.i.i.i756
  %cmp.i.i.i.i15.i.i1052 = select i1 %iszero.i.i.i.i.i13.i.i1050, i1 %cmp2.i.i.i.i14.i.i1051, i1 false
  %1794 = and i32 %.lcssa9010, 31
  %conv4.i.i.i19.i.i1056 = zext i1 %cmp.i.i.i.i15.i.i1052 to i32
  %shl.i.i.i20.i.i1057 = shl nuw i32 %conv4.i.i.i19.i.i1056, %1794
  %xor.i.i.i21.i.i1058 = xor i32 %shl.i.i.i20.i.i1057, -1
  %1795 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop5602, !amdgpu.uniform !42

1796:                                             ; preds = %ComputeEnd5603
  %div8.i.i.i16.i.i1053 = lshr i32 %.lcssa9010, 5
  %idxprom.i.i.i17.i.i1054 = zext nneg i32 %div8.i.i.i16.i.i1053 to i64
  %1797 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i17.i.i1054
  %sunkaddr7398 = getelementptr inbounds i8, ptr addrspace(1) %1797, i64 40
  %1798 = atomicrmw and ptr addrspace(1) %sunkaddr7398, i32 %.lcssa9019 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %1799, !amdgpu.uniform !42

1799:                                             ; preds = %1796, %ComputeEnd5603
  call void @llvm.amdgcn.end.cf.i64(i64 %7282)
  call void @llvm.amdgcn.wave.barrier()
  %scevgep6661 = getelementptr i8, ptr %call18, i64 -1
  br label %for.cond.i.i.i1068, !amdgpu.uniform !42

for.cond.i.i.i1068:                               ; preds = %1799, %for.cond.i.i.i1068
  %phi.broken8674 = phi i64 [ 0, %1799 ], [ %1801, %for.cond.i.i.i1068 ]
  %lsr.iv6685 = phi i64 [ %lsr.iv.next6686, %for.cond.i.i.i1068 ], [ -56, %1799 ]
  %lsr.iv6671 = phi i64 [ %lsr.iv.next6672, %for.cond.i.i.i1068 ], [ -1, %1799 ]
  %lsr.iv6662 = phi ptr [ %scevgep6664, %for.cond.i.i.i1068 ], [ %scevgep6661, %1799 ]
  %scevgep6664 = getelementptr i8, ptr %lsr.iv6662, i64 1
  %1800 = load i8, ptr %scevgep6664, align 1, !tbaa !57
  %cmp.i.i.i1070 = icmp eq i8 %1800, 0
  %lsr.iv.next6672 = add i64 %lsr.iv6671, 1
  %lsr.iv.next6686 = add i64 %lsr.iv6685, 1
  %1801 = call i64 @llvm.amdgcn.if.break.i64(i1 %cmp.i.i.i1070, i64 %phi.broken8674)
  %1802 = call i1 @llvm.amdgcn.loop.i64(i64 %1801)
  br i1 %1802, label %for.cond.i.i1.i1077.preheader, label %for.cond.i.i.i1068

for.cond.i.i1.i1077.preheader:                    ; preds = %for.cond.i.i.i1068
  %lsr.iv6685.lcssa = phi i64 [ %lsr.iv6685, %for.cond.i.i.i1068 ]
  %scevgep6664.lcssa = phi ptr [ %scevgep6664, %for.cond.i.i.i1068 ]
  %lsr.iv.next6672.lcssa = phi i64 [ %lsr.iv.next6672, %for.cond.i.i.i1068 ]
  %.lcssa9020 = phi i64 [ %1801, %for.cond.i.i.i1068 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa9020)
  br label %for.cond.i.i1.i1077, !amdgpu.uniform !42

for.cond.i.i1.i1077:                              ; preds = %for.cond.i.i1.i1077.preheader, %Flow8369
  %1803 = phi i32 [ %1840, %Flow8369 ], [ %1159, %for.cond.i.i1.i1077.preheader ]
  %1804 = phi i64 [ %1841, %Flow8369 ], [ %1160, %for.cond.i.i1.i1077.preheader ]
  %1805 = phi i32 [ %1842, %Flow8369 ], [ %1161, %for.cond.i.i1.i1077.preheader ]
  %retval.sroa.6.0.i.i.i1078 = phi i32 [ %1845, %Flow8369 ], [ undef, %for.cond.i.i1.i1077.preheader ]
  %retval.sroa.2.0.i.i.i1080 = phi i64 [ %1844, %Flow8369 ], [ undef, %for.cond.i.i1.i1077.preheader ]
  %retval.sroa.8.0.i.i.i1081 = phi i32 [ %1843, %Flow8369 ], [ undef, %for.cond.i.i1.i1077.preheader ]
  %index.0.i.i.i1082 = phi i32 [ %1847, %Flow8369 ], [ 0, %for.cond.i.i1.i1077.preheader ]
  %1806 = load i32, ptr addrspace(1) @__llvm_rpc_client, align 8, !tbaa !68
  %cmp.not.i.i.i1083 = icmp ult i32 %index.0.i.i.i1082, %1806
  %spec.store.select.i.i.i1084 = select i1 %cmp.not.i.i.i1083, i32 %index.0.i.i.i1082, i32 0
  %1807 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  %1808 = and i32 %spec.store.select.i.i.i1084, 31
  %1809 = lshr i64 %1807, %sh_prom.i.i.i.i756
  %1810 = trunc i64 %1809 to i32
  %conv4.i.i.i.i.i1088 = and i32 %1810, 1
  %shl.i.i.i.i.i1089 = shl nuw i32 %conv4.i.i.i.i.i1088, %1808
  br label %ComputeLoop5606, !amdgpu.uniform !42

1811:                                             ; preds = %ComputeEnd5607
  %sunkaddr7399 = getelementptr inbounds i8, ptr addrspace(1) %7292, i64 40
  %1812 = atomicrmw or ptr addrspace(1) %sunkaddr7399, i32 %.lcssa9021 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %1813, !amdgpu.uniform !42

1813:                                             ; preds = %1811, %ComputeEnd5607
  %1814 = phi i32 [ %1812, %1811 ], [ poison, %ComputeEnd5607 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %7301)
  %1815 = call i32 @llvm.amdgcn.readfirstlane.i32(i32 %1814)
  %1816 = or i32 %1815, %.lcssa9022
  %shl5.i.i.i.i.i1090 = shl nuw i32 1, %1808
  %and.i.i.i.i.i1091 = and i32 %shl5.i.i.i.i.i1090, %1816
  %tobool3.i.i.i.i1092 = icmp ne i32 %and.i.i.i.i.i1091, 0
  %1817 = call i64 @llvm.amdgcn.ballot.i64(i1 %tobool3.i.i.i.i1092)
  %and.i.i.i.i.i.i1093 = and i64 %1817, %1807
  %cmp.i.not.i.i.i1094 = icmp ne i64 %1807, %and.i.i.i.i.i.i1093
  br i1 %cmp.i.not.i.i.i1094, label %if.then.i.i.i.i1095, label %Flow8368, !amdgpu.uniform !42

if.then.i.i.i.i1095:                              ; preds = %1813
  fence syncscope("agent") acquire
  %1818 = load <2 x i64>, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !75
  %1819 = extractelement <2 x i64> %1818, i32 0
  %1820 = inttoptr i64 %1819 to ptr
  %1821 = extractelement <2 x i64> %1818, i32 1
  %1822 = inttoptr i64 %1821 to ptr
  %idxprom.i.i.i.i1096 = zext i32 %spec.store.select.i.i.i1084 to i64
  %arrayidx.i.i.i.i1097 = getelementptr inbounds nuw i32, ptr %1820, i64 %idxprom.i.i.i.i1096, !amdgpu.uniform !42
  %1823 = load atomic i32, ptr %arrayidx.i.i.i.i1097 monotonic, align 4
  %1824 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %1823)
  %arrayidx.i45.i.i.i1098 = getelementptr inbounds nuw i32, ptr %1822, i64 %idxprom.i.i.i.i1096, !amdgpu.uniform !42
  %1825 = load atomic i32, ptr %arrayidx.i45.i.i.i1098 monotonic, align 4
  %1826 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %1825)
  %cmp.i46.not.i.i.i1099 = icmp ne i32 %1824, %1826
  br i1 %cmp.i46.not.i.i.i1099, label %if.then12.i.i.i1100, label %Flow8366, !amdgpu.uniform !42

if.then12.i.i.i1100:                              ; preds = %if.then.i.i.i.i1095
  fence syncscope("agent") release
  %1827 = call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %1807, i1 true)
  %iszero.i.i.i.i.i.i.i1101 = icmp ne i64 %1807, 0
  %cmp2.i.i.i.i.i.i1102 = icmp eq i64 %1827, %sh_prom.i.i.i.i756
  %cmp.i.i.i.i.i.i1103 = select i1 %iszero.i.i.i.i.i.i.i1101, i1 %cmp2.i.i.i.i.i.i1102, i1 false
  %conv4.i.i51.i.i.i1104 = zext i1 %cmp.i.i.i.i.i.i1103 to i32
  %shl.i.i52.i.i.i1105 = shl nuw i32 %conv4.i.i51.i.i.i1104, %1808
  %xor.i.i.i.i.i1106 = xor i32 %shl.i.i52.i.i.i1105, -1
  %1828 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop5611, !amdgpu.uniform !42

1829:                                             ; preds = %ComputeEnd5612
  %sunkaddr7400 = getelementptr inbounds i8, ptr addrspace(1) %7292, i64 40
  %1830 = atomicrmw and ptr addrspace(1) %sunkaddr7400, i32 %.lcssa9023 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %1831, !amdgpu.uniform !42

1831:                                             ; preds = %1829, %ComputeEnd5612
  call void @llvm.amdgcn.end.cf.i64(i64 %7319)
  call void @llvm.amdgcn.wave.barrier()
  br label %Flow8366, !amdgpu.uniform !42

if.end14.i.i.i1166:                               ; preds = %Flow8366
  %1832 = call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %1807, i1 true)
  %iszero.i.i.i.i.i.i1167 = icmp ne i64 %1807, 0
  %cmp2.i.i.i.i.i1168 = icmp eq i64 %1832, %sh_prom.i.i.i.i756
  %cmp.i.i.i.i.i1169 = select i1 %iszero.i.i.i.i.i.i1167, i1 %cmp2.i.i.i.i.i1168, i1 false
  %1833 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.i.i.i.i.i1169)
  %1834 = extractvalue { i1, i64 } %1833, 0
  %1835 = extractvalue { i1, i64 } %1833, 1
  br i1 %1834, label %if.then16.i.i.i1171, label %if.end22.i.i.i1170

if.then16.i.i.i1171:                              ; preds = %if.end14.i.i.i1166
  %1836 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !76
  %opcode.i.i.i1172 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %1836, i64 %idxprom.i.i.i.i1096, i32 1
  store i32 1660944387, ptr %opcode.i.i.i1172, align 8, !tbaa !77
  %arrayidx21.i.i.i1173 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %1836, i64 %idxprom.i.i.i.i1096
  store i64 %1807, ptr %arrayidx21.i.i.i1173, align 8, !tbaa !80
  br label %if.end22.i.i.i1170, !amdgpu.uniform !42

Flow8367:                                         ; preds = %if.end22.i.i.i1170, %Flow8366
  %1837 = phi i32 [ %1824, %if.end22.i.i.i1170 ], [ %retval.sroa.8.0.i.i.i1081, %Flow8366 ]
  %1838 = phi i64 [ %1807, %if.end22.i.i.i1170 ], [ %retval.sroa.2.0.i.i.i1080, %Flow8366 ]
  %1839 = phi i32 [ %spec.store.select.i.i.i1084, %if.end22.i.i.i1170 ], [ %retval.sroa.6.0.i.i.i1078, %Flow8366 ]
  br label %cleanup26.i.i.i1107, !amdgpu.uniform !42

if.end22.i.i.i1170:                               ; preds = %if.then16.i.i.i1171, %if.end14.i.i.i1166
  call void @llvm.amdgcn.end.cf.i64(i64 %1835)
  call void @llvm.amdgcn.wave.barrier()
  br label %Flow8367, !amdgpu.uniform !42

Flow8368:                                         ; preds = %cleanup26.i.i.i1107, %1813
  %1840 = phi i32 [ %1839, %cleanup26.i.i.i1107 ], [ %1803, %1813 ]
  %1841 = phi i64 [ %1838, %cleanup26.i.i.i1107 ], [ %1804, %1813 ]
  %1842 = phi i32 [ %1837, %cleanup26.i.i.i1107 ], [ %1805, %1813 ]
  %1843 = phi i32 [ %1837, %cleanup26.i.i.i1107 ], [ %retval.sroa.8.0.i.i.i1081, %1813 ]
  %1844 = phi i64 [ %1838, %cleanup26.i.i.i1107 ], [ %retval.sroa.2.0.i.i.i1080, %1813 ]
  %1845 = phi i32 [ %1839, %cleanup26.i.i.i1107 ], [ %retval.sroa.6.0.i.i.i1078, %1813 ]
  %1846 = phi i1 [ %cmp.i46.not.i.i.i1099, %cleanup26.i.i.i1107 ], [ true, %1813 ]
  br i1 %1846, label %for.inc.i.i.i1112, label %Flow8369, !amdgpu.uniform !42

cleanup26.i.i.i1107:                              ; preds = %Flow8367
  br label %Flow8368, !amdgpu.uniform !42

for.inc.i.i.i1112:                                ; preds = %Flow8368
  %inc.i.i.i1117 = add i32 %spec.store.select.i.i.i1084, 1
  br label %Flow8369, !amdgpu.uniform !42

Flow8369:                                         ; preds = %for.inc.i.i.i1112, %Flow8368
  %1847 = phi i32 [ %inc.i.i.i1117, %for.inc.i.i.i1112 ], [ poison, %Flow8368 ]
  %1848 = phi i1 [ false, %for.inc.i.i.i1112 ], [ true, %Flow8368 ]
  br i1 %1848, label %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1118, label %for.cond.i.i1.i1077, !amdgpu.uniform !42

_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1118: ; preds = %Flow8369
  %.lcssa9026 = phi i32 [ %1840, %Flow8369 ]
  %.lcssa9025 = phi i64 [ %1841, %Flow8369 ]
  %.lcssa9024 = phi i32 [ %1842, %Flow8369 ]
  %sub.ptr.lhs.cast.i.i.i1073 = ptrtoint ptr %scevgep6664.lcssa to i64
  %sub.ptr.rhs.cast.i.i.i1074 = ptrtoint ptr %call18 to i64
  %sub.ptr.sub.i.i.i1075 = sub i64 %sub.ptr.lhs.cast.i.i.i1073, %sub.ptr.rhs.cast.i.i.i1074
  fence acquire
  %1849 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %mul.i.i.i.i.i.i1119 = shl i32 %.lcssa9026, 6
  %idxprom.i19.i.i.i.i.i1120 = zext i32 %mul.i.i.i.i.i.i1119 to i64
  %arrayidx.i20.i.i.i.i.i1121 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %1849, i64 %idxprom.i19.i.i.i.i.i1120
  %arrayidx.i22.i.i.i.i.i1122 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %arrayidx.i20.i.i.i.i.i1121, i64 %sh_prom.i.i.i.i756
  store i64 %sub.ptr.sub.i.i.i1075, ptr %arrayidx.i22.i.i.i.i.i1122, align 8, !tbaa !82
  %spec.select.i.i.i.i.i.i.i1123 = call i64 @llvm.umin.i64(i64 %sub.ptr.sub.i.i.i1075, i64 56)
  %1850 = and i64 %spec.select.i.i.i.i.i.i.i1123, 15
  %1851 = sub i64 %spec.select.i.i.i.i.i.i.i1123, %1850
  %1852 = icmp ne i64 %1851, 0
  %1853 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %1852)
  %1854 = extractvalue { i1, i64 } %1853, 0
  %1855 = extractvalue { i1, i64 } %1853, 1
  br i1 %1854, label %loop-memcpy-expansion5073.preheader, label %Flow8365

loop-memcpy-expansion5073.preheader:              ; preds = %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1118
  %scevgep6665 = getelementptr i8, ptr %1849, i64 8
  %scevgep6666 = getelementptr i8, ptr %scevgep6665, i64 %1474
  %1856 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i1120, 6
  %scevgep6667 = getelementptr i8, ptr %scevgep6666, i64 %1856
  br label %loop-memcpy-expansion5073, !amdgpu.uniform !42

Flow8365:                                         ; preds = %Flow8364, %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1118
  call void @llvm.amdgcn.end.cf.i64(i64 %1855)
  br label %loop-memcpy-residual-header5076, !amdgpu.uniform !42

loop-memcpy-expansion5073:                        ; preds = %loop-memcpy-expansion5073.preheader, %loop-memcpy-expansion5073
  %phi.broken8675 = phi i64 [ 0, %loop-memcpy-expansion5073.preheader ], [ %1861, %loop-memcpy-expansion5073 ]
  %loop-index5074 = phi i64 [ %1859, %loop-memcpy-expansion5073 ], [ 0, %loop-memcpy-expansion5073.preheader ]
  %1857 = getelementptr i8, ptr %call18, i64 %loop-index5074
  %1858 = load <4 x i32>, ptr %1857, align 1
  %scevgep6668 = getelementptr i8, ptr %scevgep6667, i64 %loop-index5074
  store <4 x i32> %1858, ptr %scevgep6668, align 1
  %1859 = add i64 %loop-index5074, 16
  %1860 = icmp uge i64 %1859, %1851
  %1861 = call i64 @llvm.amdgcn.if.break.i64(i1 %1860, i64 %phi.broken8675)
  %1862 = call i1 @llvm.amdgcn.loop.i64(i64 %1861)
  br i1 %1862, label %Flow8364, label %loop-memcpy-expansion5073

Flow8363:                                         ; preds = %Flow8361, %Flow8362
  %1863 = phi i64 [ %6332, %Flow8361 ], [ %6326, %Flow8362 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %6329)
  br label %post-loop-memcpy-expansion5072, !amdgpu.uniform !42

loop-memcpy-residual5075:                         ; preds = %loop-memcpy-residual5075.preheader, %loop-memcpy-residual5075
  %phi.broken8676 = phi i64 [ 0, %loop-memcpy-residual5075.preheader ], [ %1867, %loop-memcpy-residual5075 ]
  %residual-loop-index5077 = phi i64 [ %1865, %loop-memcpy-residual5075 ], [ 0, %loop-memcpy-residual5075.preheader ]
  %scevgep6678 = getelementptr i8, ptr %scevgep6677, i64 %residual-loop-index5077
  %1864 = load i8, ptr %scevgep6678, align 1
  %scevgep6676 = getelementptr i8, ptr %scevgep6675, i64 %residual-loop-index5077
  store i8 %1864, ptr %scevgep6676, align 1
  %1865 = add i64 %residual-loop-index5077, 1
  %1866 = icmp uge i64 %1865, %1850
  %1867 = call i64 @llvm.amdgcn.if.break.i64(i1 %1866, i64 %phi.broken8676)
  %1868 = call i1 @llvm.amdgcn.loop.i64(i64 %1867)
  br i1 %1868, label %Flow8361, label %loop-memcpy-residual5075

Flow8361:                                         ; preds = %loop-memcpy-residual5075
  %.lcssa9028 = phi i64 [ %1867, %loop-memcpy-residual5075 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa9028)
  br label %Flow8363, !amdgpu.uniform !42

post-loop-memcpy-expansion5072:                   ; preds = %Flow8363
  %tobool.not.i.i.i.i.i.i1125 = icmp eq i32 %.lcssa9024, 0
  %conv.i.i.i.i.i.i1126 = zext i1 %tobool.not.i.i.i.i.i.i1125 to i32
  fence release
  %1869 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %idxprom.i23.i.i.i.i.i1127 = zext i32 %.lcssa9026 to i64
  %arrayidx.i24.i.i.i.i.i1128 = getelementptr inbounds nuw i32, ptr %1869, i64 %idxprom.i23.i.i.i.i.i1127
  store atomic i32 %conv.i.i.i.i.i.i1126, ptr %arrayidx.i24.i.i.i.i.i1128 monotonic, align 4
  %1870 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !84
  %arrayidx.i.i5.i.i1129 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %1870, i64 %idxprom.i23.i.i.i.i.i1127
  %1871 = load i64, ptr %arrayidx.i.i5.i.i1129, align 8, !tbaa !80
  %scevgep6679 = getelementptr i8, ptr %call18, i64 56
  br label %while.cond.i.i.i.i1130, !amdgpu.uniform !42

while.cond.i.i.i.i1130:                           ; preds = %post-loop-memcpy-expansion5072, %Flow8360
  %phi.broken8680 = phi i64 [ 0, %post-loop-memcpy-expansion5072 ], [ %1893, %Flow8360 ]
  %lsr.iv6687 = phi i64 [ %1891, %Flow8360 ], [ %lsr.iv6685.lcssa, %post-loop-memcpy-expansion5072 ]
  %lsr.iv6680 = phi ptr [ %1890, %Flow8360 ], [ %scevgep6679, %post-loop-memcpy-expansion5072 ]
  %port.sroa.37.0.i.i1131 = phi i32 [ %1889, %Flow8360 ], [ %conv.i.i.i.i.i.i1126, %post-loop-memcpy-expansion5072 ]
  %idx.0.i.i.i.i1132 = phi i64 [ %1888, %Flow8360 ], [ 56, %post-loop-memcpy-expansion5072 ]
  %cmp.i.i.i.i1133 = icmp ugt i64 %sub.ptr.sub.i.i.i1075, %idx.0.i.i.i.i1132
  %1872 = call i64 @llvm.amdgcn.ballot.i64(i1 %cmp.i.i.i.i1133)
  %and.i.i.i.i6.i.i1134 = and i64 %1872, %1871
  %tobool.not.i.i.i.i1135 = icmp ne i64 %and.i.i.i.i6.i.i1134, 0
  %1873 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i22.i.i.i1136 = getelementptr inbounds nuw i32, ptr %1873, i64 %idxprom.i23.i.i.i.i.i1127
  %1874 = load atomic i32, ptr %arrayidx.i22.i.i.i1136 monotonic, align 4
  %1875 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %tobool.not.i.i.i.i1135)
  %1876 = extractvalue { i1, i64 } %1875, 0
  %1877 = extractvalue { i1, i64 } %1875, 1
  br i1 %1876, label %cond.false.i8.i.i.i.i1137, label %Flow8360

cond.false.i8.i.i.i.i1137:                        ; preds = %while.cond.i.i.i.i1130
  %1878 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %1874)
  %1879 = icmp ne i32 %1878, %port.sroa.37.0.i.i1131
  %1880 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %1879)
  %1881 = extractvalue { i1, i64 } %1880, 0
  %1882 = extractvalue { i1, i64 } %1880, 1
  br i1 %1881, label %while.body.i.i21.i.i.i.i1138.preheader, label %Flow8359

while.body.i.i21.i.i.i.i1138.preheader:           ; preds = %cond.false.i8.i.i.i.i1137
  br label %while.body.i.i21.i.i.i.i1138, !amdgpu.uniform !42

Flow8359:                                         ; preds = %Flow8358, %cond.false.i8.i.i.i.i1137
  call void @llvm.amdgcn.end.cf.i64(i64 %1882)
  br label %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i1141, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i1138:                     ; preds = %while.body.i.i21.i.i.i.i1138.preheader, %while.body.i.i21.i.i.i.i1138
  %phi.broken8677 = phi i64 [ 0, %while.body.i.i21.i.i.i.i1138.preheader ], [ %1886, %while.body.i.i21.i.i.i.i1138 ]
  call void @llvm.amdgcn.s.sleep(i32 2)
  %1883 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i22.i.i.i.i1139 = getelementptr inbounds nuw i32, ptr %1883, i64 %idxprom.i23.i.i.i.i.i1127
  %1884 = load atomic i32, ptr %arrayidx.i.i.i22.i.i.i.i1139 monotonic, align 4
  %1885 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %1884)
  %cmp.i.not.i.i23.i.i.i.i1140 = icmp eq i32 %1885, %port.sroa.37.0.i.i1131
  %1886 = call i64 @llvm.amdgcn.if.break.i64(i1 %cmp.i.not.i.i23.i.i.i.i1140, i64 %phi.broken8677)
  %1887 = call i1 @llvm.amdgcn.loop.i64(i64 %1886)
  br i1 %1887, label %Flow8358, label %while.body.i.i21.i.i.i.i1138

Flow8358:                                         ; preds = %while.body.i.i21.i.i.i.i1138
  %.lcssa9029 = phi i64 [ %1886, %while.body.i.i21.i.i.i.i1138 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa9029)
  br label %Flow8359, !amdgpu.uniform !42

Flow8360:                                         ; preds = %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1142, %while.cond.i.i.i.i1130
  %1888 = phi i64 [ %add.i.i.i.i1145, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1142 ], [ poison, %while.cond.i.i.i.i1130 ]
  %1889 = phi i32 [ %conv.i.i26.i.i.i.i1143, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1142 ], [ poison, %while.cond.i.i.i.i1130 ]
  %1890 = phi ptr [ %scevgep6681, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1142 ], [ poison, %while.cond.i.i.i.i1130 ]
  %1891 = phi i64 [ %lsr.iv.next6688, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1142 ], [ poison, %while.cond.i.i.i.i1130 ]
  %1892 = phi i1 [ false, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1142 ], [ true, %while.cond.i.i.i.i1130 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %1877)
  %1893 = call i64 @llvm.amdgcn.if.break.i64(i1 %1892, i64 %phi.broken8680)
  %1894 = call i1 @llvm.amdgcn.loop.i64(i64 %1893)
  br i1 %1894, label %cond.false.i.i.i1152, label %while.cond.i.i.i.i1130

_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i1141: ; preds = %Flow8359
  fence acquire
  %1895 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.i.i.i.i1133)
  %1896 = extractvalue { i1, i64 } %1895, 0
  %1897 = extractvalue { i1, i64 } %1895, 1
  br i1 %1896, label %if.then.i.i.i.i.i.i.i1146, label %Flow8357

if.then.i.i.i.i.i.i.i1146:                        ; preds = %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i1141
  %sub.i.i.i.i.i.i.i1148 = sub nuw i64 %sub.ptr.sub.i.i.i1075, %idx.0.i.i.i.i1132
  %spec.select.i.i.i36.i.i.i.i1149 = call i64 @llvm.umin.i64(i64 %sub.i.i.i.i.i.i.i1148, i64 64)
  %1898 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %1899 = and i64 %spec.select.i.i.i36.i.i.i.i1149, 15
  %1900 = sub i64 %spec.select.i.i.i36.i.i.i.i1149, %1899
  %1901 = icmp ne i64 %1900, 0
  %1902 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %1901)
  %1903 = extractvalue { i1, i64 } %1902, 0
  %1904 = extractvalue { i1, i64 } %1902, 1
  br i1 %1903, label %loop-memcpy-expansion5067.preheader, label %Flow8356

loop-memcpy-expansion5067.preheader:              ; preds = %if.then.i.i.i.i.i.i.i1146
  %arrayidx.i20.i35.i.i.i.i1150 = getelementptr %"struct.rpc::Buffer", ptr %1898, i64 %idxprom.i19.i.i.i.i.i1120
  %arrayidx.i22.i38.i.i.i.i1151 = getelementptr %"struct.rpc::Buffer", ptr %arrayidx.i20.i35.i.i.i.i1150, i64 %sh_prom.i.i.i.i756
  br label %loop-memcpy-expansion5067, !amdgpu.uniform !42

Flow8356:                                         ; preds = %Flow8355, %if.then.i.i.i.i.i.i.i1146
  call void @llvm.amdgcn.end.cf.i64(i64 %1904)
  br label %loop-memcpy-residual-header5070, !amdgpu.uniform !42

loop-memcpy-expansion5067:                        ; preds = %loop-memcpy-expansion5067.preheader, %loop-memcpy-expansion5067
  %phi.broken8678 = phi i64 [ 0, %loop-memcpy-expansion5067.preheader ], [ %1909, %loop-memcpy-expansion5067 ]
  %loop-index5068 = phi i64 [ %1907, %loop-memcpy-expansion5067 ], [ 0, %loop-memcpy-expansion5067.preheader ]
  %scevgep6682 = getelementptr i8, ptr %lsr.iv6680, i64 %loop-index5068
  %1905 = load <4 x i32>, ptr %scevgep6682, align 1
  %1906 = getelementptr i8, ptr %arrayidx.i22.i38.i.i.i.i1151, i64 %loop-index5068
  store <4 x i32> %1905, ptr %1906, align 1
  %1907 = add i64 %loop-index5068, 16
  %1908 = icmp uge i64 %1907, %1900
  %1909 = call i64 @llvm.amdgcn.if.break.i64(i1 %1908, i64 %phi.broken8678)
  %1910 = call i1 @llvm.amdgcn.loop.i64(i64 %1909)
  br i1 %1910, label %Flow8355, label %loop-memcpy-expansion5067

Flow8354:                                         ; preds = %Flow8353, %loop-memcpy-residual-header5070
  call void @llvm.amdgcn.end.cf.i64(i64 %6319)
  br label %Flow8357, !amdgpu.uniform !42

loop-memcpy-residual5069:                         ; preds = %loop-memcpy-residual5069.preheader, %loop-memcpy-residual5069
  %phi.broken8679 = phi i64 [ 0, %loop-memcpy-residual5069.preheader ], [ %1914, %loop-memcpy-residual5069 ]
  %residual-loop-index5071 = phi i64 [ %1912, %loop-memcpy-residual5069 ], [ 0, %loop-memcpy-residual5069.preheader ]
  %scevgep6694 = getelementptr i8, ptr %scevgep6693, i64 %residual-loop-index5071
  %1911 = load i8, ptr %scevgep6694, align 1
  %scevgep6691 = getelementptr i8, ptr %scevgep6690, i64 %residual-loop-index5071
  store i8 %1911, ptr %scevgep6691, align 1
  %1912 = add i64 %residual-loop-index5071, 1
  %1913 = icmp uge i64 %1912, %1899
  %1914 = call i64 @llvm.amdgcn.if.break.i64(i1 %1913, i64 %phi.broken8679)
  %1915 = call i1 @llvm.amdgcn.loop.i64(i64 %1914)
  br i1 %1915, label %Flow8353, label %loop-memcpy-residual5069

Flow8353:                                         ; preds = %loop-memcpy-residual5069
  %.lcssa9031 = phi i64 [ %1914, %loop-memcpy-residual5069 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa9031)
  br label %Flow8354, !amdgpu.uniform !42

_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1142: ; preds = %Flow8357
  %conv.i.i26.i.i.i.i1143 = xor i32 %port.sroa.37.0.i.i1131, 1
  fence release
  %1916 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %arrayidx.i24.i29.i.i.i.i1144 = getelementptr inbounds nuw i32, ptr %1916, i64 %idxprom.i23.i.i.i.i.i1127
  store atomic i32 %conv.i.i26.i.i.i.i1143, ptr %arrayidx.i24.i29.i.i.i.i1144 monotonic, align 4
  %add.i.i.i.i1145 = add i64 %idx.0.i.i.i.i1132, 64
  %scevgep6681 = getelementptr i8, ptr %lsr.iv6680, i64 64
  %lsr.iv.next6688 = add i64 %lsr.iv6687, -64
  br label %Flow8360, !amdgpu.uniform !42

cond.false.i.i.i1152:                             ; preds = %Flow8360
  %.lcssa9033 = phi i64 [ %1893, %Flow8360 ]
  %port.sroa.37.0.i.i1131.lcssa = phi i32 [ %port.sroa.37.0.i.i1131, %Flow8360 ]
  %.lcssa9032 = phi i32 [ %1874, %Flow8360 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa9033)
  %1917 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %.lcssa9032)
  %1918 = icmp ne i32 %1917, %port.sroa.37.0.i.i1131.lcssa
  %1919 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %1918)
  %1920 = extractvalue { i1, i64 } %1919, 0
  %1921 = extractvalue { i1, i64 } %1919, 1
  br i1 %1920, label %while.body.i.i.i.i1153.preheader, label %Flow8352

while.body.i.i.i.i1153.preheader:                 ; preds = %cond.false.i.i.i1152
  br label %while.body.i.i.i.i1153, !amdgpu.uniform !42

Flow8352:                                         ; preds = %Flow8351, %cond.false.i.i.i1152
  call void @llvm.amdgcn.end.cf.i64(i64 %1921)
  br label %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit1174, !amdgpu.uniform !42

while.body.i.i.i.i1153:                           ; preds = %while.body.i.i.i.i1153.preheader, %while.body.i.i.i.i1153
  %phi.broken8681 = phi i64 [ 0, %while.body.i.i.i.i1153.preheader ], [ %1925, %while.body.i.i.i.i1153 ]
  call void @llvm.amdgcn.s.sleep(i32 2)
  %1922 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i8.i.i1154 = getelementptr inbounds nuw i32, ptr %1922, i64 %idxprom.i23.i.i.i.i.i1127
  %1923 = load atomic i32, ptr %arrayidx.i.i.i8.i.i1154 monotonic, align 4
  %1924 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %1923)
  %cmp.i.not.i.i.i.i1155 = icmp eq i32 %1924, %port.sroa.37.0.i.i1131.lcssa
  %1925 = call i64 @llvm.amdgcn.if.break.i64(i1 %cmp.i.not.i.i.i.i1155, i64 %phi.broken8681)
  %1926 = call i1 @llvm.amdgcn.loop.i64(i64 %1925)
  br i1 %1926, label %Flow8351, label %while.body.i.i.i.i1153

Flow8351:                                         ; preds = %while.body.i.i.i.i1153
  %.lcssa9034 = phi i64 [ %1925, %while.body.i.i.i.i1153 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa9034)
  br label %Flow8352, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit1174: ; preds = %Flow8352
  fence acquire
  call void @llvm.amdgcn.wave.barrier()
  fence syncscope("agent") release
  %1927 = call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %.lcssa9025, i1 true)
  %iszero.i.i.i.i.i13.i.i1157 = icmp ne i64 %.lcssa9025, 0
  %cmp2.i.i.i.i14.i.i1158 = icmp eq i64 %1927, %sh_prom.i.i.i.i756
  %cmp.i.i.i.i15.i.i1159 = select i1 %iszero.i.i.i.i.i13.i.i1157, i1 %cmp2.i.i.i.i14.i.i1158, i1 false
  %1928 = and i32 %.lcssa9026, 31
  %conv4.i.i.i19.i.i1163 = zext i1 %cmp.i.i.i.i15.i.i1159 to i32
  %shl.i.i.i20.i.i1164 = shl nuw i32 %conv4.i.i.i19.i.i1163, %1928
  %xor.i.i.i21.i.i1165 = xor i32 %shl.i.i.i20.i.i1164, -1
  %1929 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop5615, !amdgpu.uniform !42

1930:                                             ; preds = %ComputeEnd5616
  %div8.i.i.i16.i.i1160 = lshr i32 %.lcssa9026, 5
  %idxprom.i.i.i17.i.i1161 = zext nneg i32 %div8.i.i.i16.i.i1160 to i64
  %1931 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i17.i.i1161
  %sunkaddr7401 = getelementptr inbounds i8, ptr addrspace(1) %1931, i64 40
  %1932 = atomicrmw and ptr addrspace(1) %sunkaddr7401, i32 %.lcssa9035 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %1933, !amdgpu.uniform !42

1933:                                             ; preds = %1930, %ComputeEnd5616
  call void @llvm.amdgcn.end.cf.i64(i64 %7336)
  call void @llvm.amdgcn.wave.barrier()
  br label %for.cond.i.i.i1176, !amdgpu.uniform !42

for.cond.i.i.i1176:                               ; preds = %1933, %Flow8350
  %1934 = phi i32 [ %1971, %Flow8350 ], [ %1162, %1933 ]
  %1935 = phi i64 [ %1972, %Flow8350 ], [ %1163, %1933 ]
  %1936 = phi i32 [ %1973, %Flow8350 ], [ %1164, %1933 ]
  %retval.sroa.6.0.i.i.i1177 = phi i32 [ %1976, %Flow8350 ], [ undef, %1933 ]
  %retval.sroa.2.0.i.i.i1179 = phi i64 [ %1975, %Flow8350 ], [ undef, %1933 ]
  %retval.sroa.8.0.i.i.i1180 = phi i32 [ %1974, %Flow8350 ], [ undef, %1933 ]
  %index.0.i.i.i1181 = phi i32 [ %1978, %Flow8350 ], [ 0, %1933 ]
  %1937 = load i32, ptr addrspace(1) @__llvm_rpc_client, align 8, !tbaa !68
  %cmp.not.i.i.i1182 = icmp ult i32 %index.0.i.i.i1181, %1937
  %spec.store.select.i.i.i1183 = select i1 %cmp.not.i.i.i1182, i32 %index.0.i.i.i1181, i32 0
  %1938 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  %1939 = and i32 %spec.store.select.i.i.i1183, 31
  %1940 = lshr i64 %1938, %sh_prom.i.i.i.i756
  %1941 = trunc i64 %1940 to i32
  %conv4.i.i.i.i.i1187 = and i32 %1941, 1
  %shl.i.i.i.i.i1188 = shl nuw i32 %conv4.i.i.i.i.i1187, %1939
  br label %ComputeLoop5619, !amdgpu.uniform !42

1942:                                             ; preds = %ComputeEnd5620
  %sunkaddr7402 = getelementptr inbounds i8, ptr addrspace(1) %7346, i64 40
  %1943 = atomicrmw or ptr addrspace(1) %sunkaddr7402, i32 %.lcssa9036 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %1944, !amdgpu.uniform !42

1944:                                             ; preds = %1942, %ComputeEnd5620
  %1945 = phi i32 [ %1943, %1942 ], [ poison, %ComputeEnd5620 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %7355)
  %1946 = call i32 @llvm.amdgcn.readfirstlane.i32(i32 %1945)
  %1947 = or i32 %1946, %.lcssa9037
  %shl5.i.i.i.i.i1189 = shl nuw i32 1, %1939
  %and.i.i.i.i.i1190 = and i32 %shl5.i.i.i.i.i1189, %1947
  %tobool3.i.i.i.i1191 = icmp ne i32 %and.i.i.i.i.i1190, 0
  %1948 = call i64 @llvm.amdgcn.ballot.i64(i1 %tobool3.i.i.i.i1191)
  %and.i.i.i.i.i.i1192 = and i64 %1948, %1938
  %cmp.i.not.i.i.i1193 = icmp ne i64 %1938, %and.i.i.i.i.i.i1192
  br i1 %cmp.i.not.i.i.i1193, label %if.then.i.i.i.i1194, label %Flow8349, !amdgpu.uniform !42

if.then.i.i.i.i1194:                              ; preds = %1944
  fence syncscope("agent") acquire
  %1949 = load <2 x i64>, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !75
  %1950 = extractelement <2 x i64> %1949, i32 0
  %1951 = inttoptr i64 %1950 to ptr
  %1952 = extractelement <2 x i64> %1949, i32 1
  %1953 = inttoptr i64 %1952 to ptr
  %idxprom.i.i.i.i1195 = zext i32 %spec.store.select.i.i.i1183 to i64
  %arrayidx.i.i.i.i1196 = getelementptr inbounds nuw i32, ptr %1951, i64 %idxprom.i.i.i.i1195, !amdgpu.uniform !42
  %1954 = load atomic i32, ptr %arrayidx.i.i.i.i1196 monotonic, align 4
  %1955 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %1954)
  %arrayidx.i45.i.i.i1197 = getelementptr inbounds nuw i32, ptr %1953, i64 %idxprom.i.i.i.i1195, !amdgpu.uniform !42
  %1956 = load atomic i32, ptr %arrayidx.i45.i.i.i1197 monotonic, align 4
  %1957 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %1956)
  %cmp.i46.not.i.i.i1198 = icmp ne i32 %1955, %1957
  br i1 %cmp.i46.not.i.i.i1198, label %if.then12.i.i.i1199, label %Flow8347, !amdgpu.uniform !42

if.then12.i.i.i1199:                              ; preds = %if.then.i.i.i.i1194
  fence syncscope("agent") release
  %1958 = call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %1938, i1 true)
  %iszero.i.i.i.i.i.i.i1200 = icmp ne i64 %1938, 0
  %cmp2.i.i.i.i.i.i1201 = icmp eq i64 %1958, %sh_prom.i.i.i.i756
  %cmp.i.i.i.i.i.i1202 = select i1 %iszero.i.i.i.i.i.i.i1200, i1 %cmp2.i.i.i.i.i.i1201, i1 false
  %conv4.i.i51.i.i.i1203 = zext i1 %cmp.i.i.i.i.i.i1202 to i32
  %shl.i.i52.i.i.i1204 = shl nuw i32 %conv4.i.i51.i.i.i1203, %1939
  %xor.i.i.i.i.i1205 = xor i32 %shl.i.i52.i.i.i1204, -1
  %1959 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop5624, !amdgpu.uniform !42

1960:                                             ; preds = %ComputeEnd5625
  %sunkaddr7403 = getelementptr inbounds i8, ptr addrspace(1) %7346, i64 40
  %1961 = atomicrmw and ptr addrspace(1) %sunkaddr7403, i32 %.lcssa9038 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %1962, !amdgpu.uniform !42

1962:                                             ; preds = %1960, %ComputeEnd5625
  call void @llvm.amdgcn.end.cf.i64(i64 %7373)
  call void @llvm.amdgcn.wave.barrier()
  br label %Flow8347, !amdgpu.uniform !42

if.end14.i.i.i1263:                               ; preds = %Flow8347
  %1963 = call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %1938, i1 true)
  %iszero.i.i.i.i.i.i1264 = icmp ne i64 %1938, 0
  %cmp2.i.i.i.i.i1265 = icmp eq i64 %1963, %sh_prom.i.i.i.i756
  %cmp.i.i.i.i.i1266 = select i1 %iszero.i.i.i.i.i.i1264, i1 %cmp2.i.i.i.i.i1265, i1 false
  %1964 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.i.i.i.i.i1266)
  %1965 = extractvalue { i1, i64 } %1964, 0
  %1966 = extractvalue { i1, i64 } %1964, 1
  br i1 %1965, label %if.then16.i.i.i1268, label %if.end22.i.i.i1267

if.then16.i.i.i1268:                              ; preds = %if.end14.i.i.i1263
  %1967 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !76
  %opcode.i.i.i1269 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %1967, i64 %idxprom.i.i.i.i1195, i32 1
  store i32 1660944387, ptr %opcode.i.i.i1269, align 8, !tbaa !77
  %arrayidx21.i.i.i1270 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %1967, i64 %idxprom.i.i.i.i1195
  store i64 %1938, ptr %arrayidx21.i.i.i1270, align 8, !tbaa !80
  br label %if.end22.i.i.i1267, !amdgpu.uniform !42

Flow8348:                                         ; preds = %if.end22.i.i.i1267, %Flow8347
  %1968 = phi i32 [ %1955, %if.end22.i.i.i1267 ], [ %retval.sroa.8.0.i.i.i1180, %Flow8347 ]
  %1969 = phi i64 [ %1938, %if.end22.i.i.i1267 ], [ %retval.sroa.2.0.i.i.i1179, %Flow8347 ]
  %1970 = phi i32 [ %spec.store.select.i.i.i1183, %if.end22.i.i.i1267 ], [ %retval.sroa.6.0.i.i.i1177, %Flow8347 ]
  br label %cleanup26.i.i.i1206, !amdgpu.uniform !42

if.end22.i.i.i1267:                               ; preds = %if.then16.i.i.i1268, %if.end14.i.i.i1263
  call void @llvm.amdgcn.end.cf.i64(i64 %1966)
  call void @llvm.amdgcn.wave.barrier()
  br label %Flow8348, !amdgpu.uniform !42

Flow8349:                                         ; preds = %cleanup26.i.i.i1206, %1944
  %1971 = phi i32 [ %1970, %cleanup26.i.i.i1206 ], [ %1934, %1944 ]
  %1972 = phi i64 [ %1969, %cleanup26.i.i.i1206 ], [ %1935, %1944 ]
  %1973 = phi i32 [ %1968, %cleanup26.i.i.i1206 ], [ %1936, %1944 ]
  %1974 = phi i32 [ %1968, %cleanup26.i.i.i1206 ], [ %retval.sroa.8.0.i.i.i1180, %1944 ]
  %1975 = phi i64 [ %1969, %cleanup26.i.i.i1206 ], [ %retval.sroa.2.0.i.i.i1179, %1944 ]
  %1976 = phi i32 [ %1970, %cleanup26.i.i.i1206 ], [ %retval.sroa.6.0.i.i.i1177, %1944 ]
  %1977 = phi i1 [ %cmp.i46.not.i.i.i1198, %cleanup26.i.i.i1206 ], [ true, %1944 ]
  br i1 %1977, label %for.inc.i.i.i1211, label %Flow8350, !amdgpu.uniform !42

cleanup26.i.i.i1206:                              ; preds = %Flow8348
  br label %Flow8349, !amdgpu.uniform !42

for.inc.i.i.i1211:                                ; preds = %Flow8349
  %inc.i.i.i1216 = add i32 %spec.store.select.i.i.i1183, 1
  br label %Flow8350, !amdgpu.uniform !42

Flow8350:                                         ; preds = %for.inc.i.i.i1211, %Flow8349
  %1978 = phi i32 [ %inc.i.i.i1216, %for.inc.i.i.i1211 ], [ poison, %Flow8349 ]
  %1979 = phi i1 [ false, %for.inc.i.i.i1211 ], [ true, %Flow8349 ]
  br i1 %1979, label %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1217, label %for.cond.i.i.i1176, !amdgpu.uniform !42

_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1217: ; preds = %Flow8350
  %.lcssa9041 = phi i32 [ %1971, %Flow8350 ]
  %.lcssa9040 = phi i64 [ %1972, %Flow8350 ]
  %.lcssa9039 = phi i32 [ %1973, %Flow8350 ]
  fence acquire
  %1980 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %mul.i.i.i.i.i.i1218 = shl i32 %.lcssa9041, 6
  %idxprom.i19.i.i.i.i.i1219 = zext i32 %mul.i.i.i.i.i.i1218 to i64
  %arrayidx.i20.i.i.i.i.i1220 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %1980, i64 %idxprom.i19.i.i.i.i.i1219
  %arrayidx.i22.i.i.i.i.i1221 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %arrayidx.i20.i.i.i.i.i1220, i64 %sh_prom.i.i.i.i756
  store i64 1, ptr %arrayidx.i22.i.i.i.i.i1221, align 8, !tbaa !82
  %arrayidx6.i.i.i.i.i.i.i1222 = getelementptr inbounds nuw i8, ptr %arrayidx.i22.i.i.i.i.i1221, i64 8
  store i8 10, ptr %arrayidx6.i.i.i.i.i.i.i1222, align 8
  %tobool.not.i.i.i.i.i.i1223 = icmp eq i32 %.lcssa9039, 0
  %conv.i.i.i.i.i.i1224 = zext i1 %tobool.not.i.i.i.i.i.i1223 to i32
  fence release
  %1981 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %idxprom.i23.i.i.i.i.i1225 = zext i32 %.lcssa9041 to i64
  %arrayidx.i24.i.i.i.i.i1226 = getelementptr inbounds nuw i32, ptr %1981, i64 %idxprom.i23.i.i.i.i.i1225
  store atomic i32 %conv.i.i.i.i.i.i1224, ptr %arrayidx.i24.i.i.i.i.i1226 monotonic, align 4
  %1982 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !84
  %arrayidx.i.i5.i.i1227 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %1982, i64 %idxprom.i23.i.i.i.i.i1225
  %1983 = load i64, ptr %arrayidx.i.i5.i.i1227, align 8, !tbaa !80
  br label %while.cond.i.i.i.i1228, !amdgpu.uniform !42

while.cond.i.i.i.i1228:                           ; preds = %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1217, %Flow8346
  %phi.broken8683 = phi i64 [ 0, %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1217 ], [ %2003, %Flow8346 ]
  %port.sroa.37.0.i.i1229 = phi i32 [ %2001, %Flow8346 ], [ %conv.i.i.i.i.i.i1224, %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1217 ]
  %idx.0.i.i.i.i1230 = phi i64 [ %2000, %Flow8346 ], [ 56, %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1217 ]
  %cmp.i.i.i.i1231 = icmp eq i64 %idx.0.i.i.i.i1230, 0
  %1984 = call i64 @llvm.amdgcn.ballot.i64(i1 %cmp.i.i.i.i1231)
  %and.i.i.i.i6.i.i1232 = and i64 %1984, %1983
  %tobool.not.i.i.i.i1233 = icmp ne i64 %and.i.i.i.i6.i.i1232, 0
  %1985 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i22.i.i.i1234 = getelementptr inbounds nuw i32, ptr %1985, i64 %idxprom.i23.i.i.i.i.i1225
  %1986 = load atomic i32, ptr %arrayidx.i22.i.i.i1234 monotonic, align 4
  %1987 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %tobool.not.i.i.i.i1233)
  %1988 = extractvalue { i1, i64 } %1987, 0
  %1989 = extractvalue { i1, i64 } %1987, 1
  br i1 %1988, label %cond.false.i8.i.i.i.i1235, label %Flow8346

cond.false.i8.i.i.i.i1235:                        ; preds = %while.cond.i.i.i.i1228
  %1990 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %1986)
  %1991 = icmp ne i32 %1990, %port.sroa.37.0.i.i1229
  %1992 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %1991)
  %1993 = extractvalue { i1, i64 } %1992, 0
  %1994 = extractvalue { i1, i64 } %1992, 1
  br i1 %1993, label %while.body.i.i21.i.i.i.i1236.preheader, label %Flow8345

while.body.i.i21.i.i.i.i1236.preheader:           ; preds = %cond.false.i8.i.i.i.i1235
  br label %while.body.i.i21.i.i.i.i1236, !amdgpu.uniform !42

Flow8345:                                         ; preds = %Flow8344, %cond.false.i8.i.i.i.i1235
  call void @llvm.amdgcn.end.cf.i64(i64 %1994)
  br label %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i1239, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i1236:                     ; preds = %while.body.i.i21.i.i.i.i1236.preheader, %while.body.i.i21.i.i.i.i1236
  %phi.broken8682 = phi i64 [ 0, %while.body.i.i21.i.i.i.i1236.preheader ], [ %1998, %while.body.i.i21.i.i.i.i1236 ]
  call void @llvm.amdgcn.s.sleep(i32 2)
  %1995 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i22.i.i.i.i1237 = getelementptr inbounds nuw i32, ptr %1995, i64 %idxprom.i23.i.i.i.i.i1225
  %1996 = load atomic i32, ptr %arrayidx.i.i.i22.i.i.i.i1237 monotonic, align 4
  %1997 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %1996)
  %cmp.i.not.i.i23.i.i.i.i1238 = icmp eq i32 %1997, %port.sroa.37.0.i.i1229
  %1998 = call i64 @llvm.amdgcn.if.break.i64(i1 %cmp.i.not.i.i23.i.i.i.i1238, i64 %phi.broken8682)
  %1999 = call i1 @llvm.amdgcn.loop.i64(i64 %1998)
  br i1 %1999, label %Flow8344, label %while.body.i.i21.i.i.i.i1236

Flow8344:                                         ; preds = %while.body.i.i21.i.i.i.i1236
  %.lcssa9042 = phi i64 [ %1998, %while.body.i.i21.i.i.i.i1236 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa9042)
  br label %Flow8345, !amdgpu.uniform !42

Flow8346:                                         ; preds = %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1240, %while.cond.i.i.i.i1228
  %2000 = phi i64 [ %add.i.i.i.i1243, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1240 ], [ poison, %while.cond.i.i.i.i1228 ]
  %2001 = phi i32 [ %conv.i.i26.i.i.i.i1241, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1240 ], [ poison, %while.cond.i.i.i.i1228 ]
  %2002 = phi i1 [ false, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1240 ], [ true, %while.cond.i.i.i.i1228 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %1989)
  %2003 = call i64 @llvm.amdgcn.if.break.i64(i1 %2002, i64 %phi.broken8683)
  %2004 = call i1 @llvm.amdgcn.loop.i64(i64 %2003)
  br i1 %2004, label %cond.false.i.i.i1249, label %while.cond.i.i.i.i1228

_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i1239: ; preds = %Flow8345
  fence acquire
  br i1 %cmp.i.i.i.i1231, label %if.then.i.i.i.i.i.i.i1244, label %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1240, !amdgpu.uniform !42

if.then.i.i.i.i.i.i.i1244:                        ; preds = %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i1239
  %2005 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %arrayidx.i20.i35.i.i.i.i1247 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %2005, i64 %idxprom.i19.i.i.i.i.i1219
  %arrayidx.i22.i38.i.i.i.i1248 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %arrayidx.i20.i35.i.i.i.i1247, i64 %sh_prom.i.i.i.i756
  store i8 10, ptr %arrayidx.i22.i38.i.i.i.i1248, align 1
  br label %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1240, !amdgpu.uniform !42

_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1240: ; preds = %if.then.i.i.i.i.i.i.i1244, %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i1239
  %conv.i.i26.i.i.i.i1241 = xor i32 %port.sroa.37.0.i.i1229, 1
  fence release
  %2006 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %arrayidx.i24.i29.i.i.i.i1242 = getelementptr inbounds nuw i32, ptr %2006, i64 %idxprom.i23.i.i.i.i.i1225
  store atomic i32 %conv.i.i26.i.i.i.i1241, ptr %arrayidx.i24.i29.i.i.i.i1242 monotonic, align 4
  %add.i.i.i.i1243 = add i64 %idx.0.i.i.i.i1230, 64
  br label %Flow8346, !amdgpu.uniform !42

cond.false.i.i.i1249:                             ; preds = %Flow8346
  %.lcssa9044 = phi i64 [ %2003, %Flow8346 ]
  %port.sroa.37.0.i.i1229.lcssa = phi i32 [ %port.sroa.37.0.i.i1229, %Flow8346 ]
  %.lcssa9043 = phi i32 [ %1986, %Flow8346 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa9044)
  %2007 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %.lcssa9043)
  %2008 = icmp ne i32 %2007, %port.sroa.37.0.i.i1229.lcssa
  %2009 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %2008)
  %2010 = extractvalue { i1, i64 } %2009, 0
  %2011 = extractvalue { i1, i64 } %2009, 1
  br i1 %2010, label %while.body.i.i.i.i1250.preheader, label %Flow8343

while.body.i.i.i.i1250.preheader:                 ; preds = %cond.false.i.i.i1249
  br label %while.body.i.i.i.i1250, !amdgpu.uniform !42

Flow8343:                                         ; preds = %Flow8342, %cond.false.i.i.i1249
  call void @llvm.amdgcn.end.cf.i64(i64 %2011)
  br label %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIcEERS1_T_.exit, !amdgpu.uniform !42

while.body.i.i.i.i1250:                           ; preds = %while.body.i.i.i.i1250.preheader, %while.body.i.i.i.i1250
  %phi.broken8684 = phi i64 [ 0, %while.body.i.i.i.i1250.preheader ], [ %2015, %while.body.i.i.i.i1250 ]
  call void @llvm.amdgcn.s.sleep(i32 2)
  %2012 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i8.i.i1251 = getelementptr inbounds nuw i32, ptr %2012, i64 %idxprom.i23.i.i.i.i.i1225
  %2013 = load atomic i32, ptr %arrayidx.i.i.i8.i.i1251 monotonic, align 4
  %2014 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %2013)
  %cmp.i.not.i.i.i.i1252 = icmp eq i32 %2014, %port.sroa.37.0.i.i1229.lcssa
  %2015 = call i64 @llvm.amdgcn.if.break.i64(i1 %cmp.i.not.i.i.i.i1252, i64 %phi.broken8684)
  %2016 = call i1 @llvm.amdgcn.loop.i64(i64 %2015)
  br i1 %2016, label %Flow8342, label %while.body.i.i.i.i1250

Flow8342:                                         ; preds = %while.body.i.i.i.i1250
  %.lcssa9045 = phi i64 [ %2015, %while.body.i.i.i.i1250 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa9045)
  br label %Flow8343, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIcEERS1_T_.exit: ; preds = %Flow8343
  fence acquire
  call void @llvm.amdgcn.wave.barrier()
  fence syncscope("agent") release
  %2017 = call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %.lcssa9040, i1 true)
  %iszero.i.i.i.i.i13.i.i1254 = icmp ne i64 %.lcssa9040, 0
  %cmp2.i.i.i.i14.i.i1255 = icmp eq i64 %2017, %sh_prom.i.i.i.i756
  %cmp.i.i.i.i15.i.i1256 = select i1 %iszero.i.i.i.i.i13.i.i1254, i1 %cmp2.i.i.i.i14.i.i1255, i1 false
  %2018 = and i32 %.lcssa9041, 31
  %conv4.i.i.i19.i.i1260 = zext i1 %cmp.i.i.i.i15.i.i1256 to i32
  %shl.i.i.i20.i.i1261 = shl nuw i32 %conv4.i.i.i19.i.i1260, %2018
  %xor.i.i.i21.i.i1262 = xor i32 %shl.i.i.i20.i.i1261, -1
  %2019 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop5628, !amdgpu.uniform !42

2020:                                             ; preds = %ComputeEnd5629
  %div8.i.i.i16.i.i1257 = lshr i32 %.lcssa9041, 5
  %idxprom.i.i.i17.i.i1258 = zext nneg i32 %div8.i.i.i16.i.i1257 to i64
  %2021 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i17.i.i1258
  %sunkaddr7404 = getelementptr inbounds i8, ptr addrspace(1) %2021, i64 40
  %2022 = atomicrmw and ptr addrspace(1) %sunkaddr7404, i32 %.lcssa9046 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %2023, !amdgpu.uniform !42

2023:                                             ; preds = %2020, %ComputeEnd5629
  call void @llvm.amdgcn.end.cf.i64(i64 %7390)
  call void @llvm.amdgcn.wave.barrier()
  br i1 %tobool.not.i1271, label %if.end.i, label %_ZN22__llvm_libc_22_0_0_git5clockEv.exit, !amdgpu.uniform !42

if.end.i:                                         ; preds = %2023
  %2024 = call noundef i64 @llvm.readsteadycounter()
  br i1 %cmp.i, label %if.end3.i, label %Flow8340, !amdgpu.uniform !42

if.then1.i:                                       ; preds = %Flow8340
  %2025 = or i64 %2024, %div.i
  %2026 = and i64 %2025, -4294967296
  %2027 = icmp ne i64 %2026, 0
  br i1 %2027, label %2035, label %Flow8338, !amdgpu.uniform !42

Flow8338:                                         ; preds = %2035, %if.then1.i
  %2028 = phi i64 [ %2036, %2035 ], [ poison, %if.then1.i ]
  %2029 = phi i1 [ false, %2035 ], [ true, %if.then1.i ]
  br i1 %2029, label %2030, label %Flow8339, !amdgpu.uniform !42

2030:                                             ; preds = %Flow8338
  %2031 = trunc i64 %div.i to i32
  %2032 = trunc i64 %2024 to i32
  %2033 = udiv i32 %2032, %2031
  %2034 = zext i32 %2033 to i64
  br label %Flow8339, !amdgpu.uniform !42

2035:                                             ; preds = %if.then1.i
  %2036 = udiv i64 %2024, %div.i
  br label %Flow8338, !amdgpu.uniform !42

if.end3.i:                                        ; preds = %if.end.i
  %2037 = and i64 %1107, -4294967296
  %2038 = icmp ne i64 %2037, 0
  br i1 %2038, label %2045, label %Flow8337, !amdgpu.uniform !42

Flow8337:                                         ; preds = %2045, %if.end3.i
  %2039 = phi i64 [ %2046, %2045 ], [ poison, %if.end3.i ]
  %2040 = phi i1 [ false, %2045 ], [ true, %if.end3.i ]
  br i1 %2040, label %2041, label %2049, !amdgpu.uniform !42

2041:                                             ; preds = %Flow8337
  %2042 = trunc i64 %1107 to i32
  %2043 = udiv i32 1000000, %2042
  %2044 = zext i32 %2043 to i64
  br label %2049, !amdgpu.uniform !42

2045:                                             ; preds = %if.end3.i
  %2046 = sdiv i64 1000000, %1107
  br label %Flow8337, !amdgpu.uniform !42

Flow8340:                                         ; preds = %2049, %if.end.i
  %2047 = phi i64 [ %mul.i, %2049 ], [ poison, %if.end.i ]
  %2048 = phi i1 [ false, %2049 ], [ true, %if.end.i ]
  br i1 %2048, label %if.then1.i, label %Flow8341, !amdgpu.uniform !42

2049:                                             ; preds = %2041, %Flow8337
  %2050 = phi i64 [ %2044, %2041 ], [ %2039, %Flow8337 ]
  %mul.i = mul i64 %2050, %2024
  br label %Flow8340, !amdgpu.uniform !42

Flow8339:                                         ; preds = %2030, %Flow8338
  %2051 = phi i64 [ %2034, %2030 ], [ %2028, %Flow8338 ]
  br label %Flow8341, !amdgpu.uniform !42

Flow8341:                                         ; preds = %Flow8339, %Flow8340
  %2052 = phi i64 [ %2051, %Flow8339 ], [ %2047, %Flow8340 ]
  br label %_ZN22__llvm_libc_22_0_0_git5clockEv.exit, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git5clockEv.exit:         ; preds = %2023, %Flow8341
  %retval.1.i = phi i64 [ 0, %2023 ], [ %2052, %Flow8341 ]
  call void @llvm.lifetime.start.p5(ptr addrspace(5) %Ctx) #34
  store i8 0, ptr addrspace(5) %Ctx, align 1, !tbaa !65
  %vtable43 = load ptr addrspace(1), ptr %T.0207, align 8, !tbaa !158
  %vfn44 = getelementptr inbounds nuw i8, ptr addrspace(1) %vtable43, i64 16
  %2053 = load ptr addrspace(1), ptr addrspace(1) %vfn44, align 8
  call addrspace(1) void %2053(ptr noundef nonnull align 8 dereferenceable(24) %T.0207) #33
  %Ctx.i = getelementptr inbounds nuw i8, ptr %T.0207, i64 16
  store ptr %1103, ptr %Ctx.i, align 8, !tbaa !59
  %vtable45 = load ptr addrspace(1), ptr %T.0207, align 8, !tbaa !158
  %vfn46 = getelementptr inbounds nuw i8, ptr addrspace(1) %vtable45, i64 32
  %2054 = load ptr addrspace(1), ptr addrspace(1) %vfn46, align 8
  call addrspace(1) void %2054(ptr noundef nonnull align 8 dereferenceable(24) %T.0207) #33
  %vtable47 = load ptr addrspace(1), ptr %T.0207, align 8, !tbaa !158
  %vfn48 = getelementptr inbounds nuw i8, ptr addrspace(1) %vtable47, i64 24
  %2055 = load ptr addrspace(1), ptr addrspace(1) %vfn48, align 8
  call addrspace(1) void %2055(ptr noundef nonnull align 8 dereferenceable(24) %T.0207) #33
  br i1 %tobool.not.i1271, label %if.end.i1273, label %_ZN22__llvm_libc_22_0_0_git5clockEv.exit1282, !amdgpu.uniform !42

if.end.i1273:                                     ; preds = %_ZN22__llvm_libc_22_0_0_git5clockEv.exit
  %2056 = call noundef i64 @llvm.readsteadycounter()
  br i1 %cmp.i, label %if.end3.i1275, label %Flow8335, !amdgpu.uniform !42

if.then1.i1279:                                   ; preds = %Flow8335
  %2057 = or i64 %2056, %div.i
  %2058 = and i64 %2057, -4294967296
  %2059 = icmp ne i64 %2058, 0
  br i1 %2059, label %2067, label %Flow8333, !amdgpu.uniform !42

Flow8333:                                         ; preds = %2067, %if.then1.i1279
  %2060 = phi i64 [ %2068, %2067 ], [ poison, %if.then1.i1279 ]
  %2061 = phi i1 [ false, %2067 ], [ true, %if.then1.i1279 ]
  br i1 %2061, label %2062, label %Flow8334, !amdgpu.uniform !42

2062:                                             ; preds = %Flow8333
  %2063 = trunc i64 %div.i to i32
  %2064 = trunc i64 %2056 to i32
  %2065 = udiv i32 %2064, %2063
  %2066 = zext i32 %2065 to i64
  br label %Flow8334, !amdgpu.uniform !42

2067:                                             ; preds = %if.then1.i1279
  %2068 = udiv i64 %2056, %div.i
  br label %Flow8333, !amdgpu.uniform !42

if.end3.i1275:                                    ; preds = %if.end.i1273
  %2069 = and i64 %1107, -4294967296
  %2070 = icmp ne i64 %2069, 0
  br i1 %2070, label %2077, label %Flow8332, !amdgpu.uniform !42

Flow8332:                                         ; preds = %2077, %if.end3.i1275
  %2071 = phi i64 [ %2078, %2077 ], [ poison, %if.end3.i1275 ]
  %2072 = phi i1 [ false, %2077 ], [ true, %if.end3.i1275 ]
  br i1 %2072, label %2073, label %2081, !amdgpu.uniform !42

2073:                                             ; preds = %Flow8332
  %2074 = trunc i64 %1107 to i32
  %2075 = udiv i32 1000000, %2074
  %2076 = zext i32 %2075 to i64
  br label %2081, !amdgpu.uniform !42

2077:                                             ; preds = %if.end3.i1275
  %2078 = sdiv i64 1000000, %1107
  br label %Flow8332, !amdgpu.uniform !42

Flow8335:                                         ; preds = %2081, %if.end.i1273
  %2079 = phi i64 [ %mul.i1277, %2081 ], [ poison, %if.end.i1273 ]
  %2080 = phi i1 [ false, %2081 ], [ true, %if.end.i1273 ]
  br i1 %2080, label %if.then1.i1279, label %Flow8336, !amdgpu.uniform !42

2081:                                             ; preds = %2073, %Flow8332
  %2082 = phi i64 [ %2076, %2073 ], [ %2071, %Flow8332 ]
  %mul.i1277 = mul i64 %2082, %2056
  br label %Flow8335, !amdgpu.uniform !42

Flow8334:                                         ; preds = %2062, %Flow8333
  %2083 = phi i64 [ %2066, %2062 ], [ %2060, %Flow8333 ]
  br label %Flow8336, !amdgpu.uniform !42

Flow8336:                                         ; preds = %Flow8334, %Flow8335
  %2084 = phi i64 [ %2083, %Flow8334 ], [ %2079, %Flow8335 ]
  br label %_ZN22__llvm_libc_22_0_0_git5clockEv.exit1282, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git5clockEv.exit1282:     ; preds = %_ZN22__llvm_libc_22_0_0_git5clockEv.exit, %Flow8336
  %retval.1.i1278 = phi i64 [ 0, %_ZN22__llvm_libc_22_0_0_git5clockEv.exit ], [ %2084, %Flow8336 ]
  %2085 = load i8, ptr addrspace(5) %Ctx, align 1, !tbaa !65, !range !152, !noundef !42
  %loadedv.i = trunc nuw i8 %2085 to i1
  %loadedv.i.inv = xor i1 %loadedv.i, true
  %2086 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %loadedv.i.inv)
  %2087 = extractvalue { i1, i64 } %2086, 0
  %2088 = extractvalue { i1, i64 } %2086, 1
  br i1 %2087, label %for.cond.i.i.i1809.preheader, label %Flow8330

for.cond.i.i.i1809.preheader:                     ; preds = %_ZN22__llvm_libc_22_0_0_git5clockEv.exit1282
  br label %for.cond.i.i.i1809, !amdgpu.uniform !42

for.cond.i.i.i1283.preheader:                     ; preds = %Flow8330
  br label %for.cond.i.i.i1283, !amdgpu.uniform !42

Flow8331:                                         ; preds = %2756, %Flow8330
  %2089 = phi i32 [ %inc, %2756 ], [ %FailCount.0205, %Flow8330 ]
  %2090 = phi i32 [ %1215, %2756 ], [ %2757, %Flow8330 ]
  %2091 = phi i64 [ %1214, %2756 ], [ %2758, %Flow8330 ]
  %2092 = phi i32 [ %1213, %2756 ], [ %2759, %Flow8330 ]
  %2093 = phi i32 [ %1212, %2756 ], [ %2760, %Flow8330 ]
  %2094 = phi i64 [ %1211, %2756 ], [ %2761, %Flow8330 ]
  %2095 = phi i32 [ %1210, %2756 ], [ %2762, %Flow8330 ]
  %2096 = phi i32 [ %1209, %2756 ], [ %2763, %Flow8330 ]
  %2097 = phi i64 [ %1208, %2756 ], [ %2764, %Flow8330 ]
  %2098 = phi i32 [ %1207, %2756 ], [ %2765, %Flow8330 ]
  %2099 = phi i32 [ %1206, %2756 ], [ %2766, %Flow8330 ]
  %2100 = phi i64 [ %1205, %2756 ], [ %2767, %Flow8330 ]
  %2101 = phi i32 [ %1204, %2756 ], [ %2768, %Flow8330 ]
  %2102 = phi i32 [ %1203, %2756 ], [ %2769, %Flow8330 ]
  %2103 = phi i64 [ %1202, %2756 ], [ %2770, %Flow8330 ]
  %2104 = phi i32 [ %1201, %2756 ], [ %2771, %Flow8330 ]
  %2105 = phi i32 [ %1200, %2756 ], [ %2772, %Flow8330 ]
  %2106 = phi i64 [ %1199, %2756 ], [ %2773, %Flow8330 ]
  %2107 = phi i32 [ %1198, %2756 ], [ %2774, %Flow8330 ]
  %2108 = phi i32 [ %1197, %2756 ], [ %2775, %Flow8330 ]
  %2109 = phi i64 [ %1196, %2756 ], [ %2776, %Flow8330 ]
  %2110 = phi i32 [ %1195, %2756 ], [ %2777, %Flow8330 ]
  %2111 = phi i32 [ %1194, %2756 ], [ %2778, %Flow8330 ]
  %2112 = phi i64 [ %1193, %2756 ], [ %2779, %Flow8330 ]
  %2113 = phi i32 [ %1192, %2756 ], [ %2780, %Flow8330 ]
  %2114 = phi i32 [ %1191, %2756 ], [ %2781, %Flow8330 ]
  %2115 = phi i64 [ %1190, %2756 ], [ %2782, %Flow8330 ]
  %2116 = phi i32 [ %1189, %2756 ], [ %2783, %Flow8330 ]
  %2117 = phi i32 [ %1188, %2756 ], [ %2784, %Flow8330 ]
  %2118 = phi i64 [ %1187, %2756 ], [ %2785, %Flow8330 ]
  %2119 = phi i32 [ %1186, %2756 ], [ %2786, %Flow8330 ]
  %2120 = phi i32 [ %1185, %2756 ], [ %2787, %Flow8330 ]
  %2121 = phi i64 [ %1184, %2756 ], [ %2788, %Flow8330 ]
  %2122 = phi i32 [ %1183, %2756 ], [ %2789, %Flow8330 ]
  %2123 = phi i32 [ %1182, %2756 ], [ %2790, %Flow8330 ]
  %2124 = phi i64 [ %1181, %2756 ], [ %2791, %Flow8330 ]
  %2125 = phi i32 [ %1180, %2756 ], [ %2792, %Flow8330 ]
  %2126 = phi i32 [ %.lcssa9256, %2756 ], [ %1179, %Flow8330 ]
  %2127 = phi i64 [ %.lcssa9257, %2756 ], [ %1178, %Flow8330 ]
  %2128 = phi i32 [ %.lcssa9258, %2756 ], [ %1177, %Flow8330 ]
  %2129 = phi i32 [ %.lcssa9241, %2756 ], [ %1176, %Flow8330 ]
  %2130 = phi i64 [ %.lcssa9242, %2756 ], [ %1175, %Flow8330 ]
  %2131 = phi i32 [ %.lcssa9243, %2756 ], [ %1174, %Flow8330 ]
  %2132 = phi i32 [ %.lcssa9225, %2756 ], [ %1173, %Flow8330 ]
  %2133 = phi i64 [ %.lcssa9226, %2756 ], [ %1172, %Flow8330 ]
  %2134 = phi i32 [ %.lcssa9227, %2756 ], [ %1171, %Flow8330 ]
  %2135 = phi i32 [ %.lcssa9213, %2756 ], [ %1170, %Flow8330 ]
  %2136 = phi i64 [ %.lcssa9214, %2756 ], [ %1169, %Flow8330 ]
  %2137 = phi i32 [ %.lcssa9215, %2756 ], [ %1168, %Flow8330 ]
  %2138 = phi i32 [ %.lcssa9198, %2756 ], [ %1167, %Flow8330 ]
  %2139 = phi i64 [ %.lcssa9199, %2756 ], [ %1166, %Flow8330 ]
  %2140 = phi i32 [ %.lcssa9200, %2756 ], [ %1165, %Flow8330 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %2795)
  br label %sw.epilog, !amdgpu.uniform !42

for.cond.i.i.i1283:                               ; preds = %for.cond.i.i.i1283.preheader, %for.cond.i.i.i1283
  %phi.broken8736 = phi i64 [ 0, %for.cond.i.i.i1283.preheader ], [ %2142, %for.cond.i.i.i1283 ]
  %lsr.iv7003 = phi ptr addrspace(4) [ %scevgep7005, %for.cond.i.i.i1283 ], [ %scevgep7002, %for.cond.i.i.i1283.preheader ]
  %scevgep7005 = getelementptr i8, ptr addrspace(4) %lsr.iv7003, i64 1
  %2141 = load i8, ptr addrspace(4) %scevgep7005, align 1, !tbaa !57
  %cmp.i.i.i1285 = icmp eq i8 %2141, 0
  %2142 = call i64 @llvm.amdgcn.if.break.i64(i1 %cmp.i.i.i1285, i64 %phi.broken8736)
  %2143 = call i1 @llvm.amdgcn.loop.i64(i64 %2142)
  br i1 %2143, label %for.cond.i.i1.i1292.preheader, label %for.cond.i.i.i1283

for.cond.i.i1.i1292.preheader:                    ; preds = %for.cond.i.i.i1283
  %scevgep7005.lcssa = phi ptr addrspace(4) [ %scevgep7005, %for.cond.i.i.i1283 ]
  %.lcssa9194 = phi i64 [ %2142, %for.cond.i.i.i1283 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa9194)
  br label %for.cond.i.i1.i1292, !amdgpu.uniform !42

for.cond.i.i1.i1292:                              ; preds = %for.cond.i.i1.i1292.preheader, %Flow8329
  %2144 = phi i32 [ %2181, %Flow8329 ], [ %1165, %for.cond.i.i1.i1292.preheader ]
  %2145 = phi i64 [ %2182, %Flow8329 ], [ %1166, %for.cond.i.i1.i1292.preheader ]
  %2146 = phi i32 [ %2183, %Flow8329 ], [ %1167, %for.cond.i.i1.i1292.preheader ]
  %retval.sroa.6.0.i.i.i1293 = phi i32 [ %2186, %Flow8329 ], [ undef, %for.cond.i.i1.i1292.preheader ]
  %retval.sroa.2.0.i.i.i1295 = phi i64 [ %2185, %Flow8329 ], [ undef, %for.cond.i.i1.i1292.preheader ]
  %retval.sroa.8.0.i.i.i1296 = phi i32 [ %2184, %Flow8329 ], [ undef, %for.cond.i.i1.i1292.preheader ]
  %index.0.i.i.i1297 = phi i32 [ %2188, %Flow8329 ], [ 0, %for.cond.i.i1.i1292.preheader ]
  %2147 = load i32, ptr addrspace(1) @__llvm_rpc_client, align 8, !tbaa !68
  %cmp.not.i.i.i1298 = icmp ult i32 %index.0.i.i.i1297, %2147
  %spec.store.select.i.i.i1299 = select i1 %cmp.not.i.i.i1298, i32 %index.0.i.i.i1297, i32 0
  %2148 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  %2149 = and i32 %spec.store.select.i.i.i1299, 31
  %2150 = lshr i64 %2148, %sh_prom.i.i.i.i756
  %2151 = trunc i64 %2150 to i32
  %conv4.i.i.i.i.i1303 = and i32 %2151, 1
  %shl.i.i.i.i.i1304 = shl nuw i32 %conv4.i.i.i.i.i1303, %2149
  br label %ComputeLoop5632, !amdgpu.uniform !42

2152:                                             ; preds = %ComputeEnd5633
  %sunkaddr7405 = getelementptr inbounds i8, ptr addrspace(1) %7400, i64 40
  %2153 = atomicrmw or ptr addrspace(1) %sunkaddr7405, i32 %.lcssa9195 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %2154, !amdgpu.uniform !42

2154:                                             ; preds = %2152, %ComputeEnd5633
  %2155 = phi i32 [ %2153, %2152 ], [ poison, %ComputeEnd5633 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %7409)
  %2156 = call i32 @llvm.amdgcn.readfirstlane.i32(i32 %2155)
  %2157 = or i32 %2156, %.lcssa9196
  %shl5.i.i.i.i.i1305 = shl nuw i32 1, %2149
  %and.i.i.i.i.i1306 = and i32 %shl5.i.i.i.i.i1305, %2157
  %tobool3.i.i.i.i1307 = icmp ne i32 %and.i.i.i.i.i1306, 0
  %2158 = call i64 @llvm.amdgcn.ballot.i64(i1 %tobool3.i.i.i.i1307)
  %and.i.i.i.i.i.i1308 = and i64 %2158, %2148
  %cmp.i.not.i.i.i1309 = icmp ne i64 %2148, %and.i.i.i.i.i.i1308
  br i1 %cmp.i.not.i.i.i1309, label %if.then.i.i.i.i1310, label %Flow8328, !amdgpu.uniform !42

if.then.i.i.i.i1310:                              ; preds = %2154
  fence syncscope("agent") acquire
  %2159 = load <2 x i64>, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !75
  %2160 = extractelement <2 x i64> %2159, i32 0
  %2161 = inttoptr i64 %2160 to ptr
  %2162 = extractelement <2 x i64> %2159, i32 1
  %2163 = inttoptr i64 %2162 to ptr
  %idxprom.i.i.i.i1311 = zext i32 %spec.store.select.i.i.i1299 to i64
  %arrayidx.i.i.i.i1312 = getelementptr inbounds nuw i32, ptr %2161, i64 %idxprom.i.i.i.i1311, !amdgpu.uniform !42
  %2164 = load atomic i32, ptr %arrayidx.i.i.i.i1312 monotonic, align 4
  %2165 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %2164)
  %arrayidx.i45.i.i.i1313 = getelementptr inbounds nuw i32, ptr %2163, i64 %idxprom.i.i.i.i1311, !amdgpu.uniform !42
  %2166 = load atomic i32, ptr %arrayidx.i45.i.i.i1313 monotonic, align 4
  %2167 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %2166)
  %cmp.i46.not.i.i.i1314 = icmp ne i32 %2165, %2167
  br i1 %cmp.i46.not.i.i.i1314, label %if.then12.i.i.i1315, label %Flow8326, !amdgpu.uniform !42

if.then12.i.i.i1315:                              ; preds = %if.then.i.i.i.i1310
  fence syncscope("agent") release
  %2168 = call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %2148, i1 true)
  %iszero.i.i.i.i.i.i.i1316 = icmp ne i64 %2148, 0
  %cmp2.i.i.i.i.i.i1317 = icmp eq i64 %2168, %sh_prom.i.i.i.i756
  %cmp.i.i.i.i.i.i1318 = select i1 %iszero.i.i.i.i.i.i.i1316, i1 %cmp2.i.i.i.i.i.i1317, i1 false
  %conv4.i.i51.i.i.i1319 = zext i1 %cmp.i.i.i.i.i.i1318 to i32
  %shl.i.i52.i.i.i1320 = shl nuw i32 %conv4.i.i51.i.i.i1319, %2149
  %xor.i.i.i.i.i1321 = xor i32 %shl.i.i52.i.i.i1320, -1
  %2169 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop5637, !amdgpu.uniform !42

2170:                                             ; preds = %ComputeEnd5638
  %sunkaddr7406 = getelementptr inbounds i8, ptr addrspace(1) %7400, i64 40
  %2171 = atomicrmw and ptr addrspace(1) %sunkaddr7406, i32 %.lcssa9197 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %2172, !amdgpu.uniform !42

2172:                                             ; preds = %2170, %ComputeEnd5638
  call void @llvm.amdgcn.end.cf.i64(i64 %7427)
  call void @llvm.amdgcn.wave.barrier()
  br label %Flow8326, !amdgpu.uniform !42

if.end14.i.i.i1381:                               ; preds = %Flow8326
  %2173 = call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %2148, i1 true)
  %iszero.i.i.i.i.i.i1382 = icmp ne i64 %2148, 0
  %cmp2.i.i.i.i.i1383 = icmp eq i64 %2173, %sh_prom.i.i.i.i756
  %cmp.i.i.i.i.i1384 = select i1 %iszero.i.i.i.i.i.i1382, i1 %cmp2.i.i.i.i.i1383, i1 false
  %2174 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.i.i.i.i.i1384)
  %2175 = extractvalue { i1, i64 } %2174, 0
  %2176 = extractvalue { i1, i64 } %2174, 1
  br i1 %2175, label %if.then16.i.i.i1386, label %if.end22.i.i.i1385

if.then16.i.i.i1386:                              ; preds = %if.end14.i.i.i1381
  %2177 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !76
  %opcode.i.i.i1387 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %2177, i64 %idxprom.i.i.i.i1311, i32 1
  store i32 1660944387, ptr %opcode.i.i.i1387, align 8, !tbaa !77
  %arrayidx21.i.i.i1388 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %2177, i64 %idxprom.i.i.i.i1311
  store i64 %2148, ptr %arrayidx21.i.i.i1388, align 8, !tbaa !80
  br label %if.end22.i.i.i1385, !amdgpu.uniform !42

Flow8327:                                         ; preds = %if.end22.i.i.i1385, %Flow8326
  %2178 = phi i32 [ %2165, %if.end22.i.i.i1385 ], [ %retval.sroa.8.0.i.i.i1296, %Flow8326 ]
  %2179 = phi i64 [ %2148, %if.end22.i.i.i1385 ], [ %retval.sroa.2.0.i.i.i1295, %Flow8326 ]
  %2180 = phi i32 [ %spec.store.select.i.i.i1299, %if.end22.i.i.i1385 ], [ %retval.sroa.6.0.i.i.i1293, %Flow8326 ]
  br label %cleanup26.i.i.i1322, !amdgpu.uniform !42

if.end22.i.i.i1385:                               ; preds = %if.then16.i.i.i1386, %if.end14.i.i.i1381
  call void @llvm.amdgcn.end.cf.i64(i64 %2176)
  call void @llvm.amdgcn.wave.barrier()
  br label %Flow8327, !amdgpu.uniform !42

Flow8328:                                         ; preds = %cleanup26.i.i.i1322, %2154
  %2181 = phi i32 [ %2180, %cleanup26.i.i.i1322 ], [ %2144, %2154 ]
  %2182 = phi i64 [ %2179, %cleanup26.i.i.i1322 ], [ %2145, %2154 ]
  %2183 = phi i32 [ %2178, %cleanup26.i.i.i1322 ], [ %2146, %2154 ]
  %2184 = phi i32 [ %2178, %cleanup26.i.i.i1322 ], [ %retval.sroa.8.0.i.i.i1296, %2154 ]
  %2185 = phi i64 [ %2179, %cleanup26.i.i.i1322 ], [ %retval.sroa.2.0.i.i.i1295, %2154 ]
  %2186 = phi i32 [ %2180, %cleanup26.i.i.i1322 ], [ %retval.sroa.6.0.i.i.i1293, %2154 ]
  %2187 = phi i1 [ %cmp.i46.not.i.i.i1314, %cleanup26.i.i.i1322 ], [ true, %2154 ]
  br i1 %2187, label %for.inc.i.i.i1327, label %Flow8329, !amdgpu.uniform !42

cleanup26.i.i.i1322:                              ; preds = %Flow8327
  br label %Flow8328, !amdgpu.uniform !42

for.inc.i.i.i1327:                                ; preds = %Flow8328
  %inc.i.i.i1332 = add i32 %spec.store.select.i.i.i1299, 1
  br label %Flow8329, !amdgpu.uniform !42

Flow8329:                                         ; preds = %for.inc.i.i.i1327, %Flow8328
  %2188 = phi i32 [ %inc.i.i.i1332, %for.inc.i.i.i1327 ], [ poison, %Flow8328 ]
  %2189 = phi i1 [ false, %for.inc.i.i.i1327 ], [ true, %Flow8328 ]
  br i1 %2189, label %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1333, label %for.cond.i.i1.i1292, !amdgpu.uniform !42

_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1333: ; preds = %Flow8329
  %.lcssa9200 = phi i32 [ %2181, %Flow8329 ]
  %.lcssa9199 = phi i64 [ %2182, %Flow8329 ]
  %.lcssa9198 = phi i32 [ %2183, %Flow8329 ]
  %2190 = addrspacecast ptr addrspace(4) %scevgep7005.lcssa to ptr
  %2191 = addrspacecast ptr addrspace(4) %cond3 to ptr
  %2192 = ptrtoint ptr %2191 to i64
  %sub.ptr.lhs.cast.i.i.i1288 = ptrtoint ptr %2190 to i64
  %sub.ptr.sub.i.i.i1290 = sub i64 %sub.ptr.lhs.cast.i.i.i1288, %2192
  fence acquire
  %2193 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %mul.i.i.i.i.i.i1334 = shl i32 %.lcssa9200, 6
  %idxprom.i19.i.i.i.i.i1335 = zext i32 %mul.i.i.i.i.i.i1334 to i64
  %arrayidx.i20.i.i.i.i.i1336 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %2193, i64 %idxprom.i19.i.i.i.i.i1335
  %arrayidx.i22.i.i.i.i.i1337 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %arrayidx.i20.i.i.i.i.i1336, i64 %sh_prom.i.i.i.i756
  store i64 %sub.ptr.sub.i.i.i1290, ptr %arrayidx.i22.i.i.i.i.i1337, align 8, !tbaa !82
  %spec.select.i.i.i.i.i.i.i1338 = call i64 @llvm.umin.i64(i64 %sub.ptr.sub.i.i.i1290, i64 56)
  %2194 = and i64 %spec.select.i.i.i.i.i.i.i1338, 15
  %2195 = sub i64 %spec.select.i.i.i.i.i.i.i1338, %2194
  %2196 = icmp ne i64 %2195, 0
  %2197 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %2196)
  %2198 = extractvalue { i1, i64 } %2197, 0
  %2199 = extractvalue { i1, i64 } %2197, 1
  br i1 %2198, label %loop-memcpy-expansion5283.preheader, label %Flow8325

loop-memcpy-expansion5283.preheader:              ; preds = %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1333
  %scevgep7006 = getelementptr i8, ptr %2193, i64 8
  %scevgep7007 = getelementptr i8, ptr %scevgep7006, i64 %1474
  %2200 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i1335, 6
  %scevgep7008 = getelementptr i8, ptr %scevgep7007, i64 %2200
  br label %loop-memcpy-expansion5283, !amdgpu.uniform !42

Flow8325:                                         ; preds = %Flow8324, %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1333
  call void @llvm.amdgcn.end.cf.i64(i64 %2199)
  br label %loop-memcpy-residual-header5286, !amdgpu.uniform !42

loop-memcpy-expansion5283:                        ; preds = %loop-memcpy-expansion5283.preheader, %loop-memcpy-expansion5283
  %phi.broken8737 = phi i64 [ 0, %loop-memcpy-expansion5283.preheader ], [ %2205, %loop-memcpy-expansion5283 ]
  %loop-index5284 = phi i64 [ %2203, %loop-memcpy-expansion5283 ], [ 0, %loop-memcpy-expansion5283.preheader ]
  %2201 = getelementptr i8, ptr addrspace(4) %cond3, i64 %loop-index5284
  %2202 = load <4 x i32>, ptr addrspace(4) %2201, align 1
  %scevgep7009 = getelementptr i8, ptr %scevgep7008, i64 %loop-index5284
  store <4 x i32> %2202, ptr %scevgep7009, align 1
  %2203 = add i64 %loop-index5284, 16
  %2204 = icmp uge i64 %2203, %2195
  %2205 = call i64 @llvm.amdgcn.if.break.i64(i1 %2204, i64 %phi.broken8737)
  %2206 = call i1 @llvm.amdgcn.loop.i64(i64 %2205)
  br i1 %2206, label %Flow8324, label %loop-memcpy-expansion5283

Flow8323:                                         ; preds = %Flow8321, %Flow8322
  %2207 = phi i64 [ %6572, %Flow8321 ], [ %6567, %Flow8322 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %6570)
  br label %post-loop-memcpy-expansion5282, !amdgpu.uniform !42

loop-memcpy-residual5285:                         ; preds = %loop-memcpy-residual5285.preheader, %loop-memcpy-residual5285
  %phi.broken8738 = phi i64 [ 0, %loop-memcpy-residual5285.preheader ], [ %2211, %loop-memcpy-residual5285 ]
  %residual-loop-index5287 = phi i64 [ %2209, %loop-memcpy-residual5285 ], [ 0, %loop-memcpy-residual5285.preheader ]
  %scevgep7015 = getelementptr i8, ptr addrspace(4) %scevgep7014, i64 %residual-loop-index5287
  %2208 = load i8, ptr addrspace(4) %scevgep7015, align 1
  %scevgep7013 = getelementptr i8, ptr %scevgep7012, i64 %residual-loop-index5287
  store i8 %2208, ptr %scevgep7013, align 1
  %2209 = add i64 %residual-loop-index5287, 1
  %2210 = icmp uge i64 %2209, %2194
  %2211 = call i64 @llvm.amdgcn.if.break.i64(i1 %2210, i64 %phi.broken8738)
  %2212 = call i1 @llvm.amdgcn.loop.i64(i64 %2211)
  br i1 %2212, label %Flow8321, label %loop-memcpy-residual5285

Flow8321:                                         ; preds = %loop-memcpy-residual5285
  %.lcssa9202 = phi i64 [ %2211, %loop-memcpy-residual5285 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa9202)
  br label %Flow8323, !amdgpu.uniform !42

post-loop-memcpy-expansion5282:                   ; preds = %Flow8323
  %2213 = addrspacecast ptr addrspace(4) %scevgep7005.lcssa to ptr
  %2214 = ptrtoint ptr %2213 to i64
  %tobool.not.i.i.i.i.i.i1340 = icmp eq i32 %.lcssa9198, 0
  %conv.i.i.i.i.i.i1341 = zext i1 %tobool.not.i.i.i.i.i.i1340 to i32
  fence release
  %2215 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %idxprom.i23.i.i.i.i.i1342 = zext i32 %.lcssa9200 to i64
  %arrayidx.i24.i.i.i.i.i1343 = getelementptr inbounds nuw i32, ptr %2215, i64 %idxprom.i23.i.i.i.i.i1342
  store atomic i32 %conv.i.i.i.i.i.i1341, ptr %arrayidx.i24.i.i.i.i.i1343 monotonic, align 4
  %2216 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !84
  %arrayidx.i.i5.i.i1344 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %2216, i64 %idxprom.i23.i.i.i.i.i1342
  %2217 = load i64, ptr %arrayidx.i.i5.i.i1344, align 8, !tbaa !80
  %2218 = add i64 %1142, %2214
  br label %while.cond.i.i.i.i1345, !amdgpu.uniform !42

while.cond.i.i.i.i1345:                           ; preds = %post-loop-memcpy-expansion5282, %Flow8320
  %phi.broken8742 = phi i64 [ 0, %post-loop-memcpy-expansion5282 ], [ %2240, %Flow8320 ]
  %lsr.iv7022 = phi i64 [ %2238, %Flow8320 ], [ %2218, %post-loop-memcpy-expansion5282 ]
  %lsr.iv7017 = phi ptr addrspace(4) [ %2237, %Flow8320 ], [ %scevgep7016, %post-loop-memcpy-expansion5282 ]
  %port.sroa.37.0.i.i1346 = phi i32 [ %2236, %Flow8320 ], [ %conv.i.i.i.i.i.i1341, %post-loop-memcpy-expansion5282 ]
  %idx.0.i.i.i.i1347 = phi i64 [ %2235, %Flow8320 ], [ 56, %post-loop-memcpy-expansion5282 ]
  %cmp.i.i.i.i1348 = icmp ugt i64 %sub.ptr.sub.i.i.i1290, %idx.0.i.i.i.i1347
  %2219 = call i64 @llvm.amdgcn.ballot.i64(i1 %cmp.i.i.i.i1348)
  %and.i.i.i.i6.i.i1349 = and i64 %2219, %2217
  %tobool.not.i.i.i.i1350 = icmp ne i64 %and.i.i.i.i6.i.i1349, 0
  %2220 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i22.i.i.i1351 = getelementptr inbounds nuw i32, ptr %2220, i64 %idxprom.i23.i.i.i.i.i1342
  %2221 = load atomic i32, ptr %arrayidx.i22.i.i.i1351 monotonic, align 4
  %2222 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %tobool.not.i.i.i.i1350)
  %2223 = extractvalue { i1, i64 } %2222, 0
  %2224 = extractvalue { i1, i64 } %2222, 1
  br i1 %2223, label %cond.false.i8.i.i.i.i1352, label %Flow8320

cond.false.i8.i.i.i.i1352:                        ; preds = %while.cond.i.i.i.i1345
  %2225 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %2221)
  %2226 = icmp ne i32 %2225, %port.sroa.37.0.i.i1346
  %2227 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %2226)
  %2228 = extractvalue { i1, i64 } %2227, 0
  %2229 = extractvalue { i1, i64 } %2227, 1
  br i1 %2228, label %while.body.i.i21.i.i.i.i1353.preheader, label %Flow8319

while.body.i.i21.i.i.i.i1353.preheader:           ; preds = %cond.false.i8.i.i.i.i1352
  br label %while.body.i.i21.i.i.i.i1353, !amdgpu.uniform !42

Flow8319:                                         ; preds = %Flow8318, %cond.false.i8.i.i.i.i1352
  call void @llvm.amdgcn.end.cf.i64(i64 %2229)
  br label %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i1356, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i1353:                     ; preds = %while.body.i.i21.i.i.i.i1353.preheader, %while.body.i.i21.i.i.i.i1353
  %phi.broken8739 = phi i64 [ 0, %while.body.i.i21.i.i.i.i1353.preheader ], [ %2233, %while.body.i.i21.i.i.i.i1353 ]
  call void @llvm.amdgcn.s.sleep(i32 2)
  %2230 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i22.i.i.i.i1354 = getelementptr inbounds nuw i32, ptr %2230, i64 %idxprom.i23.i.i.i.i.i1342
  %2231 = load atomic i32, ptr %arrayidx.i.i.i22.i.i.i.i1354 monotonic, align 4
  %2232 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %2231)
  %cmp.i.not.i.i23.i.i.i.i1355 = icmp eq i32 %2232, %port.sroa.37.0.i.i1346
  %2233 = call i64 @llvm.amdgcn.if.break.i64(i1 %cmp.i.not.i.i23.i.i.i.i1355, i64 %phi.broken8739)
  %2234 = call i1 @llvm.amdgcn.loop.i64(i64 %2233)
  br i1 %2234, label %Flow8318, label %while.body.i.i21.i.i.i.i1353

Flow8318:                                         ; preds = %while.body.i.i21.i.i.i.i1353
  %.lcssa9203 = phi i64 [ %2233, %while.body.i.i21.i.i.i.i1353 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa9203)
  br label %Flow8319, !amdgpu.uniform !42

Flow8320:                                         ; preds = %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1357, %while.cond.i.i.i.i1345
  %2235 = phi i64 [ %add.i.i.i.i1360, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1357 ], [ poison, %while.cond.i.i.i.i1345 ]
  %2236 = phi i32 [ %conv.i.i26.i.i.i.i1358, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1357 ], [ poison, %while.cond.i.i.i.i1345 ]
  %2237 = phi ptr addrspace(4) [ %scevgep7018, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1357 ], [ poison, %while.cond.i.i.i.i1345 ]
  %2238 = phi i64 [ %lsr.iv.next7023, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1357 ], [ poison, %while.cond.i.i.i.i1345 ]
  %2239 = phi i1 [ false, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1357 ], [ true, %while.cond.i.i.i.i1345 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %2224)
  %2240 = call i64 @llvm.amdgcn.if.break.i64(i1 %2239, i64 %phi.broken8742)
  %2241 = call i1 @llvm.amdgcn.loop.i64(i64 %2240)
  br i1 %2241, label %cond.false.i.i.i1367, label %while.cond.i.i.i.i1345

_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i1356: ; preds = %Flow8319
  fence acquire
  %2242 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.i.i.i.i1348)
  %2243 = extractvalue { i1, i64 } %2242, 0
  %2244 = extractvalue { i1, i64 } %2242, 1
  br i1 %2243, label %if.then.i.i.i.i.i.i.i1361, label %Flow8317

if.then.i.i.i.i.i.i.i1361:                        ; preds = %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i1356
  %sub.i.i.i.i.i.i.i1363 = sub nuw i64 %sub.ptr.sub.i.i.i1290, %idx.0.i.i.i.i1347
  %spec.select.i.i.i36.i.i.i.i1364 = call i64 @llvm.umin.i64(i64 %sub.i.i.i.i.i.i.i1363, i64 64)
  %2245 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %2246 = and i64 %spec.select.i.i.i36.i.i.i.i1364, 15
  %2247 = sub i64 %spec.select.i.i.i36.i.i.i.i1364, %2246
  %2248 = icmp ne i64 %2247, 0
  %2249 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %2248)
  %2250 = extractvalue { i1, i64 } %2249, 0
  %2251 = extractvalue { i1, i64 } %2249, 1
  br i1 %2250, label %loop-memcpy-expansion5289.preheader, label %Flow8316

loop-memcpy-expansion5289.preheader:              ; preds = %if.then.i.i.i.i.i.i.i1361
  %arrayidx.i20.i35.i.i.i.i1365 = getelementptr %"struct.rpc::Buffer", ptr %2245, i64 %idxprom.i19.i.i.i.i.i1335
  %arrayidx.i22.i38.i.i.i.i1366 = getelementptr %"struct.rpc::Buffer", ptr %arrayidx.i20.i35.i.i.i.i1365, i64 %sh_prom.i.i.i.i756
  br label %loop-memcpy-expansion5289, !amdgpu.uniform !42

Flow8316:                                         ; preds = %Flow8315, %if.then.i.i.i.i.i.i.i1361
  call void @llvm.amdgcn.end.cf.i64(i64 %2251)
  br label %loop-memcpy-residual-header5292, !amdgpu.uniform !42

loop-memcpy-expansion5289:                        ; preds = %loop-memcpy-expansion5289.preheader, %loop-memcpy-expansion5289
  %phi.broken8740 = phi i64 [ 0, %loop-memcpy-expansion5289.preheader ], [ %2256, %loop-memcpy-expansion5289 ]
  %loop-index5290 = phi i64 [ %2254, %loop-memcpy-expansion5289 ], [ 0, %loop-memcpy-expansion5289.preheader ]
  %scevgep7019 = getelementptr i8, ptr addrspace(4) %lsr.iv7017, i64 %loop-index5290
  %2252 = load <4 x i32>, ptr addrspace(4) %scevgep7019, align 1
  %2253 = getelementptr i8, ptr %arrayidx.i22.i38.i.i.i.i1366, i64 %loop-index5290
  store <4 x i32> %2252, ptr %2253, align 1
  %2254 = add i64 %loop-index5290, 16
  %2255 = icmp uge i64 %2254, %2247
  %2256 = call i64 @llvm.amdgcn.if.break.i64(i1 %2255, i64 %phi.broken8740)
  %2257 = call i1 @llvm.amdgcn.loop.i64(i64 %2256)
  br i1 %2257, label %Flow8315, label %loop-memcpy-expansion5289

Flow8314:                                         ; preds = %Flow8313, %loop-memcpy-residual-header5292
  call void @llvm.amdgcn.end.cf.i64(i64 %6576)
  br label %Flow8317, !amdgpu.uniform !42

loop-memcpy-residual5291:                         ; preds = %loop-memcpy-residual5291.preheader, %loop-memcpy-residual5291
  %phi.broken8741 = phi i64 [ 0, %loop-memcpy-residual5291.preheader ], [ %2261, %loop-memcpy-residual5291 ]
  %residual-loop-index5293 = phi i64 [ %2259, %loop-memcpy-residual5291 ], [ 0, %loop-memcpy-residual5291.preheader ]
  %scevgep7028 = getelementptr i8, ptr addrspace(4) %scevgep7027, i64 %residual-loop-index5293
  %2258 = load i8, ptr addrspace(4) %scevgep7028, align 1
  %scevgep7026 = getelementptr i8, ptr %scevgep7025, i64 %residual-loop-index5293
  store i8 %2258, ptr %scevgep7026, align 1
  %2259 = add i64 %residual-loop-index5293, 1
  %2260 = icmp uge i64 %2259, %2246
  %2261 = call i64 @llvm.amdgcn.if.break.i64(i1 %2260, i64 %phi.broken8741)
  %2262 = call i1 @llvm.amdgcn.loop.i64(i64 %2261)
  br i1 %2262, label %Flow8313, label %loop-memcpy-residual5291

Flow8313:                                         ; preds = %loop-memcpy-residual5291
  %.lcssa9205 = phi i64 [ %2261, %loop-memcpy-residual5291 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa9205)
  br label %Flow8314, !amdgpu.uniform !42

_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1357: ; preds = %Flow8317
  %conv.i.i26.i.i.i.i1358 = xor i32 %port.sroa.37.0.i.i1346, 1
  fence release
  %2263 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %arrayidx.i24.i29.i.i.i.i1359 = getelementptr inbounds nuw i32, ptr %2263, i64 %idxprom.i23.i.i.i.i.i1342
  store atomic i32 %conv.i.i26.i.i.i.i1358, ptr %arrayidx.i24.i29.i.i.i.i1359 monotonic, align 4
  %add.i.i.i.i1360 = add i64 %idx.0.i.i.i.i1347, 64
  %scevgep7018 = getelementptr i8, ptr addrspace(4) %lsr.iv7017, i64 64
  %lsr.iv.next7023 = add i64 %lsr.iv7022, -64
  br label %Flow8320, !amdgpu.uniform !42

cond.false.i.i.i1367:                             ; preds = %Flow8320
  %.lcssa9207 = phi i64 [ %2240, %Flow8320 ]
  %port.sroa.37.0.i.i1346.lcssa = phi i32 [ %port.sroa.37.0.i.i1346, %Flow8320 ]
  %.lcssa9206 = phi i32 [ %2221, %Flow8320 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa9207)
  %2264 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %.lcssa9206)
  %2265 = icmp ne i32 %2264, %port.sroa.37.0.i.i1346.lcssa
  %2266 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %2265)
  %2267 = extractvalue { i1, i64 } %2266, 0
  %2268 = extractvalue { i1, i64 } %2266, 1
  br i1 %2267, label %while.body.i.i.i.i1368.preheader, label %Flow8312

while.body.i.i.i.i1368.preheader:                 ; preds = %cond.false.i.i.i1367
  br label %while.body.i.i.i.i1368, !amdgpu.uniform !42

Flow8312:                                         ; preds = %Flow8311, %cond.false.i.i.i1367
  call void @llvm.amdgcn.end.cf.i64(i64 %2268)
  br label %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit1389, !amdgpu.uniform !42

while.body.i.i.i.i1368:                           ; preds = %while.body.i.i.i.i1368.preheader, %while.body.i.i.i.i1368
  %phi.broken8743 = phi i64 [ 0, %while.body.i.i.i.i1368.preheader ], [ %2272, %while.body.i.i.i.i1368 ]
  call void @llvm.amdgcn.s.sleep(i32 2)
  %2269 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i8.i.i1369 = getelementptr inbounds nuw i32, ptr %2269, i64 %idxprom.i23.i.i.i.i.i1342
  %2270 = load atomic i32, ptr %arrayidx.i.i.i8.i.i1369 monotonic, align 4
  %2271 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %2270)
  %cmp.i.not.i.i.i.i1370 = icmp eq i32 %2271, %port.sroa.37.0.i.i1346.lcssa
  %2272 = call i64 @llvm.amdgcn.if.break.i64(i1 %cmp.i.not.i.i.i.i1370, i64 %phi.broken8743)
  %2273 = call i1 @llvm.amdgcn.loop.i64(i64 %2272)
  br i1 %2273, label %Flow8311, label %while.body.i.i.i.i1368

Flow8311:                                         ; preds = %while.body.i.i.i.i1368
  %.lcssa9208 = phi i64 [ %2272, %while.body.i.i.i.i1368 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa9208)
  br label %Flow8312, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit1389: ; preds = %Flow8312
  fence acquire
  call void @llvm.amdgcn.wave.barrier()
  fence syncscope("agent") release
  %2274 = call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %.lcssa9199, i1 true)
  %iszero.i.i.i.i.i13.i.i1372 = icmp ne i64 %.lcssa9199, 0
  %cmp2.i.i.i.i14.i.i1373 = icmp eq i64 %2274, %sh_prom.i.i.i.i756
  %cmp.i.i.i.i15.i.i1374 = select i1 %iszero.i.i.i.i.i13.i.i1372, i1 %cmp2.i.i.i.i14.i.i1373, i1 false
  %2275 = and i32 %.lcssa9200, 31
  %conv4.i.i.i19.i.i1378 = zext i1 %cmp.i.i.i.i15.i.i1374 to i32
  %shl.i.i.i20.i.i1379 = shl nuw i32 %conv4.i.i.i19.i.i1378, %2275
  %xor.i.i.i21.i.i1380 = xor i32 %shl.i.i.i20.i.i1379, -1
  %2276 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop5641, !amdgpu.uniform !42

2277:                                             ; preds = %ComputeEnd5642
  %div8.i.i.i16.i.i1375 = lshr i32 %.lcssa9200, 5
  %idxprom.i.i.i17.i.i1376 = zext nneg i32 %div8.i.i.i16.i.i1375 to i64
  %2278 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i17.i.i1376
  %sunkaddr7407 = getelementptr inbounds i8, ptr addrspace(1) %2278, i64 40
  %2279 = atomicrmw and ptr addrspace(1) %sunkaddr7407, i32 %.lcssa9209 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %2280, !amdgpu.uniform !42

2280:                                             ; preds = %2277, %ComputeEnd5642
  call void @llvm.amdgcn.end.cf.i64(i64 %7444)
  call void @llvm.amdgcn.wave.barrier()
  br label %for.cond.i.i1.i1398, !amdgpu.uniform !42

for.cond.i.i1.i1398:                              ; preds = %2280, %Flow8310
  %2281 = phi i32 [ %2318, %Flow8310 ], [ %1168, %2280 ]
  %2282 = phi i64 [ %2319, %Flow8310 ], [ %1169, %2280 ]
  %2283 = phi i32 [ %2320, %Flow8310 ], [ %1170, %2280 ]
  %retval.sroa.6.0.i.i.i1399 = phi i32 [ %2323, %Flow8310 ], [ undef, %2280 ]
  %retval.sroa.2.0.i.i.i1401 = phi i64 [ %2322, %Flow8310 ], [ undef, %2280 ]
  %retval.sroa.8.0.i.i.i1402 = phi i32 [ %2321, %Flow8310 ], [ undef, %2280 ]
  %index.0.i.i.i1403 = phi i32 [ %2325, %Flow8310 ], [ 0, %2280 ]
  %2284 = load i32, ptr addrspace(1) @__llvm_rpc_client, align 8, !tbaa !68
  %cmp.not.i.i.i1404 = icmp ult i32 %index.0.i.i.i1403, %2284
  %spec.store.select.i.i.i1405 = select i1 %cmp.not.i.i.i1404, i32 %index.0.i.i.i1403, i32 0
  %2285 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  %2286 = and i32 %spec.store.select.i.i.i1405, 31
  %2287 = lshr i64 %2285, %sh_prom.i.i.i.i756
  %2288 = trunc i64 %2287 to i32
  %conv4.i.i.i.i.i1409 = and i32 %2288, 1
  %shl.i.i.i.i.i1410 = shl nuw i32 %conv4.i.i.i.i.i1409, %2286
  br label %ComputeLoop5645, !amdgpu.uniform !42

2289:                                             ; preds = %ComputeEnd5646
  %sunkaddr7408 = getelementptr inbounds i8, ptr addrspace(1) %7454, i64 40
  %2290 = atomicrmw or ptr addrspace(1) %sunkaddr7408, i32 %.lcssa9210 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %2291, !amdgpu.uniform !42

2291:                                             ; preds = %2289, %ComputeEnd5646
  %2292 = phi i32 [ %2290, %2289 ], [ poison, %ComputeEnd5646 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %7463)
  %2293 = call i32 @llvm.amdgcn.readfirstlane.i32(i32 %2292)
  %2294 = or i32 %2293, %.lcssa9211
  %shl5.i.i.i.i.i1411 = shl nuw i32 1, %2286
  %and.i.i.i.i.i1412 = and i32 %shl5.i.i.i.i.i1411, %2294
  %tobool3.i.i.i.i1413 = icmp ne i32 %and.i.i.i.i.i1412, 0
  %2295 = call i64 @llvm.amdgcn.ballot.i64(i1 %tobool3.i.i.i.i1413)
  %and.i.i.i.i.i.i1414 = and i64 %2295, %2285
  %cmp.i.not.i.i.i1415 = icmp ne i64 %2285, %and.i.i.i.i.i.i1414
  br i1 %cmp.i.not.i.i.i1415, label %if.then.i.i.i.i1416, label %Flow8309, !amdgpu.uniform !42

if.then.i.i.i.i1416:                              ; preds = %2291
  fence syncscope("agent") acquire
  %2296 = load <2 x i64>, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !75
  %2297 = extractelement <2 x i64> %2296, i32 0
  %2298 = inttoptr i64 %2297 to ptr
  %2299 = extractelement <2 x i64> %2296, i32 1
  %2300 = inttoptr i64 %2299 to ptr
  %idxprom.i.i.i.i1417 = zext i32 %spec.store.select.i.i.i1405 to i64
  %arrayidx.i.i.i.i1418 = getelementptr inbounds nuw i32, ptr %2298, i64 %idxprom.i.i.i.i1417, !amdgpu.uniform !42
  %2301 = load atomic i32, ptr %arrayidx.i.i.i.i1418 monotonic, align 4
  %2302 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %2301)
  %arrayidx.i45.i.i.i1419 = getelementptr inbounds nuw i32, ptr %2300, i64 %idxprom.i.i.i.i1417, !amdgpu.uniform !42
  %2303 = load atomic i32, ptr %arrayidx.i45.i.i.i1419 monotonic, align 4
  %2304 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %2303)
  %cmp.i46.not.i.i.i1420 = icmp ne i32 %2302, %2304
  br i1 %cmp.i46.not.i.i.i1420, label %if.then12.i.i.i1421, label %Flow8307, !amdgpu.uniform !42

if.then12.i.i.i1421:                              ; preds = %if.then.i.i.i.i1416
  fence syncscope("agent") release
  %2305 = call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %2285, i1 true)
  %iszero.i.i.i.i.i.i.i1422 = icmp ne i64 %2285, 0
  %cmp2.i.i.i.i.i.i1423 = icmp eq i64 %2305, %sh_prom.i.i.i.i756
  %cmp.i.i.i.i.i.i1424 = select i1 %iszero.i.i.i.i.i.i.i1422, i1 %cmp2.i.i.i.i.i.i1423, i1 false
  %conv4.i.i51.i.i.i1425 = zext i1 %cmp.i.i.i.i.i.i1424 to i32
  %shl.i.i52.i.i.i1426 = shl nuw i32 %conv4.i.i51.i.i.i1425, %2286
  %xor.i.i.i.i.i1427 = xor i32 %shl.i.i52.i.i.i1426, -1
  %2306 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop5650, !amdgpu.uniform !42

2307:                                             ; preds = %ComputeEnd5651
  %sunkaddr7409 = getelementptr inbounds i8, ptr addrspace(1) %7454, i64 40
  %2308 = atomicrmw and ptr addrspace(1) %sunkaddr7409, i32 %.lcssa9212 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %2309, !amdgpu.uniform !42

2309:                                             ; preds = %2307, %ComputeEnd5651
  call void @llvm.amdgcn.end.cf.i64(i64 %7481)
  call void @llvm.amdgcn.wave.barrier()
  br label %Flow8307, !amdgpu.uniform !42

if.end14.i.i.i1487:                               ; preds = %Flow8307
  %2310 = call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %2285, i1 true)
  %iszero.i.i.i.i.i.i1488 = icmp ne i64 %2285, 0
  %cmp2.i.i.i.i.i1489 = icmp eq i64 %2310, %sh_prom.i.i.i.i756
  %cmp.i.i.i.i.i1490 = select i1 %iszero.i.i.i.i.i.i1488, i1 %cmp2.i.i.i.i.i1489, i1 false
  %2311 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.i.i.i.i.i1490)
  %2312 = extractvalue { i1, i64 } %2311, 0
  %2313 = extractvalue { i1, i64 } %2311, 1
  br i1 %2312, label %if.then16.i.i.i1492, label %if.end22.i.i.i1491

if.then16.i.i.i1492:                              ; preds = %if.end14.i.i.i1487
  %2314 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !76
  %opcode.i.i.i1493 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %2314, i64 %idxprom.i.i.i.i1417, i32 1
  store i32 1660944387, ptr %opcode.i.i.i1493, align 8, !tbaa !77
  %arrayidx21.i.i.i1494 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %2314, i64 %idxprom.i.i.i.i1417
  store i64 %2285, ptr %arrayidx21.i.i.i1494, align 8, !tbaa !80
  br label %if.end22.i.i.i1491, !amdgpu.uniform !42

Flow8308:                                         ; preds = %if.end22.i.i.i1491, %Flow8307
  %2315 = phi i32 [ %2302, %if.end22.i.i.i1491 ], [ %retval.sroa.8.0.i.i.i1402, %Flow8307 ]
  %2316 = phi i64 [ %2285, %if.end22.i.i.i1491 ], [ %retval.sroa.2.0.i.i.i1401, %Flow8307 ]
  %2317 = phi i32 [ %spec.store.select.i.i.i1405, %if.end22.i.i.i1491 ], [ %retval.sroa.6.0.i.i.i1399, %Flow8307 ]
  br label %cleanup26.i.i.i1428, !amdgpu.uniform !42

if.end22.i.i.i1491:                               ; preds = %if.then16.i.i.i1492, %if.end14.i.i.i1487
  call void @llvm.amdgcn.end.cf.i64(i64 %2313)
  call void @llvm.amdgcn.wave.barrier()
  br label %Flow8308, !amdgpu.uniform !42

Flow8309:                                         ; preds = %cleanup26.i.i.i1428, %2291
  %2318 = phi i32 [ %2317, %cleanup26.i.i.i1428 ], [ %2281, %2291 ]
  %2319 = phi i64 [ %2316, %cleanup26.i.i.i1428 ], [ %2282, %2291 ]
  %2320 = phi i32 [ %2315, %cleanup26.i.i.i1428 ], [ %2283, %2291 ]
  %2321 = phi i32 [ %2315, %cleanup26.i.i.i1428 ], [ %retval.sroa.8.0.i.i.i1402, %2291 ]
  %2322 = phi i64 [ %2316, %cleanup26.i.i.i1428 ], [ %retval.sroa.2.0.i.i.i1401, %2291 ]
  %2323 = phi i32 [ %2317, %cleanup26.i.i.i1428 ], [ %retval.sroa.6.0.i.i.i1399, %2291 ]
  %2324 = phi i1 [ %cmp.i46.not.i.i.i1420, %cleanup26.i.i.i1428 ], [ true, %2291 ]
  br i1 %2324, label %for.inc.i.i.i1433, label %Flow8310, !amdgpu.uniform !42

cleanup26.i.i.i1428:                              ; preds = %Flow8308
  br label %Flow8309, !amdgpu.uniform !42

for.inc.i.i.i1433:                                ; preds = %Flow8309
  %inc.i.i.i1438 = add i32 %spec.store.select.i.i.i1405, 1
  br label %Flow8310, !amdgpu.uniform !42

Flow8310:                                         ; preds = %for.inc.i.i.i1433, %Flow8309
  %2325 = phi i32 [ %inc.i.i.i1438, %for.inc.i.i.i1433 ], [ poison, %Flow8309 ]
  %2326 = phi i1 [ false, %for.inc.i.i.i1433 ], [ true, %Flow8309 ]
  br i1 %2326, label %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1439, label %for.cond.i.i1.i1398, !amdgpu.uniform !42

_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1439: ; preds = %Flow8310
  %.lcssa9215 = phi i32 [ %2318, %Flow8310 ]
  %.lcssa9214 = phi i64 [ %2319, %Flow8310 ]
  %.lcssa9213 = phi i32 [ %2320, %Flow8310 ]
  fence acquire
  %2327 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %mul.i.i.i.i.i.i1440 = shl i32 %.lcssa9215, 6
  %idxprom.i19.i.i.i.i.i1441 = zext i32 %mul.i.i.i.i.i.i1440 to i64
  %arrayidx.i20.i.i.i.i.i1442 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %2327, i64 %idxprom.i19.i.i.i.i.i1441
  %arrayidx.i22.i.i.i.i.i1443 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %arrayidx.i20.i.i.i.i.i1442, i64 %sh_prom.i.i.i.i756
  store i64 sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.18.35 to ptr), i64 13) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.18.35 to ptr) to i64)), ptr %arrayidx.i22.i.i.i.i.i1443, align 8, !tbaa !82
  br i1 %1138, label %loop-memcpy-expansion5295.preheader, label %Flow8306, !amdgpu.uniform !42

loop-memcpy-expansion5295.preheader:              ; preds = %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1439
  %scevgep7029 = getelementptr i8, ptr %2327, i64 8
  %scevgep7030 = getelementptr i8, ptr %scevgep7029, i64 %1474
  %2328 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i1441, 6
  %scevgep7031 = getelementptr i8, ptr %scevgep7030, i64 %2328
  br label %loop-memcpy-expansion5295, !amdgpu.uniform !42

Flow8306:                                         ; preds = %Flow8305, %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1439
  br label %loop-memcpy-residual-header5298, !amdgpu.uniform !42

loop-memcpy-expansion5295:                        ; preds = %loop-memcpy-expansion5295.preheader, %loop-memcpy-expansion5295
  %loop-index5296 = phi i64 [ %2331, %loop-memcpy-expansion5295 ], [ 0, %loop-memcpy-expansion5295.preheader ]
  %2329 = getelementptr i8, ptr addrspace(4) @.str.18.35, i64 %loop-index5296, !amdgpu.uniform !42
  %2330 = load <4 x i32>, ptr addrspace(4) %2329, align 1
  %scevgep7032 = getelementptr i8, ptr %scevgep7031, i64 %loop-index5296
  store <4 x i32> %2330, ptr %scevgep7032, align 1
  %2331 = add i64 %loop-index5296, 16
  %2332 = icmp uge i64 %2331, %1137
  br i1 %2332, label %Flow8305, label %loop-memcpy-expansion5295, !amdgpu.uniform !42

Flow8304:                                         ; preds = %Flow8302, %Flow8303
  %2333 = phi i64 [ %6581, %Flow8302 ], [ %6579, %Flow8303 ]
  br label %post-loop-memcpy-expansion5294, !amdgpu.uniform !42

loop-memcpy-residual5297:                         ; preds = %loop-memcpy-residual5297.preheader, %loop-memcpy-residual5297
  %residual-loop-index5299 = phi i64 [ %2335, %loop-memcpy-residual5297 ], [ 0, %loop-memcpy-residual5297.preheader ]
  %scevgep7037 = getelementptr i8, ptr addrspace(4) @.str.18.35, i64 %residual-loop-index5299, !amdgpu.uniform !42
  %2334 = load i8, ptr addrspace(4) %scevgep7037, align 1
  %scevgep7036 = getelementptr i8, ptr %scevgep7035, i64 %residual-loop-index5299
  store i8 %2334, ptr %scevgep7036, align 1
  %2335 = add i64 %residual-loop-index5299, 1
  %2336 = icmp uge i64 %2335, %1136
  br i1 %2336, label %Flow8302, label %loop-memcpy-residual5297, !amdgpu.uniform !42

Flow8302:                                         ; preds = %loop-memcpy-residual5297
  br label %Flow8304, !amdgpu.uniform !42

post-loop-memcpy-expansion5294:                   ; preds = %Flow8304
  %tobool.not.i.i.i.i.i.i1446 = icmp eq i32 %.lcssa9213, 0
  %conv.i.i.i.i.i.i1447 = zext i1 %tobool.not.i.i.i.i.i.i1446 to i32
  fence release
  %2337 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %idxprom.i23.i.i.i.i.i1448 = zext i32 %.lcssa9215 to i64
  %arrayidx.i24.i.i.i.i.i1449 = getelementptr inbounds nuw i32, ptr %2337, i64 %idxprom.i23.i.i.i.i.i1448
  store atomic i32 %conv.i.i.i.i.i.i1447, ptr %arrayidx.i24.i.i.i.i.i1449 monotonic, align 4
  %2338 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !84
  %arrayidx.i.i5.i.i1450 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %2338, i64 %idxprom.i23.i.i.i.i.i1448
  %2339 = load i64, ptr %arrayidx.i.i5.i.i1450, align 8, !tbaa !80
  br label %while.cond.i.i.i.i1451, !amdgpu.uniform !42

while.cond.i.i.i.i1451:                           ; preds = %post-loop-memcpy-expansion5294, %Flow8301
  %phi.broken8745 = phi i64 [ 0, %post-loop-memcpy-expansion5294 ], [ %2361, %Flow8301 ]
  %lsr.iv7043 = phi i64 [ %2359, %Flow8301 ], [ -43, %post-loop-memcpy-expansion5294 ]
  %lsr.iv7038 = phi ptr addrspace(4) [ %2358, %Flow8301 ], [ getelementptr (i8, ptr addrspace(4) @.str.18.35, i64 56), %post-loop-memcpy-expansion5294 ]
  %port.sroa.37.0.i.i1452 = phi i32 [ %2357, %Flow8301 ], [ %conv.i.i.i.i.i.i1447, %post-loop-memcpy-expansion5294 ]
  %idx.0.i.i.i.i1453 = phi i64 [ %2356, %Flow8301 ], [ 56, %post-loop-memcpy-expansion5294 ]
  %cmp.i.i.i.i1454 = icmp ult i64 %idx.0.i.i.i.i1453, sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.18.35 to ptr), i64 13) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.18.35 to ptr) to i64))
  %2340 = call i64 @llvm.amdgcn.ballot.i64(i1 %cmp.i.i.i.i1454)
  %and.i.i.i.i6.i.i1455 = and i64 %2340, %2339
  %tobool.not.i.i.i.i1456 = icmp ne i64 %and.i.i.i.i6.i.i1455, 0
  %2341 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i22.i.i.i1457 = getelementptr inbounds nuw i32, ptr %2341, i64 %idxprom.i23.i.i.i.i.i1448
  %2342 = load atomic i32, ptr %arrayidx.i22.i.i.i1457 monotonic, align 4
  %2343 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %tobool.not.i.i.i.i1456)
  %2344 = extractvalue { i1, i64 } %2343, 0
  %2345 = extractvalue { i1, i64 } %2343, 1
  br i1 %2344, label %cond.false.i8.i.i.i.i1458, label %Flow8301

cond.false.i8.i.i.i.i1458:                        ; preds = %while.cond.i.i.i.i1451
  %2346 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %2342)
  %2347 = icmp ne i32 %2346, %port.sroa.37.0.i.i1452
  %2348 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %2347)
  %2349 = extractvalue { i1, i64 } %2348, 0
  %2350 = extractvalue { i1, i64 } %2348, 1
  br i1 %2349, label %while.body.i.i21.i.i.i.i1459.preheader, label %Flow8300

while.body.i.i21.i.i.i.i1459.preheader:           ; preds = %cond.false.i8.i.i.i.i1458
  br label %while.body.i.i21.i.i.i.i1459, !amdgpu.uniform !42

Flow8300:                                         ; preds = %Flow8299, %cond.false.i8.i.i.i.i1458
  call void @llvm.amdgcn.end.cf.i64(i64 %2350)
  br label %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i1462, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i1459:                     ; preds = %while.body.i.i21.i.i.i.i1459.preheader, %while.body.i.i21.i.i.i.i1459
  %phi.broken8744 = phi i64 [ 0, %while.body.i.i21.i.i.i.i1459.preheader ], [ %2354, %while.body.i.i21.i.i.i.i1459 ]
  call void @llvm.amdgcn.s.sleep(i32 2)
  %2351 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i22.i.i.i.i1460 = getelementptr inbounds nuw i32, ptr %2351, i64 %idxprom.i23.i.i.i.i.i1448
  %2352 = load atomic i32, ptr %arrayidx.i.i.i22.i.i.i.i1460 monotonic, align 4
  %2353 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %2352)
  %cmp.i.not.i.i23.i.i.i.i1461 = icmp eq i32 %2353, %port.sroa.37.0.i.i1452
  %2354 = call i64 @llvm.amdgcn.if.break.i64(i1 %cmp.i.not.i.i23.i.i.i.i1461, i64 %phi.broken8744)
  %2355 = call i1 @llvm.amdgcn.loop.i64(i64 %2354)
  br i1 %2355, label %Flow8299, label %while.body.i.i21.i.i.i.i1459

Flow8299:                                         ; preds = %while.body.i.i21.i.i.i.i1459
  %.lcssa9216 = phi i64 [ %2354, %while.body.i.i21.i.i.i.i1459 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa9216)
  br label %Flow8300, !amdgpu.uniform !42

Flow8301:                                         ; preds = %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1463, %while.cond.i.i.i.i1451
  %2356 = phi i64 [ %add.i.i.i.i1466, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1463 ], [ poison, %while.cond.i.i.i.i1451 ]
  %2357 = phi i32 [ %conv.i.i26.i.i.i.i1464, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1463 ], [ poison, %while.cond.i.i.i.i1451 ]
  %2358 = phi ptr addrspace(4) [ %scevgep7039, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1463 ], [ poison, %while.cond.i.i.i.i1451 ]
  %2359 = phi i64 [ %lsr.iv.next7044, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1463 ], [ poison, %while.cond.i.i.i.i1451 ]
  %2360 = phi i1 [ false, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1463 ], [ true, %while.cond.i.i.i.i1451 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %2345)
  %2361 = call i64 @llvm.amdgcn.if.break.i64(i1 %2360, i64 %phi.broken8745)
  %2362 = call i1 @llvm.amdgcn.loop.i64(i64 %2361)
  br i1 %2362, label %cond.false.i.i.i1473, label %while.cond.i.i.i.i1451

_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i1462: ; preds = %Flow8300
  fence acquire
  br i1 %cmp.i.i.i.i1454, label %if.then.i.i.i.i.i.i.i1467, label %Flow8298, !amdgpu.uniform !42

if.then.i.i.i.i.i.i.i1467:                        ; preds = %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i1462
  %sub.i.i.i.i.i.i.i1469 = sub nuw i64 sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.18.35 to ptr), i64 13) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.18.35 to ptr) to i64)), %idx.0.i.i.i.i1453
  %spec.select.i.i.i36.i.i.i.i1470 = call i64 @llvm.umin.i64(i64 %sub.i.i.i.i.i.i.i1469, i64 64)
  %2363 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %2364 = and i64 %spec.select.i.i.i36.i.i.i.i1470, 15
  %2365 = sub i64 %spec.select.i.i.i36.i.i.i.i1470, %2364
  %2366 = icmp ne i64 %2365, 0
  br i1 %2366, label %loop-memcpy-expansion5301.preheader, label %Flow8297, !amdgpu.uniform !42

loop-memcpy-expansion5301.preheader:              ; preds = %if.then.i.i.i.i.i.i.i1467
  %arrayidx.i20.i35.i.i.i.i1471 = getelementptr %"struct.rpc::Buffer", ptr %2363, i64 %idxprom.i19.i.i.i.i.i1441
  %arrayidx.i22.i38.i.i.i.i1472 = getelementptr %"struct.rpc::Buffer", ptr %arrayidx.i20.i35.i.i.i.i1471, i64 %sh_prom.i.i.i.i756
  br label %loop-memcpy-expansion5301, !amdgpu.uniform !42

Flow8297:                                         ; preds = %Flow8296, %if.then.i.i.i.i.i.i.i1467
  br label %loop-memcpy-residual-header5304, !amdgpu.uniform !42

loop-memcpy-expansion5301:                        ; preds = %loop-memcpy-expansion5301.preheader, %loop-memcpy-expansion5301
  %loop-index5302 = phi i64 [ %2369, %loop-memcpy-expansion5301 ], [ 0, %loop-memcpy-expansion5301.preheader ]
  %scevgep7040 = getelementptr i8, ptr addrspace(4) %lsr.iv7038, i64 %loop-index5302, !amdgpu.uniform !42
  %2367 = load <4 x i32>, ptr addrspace(4) %scevgep7040, align 1
  %2368 = getelementptr i8, ptr %arrayidx.i22.i38.i.i.i.i1472, i64 %loop-index5302
  store <4 x i32> %2367, ptr %2368, align 1
  %2369 = add i64 %loop-index5302, 16
  %2370 = icmp uge i64 %2369, %2365
  br i1 %2370, label %Flow8296, label %loop-memcpy-expansion5301, !amdgpu.uniform !42

Flow8295:                                         ; preds = %Flow8294, %loop-memcpy-residual-header5304
  br label %Flow8298, !amdgpu.uniform !42

loop-memcpy-residual5303:                         ; preds = %loop-memcpy-residual5303.preheader, %loop-memcpy-residual5303
  %residual-loop-index5305 = phi i64 [ %2372, %loop-memcpy-residual5303 ], [ 0, %loop-memcpy-residual5303.preheader ]
  %scevgep7049 = getelementptr i8, ptr addrspace(4) %scevgep7048, i64 %residual-loop-index5305, !amdgpu.uniform !42
  %2371 = load i8, ptr addrspace(4) %scevgep7049, align 1
  %scevgep7047 = getelementptr i8, ptr %scevgep7046, i64 %residual-loop-index5305
  store i8 %2371, ptr %scevgep7047, align 1
  %2372 = add i64 %residual-loop-index5305, 1
  %2373 = icmp uge i64 %2372, %2364
  br i1 %2373, label %Flow8294, label %loop-memcpy-residual5303, !amdgpu.uniform !42

Flow8294:                                         ; preds = %loop-memcpy-residual5303
  br label %Flow8295, !amdgpu.uniform !42

_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1463: ; preds = %Flow8298
  %conv.i.i26.i.i.i.i1464 = xor i32 %port.sroa.37.0.i.i1452, 1
  fence release
  %2374 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %arrayidx.i24.i29.i.i.i.i1465 = getelementptr inbounds nuw i32, ptr %2374, i64 %idxprom.i23.i.i.i.i.i1448
  store atomic i32 %conv.i.i26.i.i.i.i1464, ptr %arrayidx.i24.i29.i.i.i.i1465 monotonic, align 4
  %add.i.i.i.i1466 = add i64 %idx.0.i.i.i.i1453, 64
  %scevgep7039 = getelementptr i8, ptr addrspace(4) %lsr.iv7038, i64 64
  %lsr.iv.next7044 = add i64 %lsr.iv7043, -64
  br label %Flow8301, !amdgpu.uniform !42

cond.false.i.i.i1473:                             ; preds = %Flow8301
  %.lcssa9218 = phi i64 [ %2361, %Flow8301 ]
  %port.sroa.37.0.i.i1452.lcssa = phi i32 [ %port.sroa.37.0.i.i1452, %Flow8301 ]
  %.lcssa9217 = phi i32 [ %2342, %Flow8301 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa9218)
  %2375 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %.lcssa9217)
  %2376 = icmp ne i32 %2375, %port.sroa.37.0.i.i1452.lcssa
  %2377 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %2376)
  %2378 = extractvalue { i1, i64 } %2377, 0
  %2379 = extractvalue { i1, i64 } %2377, 1
  br i1 %2378, label %while.body.i.i.i.i1474.preheader, label %Flow8293

while.body.i.i.i.i1474.preheader:                 ; preds = %cond.false.i.i.i1473
  br label %while.body.i.i.i.i1474, !amdgpu.uniform !42

Flow8293:                                         ; preds = %Flow8292, %cond.false.i.i.i1473
  call void @llvm.amdgcn.end.cf.i64(i64 %2379)
  br label %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit1495, !amdgpu.uniform !42

while.body.i.i.i.i1474:                           ; preds = %while.body.i.i.i.i1474.preheader, %while.body.i.i.i.i1474
  %phi.broken8746 = phi i64 [ 0, %while.body.i.i.i.i1474.preheader ], [ %2383, %while.body.i.i.i.i1474 ]
  call void @llvm.amdgcn.s.sleep(i32 2)
  %2380 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i8.i.i1475 = getelementptr inbounds nuw i32, ptr %2380, i64 %idxprom.i23.i.i.i.i.i1448
  %2381 = load atomic i32, ptr %arrayidx.i.i.i8.i.i1475 monotonic, align 4
  %2382 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %2381)
  %cmp.i.not.i.i.i.i1476 = icmp eq i32 %2382, %port.sroa.37.0.i.i1452.lcssa
  %2383 = call i64 @llvm.amdgcn.if.break.i64(i1 %cmp.i.not.i.i.i.i1476, i64 %phi.broken8746)
  %2384 = call i1 @llvm.amdgcn.loop.i64(i64 %2383)
  br i1 %2384, label %Flow8292, label %while.body.i.i.i.i1474

Flow8292:                                         ; preds = %while.body.i.i.i.i1474
  %.lcssa9219 = phi i64 [ %2383, %while.body.i.i.i.i1474 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa9219)
  br label %Flow8293, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit1495: ; preds = %Flow8293
  fence acquire
  call void @llvm.amdgcn.wave.barrier()
  fence syncscope("agent") release
  %2385 = call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %.lcssa9214, i1 true)
  %iszero.i.i.i.i.i13.i.i1478 = icmp ne i64 %.lcssa9214, 0
  %cmp2.i.i.i.i14.i.i1479 = icmp eq i64 %2385, %sh_prom.i.i.i.i756
  %cmp.i.i.i.i15.i.i1480 = select i1 %iszero.i.i.i.i.i13.i.i1478, i1 %cmp2.i.i.i.i14.i.i1479, i1 false
  %2386 = and i32 %.lcssa9215, 31
  %conv4.i.i.i19.i.i1484 = zext i1 %cmp.i.i.i.i15.i.i1480 to i32
  %shl.i.i.i20.i.i1485 = shl nuw i32 %conv4.i.i.i19.i.i1484, %2386
  %xor.i.i.i21.i.i1486 = xor i32 %shl.i.i.i20.i.i1485, -1
  %2387 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop5654, !amdgpu.uniform !42

2388:                                             ; preds = %ComputeEnd5655
  %div8.i.i.i16.i.i1481 = lshr i32 %.lcssa9215, 5
  %idxprom.i.i.i17.i.i1482 = zext nneg i32 %div8.i.i.i16.i.i1481 to i64
  %2389 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i17.i.i1482
  %sunkaddr7410 = getelementptr inbounds i8, ptr addrspace(1) %2389, i64 40
  %2390 = atomicrmw and ptr addrspace(1) %sunkaddr7410, i32 %.lcssa9220 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %2391, !amdgpu.uniform !42

2391:                                             ; preds = %2388, %ComputeEnd5655
  call void @llvm.amdgcn.end.cf.i64(i64 %7498)
  call void @llvm.amdgcn.wave.barrier()
  br label %for.cond.i.i.i1496, !amdgpu.uniform !42

for.cond.i.i.i1496:                               ; preds = %2391, %for.cond.i.i.i1496
  %phi.broken8747 = phi i64 [ 0, %2391 ], [ %2393, %for.cond.i.i.i1496 ]
  %lsr.iv7050 = phi ptr addrspace(4) [ %scevgep7052, %for.cond.i.i.i1496 ], [ %scevgep6634, %2391 ]
  %scevgep7052 = getelementptr i8, ptr addrspace(4) %lsr.iv7050, i64 1
  %2392 = load i8, ptr addrspace(4) %scevgep7052, align 1, !tbaa !57
  %cmp.i.i.i1498 = icmp eq i8 %2392, 0
  %2393 = call i64 @llvm.amdgcn.if.break.i64(i1 %cmp.i.i.i1498, i64 %phi.broken8747)
  %2394 = call i1 @llvm.amdgcn.loop.i64(i64 %2393)
  br i1 %2394, label %for.cond.i.i1.i1505.preheader, label %for.cond.i.i.i1496

for.cond.i.i1.i1505.preheader:                    ; preds = %for.cond.i.i.i1496
  %scevgep7052.lcssa = phi ptr addrspace(4) [ %scevgep7052, %for.cond.i.i.i1496 ]
  %.lcssa9221 = phi i64 [ %2393, %for.cond.i.i.i1496 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa9221)
  br label %for.cond.i.i1.i1505, !amdgpu.uniform !42

for.cond.i.i1.i1505:                              ; preds = %for.cond.i.i1.i1505.preheader, %Flow8291
  %2395 = phi i32 [ %2432, %Flow8291 ], [ %1171, %for.cond.i.i1.i1505.preheader ]
  %2396 = phi i64 [ %2433, %Flow8291 ], [ %1172, %for.cond.i.i1.i1505.preheader ]
  %2397 = phi i32 [ %2434, %Flow8291 ], [ %1173, %for.cond.i.i1.i1505.preheader ]
  %retval.sroa.6.0.i.i.i1506 = phi i32 [ %2437, %Flow8291 ], [ undef, %for.cond.i.i1.i1505.preheader ]
  %retval.sroa.2.0.i.i.i1508 = phi i64 [ %2436, %Flow8291 ], [ undef, %for.cond.i.i1.i1505.preheader ]
  %retval.sroa.8.0.i.i.i1509 = phi i32 [ %2435, %Flow8291 ], [ undef, %for.cond.i.i1.i1505.preheader ]
  %index.0.i.i.i1510 = phi i32 [ %2439, %Flow8291 ], [ 0, %for.cond.i.i1.i1505.preheader ]
  %2398 = load i32, ptr addrspace(1) @__llvm_rpc_client, align 8, !tbaa !68
  %cmp.not.i.i.i1511 = icmp ult i32 %index.0.i.i.i1510, %2398
  %spec.store.select.i.i.i1512 = select i1 %cmp.not.i.i.i1511, i32 %index.0.i.i.i1510, i32 0
  %2399 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  %2400 = and i32 %spec.store.select.i.i.i1512, 31
  %2401 = lshr i64 %2399, %sh_prom.i.i.i.i756
  %2402 = trunc i64 %2401 to i32
  %conv4.i.i.i.i.i1516 = and i32 %2402, 1
  %shl.i.i.i.i.i1517 = shl nuw i32 %conv4.i.i.i.i.i1516, %2400
  br label %ComputeLoop5658, !amdgpu.uniform !42

2403:                                             ; preds = %ComputeEnd5659
  %sunkaddr7411 = getelementptr inbounds i8, ptr addrspace(1) %7508, i64 40
  %2404 = atomicrmw or ptr addrspace(1) %sunkaddr7411, i32 %.lcssa9222 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %2405, !amdgpu.uniform !42

2405:                                             ; preds = %2403, %ComputeEnd5659
  %2406 = phi i32 [ %2404, %2403 ], [ poison, %ComputeEnd5659 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %7517)
  %2407 = call i32 @llvm.amdgcn.readfirstlane.i32(i32 %2406)
  %2408 = or i32 %2407, %.lcssa9223
  %shl5.i.i.i.i.i1518 = shl nuw i32 1, %2400
  %and.i.i.i.i.i1519 = and i32 %shl5.i.i.i.i.i1518, %2408
  %tobool3.i.i.i.i1520 = icmp ne i32 %and.i.i.i.i.i1519, 0
  %2409 = call i64 @llvm.amdgcn.ballot.i64(i1 %tobool3.i.i.i.i1520)
  %and.i.i.i.i.i.i1521 = and i64 %2409, %2399
  %cmp.i.not.i.i.i1522 = icmp ne i64 %2399, %and.i.i.i.i.i.i1521
  br i1 %cmp.i.not.i.i.i1522, label %if.then.i.i.i.i1523, label %Flow8290, !amdgpu.uniform !42

if.then.i.i.i.i1523:                              ; preds = %2405
  fence syncscope("agent") acquire
  %2410 = load <2 x i64>, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !75
  %2411 = extractelement <2 x i64> %2410, i32 0
  %2412 = inttoptr i64 %2411 to ptr
  %2413 = extractelement <2 x i64> %2410, i32 1
  %2414 = inttoptr i64 %2413 to ptr
  %idxprom.i.i.i.i1524 = zext i32 %spec.store.select.i.i.i1512 to i64
  %arrayidx.i.i.i.i1525 = getelementptr inbounds nuw i32, ptr %2412, i64 %idxprom.i.i.i.i1524, !amdgpu.uniform !42
  %2415 = load atomic i32, ptr %arrayidx.i.i.i.i1525 monotonic, align 4
  %2416 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %2415)
  %arrayidx.i45.i.i.i1526 = getelementptr inbounds nuw i32, ptr %2414, i64 %idxprom.i.i.i.i1524, !amdgpu.uniform !42
  %2417 = load atomic i32, ptr %arrayidx.i45.i.i.i1526 monotonic, align 4
  %2418 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %2417)
  %cmp.i46.not.i.i.i1527 = icmp ne i32 %2416, %2418
  br i1 %cmp.i46.not.i.i.i1527, label %if.then12.i.i.i1528, label %Flow8288, !amdgpu.uniform !42

if.then12.i.i.i1528:                              ; preds = %if.then.i.i.i.i1523
  fence syncscope("agent") release
  %2419 = call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %2399, i1 true)
  %iszero.i.i.i.i.i.i.i1529 = icmp ne i64 %2399, 0
  %cmp2.i.i.i.i.i.i1530 = icmp eq i64 %2419, %sh_prom.i.i.i.i756
  %cmp.i.i.i.i.i.i1531 = select i1 %iszero.i.i.i.i.i.i.i1529, i1 %cmp2.i.i.i.i.i.i1530, i1 false
  %conv4.i.i51.i.i.i1532 = zext i1 %cmp.i.i.i.i.i.i1531 to i32
  %shl.i.i52.i.i.i1533 = shl nuw i32 %conv4.i.i51.i.i.i1532, %2400
  %xor.i.i.i.i.i1534 = xor i32 %shl.i.i52.i.i.i1533, -1
  %2420 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop5663, !amdgpu.uniform !42

2421:                                             ; preds = %ComputeEnd5664
  %sunkaddr7412 = getelementptr inbounds i8, ptr addrspace(1) %7508, i64 40
  %2422 = atomicrmw and ptr addrspace(1) %sunkaddr7412, i32 %.lcssa9224 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %2423, !amdgpu.uniform !42

2423:                                             ; preds = %2421, %ComputeEnd5664
  call void @llvm.amdgcn.end.cf.i64(i64 %7535)
  call void @llvm.amdgcn.wave.barrier()
  br label %Flow8288, !amdgpu.uniform !42

if.end14.i.i.i1594:                               ; preds = %Flow8288
  %2424 = call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %2399, i1 true)
  %iszero.i.i.i.i.i.i1595 = icmp ne i64 %2399, 0
  %cmp2.i.i.i.i.i1596 = icmp eq i64 %2424, %sh_prom.i.i.i.i756
  %cmp.i.i.i.i.i1597 = select i1 %iszero.i.i.i.i.i.i1595, i1 %cmp2.i.i.i.i.i1596, i1 false
  %2425 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.i.i.i.i.i1597)
  %2426 = extractvalue { i1, i64 } %2425, 0
  %2427 = extractvalue { i1, i64 } %2425, 1
  br i1 %2426, label %if.then16.i.i.i1599, label %if.end22.i.i.i1598

if.then16.i.i.i1599:                              ; preds = %if.end14.i.i.i1594
  %2428 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !76
  %opcode.i.i.i1600 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %2428, i64 %idxprom.i.i.i.i1524, i32 1
  store i32 1660944387, ptr %opcode.i.i.i1600, align 8, !tbaa !77
  %arrayidx21.i.i.i1601 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %2428, i64 %idxprom.i.i.i.i1524
  store i64 %2399, ptr %arrayidx21.i.i.i1601, align 8, !tbaa !80
  br label %if.end22.i.i.i1598, !amdgpu.uniform !42

Flow8289:                                         ; preds = %if.end22.i.i.i1598, %Flow8288
  %2429 = phi i32 [ %2416, %if.end22.i.i.i1598 ], [ %retval.sroa.8.0.i.i.i1509, %Flow8288 ]
  %2430 = phi i64 [ %2399, %if.end22.i.i.i1598 ], [ %retval.sroa.2.0.i.i.i1508, %Flow8288 ]
  %2431 = phi i32 [ %spec.store.select.i.i.i1512, %if.end22.i.i.i1598 ], [ %retval.sroa.6.0.i.i.i1506, %Flow8288 ]
  br label %cleanup26.i.i.i1535, !amdgpu.uniform !42

if.end22.i.i.i1598:                               ; preds = %if.then16.i.i.i1599, %if.end14.i.i.i1594
  call void @llvm.amdgcn.end.cf.i64(i64 %2427)
  call void @llvm.amdgcn.wave.barrier()
  br label %Flow8289, !amdgpu.uniform !42

Flow8290:                                         ; preds = %cleanup26.i.i.i1535, %2405
  %2432 = phi i32 [ %2431, %cleanup26.i.i.i1535 ], [ %2395, %2405 ]
  %2433 = phi i64 [ %2430, %cleanup26.i.i.i1535 ], [ %2396, %2405 ]
  %2434 = phi i32 [ %2429, %cleanup26.i.i.i1535 ], [ %2397, %2405 ]
  %2435 = phi i32 [ %2429, %cleanup26.i.i.i1535 ], [ %retval.sroa.8.0.i.i.i1509, %2405 ]
  %2436 = phi i64 [ %2430, %cleanup26.i.i.i1535 ], [ %retval.sroa.2.0.i.i.i1508, %2405 ]
  %2437 = phi i32 [ %2431, %cleanup26.i.i.i1535 ], [ %retval.sroa.6.0.i.i.i1506, %2405 ]
  %2438 = phi i1 [ %cmp.i46.not.i.i.i1527, %cleanup26.i.i.i1535 ], [ true, %2405 ]
  br i1 %2438, label %for.inc.i.i.i1540, label %Flow8291, !amdgpu.uniform !42

cleanup26.i.i.i1535:                              ; preds = %Flow8289
  br label %Flow8290, !amdgpu.uniform !42

for.inc.i.i.i1540:                                ; preds = %Flow8290
  %inc.i.i.i1545 = add i32 %spec.store.select.i.i.i1512, 1
  br label %Flow8291, !amdgpu.uniform !42

Flow8291:                                         ; preds = %for.inc.i.i.i1540, %Flow8290
  %2439 = phi i32 [ %inc.i.i.i1545, %for.inc.i.i.i1540 ], [ poison, %Flow8290 ]
  %2440 = phi i1 [ false, %for.inc.i.i.i1540 ], [ true, %Flow8290 ]
  br i1 %2440, label %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1546, label %for.cond.i.i1.i1505, !amdgpu.uniform !42

_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1546: ; preds = %Flow8291
  %.lcssa9227 = phi i32 [ %2432, %Flow8291 ]
  %.lcssa9226 = phi i64 [ %2433, %Flow8291 ]
  %.lcssa9225 = phi i32 [ %2434, %Flow8291 ]
  %2441 = addrspacecast ptr addrspace(4) %scevgep7052.lcssa to ptr
  %2442 = addrspacecast ptr addrspace(4) %cond6 to ptr
  %2443 = ptrtoint ptr %2442 to i64
  %sub.ptr.lhs.cast.i.i.i1501 = ptrtoint ptr %2441 to i64
  %sub.ptr.sub.i.i.i1503 = sub i64 %sub.ptr.lhs.cast.i.i.i1501, %2443
  fence acquire
  %2444 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %mul.i.i.i.i.i.i1547 = shl i32 %.lcssa9227, 6
  %idxprom.i19.i.i.i.i.i1548 = zext i32 %mul.i.i.i.i.i.i1547 to i64
  %arrayidx.i20.i.i.i.i.i1549 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %2444, i64 %idxprom.i19.i.i.i.i.i1548
  %arrayidx.i22.i.i.i.i.i1550 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %arrayidx.i20.i.i.i.i.i1549, i64 %sh_prom.i.i.i.i756
  store i64 %sub.ptr.sub.i.i.i1503, ptr %arrayidx.i22.i.i.i.i.i1550, align 8, !tbaa !82
  %spec.select.i.i.i.i.i.i.i1551 = call i64 @llvm.umin.i64(i64 %sub.ptr.sub.i.i.i1503, i64 56)
  %2445 = and i64 %spec.select.i.i.i.i.i.i.i1551, 15
  %2446 = sub i64 %spec.select.i.i.i.i.i.i.i1551, %2445
  %2447 = icmp ne i64 %2446, 0
  %2448 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %2447)
  %2449 = extractvalue { i1, i64 } %2448, 0
  %2450 = extractvalue { i1, i64 } %2448, 1
  br i1 %2449, label %loop-memcpy-expansion5193.preheader, label %Flow8287

loop-memcpy-expansion5193.preheader:              ; preds = %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1546
  %scevgep7053 = getelementptr i8, ptr %2444, i64 8
  %scevgep7054 = getelementptr i8, ptr %scevgep7053, i64 %1474
  %2451 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i1548, 6
  %scevgep7055 = getelementptr i8, ptr %scevgep7054, i64 %2451
  br label %loop-memcpy-expansion5193, !amdgpu.uniform !42

Flow8287:                                         ; preds = %Flow8286, %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1546
  call void @llvm.amdgcn.end.cf.i64(i64 %2450)
  br label %loop-memcpy-residual-header5196, !amdgpu.uniform !42

loop-memcpy-expansion5193:                        ; preds = %loop-memcpy-expansion5193.preheader, %loop-memcpy-expansion5193
  %phi.broken8748 = phi i64 [ 0, %loop-memcpy-expansion5193.preheader ], [ %2456, %loop-memcpy-expansion5193 ]
  %loop-index5194 = phi i64 [ %2454, %loop-memcpy-expansion5193 ], [ 0, %loop-memcpy-expansion5193.preheader ]
  %2452 = getelementptr i8, ptr addrspace(4) %cond6, i64 %loop-index5194
  %2453 = load <4 x i32>, ptr addrspace(4) %2452, align 1
  %scevgep7056 = getelementptr i8, ptr %scevgep7055, i64 %loop-index5194
  store <4 x i32> %2453, ptr %scevgep7056, align 1
  %2454 = add i64 %loop-index5194, 16
  %2455 = icmp uge i64 %2454, %2446
  %2456 = call i64 @llvm.amdgcn.if.break.i64(i1 %2455, i64 %phi.broken8748)
  %2457 = call i1 @llvm.amdgcn.loop.i64(i64 %2456)
  br i1 %2457, label %Flow8286, label %loop-memcpy-expansion5193

Flow8285:                                         ; preds = %Flow8283, %Flow8284
  %2458 = phi i64 [ %6498, %Flow8283 ], [ %6493, %Flow8284 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %6496)
  br label %post-loop-memcpy-expansion5192, !amdgpu.uniform !42

loop-memcpy-residual5195:                         ; preds = %loop-memcpy-residual5195.preheader, %loop-memcpy-residual5195
  %phi.broken8749 = phi i64 [ 0, %loop-memcpy-residual5195.preheader ], [ %2462, %loop-memcpy-residual5195 ]
  %residual-loop-index5197 = phi i64 [ %2460, %loop-memcpy-residual5195 ], [ 0, %loop-memcpy-residual5195.preheader ]
  %scevgep7062 = getelementptr i8, ptr addrspace(4) %scevgep7061, i64 %residual-loop-index5197
  %2459 = load i8, ptr addrspace(4) %scevgep7062, align 1
  %scevgep7060 = getelementptr i8, ptr %scevgep7059, i64 %residual-loop-index5197
  store i8 %2459, ptr %scevgep7060, align 1
  %2460 = add i64 %residual-loop-index5197, 1
  %2461 = icmp uge i64 %2460, %2445
  %2462 = call i64 @llvm.amdgcn.if.break.i64(i1 %2461, i64 %phi.broken8749)
  %2463 = call i1 @llvm.amdgcn.loop.i64(i64 %2462)
  br i1 %2463, label %Flow8283, label %loop-memcpy-residual5195

Flow8283:                                         ; preds = %loop-memcpy-residual5195
  %.lcssa9229 = phi i64 [ %2462, %loop-memcpy-residual5195 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa9229)
  br label %Flow8285, !amdgpu.uniform !42

post-loop-memcpy-expansion5192:                   ; preds = %Flow8285
  %2464 = addrspacecast ptr addrspace(4) %scevgep7052.lcssa to ptr
  %2465 = ptrtoint ptr %2464 to i64
  %tobool.not.i.i.i.i.i.i1553 = icmp eq i32 %.lcssa9225, 0
  %conv.i.i.i.i.i.i1554 = zext i1 %tobool.not.i.i.i.i.i.i1553 to i32
  fence release
  %2466 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %idxprom.i23.i.i.i.i.i1555 = zext i32 %.lcssa9227 to i64
  %arrayidx.i24.i.i.i.i.i1556 = getelementptr inbounds nuw i32, ptr %2466, i64 %idxprom.i23.i.i.i.i.i1555
  store atomic i32 %conv.i.i.i.i.i.i1554, ptr %arrayidx.i24.i.i.i.i.i1556 monotonic, align 4
  %2467 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !84
  %arrayidx.i.i5.i.i1557 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %2467, i64 %idxprom.i23.i.i.i.i.i1555
  %2468 = load i64, ptr %arrayidx.i.i5.i.i1557, align 8, !tbaa !80
  %2469 = add i64 %1141, %2465
  br label %while.cond.i.i.i.i1558, !amdgpu.uniform !42

while.cond.i.i.i.i1558:                           ; preds = %post-loop-memcpy-expansion5192, %Flow8282
  %phi.broken8753 = phi i64 [ 0, %post-loop-memcpy-expansion5192 ], [ %2491, %Flow8282 ]
  %lsr.iv7068 = phi i64 [ %2489, %Flow8282 ], [ %2469, %post-loop-memcpy-expansion5192 ]
  %lsr.iv7063 = phi ptr addrspace(4) [ %2488, %Flow8282 ], [ %scevgep6648, %post-loop-memcpy-expansion5192 ]
  %port.sroa.37.0.i.i1559 = phi i32 [ %2487, %Flow8282 ], [ %conv.i.i.i.i.i.i1554, %post-loop-memcpy-expansion5192 ]
  %idx.0.i.i.i.i1560 = phi i64 [ %2486, %Flow8282 ], [ 56, %post-loop-memcpy-expansion5192 ]
  %cmp.i.i.i.i1561 = icmp ugt i64 %sub.ptr.sub.i.i.i1503, %idx.0.i.i.i.i1560
  %2470 = call i64 @llvm.amdgcn.ballot.i64(i1 %cmp.i.i.i.i1561)
  %and.i.i.i.i6.i.i1562 = and i64 %2470, %2468
  %tobool.not.i.i.i.i1563 = icmp ne i64 %and.i.i.i.i6.i.i1562, 0
  %2471 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i22.i.i.i1564 = getelementptr inbounds nuw i32, ptr %2471, i64 %idxprom.i23.i.i.i.i.i1555
  %2472 = load atomic i32, ptr %arrayidx.i22.i.i.i1564 monotonic, align 4
  %2473 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %tobool.not.i.i.i.i1563)
  %2474 = extractvalue { i1, i64 } %2473, 0
  %2475 = extractvalue { i1, i64 } %2473, 1
  br i1 %2474, label %cond.false.i8.i.i.i.i1565, label %Flow8282

cond.false.i8.i.i.i.i1565:                        ; preds = %while.cond.i.i.i.i1558
  %2476 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %2472)
  %2477 = icmp ne i32 %2476, %port.sroa.37.0.i.i1559
  %2478 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %2477)
  %2479 = extractvalue { i1, i64 } %2478, 0
  %2480 = extractvalue { i1, i64 } %2478, 1
  br i1 %2479, label %while.body.i.i21.i.i.i.i1566.preheader, label %Flow8281

while.body.i.i21.i.i.i.i1566.preheader:           ; preds = %cond.false.i8.i.i.i.i1565
  br label %while.body.i.i21.i.i.i.i1566, !amdgpu.uniform !42

Flow8281:                                         ; preds = %Flow8280, %cond.false.i8.i.i.i.i1565
  call void @llvm.amdgcn.end.cf.i64(i64 %2480)
  br label %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i1569, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i1566:                     ; preds = %while.body.i.i21.i.i.i.i1566.preheader, %while.body.i.i21.i.i.i.i1566
  %phi.broken8750 = phi i64 [ 0, %while.body.i.i21.i.i.i.i1566.preheader ], [ %2484, %while.body.i.i21.i.i.i.i1566 ]
  call void @llvm.amdgcn.s.sleep(i32 2)
  %2481 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i22.i.i.i.i1567 = getelementptr inbounds nuw i32, ptr %2481, i64 %idxprom.i23.i.i.i.i.i1555
  %2482 = load atomic i32, ptr %arrayidx.i.i.i22.i.i.i.i1567 monotonic, align 4
  %2483 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %2482)
  %cmp.i.not.i.i23.i.i.i.i1568 = icmp eq i32 %2483, %port.sroa.37.0.i.i1559
  %2484 = call i64 @llvm.amdgcn.if.break.i64(i1 %cmp.i.not.i.i23.i.i.i.i1568, i64 %phi.broken8750)
  %2485 = call i1 @llvm.amdgcn.loop.i64(i64 %2484)
  br i1 %2485, label %Flow8280, label %while.body.i.i21.i.i.i.i1566

Flow8280:                                         ; preds = %while.body.i.i21.i.i.i.i1566
  %.lcssa9230 = phi i64 [ %2484, %while.body.i.i21.i.i.i.i1566 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa9230)
  br label %Flow8281, !amdgpu.uniform !42

Flow8282:                                         ; preds = %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1570, %while.cond.i.i.i.i1558
  %2486 = phi i64 [ %add.i.i.i.i1573, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1570 ], [ poison, %while.cond.i.i.i.i1558 ]
  %2487 = phi i32 [ %conv.i.i26.i.i.i.i1571, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1570 ], [ poison, %while.cond.i.i.i.i1558 ]
  %2488 = phi ptr addrspace(4) [ %scevgep7064, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1570 ], [ poison, %while.cond.i.i.i.i1558 ]
  %2489 = phi i64 [ %lsr.iv.next7069, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1570 ], [ poison, %while.cond.i.i.i.i1558 ]
  %2490 = phi i1 [ false, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1570 ], [ true, %while.cond.i.i.i.i1558 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %2475)
  %2491 = call i64 @llvm.amdgcn.if.break.i64(i1 %2490, i64 %phi.broken8753)
  %2492 = call i1 @llvm.amdgcn.loop.i64(i64 %2491)
  br i1 %2492, label %cond.false.i.i.i1580, label %while.cond.i.i.i.i1558

_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i1569: ; preds = %Flow8281
  fence acquire
  %2493 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.i.i.i.i1561)
  %2494 = extractvalue { i1, i64 } %2493, 0
  %2495 = extractvalue { i1, i64 } %2493, 1
  br i1 %2494, label %if.then.i.i.i.i.i.i.i1574, label %Flow8279

if.then.i.i.i.i.i.i.i1574:                        ; preds = %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i1569
  %sub.i.i.i.i.i.i.i1576 = sub nuw i64 %sub.ptr.sub.i.i.i1503, %idx.0.i.i.i.i1560
  %spec.select.i.i.i36.i.i.i.i1577 = call i64 @llvm.umin.i64(i64 %sub.i.i.i.i.i.i.i1576, i64 64)
  %2496 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %2497 = and i64 %spec.select.i.i.i36.i.i.i.i1577, 15
  %2498 = sub i64 %spec.select.i.i.i36.i.i.i.i1577, %2497
  %2499 = icmp ne i64 %2498, 0
  %2500 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %2499)
  %2501 = extractvalue { i1, i64 } %2500, 0
  %2502 = extractvalue { i1, i64 } %2500, 1
  br i1 %2501, label %loop-memcpy-expansion5307.preheader, label %Flow8278

loop-memcpy-expansion5307.preheader:              ; preds = %if.then.i.i.i.i.i.i.i1574
  %arrayidx.i20.i35.i.i.i.i1578 = getelementptr %"struct.rpc::Buffer", ptr %2496, i64 %idxprom.i19.i.i.i.i.i1548
  %arrayidx.i22.i38.i.i.i.i1579 = getelementptr %"struct.rpc::Buffer", ptr %arrayidx.i20.i35.i.i.i.i1578, i64 %sh_prom.i.i.i.i756
  br label %loop-memcpy-expansion5307, !amdgpu.uniform !42

Flow8278:                                         ; preds = %Flow8277, %if.then.i.i.i.i.i.i.i1574
  call void @llvm.amdgcn.end.cf.i64(i64 %2502)
  br label %loop-memcpy-residual-header5310, !amdgpu.uniform !42

loop-memcpy-expansion5307:                        ; preds = %loop-memcpy-expansion5307.preheader, %loop-memcpy-expansion5307
  %phi.broken8751 = phi i64 [ 0, %loop-memcpy-expansion5307.preheader ], [ %2507, %loop-memcpy-expansion5307 ]
  %loop-index5308 = phi i64 [ %2505, %loop-memcpy-expansion5307 ], [ 0, %loop-memcpy-expansion5307.preheader ]
  %scevgep7065 = getelementptr i8, ptr addrspace(4) %lsr.iv7063, i64 %loop-index5308
  %2503 = load <4 x i32>, ptr addrspace(4) %scevgep7065, align 1
  %2504 = getelementptr i8, ptr %arrayidx.i22.i38.i.i.i.i1579, i64 %loop-index5308
  store <4 x i32> %2503, ptr %2504, align 1
  %2505 = add i64 %loop-index5308, 16
  %2506 = icmp uge i64 %2505, %2498
  %2507 = call i64 @llvm.amdgcn.if.break.i64(i1 %2506, i64 %phi.broken8751)
  %2508 = call i1 @llvm.amdgcn.loop.i64(i64 %2507)
  br i1 %2508, label %Flow8277, label %loop-memcpy-expansion5307

Flow8276:                                         ; preds = %Flow8275, %loop-memcpy-residual-header5310
  call void @llvm.amdgcn.end.cf.i64(i64 %6588)
  br label %Flow8279, !amdgpu.uniform !42

loop-memcpy-residual5309:                         ; preds = %loop-memcpy-residual5309.preheader, %loop-memcpy-residual5309
  %phi.broken8752 = phi i64 [ 0, %loop-memcpy-residual5309.preheader ], [ %2512, %loop-memcpy-residual5309 ]
  %residual-loop-index5311 = phi i64 [ %2510, %loop-memcpy-residual5309 ], [ 0, %loop-memcpy-residual5309.preheader ]
  %scevgep7074 = getelementptr i8, ptr addrspace(4) %scevgep7073, i64 %residual-loop-index5311
  %2509 = load i8, ptr addrspace(4) %scevgep7074, align 1
  %scevgep7072 = getelementptr i8, ptr %scevgep7071, i64 %residual-loop-index5311
  store i8 %2509, ptr %scevgep7072, align 1
  %2510 = add i64 %residual-loop-index5311, 1
  %2511 = icmp uge i64 %2510, %2497
  %2512 = call i64 @llvm.amdgcn.if.break.i64(i1 %2511, i64 %phi.broken8752)
  %2513 = call i1 @llvm.amdgcn.loop.i64(i64 %2512)
  br i1 %2513, label %Flow8275, label %loop-memcpy-residual5309

Flow8275:                                         ; preds = %loop-memcpy-residual5309
  %.lcssa9232 = phi i64 [ %2512, %loop-memcpy-residual5309 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa9232)
  br label %Flow8276, !amdgpu.uniform !42

_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1570: ; preds = %Flow8279
  %conv.i.i26.i.i.i.i1571 = xor i32 %port.sroa.37.0.i.i1559, 1
  fence release
  %2514 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %arrayidx.i24.i29.i.i.i.i1572 = getelementptr inbounds nuw i32, ptr %2514, i64 %idxprom.i23.i.i.i.i.i1555
  store atomic i32 %conv.i.i26.i.i.i.i1571, ptr %arrayidx.i24.i29.i.i.i.i1572 monotonic, align 4
  %add.i.i.i.i1573 = add i64 %idx.0.i.i.i.i1560, 64
  %scevgep7064 = getelementptr i8, ptr addrspace(4) %lsr.iv7063, i64 64
  %lsr.iv.next7069 = add i64 %lsr.iv7068, -64
  br label %Flow8282, !amdgpu.uniform !42

cond.false.i.i.i1580:                             ; preds = %Flow8282
  %.lcssa9234 = phi i64 [ %2491, %Flow8282 ]
  %port.sroa.37.0.i.i1559.lcssa = phi i32 [ %port.sroa.37.0.i.i1559, %Flow8282 ]
  %.lcssa9233 = phi i32 [ %2472, %Flow8282 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa9234)
  %2515 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %.lcssa9233)
  %2516 = icmp ne i32 %2515, %port.sroa.37.0.i.i1559.lcssa
  %2517 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %2516)
  %2518 = extractvalue { i1, i64 } %2517, 0
  %2519 = extractvalue { i1, i64 } %2517, 1
  br i1 %2518, label %while.body.i.i.i.i1581.preheader, label %Flow8274

while.body.i.i.i.i1581.preheader:                 ; preds = %cond.false.i.i.i1580
  br label %while.body.i.i.i.i1581, !amdgpu.uniform !42

Flow8274:                                         ; preds = %Flow8273, %cond.false.i.i.i1580
  call void @llvm.amdgcn.end.cf.i64(i64 %2519)
  br label %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit1602, !amdgpu.uniform !42

while.body.i.i.i.i1581:                           ; preds = %while.body.i.i.i.i1581.preheader, %while.body.i.i.i.i1581
  %phi.broken8754 = phi i64 [ 0, %while.body.i.i.i.i1581.preheader ], [ %2523, %while.body.i.i.i.i1581 ]
  call void @llvm.amdgcn.s.sleep(i32 2)
  %2520 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i8.i.i1582 = getelementptr inbounds nuw i32, ptr %2520, i64 %idxprom.i23.i.i.i.i.i1555
  %2521 = load atomic i32, ptr %arrayidx.i.i.i8.i.i1582 monotonic, align 4
  %2522 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %2521)
  %cmp.i.not.i.i.i.i1583 = icmp eq i32 %2522, %port.sroa.37.0.i.i1559.lcssa
  %2523 = call i64 @llvm.amdgcn.if.break.i64(i1 %cmp.i.not.i.i.i.i1583, i64 %phi.broken8754)
  %2524 = call i1 @llvm.amdgcn.loop.i64(i64 %2523)
  br i1 %2524, label %Flow8273, label %while.body.i.i.i.i1581

Flow8273:                                         ; preds = %while.body.i.i.i.i1581
  %.lcssa9235 = phi i64 [ %2523, %while.body.i.i.i.i1581 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa9235)
  br label %Flow8274, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit1602: ; preds = %Flow8274
  fence acquire
  call void @llvm.amdgcn.wave.barrier()
  fence syncscope("agent") release
  %2525 = call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %.lcssa9226, i1 true)
  %iszero.i.i.i.i.i13.i.i1585 = icmp ne i64 %.lcssa9226, 0
  %cmp2.i.i.i.i14.i.i1586 = icmp eq i64 %2525, %sh_prom.i.i.i.i756
  %cmp.i.i.i.i15.i.i1587 = select i1 %iszero.i.i.i.i.i13.i.i1585, i1 %cmp2.i.i.i.i14.i.i1586, i1 false
  %2526 = and i32 %.lcssa9227, 31
  %conv4.i.i.i19.i.i1591 = zext i1 %cmp.i.i.i.i15.i.i1587 to i32
  %shl.i.i.i20.i.i1592 = shl nuw i32 %conv4.i.i.i19.i.i1591, %2526
  %xor.i.i.i21.i.i1593 = xor i32 %shl.i.i.i20.i.i1592, -1
  %2527 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop5667, !amdgpu.uniform !42

2528:                                             ; preds = %ComputeEnd5668
  %div8.i.i.i16.i.i1588 = lshr i32 %.lcssa9227, 5
  %idxprom.i.i.i17.i.i1589 = zext nneg i32 %div8.i.i.i16.i.i1588 to i64
  %2529 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i17.i.i1589
  %sunkaddr7413 = getelementptr inbounds i8, ptr addrspace(1) %2529, i64 40
  %2530 = atomicrmw and ptr addrspace(1) %sunkaddr7413, i32 %.lcssa9236 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %2531, !amdgpu.uniform !42

2531:                                             ; preds = %2528, %ComputeEnd5668
  call void @llvm.amdgcn.end.cf.i64(i64 %7552)
  call void @llvm.amdgcn.wave.barrier()
  br label %for.cond.i.i.i1603, !amdgpu.uniform !42

for.cond.i.i.i1603:                               ; preds = %2531, %for.cond.i.i.i1603
  %phi.broken8755 = phi i64 [ 0, %2531 ], [ %2533, %for.cond.i.i.i1603 ]
  %lsr.iv7097 = phi i64 [ %lsr.iv.next7098, %for.cond.i.i.i1603 ], [ -56, %2531 ]
  %lsr.iv7084 = phi i64 [ %lsr.iv.next7085, %for.cond.i.i.i1603 ], [ -1, %2531 ]
  %lsr.iv7075 = phi ptr [ %scevgep7077, %for.cond.i.i.i1603 ], [ %scevgep6661, %2531 ]
  %scevgep7077 = getelementptr i8, ptr %lsr.iv7075, i64 1
  %2532 = load i8, ptr %scevgep7077, align 1, !tbaa !57
  %cmp.i.i.i1605 = icmp eq i8 %2532, 0
  %lsr.iv.next7085 = add i64 %lsr.iv7084, 1
  %lsr.iv.next7098 = add i64 %lsr.iv7097, 1
  %2533 = call i64 @llvm.amdgcn.if.break.i64(i1 %cmp.i.i.i1605, i64 %phi.broken8755)
  %2534 = call i1 @llvm.amdgcn.loop.i64(i64 %2533)
  br i1 %2534, label %for.cond.i.i1.i1612.preheader, label %for.cond.i.i.i1603

for.cond.i.i1.i1612.preheader:                    ; preds = %for.cond.i.i.i1603
  %lsr.iv7097.lcssa = phi i64 [ %lsr.iv7097, %for.cond.i.i.i1603 ]
  %scevgep7077.lcssa = phi ptr [ %scevgep7077, %for.cond.i.i.i1603 ]
  %lsr.iv.next7085.lcssa = phi i64 [ %lsr.iv.next7085, %for.cond.i.i.i1603 ]
  %.lcssa9237 = phi i64 [ %2533, %for.cond.i.i.i1603 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa9237)
  br label %for.cond.i.i1.i1612, !amdgpu.uniform !42

for.cond.i.i1.i1612:                              ; preds = %for.cond.i.i1.i1612.preheader, %Flow8272
  %2535 = phi i32 [ %2572, %Flow8272 ], [ %1174, %for.cond.i.i1.i1612.preheader ]
  %2536 = phi i64 [ %2573, %Flow8272 ], [ %1175, %for.cond.i.i1.i1612.preheader ]
  %2537 = phi i32 [ %2574, %Flow8272 ], [ %1176, %for.cond.i.i1.i1612.preheader ]
  %retval.sroa.6.0.i.i.i1613 = phi i32 [ %2577, %Flow8272 ], [ undef, %for.cond.i.i1.i1612.preheader ]
  %retval.sroa.2.0.i.i.i1615 = phi i64 [ %2576, %Flow8272 ], [ undef, %for.cond.i.i1.i1612.preheader ]
  %retval.sroa.8.0.i.i.i1616 = phi i32 [ %2575, %Flow8272 ], [ undef, %for.cond.i.i1.i1612.preheader ]
  %index.0.i.i.i1617 = phi i32 [ %2579, %Flow8272 ], [ 0, %for.cond.i.i1.i1612.preheader ]
  %2538 = load i32, ptr addrspace(1) @__llvm_rpc_client, align 8, !tbaa !68
  %cmp.not.i.i.i1618 = icmp ult i32 %index.0.i.i.i1617, %2538
  %spec.store.select.i.i.i1619 = select i1 %cmp.not.i.i.i1618, i32 %index.0.i.i.i1617, i32 0
  %2539 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  %2540 = and i32 %spec.store.select.i.i.i1619, 31
  %2541 = lshr i64 %2539, %sh_prom.i.i.i.i756
  %2542 = trunc i64 %2541 to i32
  %conv4.i.i.i.i.i1623 = and i32 %2542, 1
  %shl.i.i.i.i.i1624 = shl nuw i32 %conv4.i.i.i.i.i1623, %2540
  br label %ComputeLoop5671, !amdgpu.uniform !42

2543:                                             ; preds = %ComputeEnd5672
  %sunkaddr7414 = getelementptr inbounds i8, ptr addrspace(1) %7562, i64 40
  %2544 = atomicrmw or ptr addrspace(1) %sunkaddr7414, i32 %.lcssa9238 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %2545, !amdgpu.uniform !42

2545:                                             ; preds = %2543, %ComputeEnd5672
  %2546 = phi i32 [ %2544, %2543 ], [ poison, %ComputeEnd5672 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %7571)
  %2547 = call i32 @llvm.amdgcn.readfirstlane.i32(i32 %2546)
  %2548 = or i32 %2547, %.lcssa9239
  %shl5.i.i.i.i.i1625 = shl nuw i32 1, %2540
  %and.i.i.i.i.i1626 = and i32 %shl5.i.i.i.i.i1625, %2548
  %tobool3.i.i.i.i1627 = icmp ne i32 %and.i.i.i.i.i1626, 0
  %2549 = call i64 @llvm.amdgcn.ballot.i64(i1 %tobool3.i.i.i.i1627)
  %and.i.i.i.i.i.i1628 = and i64 %2549, %2539
  %cmp.i.not.i.i.i1629 = icmp ne i64 %2539, %and.i.i.i.i.i.i1628
  br i1 %cmp.i.not.i.i.i1629, label %if.then.i.i.i.i1630, label %Flow8271, !amdgpu.uniform !42

if.then.i.i.i.i1630:                              ; preds = %2545
  fence syncscope("agent") acquire
  %2550 = load <2 x i64>, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !75
  %2551 = extractelement <2 x i64> %2550, i32 0
  %2552 = inttoptr i64 %2551 to ptr
  %2553 = extractelement <2 x i64> %2550, i32 1
  %2554 = inttoptr i64 %2553 to ptr
  %idxprom.i.i.i.i1631 = zext i32 %spec.store.select.i.i.i1619 to i64
  %arrayidx.i.i.i.i1632 = getelementptr inbounds nuw i32, ptr %2552, i64 %idxprom.i.i.i.i1631, !amdgpu.uniform !42
  %2555 = load atomic i32, ptr %arrayidx.i.i.i.i1632 monotonic, align 4
  %2556 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %2555)
  %arrayidx.i45.i.i.i1633 = getelementptr inbounds nuw i32, ptr %2554, i64 %idxprom.i.i.i.i1631, !amdgpu.uniform !42
  %2557 = load atomic i32, ptr %arrayidx.i45.i.i.i1633 monotonic, align 4
  %2558 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %2557)
  %cmp.i46.not.i.i.i1634 = icmp ne i32 %2556, %2558
  br i1 %cmp.i46.not.i.i.i1634, label %if.then12.i.i.i1635, label %Flow8269, !amdgpu.uniform !42

if.then12.i.i.i1635:                              ; preds = %if.then.i.i.i.i1630
  fence syncscope("agent") release
  %2559 = call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %2539, i1 true)
  %iszero.i.i.i.i.i.i.i1636 = icmp ne i64 %2539, 0
  %cmp2.i.i.i.i.i.i1637 = icmp eq i64 %2559, %sh_prom.i.i.i.i756
  %cmp.i.i.i.i.i.i1638 = select i1 %iszero.i.i.i.i.i.i.i1636, i1 %cmp2.i.i.i.i.i.i1637, i1 false
  %conv4.i.i51.i.i.i1639 = zext i1 %cmp.i.i.i.i.i.i1638 to i32
  %shl.i.i52.i.i.i1640 = shl nuw i32 %conv4.i.i51.i.i.i1639, %2540
  %xor.i.i.i.i.i1641 = xor i32 %shl.i.i52.i.i.i1640, -1
  %2560 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop5676, !amdgpu.uniform !42

2561:                                             ; preds = %ComputeEnd5677
  %sunkaddr7415 = getelementptr inbounds i8, ptr addrspace(1) %7562, i64 40
  %2562 = atomicrmw and ptr addrspace(1) %sunkaddr7415, i32 %.lcssa9240 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %2563, !amdgpu.uniform !42

2563:                                             ; preds = %2561, %ComputeEnd5677
  call void @llvm.amdgcn.end.cf.i64(i64 %7589)
  call void @llvm.amdgcn.wave.barrier()
  br label %Flow8269, !amdgpu.uniform !42

if.end14.i.i.i1701:                               ; preds = %Flow8269
  %2564 = call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %2539, i1 true)
  %iszero.i.i.i.i.i.i1702 = icmp ne i64 %2539, 0
  %cmp2.i.i.i.i.i1703 = icmp eq i64 %2564, %sh_prom.i.i.i.i756
  %cmp.i.i.i.i.i1704 = select i1 %iszero.i.i.i.i.i.i1702, i1 %cmp2.i.i.i.i.i1703, i1 false
  %2565 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.i.i.i.i.i1704)
  %2566 = extractvalue { i1, i64 } %2565, 0
  %2567 = extractvalue { i1, i64 } %2565, 1
  br i1 %2566, label %if.then16.i.i.i1706, label %if.end22.i.i.i1705

if.then16.i.i.i1706:                              ; preds = %if.end14.i.i.i1701
  %2568 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !76
  %opcode.i.i.i1707 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %2568, i64 %idxprom.i.i.i.i1631, i32 1
  store i32 1660944387, ptr %opcode.i.i.i1707, align 8, !tbaa !77
  %arrayidx21.i.i.i1708 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %2568, i64 %idxprom.i.i.i.i1631
  store i64 %2539, ptr %arrayidx21.i.i.i1708, align 8, !tbaa !80
  br label %if.end22.i.i.i1705, !amdgpu.uniform !42

Flow8270:                                         ; preds = %if.end22.i.i.i1705, %Flow8269
  %2569 = phi i32 [ %2556, %if.end22.i.i.i1705 ], [ %retval.sroa.8.0.i.i.i1616, %Flow8269 ]
  %2570 = phi i64 [ %2539, %if.end22.i.i.i1705 ], [ %retval.sroa.2.0.i.i.i1615, %Flow8269 ]
  %2571 = phi i32 [ %spec.store.select.i.i.i1619, %if.end22.i.i.i1705 ], [ %retval.sroa.6.0.i.i.i1613, %Flow8269 ]
  br label %cleanup26.i.i.i1642, !amdgpu.uniform !42

if.end22.i.i.i1705:                               ; preds = %if.then16.i.i.i1706, %if.end14.i.i.i1701
  call void @llvm.amdgcn.end.cf.i64(i64 %2567)
  call void @llvm.amdgcn.wave.barrier()
  br label %Flow8270, !amdgpu.uniform !42

Flow8271:                                         ; preds = %cleanup26.i.i.i1642, %2545
  %2572 = phi i32 [ %2571, %cleanup26.i.i.i1642 ], [ %2535, %2545 ]
  %2573 = phi i64 [ %2570, %cleanup26.i.i.i1642 ], [ %2536, %2545 ]
  %2574 = phi i32 [ %2569, %cleanup26.i.i.i1642 ], [ %2537, %2545 ]
  %2575 = phi i32 [ %2569, %cleanup26.i.i.i1642 ], [ %retval.sroa.8.0.i.i.i1616, %2545 ]
  %2576 = phi i64 [ %2570, %cleanup26.i.i.i1642 ], [ %retval.sroa.2.0.i.i.i1615, %2545 ]
  %2577 = phi i32 [ %2571, %cleanup26.i.i.i1642 ], [ %retval.sroa.6.0.i.i.i1613, %2545 ]
  %2578 = phi i1 [ %cmp.i46.not.i.i.i1634, %cleanup26.i.i.i1642 ], [ true, %2545 ]
  br i1 %2578, label %for.inc.i.i.i1647, label %Flow8272, !amdgpu.uniform !42

cleanup26.i.i.i1642:                              ; preds = %Flow8270
  br label %Flow8271, !amdgpu.uniform !42

for.inc.i.i.i1647:                                ; preds = %Flow8271
  %inc.i.i.i1652 = add i32 %spec.store.select.i.i.i1619, 1
  br label %Flow8272, !amdgpu.uniform !42

Flow8272:                                         ; preds = %for.inc.i.i.i1647, %Flow8271
  %2579 = phi i32 [ %inc.i.i.i1652, %for.inc.i.i.i1647 ], [ poison, %Flow8271 ]
  %2580 = phi i1 [ false, %for.inc.i.i.i1647 ], [ true, %Flow8271 ]
  br i1 %2580, label %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1653, label %for.cond.i.i1.i1612, !amdgpu.uniform !42

_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1653: ; preds = %Flow8272
  %.lcssa9243 = phi i32 [ %2572, %Flow8272 ]
  %.lcssa9242 = phi i64 [ %2573, %Flow8272 ]
  %.lcssa9241 = phi i32 [ %2574, %Flow8272 ]
  %2581 = ptrtoint ptr %call18 to i64
  %sub.ptr.lhs.cast.i.i.i1608 = ptrtoint ptr %scevgep7077.lcssa to i64
  %sub.ptr.sub.i.i.i1610 = sub i64 %sub.ptr.lhs.cast.i.i.i1608, %2581
  fence acquire
  %2582 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %mul.i.i.i.i.i.i1654 = shl i32 %.lcssa9243, 6
  %idxprom.i19.i.i.i.i.i1655 = zext i32 %mul.i.i.i.i.i.i1654 to i64
  %arrayidx.i20.i.i.i.i.i1656 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %2582, i64 %idxprom.i19.i.i.i.i.i1655
  %arrayidx.i22.i.i.i.i.i1657 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %arrayidx.i20.i.i.i.i.i1656, i64 %sh_prom.i.i.i.i756
  store i64 %sub.ptr.sub.i.i.i1610, ptr %arrayidx.i22.i.i.i.i.i1657, align 8, !tbaa !82
  %spec.select.i.i.i.i.i.i.i1658 = call i64 @llvm.umin.i64(i64 %sub.ptr.sub.i.i.i1610, i64 56)
  %2583 = and i64 %spec.select.i.i.i.i.i.i.i1658, 15
  %2584 = sub i64 %spec.select.i.i.i.i.i.i.i1658, %2583
  %2585 = icmp ne i64 %2584, 0
  %2586 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %2585)
  %2587 = extractvalue { i1, i64 } %2586, 0
  %2588 = extractvalue { i1, i64 } %2586, 1
  br i1 %2587, label %loop-memcpy-expansion5061.preheader, label %Flow8268

loop-memcpy-expansion5061.preheader:              ; preds = %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1653
  %scevgep7078 = getelementptr i8, ptr %2582, i64 8
  %scevgep7079 = getelementptr i8, ptr %scevgep7078, i64 %1474
  %2589 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i1655, 6
  %scevgep7080 = getelementptr i8, ptr %scevgep7079, i64 %2589
  br label %loop-memcpy-expansion5061, !amdgpu.uniform !42

Flow8268:                                         ; preds = %Flow8267, %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1653
  call void @llvm.amdgcn.end.cf.i64(i64 %2588)
  br label %loop-memcpy-residual-header5064, !amdgpu.uniform !42

loop-memcpy-expansion5061:                        ; preds = %loop-memcpy-expansion5061.preheader, %loop-memcpy-expansion5061
  %phi.broken8756 = phi i64 [ 0, %loop-memcpy-expansion5061.preheader ], [ %2594, %loop-memcpy-expansion5061 ]
  %loop-index5062 = phi i64 [ %2592, %loop-memcpy-expansion5061 ], [ 0, %loop-memcpy-expansion5061.preheader ]
  %2590 = getelementptr i8, ptr %call18, i64 %loop-index5062
  %2591 = load <4 x i32>, ptr %2590, align 1
  %scevgep7081 = getelementptr i8, ptr %scevgep7080, i64 %loop-index5062
  store <4 x i32> %2591, ptr %scevgep7081, align 1
  %2592 = add i64 %loop-index5062, 16
  %2593 = icmp uge i64 %2592, %2584
  %2594 = call i64 @llvm.amdgcn.if.break.i64(i1 %2593, i64 %phi.broken8756)
  %2595 = call i1 @llvm.amdgcn.loop.i64(i64 %2594)
  br i1 %2595, label %Flow8267, label %loop-memcpy-expansion5061

Flow8266:                                         ; preds = %Flow8264, %Flow8265
  %2596 = phi i64 [ %6315, %Flow8264 ], [ %6309, %Flow8265 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %6312)
  br label %post-loop-memcpy-expansion5060, !amdgpu.uniform !42

loop-memcpy-residual5063:                         ; preds = %loop-memcpy-residual5063.preheader, %loop-memcpy-residual5063
  %phi.broken8757 = phi i64 [ 0, %loop-memcpy-residual5063.preheader ], [ %2600, %loop-memcpy-residual5063 ]
  %residual-loop-index5065 = phi i64 [ %2598, %loop-memcpy-residual5063 ], [ 0, %loop-memcpy-residual5063.preheader ]
  %scevgep7091 = getelementptr i8, ptr %scevgep7090, i64 %residual-loop-index5065
  %2597 = load i8, ptr %scevgep7091, align 1
  %scevgep7089 = getelementptr i8, ptr %scevgep7088, i64 %residual-loop-index5065
  store i8 %2597, ptr %scevgep7089, align 1
  %2598 = add i64 %residual-loop-index5065, 1
  %2599 = icmp uge i64 %2598, %2583
  %2600 = call i64 @llvm.amdgcn.if.break.i64(i1 %2599, i64 %phi.broken8757)
  %2601 = call i1 @llvm.amdgcn.loop.i64(i64 %2600)
  br i1 %2601, label %Flow8264, label %loop-memcpy-residual5063

Flow8264:                                         ; preds = %loop-memcpy-residual5063
  %.lcssa9245 = phi i64 [ %2600, %loop-memcpy-residual5063 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa9245)
  br label %Flow8266, !amdgpu.uniform !42

post-loop-memcpy-expansion5060:                   ; preds = %Flow8266
  %tobool.not.i.i.i.i.i.i1660 = icmp eq i32 %.lcssa9241, 0
  %conv.i.i.i.i.i.i1661 = zext i1 %tobool.not.i.i.i.i.i.i1660 to i32
  fence release
  %2602 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %idxprom.i23.i.i.i.i.i1662 = zext i32 %.lcssa9243 to i64
  %arrayidx.i24.i.i.i.i.i1663 = getelementptr inbounds nuw i32, ptr %2602, i64 %idxprom.i23.i.i.i.i.i1662
  store atomic i32 %conv.i.i.i.i.i.i1661, ptr %arrayidx.i24.i.i.i.i.i1663 monotonic, align 4
  %2603 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !84
  %arrayidx.i.i5.i.i1664 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %2603, i64 %idxprom.i23.i.i.i.i.i1662
  %2604 = load i64, ptr %arrayidx.i.i5.i.i1664, align 8, !tbaa !80
  br label %while.cond.i.i.i.i1665, !amdgpu.uniform !42

while.cond.i.i.i.i1665:                           ; preds = %post-loop-memcpy-expansion5060, %Flow8263
  %phi.broken8761 = phi i64 [ 0, %post-loop-memcpy-expansion5060 ], [ %2626, %Flow8263 ]
  %lsr.iv7099 = phi i64 [ %2624, %Flow8263 ], [ %lsr.iv7097.lcssa, %post-loop-memcpy-expansion5060 ]
  %lsr.iv7092 = phi ptr [ %2623, %Flow8263 ], [ %scevgep6679, %post-loop-memcpy-expansion5060 ]
  %port.sroa.37.0.i.i1666 = phi i32 [ %2622, %Flow8263 ], [ %conv.i.i.i.i.i.i1661, %post-loop-memcpy-expansion5060 ]
  %idx.0.i.i.i.i1667 = phi i64 [ %2621, %Flow8263 ], [ 56, %post-loop-memcpy-expansion5060 ]
  %cmp.i.i.i.i1668 = icmp ugt i64 %sub.ptr.sub.i.i.i1610, %idx.0.i.i.i.i1667
  %2605 = call i64 @llvm.amdgcn.ballot.i64(i1 %cmp.i.i.i.i1668)
  %and.i.i.i.i6.i.i1669 = and i64 %2605, %2604
  %tobool.not.i.i.i.i1670 = icmp ne i64 %and.i.i.i.i6.i.i1669, 0
  %2606 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i22.i.i.i1671 = getelementptr inbounds nuw i32, ptr %2606, i64 %idxprom.i23.i.i.i.i.i1662
  %2607 = load atomic i32, ptr %arrayidx.i22.i.i.i1671 monotonic, align 4
  %2608 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %tobool.not.i.i.i.i1670)
  %2609 = extractvalue { i1, i64 } %2608, 0
  %2610 = extractvalue { i1, i64 } %2608, 1
  br i1 %2609, label %cond.false.i8.i.i.i.i1672, label %Flow8263

cond.false.i8.i.i.i.i1672:                        ; preds = %while.cond.i.i.i.i1665
  %2611 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %2607)
  %2612 = icmp ne i32 %2611, %port.sroa.37.0.i.i1666
  %2613 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %2612)
  %2614 = extractvalue { i1, i64 } %2613, 0
  %2615 = extractvalue { i1, i64 } %2613, 1
  br i1 %2614, label %while.body.i.i21.i.i.i.i1673.preheader, label %Flow8262

while.body.i.i21.i.i.i.i1673.preheader:           ; preds = %cond.false.i8.i.i.i.i1672
  br label %while.body.i.i21.i.i.i.i1673, !amdgpu.uniform !42

Flow8262:                                         ; preds = %Flow8261, %cond.false.i8.i.i.i.i1672
  call void @llvm.amdgcn.end.cf.i64(i64 %2615)
  br label %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i1676, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i1673:                     ; preds = %while.body.i.i21.i.i.i.i1673.preheader, %while.body.i.i21.i.i.i.i1673
  %phi.broken8758 = phi i64 [ 0, %while.body.i.i21.i.i.i.i1673.preheader ], [ %2619, %while.body.i.i21.i.i.i.i1673 ]
  call void @llvm.amdgcn.s.sleep(i32 2)
  %2616 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i22.i.i.i.i1674 = getelementptr inbounds nuw i32, ptr %2616, i64 %idxprom.i23.i.i.i.i.i1662
  %2617 = load atomic i32, ptr %arrayidx.i.i.i22.i.i.i.i1674 monotonic, align 4
  %2618 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %2617)
  %cmp.i.not.i.i23.i.i.i.i1675 = icmp eq i32 %2618, %port.sroa.37.0.i.i1666
  %2619 = call i64 @llvm.amdgcn.if.break.i64(i1 %cmp.i.not.i.i23.i.i.i.i1675, i64 %phi.broken8758)
  %2620 = call i1 @llvm.amdgcn.loop.i64(i64 %2619)
  br i1 %2620, label %Flow8261, label %while.body.i.i21.i.i.i.i1673

Flow8261:                                         ; preds = %while.body.i.i21.i.i.i.i1673
  %.lcssa9246 = phi i64 [ %2619, %while.body.i.i21.i.i.i.i1673 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa9246)
  br label %Flow8262, !amdgpu.uniform !42

Flow8263:                                         ; preds = %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1677, %while.cond.i.i.i.i1665
  %2621 = phi i64 [ %add.i.i.i.i1680, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1677 ], [ poison, %while.cond.i.i.i.i1665 ]
  %2622 = phi i32 [ %conv.i.i26.i.i.i.i1678, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1677 ], [ poison, %while.cond.i.i.i.i1665 ]
  %2623 = phi ptr [ %scevgep7093, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1677 ], [ poison, %while.cond.i.i.i.i1665 ]
  %2624 = phi i64 [ %lsr.iv.next7100, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1677 ], [ poison, %while.cond.i.i.i.i1665 ]
  %2625 = phi i1 [ false, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1677 ], [ true, %while.cond.i.i.i.i1665 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %2610)
  %2626 = call i64 @llvm.amdgcn.if.break.i64(i1 %2625, i64 %phi.broken8761)
  %2627 = call i1 @llvm.amdgcn.loop.i64(i64 %2626)
  br i1 %2627, label %cond.false.i.i.i1687, label %while.cond.i.i.i.i1665

_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i1676: ; preds = %Flow8262
  fence acquire
  %2628 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.i.i.i.i1668)
  %2629 = extractvalue { i1, i64 } %2628, 0
  %2630 = extractvalue { i1, i64 } %2628, 1
  br i1 %2629, label %if.then.i.i.i.i.i.i.i1681, label %Flow8260

if.then.i.i.i.i.i.i.i1681:                        ; preds = %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i1676
  %sub.i.i.i.i.i.i.i1683 = sub nuw i64 %sub.ptr.sub.i.i.i1610, %idx.0.i.i.i.i1667
  %spec.select.i.i.i36.i.i.i.i1684 = call i64 @llvm.umin.i64(i64 %sub.i.i.i.i.i.i.i1683, i64 64)
  %2631 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %2632 = and i64 %spec.select.i.i.i36.i.i.i.i1684, 15
  %2633 = sub i64 %spec.select.i.i.i36.i.i.i.i1684, %2632
  %2634 = icmp ne i64 %2633, 0
  %2635 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %2634)
  %2636 = extractvalue { i1, i64 } %2635, 0
  %2637 = extractvalue { i1, i64 } %2635, 1
  br i1 %2636, label %loop-memcpy-expansion5055.preheader, label %Flow8259

loop-memcpy-expansion5055.preheader:              ; preds = %if.then.i.i.i.i.i.i.i1681
  %arrayidx.i20.i35.i.i.i.i1685 = getelementptr %"struct.rpc::Buffer", ptr %2631, i64 %idxprom.i19.i.i.i.i.i1655
  %arrayidx.i22.i38.i.i.i.i1686 = getelementptr %"struct.rpc::Buffer", ptr %arrayidx.i20.i35.i.i.i.i1685, i64 %sh_prom.i.i.i.i756
  br label %loop-memcpy-expansion5055, !amdgpu.uniform !42

Flow8259:                                         ; preds = %Flow8258, %if.then.i.i.i.i.i.i.i1681
  call void @llvm.amdgcn.end.cf.i64(i64 %2637)
  br label %loop-memcpy-residual-header5058, !amdgpu.uniform !42

loop-memcpy-expansion5055:                        ; preds = %loop-memcpy-expansion5055.preheader, %loop-memcpy-expansion5055
  %phi.broken8759 = phi i64 [ 0, %loop-memcpy-expansion5055.preheader ], [ %2642, %loop-memcpy-expansion5055 ]
  %loop-index5056 = phi i64 [ %2640, %loop-memcpy-expansion5055 ], [ 0, %loop-memcpy-expansion5055.preheader ]
  %scevgep7094 = getelementptr i8, ptr %lsr.iv7092, i64 %loop-index5056
  %2638 = load <4 x i32>, ptr %scevgep7094, align 1
  %2639 = getelementptr i8, ptr %arrayidx.i22.i38.i.i.i.i1686, i64 %loop-index5056
  store <4 x i32> %2638, ptr %2639, align 1
  %2640 = add i64 %loop-index5056, 16
  %2641 = icmp uge i64 %2640, %2633
  %2642 = call i64 @llvm.amdgcn.if.break.i64(i1 %2641, i64 %phi.broken8759)
  %2643 = call i1 @llvm.amdgcn.loop.i64(i64 %2642)
  br i1 %2643, label %Flow8258, label %loop-memcpy-expansion5055

Flow8257:                                         ; preds = %Flow8256, %loop-memcpy-residual-header5058
  call void @llvm.amdgcn.end.cf.i64(i64 %6302)
  br label %Flow8260, !amdgpu.uniform !42

loop-memcpy-residual5057:                         ; preds = %loop-memcpy-residual5057.preheader, %loop-memcpy-residual5057
  %phi.broken8760 = phi i64 [ 0, %loop-memcpy-residual5057.preheader ], [ %2647, %loop-memcpy-residual5057 ]
  %residual-loop-index5059 = phi i64 [ %2645, %loop-memcpy-residual5057 ], [ 0, %loop-memcpy-residual5057.preheader ]
  %scevgep7106 = getelementptr i8, ptr %scevgep7105, i64 %residual-loop-index5059
  %2644 = load i8, ptr %scevgep7106, align 1
  %scevgep7103 = getelementptr i8, ptr %scevgep7102, i64 %residual-loop-index5059
  store i8 %2644, ptr %scevgep7103, align 1
  %2645 = add i64 %residual-loop-index5059, 1
  %2646 = icmp uge i64 %2645, %2632
  %2647 = call i64 @llvm.amdgcn.if.break.i64(i1 %2646, i64 %phi.broken8760)
  %2648 = call i1 @llvm.amdgcn.loop.i64(i64 %2647)
  br i1 %2648, label %Flow8256, label %loop-memcpy-residual5057

Flow8256:                                         ; preds = %loop-memcpy-residual5057
  %.lcssa9248 = phi i64 [ %2647, %loop-memcpy-residual5057 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa9248)
  br label %Flow8257, !amdgpu.uniform !42

_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1677: ; preds = %Flow8260
  %conv.i.i26.i.i.i.i1678 = xor i32 %port.sroa.37.0.i.i1666, 1
  fence release
  %2649 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %arrayidx.i24.i29.i.i.i.i1679 = getelementptr inbounds nuw i32, ptr %2649, i64 %idxprom.i23.i.i.i.i.i1662
  store atomic i32 %conv.i.i26.i.i.i.i1678, ptr %arrayidx.i24.i29.i.i.i.i1679 monotonic, align 4
  %add.i.i.i.i1680 = add i64 %idx.0.i.i.i.i1667, 64
  %scevgep7093 = getelementptr i8, ptr %lsr.iv7092, i64 64
  %lsr.iv.next7100 = add i64 %lsr.iv7099, -64
  br label %Flow8263, !amdgpu.uniform !42

cond.false.i.i.i1687:                             ; preds = %Flow8263
  %.lcssa9250 = phi i64 [ %2626, %Flow8263 ]
  %port.sroa.37.0.i.i1666.lcssa = phi i32 [ %port.sroa.37.0.i.i1666, %Flow8263 ]
  %.lcssa9249 = phi i32 [ %2607, %Flow8263 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa9250)
  %2650 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %.lcssa9249)
  %2651 = icmp ne i32 %2650, %port.sroa.37.0.i.i1666.lcssa
  %2652 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %2651)
  %2653 = extractvalue { i1, i64 } %2652, 0
  %2654 = extractvalue { i1, i64 } %2652, 1
  br i1 %2653, label %while.body.i.i.i.i1688.preheader, label %Flow8255

while.body.i.i.i.i1688.preheader:                 ; preds = %cond.false.i.i.i1687
  br label %while.body.i.i.i.i1688, !amdgpu.uniform !42

Flow8255:                                         ; preds = %Flow8254, %cond.false.i.i.i1687
  call void @llvm.amdgcn.end.cf.i64(i64 %2654)
  br label %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit1709, !amdgpu.uniform !42

while.body.i.i.i.i1688:                           ; preds = %while.body.i.i.i.i1688.preheader, %while.body.i.i.i.i1688
  %phi.broken8762 = phi i64 [ 0, %while.body.i.i.i.i1688.preheader ], [ %2658, %while.body.i.i.i.i1688 ]
  call void @llvm.amdgcn.s.sleep(i32 2)
  %2655 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i8.i.i1689 = getelementptr inbounds nuw i32, ptr %2655, i64 %idxprom.i23.i.i.i.i.i1662
  %2656 = load atomic i32, ptr %arrayidx.i.i.i8.i.i1689 monotonic, align 4
  %2657 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %2656)
  %cmp.i.not.i.i.i.i1690 = icmp eq i32 %2657, %port.sroa.37.0.i.i1666.lcssa
  %2658 = call i64 @llvm.amdgcn.if.break.i64(i1 %cmp.i.not.i.i.i.i1690, i64 %phi.broken8762)
  %2659 = call i1 @llvm.amdgcn.loop.i64(i64 %2658)
  br i1 %2659, label %Flow8254, label %while.body.i.i.i.i1688

Flow8254:                                         ; preds = %while.body.i.i.i.i1688
  %.lcssa9251 = phi i64 [ %2658, %while.body.i.i.i.i1688 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa9251)
  br label %Flow8255, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit1709: ; preds = %Flow8255
  fence acquire
  call void @llvm.amdgcn.wave.barrier()
  fence syncscope("agent") release
  %2660 = call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %.lcssa9242, i1 true)
  %iszero.i.i.i.i.i13.i.i1692 = icmp ne i64 %.lcssa9242, 0
  %cmp2.i.i.i.i14.i.i1693 = icmp eq i64 %2660, %sh_prom.i.i.i.i756
  %cmp.i.i.i.i15.i.i1694 = select i1 %iszero.i.i.i.i.i13.i.i1692, i1 %cmp2.i.i.i.i14.i.i1693, i1 false
  %2661 = and i32 %.lcssa9243, 31
  %conv4.i.i.i19.i.i1698 = zext i1 %cmp.i.i.i.i15.i.i1694 to i32
  %shl.i.i.i20.i.i1699 = shl nuw i32 %conv4.i.i.i19.i.i1698, %2661
  %xor.i.i.i21.i.i1700 = xor i32 %shl.i.i.i20.i.i1699, -1
  %2662 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop5680, !amdgpu.uniform !42

2663:                                             ; preds = %ComputeEnd5681
  %div8.i.i.i16.i.i1695 = lshr i32 %.lcssa9243, 5
  %idxprom.i.i.i17.i.i1696 = zext nneg i32 %div8.i.i.i16.i.i1695 to i64
  %2664 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i17.i.i1696
  %sunkaddr7416 = getelementptr inbounds i8, ptr addrspace(1) %2664, i64 40
  %2665 = atomicrmw and ptr addrspace(1) %sunkaddr7416, i32 %.lcssa9252 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %2666, !amdgpu.uniform !42

2666:                                             ; preds = %2663, %ComputeEnd5681
  call void @llvm.amdgcn.end.cf.i64(i64 %7606)
  call void @llvm.amdgcn.wave.barrier()
  br label %for.cond.i.i.i1713, !amdgpu.uniform !42

for.cond.i.i.i1713:                               ; preds = %2666, %Flow8253
  %2667 = phi i32 [ %2704, %Flow8253 ], [ %1177, %2666 ]
  %2668 = phi i64 [ %2705, %Flow8253 ], [ %1178, %2666 ]
  %2669 = phi i32 [ %2706, %Flow8253 ], [ %1179, %2666 ]
  %retval.sroa.6.0.i.i.i1714 = phi i32 [ %2709, %Flow8253 ], [ undef, %2666 ]
  %retval.sroa.2.0.i.i.i1716 = phi i64 [ %2708, %Flow8253 ], [ undef, %2666 ]
  %retval.sroa.8.0.i.i.i1717 = phi i32 [ %2707, %Flow8253 ], [ undef, %2666 ]
  %index.0.i.i.i1718 = phi i32 [ %2711, %Flow8253 ], [ 0, %2666 ]
  %2670 = load i32, ptr addrspace(1) @__llvm_rpc_client, align 8, !tbaa !68
  %cmp.not.i.i.i1719 = icmp ult i32 %index.0.i.i.i1718, %2670
  %spec.store.select.i.i.i1720 = select i1 %cmp.not.i.i.i1719, i32 %index.0.i.i.i1718, i32 0
  %2671 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  %2672 = and i32 %spec.store.select.i.i.i1720, 31
  %2673 = lshr i64 %2671, %sh_prom.i.i.i.i756
  %2674 = trunc i64 %2673 to i32
  %conv4.i.i.i.i.i1724 = and i32 %2674, 1
  %shl.i.i.i.i.i1725 = shl nuw i32 %conv4.i.i.i.i.i1724, %2672
  br label %ComputeLoop5684, !amdgpu.uniform !42

2675:                                             ; preds = %ComputeEnd5685
  %sunkaddr7417 = getelementptr inbounds i8, ptr addrspace(1) %7616, i64 40
  %2676 = atomicrmw or ptr addrspace(1) %sunkaddr7417, i32 %.lcssa9253 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %2677, !amdgpu.uniform !42

2677:                                             ; preds = %2675, %ComputeEnd5685
  %2678 = phi i32 [ %2676, %2675 ], [ poison, %ComputeEnd5685 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %7625)
  %2679 = call i32 @llvm.amdgcn.readfirstlane.i32(i32 %2678)
  %2680 = or i32 %2679, %.lcssa9254
  %shl5.i.i.i.i.i1726 = shl nuw i32 1, %2672
  %and.i.i.i.i.i1727 = and i32 %shl5.i.i.i.i.i1726, %2680
  %tobool3.i.i.i.i1728 = icmp ne i32 %and.i.i.i.i.i1727, 0
  %2681 = call i64 @llvm.amdgcn.ballot.i64(i1 %tobool3.i.i.i.i1728)
  %and.i.i.i.i.i.i1729 = and i64 %2681, %2671
  %cmp.i.not.i.i.i1730 = icmp ne i64 %2671, %and.i.i.i.i.i.i1729
  br i1 %cmp.i.not.i.i.i1730, label %if.then.i.i.i.i1731, label %Flow8252, !amdgpu.uniform !42

if.then.i.i.i.i1731:                              ; preds = %2677
  fence syncscope("agent") acquire
  %2682 = load <2 x i64>, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !75
  %2683 = extractelement <2 x i64> %2682, i32 0
  %2684 = inttoptr i64 %2683 to ptr
  %2685 = extractelement <2 x i64> %2682, i32 1
  %2686 = inttoptr i64 %2685 to ptr
  %idxprom.i.i.i.i1732 = zext i32 %spec.store.select.i.i.i1720 to i64
  %arrayidx.i.i.i.i1733 = getelementptr inbounds nuw i32, ptr %2684, i64 %idxprom.i.i.i.i1732, !amdgpu.uniform !42
  %2687 = load atomic i32, ptr %arrayidx.i.i.i.i1733 monotonic, align 4
  %2688 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %2687)
  %arrayidx.i45.i.i.i1734 = getelementptr inbounds nuw i32, ptr %2686, i64 %idxprom.i.i.i.i1732, !amdgpu.uniform !42
  %2689 = load atomic i32, ptr %arrayidx.i45.i.i.i1734 monotonic, align 4
  %2690 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %2689)
  %cmp.i46.not.i.i.i1735 = icmp ne i32 %2688, %2690
  br i1 %cmp.i46.not.i.i.i1735, label %if.then12.i.i.i1736, label %Flow8250, !amdgpu.uniform !42

if.then12.i.i.i1736:                              ; preds = %if.then.i.i.i.i1731
  fence syncscope("agent") release
  %2691 = call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %2671, i1 true)
  %iszero.i.i.i.i.i.i.i1737 = icmp ne i64 %2671, 0
  %cmp2.i.i.i.i.i.i1738 = icmp eq i64 %2691, %sh_prom.i.i.i.i756
  %cmp.i.i.i.i.i.i1739 = select i1 %iszero.i.i.i.i.i.i.i1737, i1 %cmp2.i.i.i.i.i.i1738, i1 false
  %conv4.i.i51.i.i.i1740 = zext i1 %cmp.i.i.i.i.i.i1739 to i32
  %shl.i.i52.i.i.i1741 = shl nuw i32 %conv4.i.i51.i.i.i1740, %2672
  %xor.i.i.i.i.i1742 = xor i32 %shl.i.i52.i.i.i1741, -1
  %2692 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop5689, !amdgpu.uniform !42

2693:                                             ; preds = %ComputeEnd5690
  %sunkaddr7418 = getelementptr inbounds i8, ptr addrspace(1) %7616, i64 40
  %2694 = atomicrmw and ptr addrspace(1) %sunkaddr7418, i32 %.lcssa9255 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %2695, !amdgpu.uniform !42

2695:                                             ; preds = %2693, %ComputeEnd5690
  call void @llvm.amdgcn.end.cf.i64(i64 %7643)
  call void @llvm.amdgcn.wave.barrier()
  br label %Flow8250, !amdgpu.uniform !42

if.end14.i.i.i1800:                               ; preds = %Flow8250
  %2696 = call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %2671, i1 true)
  %iszero.i.i.i.i.i.i1801 = icmp ne i64 %2671, 0
  %cmp2.i.i.i.i.i1802 = icmp eq i64 %2696, %sh_prom.i.i.i.i756
  %cmp.i.i.i.i.i1803 = select i1 %iszero.i.i.i.i.i.i1801, i1 %cmp2.i.i.i.i.i1802, i1 false
  %2697 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.i.i.i.i.i1803)
  %2698 = extractvalue { i1, i64 } %2697, 0
  %2699 = extractvalue { i1, i64 } %2697, 1
  br i1 %2698, label %if.then16.i.i.i1805, label %if.end22.i.i.i1804

if.then16.i.i.i1805:                              ; preds = %if.end14.i.i.i1800
  %2700 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !76
  %opcode.i.i.i1806 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %2700, i64 %idxprom.i.i.i.i1732, i32 1
  store i32 1660944387, ptr %opcode.i.i.i1806, align 8, !tbaa !77
  %arrayidx21.i.i.i1807 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %2700, i64 %idxprom.i.i.i.i1732
  store i64 %2671, ptr %arrayidx21.i.i.i1807, align 8, !tbaa !80
  br label %if.end22.i.i.i1804, !amdgpu.uniform !42

Flow8251:                                         ; preds = %if.end22.i.i.i1804, %Flow8250
  %2701 = phi i32 [ %2688, %if.end22.i.i.i1804 ], [ %retval.sroa.8.0.i.i.i1717, %Flow8250 ]
  %2702 = phi i64 [ %2671, %if.end22.i.i.i1804 ], [ %retval.sroa.2.0.i.i.i1716, %Flow8250 ]
  %2703 = phi i32 [ %spec.store.select.i.i.i1720, %if.end22.i.i.i1804 ], [ %retval.sroa.6.0.i.i.i1714, %Flow8250 ]
  br label %cleanup26.i.i.i1743, !amdgpu.uniform !42

if.end22.i.i.i1804:                               ; preds = %if.then16.i.i.i1805, %if.end14.i.i.i1800
  call void @llvm.amdgcn.end.cf.i64(i64 %2699)
  call void @llvm.amdgcn.wave.barrier()
  br label %Flow8251, !amdgpu.uniform !42

Flow8252:                                         ; preds = %cleanup26.i.i.i1743, %2677
  %2704 = phi i32 [ %2703, %cleanup26.i.i.i1743 ], [ %2667, %2677 ]
  %2705 = phi i64 [ %2702, %cleanup26.i.i.i1743 ], [ %2668, %2677 ]
  %2706 = phi i32 [ %2701, %cleanup26.i.i.i1743 ], [ %2669, %2677 ]
  %2707 = phi i32 [ %2701, %cleanup26.i.i.i1743 ], [ %retval.sroa.8.0.i.i.i1717, %2677 ]
  %2708 = phi i64 [ %2702, %cleanup26.i.i.i1743 ], [ %retval.sroa.2.0.i.i.i1716, %2677 ]
  %2709 = phi i32 [ %2703, %cleanup26.i.i.i1743 ], [ %retval.sroa.6.0.i.i.i1714, %2677 ]
  %2710 = phi i1 [ %cmp.i46.not.i.i.i1735, %cleanup26.i.i.i1743 ], [ true, %2677 ]
  br i1 %2710, label %for.inc.i.i.i1748, label %Flow8253, !amdgpu.uniform !42

cleanup26.i.i.i1743:                              ; preds = %Flow8251
  br label %Flow8252, !amdgpu.uniform !42

for.inc.i.i.i1748:                                ; preds = %Flow8252
  %inc.i.i.i1753 = add i32 %spec.store.select.i.i.i1720, 1
  br label %Flow8253, !amdgpu.uniform !42

Flow8253:                                         ; preds = %for.inc.i.i.i1748, %Flow8252
  %2711 = phi i32 [ %inc.i.i.i1753, %for.inc.i.i.i1748 ], [ poison, %Flow8252 ]
  %2712 = phi i1 [ false, %for.inc.i.i.i1748 ], [ true, %Flow8252 ]
  br i1 %2712, label %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1754, label %for.cond.i.i.i1713, !amdgpu.uniform !42

_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1754: ; preds = %Flow8253
  %.lcssa9258 = phi i32 [ %2704, %Flow8253 ]
  %.lcssa9257 = phi i64 [ %2705, %Flow8253 ]
  %.lcssa9256 = phi i32 [ %2706, %Flow8253 ]
  fence acquire
  %2713 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %mul.i.i.i.i.i.i1755 = shl i32 %.lcssa9258, 6
  %idxprom.i19.i.i.i.i.i1756 = zext i32 %mul.i.i.i.i.i.i1755 to i64
  %arrayidx.i20.i.i.i.i.i1757 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %2713, i64 %idxprom.i19.i.i.i.i.i1756
  %arrayidx.i22.i.i.i.i.i1758 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %arrayidx.i20.i.i.i.i.i1757, i64 %sh_prom.i.i.i.i756
  store i64 1, ptr %arrayidx.i22.i.i.i.i.i1758, align 8, !tbaa !82
  %arrayidx6.i.i.i.i.i.i.i1759 = getelementptr inbounds nuw i8, ptr %arrayidx.i22.i.i.i.i.i1758, i64 8
  store i8 10, ptr %arrayidx6.i.i.i.i.i.i.i1759, align 8
  %tobool.not.i.i.i.i.i.i1760 = icmp eq i32 %.lcssa9256, 0
  %conv.i.i.i.i.i.i1761 = zext i1 %tobool.not.i.i.i.i.i.i1760 to i32
  fence release
  %2714 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %idxprom.i23.i.i.i.i.i1762 = zext i32 %.lcssa9258 to i64
  %arrayidx.i24.i.i.i.i.i1763 = getelementptr inbounds nuw i32, ptr %2714, i64 %idxprom.i23.i.i.i.i.i1762
  store atomic i32 %conv.i.i.i.i.i.i1761, ptr %arrayidx.i24.i.i.i.i.i1763 monotonic, align 4
  %2715 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !84
  %arrayidx.i.i5.i.i1764 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %2715, i64 %idxprom.i23.i.i.i.i.i1762
  %2716 = load i64, ptr %arrayidx.i.i5.i.i1764, align 8, !tbaa !80
  br label %while.cond.i.i.i.i1765, !amdgpu.uniform !42

while.cond.i.i.i.i1765:                           ; preds = %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1754, %Flow8249
  %phi.broken8764 = phi i64 [ 0, %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1754 ], [ %2736, %Flow8249 ]
  %port.sroa.37.0.i.i1766 = phi i32 [ %2734, %Flow8249 ], [ %conv.i.i.i.i.i.i1761, %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1754 ]
  %idx.0.i.i.i.i1767 = phi i64 [ %2733, %Flow8249 ], [ 56, %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1754 ]
  %cmp.i.i.i.i1768 = icmp eq i64 %idx.0.i.i.i.i1767, 0
  %2717 = call i64 @llvm.amdgcn.ballot.i64(i1 %cmp.i.i.i.i1768)
  %and.i.i.i.i6.i.i1769 = and i64 %2717, %2716
  %tobool.not.i.i.i.i1770 = icmp ne i64 %and.i.i.i.i6.i.i1769, 0
  %2718 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i22.i.i.i1771 = getelementptr inbounds nuw i32, ptr %2718, i64 %idxprom.i23.i.i.i.i.i1762
  %2719 = load atomic i32, ptr %arrayidx.i22.i.i.i1771 monotonic, align 4
  %2720 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %tobool.not.i.i.i.i1770)
  %2721 = extractvalue { i1, i64 } %2720, 0
  %2722 = extractvalue { i1, i64 } %2720, 1
  br i1 %2721, label %cond.false.i8.i.i.i.i1772, label %Flow8249

cond.false.i8.i.i.i.i1772:                        ; preds = %while.cond.i.i.i.i1765
  %2723 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %2719)
  %2724 = icmp ne i32 %2723, %port.sroa.37.0.i.i1766
  %2725 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %2724)
  %2726 = extractvalue { i1, i64 } %2725, 0
  %2727 = extractvalue { i1, i64 } %2725, 1
  br i1 %2726, label %while.body.i.i21.i.i.i.i1773.preheader, label %Flow8248

while.body.i.i21.i.i.i.i1773.preheader:           ; preds = %cond.false.i8.i.i.i.i1772
  br label %while.body.i.i21.i.i.i.i1773, !amdgpu.uniform !42

Flow8248:                                         ; preds = %Flow8247, %cond.false.i8.i.i.i.i1772
  call void @llvm.amdgcn.end.cf.i64(i64 %2727)
  br label %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i1776, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i1773:                     ; preds = %while.body.i.i21.i.i.i.i1773.preheader, %while.body.i.i21.i.i.i.i1773
  %phi.broken8763 = phi i64 [ 0, %while.body.i.i21.i.i.i.i1773.preheader ], [ %2731, %while.body.i.i21.i.i.i.i1773 ]
  call void @llvm.amdgcn.s.sleep(i32 2)
  %2728 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i22.i.i.i.i1774 = getelementptr inbounds nuw i32, ptr %2728, i64 %idxprom.i23.i.i.i.i.i1762
  %2729 = load atomic i32, ptr %arrayidx.i.i.i22.i.i.i.i1774 monotonic, align 4
  %2730 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %2729)
  %cmp.i.not.i.i23.i.i.i.i1775 = icmp eq i32 %2730, %port.sroa.37.0.i.i1766
  %2731 = call i64 @llvm.amdgcn.if.break.i64(i1 %cmp.i.not.i.i23.i.i.i.i1775, i64 %phi.broken8763)
  %2732 = call i1 @llvm.amdgcn.loop.i64(i64 %2731)
  br i1 %2732, label %Flow8247, label %while.body.i.i21.i.i.i.i1773

Flow8247:                                         ; preds = %while.body.i.i21.i.i.i.i1773
  %.lcssa9259 = phi i64 [ %2731, %while.body.i.i21.i.i.i.i1773 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa9259)
  br label %Flow8248, !amdgpu.uniform !42

Flow8249:                                         ; preds = %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1777, %while.cond.i.i.i.i1765
  %2733 = phi i64 [ %add.i.i.i.i1780, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1777 ], [ poison, %while.cond.i.i.i.i1765 ]
  %2734 = phi i32 [ %conv.i.i26.i.i.i.i1778, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1777 ], [ poison, %while.cond.i.i.i.i1765 ]
  %2735 = phi i1 [ false, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1777 ], [ true, %while.cond.i.i.i.i1765 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %2722)
  %2736 = call i64 @llvm.amdgcn.if.break.i64(i1 %2735, i64 %phi.broken8764)
  %2737 = call i1 @llvm.amdgcn.loop.i64(i64 %2736)
  br i1 %2737, label %cond.false.i.i.i1786, label %while.cond.i.i.i.i1765

_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i1776: ; preds = %Flow8248
  fence acquire
  br i1 %cmp.i.i.i.i1768, label %if.then.i.i.i.i.i.i.i1781, label %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1777, !amdgpu.uniform !42

if.then.i.i.i.i.i.i.i1781:                        ; preds = %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i1776
  %2738 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %arrayidx.i20.i35.i.i.i.i1784 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %2738, i64 %idxprom.i19.i.i.i.i.i1756
  %arrayidx.i22.i38.i.i.i.i1785 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %arrayidx.i20.i35.i.i.i.i1784, i64 %sh_prom.i.i.i.i756
  store i8 10, ptr %arrayidx.i22.i38.i.i.i.i1785, align 1
  br label %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1777, !amdgpu.uniform !42

_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1777: ; preds = %if.then.i.i.i.i.i.i.i1781, %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i1776
  %conv.i.i26.i.i.i.i1778 = xor i32 %port.sroa.37.0.i.i1766, 1
  fence release
  %2739 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %arrayidx.i24.i29.i.i.i.i1779 = getelementptr inbounds nuw i32, ptr %2739, i64 %idxprom.i23.i.i.i.i.i1762
  store atomic i32 %conv.i.i26.i.i.i.i1778, ptr %arrayidx.i24.i29.i.i.i.i1779 monotonic, align 4
  %add.i.i.i.i1780 = add i64 %idx.0.i.i.i.i1767, 64
  br label %Flow8249, !amdgpu.uniform !42

cond.false.i.i.i1786:                             ; preds = %Flow8249
  %.lcssa9261 = phi i64 [ %2736, %Flow8249 ]
  %port.sroa.37.0.i.i1766.lcssa = phi i32 [ %port.sroa.37.0.i.i1766, %Flow8249 ]
  %.lcssa9260 = phi i32 [ %2719, %Flow8249 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa9261)
  %2740 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %.lcssa9260)
  %2741 = icmp ne i32 %2740, %port.sroa.37.0.i.i1766.lcssa
  %2742 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %2741)
  %2743 = extractvalue { i1, i64 } %2742, 0
  %2744 = extractvalue { i1, i64 } %2742, 1
  br i1 %2743, label %while.body.i.i.i.i1787.preheader, label %Flow8246

while.body.i.i.i.i1787.preheader:                 ; preds = %cond.false.i.i.i1786
  br label %while.body.i.i.i.i1787, !amdgpu.uniform !42

Flow8246:                                         ; preds = %Flow8245, %cond.false.i.i.i1786
  call void @llvm.amdgcn.end.cf.i64(i64 %2744)
  br label %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIcEERS1_T_.exit1808, !amdgpu.uniform !42

while.body.i.i.i.i1787:                           ; preds = %while.body.i.i.i.i1787.preheader, %while.body.i.i.i.i1787
  %phi.broken8765 = phi i64 [ 0, %while.body.i.i.i.i1787.preheader ], [ %2748, %while.body.i.i.i.i1787 ]
  call void @llvm.amdgcn.s.sleep(i32 2)
  %2745 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i8.i.i1788 = getelementptr inbounds nuw i32, ptr %2745, i64 %idxprom.i23.i.i.i.i.i1762
  %2746 = load atomic i32, ptr %arrayidx.i.i.i8.i.i1788 monotonic, align 4
  %2747 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %2746)
  %cmp.i.not.i.i.i.i1789 = icmp eq i32 %2747, %port.sroa.37.0.i.i1766.lcssa
  %2748 = call i64 @llvm.amdgcn.if.break.i64(i1 %cmp.i.not.i.i.i.i1789, i64 %phi.broken8765)
  %2749 = call i1 @llvm.amdgcn.loop.i64(i64 %2748)
  br i1 %2749, label %Flow8245, label %while.body.i.i.i.i1787

Flow8245:                                         ; preds = %while.body.i.i.i.i1787
  %.lcssa9262 = phi i64 [ %2748, %while.body.i.i.i.i1787 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa9262)
  br label %Flow8246, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIcEERS1_T_.exit1808: ; preds = %Flow8246
  fence acquire
  call void @llvm.amdgcn.wave.barrier()
  fence syncscope("agent") release
  %2750 = call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %.lcssa9257, i1 true)
  %iszero.i.i.i.i.i13.i.i1791 = icmp ne i64 %.lcssa9257, 0
  %cmp2.i.i.i.i14.i.i1792 = icmp eq i64 %2750, %sh_prom.i.i.i.i756
  %cmp.i.i.i.i15.i.i1793 = select i1 %iszero.i.i.i.i.i13.i.i1791, i1 %cmp2.i.i.i.i14.i.i1792, i1 false
  %2751 = and i32 %.lcssa9258, 31
  %conv4.i.i.i19.i.i1797 = zext i1 %cmp.i.i.i.i15.i.i1793 to i32
  %shl.i.i.i20.i.i1798 = shl nuw i32 %conv4.i.i.i19.i.i1797, %2751
  %xor.i.i.i21.i.i1799 = xor i32 %shl.i.i.i20.i.i1798, -1
  %2752 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop5693, !amdgpu.uniform !42

2753:                                             ; preds = %ComputeEnd5694
  %div8.i.i.i16.i.i1794 = lshr i32 %.lcssa9258, 5
  %idxprom.i.i.i17.i.i1795 = zext nneg i32 %div8.i.i.i16.i.i1794 to i64
  %2754 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i17.i.i1795
  %sunkaddr7419 = getelementptr inbounds i8, ptr addrspace(1) %2754, i64 40
  %2755 = atomicrmw and ptr addrspace(1) %sunkaddr7419, i32 %.lcssa9263 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %2756, !amdgpu.uniform !42

2756:                                             ; preds = %2753, %ComputeEnd5694
  call void @llvm.amdgcn.end.cf.i64(i64 %7660)
  call void @llvm.amdgcn.wave.barrier()
  %inc = add nsw i32 %FailCount.0205, 1
  br label %Flow8331, !amdgpu.uniform !42

Flow8330:                                         ; preds = %Flow8149, %_ZN22__llvm_libc_22_0_0_git5clockEv.exit1282
  %2757 = phi i32 [ %3452, %Flow8149 ], [ poison, %_ZN22__llvm_libc_22_0_0_git5clockEv.exit1282 ]
  %2758 = phi i64 [ %3453, %Flow8149 ], [ poison, %_ZN22__llvm_libc_22_0_0_git5clockEv.exit1282 ]
  %2759 = phi i32 [ %3454, %Flow8149 ], [ poison, %_ZN22__llvm_libc_22_0_0_git5clockEv.exit1282 ]
  %2760 = phi i32 [ %3455, %Flow8149 ], [ poison, %_ZN22__llvm_libc_22_0_0_git5clockEv.exit1282 ]
  %2761 = phi i64 [ %3456, %Flow8149 ], [ poison, %_ZN22__llvm_libc_22_0_0_git5clockEv.exit1282 ]
  %2762 = phi i32 [ %3457, %Flow8149 ], [ poison, %_ZN22__llvm_libc_22_0_0_git5clockEv.exit1282 ]
  %2763 = phi i32 [ %3458, %Flow8149 ], [ poison, %_ZN22__llvm_libc_22_0_0_git5clockEv.exit1282 ]
  %2764 = phi i64 [ %3459, %Flow8149 ], [ poison, %_ZN22__llvm_libc_22_0_0_git5clockEv.exit1282 ]
  %2765 = phi i32 [ %3460, %Flow8149 ], [ poison, %_ZN22__llvm_libc_22_0_0_git5clockEv.exit1282 ]
  %2766 = phi i32 [ %3461, %Flow8149 ], [ poison, %_ZN22__llvm_libc_22_0_0_git5clockEv.exit1282 ]
  %2767 = phi i64 [ %3462, %Flow8149 ], [ poison, %_ZN22__llvm_libc_22_0_0_git5clockEv.exit1282 ]
  %2768 = phi i32 [ %3463, %Flow8149 ], [ poison, %_ZN22__llvm_libc_22_0_0_git5clockEv.exit1282 ]
  %2769 = phi i32 [ %3464, %Flow8149 ], [ poison, %_ZN22__llvm_libc_22_0_0_git5clockEv.exit1282 ]
  %2770 = phi i64 [ %3465, %Flow8149 ], [ poison, %_ZN22__llvm_libc_22_0_0_git5clockEv.exit1282 ]
  %2771 = phi i32 [ %3466, %Flow8149 ], [ poison, %_ZN22__llvm_libc_22_0_0_git5clockEv.exit1282 ]
  %2772 = phi i32 [ %3467, %Flow8149 ], [ poison, %_ZN22__llvm_libc_22_0_0_git5clockEv.exit1282 ]
  %2773 = phi i64 [ %3468, %Flow8149 ], [ poison, %_ZN22__llvm_libc_22_0_0_git5clockEv.exit1282 ]
  %2774 = phi i32 [ %3469, %Flow8149 ], [ poison, %_ZN22__llvm_libc_22_0_0_git5clockEv.exit1282 ]
  %2775 = phi i32 [ %3470, %Flow8149 ], [ poison, %_ZN22__llvm_libc_22_0_0_git5clockEv.exit1282 ]
  %2776 = phi i64 [ %3471, %Flow8149 ], [ poison, %_ZN22__llvm_libc_22_0_0_git5clockEv.exit1282 ]
  %2777 = phi i32 [ %3472, %Flow8149 ], [ poison, %_ZN22__llvm_libc_22_0_0_git5clockEv.exit1282 ]
  %2778 = phi i32 [ %.lcssa9109, %Flow8149 ], [ poison, %_ZN22__llvm_libc_22_0_0_git5clockEv.exit1282 ]
  %2779 = phi i64 [ %.lcssa9110, %Flow8149 ], [ poison, %_ZN22__llvm_libc_22_0_0_git5clockEv.exit1282 ]
  %2780 = phi i32 [ %.lcssa9111, %Flow8149 ], [ poison, %_ZN22__llvm_libc_22_0_0_git5clockEv.exit1282 ]
  %2781 = phi i32 [ %.lcssa9094, %Flow8149 ], [ poison, %_ZN22__llvm_libc_22_0_0_git5clockEv.exit1282 ]
  %2782 = phi i64 [ %.lcssa9095, %Flow8149 ], [ poison, %_ZN22__llvm_libc_22_0_0_git5clockEv.exit1282 ]
  %2783 = phi i32 [ %.lcssa9096, %Flow8149 ], [ poison, %_ZN22__llvm_libc_22_0_0_git5clockEv.exit1282 ]
  %2784 = phi i32 [ %.lcssa9078, %Flow8149 ], [ poison, %_ZN22__llvm_libc_22_0_0_git5clockEv.exit1282 ]
  %2785 = phi i64 [ %.lcssa9079, %Flow8149 ], [ poison, %_ZN22__llvm_libc_22_0_0_git5clockEv.exit1282 ]
  %2786 = phi i32 [ %.lcssa9080, %Flow8149 ], [ poison, %_ZN22__llvm_libc_22_0_0_git5clockEv.exit1282 ]
  %2787 = phi i32 [ %.lcssa9066, %Flow8149 ], [ poison, %_ZN22__llvm_libc_22_0_0_git5clockEv.exit1282 ]
  %2788 = phi i64 [ %.lcssa9067, %Flow8149 ], [ poison, %_ZN22__llvm_libc_22_0_0_git5clockEv.exit1282 ]
  %2789 = phi i32 [ %.lcssa9068, %Flow8149 ], [ poison, %_ZN22__llvm_libc_22_0_0_git5clockEv.exit1282 ]
  %2790 = phi i32 [ %.lcssa9051, %Flow8149 ], [ poison, %_ZN22__llvm_libc_22_0_0_git5clockEv.exit1282 ]
  %2791 = phi i64 [ %.lcssa9052, %Flow8149 ], [ poison, %_ZN22__llvm_libc_22_0_0_git5clockEv.exit1282 ]
  %2792 = phi i32 [ %.lcssa9053, %Flow8149 ], [ poison, %_ZN22__llvm_libc_22_0_0_git5clockEv.exit1282 ]
  %2793 = call { i1, i64 } @llvm.amdgcn.else.i64.i64(i64 %2088)
  %2794 = extractvalue { i1, i64 } %2793, 0
  %2795 = extractvalue { i1, i64 } %2793, 1
  br i1 %2794, label %for.cond.i.i.i1283.preheader, label %Flow8331

for.cond.i.i.i1809:                               ; preds = %for.cond.i.i.i1809.preheader, %for.cond.i.i.i1809
  %phi.broken8685 = phi i64 [ 0, %for.cond.i.i.i1809.preheader ], [ %2797, %for.cond.i.i.i1809 ]
  %lsr.iv6695 = phi ptr addrspace(4) [ %scevgep6697, %for.cond.i.i.i1809 ], [ %scevgep6586, %for.cond.i.i.i1809.preheader ]
  %scevgep6697 = getelementptr i8, ptr addrspace(4) %lsr.iv6695, i64 1
  %2796 = load i8, ptr addrspace(4) %scevgep6697, align 1, !tbaa !57
  %cmp.i.i.i1811 = icmp eq i8 %2796, 0
  %2797 = call i64 @llvm.amdgcn.if.break.i64(i1 %cmp.i.i.i1811, i64 %phi.broken8685)
  %2798 = call i1 @llvm.amdgcn.loop.i64(i64 %2797)
  br i1 %2798, label %for.cond.i.i1.i1818.preheader, label %for.cond.i.i.i1809

for.cond.i.i1.i1818.preheader:                    ; preds = %for.cond.i.i.i1809
  %scevgep6697.lcssa = phi ptr addrspace(4) [ %scevgep6697, %for.cond.i.i.i1809 ]
  %.lcssa9047 = phi i64 [ %2797, %for.cond.i.i.i1809 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa9047)
  br label %for.cond.i.i1.i1818, !amdgpu.uniform !42

for.cond.i.i1.i1818:                              ; preds = %for.cond.i.i1.i1818.preheader, %Flow8244
  %2799 = phi i32 [ %2836, %Flow8244 ], [ %1180, %for.cond.i.i1.i1818.preheader ]
  %2800 = phi i64 [ %2837, %Flow8244 ], [ %1181, %for.cond.i.i1.i1818.preheader ]
  %2801 = phi i32 [ %2838, %Flow8244 ], [ %1182, %for.cond.i.i1.i1818.preheader ]
  %retval.sroa.6.0.i.i.i1819 = phi i32 [ %2841, %Flow8244 ], [ undef, %for.cond.i.i1.i1818.preheader ]
  %retval.sroa.2.0.i.i.i1821 = phi i64 [ %2840, %Flow8244 ], [ undef, %for.cond.i.i1.i1818.preheader ]
  %retval.sroa.8.0.i.i.i1822 = phi i32 [ %2839, %Flow8244 ], [ undef, %for.cond.i.i1.i1818.preheader ]
  %index.0.i.i.i1823 = phi i32 [ %2843, %Flow8244 ], [ 0, %for.cond.i.i1.i1818.preheader ]
  %2802 = load i32, ptr addrspace(1) @__llvm_rpc_client, align 8, !tbaa !68
  %cmp.not.i.i.i1824 = icmp ult i32 %index.0.i.i.i1823, %2802
  %spec.store.select.i.i.i1825 = select i1 %cmp.not.i.i.i1824, i32 %index.0.i.i.i1823, i32 0
  %2803 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  %2804 = and i32 %spec.store.select.i.i.i1825, 31
  %2805 = lshr i64 %2803, %sh_prom.i.i.i.i756
  %2806 = trunc i64 %2805 to i32
  %conv4.i.i.i.i.i1829 = and i32 %2806, 1
  %shl.i.i.i.i.i1830 = shl nuw i32 %conv4.i.i.i.i.i1829, %2804
  br label %ComputeLoop5697, !amdgpu.uniform !42

2807:                                             ; preds = %ComputeEnd5698
  %sunkaddr7420 = getelementptr inbounds i8, ptr addrspace(1) %7670, i64 40
  %2808 = atomicrmw or ptr addrspace(1) %sunkaddr7420, i32 %.lcssa9048 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %2809, !amdgpu.uniform !42

2809:                                             ; preds = %2807, %ComputeEnd5698
  %2810 = phi i32 [ %2808, %2807 ], [ poison, %ComputeEnd5698 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %7679)
  %2811 = call i32 @llvm.amdgcn.readfirstlane.i32(i32 %2810)
  %2812 = or i32 %2811, %.lcssa9049
  %shl5.i.i.i.i.i1831 = shl nuw i32 1, %2804
  %and.i.i.i.i.i1832 = and i32 %shl5.i.i.i.i.i1831, %2812
  %tobool3.i.i.i.i1833 = icmp ne i32 %and.i.i.i.i.i1832, 0
  %2813 = call i64 @llvm.amdgcn.ballot.i64(i1 %tobool3.i.i.i.i1833)
  %and.i.i.i.i.i.i1834 = and i64 %2813, %2803
  %cmp.i.not.i.i.i1835 = icmp ne i64 %2803, %and.i.i.i.i.i.i1834
  br i1 %cmp.i.not.i.i.i1835, label %if.then.i.i.i.i1836, label %Flow8243, !amdgpu.uniform !42

if.then.i.i.i.i1836:                              ; preds = %2809
  fence syncscope("agent") acquire
  %2814 = load <2 x i64>, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !75
  %2815 = extractelement <2 x i64> %2814, i32 0
  %2816 = inttoptr i64 %2815 to ptr
  %2817 = extractelement <2 x i64> %2814, i32 1
  %2818 = inttoptr i64 %2817 to ptr
  %idxprom.i.i.i.i1837 = zext i32 %spec.store.select.i.i.i1825 to i64
  %arrayidx.i.i.i.i1838 = getelementptr inbounds nuw i32, ptr %2816, i64 %idxprom.i.i.i.i1837, !amdgpu.uniform !42
  %2819 = load atomic i32, ptr %arrayidx.i.i.i.i1838 monotonic, align 4
  %2820 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %2819)
  %arrayidx.i45.i.i.i1839 = getelementptr inbounds nuw i32, ptr %2818, i64 %idxprom.i.i.i.i1837, !amdgpu.uniform !42
  %2821 = load atomic i32, ptr %arrayidx.i45.i.i.i1839 monotonic, align 4
  %2822 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %2821)
  %cmp.i46.not.i.i.i1840 = icmp ne i32 %2820, %2822
  br i1 %cmp.i46.not.i.i.i1840, label %if.then12.i.i.i1841, label %Flow8241, !amdgpu.uniform !42

if.then12.i.i.i1841:                              ; preds = %if.then.i.i.i.i1836
  fence syncscope("agent") release
  %2823 = call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %2803, i1 true)
  %iszero.i.i.i.i.i.i.i1842 = icmp ne i64 %2803, 0
  %cmp2.i.i.i.i.i.i1843 = icmp eq i64 %2823, %sh_prom.i.i.i.i756
  %cmp.i.i.i.i.i.i1844 = select i1 %iszero.i.i.i.i.i.i.i1842, i1 %cmp2.i.i.i.i.i.i1843, i1 false
  %conv4.i.i51.i.i.i1845 = zext i1 %cmp.i.i.i.i.i.i1844 to i32
  %shl.i.i52.i.i.i1846 = shl nuw i32 %conv4.i.i51.i.i.i1845, %2804
  %xor.i.i.i.i.i1847 = xor i32 %shl.i.i52.i.i.i1846, -1
  %2824 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop5702, !amdgpu.uniform !42

2825:                                             ; preds = %ComputeEnd5703
  %sunkaddr7421 = getelementptr inbounds i8, ptr addrspace(1) %7670, i64 40
  %2826 = atomicrmw and ptr addrspace(1) %sunkaddr7421, i32 %.lcssa9050 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %2827, !amdgpu.uniform !42

2827:                                             ; preds = %2825, %ComputeEnd5703
  call void @llvm.amdgcn.end.cf.i64(i64 %7697)
  call void @llvm.amdgcn.wave.barrier()
  br label %Flow8241, !amdgpu.uniform !42

if.end14.i.i.i1907:                               ; preds = %Flow8241
  %2828 = call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %2803, i1 true)
  %iszero.i.i.i.i.i.i1908 = icmp ne i64 %2803, 0
  %cmp2.i.i.i.i.i1909 = icmp eq i64 %2828, %sh_prom.i.i.i.i756
  %cmp.i.i.i.i.i1910 = select i1 %iszero.i.i.i.i.i.i1908, i1 %cmp2.i.i.i.i.i1909, i1 false
  %2829 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.i.i.i.i.i1910)
  %2830 = extractvalue { i1, i64 } %2829, 0
  %2831 = extractvalue { i1, i64 } %2829, 1
  br i1 %2830, label %if.then16.i.i.i1912, label %if.end22.i.i.i1911

if.then16.i.i.i1912:                              ; preds = %if.end14.i.i.i1907
  %2832 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !76
  %opcode.i.i.i1913 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %2832, i64 %idxprom.i.i.i.i1837, i32 1
  store i32 1660944387, ptr %opcode.i.i.i1913, align 8, !tbaa !77
  %arrayidx21.i.i.i1914 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %2832, i64 %idxprom.i.i.i.i1837
  store i64 %2803, ptr %arrayidx21.i.i.i1914, align 8, !tbaa !80
  br label %if.end22.i.i.i1911, !amdgpu.uniform !42

Flow8242:                                         ; preds = %if.end22.i.i.i1911, %Flow8241
  %2833 = phi i32 [ %2820, %if.end22.i.i.i1911 ], [ %retval.sroa.8.0.i.i.i1822, %Flow8241 ]
  %2834 = phi i64 [ %2803, %if.end22.i.i.i1911 ], [ %retval.sroa.2.0.i.i.i1821, %Flow8241 ]
  %2835 = phi i32 [ %spec.store.select.i.i.i1825, %if.end22.i.i.i1911 ], [ %retval.sroa.6.0.i.i.i1819, %Flow8241 ]
  br label %cleanup26.i.i.i1848, !amdgpu.uniform !42

if.end22.i.i.i1911:                               ; preds = %if.then16.i.i.i1912, %if.end14.i.i.i1907
  call void @llvm.amdgcn.end.cf.i64(i64 %2831)
  call void @llvm.amdgcn.wave.barrier()
  br label %Flow8242, !amdgpu.uniform !42

Flow8243:                                         ; preds = %cleanup26.i.i.i1848, %2809
  %2836 = phi i32 [ %2835, %cleanup26.i.i.i1848 ], [ %2799, %2809 ]
  %2837 = phi i64 [ %2834, %cleanup26.i.i.i1848 ], [ %2800, %2809 ]
  %2838 = phi i32 [ %2833, %cleanup26.i.i.i1848 ], [ %2801, %2809 ]
  %2839 = phi i32 [ %2833, %cleanup26.i.i.i1848 ], [ %retval.sroa.8.0.i.i.i1822, %2809 ]
  %2840 = phi i64 [ %2834, %cleanup26.i.i.i1848 ], [ %retval.sroa.2.0.i.i.i1821, %2809 ]
  %2841 = phi i32 [ %2835, %cleanup26.i.i.i1848 ], [ %retval.sroa.6.0.i.i.i1819, %2809 ]
  %2842 = phi i1 [ %cmp.i46.not.i.i.i1840, %cleanup26.i.i.i1848 ], [ true, %2809 ]
  br i1 %2842, label %for.inc.i.i.i1853, label %Flow8244, !amdgpu.uniform !42

cleanup26.i.i.i1848:                              ; preds = %Flow8242
  br label %Flow8243, !amdgpu.uniform !42

for.inc.i.i.i1853:                                ; preds = %Flow8243
  %inc.i.i.i1858 = add i32 %spec.store.select.i.i.i1825, 1
  br label %Flow8244, !amdgpu.uniform !42

Flow8244:                                         ; preds = %for.inc.i.i.i1853, %Flow8243
  %2843 = phi i32 [ %inc.i.i.i1858, %for.inc.i.i.i1853 ], [ poison, %Flow8243 ]
  %2844 = phi i1 [ false, %for.inc.i.i.i1853 ], [ true, %Flow8243 ]
  br i1 %2844, label %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1859, label %for.cond.i.i1.i1818, !amdgpu.uniform !42

_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1859: ; preds = %Flow8244
  %.lcssa9053 = phi i32 [ %2836, %Flow8244 ]
  %.lcssa9052 = phi i64 [ %2837, %Flow8244 ]
  %.lcssa9051 = phi i32 [ %2838, %Flow8244 ]
  %2845 = addrspacecast ptr addrspace(4) %scevgep6697.lcssa to ptr
  %2846 = addrspacecast ptr addrspace(4) %cond to ptr
  %2847 = ptrtoint ptr %2846 to i64
  %sub.ptr.lhs.cast.i.i.i1814 = ptrtoint ptr %2845 to i64
  %sub.ptr.sub.i.i.i1816 = sub i64 %sub.ptr.lhs.cast.i.i.i1814, %2847
  fence acquire
  %2848 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %mul.i.i.i.i.i.i1860 = shl i32 %.lcssa9053, 6
  %idxprom.i19.i.i.i.i.i1861 = zext i32 %mul.i.i.i.i.i.i1860 to i64
  %arrayidx.i20.i.i.i.i.i1862 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %2848, i64 %idxprom.i19.i.i.i.i.i1861
  %arrayidx.i22.i.i.i.i.i1863 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %arrayidx.i20.i.i.i.i.i1862, i64 %sh_prom.i.i.i.i756
  store i64 %sub.ptr.sub.i.i.i1816, ptr %arrayidx.i22.i.i.i.i.i1863, align 8, !tbaa !82
  %spec.select.i.i.i.i.i.i.i1864 = call i64 @llvm.umin.i64(i64 %sub.ptr.sub.i.i.i1816, i64 56)
  %2849 = and i64 %spec.select.i.i.i.i.i.i.i1864, 15
  %2850 = sub i64 %spec.select.i.i.i.i.i.i.i1864, %2849
  %2851 = icmp ne i64 %2850, 0
  %2852 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %2851)
  %2853 = extractvalue { i1, i64 } %2852, 0
  %2854 = extractvalue { i1, i64 } %2852, 1
  br i1 %2853, label %loop-memcpy-expansion5157.preheader, label %Flow8240

loop-memcpy-expansion5157.preheader:              ; preds = %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1859
  %scevgep6698 = getelementptr i8, ptr %2848, i64 8
  %scevgep6699 = getelementptr i8, ptr %scevgep6698, i64 %1474
  %2855 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i1861, 6
  %scevgep6700 = getelementptr i8, ptr %scevgep6699, i64 %2855
  br label %loop-memcpy-expansion5157, !amdgpu.uniform !42

Flow8240:                                         ; preds = %Flow8239, %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1859
  call void @llvm.amdgcn.end.cf.i64(i64 %2854)
  br label %loop-memcpy-residual-header5160, !amdgpu.uniform !42

loop-memcpy-expansion5157:                        ; preds = %loop-memcpy-expansion5157.preheader, %loop-memcpy-expansion5157
  %phi.broken8686 = phi i64 [ 0, %loop-memcpy-expansion5157.preheader ], [ %2860, %loop-memcpy-expansion5157 ]
  %loop-index5158 = phi i64 [ %2858, %loop-memcpy-expansion5157 ], [ 0, %loop-memcpy-expansion5157.preheader ]
  %2856 = getelementptr i8, ptr addrspace(4) %cond, i64 %loop-index5158
  %2857 = load <4 x i32>, ptr addrspace(4) %2856, align 1
  %scevgep6701 = getelementptr i8, ptr %scevgep6700, i64 %loop-index5158
  store <4 x i32> %2857, ptr %scevgep6701, align 1
  %2858 = add i64 %loop-index5158, 16
  %2859 = icmp uge i64 %2858, %2850
  %2860 = call i64 @llvm.amdgcn.if.break.i64(i1 %2859, i64 %phi.broken8686)
  %2861 = call i1 @llvm.amdgcn.loop.i64(i64 %2860)
  br i1 %2861, label %Flow8239, label %loop-memcpy-expansion5157

Flow8238:                                         ; preds = %Flow8236, %Flow8237
  %2862 = phi i64 [ %6451, %Flow8236 ], [ %6446, %Flow8237 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %6449)
  br label %post-loop-memcpy-expansion5156, !amdgpu.uniform !42

loop-memcpy-residual5159:                         ; preds = %loop-memcpy-residual5159.preheader, %loop-memcpy-residual5159
  %phi.broken8687 = phi i64 [ 0, %loop-memcpy-residual5159.preheader ], [ %2866, %loop-memcpy-residual5159 ]
  %residual-loop-index5161 = phi i64 [ %2864, %loop-memcpy-residual5159 ], [ 0, %loop-memcpy-residual5159.preheader ]
  %scevgep6707 = getelementptr i8, ptr addrspace(4) %scevgep6706, i64 %residual-loop-index5161
  %2863 = load i8, ptr addrspace(4) %scevgep6707, align 1
  %scevgep6705 = getelementptr i8, ptr %scevgep6704, i64 %residual-loop-index5161
  store i8 %2863, ptr %scevgep6705, align 1
  %2864 = add i64 %residual-loop-index5161, 1
  %2865 = icmp uge i64 %2864, %2849
  %2866 = call i64 @llvm.amdgcn.if.break.i64(i1 %2865, i64 %phi.broken8687)
  %2867 = call i1 @llvm.amdgcn.loop.i64(i64 %2866)
  br i1 %2867, label %Flow8236, label %loop-memcpy-residual5159

Flow8236:                                         ; preds = %loop-memcpy-residual5159
  %.lcssa9055 = phi i64 [ %2866, %loop-memcpy-residual5159 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa9055)
  br label %Flow8238, !amdgpu.uniform !42

post-loop-memcpy-expansion5156:                   ; preds = %Flow8238
  %2868 = addrspacecast ptr addrspace(4) %scevgep6697.lcssa to ptr
  %2869 = ptrtoint ptr %2868 to i64
  %tobool.not.i.i.i.i.i.i1866 = icmp eq i32 %.lcssa9051, 0
  %conv.i.i.i.i.i.i1867 = zext i1 %tobool.not.i.i.i.i.i.i1866 to i32
  fence release
  %2870 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %idxprom.i23.i.i.i.i.i1868 = zext i32 %.lcssa9053 to i64
  %arrayidx.i24.i.i.i.i.i1869 = getelementptr inbounds nuw i32, ptr %2870, i64 %idxprom.i23.i.i.i.i.i1868
  store atomic i32 %conv.i.i.i.i.i.i1867, ptr %arrayidx.i24.i.i.i.i.i1869 monotonic, align 4
  %2871 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !84
  %arrayidx.i.i5.i.i1870 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %2871, i64 %idxprom.i23.i.i.i.i.i1868
  %2872 = load i64, ptr %arrayidx.i.i5.i.i1870, align 8, !tbaa !80
  %2873 = add i64 %1140, %2869
  br label %while.cond.i.i.i.i1871, !amdgpu.uniform !42

while.cond.i.i.i.i1871:                           ; preds = %post-loop-memcpy-expansion5156, %Flow8235
  %phi.broken8691 = phi i64 [ 0, %post-loop-memcpy-expansion5156 ], [ %2895, %Flow8235 ]
  %lsr.iv6713 = phi i64 [ %2893, %Flow8235 ], [ %2873, %post-loop-memcpy-expansion5156 ]
  %lsr.iv6708 = phi ptr addrspace(4) [ %2892, %Flow8235 ], [ %scevgep6600, %post-loop-memcpy-expansion5156 ]
  %port.sroa.37.0.i.i1872 = phi i32 [ %2891, %Flow8235 ], [ %conv.i.i.i.i.i.i1867, %post-loop-memcpy-expansion5156 ]
  %idx.0.i.i.i.i1873 = phi i64 [ %2890, %Flow8235 ], [ 56, %post-loop-memcpy-expansion5156 ]
  %cmp.i.i.i.i1874 = icmp ugt i64 %sub.ptr.sub.i.i.i1816, %idx.0.i.i.i.i1873
  %2874 = call i64 @llvm.amdgcn.ballot.i64(i1 %cmp.i.i.i.i1874)
  %and.i.i.i.i6.i.i1875 = and i64 %2874, %2872
  %tobool.not.i.i.i.i1876 = icmp ne i64 %and.i.i.i.i6.i.i1875, 0
  %2875 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i22.i.i.i1877 = getelementptr inbounds nuw i32, ptr %2875, i64 %idxprom.i23.i.i.i.i.i1868
  %2876 = load atomic i32, ptr %arrayidx.i22.i.i.i1877 monotonic, align 4
  %2877 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %tobool.not.i.i.i.i1876)
  %2878 = extractvalue { i1, i64 } %2877, 0
  %2879 = extractvalue { i1, i64 } %2877, 1
  br i1 %2878, label %cond.false.i8.i.i.i.i1878, label %Flow8235

cond.false.i8.i.i.i.i1878:                        ; preds = %while.cond.i.i.i.i1871
  %2880 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %2876)
  %2881 = icmp ne i32 %2880, %port.sroa.37.0.i.i1872
  %2882 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %2881)
  %2883 = extractvalue { i1, i64 } %2882, 0
  %2884 = extractvalue { i1, i64 } %2882, 1
  br i1 %2883, label %while.body.i.i21.i.i.i.i1879.preheader, label %Flow8234

while.body.i.i21.i.i.i.i1879.preheader:           ; preds = %cond.false.i8.i.i.i.i1878
  br label %while.body.i.i21.i.i.i.i1879, !amdgpu.uniform !42

Flow8234:                                         ; preds = %Flow8233, %cond.false.i8.i.i.i.i1878
  call void @llvm.amdgcn.end.cf.i64(i64 %2884)
  br label %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i1882, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i1879:                     ; preds = %while.body.i.i21.i.i.i.i1879.preheader, %while.body.i.i21.i.i.i.i1879
  %phi.broken8688 = phi i64 [ 0, %while.body.i.i21.i.i.i.i1879.preheader ], [ %2888, %while.body.i.i21.i.i.i.i1879 ]
  call void @llvm.amdgcn.s.sleep(i32 2)
  %2885 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i22.i.i.i.i1880 = getelementptr inbounds nuw i32, ptr %2885, i64 %idxprom.i23.i.i.i.i.i1868
  %2886 = load atomic i32, ptr %arrayidx.i.i.i22.i.i.i.i1880 monotonic, align 4
  %2887 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %2886)
  %cmp.i.not.i.i23.i.i.i.i1881 = icmp eq i32 %2887, %port.sroa.37.0.i.i1872
  %2888 = call i64 @llvm.amdgcn.if.break.i64(i1 %cmp.i.not.i.i23.i.i.i.i1881, i64 %phi.broken8688)
  %2889 = call i1 @llvm.amdgcn.loop.i64(i64 %2888)
  br i1 %2889, label %Flow8233, label %while.body.i.i21.i.i.i.i1879

Flow8233:                                         ; preds = %while.body.i.i21.i.i.i.i1879
  %.lcssa9056 = phi i64 [ %2888, %while.body.i.i21.i.i.i.i1879 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa9056)
  br label %Flow8234, !amdgpu.uniform !42

Flow8235:                                         ; preds = %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1883, %while.cond.i.i.i.i1871
  %2890 = phi i64 [ %add.i.i.i.i1886, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1883 ], [ poison, %while.cond.i.i.i.i1871 ]
  %2891 = phi i32 [ %conv.i.i26.i.i.i.i1884, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1883 ], [ poison, %while.cond.i.i.i.i1871 ]
  %2892 = phi ptr addrspace(4) [ %scevgep6709, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1883 ], [ poison, %while.cond.i.i.i.i1871 ]
  %2893 = phi i64 [ %lsr.iv.next6714, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1883 ], [ poison, %while.cond.i.i.i.i1871 ]
  %2894 = phi i1 [ false, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1883 ], [ true, %while.cond.i.i.i.i1871 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %2879)
  %2895 = call i64 @llvm.amdgcn.if.break.i64(i1 %2894, i64 %phi.broken8691)
  %2896 = call i1 @llvm.amdgcn.loop.i64(i64 %2895)
  br i1 %2896, label %cond.false.i.i.i1893, label %while.cond.i.i.i.i1871

_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i1882: ; preds = %Flow8234
  fence acquire
  %2897 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.i.i.i.i1874)
  %2898 = extractvalue { i1, i64 } %2897, 0
  %2899 = extractvalue { i1, i64 } %2897, 1
  br i1 %2898, label %if.then.i.i.i.i.i.i.i1887, label %Flow8232

if.then.i.i.i.i.i.i.i1887:                        ; preds = %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i1882
  %sub.i.i.i.i.i.i.i1889 = sub nuw i64 %sub.ptr.sub.i.i.i1816, %idx.0.i.i.i.i1873
  %spec.select.i.i.i36.i.i.i.i1890 = call i64 @llvm.umin.i64(i64 %sub.i.i.i.i.i.i.i1889, i64 64)
  %2900 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %2901 = and i64 %spec.select.i.i.i36.i.i.i.i1890, 15
  %2902 = sub i64 %spec.select.i.i.i36.i.i.i.i1890, %2901
  %2903 = icmp ne i64 %2902, 0
  %2904 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %2903)
  %2905 = extractvalue { i1, i64 } %2904, 0
  %2906 = extractvalue { i1, i64 } %2904, 1
  br i1 %2905, label %loop-memcpy-expansion5313.preheader, label %Flow8231

loop-memcpy-expansion5313.preheader:              ; preds = %if.then.i.i.i.i.i.i.i1887
  %arrayidx.i20.i35.i.i.i.i1891 = getelementptr %"struct.rpc::Buffer", ptr %2900, i64 %idxprom.i19.i.i.i.i.i1861
  %arrayidx.i22.i38.i.i.i.i1892 = getelementptr %"struct.rpc::Buffer", ptr %arrayidx.i20.i35.i.i.i.i1891, i64 %sh_prom.i.i.i.i756
  br label %loop-memcpy-expansion5313, !amdgpu.uniform !42

Flow8231:                                         ; preds = %Flow8230, %if.then.i.i.i.i.i.i.i1887
  call void @llvm.amdgcn.end.cf.i64(i64 %2906)
  br label %loop-memcpy-residual-header5316, !amdgpu.uniform !42

loop-memcpy-expansion5313:                        ; preds = %loop-memcpy-expansion5313.preheader, %loop-memcpy-expansion5313
  %phi.broken8689 = phi i64 [ 0, %loop-memcpy-expansion5313.preheader ], [ %2911, %loop-memcpy-expansion5313 ]
  %loop-index5314 = phi i64 [ %2909, %loop-memcpy-expansion5313 ], [ 0, %loop-memcpy-expansion5313.preheader ]
  %scevgep6710 = getelementptr i8, ptr addrspace(4) %lsr.iv6708, i64 %loop-index5314
  %2907 = load <4 x i32>, ptr addrspace(4) %scevgep6710, align 1
  %2908 = getelementptr i8, ptr %arrayidx.i22.i38.i.i.i.i1892, i64 %loop-index5314
  store <4 x i32> %2907, ptr %2908, align 1
  %2909 = add i64 %loop-index5314, 16
  %2910 = icmp uge i64 %2909, %2902
  %2911 = call i64 @llvm.amdgcn.if.break.i64(i1 %2910, i64 %phi.broken8689)
  %2912 = call i1 @llvm.amdgcn.loop.i64(i64 %2911)
  br i1 %2912, label %Flow8230, label %loop-memcpy-expansion5313

Flow8229:                                         ; preds = %Flow8228, %loop-memcpy-residual-header5316
  call void @llvm.amdgcn.end.cf.i64(i64 %6594)
  br label %Flow8232, !amdgpu.uniform !42

loop-memcpy-residual5315:                         ; preds = %loop-memcpy-residual5315.preheader, %loop-memcpy-residual5315
  %phi.broken8690 = phi i64 [ 0, %loop-memcpy-residual5315.preheader ], [ %2916, %loop-memcpy-residual5315 ]
  %residual-loop-index5317 = phi i64 [ %2914, %loop-memcpy-residual5315 ], [ 0, %loop-memcpy-residual5315.preheader ]
  %scevgep6719 = getelementptr i8, ptr addrspace(4) %scevgep6718, i64 %residual-loop-index5317
  %2913 = load i8, ptr addrspace(4) %scevgep6719, align 1
  %scevgep6717 = getelementptr i8, ptr %scevgep6716, i64 %residual-loop-index5317
  store i8 %2913, ptr %scevgep6717, align 1
  %2914 = add i64 %residual-loop-index5317, 1
  %2915 = icmp uge i64 %2914, %2901
  %2916 = call i64 @llvm.amdgcn.if.break.i64(i1 %2915, i64 %phi.broken8690)
  %2917 = call i1 @llvm.amdgcn.loop.i64(i64 %2916)
  br i1 %2917, label %Flow8228, label %loop-memcpy-residual5315

Flow8228:                                         ; preds = %loop-memcpy-residual5315
  %.lcssa9058 = phi i64 [ %2916, %loop-memcpy-residual5315 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa9058)
  br label %Flow8229, !amdgpu.uniform !42

_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1883: ; preds = %Flow8232
  %conv.i.i26.i.i.i.i1884 = xor i32 %port.sroa.37.0.i.i1872, 1
  fence release
  %2918 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %arrayidx.i24.i29.i.i.i.i1885 = getelementptr inbounds nuw i32, ptr %2918, i64 %idxprom.i23.i.i.i.i.i1868
  store atomic i32 %conv.i.i26.i.i.i.i1884, ptr %arrayidx.i24.i29.i.i.i.i1885 monotonic, align 4
  %add.i.i.i.i1886 = add i64 %idx.0.i.i.i.i1873, 64
  %scevgep6709 = getelementptr i8, ptr addrspace(4) %lsr.iv6708, i64 64
  %lsr.iv.next6714 = add i64 %lsr.iv6713, -64
  br label %Flow8235, !amdgpu.uniform !42

cond.false.i.i.i1893:                             ; preds = %Flow8235
  %.lcssa9060 = phi i64 [ %2895, %Flow8235 ]
  %port.sroa.37.0.i.i1872.lcssa = phi i32 [ %port.sroa.37.0.i.i1872, %Flow8235 ]
  %.lcssa9059 = phi i32 [ %2876, %Flow8235 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa9060)
  %2919 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %.lcssa9059)
  %2920 = icmp ne i32 %2919, %port.sroa.37.0.i.i1872.lcssa
  %2921 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %2920)
  %2922 = extractvalue { i1, i64 } %2921, 0
  %2923 = extractvalue { i1, i64 } %2921, 1
  br i1 %2922, label %while.body.i.i.i.i1894.preheader, label %Flow8227

while.body.i.i.i.i1894.preheader:                 ; preds = %cond.false.i.i.i1893
  br label %while.body.i.i.i.i1894, !amdgpu.uniform !42

Flow8227:                                         ; preds = %Flow8226, %cond.false.i.i.i1893
  call void @llvm.amdgcn.end.cf.i64(i64 %2923)
  br label %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit1915, !amdgpu.uniform !42

while.body.i.i.i.i1894:                           ; preds = %while.body.i.i.i.i1894.preheader, %while.body.i.i.i.i1894
  %phi.broken8692 = phi i64 [ 0, %while.body.i.i.i.i1894.preheader ], [ %2927, %while.body.i.i.i.i1894 ]
  call void @llvm.amdgcn.s.sleep(i32 2)
  %2924 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i8.i.i1895 = getelementptr inbounds nuw i32, ptr %2924, i64 %idxprom.i23.i.i.i.i.i1868
  %2925 = load atomic i32, ptr %arrayidx.i.i.i8.i.i1895 monotonic, align 4
  %2926 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %2925)
  %cmp.i.not.i.i.i.i1896 = icmp eq i32 %2926, %port.sroa.37.0.i.i1872.lcssa
  %2927 = call i64 @llvm.amdgcn.if.break.i64(i1 %cmp.i.not.i.i.i.i1896, i64 %phi.broken8692)
  %2928 = call i1 @llvm.amdgcn.loop.i64(i64 %2927)
  br i1 %2928, label %Flow8226, label %while.body.i.i.i.i1894

Flow8226:                                         ; preds = %while.body.i.i.i.i1894
  %.lcssa9061 = phi i64 [ %2927, %while.body.i.i.i.i1894 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa9061)
  br label %Flow8227, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit1915: ; preds = %Flow8227
  fence acquire
  call void @llvm.amdgcn.wave.barrier()
  fence syncscope("agent") release
  %2929 = call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %.lcssa9052, i1 true)
  %iszero.i.i.i.i.i13.i.i1898 = icmp ne i64 %.lcssa9052, 0
  %cmp2.i.i.i.i14.i.i1899 = icmp eq i64 %2929, %sh_prom.i.i.i.i756
  %cmp.i.i.i.i15.i.i1900 = select i1 %iszero.i.i.i.i.i13.i.i1898, i1 %cmp2.i.i.i.i14.i.i1899, i1 false
  %2930 = and i32 %.lcssa9053, 31
  %conv4.i.i.i19.i.i1904 = zext i1 %cmp.i.i.i.i15.i.i1900 to i32
  %shl.i.i.i20.i.i1905 = shl nuw i32 %conv4.i.i.i19.i.i1904, %2930
  %xor.i.i.i21.i.i1906 = xor i32 %shl.i.i.i20.i.i1905, -1
  %2931 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop5706, !amdgpu.uniform !42

2932:                                             ; preds = %ComputeEnd5707
  %div8.i.i.i16.i.i1901 = lshr i32 %.lcssa9053, 5
  %idxprom.i.i.i17.i.i1902 = zext nneg i32 %div8.i.i.i16.i.i1901 to i64
  %2933 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i17.i.i1902
  %sunkaddr7422 = getelementptr inbounds i8, ptr addrspace(1) %2933, i64 40
  %2934 = atomicrmw and ptr addrspace(1) %sunkaddr7422, i32 %.lcssa9062 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %2935, !amdgpu.uniform !42

2935:                                             ; preds = %2932, %ComputeEnd5707
  call void @llvm.amdgcn.end.cf.i64(i64 %7714)
  call void @llvm.amdgcn.wave.barrier()
  br label %for.cond.i.i1.i1924, !amdgpu.uniform !42

for.cond.i.i1.i1924:                              ; preds = %2935, %Flow8225
  %2936 = phi i32 [ %2973, %Flow8225 ], [ %1183, %2935 ]
  %2937 = phi i64 [ %2974, %Flow8225 ], [ %1184, %2935 ]
  %2938 = phi i32 [ %2975, %Flow8225 ], [ %1185, %2935 ]
  %retval.sroa.6.0.i.i.i1925 = phi i32 [ %2978, %Flow8225 ], [ undef, %2935 ]
  %retval.sroa.2.0.i.i.i1927 = phi i64 [ %2977, %Flow8225 ], [ undef, %2935 ]
  %retval.sroa.8.0.i.i.i1928 = phi i32 [ %2976, %Flow8225 ], [ undef, %2935 ]
  %index.0.i.i.i1929 = phi i32 [ %2980, %Flow8225 ], [ 0, %2935 ]
  %2939 = load i32, ptr addrspace(1) @__llvm_rpc_client, align 8, !tbaa !68
  %cmp.not.i.i.i1930 = icmp ult i32 %index.0.i.i.i1929, %2939
  %spec.store.select.i.i.i1931 = select i1 %cmp.not.i.i.i1930, i32 %index.0.i.i.i1929, i32 0
  %2940 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  %2941 = and i32 %spec.store.select.i.i.i1931, 31
  %2942 = lshr i64 %2940, %sh_prom.i.i.i.i756
  %2943 = trunc i64 %2942 to i32
  %conv4.i.i.i.i.i1935 = and i32 %2943, 1
  %shl.i.i.i.i.i1936 = shl nuw i32 %conv4.i.i.i.i.i1935, %2941
  br label %ComputeLoop5710, !amdgpu.uniform !42

2944:                                             ; preds = %ComputeEnd5711
  %sunkaddr7423 = getelementptr inbounds i8, ptr addrspace(1) %7724, i64 40
  %2945 = atomicrmw or ptr addrspace(1) %sunkaddr7423, i32 %.lcssa9063 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %2946, !amdgpu.uniform !42

2946:                                             ; preds = %2944, %ComputeEnd5711
  %2947 = phi i32 [ %2945, %2944 ], [ poison, %ComputeEnd5711 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %7733)
  %2948 = call i32 @llvm.amdgcn.readfirstlane.i32(i32 %2947)
  %2949 = or i32 %2948, %.lcssa9064
  %shl5.i.i.i.i.i1937 = shl nuw i32 1, %2941
  %and.i.i.i.i.i1938 = and i32 %shl5.i.i.i.i.i1937, %2949
  %tobool3.i.i.i.i1939 = icmp ne i32 %and.i.i.i.i.i1938, 0
  %2950 = call i64 @llvm.amdgcn.ballot.i64(i1 %tobool3.i.i.i.i1939)
  %and.i.i.i.i.i.i1940 = and i64 %2950, %2940
  %cmp.i.not.i.i.i1941 = icmp ne i64 %2940, %and.i.i.i.i.i.i1940
  br i1 %cmp.i.not.i.i.i1941, label %if.then.i.i.i.i1942, label %Flow8224, !amdgpu.uniform !42

if.then.i.i.i.i1942:                              ; preds = %2946
  fence syncscope("agent") acquire
  %2951 = load <2 x i64>, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !75
  %2952 = extractelement <2 x i64> %2951, i32 0
  %2953 = inttoptr i64 %2952 to ptr
  %2954 = extractelement <2 x i64> %2951, i32 1
  %2955 = inttoptr i64 %2954 to ptr
  %idxprom.i.i.i.i1943 = zext i32 %spec.store.select.i.i.i1931 to i64
  %arrayidx.i.i.i.i1944 = getelementptr inbounds nuw i32, ptr %2953, i64 %idxprom.i.i.i.i1943, !amdgpu.uniform !42
  %2956 = load atomic i32, ptr %arrayidx.i.i.i.i1944 monotonic, align 4
  %2957 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %2956)
  %arrayidx.i45.i.i.i1945 = getelementptr inbounds nuw i32, ptr %2955, i64 %idxprom.i.i.i.i1943, !amdgpu.uniform !42
  %2958 = load atomic i32, ptr %arrayidx.i45.i.i.i1945 monotonic, align 4
  %2959 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %2958)
  %cmp.i46.not.i.i.i1946 = icmp ne i32 %2957, %2959
  br i1 %cmp.i46.not.i.i.i1946, label %if.then12.i.i.i1947, label %Flow8222, !amdgpu.uniform !42

if.then12.i.i.i1947:                              ; preds = %if.then.i.i.i.i1942
  fence syncscope("agent") release
  %2960 = call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %2940, i1 true)
  %iszero.i.i.i.i.i.i.i1948 = icmp ne i64 %2940, 0
  %cmp2.i.i.i.i.i.i1949 = icmp eq i64 %2960, %sh_prom.i.i.i.i756
  %cmp.i.i.i.i.i.i1950 = select i1 %iszero.i.i.i.i.i.i.i1948, i1 %cmp2.i.i.i.i.i.i1949, i1 false
  %conv4.i.i51.i.i.i1951 = zext i1 %cmp.i.i.i.i.i.i1950 to i32
  %shl.i.i52.i.i.i1952 = shl nuw i32 %conv4.i.i51.i.i.i1951, %2941
  %xor.i.i.i.i.i1953 = xor i32 %shl.i.i52.i.i.i1952, -1
  %2961 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop5715, !amdgpu.uniform !42

2962:                                             ; preds = %ComputeEnd5716
  %sunkaddr7424 = getelementptr inbounds i8, ptr addrspace(1) %7724, i64 40
  %2963 = atomicrmw and ptr addrspace(1) %sunkaddr7424, i32 %.lcssa9065 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %2964, !amdgpu.uniform !42

2964:                                             ; preds = %2962, %ComputeEnd5716
  call void @llvm.amdgcn.end.cf.i64(i64 %7751)
  call void @llvm.amdgcn.wave.barrier()
  br label %Flow8222, !amdgpu.uniform !42

if.end14.i.i.i2013:                               ; preds = %Flow8222
  %2965 = call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %2940, i1 true)
  %iszero.i.i.i.i.i.i2014 = icmp ne i64 %2940, 0
  %cmp2.i.i.i.i.i2015 = icmp eq i64 %2965, %sh_prom.i.i.i.i756
  %cmp.i.i.i.i.i2016 = select i1 %iszero.i.i.i.i.i.i2014, i1 %cmp2.i.i.i.i.i2015, i1 false
  %2966 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.i.i.i.i.i2016)
  %2967 = extractvalue { i1, i64 } %2966, 0
  %2968 = extractvalue { i1, i64 } %2966, 1
  br i1 %2967, label %if.then16.i.i.i2018, label %if.end22.i.i.i2017

if.then16.i.i.i2018:                              ; preds = %if.end14.i.i.i2013
  %2969 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !76
  %opcode.i.i.i2019 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %2969, i64 %idxprom.i.i.i.i1943, i32 1
  store i32 1660944387, ptr %opcode.i.i.i2019, align 8, !tbaa !77
  %arrayidx21.i.i.i2020 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %2969, i64 %idxprom.i.i.i.i1943
  store i64 %2940, ptr %arrayidx21.i.i.i2020, align 8, !tbaa !80
  br label %if.end22.i.i.i2017, !amdgpu.uniform !42

Flow8223:                                         ; preds = %if.end22.i.i.i2017, %Flow8222
  %2970 = phi i32 [ %2957, %if.end22.i.i.i2017 ], [ %retval.sroa.8.0.i.i.i1928, %Flow8222 ]
  %2971 = phi i64 [ %2940, %if.end22.i.i.i2017 ], [ %retval.sroa.2.0.i.i.i1927, %Flow8222 ]
  %2972 = phi i32 [ %spec.store.select.i.i.i1931, %if.end22.i.i.i2017 ], [ %retval.sroa.6.0.i.i.i1925, %Flow8222 ]
  br label %cleanup26.i.i.i1954, !amdgpu.uniform !42

if.end22.i.i.i2017:                               ; preds = %if.then16.i.i.i2018, %if.end14.i.i.i2013
  call void @llvm.amdgcn.end.cf.i64(i64 %2968)
  call void @llvm.amdgcn.wave.barrier()
  br label %Flow8223, !amdgpu.uniform !42

Flow8224:                                         ; preds = %cleanup26.i.i.i1954, %2946
  %2973 = phi i32 [ %2972, %cleanup26.i.i.i1954 ], [ %2936, %2946 ]
  %2974 = phi i64 [ %2971, %cleanup26.i.i.i1954 ], [ %2937, %2946 ]
  %2975 = phi i32 [ %2970, %cleanup26.i.i.i1954 ], [ %2938, %2946 ]
  %2976 = phi i32 [ %2970, %cleanup26.i.i.i1954 ], [ %retval.sroa.8.0.i.i.i1928, %2946 ]
  %2977 = phi i64 [ %2971, %cleanup26.i.i.i1954 ], [ %retval.sroa.2.0.i.i.i1927, %2946 ]
  %2978 = phi i32 [ %2972, %cleanup26.i.i.i1954 ], [ %retval.sroa.6.0.i.i.i1925, %2946 ]
  %2979 = phi i1 [ %cmp.i46.not.i.i.i1946, %cleanup26.i.i.i1954 ], [ true, %2946 ]
  br i1 %2979, label %for.inc.i.i.i1959, label %Flow8225, !amdgpu.uniform !42

cleanup26.i.i.i1954:                              ; preds = %Flow8223
  br label %Flow8224, !amdgpu.uniform !42

for.inc.i.i.i1959:                                ; preds = %Flow8224
  %inc.i.i.i1964 = add i32 %spec.store.select.i.i.i1931, 1
  br label %Flow8225, !amdgpu.uniform !42

Flow8225:                                         ; preds = %for.inc.i.i.i1959, %Flow8224
  %2980 = phi i32 [ %inc.i.i.i1964, %for.inc.i.i.i1959 ], [ poison, %Flow8224 ]
  %2981 = phi i1 [ false, %for.inc.i.i.i1959 ], [ true, %Flow8224 ]
  br i1 %2981, label %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1965, label %for.cond.i.i1.i1924, !amdgpu.uniform !42

_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1965: ; preds = %Flow8225
  %.lcssa9068 = phi i32 [ %2973, %Flow8225 ]
  %.lcssa9067 = phi i64 [ %2974, %Flow8225 ]
  %.lcssa9066 = phi i32 [ %2975, %Flow8225 ]
  fence acquire
  %2982 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %mul.i.i.i.i.i.i1966 = shl i32 %.lcssa9068, 6
  %idxprom.i19.i.i.i.i.i1967 = zext i32 %mul.i.i.i.i.i.i1966 to i64
  %arrayidx.i20.i.i.i.i.i1968 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %2982, i64 %idxprom.i19.i.i.i.i.i1967
  %arrayidx.i22.i.i.i.i.i1969 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %arrayidx.i20.i.i.i.i.i1968, i64 %sh_prom.i.i.i.i756
  store i64 sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.19.36 to ptr), i64 13) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.19.36 to ptr) to i64)), ptr %arrayidx.i22.i.i.i.i.i1969, align 8, !tbaa !82
  br i1 %1114, label %loop-memcpy-expansion5319.preheader, label %Flow8221, !amdgpu.uniform !42

loop-memcpy-expansion5319.preheader:              ; preds = %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1965
  %scevgep6720 = getelementptr i8, ptr %2982, i64 8
  %scevgep6721 = getelementptr i8, ptr %scevgep6720, i64 %1474
  %2983 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i1967, 6
  %scevgep6722 = getelementptr i8, ptr %scevgep6721, i64 %2983
  br label %loop-memcpy-expansion5319, !amdgpu.uniform !42

Flow8221:                                         ; preds = %Flow8220, %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i1965
  br label %loop-memcpy-residual-header5322, !amdgpu.uniform !42

loop-memcpy-expansion5319:                        ; preds = %loop-memcpy-expansion5319.preheader, %loop-memcpy-expansion5319
  %loop-index5320 = phi i64 [ %2986, %loop-memcpy-expansion5319 ], [ 0, %loop-memcpy-expansion5319.preheader ]
  %2984 = getelementptr i8, ptr addrspace(4) @.str.19.36, i64 %loop-index5320, !amdgpu.uniform !42
  %2985 = load <4 x i32>, ptr addrspace(4) %2984, align 1
  %scevgep6723 = getelementptr i8, ptr %scevgep6722, i64 %loop-index5320
  store <4 x i32> %2985, ptr %scevgep6723, align 1
  %2986 = add i64 %loop-index5320, 16
  %2987 = icmp uge i64 %2986, %1113
  br i1 %2987, label %Flow8220, label %loop-memcpy-expansion5319, !amdgpu.uniform !42

Flow8219:                                         ; preds = %Flow8217, %Flow8218
  %2988 = phi i64 [ %6599, %Flow8217 ], [ %6597, %Flow8218 ]
  br label %post-loop-memcpy-expansion5318, !amdgpu.uniform !42

loop-memcpy-residual5321:                         ; preds = %loop-memcpy-residual5321.preheader, %loop-memcpy-residual5321
  %residual-loop-index5323 = phi i64 [ %2990, %loop-memcpy-residual5321 ], [ 0, %loop-memcpy-residual5321.preheader ]
  %scevgep6728 = getelementptr i8, ptr addrspace(4) @.str.19.36, i64 %residual-loop-index5323, !amdgpu.uniform !42
  %2989 = load i8, ptr addrspace(4) %scevgep6728, align 1
  %scevgep6727 = getelementptr i8, ptr %scevgep6726, i64 %residual-loop-index5323
  store i8 %2989, ptr %scevgep6727, align 1
  %2990 = add i64 %residual-loop-index5323, 1
  %2991 = icmp uge i64 %2990, %1112
  br i1 %2991, label %Flow8217, label %loop-memcpy-residual5321, !amdgpu.uniform !42

Flow8217:                                         ; preds = %loop-memcpy-residual5321
  br label %Flow8219, !amdgpu.uniform !42

post-loop-memcpy-expansion5318:                   ; preds = %Flow8219
  %tobool.not.i.i.i.i.i.i1972 = icmp eq i32 %.lcssa9066, 0
  %conv.i.i.i.i.i.i1973 = zext i1 %tobool.not.i.i.i.i.i.i1972 to i32
  fence release
  %2992 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %idxprom.i23.i.i.i.i.i1974 = zext i32 %.lcssa9068 to i64
  %arrayidx.i24.i.i.i.i.i1975 = getelementptr inbounds nuw i32, ptr %2992, i64 %idxprom.i23.i.i.i.i.i1974
  store atomic i32 %conv.i.i.i.i.i.i1973, ptr %arrayidx.i24.i.i.i.i.i1975 monotonic, align 4
  %2993 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !84
  %arrayidx.i.i5.i.i1976 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %2993, i64 %idxprom.i23.i.i.i.i.i1974
  %2994 = load i64, ptr %arrayidx.i.i5.i.i1976, align 8, !tbaa !80
  br label %while.cond.i.i.i.i1977, !amdgpu.uniform !42

while.cond.i.i.i.i1977:                           ; preds = %post-loop-memcpy-expansion5318, %Flow8216
  %phi.broken8694 = phi i64 [ 0, %post-loop-memcpy-expansion5318 ], [ %3016, %Flow8216 ]
  %lsr.iv6734 = phi i64 [ %3014, %Flow8216 ], [ -43, %post-loop-memcpy-expansion5318 ]
  %lsr.iv6729 = phi ptr addrspace(4) [ %3013, %Flow8216 ], [ getelementptr (i8, ptr addrspace(4) @.str.19.36, i64 56), %post-loop-memcpy-expansion5318 ]
  %port.sroa.37.0.i.i1978 = phi i32 [ %3012, %Flow8216 ], [ %conv.i.i.i.i.i.i1973, %post-loop-memcpy-expansion5318 ]
  %idx.0.i.i.i.i1979 = phi i64 [ %3011, %Flow8216 ], [ 56, %post-loop-memcpy-expansion5318 ]
  %cmp.i.i.i.i1980 = icmp ult i64 %idx.0.i.i.i.i1979, sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.19.36 to ptr), i64 13) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.19.36 to ptr) to i64))
  %2995 = call i64 @llvm.amdgcn.ballot.i64(i1 %cmp.i.i.i.i1980)
  %and.i.i.i.i6.i.i1981 = and i64 %2995, %2994
  %tobool.not.i.i.i.i1982 = icmp ne i64 %and.i.i.i.i6.i.i1981, 0
  %2996 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i22.i.i.i1983 = getelementptr inbounds nuw i32, ptr %2996, i64 %idxprom.i23.i.i.i.i.i1974
  %2997 = load atomic i32, ptr %arrayidx.i22.i.i.i1983 monotonic, align 4
  %2998 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %tobool.not.i.i.i.i1982)
  %2999 = extractvalue { i1, i64 } %2998, 0
  %3000 = extractvalue { i1, i64 } %2998, 1
  br i1 %2999, label %cond.false.i8.i.i.i.i1984, label %Flow8216

cond.false.i8.i.i.i.i1984:                        ; preds = %while.cond.i.i.i.i1977
  %3001 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %2997)
  %3002 = icmp ne i32 %3001, %port.sroa.37.0.i.i1978
  %3003 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3002)
  %3004 = extractvalue { i1, i64 } %3003, 0
  %3005 = extractvalue { i1, i64 } %3003, 1
  br i1 %3004, label %while.body.i.i21.i.i.i.i1985.preheader, label %Flow8215

while.body.i.i21.i.i.i.i1985.preheader:           ; preds = %cond.false.i8.i.i.i.i1984
  br label %while.body.i.i21.i.i.i.i1985, !amdgpu.uniform !42

Flow8215:                                         ; preds = %Flow8214, %cond.false.i8.i.i.i.i1984
  call void @llvm.amdgcn.end.cf.i64(i64 %3005)
  br label %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i1988, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i1985:                     ; preds = %while.body.i.i21.i.i.i.i1985.preheader, %while.body.i.i21.i.i.i.i1985
  %phi.broken8693 = phi i64 [ 0, %while.body.i.i21.i.i.i.i1985.preheader ], [ %3009, %while.body.i.i21.i.i.i.i1985 ]
  call void @llvm.amdgcn.s.sleep(i32 2)
  %3006 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i22.i.i.i.i1986 = getelementptr inbounds nuw i32, ptr %3006, i64 %idxprom.i23.i.i.i.i.i1974
  %3007 = load atomic i32, ptr %arrayidx.i.i.i22.i.i.i.i1986 monotonic, align 4
  %3008 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %3007)
  %cmp.i.not.i.i23.i.i.i.i1987 = icmp eq i32 %3008, %port.sroa.37.0.i.i1978
  %3009 = call i64 @llvm.amdgcn.if.break.i64(i1 %cmp.i.not.i.i23.i.i.i.i1987, i64 %phi.broken8693)
  %3010 = call i1 @llvm.amdgcn.loop.i64(i64 %3009)
  br i1 %3010, label %Flow8214, label %while.body.i.i21.i.i.i.i1985

Flow8214:                                         ; preds = %while.body.i.i21.i.i.i.i1985
  %.lcssa9069 = phi i64 [ %3009, %while.body.i.i21.i.i.i.i1985 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa9069)
  br label %Flow8215, !amdgpu.uniform !42

Flow8216:                                         ; preds = %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1989, %while.cond.i.i.i.i1977
  %3011 = phi i64 [ %add.i.i.i.i1992, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1989 ], [ poison, %while.cond.i.i.i.i1977 ]
  %3012 = phi i32 [ %conv.i.i26.i.i.i.i1990, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1989 ], [ poison, %while.cond.i.i.i.i1977 ]
  %3013 = phi ptr addrspace(4) [ %scevgep6730, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1989 ], [ poison, %while.cond.i.i.i.i1977 ]
  %3014 = phi i64 [ %lsr.iv.next6735, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1989 ], [ poison, %while.cond.i.i.i.i1977 ]
  %3015 = phi i1 [ false, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1989 ], [ true, %while.cond.i.i.i.i1977 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %3000)
  %3016 = call i64 @llvm.amdgcn.if.break.i64(i1 %3015, i64 %phi.broken8694)
  %3017 = call i1 @llvm.amdgcn.loop.i64(i64 %3016)
  br i1 %3017, label %cond.false.i.i.i1999, label %while.cond.i.i.i.i1977

_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i1988: ; preds = %Flow8215
  fence acquire
  br i1 %cmp.i.i.i.i1980, label %if.then.i.i.i.i.i.i.i1993, label %Flow8213, !amdgpu.uniform !42

if.then.i.i.i.i.i.i.i1993:                        ; preds = %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i1988
  %sub.i.i.i.i.i.i.i1995 = sub nuw i64 sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.19.36 to ptr), i64 13) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.19.36 to ptr) to i64)), %idx.0.i.i.i.i1979
  %spec.select.i.i.i36.i.i.i.i1996 = call i64 @llvm.umin.i64(i64 %sub.i.i.i.i.i.i.i1995, i64 64)
  %3018 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %3019 = and i64 %spec.select.i.i.i36.i.i.i.i1996, 15
  %3020 = sub i64 %spec.select.i.i.i36.i.i.i.i1996, %3019
  %3021 = icmp ne i64 %3020, 0
  br i1 %3021, label %loop-memcpy-expansion5325.preheader, label %Flow8212, !amdgpu.uniform !42

loop-memcpy-expansion5325.preheader:              ; preds = %if.then.i.i.i.i.i.i.i1993
  %arrayidx.i20.i35.i.i.i.i1997 = getelementptr %"struct.rpc::Buffer", ptr %3018, i64 %idxprom.i19.i.i.i.i.i1967
  %arrayidx.i22.i38.i.i.i.i1998 = getelementptr %"struct.rpc::Buffer", ptr %arrayidx.i20.i35.i.i.i.i1997, i64 %sh_prom.i.i.i.i756
  br label %loop-memcpy-expansion5325, !amdgpu.uniform !42

Flow8212:                                         ; preds = %Flow8211, %if.then.i.i.i.i.i.i.i1993
  br label %loop-memcpy-residual-header5328, !amdgpu.uniform !42

loop-memcpy-expansion5325:                        ; preds = %loop-memcpy-expansion5325.preheader, %loop-memcpy-expansion5325
  %loop-index5326 = phi i64 [ %3024, %loop-memcpy-expansion5325 ], [ 0, %loop-memcpy-expansion5325.preheader ]
  %scevgep6731 = getelementptr i8, ptr addrspace(4) %lsr.iv6729, i64 %loop-index5326, !amdgpu.uniform !42
  %3022 = load <4 x i32>, ptr addrspace(4) %scevgep6731, align 1
  %3023 = getelementptr i8, ptr %arrayidx.i22.i38.i.i.i.i1998, i64 %loop-index5326
  store <4 x i32> %3022, ptr %3023, align 1
  %3024 = add i64 %loop-index5326, 16
  %3025 = icmp uge i64 %3024, %3020
  br i1 %3025, label %Flow8211, label %loop-memcpy-expansion5325, !amdgpu.uniform !42

Flow8210:                                         ; preds = %Flow8209, %loop-memcpy-residual-header5328
  br label %Flow8213, !amdgpu.uniform !42

loop-memcpy-residual5327:                         ; preds = %loop-memcpy-residual5327.preheader, %loop-memcpy-residual5327
  %residual-loop-index5329 = phi i64 [ %3027, %loop-memcpy-residual5327 ], [ 0, %loop-memcpy-residual5327.preheader ]
  %scevgep6740 = getelementptr i8, ptr addrspace(4) %scevgep6739, i64 %residual-loop-index5329, !amdgpu.uniform !42
  %3026 = load i8, ptr addrspace(4) %scevgep6740, align 1
  %scevgep6738 = getelementptr i8, ptr %scevgep6737, i64 %residual-loop-index5329
  store i8 %3026, ptr %scevgep6738, align 1
  %3027 = add i64 %residual-loop-index5329, 1
  %3028 = icmp uge i64 %3027, %3019
  br i1 %3028, label %Flow8209, label %loop-memcpy-residual5327, !amdgpu.uniform !42

Flow8209:                                         ; preds = %loop-memcpy-residual5327
  br label %Flow8210, !amdgpu.uniform !42

_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1989: ; preds = %Flow8213
  %conv.i.i26.i.i.i.i1990 = xor i32 %port.sroa.37.0.i.i1978, 1
  fence release
  %3029 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %arrayidx.i24.i29.i.i.i.i1991 = getelementptr inbounds nuw i32, ptr %3029, i64 %idxprom.i23.i.i.i.i.i1974
  store atomic i32 %conv.i.i26.i.i.i.i1990, ptr %arrayidx.i24.i29.i.i.i.i1991 monotonic, align 4
  %add.i.i.i.i1992 = add i64 %idx.0.i.i.i.i1979, 64
  %scevgep6730 = getelementptr i8, ptr addrspace(4) %lsr.iv6729, i64 64
  %lsr.iv.next6735 = add i64 %lsr.iv6734, -64
  br label %Flow8216, !amdgpu.uniform !42

cond.false.i.i.i1999:                             ; preds = %Flow8216
  %.lcssa9071 = phi i64 [ %3016, %Flow8216 ]
  %port.sroa.37.0.i.i1978.lcssa = phi i32 [ %port.sroa.37.0.i.i1978, %Flow8216 ]
  %.lcssa9070 = phi i32 [ %2997, %Flow8216 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa9071)
  %3030 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %.lcssa9070)
  %3031 = icmp ne i32 %3030, %port.sroa.37.0.i.i1978.lcssa
  %3032 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3031)
  %3033 = extractvalue { i1, i64 } %3032, 0
  %3034 = extractvalue { i1, i64 } %3032, 1
  br i1 %3033, label %while.body.i.i.i.i2000.preheader, label %Flow8208

while.body.i.i.i.i2000.preheader:                 ; preds = %cond.false.i.i.i1999
  br label %while.body.i.i.i.i2000, !amdgpu.uniform !42

Flow8208:                                         ; preds = %Flow8207, %cond.false.i.i.i1999
  call void @llvm.amdgcn.end.cf.i64(i64 %3034)
  br label %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit2021, !amdgpu.uniform !42

while.body.i.i.i.i2000:                           ; preds = %while.body.i.i.i.i2000.preheader, %while.body.i.i.i.i2000
  %phi.broken8695 = phi i64 [ 0, %while.body.i.i.i.i2000.preheader ], [ %3038, %while.body.i.i.i.i2000 ]
  call void @llvm.amdgcn.s.sleep(i32 2)
  %3035 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i8.i.i2001 = getelementptr inbounds nuw i32, ptr %3035, i64 %idxprom.i23.i.i.i.i.i1974
  %3036 = load atomic i32, ptr %arrayidx.i.i.i8.i.i2001 monotonic, align 4
  %3037 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %3036)
  %cmp.i.not.i.i.i.i2002 = icmp eq i32 %3037, %port.sroa.37.0.i.i1978.lcssa
  %3038 = call i64 @llvm.amdgcn.if.break.i64(i1 %cmp.i.not.i.i.i.i2002, i64 %phi.broken8695)
  %3039 = call i1 @llvm.amdgcn.loop.i64(i64 %3038)
  br i1 %3039, label %Flow8207, label %while.body.i.i.i.i2000

Flow8207:                                         ; preds = %while.body.i.i.i.i2000
  %.lcssa9072 = phi i64 [ %3038, %while.body.i.i.i.i2000 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa9072)
  br label %Flow8208, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit2021: ; preds = %Flow8208
  fence acquire
  call void @llvm.amdgcn.wave.barrier()
  fence syncscope("agent") release
  %3040 = call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %.lcssa9067, i1 true)
  %iszero.i.i.i.i.i13.i.i2004 = icmp ne i64 %.lcssa9067, 0
  %cmp2.i.i.i.i14.i.i2005 = icmp eq i64 %3040, %sh_prom.i.i.i.i756
  %cmp.i.i.i.i15.i.i2006 = select i1 %iszero.i.i.i.i.i13.i.i2004, i1 %cmp2.i.i.i.i14.i.i2005, i1 false
  %3041 = and i32 %.lcssa9068, 31
  %conv4.i.i.i19.i.i2010 = zext i1 %cmp.i.i.i.i15.i.i2006 to i32
  %shl.i.i.i20.i.i2011 = shl nuw i32 %conv4.i.i.i19.i.i2010, %3041
  %xor.i.i.i21.i.i2012 = xor i32 %shl.i.i.i20.i.i2011, -1
  %3042 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop5719, !amdgpu.uniform !42

3043:                                             ; preds = %ComputeEnd5720
  %div8.i.i.i16.i.i2007 = lshr i32 %.lcssa9068, 5
  %idxprom.i.i.i17.i.i2008 = zext nneg i32 %div8.i.i.i16.i.i2007 to i64
  %3044 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i17.i.i2008
  %sunkaddr7425 = getelementptr inbounds i8, ptr addrspace(1) %3044, i64 40
  %3045 = atomicrmw and ptr addrspace(1) %sunkaddr7425, i32 %.lcssa9073 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %3046, !amdgpu.uniform !42

3046:                                             ; preds = %3043, %ComputeEnd5720
  call void @llvm.amdgcn.end.cf.i64(i64 %7768)
  call void @llvm.amdgcn.wave.barrier()
  br label %for.cond.i.i.i2022, !amdgpu.uniform !42

for.cond.i.i.i2022:                               ; preds = %3046, %for.cond.i.i.i2022
  %phi.broken8696 = phi i64 [ 0, %3046 ], [ %3048, %for.cond.i.i.i2022 ]
  %lsr.iv6741 = phi ptr addrspace(4) [ %scevgep6743, %for.cond.i.i.i2022 ], [ %scevgep6634, %3046 ]
  %scevgep6743 = getelementptr i8, ptr addrspace(4) %lsr.iv6741, i64 1
  %3047 = load i8, ptr addrspace(4) %scevgep6743, align 1, !tbaa !57
  %cmp.i.i.i2024 = icmp eq i8 %3047, 0
  %3048 = call i64 @llvm.amdgcn.if.break.i64(i1 %cmp.i.i.i2024, i64 %phi.broken8696)
  %3049 = call i1 @llvm.amdgcn.loop.i64(i64 %3048)
  br i1 %3049, label %for.cond.i.i1.i2031.preheader, label %for.cond.i.i.i2022

for.cond.i.i1.i2031.preheader:                    ; preds = %for.cond.i.i.i2022
  %scevgep6743.lcssa = phi ptr addrspace(4) [ %scevgep6743, %for.cond.i.i.i2022 ]
  %.lcssa9074 = phi i64 [ %3048, %for.cond.i.i.i2022 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa9074)
  br label %for.cond.i.i1.i2031, !amdgpu.uniform !42

for.cond.i.i1.i2031:                              ; preds = %for.cond.i.i1.i2031.preheader, %Flow8206
  %3050 = phi i32 [ %3087, %Flow8206 ], [ %1186, %for.cond.i.i1.i2031.preheader ]
  %3051 = phi i64 [ %3088, %Flow8206 ], [ %1187, %for.cond.i.i1.i2031.preheader ]
  %3052 = phi i32 [ %3089, %Flow8206 ], [ %1188, %for.cond.i.i1.i2031.preheader ]
  %retval.sroa.6.0.i.i.i2032 = phi i32 [ %3092, %Flow8206 ], [ undef, %for.cond.i.i1.i2031.preheader ]
  %retval.sroa.2.0.i.i.i2034 = phi i64 [ %3091, %Flow8206 ], [ undef, %for.cond.i.i1.i2031.preheader ]
  %retval.sroa.8.0.i.i.i2035 = phi i32 [ %3090, %Flow8206 ], [ undef, %for.cond.i.i1.i2031.preheader ]
  %index.0.i.i.i2036 = phi i32 [ %3094, %Flow8206 ], [ 0, %for.cond.i.i1.i2031.preheader ]
  %3053 = load i32, ptr addrspace(1) @__llvm_rpc_client, align 8, !tbaa !68
  %cmp.not.i.i.i2037 = icmp ult i32 %index.0.i.i.i2036, %3053
  %spec.store.select.i.i.i2038 = select i1 %cmp.not.i.i.i2037, i32 %index.0.i.i.i2036, i32 0
  %3054 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  %3055 = and i32 %spec.store.select.i.i.i2038, 31
  %3056 = lshr i64 %3054, %sh_prom.i.i.i.i756
  %3057 = trunc i64 %3056 to i32
  %conv4.i.i.i.i.i2042 = and i32 %3057, 1
  %shl.i.i.i.i.i2043 = shl nuw i32 %conv4.i.i.i.i.i2042, %3055
  br label %ComputeLoop5723, !amdgpu.uniform !42

3058:                                             ; preds = %ComputeEnd5724
  %sunkaddr7426 = getelementptr inbounds i8, ptr addrspace(1) %7778, i64 40
  %3059 = atomicrmw or ptr addrspace(1) %sunkaddr7426, i32 %.lcssa9075 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %3060, !amdgpu.uniform !42

3060:                                             ; preds = %3058, %ComputeEnd5724
  %3061 = phi i32 [ %3059, %3058 ], [ poison, %ComputeEnd5724 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %7787)
  %3062 = call i32 @llvm.amdgcn.readfirstlane.i32(i32 %3061)
  %3063 = or i32 %3062, %.lcssa9076
  %shl5.i.i.i.i.i2044 = shl nuw i32 1, %3055
  %and.i.i.i.i.i2045 = and i32 %shl5.i.i.i.i.i2044, %3063
  %tobool3.i.i.i.i2046 = icmp ne i32 %and.i.i.i.i.i2045, 0
  %3064 = call i64 @llvm.amdgcn.ballot.i64(i1 %tobool3.i.i.i.i2046)
  %and.i.i.i.i.i.i2047 = and i64 %3064, %3054
  %cmp.i.not.i.i.i2048 = icmp ne i64 %3054, %and.i.i.i.i.i.i2047
  br i1 %cmp.i.not.i.i.i2048, label %if.then.i.i.i.i2049, label %Flow8205, !amdgpu.uniform !42

if.then.i.i.i.i2049:                              ; preds = %3060
  fence syncscope("agent") acquire
  %3065 = load <2 x i64>, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !75
  %3066 = extractelement <2 x i64> %3065, i32 0
  %3067 = inttoptr i64 %3066 to ptr
  %3068 = extractelement <2 x i64> %3065, i32 1
  %3069 = inttoptr i64 %3068 to ptr
  %idxprom.i.i.i.i2050 = zext i32 %spec.store.select.i.i.i2038 to i64
  %arrayidx.i.i.i.i2051 = getelementptr inbounds nuw i32, ptr %3067, i64 %idxprom.i.i.i.i2050, !amdgpu.uniform !42
  %3070 = load atomic i32, ptr %arrayidx.i.i.i.i2051 monotonic, align 4
  %3071 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %3070)
  %arrayidx.i45.i.i.i2052 = getelementptr inbounds nuw i32, ptr %3069, i64 %idxprom.i.i.i.i2050, !amdgpu.uniform !42
  %3072 = load atomic i32, ptr %arrayidx.i45.i.i.i2052 monotonic, align 4
  %3073 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %3072)
  %cmp.i46.not.i.i.i2053 = icmp ne i32 %3071, %3073
  br i1 %cmp.i46.not.i.i.i2053, label %if.then12.i.i.i2054, label %Flow8203, !amdgpu.uniform !42

if.then12.i.i.i2054:                              ; preds = %if.then.i.i.i.i2049
  fence syncscope("agent") release
  %3074 = call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %3054, i1 true)
  %iszero.i.i.i.i.i.i.i2055 = icmp ne i64 %3054, 0
  %cmp2.i.i.i.i.i.i2056 = icmp eq i64 %3074, %sh_prom.i.i.i.i756
  %cmp.i.i.i.i.i.i2057 = select i1 %iszero.i.i.i.i.i.i.i2055, i1 %cmp2.i.i.i.i.i.i2056, i1 false
  %conv4.i.i51.i.i.i2058 = zext i1 %cmp.i.i.i.i.i.i2057 to i32
  %shl.i.i52.i.i.i2059 = shl nuw i32 %conv4.i.i51.i.i.i2058, %3055
  %xor.i.i.i.i.i2060 = xor i32 %shl.i.i52.i.i.i2059, -1
  %3075 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop5728, !amdgpu.uniform !42

3076:                                             ; preds = %ComputeEnd5729
  %sunkaddr7427 = getelementptr inbounds i8, ptr addrspace(1) %7778, i64 40
  %3077 = atomicrmw and ptr addrspace(1) %sunkaddr7427, i32 %.lcssa9077 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %3078, !amdgpu.uniform !42

3078:                                             ; preds = %3076, %ComputeEnd5729
  call void @llvm.amdgcn.end.cf.i64(i64 %7805)
  call void @llvm.amdgcn.wave.barrier()
  br label %Flow8203, !amdgpu.uniform !42

if.end14.i.i.i2120:                               ; preds = %Flow8203
  %3079 = call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %3054, i1 true)
  %iszero.i.i.i.i.i.i2121 = icmp ne i64 %3054, 0
  %cmp2.i.i.i.i.i2122 = icmp eq i64 %3079, %sh_prom.i.i.i.i756
  %cmp.i.i.i.i.i2123 = select i1 %iszero.i.i.i.i.i.i2121, i1 %cmp2.i.i.i.i.i2122, i1 false
  %3080 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.i.i.i.i.i2123)
  %3081 = extractvalue { i1, i64 } %3080, 0
  %3082 = extractvalue { i1, i64 } %3080, 1
  br i1 %3081, label %if.then16.i.i.i2125, label %if.end22.i.i.i2124

if.then16.i.i.i2125:                              ; preds = %if.end14.i.i.i2120
  %3083 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !76
  %opcode.i.i.i2126 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %3083, i64 %idxprom.i.i.i.i2050, i32 1
  store i32 1660944387, ptr %opcode.i.i.i2126, align 8, !tbaa !77
  %arrayidx21.i.i.i2127 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %3083, i64 %idxprom.i.i.i.i2050
  store i64 %3054, ptr %arrayidx21.i.i.i2127, align 8, !tbaa !80
  br label %if.end22.i.i.i2124, !amdgpu.uniform !42

Flow8204:                                         ; preds = %if.end22.i.i.i2124, %Flow8203
  %3084 = phi i32 [ %3071, %if.end22.i.i.i2124 ], [ %retval.sroa.8.0.i.i.i2035, %Flow8203 ]
  %3085 = phi i64 [ %3054, %if.end22.i.i.i2124 ], [ %retval.sroa.2.0.i.i.i2034, %Flow8203 ]
  %3086 = phi i32 [ %spec.store.select.i.i.i2038, %if.end22.i.i.i2124 ], [ %retval.sroa.6.0.i.i.i2032, %Flow8203 ]
  br label %cleanup26.i.i.i2061, !amdgpu.uniform !42

if.end22.i.i.i2124:                               ; preds = %if.then16.i.i.i2125, %if.end14.i.i.i2120
  call void @llvm.amdgcn.end.cf.i64(i64 %3082)
  call void @llvm.amdgcn.wave.barrier()
  br label %Flow8204, !amdgpu.uniform !42

Flow8205:                                         ; preds = %cleanup26.i.i.i2061, %3060
  %3087 = phi i32 [ %3086, %cleanup26.i.i.i2061 ], [ %3050, %3060 ]
  %3088 = phi i64 [ %3085, %cleanup26.i.i.i2061 ], [ %3051, %3060 ]
  %3089 = phi i32 [ %3084, %cleanup26.i.i.i2061 ], [ %3052, %3060 ]
  %3090 = phi i32 [ %3084, %cleanup26.i.i.i2061 ], [ %retval.sroa.8.0.i.i.i2035, %3060 ]
  %3091 = phi i64 [ %3085, %cleanup26.i.i.i2061 ], [ %retval.sroa.2.0.i.i.i2034, %3060 ]
  %3092 = phi i32 [ %3086, %cleanup26.i.i.i2061 ], [ %retval.sroa.6.0.i.i.i2032, %3060 ]
  %3093 = phi i1 [ %cmp.i46.not.i.i.i2053, %cleanup26.i.i.i2061 ], [ true, %3060 ]
  br i1 %3093, label %for.inc.i.i.i2066, label %Flow8206, !amdgpu.uniform !42

cleanup26.i.i.i2061:                              ; preds = %Flow8204
  br label %Flow8205, !amdgpu.uniform !42

for.inc.i.i.i2066:                                ; preds = %Flow8205
  %inc.i.i.i2071 = add i32 %spec.store.select.i.i.i2038, 1
  br label %Flow8206, !amdgpu.uniform !42

Flow8206:                                         ; preds = %for.inc.i.i.i2066, %Flow8205
  %3094 = phi i32 [ %inc.i.i.i2071, %for.inc.i.i.i2066 ], [ poison, %Flow8205 ]
  %3095 = phi i1 [ false, %for.inc.i.i.i2066 ], [ true, %Flow8205 ]
  br i1 %3095, label %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i2072, label %for.cond.i.i1.i2031, !amdgpu.uniform !42

_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i2072: ; preds = %Flow8206
  %.lcssa9080 = phi i32 [ %3087, %Flow8206 ]
  %.lcssa9079 = phi i64 [ %3088, %Flow8206 ]
  %.lcssa9078 = phi i32 [ %3089, %Flow8206 ]
  %3096 = addrspacecast ptr addrspace(4) %scevgep6743.lcssa to ptr
  %3097 = addrspacecast ptr addrspace(4) %cond6 to ptr
  %3098 = ptrtoint ptr %3097 to i64
  %sub.ptr.lhs.cast.i.i.i2027 = ptrtoint ptr %3096 to i64
  %sub.ptr.sub.i.i.i2029 = sub i64 %sub.ptr.lhs.cast.i.i.i2027, %3098
  fence acquire
  %3099 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %mul.i.i.i.i.i.i2073 = shl i32 %.lcssa9080, 6
  %idxprom.i19.i.i.i.i.i2074 = zext i32 %mul.i.i.i.i.i.i2073 to i64
  %arrayidx.i20.i.i.i.i.i2075 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %3099, i64 %idxprom.i19.i.i.i.i.i2074
  %arrayidx.i22.i.i.i.i.i2076 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %arrayidx.i20.i.i.i.i.i2075, i64 %sh_prom.i.i.i.i756
  store i64 %sub.ptr.sub.i.i.i2029, ptr %arrayidx.i22.i.i.i.i.i2076, align 8, !tbaa !82
  %spec.select.i.i.i.i.i.i.i2077 = call i64 @llvm.umin.i64(i64 %sub.ptr.sub.i.i.i2029, i64 56)
  %3100 = and i64 %spec.select.i.i.i.i.i.i.i2077, 15
  %3101 = sub i64 %spec.select.i.i.i.i.i.i.i2077, %3100
  %3102 = icmp ne i64 %3101, 0
  %3103 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3102)
  %3104 = extractvalue { i1, i64 } %3103, 0
  %3105 = extractvalue { i1, i64 } %3103, 1
  br i1 %3104, label %loop-memcpy-expansion5199.preheader, label %Flow8202

loop-memcpy-expansion5199.preheader:              ; preds = %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i2072
  %scevgep6744 = getelementptr i8, ptr %3099, i64 8
  %scevgep6745 = getelementptr i8, ptr %scevgep6744, i64 %1474
  %3106 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i2074, 6
  %scevgep6746 = getelementptr i8, ptr %scevgep6745, i64 %3106
  br label %loop-memcpy-expansion5199, !amdgpu.uniform !42

Flow8202:                                         ; preds = %Flow8201, %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i2072
  call void @llvm.amdgcn.end.cf.i64(i64 %3105)
  br label %loop-memcpy-residual-header5202, !amdgpu.uniform !42

loop-memcpy-expansion5199:                        ; preds = %loop-memcpy-expansion5199.preheader, %loop-memcpy-expansion5199
  %phi.broken8697 = phi i64 [ 0, %loop-memcpy-expansion5199.preheader ], [ %3111, %loop-memcpy-expansion5199 ]
  %loop-index5200 = phi i64 [ %3109, %loop-memcpy-expansion5199 ], [ 0, %loop-memcpy-expansion5199.preheader ]
  %3107 = getelementptr i8, ptr addrspace(4) %cond6, i64 %loop-index5200
  %3108 = load <4 x i32>, ptr addrspace(4) %3107, align 1
  %scevgep6747 = getelementptr i8, ptr %scevgep6746, i64 %loop-index5200
  store <4 x i32> %3108, ptr %scevgep6747, align 1
  %3109 = add i64 %loop-index5200, 16
  %3110 = icmp uge i64 %3109, %3101
  %3111 = call i64 @llvm.amdgcn.if.break.i64(i1 %3110, i64 %phi.broken8697)
  %3112 = call i1 @llvm.amdgcn.loop.i64(i64 %3111)
  br i1 %3112, label %Flow8201, label %loop-memcpy-expansion5199

Flow8200:                                         ; preds = %Flow8198, %Flow8199
  %3113 = phi i64 [ %6508, %Flow8198 ], [ %6503, %Flow8199 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %6506)
  br label %post-loop-memcpy-expansion5198, !amdgpu.uniform !42

loop-memcpy-residual5201:                         ; preds = %loop-memcpy-residual5201.preheader, %loop-memcpy-residual5201
  %phi.broken8698 = phi i64 [ 0, %loop-memcpy-residual5201.preheader ], [ %3117, %loop-memcpy-residual5201 ]
  %residual-loop-index5203 = phi i64 [ %3115, %loop-memcpy-residual5201 ], [ 0, %loop-memcpy-residual5201.preheader ]
  %scevgep6753 = getelementptr i8, ptr addrspace(4) %scevgep6752, i64 %residual-loop-index5203
  %3114 = load i8, ptr addrspace(4) %scevgep6753, align 1
  %scevgep6751 = getelementptr i8, ptr %scevgep6750, i64 %residual-loop-index5203
  store i8 %3114, ptr %scevgep6751, align 1
  %3115 = add i64 %residual-loop-index5203, 1
  %3116 = icmp uge i64 %3115, %3100
  %3117 = call i64 @llvm.amdgcn.if.break.i64(i1 %3116, i64 %phi.broken8698)
  %3118 = call i1 @llvm.amdgcn.loop.i64(i64 %3117)
  br i1 %3118, label %Flow8198, label %loop-memcpy-residual5201

Flow8198:                                         ; preds = %loop-memcpy-residual5201
  %.lcssa9082 = phi i64 [ %3117, %loop-memcpy-residual5201 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa9082)
  br label %Flow8200, !amdgpu.uniform !42

post-loop-memcpy-expansion5198:                   ; preds = %Flow8200
  %3119 = addrspacecast ptr addrspace(4) %scevgep6743.lcssa to ptr
  %3120 = ptrtoint ptr %3119 to i64
  %tobool.not.i.i.i.i.i.i2079 = icmp eq i32 %.lcssa9078, 0
  %conv.i.i.i.i.i.i2080 = zext i1 %tobool.not.i.i.i.i.i.i2079 to i32
  fence release
  %3121 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %idxprom.i23.i.i.i.i.i2081 = zext i32 %.lcssa9080 to i64
  %arrayidx.i24.i.i.i.i.i2082 = getelementptr inbounds nuw i32, ptr %3121, i64 %idxprom.i23.i.i.i.i.i2081
  store atomic i32 %conv.i.i.i.i.i.i2080, ptr %arrayidx.i24.i.i.i.i.i2082 monotonic, align 4
  %3122 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !84
  %arrayidx.i.i5.i.i2083 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %3122, i64 %idxprom.i23.i.i.i.i.i2081
  %3123 = load i64, ptr %arrayidx.i.i5.i.i2083, align 8, !tbaa !80
  %3124 = add i64 %1141, %3120
  br label %while.cond.i.i.i.i2084, !amdgpu.uniform !42

while.cond.i.i.i.i2084:                           ; preds = %post-loop-memcpy-expansion5198, %Flow8197
  %phi.broken8702 = phi i64 [ 0, %post-loop-memcpy-expansion5198 ], [ %3146, %Flow8197 ]
  %lsr.iv6759 = phi i64 [ %3144, %Flow8197 ], [ %3124, %post-loop-memcpy-expansion5198 ]
  %lsr.iv6754 = phi ptr addrspace(4) [ %3143, %Flow8197 ], [ %scevgep6648, %post-loop-memcpy-expansion5198 ]
  %port.sroa.37.0.i.i2085 = phi i32 [ %3142, %Flow8197 ], [ %conv.i.i.i.i.i.i2080, %post-loop-memcpy-expansion5198 ]
  %idx.0.i.i.i.i2086 = phi i64 [ %3141, %Flow8197 ], [ 56, %post-loop-memcpy-expansion5198 ]
  %cmp.i.i.i.i2087 = icmp ugt i64 %sub.ptr.sub.i.i.i2029, %idx.0.i.i.i.i2086
  %3125 = call i64 @llvm.amdgcn.ballot.i64(i1 %cmp.i.i.i.i2087)
  %and.i.i.i.i6.i.i2088 = and i64 %3125, %3123
  %tobool.not.i.i.i.i2089 = icmp ne i64 %and.i.i.i.i6.i.i2088, 0
  %3126 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i22.i.i.i2090 = getelementptr inbounds nuw i32, ptr %3126, i64 %idxprom.i23.i.i.i.i.i2081
  %3127 = load atomic i32, ptr %arrayidx.i22.i.i.i2090 monotonic, align 4
  %3128 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %tobool.not.i.i.i.i2089)
  %3129 = extractvalue { i1, i64 } %3128, 0
  %3130 = extractvalue { i1, i64 } %3128, 1
  br i1 %3129, label %cond.false.i8.i.i.i.i2091, label %Flow8197

cond.false.i8.i.i.i.i2091:                        ; preds = %while.cond.i.i.i.i2084
  %3131 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %3127)
  %3132 = icmp ne i32 %3131, %port.sroa.37.0.i.i2085
  %3133 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3132)
  %3134 = extractvalue { i1, i64 } %3133, 0
  %3135 = extractvalue { i1, i64 } %3133, 1
  br i1 %3134, label %while.body.i.i21.i.i.i.i2092.preheader, label %Flow8196

while.body.i.i21.i.i.i.i2092.preheader:           ; preds = %cond.false.i8.i.i.i.i2091
  br label %while.body.i.i21.i.i.i.i2092, !amdgpu.uniform !42

Flow8196:                                         ; preds = %Flow8195, %cond.false.i8.i.i.i.i2091
  call void @llvm.amdgcn.end.cf.i64(i64 %3135)
  br label %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i2095, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i2092:                     ; preds = %while.body.i.i21.i.i.i.i2092.preheader, %while.body.i.i21.i.i.i.i2092
  %phi.broken8699 = phi i64 [ 0, %while.body.i.i21.i.i.i.i2092.preheader ], [ %3139, %while.body.i.i21.i.i.i.i2092 ]
  call void @llvm.amdgcn.s.sleep(i32 2)
  %3136 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i22.i.i.i.i2093 = getelementptr inbounds nuw i32, ptr %3136, i64 %idxprom.i23.i.i.i.i.i2081
  %3137 = load atomic i32, ptr %arrayidx.i.i.i22.i.i.i.i2093 monotonic, align 4
  %3138 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %3137)
  %cmp.i.not.i.i23.i.i.i.i2094 = icmp eq i32 %3138, %port.sroa.37.0.i.i2085
  %3139 = call i64 @llvm.amdgcn.if.break.i64(i1 %cmp.i.not.i.i23.i.i.i.i2094, i64 %phi.broken8699)
  %3140 = call i1 @llvm.amdgcn.loop.i64(i64 %3139)
  br i1 %3140, label %Flow8195, label %while.body.i.i21.i.i.i.i2092

Flow8195:                                         ; preds = %while.body.i.i21.i.i.i.i2092
  %.lcssa9083 = phi i64 [ %3139, %while.body.i.i21.i.i.i.i2092 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa9083)
  br label %Flow8196, !amdgpu.uniform !42

Flow8197:                                         ; preds = %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i2096, %while.cond.i.i.i.i2084
  %3141 = phi i64 [ %add.i.i.i.i2099, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i2096 ], [ poison, %while.cond.i.i.i.i2084 ]
  %3142 = phi i32 [ %conv.i.i26.i.i.i.i2097, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i2096 ], [ poison, %while.cond.i.i.i.i2084 ]
  %3143 = phi ptr addrspace(4) [ %scevgep6755, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i2096 ], [ poison, %while.cond.i.i.i.i2084 ]
  %3144 = phi i64 [ %lsr.iv.next6760, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i2096 ], [ poison, %while.cond.i.i.i.i2084 ]
  %3145 = phi i1 [ false, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i2096 ], [ true, %while.cond.i.i.i.i2084 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %3130)
  %3146 = call i64 @llvm.amdgcn.if.break.i64(i1 %3145, i64 %phi.broken8702)
  %3147 = call i1 @llvm.amdgcn.loop.i64(i64 %3146)
  br i1 %3147, label %cond.false.i.i.i2106, label %while.cond.i.i.i.i2084

_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i2095: ; preds = %Flow8196
  fence acquire
  %3148 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.i.i.i.i2087)
  %3149 = extractvalue { i1, i64 } %3148, 0
  %3150 = extractvalue { i1, i64 } %3148, 1
  br i1 %3149, label %if.then.i.i.i.i.i.i.i2100, label %Flow8194

if.then.i.i.i.i.i.i.i2100:                        ; preds = %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i2095
  %sub.i.i.i.i.i.i.i2102 = sub nuw i64 %sub.ptr.sub.i.i.i2029, %idx.0.i.i.i.i2086
  %spec.select.i.i.i36.i.i.i.i2103 = call i64 @llvm.umin.i64(i64 %sub.i.i.i.i.i.i.i2102, i64 64)
  %3151 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %3152 = and i64 %spec.select.i.i.i36.i.i.i.i2103, 15
  %3153 = sub i64 %spec.select.i.i.i36.i.i.i.i2103, %3152
  %3154 = icmp ne i64 %3153, 0
  %3155 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3154)
  %3156 = extractvalue { i1, i64 } %3155, 0
  %3157 = extractvalue { i1, i64 } %3155, 1
  br i1 %3156, label %loop-memcpy-expansion5331.preheader, label %Flow8193

loop-memcpy-expansion5331.preheader:              ; preds = %if.then.i.i.i.i.i.i.i2100
  %arrayidx.i20.i35.i.i.i.i2104 = getelementptr %"struct.rpc::Buffer", ptr %3151, i64 %idxprom.i19.i.i.i.i.i2074
  %arrayidx.i22.i38.i.i.i.i2105 = getelementptr %"struct.rpc::Buffer", ptr %arrayidx.i20.i35.i.i.i.i2104, i64 %sh_prom.i.i.i.i756
  br label %loop-memcpy-expansion5331, !amdgpu.uniform !42

Flow8193:                                         ; preds = %Flow8192, %if.then.i.i.i.i.i.i.i2100
  call void @llvm.amdgcn.end.cf.i64(i64 %3157)
  br label %loop-memcpy-residual-header5334, !amdgpu.uniform !42

loop-memcpy-expansion5331:                        ; preds = %loop-memcpy-expansion5331.preheader, %loop-memcpy-expansion5331
  %phi.broken8700 = phi i64 [ 0, %loop-memcpy-expansion5331.preheader ], [ %3162, %loop-memcpy-expansion5331 ]
  %loop-index5332 = phi i64 [ %3160, %loop-memcpy-expansion5331 ], [ 0, %loop-memcpy-expansion5331.preheader ]
  %scevgep6756 = getelementptr i8, ptr addrspace(4) %lsr.iv6754, i64 %loop-index5332
  %3158 = load <4 x i32>, ptr addrspace(4) %scevgep6756, align 1
  %3159 = getelementptr i8, ptr %arrayidx.i22.i38.i.i.i.i2105, i64 %loop-index5332
  store <4 x i32> %3158, ptr %3159, align 1
  %3160 = add i64 %loop-index5332, 16
  %3161 = icmp uge i64 %3160, %3153
  %3162 = call i64 @llvm.amdgcn.if.break.i64(i1 %3161, i64 %phi.broken8700)
  %3163 = call i1 @llvm.amdgcn.loop.i64(i64 %3162)
  br i1 %3163, label %Flow8192, label %loop-memcpy-expansion5331

Flow8191:                                         ; preds = %Flow8190, %loop-memcpy-residual-header5334
  call void @llvm.amdgcn.end.cf.i64(i64 %6606)
  br label %Flow8194, !amdgpu.uniform !42

loop-memcpy-residual5333:                         ; preds = %loop-memcpy-residual5333.preheader, %loop-memcpy-residual5333
  %phi.broken8701 = phi i64 [ 0, %loop-memcpy-residual5333.preheader ], [ %3167, %loop-memcpy-residual5333 ]
  %residual-loop-index5335 = phi i64 [ %3165, %loop-memcpy-residual5333 ], [ 0, %loop-memcpy-residual5333.preheader ]
  %scevgep6765 = getelementptr i8, ptr addrspace(4) %scevgep6764, i64 %residual-loop-index5335
  %3164 = load i8, ptr addrspace(4) %scevgep6765, align 1
  %scevgep6763 = getelementptr i8, ptr %scevgep6762, i64 %residual-loop-index5335
  store i8 %3164, ptr %scevgep6763, align 1
  %3165 = add i64 %residual-loop-index5335, 1
  %3166 = icmp uge i64 %3165, %3152
  %3167 = call i64 @llvm.amdgcn.if.break.i64(i1 %3166, i64 %phi.broken8701)
  %3168 = call i1 @llvm.amdgcn.loop.i64(i64 %3167)
  br i1 %3168, label %Flow8190, label %loop-memcpy-residual5333

Flow8190:                                         ; preds = %loop-memcpy-residual5333
  %.lcssa9085 = phi i64 [ %3167, %loop-memcpy-residual5333 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa9085)
  br label %Flow8191, !amdgpu.uniform !42

_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i2096: ; preds = %Flow8194
  %conv.i.i26.i.i.i.i2097 = xor i32 %port.sroa.37.0.i.i2085, 1
  fence release
  %3169 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %arrayidx.i24.i29.i.i.i.i2098 = getelementptr inbounds nuw i32, ptr %3169, i64 %idxprom.i23.i.i.i.i.i2081
  store atomic i32 %conv.i.i26.i.i.i.i2097, ptr %arrayidx.i24.i29.i.i.i.i2098 monotonic, align 4
  %add.i.i.i.i2099 = add i64 %idx.0.i.i.i.i2086, 64
  %scevgep6755 = getelementptr i8, ptr addrspace(4) %lsr.iv6754, i64 64
  %lsr.iv.next6760 = add i64 %lsr.iv6759, -64
  br label %Flow8197, !amdgpu.uniform !42

cond.false.i.i.i2106:                             ; preds = %Flow8197
  %.lcssa9087 = phi i64 [ %3146, %Flow8197 ]
  %port.sroa.37.0.i.i2085.lcssa = phi i32 [ %port.sroa.37.0.i.i2085, %Flow8197 ]
  %.lcssa9086 = phi i32 [ %3127, %Flow8197 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa9087)
  %3170 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %.lcssa9086)
  %3171 = icmp ne i32 %3170, %port.sroa.37.0.i.i2085.lcssa
  %3172 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3171)
  %3173 = extractvalue { i1, i64 } %3172, 0
  %3174 = extractvalue { i1, i64 } %3172, 1
  br i1 %3173, label %while.body.i.i.i.i2107.preheader, label %Flow8189

while.body.i.i.i.i2107.preheader:                 ; preds = %cond.false.i.i.i2106
  br label %while.body.i.i.i.i2107, !amdgpu.uniform !42

Flow8189:                                         ; preds = %Flow8188, %cond.false.i.i.i2106
  call void @llvm.amdgcn.end.cf.i64(i64 %3174)
  br label %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit2128, !amdgpu.uniform !42

while.body.i.i.i.i2107:                           ; preds = %while.body.i.i.i.i2107.preheader, %while.body.i.i.i.i2107
  %phi.broken8703 = phi i64 [ 0, %while.body.i.i.i.i2107.preheader ], [ %3178, %while.body.i.i.i.i2107 ]
  call void @llvm.amdgcn.s.sleep(i32 2)
  %3175 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i8.i.i2108 = getelementptr inbounds nuw i32, ptr %3175, i64 %idxprom.i23.i.i.i.i.i2081
  %3176 = load atomic i32, ptr %arrayidx.i.i.i8.i.i2108 monotonic, align 4
  %3177 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %3176)
  %cmp.i.not.i.i.i.i2109 = icmp eq i32 %3177, %port.sroa.37.0.i.i2085.lcssa
  %3178 = call i64 @llvm.amdgcn.if.break.i64(i1 %cmp.i.not.i.i.i.i2109, i64 %phi.broken8703)
  %3179 = call i1 @llvm.amdgcn.loop.i64(i64 %3178)
  br i1 %3179, label %Flow8188, label %while.body.i.i.i.i2107

Flow8188:                                         ; preds = %while.body.i.i.i.i2107
  %.lcssa9088 = phi i64 [ %3178, %while.body.i.i.i.i2107 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa9088)
  br label %Flow8189, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit2128: ; preds = %Flow8189
  fence acquire
  call void @llvm.amdgcn.wave.barrier()
  fence syncscope("agent") release
  %3180 = call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %.lcssa9079, i1 true)
  %iszero.i.i.i.i.i13.i.i2111 = icmp ne i64 %.lcssa9079, 0
  %cmp2.i.i.i.i14.i.i2112 = icmp eq i64 %3180, %sh_prom.i.i.i.i756
  %cmp.i.i.i.i15.i.i2113 = select i1 %iszero.i.i.i.i.i13.i.i2111, i1 %cmp2.i.i.i.i14.i.i2112, i1 false
  %3181 = and i32 %.lcssa9080, 31
  %conv4.i.i.i19.i.i2117 = zext i1 %cmp.i.i.i.i15.i.i2113 to i32
  %shl.i.i.i20.i.i2118 = shl nuw i32 %conv4.i.i.i19.i.i2117, %3181
  %xor.i.i.i21.i.i2119 = xor i32 %shl.i.i.i20.i.i2118, -1
  %3182 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop5732, !amdgpu.uniform !42

3183:                                             ; preds = %ComputeEnd5733
  %div8.i.i.i16.i.i2114 = lshr i32 %.lcssa9080, 5
  %idxprom.i.i.i17.i.i2115 = zext nneg i32 %div8.i.i.i16.i.i2114 to i64
  %3184 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i17.i.i2115
  %sunkaddr7428 = getelementptr inbounds i8, ptr addrspace(1) %3184, i64 40
  %3185 = atomicrmw and ptr addrspace(1) %sunkaddr7428, i32 %.lcssa9089 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %3186, !amdgpu.uniform !42

3186:                                             ; preds = %3183, %ComputeEnd5733
  call void @llvm.amdgcn.end.cf.i64(i64 %7822)
  call void @llvm.amdgcn.wave.barrier()
  br label %for.cond.i.i.i2129, !amdgpu.uniform !42

for.cond.i.i.i2129:                               ; preds = %3186, %for.cond.i.i.i2129
  %phi.broken8704 = phi i64 [ 0, %3186 ], [ %3188, %for.cond.i.i.i2129 ]
  %lsr.iv6788 = phi i64 [ %lsr.iv.next6789, %for.cond.i.i.i2129 ], [ -56, %3186 ]
  %lsr.iv6775 = phi i64 [ %lsr.iv.next6776, %for.cond.i.i.i2129 ], [ -1, %3186 ]
  %lsr.iv6766 = phi ptr [ %scevgep6768, %for.cond.i.i.i2129 ], [ %scevgep6661, %3186 ]
  %scevgep6768 = getelementptr i8, ptr %lsr.iv6766, i64 1
  %3187 = load i8, ptr %scevgep6768, align 1, !tbaa !57
  %cmp.i.i.i2131 = icmp eq i8 %3187, 0
  %lsr.iv.next6776 = add i64 %lsr.iv6775, 1
  %lsr.iv.next6789 = add i64 %lsr.iv6788, 1
  %3188 = call i64 @llvm.amdgcn.if.break.i64(i1 %cmp.i.i.i2131, i64 %phi.broken8704)
  %3189 = call i1 @llvm.amdgcn.loop.i64(i64 %3188)
  br i1 %3189, label %for.cond.i.i1.i2138.preheader, label %for.cond.i.i.i2129

for.cond.i.i1.i2138.preheader:                    ; preds = %for.cond.i.i.i2129
  %lsr.iv6788.lcssa = phi i64 [ %lsr.iv6788, %for.cond.i.i.i2129 ]
  %scevgep6768.lcssa = phi ptr [ %scevgep6768, %for.cond.i.i.i2129 ]
  %lsr.iv.next6776.lcssa = phi i64 [ %lsr.iv.next6776, %for.cond.i.i.i2129 ]
  %.lcssa9090 = phi i64 [ %3188, %for.cond.i.i.i2129 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa9090)
  br label %for.cond.i.i1.i2138, !amdgpu.uniform !42

for.cond.i.i1.i2138:                              ; preds = %for.cond.i.i1.i2138.preheader, %Flow8187
  %3190 = phi i32 [ %3227, %Flow8187 ], [ %1189, %for.cond.i.i1.i2138.preheader ]
  %3191 = phi i64 [ %3228, %Flow8187 ], [ %1190, %for.cond.i.i1.i2138.preheader ]
  %3192 = phi i32 [ %3229, %Flow8187 ], [ %1191, %for.cond.i.i1.i2138.preheader ]
  %retval.sroa.6.0.i.i.i2139 = phi i32 [ %3232, %Flow8187 ], [ undef, %for.cond.i.i1.i2138.preheader ]
  %retval.sroa.2.0.i.i.i2141 = phi i64 [ %3231, %Flow8187 ], [ undef, %for.cond.i.i1.i2138.preheader ]
  %retval.sroa.8.0.i.i.i2142 = phi i32 [ %3230, %Flow8187 ], [ undef, %for.cond.i.i1.i2138.preheader ]
  %index.0.i.i.i2143 = phi i32 [ %3234, %Flow8187 ], [ 0, %for.cond.i.i1.i2138.preheader ]
  %3193 = load i32, ptr addrspace(1) @__llvm_rpc_client, align 8, !tbaa !68
  %cmp.not.i.i.i2144 = icmp ult i32 %index.0.i.i.i2143, %3193
  %spec.store.select.i.i.i2145 = select i1 %cmp.not.i.i.i2144, i32 %index.0.i.i.i2143, i32 0
  %3194 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  %3195 = and i32 %spec.store.select.i.i.i2145, 31
  %3196 = lshr i64 %3194, %sh_prom.i.i.i.i756
  %3197 = trunc i64 %3196 to i32
  %conv4.i.i.i.i.i2149 = and i32 %3197, 1
  %shl.i.i.i.i.i2150 = shl nuw i32 %conv4.i.i.i.i.i2149, %3195
  br label %ComputeLoop5736, !amdgpu.uniform !42

3198:                                             ; preds = %ComputeEnd5737
  %sunkaddr7429 = getelementptr inbounds i8, ptr addrspace(1) %7832, i64 40
  %3199 = atomicrmw or ptr addrspace(1) %sunkaddr7429, i32 %.lcssa9091 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %3200, !amdgpu.uniform !42

3200:                                             ; preds = %3198, %ComputeEnd5737
  %3201 = phi i32 [ %3199, %3198 ], [ poison, %ComputeEnd5737 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %7841)
  %3202 = call i32 @llvm.amdgcn.readfirstlane.i32(i32 %3201)
  %3203 = or i32 %3202, %.lcssa9092
  %shl5.i.i.i.i.i2151 = shl nuw i32 1, %3195
  %and.i.i.i.i.i2152 = and i32 %shl5.i.i.i.i.i2151, %3203
  %tobool3.i.i.i.i2153 = icmp ne i32 %and.i.i.i.i.i2152, 0
  %3204 = call i64 @llvm.amdgcn.ballot.i64(i1 %tobool3.i.i.i.i2153)
  %and.i.i.i.i.i.i2154 = and i64 %3204, %3194
  %cmp.i.not.i.i.i2155 = icmp ne i64 %3194, %and.i.i.i.i.i.i2154
  br i1 %cmp.i.not.i.i.i2155, label %if.then.i.i.i.i2156, label %Flow8186, !amdgpu.uniform !42

if.then.i.i.i.i2156:                              ; preds = %3200
  fence syncscope("agent") acquire
  %3205 = load <2 x i64>, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !75
  %3206 = extractelement <2 x i64> %3205, i32 0
  %3207 = inttoptr i64 %3206 to ptr
  %3208 = extractelement <2 x i64> %3205, i32 1
  %3209 = inttoptr i64 %3208 to ptr
  %idxprom.i.i.i.i2157 = zext i32 %spec.store.select.i.i.i2145 to i64
  %arrayidx.i.i.i.i2158 = getelementptr inbounds nuw i32, ptr %3207, i64 %idxprom.i.i.i.i2157, !amdgpu.uniform !42
  %3210 = load atomic i32, ptr %arrayidx.i.i.i.i2158 monotonic, align 4
  %3211 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %3210)
  %arrayidx.i45.i.i.i2159 = getelementptr inbounds nuw i32, ptr %3209, i64 %idxprom.i.i.i.i2157, !amdgpu.uniform !42
  %3212 = load atomic i32, ptr %arrayidx.i45.i.i.i2159 monotonic, align 4
  %3213 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %3212)
  %cmp.i46.not.i.i.i2160 = icmp ne i32 %3211, %3213
  br i1 %cmp.i46.not.i.i.i2160, label %if.then12.i.i.i2161, label %Flow8184, !amdgpu.uniform !42

if.then12.i.i.i2161:                              ; preds = %if.then.i.i.i.i2156
  fence syncscope("agent") release
  %3214 = call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %3194, i1 true)
  %iszero.i.i.i.i.i.i.i2162 = icmp ne i64 %3194, 0
  %cmp2.i.i.i.i.i.i2163 = icmp eq i64 %3214, %sh_prom.i.i.i.i756
  %cmp.i.i.i.i.i.i2164 = select i1 %iszero.i.i.i.i.i.i.i2162, i1 %cmp2.i.i.i.i.i.i2163, i1 false
  %conv4.i.i51.i.i.i2165 = zext i1 %cmp.i.i.i.i.i.i2164 to i32
  %shl.i.i52.i.i.i2166 = shl nuw i32 %conv4.i.i51.i.i.i2165, %3195
  %xor.i.i.i.i.i2167 = xor i32 %shl.i.i52.i.i.i2166, -1
  %3215 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop5741, !amdgpu.uniform !42

3216:                                             ; preds = %ComputeEnd5742
  %sunkaddr7430 = getelementptr inbounds i8, ptr addrspace(1) %7832, i64 40
  %3217 = atomicrmw and ptr addrspace(1) %sunkaddr7430, i32 %.lcssa9093 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %3218, !amdgpu.uniform !42

3218:                                             ; preds = %3216, %ComputeEnd5742
  call void @llvm.amdgcn.end.cf.i64(i64 %7859)
  call void @llvm.amdgcn.wave.barrier()
  br label %Flow8184, !amdgpu.uniform !42

if.end14.i.i.i2227:                               ; preds = %Flow8184
  %3219 = call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %3194, i1 true)
  %iszero.i.i.i.i.i.i2228 = icmp ne i64 %3194, 0
  %cmp2.i.i.i.i.i2229 = icmp eq i64 %3219, %sh_prom.i.i.i.i756
  %cmp.i.i.i.i.i2230 = select i1 %iszero.i.i.i.i.i.i2228, i1 %cmp2.i.i.i.i.i2229, i1 false
  %3220 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.i.i.i.i.i2230)
  %3221 = extractvalue { i1, i64 } %3220, 0
  %3222 = extractvalue { i1, i64 } %3220, 1
  br i1 %3221, label %if.then16.i.i.i2232, label %if.end22.i.i.i2231

if.then16.i.i.i2232:                              ; preds = %if.end14.i.i.i2227
  %3223 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !76
  %opcode.i.i.i2233 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %3223, i64 %idxprom.i.i.i.i2157, i32 1
  store i32 1660944387, ptr %opcode.i.i.i2233, align 8, !tbaa !77
  %arrayidx21.i.i.i2234 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %3223, i64 %idxprom.i.i.i.i2157
  store i64 %3194, ptr %arrayidx21.i.i.i2234, align 8, !tbaa !80
  br label %if.end22.i.i.i2231, !amdgpu.uniform !42

Flow8185:                                         ; preds = %if.end22.i.i.i2231, %Flow8184
  %3224 = phi i32 [ %3211, %if.end22.i.i.i2231 ], [ %retval.sroa.8.0.i.i.i2142, %Flow8184 ]
  %3225 = phi i64 [ %3194, %if.end22.i.i.i2231 ], [ %retval.sroa.2.0.i.i.i2141, %Flow8184 ]
  %3226 = phi i32 [ %spec.store.select.i.i.i2145, %if.end22.i.i.i2231 ], [ %retval.sroa.6.0.i.i.i2139, %Flow8184 ]
  br label %cleanup26.i.i.i2168, !amdgpu.uniform !42

if.end22.i.i.i2231:                               ; preds = %if.then16.i.i.i2232, %if.end14.i.i.i2227
  call void @llvm.amdgcn.end.cf.i64(i64 %3222)
  call void @llvm.amdgcn.wave.barrier()
  br label %Flow8185, !amdgpu.uniform !42

Flow8186:                                         ; preds = %cleanup26.i.i.i2168, %3200
  %3227 = phi i32 [ %3226, %cleanup26.i.i.i2168 ], [ %3190, %3200 ]
  %3228 = phi i64 [ %3225, %cleanup26.i.i.i2168 ], [ %3191, %3200 ]
  %3229 = phi i32 [ %3224, %cleanup26.i.i.i2168 ], [ %3192, %3200 ]
  %3230 = phi i32 [ %3224, %cleanup26.i.i.i2168 ], [ %retval.sroa.8.0.i.i.i2142, %3200 ]
  %3231 = phi i64 [ %3225, %cleanup26.i.i.i2168 ], [ %retval.sroa.2.0.i.i.i2141, %3200 ]
  %3232 = phi i32 [ %3226, %cleanup26.i.i.i2168 ], [ %retval.sroa.6.0.i.i.i2139, %3200 ]
  %3233 = phi i1 [ %cmp.i46.not.i.i.i2160, %cleanup26.i.i.i2168 ], [ true, %3200 ]
  br i1 %3233, label %for.inc.i.i.i2173, label %Flow8187, !amdgpu.uniform !42

cleanup26.i.i.i2168:                              ; preds = %Flow8185
  br label %Flow8186, !amdgpu.uniform !42

for.inc.i.i.i2173:                                ; preds = %Flow8186
  %inc.i.i.i2178 = add i32 %spec.store.select.i.i.i2145, 1
  br label %Flow8187, !amdgpu.uniform !42

Flow8187:                                         ; preds = %for.inc.i.i.i2173, %Flow8186
  %3234 = phi i32 [ %inc.i.i.i2178, %for.inc.i.i.i2173 ], [ poison, %Flow8186 ]
  %3235 = phi i1 [ false, %for.inc.i.i.i2173 ], [ true, %Flow8186 ]
  br i1 %3235, label %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i2179, label %for.cond.i.i1.i2138, !amdgpu.uniform !42

_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i2179: ; preds = %Flow8187
  %.lcssa9096 = phi i32 [ %3227, %Flow8187 ]
  %.lcssa9095 = phi i64 [ %3228, %Flow8187 ]
  %.lcssa9094 = phi i32 [ %3229, %Flow8187 ]
  %3236 = ptrtoint ptr %call18 to i64
  %sub.ptr.lhs.cast.i.i.i2134 = ptrtoint ptr %scevgep6768.lcssa to i64
  %sub.ptr.sub.i.i.i2136 = sub i64 %sub.ptr.lhs.cast.i.i.i2134, %3236
  fence acquire
  %3237 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %mul.i.i.i.i.i.i2180 = shl i32 %.lcssa9096, 6
  %idxprom.i19.i.i.i.i.i2181 = zext i32 %mul.i.i.i.i.i.i2180 to i64
  %arrayidx.i20.i.i.i.i.i2182 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %3237, i64 %idxprom.i19.i.i.i.i.i2181
  %arrayidx.i22.i.i.i.i.i2183 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %arrayidx.i20.i.i.i.i.i2182, i64 %sh_prom.i.i.i.i756
  store i64 %sub.ptr.sub.i.i.i2136, ptr %arrayidx.i22.i.i.i.i.i2183, align 8, !tbaa !82
  %spec.select.i.i.i.i.i.i.i2184 = call i64 @llvm.umin.i64(i64 %sub.ptr.sub.i.i.i2136, i64 56)
  %3238 = and i64 %spec.select.i.i.i.i.i.i.i2184, 15
  %3239 = sub i64 %spec.select.i.i.i.i.i.i.i2184, %3238
  %3240 = icmp ne i64 %3239, 0
  %3241 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3240)
  %3242 = extractvalue { i1, i64 } %3241, 0
  %3243 = extractvalue { i1, i64 } %3241, 1
  br i1 %3242, label %loop-memcpy-expansion5049.preheader, label %Flow8183

loop-memcpy-expansion5049.preheader:              ; preds = %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i2179
  %scevgep6769 = getelementptr i8, ptr %3237, i64 8
  %scevgep6770 = getelementptr i8, ptr %scevgep6769, i64 %1474
  %3244 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i2181, 6
  %scevgep6771 = getelementptr i8, ptr %scevgep6770, i64 %3244
  br label %loop-memcpy-expansion5049, !amdgpu.uniform !42

Flow8183:                                         ; preds = %Flow8182, %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i2179
  call void @llvm.amdgcn.end.cf.i64(i64 %3243)
  br label %loop-memcpy-residual-header5052, !amdgpu.uniform !42

loop-memcpy-expansion5049:                        ; preds = %loop-memcpy-expansion5049.preheader, %loop-memcpy-expansion5049
  %phi.broken8705 = phi i64 [ 0, %loop-memcpy-expansion5049.preheader ], [ %3249, %loop-memcpy-expansion5049 ]
  %loop-index5050 = phi i64 [ %3247, %loop-memcpy-expansion5049 ], [ 0, %loop-memcpy-expansion5049.preheader ]
  %3245 = getelementptr i8, ptr %call18, i64 %loop-index5050
  %3246 = load <4 x i32>, ptr %3245, align 1
  %scevgep6772 = getelementptr i8, ptr %scevgep6771, i64 %loop-index5050
  store <4 x i32> %3246, ptr %scevgep6772, align 1
  %3247 = add i64 %loop-index5050, 16
  %3248 = icmp uge i64 %3247, %3239
  %3249 = call i64 @llvm.amdgcn.if.break.i64(i1 %3248, i64 %phi.broken8705)
  %3250 = call i1 @llvm.amdgcn.loop.i64(i64 %3249)
  br i1 %3250, label %Flow8182, label %loop-memcpy-expansion5049

Flow8181:                                         ; preds = %Flow8179, %Flow8180
  %3251 = phi i64 [ %6298, %Flow8179 ], [ %6292, %Flow8180 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %6295)
  br label %post-loop-memcpy-expansion5048, !amdgpu.uniform !42

loop-memcpy-residual5051:                         ; preds = %loop-memcpy-residual5051.preheader, %loop-memcpy-residual5051
  %phi.broken8706 = phi i64 [ 0, %loop-memcpy-residual5051.preheader ], [ %3255, %loop-memcpy-residual5051 ]
  %residual-loop-index5053 = phi i64 [ %3253, %loop-memcpy-residual5051 ], [ 0, %loop-memcpy-residual5051.preheader ]
  %scevgep6782 = getelementptr i8, ptr %scevgep6781, i64 %residual-loop-index5053
  %3252 = load i8, ptr %scevgep6782, align 1
  %scevgep6780 = getelementptr i8, ptr %scevgep6779, i64 %residual-loop-index5053
  store i8 %3252, ptr %scevgep6780, align 1
  %3253 = add i64 %residual-loop-index5053, 1
  %3254 = icmp uge i64 %3253, %3238
  %3255 = call i64 @llvm.amdgcn.if.break.i64(i1 %3254, i64 %phi.broken8706)
  %3256 = call i1 @llvm.amdgcn.loop.i64(i64 %3255)
  br i1 %3256, label %Flow8179, label %loop-memcpy-residual5051

Flow8179:                                         ; preds = %loop-memcpy-residual5051
  %.lcssa9098 = phi i64 [ %3255, %loop-memcpy-residual5051 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa9098)
  br label %Flow8181, !amdgpu.uniform !42

post-loop-memcpy-expansion5048:                   ; preds = %Flow8181
  %tobool.not.i.i.i.i.i.i2186 = icmp eq i32 %.lcssa9094, 0
  %conv.i.i.i.i.i.i2187 = zext i1 %tobool.not.i.i.i.i.i.i2186 to i32
  fence release
  %3257 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %idxprom.i23.i.i.i.i.i2188 = zext i32 %.lcssa9096 to i64
  %arrayidx.i24.i.i.i.i.i2189 = getelementptr inbounds nuw i32, ptr %3257, i64 %idxprom.i23.i.i.i.i.i2188
  store atomic i32 %conv.i.i.i.i.i.i2187, ptr %arrayidx.i24.i.i.i.i.i2189 monotonic, align 4
  %3258 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !84
  %arrayidx.i.i5.i.i2190 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %3258, i64 %idxprom.i23.i.i.i.i.i2188
  %3259 = load i64, ptr %arrayidx.i.i5.i.i2190, align 8, !tbaa !80
  br label %while.cond.i.i.i.i2191, !amdgpu.uniform !42

while.cond.i.i.i.i2191:                           ; preds = %post-loop-memcpy-expansion5048, %Flow8178
  %phi.broken8710 = phi i64 [ 0, %post-loop-memcpy-expansion5048 ], [ %3281, %Flow8178 ]
  %lsr.iv6790 = phi i64 [ %3279, %Flow8178 ], [ %lsr.iv6788.lcssa, %post-loop-memcpy-expansion5048 ]
  %lsr.iv6783 = phi ptr [ %3278, %Flow8178 ], [ %scevgep6679, %post-loop-memcpy-expansion5048 ]
  %port.sroa.37.0.i.i2192 = phi i32 [ %3277, %Flow8178 ], [ %conv.i.i.i.i.i.i2187, %post-loop-memcpy-expansion5048 ]
  %idx.0.i.i.i.i2193 = phi i64 [ %3276, %Flow8178 ], [ 56, %post-loop-memcpy-expansion5048 ]
  %cmp.i.i.i.i2194 = icmp ugt i64 %sub.ptr.sub.i.i.i2136, %idx.0.i.i.i.i2193
  %3260 = call i64 @llvm.amdgcn.ballot.i64(i1 %cmp.i.i.i.i2194)
  %and.i.i.i.i6.i.i2195 = and i64 %3260, %3259
  %tobool.not.i.i.i.i2196 = icmp ne i64 %and.i.i.i.i6.i.i2195, 0
  %3261 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i22.i.i.i2197 = getelementptr inbounds nuw i32, ptr %3261, i64 %idxprom.i23.i.i.i.i.i2188
  %3262 = load atomic i32, ptr %arrayidx.i22.i.i.i2197 monotonic, align 4
  %3263 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %tobool.not.i.i.i.i2196)
  %3264 = extractvalue { i1, i64 } %3263, 0
  %3265 = extractvalue { i1, i64 } %3263, 1
  br i1 %3264, label %cond.false.i8.i.i.i.i2198, label %Flow8178

cond.false.i8.i.i.i.i2198:                        ; preds = %while.cond.i.i.i.i2191
  %3266 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %3262)
  %3267 = icmp ne i32 %3266, %port.sroa.37.0.i.i2192
  %3268 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3267)
  %3269 = extractvalue { i1, i64 } %3268, 0
  %3270 = extractvalue { i1, i64 } %3268, 1
  br i1 %3269, label %while.body.i.i21.i.i.i.i2199.preheader, label %Flow8177

while.body.i.i21.i.i.i.i2199.preheader:           ; preds = %cond.false.i8.i.i.i.i2198
  br label %while.body.i.i21.i.i.i.i2199, !amdgpu.uniform !42

Flow8177:                                         ; preds = %Flow8176, %cond.false.i8.i.i.i.i2198
  call void @llvm.amdgcn.end.cf.i64(i64 %3270)
  br label %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i2202, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i2199:                     ; preds = %while.body.i.i21.i.i.i.i2199.preheader, %while.body.i.i21.i.i.i.i2199
  %phi.broken8707 = phi i64 [ 0, %while.body.i.i21.i.i.i.i2199.preheader ], [ %3274, %while.body.i.i21.i.i.i.i2199 ]
  call void @llvm.amdgcn.s.sleep(i32 2)
  %3271 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i22.i.i.i.i2200 = getelementptr inbounds nuw i32, ptr %3271, i64 %idxprom.i23.i.i.i.i.i2188
  %3272 = load atomic i32, ptr %arrayidx.i.i.i22.i.i.i.i2200 monotonic, align 4
  %3273 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %3272)
  %cmp.i.not.i.i23.i.i.i.i2201 = icmp eq i32 %3273, %port.sroa.37.0.i.i2192
  %3274 = call i64 @llvm.amdgcn.if.break.i64(i1 %cmp.i.not.i.i23.i.i.i.i2201, i64 %phi.broken8707)
  %3275 = call i1 @llvm.amdgcn.loop.i64(i64 %3274)
  br i1 %3275, label %Flow8176, label %while.body.i.i21.i.i.i.i2199

Flow8176:                                         ; preds = %while.body.i.i21.i.i.i.i2199
  %.lcssa9099 = phi i64 [ %3274, %while.body.i.i21.i.i.i.i2199 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa9099)
  br label %Flow8177, !amdgpu.uniform !42

Flow8178:                                         ; preds = %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i2203, %while.cond.i.i.i.i2191
  %3276 = phi i64 [ %add.i.i.i.i2206, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i2203 ], [ poison, %while.cond.i.i.i.i2191 ]
  %3277 = phi i32 [ %conv.i.i26.i.i.i.i2204, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i2203 ], [ poison, %while.cond.i.i.i.i2191 ]
  %3278 = phi ptr [ %scevgep6784, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i2203 ], [ poison, %while.cond.i.i.i.i2191 ]
  %3279 = phi i64 [ %lsr.iv.next6791, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i2203 ], [ poison, %while.cond.i.i.i.i2191 ]
  %3280 = phi i1 [ false, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i2203 ], [ true, %while.cond.i.i.i.i2191 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %3265)
  %3281 = call i64 @llvm.amdgcn.if.break.i64(i1 %3280, i64 %phi.broken8710)
  %3282 = call i1 @llvm.amdgcn.loop.i64(i64 %3281)
  br i1 %3282, label %cond.false.i.i.i2213, label %while.cond.i.i.i.i2191

_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i2202: ; preds = %Flow8177
  fence acquire
  %3283 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.i.i.i.i2194)
  %3284 = extractvalue { i1, i64 } %3283, 0
  %3285 = extractvalue { i1, i64 } %3283, 1
  br i1 %3284, label %if.then.i.i.i.i.i.i.i2207, label %Flow8175

if.then.i.i.i.i.i.i.i2207:                        ; preds = %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i2202
  %sub.i.i.i.i.i.i.i2209 = sub nuw i64 %sub.ptr.sub.i.i.i2136, %idx.0.i.i.i.i2193
  %spec.select.i.i.i36.i.i.i.i2210 = call i64 @llvm.umin.i64(i64 %sub.i.i.i.i.i.i.i2209, i64 64)
  %3286 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %3287 = and i64 %spec.select.i.i.i36.i.i.i.i2210, 15
  %3288 = sub i64 %spec.select.i.i.i36.i.i.i.i2210, %3287
  %3289 = icmp ne i64 %3288, 0
  %3290 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3289)
  %3291 = extractvalue { i1, i64 } %3290, 0
  %3292 = extractvalue { i1, i64 } %3290, 1
  br i1 %3291, label %loop-memcpy-expansion5043.preheader, label %Flow8174

loop-memcpy-expansion5043.preheader:              ; preds = %if.then.i.i.i.i.i.i.i2207
  %arrayidx.i20.i35.i.i.i.i2211 = getelementptr %"struct.rpc::Buffer", ptr %3286, i64 %idxprom.i19.i.i.i.i.i2181
  %arrayidx.i22.i38.i.i.i.i2212 = getelementptr %"struct.rpc::Buffer", ptr %arrayidx.i20.i35.i.i.i.i2211, i64 %sh_prom.i.i.i.i756
  br label %loop-memcpy-expansion5043, !amdgpu.uniform !42

Flow8174:                                         ; preds = %Flow8173, %if.then.i.i.i.i.i.i.i2207
  call void @llvm.amdgcn.end.cf.i64(i64 %3292)
  br label %loop-memcpy-residual-header5046, !amdgpu.uniform !42

loop-memcpy-expansion5043:                        ; preds = %loop-memcpy-expansion5043.preheader, %loop-memcpy-expansion5043
  %phi.broken8708 = phi i64 [ 0, %loop-memcpy-expansion5043.preheader ], [ %3297, %loop-memcpy-expansion5043 ]
  %loop-index5044 = phi i64 [ %3295, %loop-memcpy-expansion5043 ], [ 0, %loop-memcpy-expansion5043.preheader ]
  %scevgep6785 = getelementptr i8, ptr %lsr.iv6783, i64 %loop-index5044
  %3293 = load <4 x i32>, ptr %scevgep6785, align 1
  %3294 = getelementptr i8, ptr %arrayidx.i22.i38.i.i.i.i2212, i64 %loop-index5044
  store <4 x i32> %3293, ptr %3294, align 1
  %3295 = add i64 %loop-index5044, 16
  %3296 = icmp uge i64 %3295, %3288
  %3297 = call i64 @llvm.amdgcn.if.break.i64(i1 %3296, i64 %phi.broken8708)
  %3298 = call i1 @llvm.amdgcn.loop.i64(i64 %3297)
  br i1 %3298, label %Flow8173, label %loop-memcpy-expansion5043

Flow8172:                                         ; preds = %Flow8171, %loop-memcpy-residual-header5046
  call void @llvm.amdgcn.end.cf.i64(i64 %6285)
  br label %Flow8175, !amdgpu.uniform !42

loop-memcpy-residual5045:                         ; preds = %loop-memcpy-residual5045.preheader, %loop-memcpy-residual5045
  %phi.broken8709 = phi i64 [ 0, %loop-memcpy-residual5045.preheader ], [ %3302, %loop-memcpy-residual5045 ]
  %residual-loop-index5047 = phi i64 [ %3300, %loop-memcpy-residual5045 ], [ 0, %loop-memcpy-residual5045.preheader ]
  %scevgep6797 = getelementptr i8, ptr %scevgep6796, i64 %residual-loop-index5047
  %3299 = load i8, ptr %scevgep6797, align 1
  %scevgep6794 = getelementptr i8, ptr %scevgep6793, i64 %residual-loop-index5047
  store i8 %3299, ptr %scevgep6794, align 1
  %3300 = add i64 %residual-loop-index5047, 1
  %3301 = icmp uge i64 %3300, %3287
  %3302 = call i64 @llvm.amdgcn.if.break.i64(i1 %3301, i64 %phi.broken8709)
  %3303 = call i1 @llvm.amdgcn.loop.i64(i64 %3302)
  br i1 %3303, label %Flow8171, label %loop-memcpy-residual5045

Flow8171:                                         ; preds = %loop-memcpy-residual5045
  %.lcssa9101 = phi i64 [ %3302, %loop-memcpy-residual5045 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa9101)
  br label %Flow8172, !amdgpu.uniform !42

_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i2203: ; preds = %Flow8175
  %conv.i.i26.i.i.i.i2204 = xor i32 %port.sroa.37.0.i.i2192, 1
  fence release
  %3304 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %arrayidx.i24.i29.i.i.i.i2205 = getelementptr inbounds nuw i32, ptr %3304, i64 %idxprom.i23.i.i.i.i.i2188
  store atomic i32 %conv.i.i26.i.i.i.i2204, ptr %arrayidx.i24.i29.i.i.i.i2205 monotonic, align 4
  %add.i.i.i.i2206 = add i64 %idx.0.i.i.i.i2193, 64
  %scevgep6784 = getelementptr i8, ptr %lsr.iv6783, i64 64
  %lsr.iv.next6791 = add i64 %lsr.iv6790, -64
  br label %Flow8178, !amdgpu.uniform !42

cond.false.i.i.i2213:                             ; preds = %Flow8178
  %.lcssa9103 = phi i64 [ %3281, %Flow8178 ]
  %port.sroa.37.0.i.i2192.lcssa = phi i32 [ %port.sroa.37.0.i.i2192, %Flow8178 ]
  %.lcssa9102 = phi i32 [ %3262, %Flow8178 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa9103)
  %3305 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %.lcssa9102)
  %3306 = icmp ne i32 %3305, %port.sroa.37.0.i.i2192.lcssa
  %3307 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3306)
  %3308 = extractvalue { i1, i64 } %3307, 0
  %3309 = extractvalue { i1, i64 } %3307, 1
  br i1 %3308, label %while.body.i.i.i.i2214.preheader, label %Flow8170

while.body.i.i.i.i2214.preheader:                 ; preds = %cond.false.i.i.i2213
  br label %while.body.i.i.i.i2214, !amdgpu.uniform !42

Flow8170:                                         ; preds = %Flow8169, %cond.false.i.i.i2213
  call void @llvm.amdgcn.end.cf.i64(i64 %3309)
  br label %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit2235, !amdgpu.uniform !42

while.body.i.i.i.i2214:                           ; preds = %while.body.i.i.i.i2214.preheader, %while.body.i.i.i.i2214
  %phi.broken8711 = phi i64 [ 0, %while.body.i.i.i.i2214.preheader ], [ %3313, %while.body.i.i.i.i2214 ]
  call void @llvm.amdgcn.s.sleep(i32 2)
  %3310 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i8.i.i2215 = getelementptr inbounds nuw i32, ptr %3310, i64 %idxprom.i23.i.i.i.i.i2188
  %3311 = load atomic i32, ptr %arrayidx.i.i.i8.i.i2215 monotonic, align 4
  %3312 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %3311)
  %cmp.i.not.i.i.i.i2216 = icmp eq i32 %3312, %port.sroa.37.0.i.i2192.lcssa
  %3313 = call i64 @llvm.amdgcn.if.break.i64(i1 %cmp.i.not.i.i.i.i2216, i64 %phi.broken8711)
  %3314 = call i1 @llvm.amdgcn.loop.i64(i64 %3313)
  br i1 %3314, label %Flow8169, label %while.body.i.i.i.i2214

Flow8169:                                         ; preds = %while.body.i.i.i.i2214
  %.lcssa9104 = phi i64 [ %3313, %while.body.i.i.i.i2214 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa9104)
  br label %Flow8170, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit2235: ; preds = %Flow8170
  fence acquire
  call void @llvm.amdgcn.wave.barrier()
  fence syncscope("agent") release
  %3315 = call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %.lcssa9095, i1 true)
  %iszero.i.i.i.i.i13.i.i2218 = icmp ne i64 %.lcssa9095, 0
  %cmp2.i.i.i.i14.i.i2219 = icmp eq i64 %3315, %sh_prom.i.i.i.i756
  %cmp.i.i.i.i15.i.i2220 = select i1 %iszero.i.i.i.i.i13.i.i2218, i1 %cmp2.i.i.i.i14.i.i2219, i1 false
  %3316 = and i32 %.lcssa9096, 31
  %conv4.i.i.i19.i.i2224 = zext i1 %cmp.i.i.i.i15.i.i2220 to i32
  %shl.i.i.i20.i.i2225 = shl nuw i32 %conv4.i.i.i19.i.i2224, %3316
  %xor.i.i.i21.i.i2226 = xor i32 %shl.i.i.i20.i.i2225, -1
  %3317 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop5745, !amdgpu.uniform !42

3318:                                             ; preds = %ComputeEnd5746
  %div8.i.i.i16.i.i2221 = lshr i32 %.lcssa9096, 5
  %idxprom.i.i.i17.i.i2222 = zext nneg i32 %div8.i.i.i16.i.i2221 to i64
  %3319 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i17.i.i2222
  %sunkaddr7431 = getelementptr inbounds i8, ptr addrspace(1) %3319, i64 40
  %3320 = atomicrmw and ptr addrspace(1) %sunkaddr7431, i32 %.lcssa9105 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %3321, !amdgpu.uniform !42

3321:                                             ; preds = %3318, %ComputeEnd5746
  call void @llvm.amdgcn.end.cf.i64(i64 %7876)
  call void @llvm.amdgcn.wave.barrier()
  br label %for.cond.i.i1.i2244, !amdgpu.uniform !42

for.cond.i.i1.i2244:                              ; preds = %3321, %Flow8168
  %3322 = phi i32 [ %3359, %Flow8168 ], [ %1192, %3321 ]
  %3323 = phi i64 [ %3360, %Flow8168 ], [ %1193, %3321 ]
  %3324 = phi i32 [ %3361, %Flow8168 ], [ %1194, %3321 ]
  %retval.sroa.6.0.i.i.i2245 = phi i32 [ %3364, %Flow8168 ], [ undef, %3321 ]
  %retval.sroa.2.0.i.i.i2247 = phi i64 [ %3363, %Flow8168 ], [ undef, %3321 ]
  %retval.sroa.8.0.i.i.i2248 = phi i32 [ %3362, %Flow8168 ], [ undef, %3321 ]
  %index.0.i.i.i2249 = phi i32 [ %3366, %Flow8168 ], [ 0, %3321 ]
  %3325 = load i32, ptr addrspace(1) @__llvm_rpc_client, align 8, !tbaa !68
  %cmp.not.i.i.i2250 = icmp ult i32 %index.0.i.i.i2249, %3325
  %spec.store.select.i.i.i2251 = select i1 %cmp.not.i.i.i2250, i32 %index.0.i.i.i2249, i32 0
  %3326 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  %3327 = and i32 %spec.store.select.i.i.i2251, 31
  %3328 = lshr i64 %3326, %sh_prom.i.i.i.i756
  %3329 = trunc i64 %3328 to i32
  %conv4.i.i.i.i.i2255 = and i32 %3329, 1
  %shl.i.i.i.i.i2256 = shl nuw i32 %conv4.i.i.i.i.i2255, %3327
  br label %ComputeLoop5749, !amdgpu.uniform !42

3330:                                             ; preds = %ComputeEnd5750
  %sunkaddr7432 = getelementptr inbounds i8, ptr addrspace(1) %7886, i64 40
  %3331 = atomicrmw or ptr addrspace(1) %sunkaddr7432, i32 %.lcssa9106 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %3332, !amdgpu.uniform !42

3332:                                             ; preds = %3330, %ComputeEnd5750
  %3333 = phi i32 [ %3331, %3330 ], [ poison, %ComputeEnd5750 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %7895)
  %3334 = call i32 @llvm.amdgcn.readfirstlane.i32(i32 %3333)
  %3335 = or i32 %3334, %.lcssa9107
  %shl5.i.i.i.i.i2257 = shl nuw i32 1, %3327
  %and.i.i.i.i.i2258 = and i32 %shl5.i.i.i.i.i2257, %3335
  %tobool3.i.i.i.i2259 = icmp ne i32 %and.i.i.i.i.i2258, 0
  %3336 = call i64 @llvm.amdgcn.ballot.i64(i1 %tobool3.i.i.i.i2259)
  %and.i.i.i.i.i.i2260 = and i64 %3336, %3326
  %cmp.i.not.i.i.i2261 = icmp ne i64 %3326, %and.i.i.i.i.i.i2260
  br i1 %cmp.i.not.i.i.i2261, label %if.then.i.i.i.i2262, label %Flow8167, !amdgpu.uniform !42

if.then.i.i.i.i2262:                              ; preds = %3332
  fence syncscope("agent") acquire
  %3337 = load <2 x i64>, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !75
  %3338 = extractelement <2 x i64> %3337, i32 0
  %3339 = inttoptr i64 %3338 to ptr
  %3340 = extractelement <2 x i64> %3337, i32 1
  %3341 = inttoptr i64 %3340 to ptr
  %idxprom.i.i.i.i2263 = zext i32 %spec.store.select.i.i.i2251 to i64
  %arrayidx.i.i.i.i2264 = getelementptr inbounds nuw i32, ptr %3339, i64 %idxprom.i.i.i.i2263, !amdgpu.uniform !42
  %3342 = load atomic i32, ptr %arrayidx.i.i.i.i2264 monotonic, align 4
  %3343 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %3342)
  %arrayidx.i45.i.i.i2265 = getelementptr inbounds nuw i32, ptr %3341, i64 %idxprom.i.i.i.i2263, !amdgpu.uniform !42
  %3344 = load atomic i32, ptr %arrayidx.i45.i.i.i2265 monotonic, align 4
  %3345 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %3344)
  %cmp.i46.not.i.i.i2266 = icmp ne i32 %3343, %3345
  br i1 %cmp.i46.not.i.i.i2266, label %if.then12.i.i.i2267, label %Flow8165, !amdgpu.uniform !42

if.then12.i.i.i2267:                              ; preds = %if.then.i.i.i.i2262
  fence syncscope("agent") release
  %3346 = call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %3326, i1 true)
  %iszero.i.i.i.i.i.i.i2268 = icmp ne i64 %3326, 0
  %cmp2.i.i.i.i.i.i2269 = icmp eq i64 %3346, %sh_prom.i.i.i.i756
  %cmp.i.i.i.i.i.i2270 = select i1 %iszero.i.i.i.i.i.i.i2268, i1 %cmp2.i.i.i.i.i.i2269, i1 false
  %conv4.i.i51.i.i.i2271 = zext i1 %cmp.i.i.i.i.i.i2270 to i32
  %shl.i.i52.i.i.i2272 = shl nuw i32 %conv4.i.i51.i.i.i2271, %3327
  %xor.i.i.i.i.i2273 = xor i32 %shl.i.i52.i.i.i2272, -1
  %3347 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop5754, !amdgpu.uniform !42

3348:                                             ; preds = %ComputeEnd5755
  %sunkaddr7433 = getelementptr inbounds i8, ptr addrspace(1) %7886, i64 40
  %3349 = atomicrmw and ptr addrspace(1) %sunkaddr7433, i32 %.lcssa9108 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %3350, !amdgpu.uniform !42

3350:                                             ; preds = %3348, %ComputeEnd5755
  call void @llvm.amdgcn.end.cf.i64(i64 %7913)
  call void @llvm.amdgcn.wave.barrier()
  br label %Flow8165, !amdgpu.uniform !42

if.end14.i.i.i2333:                               ; preds = %Flow8165
  %3351 = call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %3326, i1 true)
  %iszero.i.i.i.i.i.i2334 = icmp ne i64 %3326, 0
  %cmp2.i.i.i.i.i2335 = icmp eq i64 %3351, %sh_prom.i.i.i.i756
  %cmp.i.i.i.i.i2336 = select i1 %iszero.i.i.i.i.i.i2334, i1 %cmp2.i.i.i.i.i2335, i1 false
  %3352 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.i.i.i.i.i2336)
  %3353 = extractvalue { i1, i64 } %3352, 0
  %3354 = extractvalue { i1, i64 } %3352, 1
  br i1 %3353, label %if.then16.i.i.i2338, label %if.end22.i.i.i2337

if.then16.i.i.i2338:                              ; preds = %if.end14.i.i.i2333
  %3355 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !76
  %opcode.i.i.i2339 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %3355, i64 %idxprom.i.i.i.i2263, i32 1
  store i32 1660944387, ptr %opcode.i.i.i2339, align 8, !tbaa !77
  %arrayidx21.i.i.i2340 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %3355, i64 %idxprom.i.i.i.i2263
  store i64 %3326, ptr %arrayidx21.i.i.i2340, align 8, !tbaa !80
  br label %if.end22.i.i.i2337, !amdgpu.uniform !42

Flow8166:                                         ; preds = %if.end22.i.i.i2337, %Flow8165
  %3356 = phi i32 [ %3343, %if.end22.i.i.i2337 ], [ %retval.sroa.8.0.i.i.i2248, %Flow8165 ]
  %3357 = phi i64 [ %3326, %if.end22.i.i.i2337 ], [ %retval.sroa.2.0.i.i.i2247, %Flow8165 ]
  %3358 = phi i32 [ %spec.store.select.i.i.i2251, %if.end22.i.i.i2337 ], [ %retval.sroa.6.0.i.i.i2245, %Flow8165 ]
  br label %cleanup26.i.i.i2274, !amdgpu.uniform !42

if.end22.i.i.i2337:                               ; preds = %if.then16.i.i.i2338, %if.end14.i.i.i2333
  call void @llvm.amdgcn.end.cf.i64(i64 %3354)
  call void @llvm.amdgcn.wave.barrier()
  br label %Flow8166, !amdgpu.uniform !42

Flow8167:                                         ; preds = %cleanup26.i.i.i2274, %3332
  %3359 = phi i32 [ %3358, %cleanup26.i.i.i2274 ], [ %3322, %3332 ]
  %3360 = phi i64 [ %3357, %cleanup26.i.i.i2274 ], [ %3323, %3332 ]
  %3361 = phi i32 [ %3356, %cleanup26.i.i.i2274 ], [ %3324, %3332 ]
  %3362 = phi i32 [ %3356, %cleanup26.i.i.i2274 ], [ %retval.sroa.8.0.i.i.i2248, %3332 ]
  %3363 = phi i64 [ %3357, %cleanup26.i.i.i2274 ], [ %retval.sroa.2.0.i.i.i2247, %3332 ]
  %3364 = phi i32 [ %3358, %cleanup26.i.i.i2274 ], [ %retval.sroa.6.0.i.i.i2245, %3332 ]
  %3365 = phi i1 [ %cmp.i46.not.i.i.i2266, %cleanup26.i.i.i2274 ], [ true, %3332 ]
  br i1 %3365, label %for.inc.i.i.i2279, label %Flow8168, !amdgpu.uniform !42

cleanup26.i.i.i2274:                              ; preds = %Flow8166
  br label %Flow8167, !amdgpu.uniform !42

for.inc.i.i.i2279:                                ; preds = %Flow8167
  %inc.i.i.i2284 = add i32 %spec.store.select.i.i.i2251, 1
  br label %Flow8168, !amdgpu.uniform !42

Flow8168:                                         ; preds = %for.inc.i.i.i2279, %Flow8167
  %3366 = phi i32 [ %inc.i.i.i2284, %for.inc.i.i.i2279 ], [ poison, %Flow8167 ]
  %3367 = phi i1 [ false, %for.inc.i.i.i2279 ], [ true, %Flow8167 ]
  br i1 %3367, label %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i2285, label %for.cond.i.i1.i2244, !amdgpu.uniform !42

_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i2285: ; preds = %Flow8168
  %.lcssa9111 = phi i32 [ %3359, %Flow8168 ]
  %.lcssa9110 = phi i64 [ %3360, %Flow8168 ]
  %.lcssa9109 = phi i32 [ %3361, %Flow8168 ]
  fence acquire
  %3368 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %mul.i.i.i.i.i.i2286 = shl i32 %.lcssa9111, 6
  %idxprom.i19.i.i.i.i.i2287 = zext i32 %mul.i.i.i.i.i.i2286 to i64
  %arrayidx.i20.i.i.i.i.i2288 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %3368, i64 %idxprom.i19.i.i.i.i.i2287
  %arrayidx.i22.i.i.i.i.i2289 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %arrayidx.i20.i.i.i.i.i2288, i64 %sh_prom.i.i.i.i756
  store i64 sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.20.37 to ptr), i64 2) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.20.37 to ptr) to i64)), ptr %arrayidx.i22.i.i.i.i.i2289, align 8, !tbaa !82
  br i1 %1118, label %loop-memcpy-expansion5337.preheader, label %Flow8164, !amdgpu.uniform !42

loop-memcpy-expansion5337.preheader:              ; preds = %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i2285
  %scevgep6798 = getelementptr i8, ptr %3368, i64 8
  %scevgep6799 = getelementptr i8, ptr %scevgep6798, i64 %1474
  %3369 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i2287, 6
  %scevgep6800 = getelementptr i8, ptr %scevgep6799, i64 %3369
  br label %loop-memcpy-expansion5337, !amdgpu.uniform !42

Flow8164:                                         ; preds = %Flow8163, %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i2285
  br label %loop-memcpy-residual-header5340, !amdgpu.uniform !42

loop-memcpy-expansion5337:                        ; preds = %loop-memcpy-expansion5337.preheader, %loop-memcpy-expansion5337
  %loop-index5338 = phi i64 [ %3372, %loop-memcpy-expansion5337 ], [ 0, %loop-memcpy-expansion5337.preheader ]
  %3370 = getelementptr i8, ptr addrspace(4) @.str.20.37, i64 %loop-index5338, !amdgpu.uniform !42
  %3371 = load <4 x i32>, ptr addrspace(4) %3370, align 1
  %scevgep6801 = getelementptr i8, ptr %scevgep6800, i64 %loop-index5338
  store <4 x i32> %3371, ptr %scevgep6801, align 1
  %3372 = add i64 %loop-index5338, 16
  %3373 = icmp uge i64 %3372, %1117
  br i1 %3373, label %Flow8163, label %loop-memcpy-expansion5337, !amdgpu.uniform !42

Flow8162:                                         ; preds = %Flow8160, %Flow8161
  %3374 = phi i64 [ %6611, %Flow8160 ], [ %6609, %Flow8161 ]
  br label %post-loop-memcpy-expansion5336, !amdgpu.uniform !42

loop-memcpy-residual5339:                         ; preds = %loop-memcpy-residual5339.preheader, %loop-memcpy-residual5339
  %residual-loop-index5341 = phi i64 [ %3376, %loop-memcpy-residual5339 ], [ 0, %loop-memcpy-residual5339.preheader ]
  %scevgep6806 = getelementptr i8, ptr addrspace(4) @.str.20.37, i64 %residual-loop-index5341, !amdgpu.uniform !42
  %3375 = load i8, ptr addrspace(4) %scevgep6806, align 1
  %scevgep6805 = getelementptr i8, ptr %scevgep6804, i64 %residual-loop-index5341
  store i8 %3375, ptr %scevgep6805, align 1
  %3376 = add i64 %residual-loop-index5341, 1
  %3377 = icmp uge i64 %3376, %1116
  br i1 %3377, label %Flow8160, label %loop-memcpy-residual5339, !amdgpu.uniform !42

Flow8160:                                         ; preds = %loop-memcpy-residual5339
  br label %Flow8162, !amdgpu.uniform !42

post-loop-memcpy-expansion5336:                   ; preds = %Flow8162
  %tobool.not.i.i.i.i.i.i2292 = icmp eq i32 %.lcssa9109, 0
  %conv.i.i.i.i.i.i2293 = zext i1 %tobool.not.i.i.i.i.i.i2292 to i32
  fence release
  %3378 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %idxprom.i23.i.i.i.i.i2294 = zext i32 %.lcssa9111 to i64
  %arrayidx.i24.i.i.i.i.i2295 = getelementptr inbounds nuw i32, ptr %3378, i64 %idxprom.i23.i.i.i.i.i2294
  store atomic i32 %conv.i.i.i.i.i.i2293, ptr %arrayidx.i24.i.i.i.i.i2295 monotonic, align 4
  %3379 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !84
  %arrayidx.i.i5.i.i2296 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %3379, i64 %idxprom.i23.i.i.i.i.i2294
  %3380 = load i64, ptr %arrayidx.i.i5.i.i2296, align 8, !tbaa !80
  br label %while.cond.i.i.i.i2297, !amdgpu.uniform !42

while.cond.i.i.i.i2297:                           ; preds = %post-loop-memcpy-expansion5336, %Flow8159
  %phi.broken8713 = phi i64 [ 0, %post-loop-memcpy-expansion5336 ], [ %3402, %Flow8159 ]
  %lsr.iv6812 = phi i64 [ %3400, %Flow8159 ], [ -54, %post-loop-memcpy-expansion5336 ]
  %lsr.iv6807 = phi ptr addrspace(4) [ %3399, %Flow8159 ], [ getelementptr (i8, ptr addrspace(4) @.str.20.37, i64 56), %post-loop-memcpy-expansion5336 ]
  %port.sroa.37.0.i.i2298 = phi i32 [ %3398, %Flow8159 ], [ %conv.i.i.i.i.i.i2293, %post-loop-memcpy-expansion5336 ]
  %idx.0.i.i.i.i2299 = phi i64 [ %3397, %Flow8159 ], [ 56, %post-loop-memcpy-expansion5336 ]
  %cmp.i.i.i.i2300 = icmp ult i64 %idx.0.i.i.i.i2299, sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.20.37 to ptr), i64 2) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.20.37 to ptr) to i64))
  %3381 = call i64 @llvm.amdgcn.ballot.i64(i1 %cmp.i.i.i.i2300)
  %and.i.i.i.i6.i.i2301 = and i64 %3381, %3380
  %tobool.not.i.i.i.i2302 = icmp ne i64 %and.i.i.i.i6.i.i2301, 0
  %3382 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i22.i.i.i2303 = getelementptr inbounds nuw i32, ptr %3382, i64 %idxprom.i23.i.i.i.i.i2294
  %3383 = load atomic i32, ptr %arrayidx.i22.i.i.i2303 monotonic, align 4
  %3384 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %tobool.not.i.i.i.i2302)
  %3385 = extractvalue { i1, i64 } %3384, 0
  %3386 = extractvalue { i1, i64 } %3384, 1
  br i1 %3385, label %cond.false.i8.i.i.i.i2304, label %Flow8159

cond.false.i8.i.i.i.i2304:                        ; preds = %while.cond.i.i.i.i2297
  %3387 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %3383)
  %3388 = icmp ne i32 %3387, %port.sroa.37.0.i.i2298
  %3389 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3388)
  %3390 = extractvalue { i1, i64 } %3389, 0
  %3391 = extractvalue { i1, i64 } %3389, 1
  br i1 %3390, label %while.body.i.i21.i.i.i.i2305.preheader, label %Flow8158

while.body.i.i21.i.i.i.i2305.preheader:           ; preds = %cond.false.i8.i.i.i.i2304
  br label %while.body.i.i21.i.i.i.i2305, !amdgpu.uniform !42

Flow8158:                                         ; preds = %Flow8157, %cond.false.i8.i.i.i.i2304
  call void @llvm.amdgcn.end.cf.i64(i64 %3391)
  br label %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i2308, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i2305:                     ; preds = %while.body.i.i21.i.i.i.i2305.preheader, %while.body.i.i21.i.i.i.i2305
  %phi.broken8712 = phi i64 [ 0, %while.body.i.i21.i.i.i.i2305.preheader ], [ %3395, %while.body.i.i21.i.i.i.i2305 ]
  call void @llvm.amdgcn.s.sleep(i32 2)
  %3392 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i22.i.i.i.i2306 = getelementptr inbounds nuw i32, ptr %3392, i64 %idxprom.i23.i.i.i.i.i2294
  %3393 = load atomic i32, ptr %arrayidx.i.i.i22.i.i.i.i2306 monotonic, align 4
  %3394 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %3393)
  %cmp.i.not.i.i23.i.i.i.i2307 = icmp eq i32 %3394, %port.sroa.37.0.i.i2298
  %3395 = call i64 @llvm.amdgcn.if.break.i64(i1 %cmp.i.not.i.i23.i.i.i.i2307, i64 %phi.broken8712)
  %3396 = call i1 @llvm.amdgcn.loop.i64(i64 %3395)
  br i1 %3396, label %Flow8157, label %while.body.i.i21.i.i.i.i2305

Flow8157:                                         ; preds = %while.body.i.i21.i.i.i.i2305
  %.lcssa9112 = phi i64 [ %3395, %while.body.i.i21.i.i.i.i2305 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa9112)
  br label %Flow8158, !amdgpu.uniform !42

Flow8159:                                         ; preds = %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i2309, %while.cond.i.i.i.i2297
  %3397 = phi i64 [ %add.i.i.i.i2312, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i2309 ], [ poison, %while.cond.i.i.i.i2297 ]
  %3398 = phi i32 [ %conv.i.i26.i.i.i.i2310, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i2309 ], [ poison, %while.cond.i.i.i.i2297 ]
  %3399 = phi ptr addrspace(4) [ %scevgep6808, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i2309 ], [ poison, %while.cond.i.i.i.i2297 ]
  %3400 = phi i64 [ %lsr.iv.next6813, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i2309 ], [ poison, %while.cond.i.i.i.i2297 ]
  %3401 = phi i1 [ false, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i2309 ], [ true, %while.cond.i.i.i.i2297 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %3386)
  %3402 = call i64 @llvm.amdgcn.if.break.i64(i1 %3401, i64 %phi.broken8713)
  %3403 = call i1 @llvm.amdgcn.loop.i64(i64 %3402)
  br i1 %3403, label %cond.false.i.i.i2319, label %while.cond.i.i.i.i2297

_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i2308: ; preds = %Flow8158
  fence acquire
  br i1 %cmp.i.i.i.i2300, label %if.then.i.i.i.i.i.i.i2313, label %Flow8156, !amdgpu.uniform !42

if.then.i.i.i.i.i.i.i2313:                        ; preds = %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i2308
  %sub.i.i.i.i.i.i.i2315 = sub nuw i64 sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.20.37 to ptr), i64 2) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.20.37 to ptr) to i64)), %idx.0.i.i.i.i2299
  %spec.select.i.i.i36.i.i.i.i2316 = call i64 @llvm.umin.i64(i64 %sub.i.i.i.i.i.i.i2315, i64 64)
  %3404 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %3405 = and i64 %spec.select.i.i.i36.i.i.i.i2316, 15
  %3406 = sub i64 %spec.select.i.i.i36.i.i.i.i2316, %3405
  %3407 = icmp ne i64 %3406, 0
  br i1 %3407, label %loop-memcpy-expansion5343.preheader, label %Flow8155, !amdgpu.uniform !42

loop-memcpy-expansion5343.preheader:              ; preds = %if.then.i.i.i.i.i.i.i2313
  %arrayidx.i20.i35.i.i.i.i2317 = getelementptr %"struct.rpc::Buffer", ptr %3404, i64 %idxprom.i19.i.i.i.i.i2287
  %arrayidx.i22.i38.i.i.i.i2318 = getelementptr %"struct.rpc::Buffer", ptr %arrayidx.i20.i35.i.i.i.i2317, i64 %sh_prom.i.i.i.i756
  br label %loop-memcpy-expansion5343, !amdgpu.uniform !42

Flow8155:                                         ; preds = %Flow8154, %if.then.i.i.i.i.i.i.i2313
  br label %loop-memcpy-residual-header5346, !amdgpu.uniform !42

loop-memcpy-expansion5343:                        ; preds = %loop-memcpy-expansion5343.preheader, %loop-memcpy-expansion5343
  %loop-index5344 = phi i64 [ %3410, %loop-memcpy-expansion5343 ], [ 0, %loop-memcpy-expansion5343.preheader ]
  %scevgep6809 = getelementptr i8, ptr addrspace(4) %lsr.iv6807, i64 %loop-index5344, !amdgpu.uniform !42
  %3408 = load <4 x i32>, ptr addrspace(4) %scevgep6809, align 1
  %3409 = getelementptr i8, ptr %arrayidx.i22.i38.i.i.i.i2318, i64 %loop-index5344
  store <4 x i32> %3408, ptr %3409, align 1
  %3410 = add i64 %loop-index5344, 16
  %3411 = icmp uge i64 %3410, %3406
  br i1 %3411, label %Flow8154, label %loop-memcpy-expansion5343, !amdgpu.uniform !42

Flow8153:                                         ; preds = %Flow8152, %loop-memcpy-residual-header5346
  br label %Flow8156, !amdgpu.uniform !42

loop-memcpy-residual5345:                         ; preds = %loop-memcpy-residual5345.preheader, %loop-memcpy-residual5345
  %residual-loop-index5347 = phi i64 [ %3413, %loop-memcpy-residual5345 ], [ 0, %loop-memcpy-residual5345.preheader ]
  %scevgep6818 = getelementptr i8, ptr addrspace(4) %scevgep6817, i64 %residual-loop-index5347, !amdgpu.uniform !42
  %3412 = load i8, ptr addrspace(4) %scevgep6818, align 1
  %scevgep6816 = getelementptr i8, ptr %scevgep6815, i64 %residual-loop-index5347
  store i8 %3412, ptr %scevgep6816, align 1
  %3413 = add i64 %residual-loop-index5347, 1
  %3414 = icmp uge i64 %3413, %3405
  br i1 %3414, label %Flow8152, label %loop-memcpy-residual5345, !amdgpu.uniform !42

Flow8152:                                         ; preds = %loop-memcpy-residual5345
  br label %Flow8153, !amdgpu.uniform !42

_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i2309: ; preds = %Flow8156
  %conv.i.i26.i.i.i.i2310 = xor i32 %port.sroa.37.0.i.i2298, 1
  fence release
  %3415 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %arrayidx.i24.i29.i.i.i.i2311 = getelementptr inbounds nuw i32, ptr %3415, i64 %idxprom.i23.i.i.i.i.i2294
  store atomic i32 %conv.i.i26.i.i.i.i2310, ptr %arrayidx.i24.i29.i.i.i.i2311 monotonic, align 4
  %add.i.i.i.i2312 = add i64 %idx.0.i.i.i.i2299, 64
  %scevgep6808 = getelementptr i8, ptr addrspace(4) %lsr.iv6807, i64 64
  %lsr.iv.next6813 = add i64 %lsr.iv6812, -64
  br label %Flow8159, !amdgpu.uniform !42

cond.false.i.i.i2319:                             ; preds = %Flow8159
  %.lcssa9114 = phi i64 [ %3402, %Flow8159 ]
  %port.sroa.37.0.i.i2298.lcssa = phi i32 [ %port.sroa.37.0.i.i2298, %Flow8159 ]
  %.lcssa9113 = phi i32 [ %3383, %Flow8159 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa9114)
  %3416 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %.lcssa9113)
  %3417 = icmp ne i32 %3416, %port.sroa.37.0.i.i2298.lcssa
  %3418 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3417)
  %3419 = extractvalue { i1, i64 } %3418, 0
  %3420 = extractvalue { i1, i64 } %3418, 1
  br i1 %3419, label %while.body.i.i.i.i2320.preheader, label %Flow8151

while.body.i.i.i.i2320.preheader:                 ; preds = %cond.false.i.i.i2319
  br label %while.body.i.i.i.i2320, !amdgpu.uniform !42

Flow8151:                                         ; preds = %Flow8150, %cond.false.i.i.i2319
  call void @llvm.amdgcn.end.cf.i64(i64 %3420)
  br label %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit2341, !amdgpu.uniform !42

while.body.i.i.i.i2320:                           ; preds = %while.body.i.i.i.i2320.preheader, %while.body.i.i.i.i2320
  %phi.broken8714 = phi i64 [ 0, %while.body.i.i.i.i2320.preheader ], [ %3424, %while.body.i.i.i.i2320 ]
  call void @llvm.amdgcn.s.sleep(i32 2)
  %3421 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i8.i.i2321 = getelementptr inbounds nuw i32, ptr %3421, i64 %idxprom.i23.i.i.i.i.i2294
  %3422 = load atomic i32, ptr %arrayidx.i.i.i8.i.i2321 monotonic, align 4
  %3423 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %3422)
  %cmp.i.not.i.i.i.i2322 = icmp eq i32 %3423, %port.sroa.37.0.i.i2298.lcssa
  %3424 = call i64 @llvm.amdgcn.if.break.i64(i1 %cmp.i.not.i.i.i.i2322, i64 %phi.broken8714)
  %3425 = call i1 @llvm.amdgcn.loop.i64(i64 %3424)
  br i1 %3425, label %Flow8150, label %while.body.i.i.i.i2320

Flow8150:                                         ; preds = %while.body.i.i.i.i2320
  %.lcssa9115 = phi i64 [ %3424, %while.body.i.i.i.i2320 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa9115)
  br label %Flow8151, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit2341: ; preds = %Flow8151
  fence acquire
  call void @llvm.amdgcn.wave.barrier()
  fence syncscope("agent") release
  %3426 = call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %.lcssa9110, i1 true)
  %iszero.i.i.i.i.i13.i.i2324 = icmp ne i64 %.lcssa9110, 0
  %cmp2.i.i.i.i14.i.i2325 = icmp eq i64 %3426, %sh_prom.i.i.i.i756
  %cmp.i.i.i.i15.i.i2326 = select i1 %iszero.i.i.i.i.i13.i.i2324, i1 %cmp2.i.i.i.i14.i.i2325, i1 false
  %3427 = and i32 %.lcssa9111, 31
  %conv4.i.i.i19.i.i2330 = zext i1 %cmp.i.i.i.i15.i.i2326 to i32
  %shl.i.i.i20.i.i2331 = shl nuw i32 %conv4.i.i.i19.i.i2330, %3427
  %xor.i.i.i21.i.i2332 = xor i32 %shl.i.i.i20.i.i2331, -1
  %3428 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop5758, !amdgpu.uniform !42

3429:                                             ; preds = %ComputeEnd5759
  %div8.i.i.i16.i.i2327 = lshr i32 %.lcssa9111, 5
  %idxprom.i.i.i17.i.i2328 = zext nneg i32 %div8.i.i.i16.i.i2327 to i64
  %3430 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i17.i.i2328
  %sunkaddr7434 = getelementptr inbounds i8, ptr addrspace(1) %3430, i64 40
  %3431 = atomicrmw and ptr addrspace(1) %sunkaddr7434, i32 %.lcssa9116 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %3432, !amdgpu.uniform !42

3432:                                             ; preds = %3429, %ComputeEnd5759
  call void @llvm.amdgcn.end.cf.i64(i64 %7930)
  call void @llvm.amdgcn.wave.barrier()
  %cmp62 = icmp uge i64 %retval.1.i1278, %retval.1.i
  br i1 %cmp62, label %if.else, label %Flow8148, !amdgpu.uniform !42

Flow8148:                                         ; preds = %Flow8128, %3432
  %3433 = phi i32 [ %3629, %Flow8128 ], [ poison, %3432 ]
  %3434 = phi i64 [ %3630, %Flow8128 ], [ poison, %3432 ]
  %3435 = phi i32 [ %3631, %Flow8128 ], [ poison, %3432 ]
  %3436 = phi i32 [ %3632, %Flow8128 ], [ poison, %3432 ]
  %3437 = phi i64 [ %3633, %Flow8128 ], [ poison, %3432 ]
  %3438 = phi i32 [ %3634, %Flow8128 ], [ poison, %3432 ]
  %3439 = phi i32 [ %3635, %Flow8128 ], [ poison, %3432 ]
  %3440 = phi i64 [ %3636, %Flow8128 ], [ poison, %3432 ]
  %3441 = phi i32 [ %3637, %Flow8128 ], [ poison, %3432 ]
  %3442 = phi i32 [ %3638, %Flow8128 ], [ poison, %3432 ]
  %3443 = phi i64 [ %3639, %Flow8128 ], [ poison, %3432 ]
  %3444 = phi i32 [ %3640, %Flow8128 ], [ poison, %3432 ]
  %3445 = phi i32 [ %3641, %Flow8128 ], [ poison, %3432 ]
  %3446 = phi i64 [ %3642, %Flow8128 ], [ poison, %3432 ]
  %3447 = phi i32 [ %3643, %Flow8128 ], [ poison, %3432 ]
  %3448 = phi i32 [ %3644, %Flow8128 ], [ poison, %3432 ]
  %3449 = phi i64 [ %3645, %Flow8128 ], [ poison, %3432 ]
  %3450 = phi i32 [ %3646, %Flow8128 ], [ poison, %3432 ]
  %3451 = phi i1 [ false, %Flow8128 ], [ true, %3432 ]
  br i1 %3451, label %for.cond.i.i1.i2350.preheader, label %Flow8149, !amdgpu.uniform !42

for.cond.i.i1.i2350.preheader:                    ; preds = %Flow8148
  br label %for.cond.i.i1.i2350, !amdgpu.uniform !42

Flow8149:                                         ; preds = %3583, %Flow8148
  %3452 = phi i32 [ %1215, %3583 ], [ %3433, %Flow8148 ]
  %3453 = phi i64 [ %1214, %3583 ], [ %3434, %Flow8148 ]
  %3454 = phi i32 [ %1213, %3583 ], [ %3435, %Flow8148 ]
  %3455 = phi i32 [ %1212, %3583 ], [ %3436, %Flow8148 ]
  %3456 = phi i64 [ %1211, %3583 ], [ %3437, %Flow8148 ]
  %3457 = phi i32 [ %1210, %3583 ], [ %3438, %Flow8148 ]
  %3458 = phi i32 [ %1209, %3583 ], [ %3439, %Flow8148 ]
  %3459 = phi i64 [ %1208, %3583 ], [ %3440, %Flow8148 ]
  %3460 = phi i32 [ %1207, %3583 ], [ %3441, %Flow8148 ]
  %3461 = phi i32 [ %1206, %3583 ], [ %3442, %Flow8148 ]
  %3462 = phi i64 [ %1205, %3583 ], [ %3443, %Flow8148 ]
  %3463 = phi i32 [ %1204, %3583 ], [ %3444, %Flow8148 ]
  %3464 = phi i32 [ %1203, %3583 ], [ %3445, %Flow8148 ]
  %3465 = phi i64 [ %1202, %3583 ], [ %3446, %Flow8148 ]
  %3466 = phi i32 [ %1201, %3583 ], [ %3447, %Flow8148 ]
  %3467 = phi i32 [ %1200, %3583 ], [ %3448, %Flow8148 ]
  %3468 = phi i64 [ %1199, %3583 ], [ %3449, %Flow8148 ]
  %3469 = phi i32 [ %1198, %3583 ], [ %3450, %Flow8148 ]
  %3470 = phi i32 [ %.lcssa9186, %3583 ], [ %1197, %Flow8148 ]
  %3471 = phi i64 [ %.lcssa9187, %3583 ], [ %1196, %Flow8148 ]
  %3472 = phi i32 [ %.lcssa9188, %3583 ], [ %1195, %Flow8148 ]
  br label %Flow8330, !amdgpu.uniform !42

for.cond.i.i1.i2350:                              ; preds = %for.cond.i.i1.i2350.preheader, %Flow8147
  %3473 = phi i32 [ %3510, %Flow8147 ], [ %1195, %for.cond.i.i1.i2350.preheader ]
  %3474 = phi i64 [ %3511, %Flow8147 ], [ %1196, %for.cond.i.i1.i2350.preheader ]
  %3475 = phi i32 [ %3512, %Flow8147 ], [ %1197, %for.cond.i.i1.i2350.preheader ]
  %retval.sroa.6.0.i.i.i2351 = phi i32 [ %3515, %Flow8147 ], [ undef, %for.cond.i.i1.i2350.preheader ]
  %retval.sroa.2.0.i.i.i2353 = phi i64 [ %3514, %Flow8147 ], [ undef, %for.cond.i.i1.i2350.preheader ]
  %retval.sroa.8.0.i.i.i2354 = phi i32 [ %3513, %Flow8147 ], [ undef, %for.cond.i.i1.i2350.preheader ]
  %index.0.i.i.i2355 = phi i32 [ %3517, %Flow8147 ], [ 0, %for.cond.i.i1.i2350.preheader ]
  %3476 = load i32, ptr addrspace(1) @__llvm_rpc_client, align 8, !tbaa !68
  %cmp.not.i.i.i2356 = icmp ult i32 %index.0.i.i.i2355, %3476
  %spec.store.select.i.i.i2357 = select i1 %cmp.not.i.i.i2356, i32 %index.0.i.i.i2355, i32 0
  %3477 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  %3478 = and i32 %spec.store.select.i.i.i2357, 31
  %3479 = lshr i64 %3477, %sh_prom.i.i.i.i756
  %3480 = trunc i64 %3479 to i32
  %conv4.i.i.i.i.i2361 = and i32 %3480, 1
  %shl.i.i.i.i.i2362 = shl nuw i32 %conv4.i.i.i.i.i2361, %3478
  br label %ComputeLoop5762, !amdgpu.uniform !42

3481:                                             ; preds = %ComputeEnd5763
  %sunkaddr7435 = getelementptr inbounds i8, ptr addrspace(1) %7940, i64 40
  %3482 = atomicrmw or ptr addrspace(1) %sunkaddr7435, i32 %.lcssa9183 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %3483, !amdgpu.uniform !42

3483:                                             ; preds = %3481, %ComputeEnd5763
  %3484 = phi i32 [ %3482, %3481 ], [ poison, %ComputeEnd5763 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %7949)
  %3485 = call i32 @llvm.amdgcn.readfirstlane.i32(i32 %3484)
  %3486 = or i32 %3485, %.lcssa9184
  %shl5.i.i.i.i.i2363 = shl nuw i32 1, %3478
  %and.i.i.i.i.i2364 = and i32 %shl5.i.i.i.i.i2363, %3486
  %tobool3.i.i.i.i2365 = icmp ne i32 %and.i.i.i.i.i2364, 0
  %3487 = call i64 @llvm.amdgcn.ballot.i64(i1 %tobool3.i.i.i.i2365)
  %and.i.i.i.i.i.i2366 = and i64 %3487, %3477
  %cmp.i.not.i.i.i2367 = icmp ne i64 %3477, %and.i.i.i.i.i.i2366
  br i1 %cmp.i.not.i.i.i2367, label %if.then.i.i.i.i2368, label %Flow8146, !amdgpu.uniform !42

if.then.i.i.i.i2368:                              ; preds = %3483
  fence syncscope("agent") acquire
  %3488 = load <2 x i64>, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !75
  %3489 = extractelement <2 x i64> %3488, i32 0
  %3490 = inttoptr i64 %3489 to ptr
  %3491 = extractelement <2 x i64> %3488, i32 1
  %3492 = inttoptr i64 %3491 to ptr
  %idxprom.i.i.i.i2369 = zext i32 %spec.store.select.i.i.i2357 to i64
  %arrayidx.i.i.i.i2370 = getelementptr inbounds nuw i32, ptr %3490, i64 %idxprom.i.i.i.i2369, !amdgpu.uniform !42
  %3493 = load atomic i32, ptr %arrayidx.i.i.i.i2370 monotonic, align 4
  %3494 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %3493)
  %arrayidx.i45.i.i.i2371 = getelementptr inbounds nuw i32, ptr %3492, i64 %idxprom.i.i.i.i2369, !amdgpu.uniform !42
  %3495 = load atomic i32, ptr %arrayidx.i45.i.i.i2371 monotonic, align 4
  %3496 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %3495)
  %cmp.i46.not.i.i.i2372 = icmp ne i32 %3494, %3496
  br i1 %cmp.i46.not.i.i.i2372, label %if.then12.i.i.i2373, label %Flow8144, !amdgpu.uniform !42

if.then12.i.i.i2373:                              ; preds = %if.then.i.i.i.i2368
  fence syncscope("agent") release
  %3497 = call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %3477, i1 true)
  %iszero.i.i.i.i.i.i.i2374 = icmp ne i64 %3477, 0
  %cmp2.i.i.i.i.i.i2375 = icmp eq i64 %3497, %sh_prom.i.i.i.i756
  %cmp.i.i.i.i.i.i2376 = select i1 %iszero.i.i.i.i.i.i.i2374, i1 %cmp2.i.i.i.i.i.i2375, i1 false
  %conv4.i.i51.i.i.i2377 = zext i1 %cmp.i.i.i.i.i.i2376 to i32
  %shl.i.i52.i.i.i2378 = shl nuw i32 %conv4.i.i51.i.i.i2377, %3478
  %xor.i.i.i.i.i2379 = xor i32 %shl.i.i52.i.i.i2378, -1
  %3498 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop5767, !amdgpu.uniform !42

3499:                                             ; preds = %ComputeEnd5768
  %sunkaddr7436 = getelementptr inbounds i8, ptr addrspace(1) %7940, i64 40
  %3500 = atomicrmw and ptr addrspace(1) %sunkaddr7436, i32 %.lcssa9185 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %3501, !amdgpu.uniform !42

3501:                                             ; preds = %3499, %ComputeEnd5768
  call void @llvm.amdgcn.end.cf.i64(i64 %7967)
  call void @llvm.amdgcn.wave.barrier()
  br label %Flow8144, !amdgpu.uniform !42

if.end14.i.i.i2439:                               ; preds = %Flow8144
  %3502 = call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %3477, i1 true)
  %iszero.i.i.i.i.i.i2440 = icmp ne i64 %3477, 0
  %cmp2.i.i.i.i.i2441 = icmp eq i64 %3502, %sh_prom.i.i.i.i756
  %cmp.i.i.i.i.i2442 = select i1 %iszero.i.i.i.i.i.i2440, i1 %cmp2.i.i.i.i.i2441, i1 false
  %3503 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.i.i.i.i.i2442)
  %3504 = extractvalue { i1, i64 } %3503, 0
  %3505 = extractvalue { i1, i64 } %3503, 1
  br i1 %3504, label %if.then16.i.i.i2444, label %if.end22.i.i.i2443

if.then16.i.i.i2444:                              ; preds = %if.end14.i.i.i2439
  %3506 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !76
  %opcode.i.i.i2445 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %3506, i64 %idxprom.i.i.i.i2369, i32 1
  store i32 1660944387, ptr %opcode.i.i.i2445, align 8, !tbaa !77
  %arrayidx21.i.i.i2446 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %3506, i64 %idxprom.i.i.i.i2369
  store i64 %3477, ptr %arrayidx21.i.i.i2446, align 8, !tbaa !80
  br label %if.end22.i.i.i2443, !amdgpu.uniform !42

Flow8145:                                         ; preds = %if.end22.i.i.i2443, %Flow8144
  %3507 = phi i32 [ %3494, %if.end22.i.i.i2443 ], [ %retval.sroa.8.0.i.i.i2354, %Flow8144 ]
  %3508 = phi i64 [ %3477, %if.end22.i.i.i2443 ], [ %retval.sroa.2.0.i.i.i2353, %Flow8144 ]
  %3509 = phi i32 [ %spec.store.select.i.i.i2357, %if.end22.i.i.i2443 ], [ %retval.sroa.6.0.i.i.i2351, %Flow8144 ]
  br label %cleanup26.i.i.i2380, !amdgpu.uniform !42

if.end22.i.i.i2443:                               ; preds = %if.then16.i.i.i2444, %if.end14.i.i.i2439
  call void @llvm.amdgcn.end.cf.i64(i64 %3505)
  call void @llvm.amdgcn.wave.barrier()
  br label %Flow8145, !amdgpu.uniform !42

Flow8146:                                         ; preds = %cleanup26.i.i.i2380, %3483
  %3510 = phi i32 [ %3509, %cleanup26.i.i.i2380 ], [ %3473, %3483 ]
  %3511 = phi i64 [ %3508, %cleanup26.i.i.i2380 ], [ %3474, %3483 ]
  %3512 = phi i32 [ %3507, %cleanup26.i.i.i2380 ], [ %3475, %3483 ]
  %3513 = phi i32 [ %3507, %cleanup26.i.i.i2380 ], [ %retval.sroa.8.0.i.i.i2354, %3483 ]
  %3514 = phi i64 [ %3508, %cleanup26.i.i.i2380 ], [ %retval.sroa.2.0.i.i.i2353, %3483 ]
  %3515 = phi i32 [ %3509, %cleanup26.i.i.i2380 ], [ %retval.sroa.6.0.i.i.i2351, %3483 ]
  %3516 = phi i1 [ %cmp.i46.not.i.i.i2372, %cleanup26.i.i.i2380 ], [ true, %3483 ]
  br i1 %3516, label %for.inc.i.i.i2385, label %Flow8147, !amdgpu.uniform !42

cleanup26.i.i.i2380:                              ; preds = %Flow8145
  br label %Flow8146, !amdgpu.uniform !42

for.inc.i.i.i2385:                                ; preds = %Flow8146
  %inc.i.i.i2390 = add i32 %spec.store.select.i.i.i2357, 1
  br label %Flow8147, !amdgpu.uniform !42

Flow8147:                                         ; preds = %for.inc.i.i.i2385, %Flow8146
  %3517 = phi i32 [ %inc.i.i.i2390, %for.inc.i.i.i2385 ], [ poison, %Flow8146 ]
  %3518 = phi i1 [ false, %for.inc.i.i.i2385 ], [ true, %Flow8146 ]
  br i1 %3518, label %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i2391, label %for.cond.i.i1.i2350, !amdgpu.uniform !42

_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i2391: ; preds = %Flow8147
  %.lcssa9188 = phi i32 [ %3510, %Flow8147 ]
  %.lcssa9187 = phi i64 [ %3511, %Flow8147 ]
  %.lcssa9186 = phi i32 [ %3512, %Flow8147 ]
  fence acquire
  %3519 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %mul.i.i.i.i.i.i2392 = shl i32 %.lcssa9188, 6
  %idxprom.i19.i.i.i.i.i2393 = zext i32 %mul.i.i.i.i.i.i2392 to i64
  %arrayidx.i20.i.i.i.i.i2394 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %3519, i64 %idxprom.i19.i.i.i.i.i2393
  %arrayidx.i22.i.i.i.i.i2395 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %arrayidx.i20.i.i.i.i.i2394, i64 %sh_prom.i.i.i.i756
  store i64 sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.21.38 to ptr), i64 25) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.21.38 to ptr) to i64)), ptr %arrayidx.i22.i.i.i.i.i2395, align 8, !tbaa !82
  br i1 %1134, label %loop-memcpy-expansion5349.preheader, label %Flow8143, !amdgpu.uniform !42

loop-memcpy-expansion5349.preheader:              ; preds = %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i2391
  %scevgep6981 = getelementptr i8, ptr %3519, i64 8
  %scevgep6982 = getelementptr i8, ptr %scevgep6981, i64 %1474
  %3520 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i2393, 6
  %scevgep6983 = getelementptr i8, ptr %scevgep6982, i64 %3520
  br label %loop-memcpy-expansion5349, !amdgpu.uniform !42

Flow8143:                                         ; preds = %Flow8142, %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i2391
  br label %loop-memcpy-residual-header5352, !amdgpu.uniform !42

loop-memcpy-expansion5349:                        ; preds = %loop-memcpy-expansion5349.preheader, %loop-memcpy-expansion5349
  %loop-index5350 = phi i64 [ %3523, %loop-memcpy-expansion5349 ], [ 0, %loop-memcpy-expansion5349.preheader ]
  %3521 = getelementptr i8, ptr addrspace(4) @.str.21.38, i64 %loop-index5350, !amdgpu.uniform !42
  %3522 = load <4 x i32>, ptr addrspace(4) %3521, align 1
  %scevgep6984 = getelementptr i8, ptr %scevgep6983, i64 %loop-index5350
  store <4 x i32> %3522, ptr %scevgep6984, align 1
  %3523 = add i64 %loop-index5350, 16
  %3524 = icmp uge i64 %3523, %1133
  br i1 %3524, label %Flow8142, label %loop-memcpy-expansion5349, !amdgpu.uniform !42

Flow8141:                                         ; preds = %Flow8139, %Flow8140
  %3525 = phi i64 [ %6617, %Flow8139 ], [ %6615, %Flow8140 ]
  br label %post-loop-memcpy-expansion5348, !amdgpu.uniform !42

loop-memcpy-residual5351:                         ; preds = %loop-memcpy-residual5351.preheader, %loop-memcpy-residual5351
  %residual-loop-index5353 = phi i64 [ %3527, %loop-memcpy-residual5351 ], [ 0, %loop-memcpy-residual5351.preheader ]
  %scevgep6989 = getelementptr i8, ptr addrspace(4) getelementptr inbounds nuw (i8, ptr addrspace(4) @.str.21.38, i64 16), i64 %residual-loop-index5353, !amdgpu.uniform !42
  %3526 = load i8, ptr addrspace(4) %scevgep6989, align 1
  %scevgep6988 = getelementptr i8, ptr %scevgep6987, i64 %residual-loop-index5353
  store i8 %3526, ptr %scevgep6988, align 1
  %3527 = add i64 %residual-loop-index5353, 1
  %3528 = icmp uge i64 %3527, %1132
  br i1 %3528, label %Flow8139, label %loop-memcpy-residual5351, !amdgpu.uniform !42

Flow8139:                                         ; preds = %loop-memcpy-residual5351
  br label %Flow8141, !amdgpu.uniform !42

post-loop-memcpy-expansion5348:                   ; preds = %Flow8141
  %tobool.not.i.i.i.i.i.i2398 = icmp eq i32 %.lcssa9186, 0
  %conv.i.i.i.i.i.i2399 = zext i1 %tobool.not.i.i.i.i.i.i2398 to i32
  fence release
  %3529 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %idxprom.i23.i.i.i.i.i2400 = zext i32 %.lcssa9188 to i64
  %arrayidx.i24.i.i.i.i.i2401 = getelementptr inbounds nuw i32, ptr %3529, i64 %idxprom.i23.i.i.i.i.i2400
  store atomic i32 %conv.i.i.i.i.i.i2399, ptr %arrayidx.i24.i.i.i.i.i2401 monotonic, align 4
  %3530 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !84
  %arrayidx.i.i5.i.i2402 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %3530, i64 %idxprom.i23.i.i.i.i.i2400
  %3531 = load i64, ptr %arrayidx.i.i5.i.i2402, align 8, !tbaa !80
  br label %while.cond.i.i.i.i2403, !amdgpu.uniform !42

while.cond.i.i.i.i2403:                           ; preds = %post-loop-memcpy-expansion5348, %Flow8138
  %phi.broken8734 = phi i64 [ 0, %post-loop-memcpy-expansion5348 ], [ %3553, %Flow8138 ]
  %lsr.iv6995 = phi i64 [ %3551, %Flow8138 ], [ -31, %post-loop-memcpy-expansion5348 ]
  %lsr.iv6990 = phi ptr addrspace(4) [ %3550, %Flow8138 ], [ getelementptr (i8, ptr addrspace(4) @.str.21.38, i64 56), %post-loop-memcpy-expansion5348 ]
  %port.sroa.37.0.i.i2404 = phi i32 [ %3549, %Flow8138 ], [ %conv.i.i.i.i.i.i2399, %post-loop-memcpy-expansion5348 ]
  %idx.0.i.i.i.i2405 = phi i64 [ %3548, %Flow8138 ], [ 56, %post-loop-memcpy-expansion5348 ]
  %cmp.i.i.i.i2406 = icmp ult i64 %idx.0.i.i.i.i2405, sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.21.38 to ptr), i64 25) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.21.38 to ptr) to i64))
  %3532 = call i64 @llvm.amdgcn.ballot.i64(i1 %cmp.i.i.i.i2406)
  %and.i.i.i.i6.i.i2407 = and i64 %3532, %3531
  %tobool.not.i.i.i.i2408 = icmp ne i64 %and.i.i.i.i6.i.i2407, 0
  %3533 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i22.i.i.i2409 = getelementptr inbounds nuw i32, ptr %3533, i64 %idxprom.i23.i.i.i.i.i2400
  %3534 = load atomic i32, ptr %arrayidx.i22.i.i.i2409 monotonic, align 4
  %3535 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %tobool.not.i.i.i.i2408)
  %3536 = extractvalue { i1, i64 } %3535, 0
  %3537 = extractvalue { i1, i64 } %3535, 1
  br i1 %3536, label %cond.false.i8.i.i.i.i2410, label %Flow8138

cond.false.i8.i.i.i.i2410:                        ; preds = %while.cond.i.i.i.i2403
  %3538 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %3534)
  %3539 = icmp ne i32 %3538, %port.sroa.37.0.i.i2404
  %3540 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3539)
  %3541 = extractvalue { i1, i64 } %3540, 0
  %3542 = extractvalue { i1, i64 } %3540, 1
  br i1 %3541, label %while.body.i.i21.i.i.i.i2411.preheader, label %Flow8137

while.body.i.i21.i.i.i.i2411.preheader:           ; preds = %cond.false.i8.i.i.i.i2410
  br label %while.body.i.i21.i.i.i.i2411, !amdgpu.uniform !42

Flow8137:                                         ; preds = %Flow8136, %cond.false.i8.i.i.i.i2410
  call void @llvm.amdgcn.end.cf.i64(i64 %3542)
  br label %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i2414, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i2411:                     ; preds = %while.body.i.i21.i.i.i.i2411.preheader, %while.body.i.i21.i.i.i.i2411
  %phi.broken8733 = phi i64 [ 0, %while.body.i.i21.i.i.i.i2411.preheader ], [ %3546, %while.body.i.i21.i.i.i.i2411 ]
  call void @llvm.amdgcn.s.sleep(i32 2)
  %3543 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i22.i.i.i.i2412 = getelementptr inbounds nuw i32, ptr %3543, i64 %idxprom.i23.i.i.i.i.i2400
  %3544 = load atomic i32, ptr %arrayidx.i.i.i22.i.i.i.i2412 monotonic, align 4
  %3545 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %3544)
  %cmp.i.not.i.i23.i.i.i.i2413 = icmp eq i32 %3545, %port.sroa.37.0.i.i2404
  %3546 = call i64 @llvm.amdgcn.if.break.i64(i1 %cmp.i.not.i.i23.i.i.i.i2413, i64 %phi.broken8733)
  %3547 = call i1 @llvm.amdgcn.loop.i64(i64 %3546)
  br i1 %3547, label %Flow8136, label %while.body.i.i21.i.i.i.i2411

Flow8136:                                         ; preds = %while.body.i.i21.i.i.i.i2411
  %.lcssa9189 = phi i64 [ %3546, %while.body.i.i21.i.i.i.i2411 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa9189)
  br label %Flow8137, !amdgpu.uniform !42

Flow8138:                                         ; preds = %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i2415, %while.cond.i.i.i.i2403
  %3548 = phi i64 [ %add.i.i.i.i2418, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i2415 ], [ poison, %while.cond.i.i.i.i2403 ]
  %3549 = phi i32 [ %conv.i.i26.i.i.i.i2416, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i2415 ], [ poison, %while.cond.i.i.i.i2403 ]
  %3550 = phi ptr addrspace(4) [ %scevgep6991, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i2415 ], [ poison, %while.cond.i.i.i.i2403 ]
  %3551 = phi i64 [ %lsr.iv.next6996, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i2415 ], [ poison, %while.cond.i.i.i.i2403 ]
  %3552 = phi i1 [ false, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i2415 ], [ true, %while.cond.i.i.i.i2403 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %3537)
  %3553 = call i64 @llvm.amdgcn.if.break.i64(i1 %3552, i64 %phi.broken8734)
  %3554 = call i1 @llvm.amdgcn.loop.i64(i64 %3553)
  br i1 %3554, label %cond.false.i.i.i2425, label %while.cond.i.i.i.i2403

_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i2414: ; preds = %Flow8137
  fence acquire
  br i1 %cmp.i.i.i.i2406, label %if.then.i.i.i.i.i.i.i2419, label %Flow8135, !amdgpu.uniform !42

if.then.i.i.i.i.i.i.i2419:                        ; preds = %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i2414
  %sub.i.i.i.i.i.i.i2421 = sub nuw i64 sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.21.38 to ptr), i64 25) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.21.38 to ptr) to i64)), %idx.0.i.i.i.i2405
  %spec.select.i.i.i36.i.i.i.i2422 = call i64 @llvm.umin.i64(i64 %sub.i.i.i.i.i.i.i2421, i64 64)
  %3555 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %3556 = and i64 %spec.select.i.i.i36.i.i.i.i2422, 15
  %3557 = sub i64 %spec.select.i.i.i36.i.i.i.i2422, %3556
  %3558 = icmp ne i64 %3557, 0
  br i1 %3558, label %loop-memcpy-expansion5355.preheader, label %Flow8134, !amdgpu.uniform !42

loop-memcpy-expansion5355.preheader:              ; preds = %if.then.i.i.i.i.i.i.i2419
  %arrayidx.i20.i35.i.i.i.i2423 = getelementptr %"struct.rpc::Buffer", ptr %3555, i64 %idxprom.i19.i.i.i.i.i2393
  %arrayidx.i22.i38.i.i.i.i2424 = getelementptr %"struct.rpc::Buffer", ptr %arrayidx.i20.i35.i.i.i.i2423, i64 %sh_prom.i.i.i.i756
  br label %loop-memcpy-expansion5355, !amdgpu.uniform !42

Flow8134:                                         ; preds = %Flow8133, %if.then.i.i.i.i.i.i.i2419
  br label %loop-memcpy-residual-header5358, !amdgpu.uniform !42

loop-memcpy-expansion5355:                        ; preds = %loop-memcpy-expansion5355.preheader, %loop-memcpy-expansion5355
  %loop-index5356 = phi i64 [ %3561, %loop-memcpy-expansion5355 ], [ 0, %loop-memcpy-expansion5355.preheader ]
  %scevgep6992 = getelementptr i8, ptr addrspace(4) %lsr.iv6990, i64 %loop-index5356, !amdgpu.uniform !42
  %3559 = load <4 x i32>, ptr addrspace(4) %scevgep6992, align 1
  %3560 = getelementptr i8, ptr %arrayidx.i22.i38.i.i.i.i2424, i64 %loop-index5356
  store <4 x i32> %3559, ptr %3560, align 1
  %3561 = add i64 %loop-index5356, 16
  %3562 = icmp uge i64 %3561, %3557
  br i1 %3562, label %Flow8133, label %loop-memcpy-expansion5355, !amdgpu.uniform !42

Flow8132:                                         ; preds = %Flow8131, %loop-memcpy-residual-header5358
  br label %Flow8135, !amdgpu.uniform !42

loop-memcpy-residual5357:                         ; preds = %loop-memcpy-residual5357.preheader, %loop-memcpy-residual5357
  %residual-loop-index5359 = phi i64 [ %3564, %loop-memcpy-residual5357 ], [ 0, %loop-memcpy-residual5357.preheader ]
  %scevgep7001 = getelementptr i8, ptr addrspace(4) %scevgep7000, i64 %residual-loop-index5359, !amdgpu.uniform !42
  %3563 = load i8, ptr addrspace(4) %scevgep7001, align 1
  %scevgep6999 = getelementptr i8, ptr %scevgep6998, i64 %residual-loop-index5359
  store i8 %3563, ptr %scevgep6999, align 1
  %3564 = add i64 %residual-loop-index5359, 1
  %3565 = icmp uge i64 %3564, %3556
  br i1 %3565, label %Flow8131, label %loop-memcpy-residual5357, !amdgpu.uniform !42

Flow8131:                                         ; preds = %loop-memcpy-residual5357
  br label %Flow8132, !amdgpu.uniform !42

_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i2415: ; preds = %Flow8135
  %conv.i.i26.i.i.i.i2416 = xor i32 %port.sroa.37.0.i.i2404, 1
  fence release
  %3566 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %arrayidx.i24.i29.i.i.i.i2417 = getelementptr inbounds nuw i32, ptr %3566, i64 %idxprom.i23.i.i.i.i.i2400
  store atomic i32 %conv.i.i26.i.i.i.i2416, ptr %arrayidx.i24.i29.i.i.i.i2417 monotonic, align 4
  %add.i.i.i.i2418 = add i64 %idx.0.i.i.i.i2405, 64
  %scevgep6991 = getelementptr i8, ptr addrspace(4) %lsr.iv6990, i64 64
  %lsr.iv.next6996 = add i64 %lsr.iv6995, -64
  br label %Flow8138, !amdgpu.uniform !42

cond.false.i.i.i2425:                             ; preds = %Flow8138
  %.lcssa9191 = phi i64 [ %3553, %Flow8138 ]
  %port.sroa.37.0.i.i2404.lcssa = phi i32 [ %port.sroa.37.0.i.i2404, %Flow8138 ]
  %.lcssa9190 = phi i32 [ %3534, %Flow8138 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa9191)
  %3567 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %.lcssa9190)
  %3568 = icmp ne i32 %3567, %port.sroa.37.0.i.i2404.lcssa
  %3569 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3568)
  %3570 = extractvalue { i1, i64 } %3569, 0
  %3571 = extractvalue { i1, i64 } %3569, 1
  br i1 %3570, label %while.body.i.i.i.i2426.preheader, label %Flow8130

while.body.i.i.i.i2426.preheader:                 ; preds = %cond.false.i.i.i2425
  br label %while.body.i.i.i.i2426, !amdgpu.uniform !42

Flow8130:                                         ; preds = %Flow8129, %cond.false.i.i.i2425
  call void @llvm.amdgcn.end.cf.i64(i64 %3571)
  br label %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit2447, !amdgpu.uniform !42

while.body.i.i.i.i2426:                           ; preds = %while.body.i.i.i.i2426.preheader, %while.body.i.i.i.i2426
  %phi.broken8735 = phi i64 [ 0, %while.body.i.i.i.i2426.preheader ], [ %3575, %while.body.i.i.i.i2426 ]
  call void @llvm.amdgcn.s.sleep(i32 2)
  %3572 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i8.i.i2427 = getelementptr inbounds nuw i32, ptr %3572, i64 %idxprom.i23.i.i.i.i.i2400
  %3573 = load atomic i32, ptr %arrayidx.i.i.i8.i.i2427 monotonic, align 4
  %3574 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %3573)
  %cmp.i.not.i.i.i.i2428 = icmp eq i32 %3574, %port.sroa.37.0.i.i2404.lcssa
  %3575 = call i64 @llvm.amdgcn.if.break.i64(i1 %cmp.i.not.i.i.i.i2428, i64 %phi.broken8735)
  %3576 = call i1 @llvm.amdgcn.loop.i64(i64 %3575)
  br i1 %3576, label %Flow8129, label %while.body.i.i.i.i2426

Flow8129:                                         ; preds = %while.body.i.i.i.i2426
  %.lcssa9192 = phi i64 [ %3575, %while.body.i.i.i.i2426 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa9192)
  br label %Flow8130, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit2447: ; preds = %Flow8130
  fence acquire
  call void @llvm.amdgcn.wave.barrier()
  fence syncscope("agent") release
  %3577 = call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %.lcssa9187, i1 true)
  %iszero.i.i.i.i.i13.i.i2430 = icmp ne i64 %.lcssa9187, 0
  %cmp2.i.i.i.i14.i.i2431 = icmp eq i64 %3577, %sh_prom.i.i.i.i756
  %cmp.i.i.i.i15.i.i2432 = select i1 %iszero.i.i.i.i.i13.i.i2430, i1 %cmp2.i.i.i.i14.i.i2431, i1 false
  %3578 = and i32 %.lcssa9188, 31
  %conv4.i.i.i19.i.i2436 = zext i1 %cmp.i.i.i.i15.i.i2432 to i32
  %shl.i.i.i20.i.i2437 = shl nuw i32 %conv4.i.i.i19.i.i2436, %3578
  %xor.i.i.i21.i.i2438 = xor i32 %shl.i.i.i20.i.i2437, -1
  %3579 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop5771, !amdgpu.uniform !42

3580:                                             ; preds = %ComputeEnd5772
  %div8.i.i.i16.i.i2433 = lshr i32 %.lcssa9188, 5
  %idxprom.i.i.i17.i.i2434 = zext nneg i32 %div8.i.i.i16.i.i2433 to i64
  %3581 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i17.i.i2434
  %sunkaddr7437 = getelementptr inbounds i8, ptr addrspace(1) %3581, i64 40
  %3582 = atomicrmw and ptr addrspace(1) %sunkaddr7437, i32 %.lcssa9193 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %3583, !amdgpu.uniform !42

3583:                                             ; preds = %3580, %ComputeEnd5772
  call void @llvm.amdgcn.end.cf.i64(i64 %7984)
  call void @llvm.amdgcn.wave.barrier()
  br label %Flow8149, !amdgpu.uniform !42

if.else:                                          ; preds = %3432
  %sub = sub nuw i64 %retval.1.i1278, %retval.1.i
  %mul = mul i64 %sub, 1000
  %sunkaddr7438 = getelementptr inbounds i8, ptr %Options, i64 9
  %3584 = load i8, ptr %sunkaddr7438, align 1, !tbaa !161, !range !152, !noundef !42
  %loadedv72 = trunc nuw i8 %3584 to i1
  %cmp73 = icmp ugt i64 %mul, 999999
  %or.cond = select i1 %loadedv72, i1 true, i1 %cmp73
  %or.cond.inv = xor i1 %or.cond, true
  %3585 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %or.cond.inv)
  %3586 = extractvalue { i1, i64 } %3585, 0
  %3587 = extractvalue { i1, i64 } %3585, 1
  br i1 %3586, label %if.else77, label %Flow8127

Flow8127:                                         ; preds = %Flow8063, %if.else
  %3588 = phi i32 [ %4196, %Flow8063 ], [ poison, %if.else ]
  %3589 = phi i64 [ %4197, %Flow8063 ], [ poison, %if.else ]
  %3590 = phi i32 [ %4198, %Flow8063 ], [ poison, %if.else ]
  %3591 = phi i32 [ %4199, %Flow8063 ], [ poison, %if.else ]
  %3592 = phi i64 [ %4200, %Flow8063 ], [ poison, %if.else ]
  %3593 = phi i32 [ %4201, %Flow8063 ], [ poison, %if.else ]
  %3594 = phi i32 [ %4202, %Flow8063 ], [ poison, %if.else ]
  %3595 = phi i64 [ %4203, %Flow8063 ], [ poison, %if.else ]
  %3596 = phi i32 [ %4204, %Flow8063 ], [ poison, %if.else ]
  %3597 = phi i32 [ %4205, %Flow8063 ], [ poison, %if.else ]
  %3598 = phi i64 [ %4206, %Flow8063 ], [ poison, %if.else ]
  %3599 = phi i32 [ %4207, %Flow8063 ], [ poison, %if.else ]
  %3600 = call { i1, i64 } @llvm.amdgcn.else.i64.i64(i64 %3587)
  %3601 = extractvalue { i1, i64 } %3600, 0
  %3602 = extractvalue { i1, i64 } %3600, 1
  br i1 %3601, label %if.then74, label %Flow8128

if.then74:                                        ; preds = %Flow8127
  call void @llvm.lifetime.start.p5(ptr addrspace(5) %buffer.i.i.i2988) #34, !noalias !162
  %cmp7.i.i.i.i.i.i2449 = icmp uge i64 %mul, 1000000
  br i1 %cmp7.i.i.i.i.i.i2449, label %while.body.i.i.preheader.i.i.i.i, label %Flow8126, !amdgpu.uniform !42

while.body.i.i.preheader.i.i.i.i:                 ; preds = %if.then74
  %div = udiv i64 %mul, 1000000
  br label %while.body.i.i.i.i.i.i2450, !amdgpu.uniform !42

Flow8126:                                         ; preds = %Flow8125, %if.then74
  %3603 = phi i1 [ %cmp.i.i.i.i.i.i.i.i2459.lcssa, %Flow8125 ], [ false, %if.then74 ]
  %3604 = phi i64 [ %writer.sroa.8.1.i.i.i.i.lcssa, %Flow8125 ], [ 0, %if.then74 ]
  br label %_ZN22__llvm_libc_22_0_0_git15IntegerToStringIyNS_7details3FmtILh10ELb0ELb0ELb0ELm1EEEE13IntegerWriter25write_unsigned_number_decEyRNS1_22StringBufferWriterImplILb0EEE.exit.i.i.i.i.i, !amdgpu.uniform !42

while.body.i.i.i.i.i.i2450:                       ; preds = %while.body.i.i.preheader.i.i.i.i, %_ZN22__llvm_libc_22_0_0_git7details22StringBufferWriterImplILb0EE4pushEc.exit.i.i.i.i.i.i2463
  %writer.sroa.8.0.i.i.i.i = phi i64 [ %writer.sroa.8.1.i.i.i.i, %_ZN22__llvm_libc_22_0_0_git7details22StringBufferWriterImplILb0EE4pushEc.exit.i.i.i.i.i.i2463 ], [ 0, %while.body.i.i.preheader.i.i.i.i ]
  %value.addr.09.i.i.i.i.i.i2451 = phi i64 [ %div.i.i.i.i.i.i.i2452, %_ZN22__llvm_libc_22_0_0_git7details22StringBufferWriterImplILb0EE4pushEc.exit.i.i.i.i.i.i2463 ], [ %div, %while.body.i.i.preheader.i.i.i.i ]
  br label %NodeBlock7541, !amdgpu.uniform !42

NodeBlock7541:                                    ; preds = %while.body.i.i.i.i.i.i2450
  %div.i.i.i.i.i.i.i2452 = udiv i64 %value.addr.09.i.i.i.i.i.i2451, 10
  %.neg.i2453 = mul nuw nsw i64 %div.i.i.i.i.i.i.i2452, 246
  %rem.i.i.i.i.i.i.decomposed.i2454 = add nuw nsw i64 %.neg.i2453, %value.addr.09.i.i.i.i.i.i2451
  %conv.i.i.i.i.i.i.i2455 = trunc i64 %rem.i.i.i.i.i.i.decomposed.i2454 to i8
  %3605 = zext i8 %conv.i.i.i.i.i.i.i2455 to i16
  %Pivot7542 = icmp sge i16 %3605, 5
  br i1 %Pivot7542, label %NodeBlock7539, label %Flow8124, !amdgpu.uniform !42

NodeBlock7539:                                    ; preds = %NodeBlock7541
  %Pivot7540 = icmp sge i16 %3605, 7
  br i1 %Pivot7540, label %NodeBlock7537, label %Flow8115, !amdgpu.uniform !42

NodeBlock7537:                                    ; preds = %NodeBlock7539
  %Pivot7538 = icmp sge i16 %3605, 8
  br i1 %Pivot7538, label %NodeBlock7535, label %Flow8111, !amdgpu.uniform !42

NodeBlock7535:                                    ; preds = %NodeBlock7537
  %Pivot7536 = icmp sge i16 %3605, 9
  br i1 %Pivot7536, label %sw.bb9.i.i.i.i.i.i.i.i2456, label %Flow8109, !amdgpu.uniform !42

Flow8115:                                         ; preds = %Flow8112, %NodeBlock7539
  %3606 = phi i8 [ %3619, %Flow8112 ], [ poison, %NodeBlock7539 ]
  %3607 = phi i1 [ false, %Flow8112 ], [ true, %NodeBlock7539 ]
  br i1 %3607, label %NodeBlock7533, label %Flow8116, !amdgpu.uniform !42

NodeBlock7533:                                    ; preds = %Flow8115
  %Pivot7534 = icmp sge i16 %3605, 6
  br i1 %Pivot7534, label %sw.bb6.i.i.i.i.i.i.i.i2600, label %Flow8113, !amdgpu.uniform !42

Flow8124:                                         ; preds = %Flow8116, %NodeBlock7541
  %3608 = phi i8 [ %3621, %Flow8116 ], [ poison, %NodeBlock7541 ]
  %3609 = phi i1 [ false, %Flow8116 ], [ true, %NodeBlock7541 ]
  br i1 %3609, label %NodeBlock7531, label %if.then.i.i.i.i.i.i.i2457, !amdgpu.uniform !42

NodeBlock7531:                                    ; preds = %Flow8124
  %Pivot7532 = icmp sge i16 %3605, 3
  br i1 %Pivot7532, label %NodeBlock7529, label %Flow8122, !amdgpu.uniform !42

NodeBlock7529:                                    ; preds = %NodeBlock7531
  %Pivot7530 = icmp sge i16 %3605, 4
  br i1 %Pivot7530, label %sw.bb4.i.i.i.i.i.i.i.i2602, label %Flow8117, !amdgpu.uniform !42

Flow8122:                                         ; preds = %Flow8118, %NodeBlock7531
  %3610 = phi i8 [ %3622, %Flow8118 ], [ poison, %NodeBlock7531 ]
  %3611 = phi i1 [ false, %Flow8118 ], [ true, %NodeBlock7531 ]
  br i1 %3611, label %NodeBlock7527, label %Flow8123, !amdgpu.uniform !42

NodeBlock7527:                                    ; preds = %Flow8122
  %Pivot7528 = icmp sge i16 %3605, 2
  br i1 %Pivot7528, label %sw.bb2.i.i.i.i.i.i.i.i2604, label %Flow8120, !amdgpu.uniform !42

Flow8120:                                         ; preds = %sw.bb2.i.i.i.i.i.i.i.i2604, %NodeBlock7527
  %3612 = phi i1 [ false, %sw.bb2.i.i.i.i.i.i.i.i2604 ], [ true, %NodeBlock7527 ]
  br i1 %3612, label %LeafBlock7525, label %Flow8121, !amdgpu.uniform !42

LeafBlock7525:                                    ; preds = %Flow8120
  %SwitchLeaf7526 = icmp eq i16 %3605, 1
  br i1 %SwitchLeaf7526, label %sw.bb1.i.i.i.i.i.i.i.i2605, label %Flow8119, !amdgpu.uniform !42

sw.bb1.i.i.i.i.i.i.i.i2605:                       ; preds = %LeafBlock7525
  br label %Flow8119, !amdgpu.uniform !42

sw.bb2.i.i.i.i.i.i.i.i2604:                       ; preds = %NodeBlock7527
  br label %Flow8120, !amdgpu.uniform !42

Flow8117:                                         ; preds = %sw.bb4.i.i.i.i.i.i.i.i2602, %NodeBlock7529
  %3613 = phi i1 [ false, %sw.bb4.i.i.i.i.i.i.i.i2602 ], [ true, %NodeBlock7529 ]
  br i1 %3613, label %sw.bb3.i.i.i.i.i.i.i.i2603, label %Flow8118, !amdgpu.uniform !42

sw.bb3.i.i.i.i.i.i.i.i2603:                       ; preds = %Flow8117
  br label %Flow8118, !amdgpu.uniform !42

sw.bb4.i.i.i.i.i.i.i.i2602:                       ; preds = %NodeBlock7529
  br label %Flow8117, !amdgpu.uniform !42

Flow8113:                                         ; preds = %sw.bb6.i.i.i.i.i.i.i.i2600, %NodeBlock7533
  %3614 = phi i1 [ false, %sw.bb6.i.i.i.i.i.i.i.i2600 ], [ true, %NodeBlock7533 ]
  br i1 %3614, label %sw.bb5.i.i.i.i.i.i.i.i2601, label %Flow8114, !amdgpu.uniform !42

sw.bb5.i.i.i.i.i.i.i.i2601:                       ; preds = %Flow8113
  br label %Flow8114, !amdgpu.uniform !42

sw.bb6.i.i.i.i.i.i.i.i2600:                       ; preds = %NodeBlock7533
  br label %Flow8113, !amdgpu.uniform !42

Flow8111:                                         ; preds = %Flow8110, %NodeBlock7537
  %3615 = phi i8 [ %3618, %Flow8110 ], [ poison, %NodeBlock7537 ]
  %3616 = phi i1 [ false, %Flow8110 ], [ true, %NodeBlock7537 ]
  br i1 %3616, label %sw.bb7.i.i.i.i.i.i.i.i2599, label %Flow8112, !amdgpu.uniform !42

sw.bb7.i.i.i.i.i.i.i.i2599:                       ; preds = %Flow8111
  br label %Flow8112, !amdgpu.uniform !42

Flow8109:                                         ; preds = %sw.bb9.i.i.i.i.i.i.i.i2456, %NodeBlock7535
  %3617 = phi i1 [ false, %sw.bb9.i.i.i.i.i.i.i.i2456 ], [ true, %NodeBlock7535 ]
  br i1 %3617, label %sw.bb8.i.i.i.i.i.i.i.i2598, label %Flow8110, !amdgpu.uniform !42

sw.bb8.i.i.i.i.i.i.i.i2598:                       ; preds = %Flow8109
  br label %Flow8110, !amdgpu.uniform !42

sw.bb9.i.i.i.i.i.i.i.i2456:                       ; preds = %NodeBlock7535
  br label %Flow8109, !amdgpu.uniform !42

Flow8110:                                         ; preds = %sw.bb8.i.i.i.i.i.i.i.i2598, %Flow8109
  %3618 = phi i8 [ 56, %sw.bb8.i.i.i.i.i.i.i.i2598 ], [ 57, %Flow8109 ]
  br label %Flow8111, !amdgpu.uniform !42

Flow8112:                                         ; preds = %sw.bb7.i.i.i.i.i.i.i.i2599, %Flow8111
  %3619 = phi i8 [ 55, %sw.bb7.i.i.i.i.i.i.i.i2599 ], [ %3615, %Flow8111 ]
  br label %Flow8115, !amdgpu.uniform !42

Flow8114:                                         ; preds = %sw.bb5.i.i.i.i.i.i.i.i2601, %Flow8113
  %3620 = phi i8 [ 53, %sw.bb5.i.i.i.i.i.i.i.i2601 ], [ 54, %Flow8113 ]
  br label %Flow8116, !amdgpu.uniform !42

Flow8116:                                         ; preds = %Flow8114, %Flow8115
  %3621 = phi i8 [ %3620, %Flow8114 ], [ %3606, %Flow8115 ]
  br label %Flow8124, !amdgpu.uniform !42

Flow8118:                                         ; preds = %sw.bb3.i.i.i.i.i.i.i.i2603, %Flow8117
  %3622 = phi i8 [ 51, %sw.bb3.i.i.i.i.i.i.i.i2603 ], [ 52, %Flow8117 ]
  br label %Flow8122, !amdgpu.uniform !42

Flow8119:                                         ; preds = %sw.bb1.i.i.i.i.i.i.i.i2605, %LeafBlock7525
  %3623 = phi i8 [ 49, %sw.bb1.i.i.i.i.i.i.i.i2605 ], [ 48, %LeafBlock7525 ]
  br label %Flow8121, !amdgpu.uniform !42

Flow8121:                                         ; preds = %Flow8119, %Flow8120
  %3624 = phi i8 [ %3623, %Flow8119 ], [ 50, %Flow8120 ]
  br label %Flow8123, !amdgpu.uniform !42

Flow8123:                                         ; preds = %Flow8121, %Flow8122
  %3625 = phi i8 [ %3624, %Flow8121 ], [ %3610, %Flow8122 ]
  br label %if.then.i.i.i.i.i.i.i2457, !amdgpu.uniform !42

if.then.i.i.i.i.i.i.i2457:                        ; preds = %Flow8124, %Flow8123
  %retval.0.i.i.i.i.i.i.i.i2458 = phi i8 [ %3608, %Flow8124 ], [ %3625, %Flow8123 ]
  %cmp.i.i.i.i.i.i.i.i2459 = icmp eq i64 %writer.sroa.8.0.i.i.i.i, 21
  %cmp.i.i.i.i.i.i.i.i2459.inv = xor i1 %cmp.i.i.i.i.i.i.i.i2459, true
  br i1 %cmp.i.i.i.i.i.i.i.i2459.inv, label %if.then3.i.i.i.i.i.i.i2460, label %_ZN22__llvm_libc_22_0_0_git7details22StringBufferWriterImplILb0EE4pushEc.exit.i.i.i.i.i.i2463, !amdgpu.uniform !42

if.then3.i.i.i.i.i.i.i2460:                       ; preds = %if.then.i.i.i.i.i.i.i2457
  %3626 = trunc i64 %writer.sroa.8.0.i.i.i.i to i32
  %3627 = xor i32 %3626, -1
  %3628 = getelementptr i8, ptr addrspace(5) %buffer.i.i.i2988, i32 %3627
  %arrayidx.i.i.i.i.i.i.i.i24616065 = getelementptr i8, ptr addrspace(5) %3628, i32 21
  %inc.i.i.i.i.i.i.i2462 = add i64 %writer.sroa.8.0.i.i.i.i, 1
  store i8 %retval.0.i.i.i.i.i.i.i.i2458, ptr addrspace(5) %arrayidx.i.i.i.i.i.i.i.i24616065, align 1, !tbaa !57, !noalias !162
  br label %_ZN22__llvm_libc_22_0_0_git7details22StringBufferWriterImplILb0EE4pushEc.exit.i.i.i.i.i.i2463, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git7details22StringBufferWriterImplILb0EE4pushEc.exit.i.i.i.i.i.i2463: ; preds = %if.then3.i.i.i.i.i.i.i2460, %if.then.i.i.i.i.i.i.i2457
  %writer.sroa.8.1.i.i.i.i = phi i64 [ %inc.i.i.i.i.i.i.i2462, %if.then3.i.i.i.i.i.i.i2460 ], [ 21, %if.then.i.i.i.i.i.i.i2457 ]
  %cmp.i.i.i.i.i.i2464 = icmp samesign ult i64 %value.addr.09.i.i.i.i.i.i2451, 10
  %.not.i.i.i.i.i.i2465 = or i1 %cmp.i.i.i.i.i.i.i.i2459, %cmp.i.i.i.i.i.i2464
  br i1 %.not.i.i.i.i.i.i2465, label %Flow8125, label %while.body.i.i.i.i.i.i2450, !amdgpu.uniform !42

Flow8125:                                         ; preds = %_ZN22__llvm_libc_22_0_0_git7details22StringBufferWriterImplILb0EE4pushEc.exit.i.i.i.i.i.i2463
  %writer.sroa.8.1.i.i.i.i.lcssa = phi i64 [ %writer.sroa.8.1.i.i.i.i, %_ZN22__llvm_libc_22_0_0_git7details22StringBufferWriterImplILb0EE4pushEc.exit.i.i.i.i.i.i2463 ]
  %cmp.i.i.i.i.i.i.i.i2459.lcssa = phi i1 [ %cmp.i.i.i.i.i.i.i.i2459, %_ZN22__llvm_libc_22_0_0_git7details22StringBufferWriterImplILb0EE4pushEc.exit.i.i.i.i.i.i2463 ]
  br label %Flow8126, !amdgpu.uniform !42

Flow8128:                                         ; preds = %3892, %Flow8127
  %3629 = phi i32 [ %1215, %3892 ], [ %3588, %Flow8127 ]
  %3630 = phi i64 [ %1214, %3892 ], [ %3589, %Flow8127 ]
  %3631 = phi i32 [ %1213, %3892 ], [ %3590, %Flow8127 ]
  %3632 = phi i32 [ %1212, %3892 ], [ %3591, %Flow8127 ]
  %3633 = phi i64 [ %1211, %3892 ], [ %3592, %Flow8127 ]
  %3634 = phi i32 [ %1210, %3892 ], [ %3593, %Flow8127 ]
  %3635 = phi i32 [ %1209, %3892 ], [ %3594, %Flow8127 ]
  %3636 = phi i64 [ %1208, %3892 ], [ %3595, %Flow8127 ]
  %3637 = phi i32 [ %1207, %3892 ], [ %3596, %Flow8127 ]
  %3638 = phi i32 [ %1206, %3892 ], [ %3597, %Flow8127 ]
  %3639 = phi i64 [ %1205, %3892 ], [ %3598, %Flow8127 ]
  %3640 = phi i32 [ %1204, %3892 ], [ %3599, %Flow8127 ]
  %3641 = phi i32 [ %.lcssa9175, %3892 ], [ %1203, %Flow8127 ]
  %3642 = phi i64 [ %.lcssa9176, %3892 ], [ %1202, %Flow8127 ]
  %3643 = phi i32 [ %.lcssa9177, %3892 ], [ %1201, %Flow8127 ]
  %3644 = phi i32 [ %.lcssa9164, %3892 ], [ %1200, %Flow8127 ]
  %3645 = phi i64 [ %.lcssa9165, %3892 ], [ %1199, %Flow8127 ]
  %3646 = phi i32 [ %.lcssa9166, %3892 ], [ %1198, %Flow8127 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %3602)
  br label %Flow8148, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git15IntegerToStringIyNS_7details3FmtILh10ELb0ELb0ELb0ELm1EEEE13IntegerWriter25write_unsigned_number_decEyRNS1_22StringBufferWriterImplILb0EEE.exit.i.i.i.i.i: ; preds = %Flow8126
  %cmp.i15.i.i.i.i = icmp ne i64 %3604, 0
  %or.cond.not.i.i.i.i2467 = select i1 %3603, i1 true, i1 %cmp.i15.i.i.i.i
  %or.cond.not.i.i.i.i2467.inv = xor i1 %or.cond.not.i.i.i.i2467, true
  br i1 %or.cond.not.i.i.i.i2467.inv, label %if.then3.i.i.lr.ph.i.i.i.i2468, label %_ZN22__llvm_libc_22_0_0_git15IntegerToStringIyNS_7details3FmtILh10ELb0ELb0ELb0ELm1EEEEC2Ey.exit.i.i.i, !amdgpu.uniform !42

if.then3.i.i.lr.ph.i.i.i.i2468:                   ; preds = %_ZN22__llvm_libc_22_0_0_git15IntegerToStringIyNS_7details3FmtILh10ELb0ELb0ELb0ELm1EEEE13IntegerWriter25write_unsigned_number_decEyRNS1_22StringBufferWriterImplILb0EEE.exit.i.i.i.i.i
  %sunkaddr7439 = getelementptr inbounds i8, ptr addrspace(5) %buffer.i.i.i2988, i32 20
  store i8 48, ptr addrspace(5) %sunkaddr7439, align 4, !tbaa !57, !noalias !162
  br label %_ZN22__llvm_libc_22_0_0_git15IntegerToStringIyNS_7details3FmtILh10ELb0ELb0ELb0ELm1EEEEC2Ey.exit.i.i.i, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git15IntegerToStringIyNS_7details3FmtILh10ELb0ELb0ELb0ELm1EEEEC2Ey.exit.i.i.i: ; preds = %if.then3.i.i.lr.ph.i.i.i.i2468, %_ZN22__llvm_libc_22_0_0_git15IntegerToStringIyNS_7details3FmtILh10ELb0ELb0ELb0ELm1EEEE13IntegerWriter25write_unsigned_number_decEyRNS1_22StringBufferWriterImplILb0EEE.exit.i.i.i.i.i
  %writer.sroa.8.4.i.i.i.i = phi i64 [ 1, %if.then3.i.i.lr.ph.i.i.i.i2468 ], [ %3604, %_ZN22__llvm_libc_22_0_0_git15IntegerToStringIyNS_7details3FmtILh10ELb0ELb0ELb0ELm1EEEE13IntegerWriter25write_unsigned_number_decEyRNS1_22StringBufferWriterImplILb0EEE.exit.i.i.i.i.i ]
  %sunkaddr7440 = getelementptr inbounds i8, ptr addrspace(5) %buffer.i.i.i2988, i32 24
  store i64 %writer.sroa.8.4.i.i.i.i, ptr addrspace(5) %sunkaddr7440, align 8, !tbaa !167, !noalias !162
  %cmp.i.not.i.i.i.i.i2471 = icmp ne i64 %writer.sroa.8.4.i.i.i.i, 0
  br i1 %cmp.i.not.i.i.i.i.i2471, label %if.then.i.i.i.i.i.i2472, label %Flow8108, !amdgpu.uniform !42

if.then.i.i.i.i.i.i2472:                          ; preds = %_ZN22__llvm_libc_22_0_0_git15IntegerToStringIyNS_7details3FmtILh10ELb0ELb0ELb0ELm1EEEEC2Ey.exit.i.i.i
  %3647 = call { i64, i1 } @llvm.uadd.with.overflow.i64(i64 %writer.sroa.8.4.i.i.i.i, i64 1)
  %ov7442 = extractvalue { i64, i1 } %3647, 1
  %ov7442.inv = xor i1 %ov7442, true
  br i1 %ov7442.inv, label %if.end.i.i.i.i.i.i.i2475, label %Flow8107, !amdgpu.uniform !42

if.end.i.i.i.i.i.i.i2475:                         ; preds = %if.then.i.i.i.i.i.i2472
  %math7441 = extractvalue { i64, i1 } %3647, 0
  %cmp2.i.i.i.i.i.i.i2476 = icmp ult i64 %math7441, 1676976733973595601
  br i1 %cmp2.i.i.i.i.i.i.i2476, label %if.then3.i.i.i.i4.i.i.i2595, label %if.end4.i.i.i.i.i.i.i2477, !amdgpu.uniform !42

if.then3.i.i.i.i4.i.i.i2595:                      ; preds = %if.end.i.i.i.i.i.i.i2475
  %mul.i.i.i.i.i.i.i2596 = mul nuw i64 %math7441, 11
  %div18.i.i.i.i.i.i.i2597 = lshr i64 %mul.i.i.i.i.i.i.i2596, 3
  br label %if.end4.i.i.i.i.i.i.i2477, !amdgpu.uniform !42

Flow8107:                                         ; preds = %if.end4.i.i.i.i.i.i.i2477, %if.then.i.i.i.i.i.i2472
  %3648 = phi ptr [ %cond.i.i.i2485, %if.end4.i.i.i.i.i.i.i2477 ], [ addrspacecast (ptr addrspace(4) @_ZN22__llvm_libc_22_0_0_git3cpp6string14NULL_CHARACTERE to ptr), %if.then.i.i.i.i.i.i2472 ]
  br label %_ZN22__llvm_libc_22_0_0_git3cpp6string7reserveEm.exit.i.i.i.i.i.i2487, !amdgpu.uniform !42

if.end4.i.i.i.i.i.i.i2477:                        ; preds = %if.then3.i.i.i.i4.i.i.i2595, %if.end.i.i.i.i.i.i.i2475
  %new_capacity.addr.0.i.i.i.i.i.i.i2478 = phi i64 [ %div18.i.i.i.i.i.i.i2597, %if.then3.i.i.i.i4.i.i.i2595 ], [ %math7441, %if.end.i.i.i.i.i.i.i2475 ]
  %sub.i.i.i2479 = add i64 %new_capacity.addr.0.i.i.i.i.i.i.i2478, 7
  %div2.i.i.i2480 = and i64 %sub.i.i.i2479, -8
  %3649 = load ptr, ptr addrspace(1) @_ZN12_GLOBAL__N_13ptrE, align 8, !tbaa !55
  %add.ptr.i.i.i2481 = getelementptr inbounds nuw i8, ptr %3649, i64 %div2.i.i.i2480
  store ptr %add.ptr.i.i.i2481, ptr addrspace(1) @_ZN12_GLOBAL__N_13ptrE, align 8, !tbaa !55
  %sub.ptr.lhs.cast.i.i.i2482 = ptrtoint ptr %add.ptr.i.i.i2481 to i64
  %sub.ptr.sub.i.i.i2483 = sub i64 %sub.ptr.lhs.cast.i.i.i2482, ptrtoint (ptr addrspacecast (ptr addrspace(1) @_ZN12_GLOBAL__N_16memoryE to ptr) to i64)
  %cmp.i.i.i2484 = icmp ugt i64 %sub.ptr.sub.i.i.i2483, 65335
  %cond.i.i.i2485 = select i1 %cmp.i.i.i2484, ptr null, ptr %3649
  br label %Flow8107, !amdgpu.uniform !42

Flow8108:                                         ; preds = %Flow8106, %_ZN22__llvm_libc_22_0_0_git15IntegerToStringIyNS_7details3FmtILh10ELb0ELb0ELb0ELm1EEEEC2Ey.exit.i.i.i
  %3650 = phi ptr [ %3648, %Flow8106 ], [ addrspacecast (ptr addrspace(4) @_ZN22__llvm_libc_22_0_0_git3cpp6string14NULL_CHARACTERE to ptr), %_ZN22__llvm_libc_22_0_0_git15IntegerToStringIyNS_7details3FmtILh10ELb0ELb0ELb0ELm1EEEEC2Ey.exit.i.i.i ]
  br label %if.end.i.i.i.i.i.i2489, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git3cpp6string7reserveEm.exit.i.i.i.i.i.i2487: ; preds = %Flow8107
  br label %loadstoreloop4957, !amdgpu.uniform !42

loadstoreloop4957:                                ; preds = %_ZN22__llvm_libc_22_0_0_git3cpp6string7reserveEm.exit.i.i.i.i.i.i2487, %loadstoreloop4957
  %3651 = phi i64 [ %3653, %loadstoreloop4957 ], [ 0, %_ZN22__llvm_libc_22_0_0_git3cpp6string7reserveEm.exit.i.i.i.i.i.i2487 ]
  %3652 = getelementptr i8, ptr %3648, i64 %3651
  store i8 0, ptr %3652, align 1
  %3653 = add i64 %3651, 1
  %3654 = icmp uge i64 %3653, %writer.sroa.8.4.i.i.i.i
  br i1 %3654, label %Flow8106, label %loadstoreloop4957, !amdgpu.uniform !42

Flow8106:                                         ; preds = %loadstoreloop4957
  br label %Flow8108, !amdgpu.uniform !42

if.end.i.i.i.i.i.i2489:                           ; preds = %Flow8108
  %cmp.not.i6.i.i.i.i.i.i2491 = icmp ne ptr %3650, addrspacecast (ptr addrspace(4) @_ZN22__llvm_libc_22_0_0_git3cpp6string14NULL_CHARACTERE to ptr)
  br i1 %cmp.not.i6.i.i.i.i.i.i2491, label %if.then.i.i.i.i2.i.i.i2492, label %_ZN22__llvm_libc_22_0_0_git3cpp9to_stringEy.exit.i, !amdgpu.uniform !42

if.then.i.i.i.i2.i.i.i2492:                       ; preds = %if.end.i.i.i.i.i.i2489
  %arrayidx.i.i.i.i3.i.i.i2493 = getelementptr inbounds nuw i8, ptr %3650, i64 %writer.sroa.8.4.i.i.i.i
  store i8 0, ptr %arrayidx.i.i.i.i3.i.i.i2493, align 1, !tbaa !57, !noalias !162
  br label %_ZN22__llvm_libc_22_0_0_git3cpp9to_stringEy.exit.i, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git3cpp9to_stringEy.exit.i: ; preds = %if.then.i.i.i.i2.i.i.i2492, %if.end.i.i.i.i.i.i2489
  %3655 = trunc i64 %writer.sroa.8.4.i.i.i.i to i32
  %3656 = and i64 %writer.sroa.8.4.i.i.i.i, 15
  %3657 = sub i64 %writer.sroa.8.4.i.i.i.i, %3656
  %3658 = icmp ne i64 %3657, 0
  br i1 %3658, label %loop-memcpy-expansion5133.preheader, label %Flow8105, !amdgpu.uniform !42

loop-memcpy-expansion5133.preheader:              ; preds = %_ZN22__llvm_libc_22_0_0_git3cpp9to_stringEy.exit.i
  %3659 = sub i32 0, %3655
  %scevgep6928 = getelementptr i8, ptr addrspace(5) %scevgep6927, i32 %3659
  br label %loop-memcpy-expansion5133, !amdgpu.uniform !42

Flow8105:                                         ; preds = %Flow8104, %_ZN22__llvm_libc_22_0_0_git3cpp9to_stringEy.exit.i
  br label %loop-memcpy-residual-header5136, !amdgpu.uniform !42

loop-memcpy-expansion5133:                        ; preds = %loop-memcpy-expansion5133.preheader, %loop-memcpy-expansion5133
  %lsr.iv6929 = phi ptr addrspace(5) [ %scevgep6930, %loop-memcpy-expansion5133 ], [ %scevgep6928, %loop-memcpy-expansion5133.preheader ], !amdgpu.uniform !42
  %loop-index5134 = phi i64 [ %3662, %loop-memcpy-expansion5133 ], [ 0, %loop-memcpy-expansion5133.preheader ]
  %3660 = load <4 x i32>, ptr addrspace(5) %lsr.iv6929, align 1
  %3661 = getelementptr i8, ptr %3650, i64 %loop-index5134
  store <4 x i32> %3660, ptr %3661, align 1
  %3662 = add i64 %loop-index5134, 16
  %scevgep6930 = getelementptr i8, ptr addrspace(5) %lsr.iv6929, i32 16
  %3663 = icmp uge i64 %3662, %3657
  br i1 %3663, label %Flow8104, label %loop-memcpy-expansion5133, !amdgpu.uniform !42

Flow8103:                                         ; preds = %Flow8102, %loop-memcpy-residual-header5136
  br label %post-loop-memcpy-expansion5132, !amdgpu.uniform !42

loop-memcpy-residual5135:                         ; preds = %loop-memcpy-residual5135.preheader, %loop-memcpy-residual5135
  %lsr.iv6933 = phi ptr addrspace(5) [ %scevgep6934, %loop-memcpy-residual5135 ], [ %scevgep6932, %loop-memcpy-residual5135.preheader ], !amdgpu.uniform !42
  %residual-loop-index5137 = phi i64 [ %3665, %loop-memcpy-residual5135 ], [ 0, %loop-memcpy-residual5135.preheader ]
  %3664 = load i8, ptr addrspace(5) %lsr.iv6933, align 1
  %scevgep6936 = getelementptr i8, ptr %scevgep6935, i64 %residual-loop-index5137
  store i8 %3664, ptr %scevgep6936, align 1
  %3665 = add i64 %residual-loop-index5137, 1
  %scevgep6934 = getelementptr i8, ptr addrspace(5) %lsr.iv6933, i32 1
  %3666 = icmp uge i64 %3665, %3656
  br i1 %3666, label %Flow8102, label %loop-memcpy-residual5135, !amdgpu.uniform !42

Flow8102:                                         ; preds = %loop-memcpy-residual5135
  br label %Flow8103, !amdgpu.uniform !42

post-loop-memcpy-expansion5132:                   ; preds = %Flow8103
  call void @llvm.lifetime.end.p5(ptr addrspace(5) %buffer.i.i.i2988) #34, !noalias !162
  br label %for.cond.i.i.i2497, !amdgpu.uniform !42

for.cond.i.i.i2497:                               ; preds = %post-loop-memcpy-expansion5132, %Flow8101
  %3667 = phi i32 [ %3704, %Flow8101 ], [ %1198, %post-loop-memcpy-expansion5132 ]
  %3668 = phi i64 [ %3705, %Flow8101 ], [ %1199, %post-loop-memcpy-expansion5132 ]
  %3669 = phi i32 [ %3706, %Flow8101 ], [ %1200, %post-loop-memcpy-expansion5132 ]
  %retval.sroa.6.0.i.i.i2498 = phi i32 [ %3709, %Flow8101 ], [ undef, %post-loop-memcpy-expansion5132 ]
  %retval.sroa.2.0.i.i.i2500 = phi i64 [ %3708, %Flow8101 ], [ undef, %post-loop-memcpy-expansion5132 ]
  %retval.sroa.8.0.i.i.i2501 = phi i32 [ %3707, %Flow8101 ], [ undef, %post-loop-memcpy-expansion5132 ]
  %index.0.i.i.i2502 = phi i32 [ %3711, %Flow8101 ], [ 0, %post-loop-memcpy-expansion5132 ]
  %3670 = load i32, ptr addrspace(1) @__llvm_rpc_client, align 8, !tbaa !68
  %cmp.not.i.i.i2503 = icmp ult i32 %index.0.i.i.i2502, %3670
  %spec.store.select.i.i.i2504 = select i1 %cmp.not.i.i.i2503, i32 %index.0.i.i.i2502, i32 0
  %3671 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  %3672 = and i32 %spec.store.select.i.i.i2504, 31
  %3673 = lshr i64 %3671, %sh_prom.i.i.i.i756
  %3674 = trunc i64 %3673 to i32
  %conv4.i.i.i.i.i2508 = and i32 %3674, 1
  %shl.i.i.i.i.i2509 = shl nuw i32 %conv4.i.i.i.i.i2508, %3672
  br label %ComputeLoop5775, !amdgpu.uniform !42

3675:                                             ; preds = %ComputeEnd5776
  %sunkaddr7443 = getelementptr inbounds i8, ptr addrspace(1) %7994, i64 40
  %3676 = atomicrmw or ptr addrspace(1) %sunkaddr7443, i32 %.lcssa9161 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %3677, !amdgpu.uniform !42

3677:                                             ; preds = %3675, %ComputeEnd5776
  %3678 = phi i32 [ %3676, %3675 ], [ poison, %ComputeEnd5776 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %8003)
  %3679 = call i32 @llvm.amdgcn.readfirstlane.i32(i32 %3678)
  %3680 = or i32 %3679, %.lcssa9162
  %shl5.i.i.i.i.i2510 = shl nuw i32 1, %3672
  %and.i.i.i.i.i2511 = and i32 %shl5.i.i.i.i.i2510, %3680
  %tobool3.i.i.i.i2512 = icmp ne i32 %and.i.i.i.i.i2511, 0
  %3681 = call i64 @llvm.amdgcn.ballot.i64(i1 %tobool3.i.i.i.i2512)
  %and.i.i.i.i.i.i2513 = and i64 %3681, %3671
  %cmp.i.not.i.i.i2514 = icmp ne i64 %3671, %and.i.i.i.i.i.i2513
  br i1 %cmp.i.not.i.i.i2514, label %if.then.i.i.i.i2515, label %Flow8100, !amdgpu.uniform !42

if.then.i.i.i.i2515:                              ; preds = %3677
  fence syncscope("agent") acquire
  %3682 = load <2 x i64>, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !75
  %3683 = extractelement <2 x i64> %3682, i32 0
  %3684 = inttoptr i64 %3683 to ptr
  %3685 = extractelement <2 x i64> %3682, i32 1
  %3686 = inttoptr i64 %3685 to ptr
  %idxprom.i.i.i.i2516 = zext i32 %spec.store.select.i.i.i2504 to i64
  %arrayidx.i.i.i.i2517 = getelementptr inbounds nuw i32, ptr %3684, i64 %idxprom.i.i.i.i2516, !amdgpu.uniform !42
  %3687 = load atomic i32, ptr %arrayidx.i.i.i.i2517 monotonic, align 4
  %3688 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %3687)
  %arrayidx.i45.i.i.i2518 = getelementptr inbounds nuw i32, ptr %3686, i64 %idxprom.i.i.i.i2516, !amdgpu.uniform !42
  %3689 = load atomic i32, ptr %arrayidx.i45.i.i.i2518 monotonic, align 4
  %3690 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %3689)
  %cmp.i46.not.i.i.i2519 = icmp ne i32 %3688, %3690
  br i1 %cmp.i46.not.i.i.i2519, label %if.then12.i.i.i2520, label %Flow8098, !amdgpu.uniform !42

if.then12.i.i.i2520:                              ; preds = %if.then.i.i.i.i2515
  fence syncscope("agent") release
  %3691 = call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %3671, i1 true)
  %iszero.i.i.i.i.i.i.i2521 = icmp ne i64 %3671, 0
  %cmp2.i.i.i.i.i.i2522 = icmp eq i64 %3691, %sh_prom.i.i.i.i756
  %cmp.i.i.i.i.i2.i2523 = select i1 %iszero.i.i.i.i.i.i.i2521, i1 %cmp2.i.i.i.i.i.i2522, i1 false
  %conv4.i.i51.i.i.i2524 = zext i1 %cmp.i.i.i.i.i2.i2523 to i32
  %shl.i.i52.i.i.i2525 = shl nuw i32 %conv4.i.i51.i.i.i2524, %3672
  %xor.i.i.i.i.i2526 = xor i32 %shl.i.i52.i.i.i2525, -1
  %3692 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop5780, !amdgpu.uniform !42

3693:                                             ; preds = %ComputeEnd5781
  %sunkaddr7444 = getelementptr inbounds i8, ptr addrspace(1) %7994, i64 40
  %3694 = atomicrmw and ptr addrspace(1) %sunkaddr7444, i32 %.lcssa9163 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %3695, !amdgpu.uniform !42

3695:                                             ; preds = %3693, %ComputeEnd5781
  call void @llvm.amdgcn.end.cf.i64(i64 %8021)
  call void @llvm.amdgcn.wave.barrier()
  br label %Flow8098, !amdgpu.uniform !42

if.end14.i.i.i2587:                               ; preds = %Flow8098
  %3696 = call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %3671, i1 true)
  %iszero.i.i.i.i.i.i2588 = icmp ne i64 %3671, 0
  %cmp2.i.i.i.i.i2589 = icmp eq i64 %3696, %sh_prom.i.i.i.i756
  %cmp.i.i.i.i.i2590 = select i1 %iszero.i.i.i.i.i.i2588, i1 %cmp2.i.i.i.i.i2589, i1 false
  %3697 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.i.i.i.i.i2590)
  %3698 = extractvalue { i1, i64 } %3697, 0
  %3699 = extractvalue { i1, i64 } %3697, 1
  br i1 %3698, label %if.then16.i.i.i2592, label %if.end22.i.i.i2591

if.then16.i.i.i2592:                              ; preds = %if.end14.i.i.i2587
  %3700 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !76
  %opcode.i.i.i2593 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %3700, i64 %idxprom.i.i.i.i2516, i32 1
  store i32 1660944387, ptr %opcode.i.i.i2593, align 8, !tbaa !77
  %arrayidx21.i.i.i2594 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %3700, i64 %idxprom.i.i.i.i2516
  store i64 %3671, ptr %arrayidx21.i.i.i2594, align 8, !tbaa !80
  br label %if.end22.i.i.i2591, !amdgpu.uniform !42

Flow8099:                                         ; preds = %if.end22.i.i.i2591, %Flow8098
  %3701 = phi i32 [ %3688, %if.end22.i.i.i2591 ], [ %retval.sroa.8.0.i.i.i2501, %Flow8098 ]
  %3702 = phi i64 [ %3671, %if.end22.i.i.i2591 ], [ %retval.sroa.2.0.i.i.i2500, %Flow8098 ]
  %3703 = phi i32 [ %spec.store.select.i.i.i2504, %if.end22.i.i.i2591 ], [ %retval.sroa.6.0.i.i.i2498, %Flow8098 ]
  br label %cleanup26.i.i.i2527, !amdgpu.uniform !42

if.end22.i.i.i2591:                               ; preds = %if.then16.i.i.i2592, %if.end14.i.i.i2587
  call void @llvm.amdgcn.end.cf.i64(i64 %3699)
  call void @llvm.amdgcn.wave.barrier()
  br label %Flow8099, !amdgpu.uniform !42

Flow8100:                                         ; preds = %cleanup26.i.i.i2527, %3677
  %3704 = phi i32 [ %3703, %cleanup26.i.i.i2527 ], [ %3667, %3677 ]
  %3705 = phi i64 [ %3702, %cleanup26.i.i.i2527 ], [ %3668, %3677 ]
  %3706 = phi i32 [ %3701, %cleanup26.i.i.i2527 ], [ %3669, %3677 ]
  %3707 = phi i32 [ %3701, %cleanup26.i.i.i2527 ], [ %retval.sroa.8.0.i.i.i2501, %3677 ]
  %3708 = phi i64 [ %3702, %cleanup26.i.i.i2527 ], [ %retval.sroa.2.0.i.i.i2500, %3677 ]
  %3709 = phi i32 [ %3703, %cleanup26.i.i.i2527 ], [ %retval.sroa.6.0.i.i.i2498, %3677 ]
  %3710 = phi i1 [ %cmp.i46.not.i.i.i2519, %cleanup26.i.i.i2527 ], [ true, %3677 ]
  br i1 %3710, label %for.inc.i.i.i2532, label %Flow8101, !amdgpu.uniform !42

cleanup26.i.i.i2527:                              ; preds = %Flow8099
  br label %Flow8100, !amdgpu.uniform !42

for.inc.i.i.i2532:                                ; preds = %Flow8100
  %inc.i.i.i2537 = add i32 %spec.store.select.i.i.i2504, 1
  br label %Flow8101, !amdgpu.uniform !42

Flow8101:                                         ; preds = %for.inc.i.i.i2532, %Flow8100
  %3711 = phi i32 [ %inc.i.i.i2537, %for.inc.i.i.i2532 ], [ poison, %Flow8100 ]
  %3712 = phi i1 [ false, %for.inc.i.i.i2532 ], [ true, %Flow8100 ]
  br i1 %3712, label %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i2538, label %for.cond.i.i.i2497, !amdgpu.uniform !42

_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i2538: ; preds = %Flow8101
  %.lcssa9166 = phi i32 [ %3704, %Flow8101 ]
  %.lcssa9165 = phi i64 [ %3705, %Flow8101 ]
  %.lcssa9164 = phi i32 [ %3706, %Flow8101 ]
  fence acquire
  %3713 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %mul.i.i.i.i.i.i2539 = shl i32 %.lcssa9166, 6
  %idxprom.i19.i.i.i.i.i2540 = zext i32 %mul.i.i.i.i.i.i2539 to i64
  %arrayidx.i20.i.i.i.i.i2541 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %3713, i64 %idxprom.i19.i.i.i.i.i2540
  %arrayidx.i22.i.i.i.i.i2542 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %arrayidx.i20.i.i.i.i.i2541, i64 %sh_prom.i.i.i.i756
  store i64 %writer.sroa.8.4.i.i.i.i, ptr %arrayidx.i22.i.i.i.i.i2542, align 8, !tbaa !82
  %spec.select.i.i.i.i.i.i.i2543 = call i64 @llvm.umin.i64(i64 %writer.sroa.8.4.i.i.i.i, i64 56)
  %3714 = and i64 %spec.select.i.i.i.i.i.i.i2543, 15
  %3715 = sub i64 %spec.select.i.i.i.i.i.i.i2543, %3714
  %3716 = icmp ne i64 %3715, 0
  br i1 %3716, label %loop-memcpy-expansion5037.preheader, label %Flow8097, !amdgpu.uniform !42

loop-memcpy-expansion5037.preheader:              ; preds = %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i2538
  %scevgep6937 = getelementptr i8, ptr %3713, i64 8
  %scevgep6938 = getelementptr i8, ptr %scevgep6937, i64 %1474
  %3717 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i2540, 6
  %scevgep6939 = getelementptr i8, ptr %scevgep6938, i64 %3717
  br label %loop-memcpy-expansion5037, !amdgpu.uniform !42

Flow8097:                                         ; preds = %Flow8096, %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i2538
  br label %loop-memcpy-residual-header5040, !amdgpu.uniform !42

loop-memcpy-expansion5037:                        ; preds = %loop-memcpy-expansion5037.preheader, %loop-memcpy-expansion5037
  %loop-index5038 = phi i64 [ %3720, %loop-memcpy-expansion5037 ], [ 0, %loop-memcpy-expansion5037.preheader ]
  %3718 = getelementptr i8, ptr %3650, i64 %loop-index5038, !amdgpu.uniform !42
  %3719 = load <4 x i32>, ptr %3718, align 1
  %scevgep6940 = getelementptr i8, ptr %scevgep6939, i64 %loop-index5038
  store <4 x i32> %3719, ptr %scevgep6940, align 1
  %3720 = add i64 %loop-index5038, 16
  %3721 = icmp uge i64 %3720, %3715
  br i1 %3721, label %Flow8096, label %loop-memcpy-expansion5037, !amdgpu.uniform !42

Flow8095:                                         ; preds = %Flow8093, %Flow8094
  %3722 = phi i64 [ %6281, %Flow8093 ], [ %6278, %Flow8094 ]
  br label %post-loop-memcpy-expansion5036, !amdgpu.uniform !42

loop-memcpy-residual5039:                         ; preds = %loop-memcpy-residual5039.preheader, %loop-memcpy-residual5039
  %residual-loop-index5041 = phi i64 [ %3724, %loop-memcpy-residual5039 ], [ 0, %loop-memcpy-residual5039.preheader ]
  %scevgep6946 = getelementptr i8, ptr %scevgep6945, i64 %residual-loop-index5041, !amdgpu.uniform !42
  %3723 = load i8, ptr %scevgep6946, align 1
  %scevgep6944 = getelementptr i8, ptr %scevgep6943, i64 %residual-loop-index5041
  store i8 %3723, ptr %scevgep6944, align 1
  %3724 = add i64 %residual-loop-index5041, 1
  %3725 = icmp uge i64 %3724, %3714
  br i1 %3725, label %Flow8093, label %loop-memcpy-residual5039, !amdgpu.uniform !42

Flow8093:                                         ; preds = %loop-memcpy-residual5039
  br label %Flow8095, !amdgpu.uniform !42

post-loop-memcpy-expansion5036:                   ; preds = %Flow8095
  %tobool.not.i.i.i.i.i.i2545 = icmp eq i32 %.lcssa9164, 0
  %conv.i.i.i.i.i.i2546 = zext i1 %tobool.not.i.i.i.i.i.i2545 to i32
  fence release
  %3726 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %idxprom.i23.i.i.i.i.i2547 = zext i32 %.lcssa9166 to i64
  %arrayidx.i24.i.i.i.i.i2548 = getelementptr inbounds nuw i32, ptr %3726, i64 %idxprom.i23.i.i.i.i.i2547
  store atomic i32 %conv.i.i.i.i.i.i2546, ptr %arrayidx.i24.i.i.i.i.i2548 monotonic, align 4
  %3727 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !84
  %arrayidx.i.i5.i.i2549 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %3727, i64 %idxprom.i23.i.i.i.i.i2547
  %3728 = load i64, ptr %arrayidx.i.i5.i.i2549, align 8, !tbaa !80
  %scevgep6947 = getelementptr i8, ptr %3650, i64 56
  %3729 = add i64 %writer.sroa.8.4.i.i.i.i, -56
  br label %while.cond.i.i.i.i2550, !amdgpu.uniform !42

while.cond.i.i.i.i2550:                           ; preds = %post-loop-memcpy-expansion5036, %Flow8092
  %phi.broken8728 = phi i64 [ 0, %post-loop-memcpy-expansion5036 ], [ %3751, %Flow8092 ]
  %lsr.iv6953 = phi i64 [ %3749, %Flow8092 ], [ %3729, %post-loop-memcpy-expansion5036 ]
  %lsr.iv6948 = phi ptr [ %3748, %Flow8092 ], [ %scevgep6947, %post-loop-memcpy-expansion5036 ]
  %port.sroa.37.0.i.i2551 = phi i32 [ %3747, %Flow8092 ], [ %conv.i.i.i.i.i.i2546, %post-loop-memcpy-expansion5036 ]
  %idx.0.i.i.i.i2552 = phi i64 [ %3746, %Flow8092 ], [ 56, %post-loop-memcpy-expansion5036 ]
  %cmp.i.i.i.i2553 = icmp ugt i64 %writer.sroa.8.4.i.i.i.i, %idx.0.i.i.i.i2552
  %3730 = call i64 @llvm.amdgcn.ballot.i64(i1 %cmp.i.i.i.i2553)
  %and.i.i.i.i6.i.i2554 = and i64 %3730, %3728
  %tobool.not.i.i.i.i2555 = icmp ne i64 %and.i.i.i.i6.i.i2554, 0
  %3731 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i22.i.i.i2556 = getelementptr inbounds nuw i32, ptr %3731, i64 %idxprom.i23.i.i.i.i.i2547
  %3732 = load atomic i32, ptr %arrayidx.i22.i.i.i2556 monotonic, align 4
  %3733 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %tobool.not.i.i.i.i2555)
  %3734 = extractvalue { i1, i64 } %3733, 0
  %3735 = extractvalue { i1, i64 } %3733, 1
  br i1 %3734, label %cond.false.i8.i.i.i.i2557, label %Flow8092

cond.false.i8.i.i.i.i2557:                        ; preds = %while.cond.i.i.i.i2550
  %3736 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %3732)
  %3737 = icmp ne i32 %3736, %port.sroa.37.0.i.i2551
  %3738 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3737)
  %3739 = extractvalue { i1, i64 } %3738, 0
  %3740 = extractvalue { i1, i64 } %3738, 1
  br i1 %3739, label %while.body.i.i21.i.i.i.i2558.preheader, label %Flow8091

while.body.i.i21.i.i.i.i2558.preheader:           ; preds = %cond.false.i8.i.i.i.i2557
  br label %while.body.i.i21.i.i.i.i2558, !amdgpu.uniform !42

Flow8091:                                         ; preds = %Flow8090, %cond.false.i8.i.i.i.i2557
  call void @llvm.amdgcn.end.cf.i64(i64 %3740)
  br label %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i2561, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i2558:                     ; preds = %while.body.i.i21.i.i.i.i2558.preheader, %while.body.i.i21.i.i.i.i2558
  %phi.broken8727 = phi i64 [ 0, %while.body.i.i21.i.i.i.i2558.preheader ], [ %3744, %while.body.i.i21.i.i.i.i2558 ]
  call void @llvm.amdgcn.s.sleep(i32 2)
  %3741 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i22.i.i.i.i2559 = getelementptr inbounds nuw i32, ptr %3741, i64 %idxprom.i23.i.i.i.i.i2547
  %3742 = load atomic i32, ptr %arrayidx.i.i.i22.i.i.i.i2559 monotonic, align 4
  %3743 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %3742)
  %cmp.i.not.i.i23.i.i.i.i2560 = icmp eq i32 %3743, %port.sroa.37.0.i.i2551
  %3744 = call i64 @llvm.amdgcn.if.break.i64(i1 %cmp.i.not.i.i23.i.i.i.i2560, i64 %phi.broken8727)
  %3745 = call i1 @llvm.amdgcn.loop.i64(i64 %3744)
  br i1 %3745, label %Flow8090, label %while.body.i.i21.i.i.i.i2558

Flow8090:                                         ; preds = %while.body.i.i21.i.i.i.i2558
  %.lcssa9167 = phi i64 [ %3744, %while.body.i.i21.i.i.i.i2558 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa9167)
  br label %Flow8091, !amdgpu.uniform !42

Flow8092:                                         ; preds = %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i2562, %while.cond.i.i.i.i2550
  %3746 = phi i64 [ %add.i.i.i.i2565, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i2562 ], [ poison, %while.cond.i.i.i.i2550 ]
  %3747 = phi i32 [ %conv.i.i26.i.i.i.i2563, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i2562 ], [ poison, %while.cond.i.i.i.i2550 ]
  %3748 = phi ptr [ %scevgep6949, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i2562 ], [ poison, %while.cond.i.i.i.i2550 ]
  %3749 = phi i64 [ %lsr.iv.next6954, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i2562 ], [ poison, %while.cond.i.i.i.i2550 ]
  %3750 = phi i1 [ false, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i2562 ], [ true, %while.cond.i.i.i.i2550 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %3735)
  %3751 = call i64 @llvm.amdgcn.if.break.i64(i1 %3750, i64 %phi.broken8728)
  %3752 = call i1 @llvm.amdgcn.loop.i64(i64 %3751)
  br i1 %3752, label %cond.false.i.i.i2572, label %while.cond.i.i.i.i2550

_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i2561: ; preds = %Flow8091
  fence acquire
  br i1 %cmp.i.i.i.i2553, label %if.then.i.i.i.i.i.i3.i2566, label %Flow8089, !amdgpu.uniform !42

if.then.i.i.i.i.i.i3.i2566:                       ; preds = %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i2561
  %sub.i.i.i.i.i.i.i2568 = sub nuw i64 %writer.sroa.8.4.i.i.i.i, %idx.0.i.i.i.i2552
  %spec.select.i.i.i36.i.i.i.i2569 = call i64 @llvm.umin.i64(i64 %sub.i.i.i.i.i.i.i2568, i64 64)
  %3753 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %3754 = and i64 %spec.select.i.i.i36.i.i.i.i2569, 15
  %3755 = sub i64 %spec.select.i.i.i36.i.i.i.i2569, %3754
  %3756 = icmp ne i64 %3755, 0
  br i1 %3756, label %loop-memcpy-expansion5031.preheader, label %Flow8088, !amdgpu.uniform !42

loop-memcpy-expansion5031.preheader:              ; preds = %if.then.i.i.i.i.i.i3.i2566
  %arrayidx.i20.i35.i.i.i.i2570 = getelementptr %"struct.rpc::Buffer", ptr %3753, i64 %idxprom.i19.i.i.i.i.i2540
  %arrayidx.i22.i38.i.i.i.i2571 = getelementptr %"struct.rpc::Buffer", ptr %arrayidx.i20.i35.i.i.i.i2570, i64 %sh_prom.i.i.i.i756
  br label %loop-memcpy-expansion5031, !amdgpu.uniform !42

Flow8088:                                         ; preds = %Flow8087, %if.then.i.i.i.i.i.i3.i2566
  br label %loop-memcpy-residual-header5034, !amdgpu.uniform !42

loop-memcpy-expansion5031:                        ; preds = %loop-memcpy-expansion5031.preheader, %loop-memcpy-expansion5031
  %loop-index5032 = phi i64 [ %3759, %loop-memcpy-expansion5031 ], [ 0, %loop-memcpy-expansion5031.preheader ]
  %scevgep6950 = getelementptr i8, ptr %lsr.iv6948, i64 %loop-index5032, !amdgpu.uniform !42
  %3757 = load <4 x i32>, ptr %scevgep6950, align 1
  %3758 = getelementptr i8, ptr %arrayidx.i22.i38.i.i.i.i2571, i64 %loop-index5032
  store <4 x i32> %3757, ptr %3758, align 1
  %3759 = add i64 %loop-index5032, 16
  %3760 = icmp uge i64 %3759, %3755
  br i1 %3760, label %Flow8087, label %loop-memcpy-expansion5031, !amdgpu.uniform !42

Flow8086:                                         ; preds = %Flow8085, %loop-memcpy-residual-header5034
  br label %Flow8089, !amdgpu.uniform !42

loop-memcpy-residual5033:                         ; preds = %loop-memcpy-residual5033.preheader, %loop-memcpy-residual5033
  %residual-loop-index5035 = phi i64 [ %3762, %loop-memcpy-residual5033 ], [ 0, %loop-memcpy-residual5033.preheader ]
  %scevgep6959 = getelementptr i8, ptr %scevgep6958, i64 %residual-loop-index5035, !amdgpu.uniform !42
  %3761 = load i8, ptr %scevgep6959, align 1
  %scevgep6957 = getelementptr i8, ptr %scevgep6956, i64 %residual-loop-index5035
  store i8 %3761, ptr %scevgep6957, align 1
  %3762 = add i64 %residual-loop-index5035, 1
  %3763 = icmp uge i64 %3762, %3754
  br i1 %3763, label %Flow8085, label %loop-memcpy-residual5033, !amdgpu.uniform !42

Flow8085:                                         ; preds = %loop-memcpy-residual5033
  br label %Flow8086, !amdgpu.uniform !42

_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i2562: ; preds = %Flow8089
  %conv.i.i26.i.i.i.i2563 = xor i32 %port.sroa.37.0.i.i2551, 1
  fence release
  %3764 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %arrayidx.i24.i29.i.i.i.i2564 = getelementptr inbounds nuw i32, ptr %3764, i64 %idxprom.i23.i.i.i.i.i2547
  store atomic i32 %conv.i.i26.i.i.i.i2563, ptr %arrayidx.i24.i29.i.i.i.i2564 monotonic, align 4
  %add.i.i.i.i2565 = add i64 %idx.0.i.i.i.i2552, 64
  %scevgep6949 = getelementptr i8, ptr %lsr.iv6948, i64 64
  %lsr.iv.next6954 = add i64 %lsr.iv6953, -64
  br label %Flow8092, !amdgpu.uniform !42

cond.false.i.i.i2572:                             ; preds = %Flow8092
  %.lcssa9169 = phi i64 [ %3751, %Flow8092 ]
  %port.sroa.37.0.i.i2551.lcssa = phi i32 [ %port.sroa.37.0.i.i2551, %Flow8092 ]
  %.lcssa9168 = phi i32 [ %3732, %Flow8092 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa9169)
  %3765 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %.lcssa9168)
  %3766 = icmp ne i32 %3765, %port.sroa.37.0.i.i2551.lcssa
  %3767 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3766)
  %3768 = extractvalue { i1, i64 } %3767, 0
  %3769 = extractvalue { i1, i64 } %3767, 1
  br i1 %3768, label %while.body.i.i.i.i2573.preheader, label %Flow8084

while.body.i.i.i.i2573.preheader:                 ; preds = %cond.false.i.i.i2572
  br label %while.body.i.i.i.i2573, !amdgpu.uniform !42

Flow8084:                                         ; preds = %Flow8083, %cond.false.i.i.i2572
  call void @llvm.amdgcn.end.cf.i64(i64 %3769)
  br label %_ZN22__llvm_libc_22_0_0_git15write_to_stderrENS_3cpp11string_viewE.exit.i2576, !amdgpu.uniform !42

while.body.i.i.i.i2573:                           ; preds = %while.body.i.i.i.i2573.preheader, %while.body.i.i.i.i2573
  %phi.broken8729 = phi i64 [ 0, %while.body.i.i.i.i2573.preheader ], [ %3773, %while.body.i.i.i.i2573 ]
  call void @llvm.amdgcn.s.sleep(i32 2)
  %3770 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i8.i.i2574 = getelementptr inbounds nuw i32, ptr %3770, i64 %idxprom.i23.i.i.i.i.i2547
  %3771 = load atomic i32, ptr %arrayidx.i.i.i8.i.i2574 monotonic, align 4
  %3772 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %3771)
  %cmp.i.not.i.i.i.i2575 = icmp eq i32 %3772, %port.sroa.37.0.i.i2551.lcssa
  %3773 = call i64 @llvm.amdgcn.if.break.i64(i1 %cmp.i.not.i.i.i.i2575, i64 %phi.broken8729)
  %3774 = call i1 @llvm.amdgcn.loop.i64(i64 %3773)
  br i1 %3774, label %Flow8083, label %while.body.i.i.i.i2573

Flow8083:                                         ; preds = %while.body.i.i.i.i2573
  %.lcssa9170 = phi i64 [ %3773, %while.body.i.i.i.i2573 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa9170)
  br label %Flow8084, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git15write_to_stderrENS_3cpp11string_viewE.exit.i2576: ; preds = %Flow8084
  fence acquire
  call void @llvm.amdgcn.wave.barrier()
  fence syncscope("agent") release
  %3775 = call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %.lcssa9165, i1 true)
  %iszero.i.i.i.i.i13.i.i2577 = icmp ne i64 %.lcssa9165, 0
  %cmp2.i.i.i.i14.i.i2578 = icmp eq i64 %3775, %sh_prom.i.i.i.i756
  %cmp.i.i.i.i15.i.i2579 = select i1 %iszero.i.i.i.i.i13.i.i2577, i1 %cmp2.i.i.i.i14.i.i2578, i1 false
  %3776 = and i32 %.lcssa9166, 31
  %conv4.i.i.i19.i.i2583 = zext i1 %cmp.i.i.i.i15.i.i2579 to i32
  %shl.i.i.i20.i.i2584 = shl nuw i32 %conv4.i.i.i19.i.i2583, %3776
  %xor.i.i.i21.i.i2585 = xor i32 %shl.i.i.i20.i.i2584, -1
  %3777 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop5784, !amdgpu.uniform !42

3778:                                             ; preds = %ComputeEnd5785
  %div8.i.i.i16.i.i2580 = lshr i32 %.lcssa9166, 5
  %idxprom.i.i.i17.i.i2581 = zext nneg i32 %div8.i.i.i16.i.i2580 to i64
  %3779 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i17.i.i2581
  %sunkaddr7445 = getelementptr inbounds i8, ptr addrspace(1) %3779, i64 40
  %3780 = atomicrmw and ptr addrspace(1) %sunkaddr7445, i32 %.lcssa9171 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %3781, !amdgpu.uniform !42

3781:                                             ; preds = %3778, %ComputeEnd5785
  call void @llvm.amdgcn.end.cf.i64(i64 %8038)
  call void @llvm.amdgcn.wave.barrier()
  br label %for.cond.i.i1.i2615, !amdgpu.uniform !42

for.cond.i.i1.i2615:                              ; preds = %3781, %Flow8082
  %3782 = phi i32 [ %3819, %Flow8082 ], [ %1201, %3781 ]
  %3783 = phi i64 [ %3820, %Flow8082 ], [ %1202, %3781 ]
  %3784 = phi i32 [ %3821, %Flow8082 ], [ %1203, %3781 ]
  %retval.sroa.6.0.i.i.i2616 = phi i32 [ %3824, %Flow8082 ], [ undef, %3781 ]
  %retval.sroa.2.0.i.i.i2618 = phi i64 [ %3823, %Flow8082 ], [ undef, %3781 ]
  %retval.sroa.8.0.i.i.i2619 = phi i32 [ %3822, %Flow8082 ], [ undef, %3781 ]
  %index.0.i.i.i2620 = phi i32 [ %3826, %Flow8082 ], [ 0, %3781 ]
  %3785 = load i32, ptr addrspace(1) @__llvm_rpc_client, align 8, !tbaa !68
  %cmp.not.i.i.i2621 = icmp ult i32 %index.0.i.i.i2620, %3785
  %spec.store.select.i.i.i2622 = select i1 %cmp.not.i.i.i2621, i32 %index.0.i.i.i2620, i32 0
  %3786 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  %3787 = and i32 %spec.store.select.i.i.i2622, 31
  %3788 = lshr i64 %3786, %sh_prom.i.i.i.i756
  %3789 = trunc i64 %3788 to i32
  %conv4.i.i.i.i.i2626 = and i32 %3789, 1
  %shl.i.i.i.i.i2627 = shl nuw i32 %conv4.i.i.i.i.i2626, %3787
  br label %ComputeLoop5788, !amdgpu.uniform !42

3790:                                             ; preds = %ComputeEnd5789
  %sunkaddr7446 = getelementptr inbounds i8, ptr addrspace(1) %8048, i64 40
  %3791 = atomicrmw or ptr addrspace(1) %sunkaddr7446, i32 %.lcssa9172 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %3792, !amdgpu.uniform !42

3792:                                             ; preds = %3790, %ComputeEnd5789
  %3793 = phi i32 [ %3791, %3790 ], [ poison, %ComputeEnd5789 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %8057)
  %3794 = call i32 @llvm.amdgcn.readfirstlane.i32(i32 %3793)
  %3795 = or i32 %3794, %.lcssa9173
  %shl5.i.i.i.i.i2628 = shl nuw i32 1, %3787
  %and.i.i.i.i.i2629 = and i32 %shl5.i.i.i.i.i2628, %3795
  %tobool3.i.i.i.i2630 = icmp ne i32 %and.i.i.i.i.i2629, 0
  %3796 = call i64 @llvm.amdgcn.ballot.i64(i1 %tobool3.i.i.i.i2630)
  %and.i.i.i.i.i.i2631 = and i64 %3796, %3786
  %cmp.i.not.i.i.i2632 = icmp ne i64 %3786, %and.i.i.i.i.i.i2631
  br i1 %cmp.i.not.i.i.i2632, label %if.then.i.i.i.i2633, label %Flow8081, !amdgpu.uniform !42

if.then.i.i.i.i2633:                              ; preds = %3792
  fence syncscope("agent") acquire
  %3797 = load <2 x i64>, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !75
  %3798 = extractelement <2 x i64> %3797, i32 0
  %3799 = inttoptr i64 %3798 to ptr
  %3800 = extractelement <2 x i64> %3797, i32 1
  %3801 = inttoptr i64 %3800 to ptr
  %idxprom.i.i.i.i2634 = zext i32 %spec.store.select.i.i.i2622 to i64
  %arrayidx.i.i.i.i2635 = getelementptr inbounds nuw i32, ptr %3799, i64 %idxprom.i.i.i.i2634, !amdgpu.uniform !42
  %3802 = load atomic i32, ptr %arrayidx.i.i.i.i2635 monotonic, align 4
  %3803 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %3802)
  %arrayidx.i45.i.i.i2636 = getelementptr inbounds nuw i32, ptr %3801, i64 %idxprom.i.i.i.i2634, !amdgpu.uniform !42
  %3804 = load atomic i32, ptr %arrayidx.i45.i.i.i2636 monotonic, align 4
  %3805 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %3804)
  %cmp.i46.not.i.i.i2637 = icmp ne i32 %3803, %3805
  br i1 %cmp.i46.not.i.i.i2637, label %if.then12.i.i.i2638, label %Flow8079, !amdgpu.uniform !42

if.then12.i.i.i2638:                              ; preds = %if.then.i.i.i.i2633
  fence syncscope("agent") release
  %3806 = call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %3786, i1 true)
  %iszero.i.i.i.i.i.i.i2639 = icmp ne i64 %3786, 0
  %cmp2.i.i.i.i.i.i2640 = icmp eq i64 %3806, %sh_prom.i.i.i.i756
  %cmp.i.i.i.i.i.i2641 = select i1 %iszero.i.i.i.i.i.i.i2639, i1 %cmp2.i.i.i.i.i.i2640, i1 false
  %conv4.i.i51.i.i.i2642 = zext i1 %cmp.i.i.i.i.i.i2641 to i32
  %shl.i.i52.i.i.i2643 = shl nuw i32 %conv4.i.i51.i.i.i2642, %3787
  %xor.i.i.i.i.i2644 = xor i32 %shl.i.i52.i.i.i2643, -1
  %3807 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop5793, !amdgpu.uniform !42

3808:                                             ; preds = %ComputeEnd5794
  %sunkaddr7447 = getelementptr inbounds i8, ptr addrspace(1) %8048, i64 40
  %3809 = atomicrmw and ptr addrspace(1) %sunkaddr7447, i32 %.lcssa9174 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %3810, !amdgpu.uniform !42

3810:                                             ; preds = %3808, %ComputeEnd5794
  call void @llvm.amdgcn.end.cf.i64(i64 %8075)
  call void @llvm.amdgcn.wave.barrier()
  br label %Flow8079, !amdgpu.uniform !42

if.end14.i.i.i2704:                               ; preds = %Flow8079
  %3811 = call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %3786, i1 true)
  %iszero.i.i.i.i.i.i2705 = icmp ne i64 %3786, 0
  %cmp2.i.i.i.i.i2706 = icmp eq i64 %3811, %sh_prom.i.i.i.i756
  %cmp.i.i.i.i.i2707 = select i1 %iszero.i.i.i.i.i.i2705, i1 %cmp2.i.i.i.i.i2706, i1 false
  %3812 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.i.i.i.i.i2707)
  %3813 = extractvalue { i1, i64 } %3812, 0
  %3814 = extractvalue { i1, i64 } %3812, 1
  br i1 %3813, label %if.then16.i.i.i2709, label %if.end22.i.i.i2708

if.then16.i.i.i2709:                              ; preds = %if.end14.i.i.i2704
  %3815 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !76
  %opcode.i.i.i2710 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %3815, i64 %idxprom.i.i.i.i2634, i32 1
  store i32 1660944387, ptr %opcode.i.i.i2710, align 8, !tbaa !77
  %arrayidx21.i.i.i2711 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %3815, i64 %idxprom.i.i.i.i2634
  store i64 %3786, ptr %arrayidx21.i.i.i2711, align 8, !tbaa !80
  br label %if.end22.i.i.i2708, !amdgpu.uniform !42

Flow8080:                                         ; preds = %if.end22.i.i.i2708, %Flow8079
  %3816 = phi i32 [ %3803, %if.end22.i.i.i2708 ], [ %retval.sroa.8.0.i.i.i2619, %Flow8079 ]
  %3817 = phi i64 [ %3786, %if.end22.i.i.i2708 ], [ %retval.sroa.2.0.i.i.i2618, %Flow8079 ]
  %3818 = phi i32 [ %spec.store.select.i.i.i2622, %if.end22.i.i.i2708 ], [ %retval.sroa.6.0.i.i.i2616, %Flow8079 ]
  br label %cleanup26.i.i.i2645, !amdgpu.uniform !42

if.end22.i.i.i2708:                               ; preds = %if.then16.i.i.i2709, %if.end14.i.i.i2704
  call void @llvm.amdgcn.end.cf.i64(i64 %3814)
  call void @llvm.amdgcn.wave.barrier()
  br label %Flow8080, !amdgpu.uniform !42

Flow8081:                                         ; preds = %cleanup26.i.i.i2645, %3792
  %3819 = phi i32 [ %3818, %cleanup26.i.i.i2645 ], [ %3782, %3792 ]
  %3820 = phi i64 [ %3817, %cleanup26.i.i.i2645 ], [ %3783, %3792 ]
  %3821 = phi i32 [ %3816, %cleanup26.i.i.i2645 ], [ %3784, %3792 ]
  %3822 = phi i32 [ %3816, %cleanup26.i.i.i2645 ], [ %retval.sroa.8.0.i.i.i2619, %3792 ]
  %3823 = phi i64 [ %3817, %cleanup26.i.i.i2645 ], [ %retval.sroa.2.0.i.i.i2618, %3792 ]
  %3824 = phi i32 [ %3818, %cleanup26.i.i.i2645 ], [ %retval.sroa.6.0.i.i.i2616, %3792 ]
  %3825 = phi i1 [ %cmp.i46.not.i.i.i2637, %cleanup26.i.i.i2645 ], [ true, %3792 ]
  br i1 %3825, label %for.inc.i.i.i2650, label %Flow8082, !amdgpu.uniform !42

cleanup26.i.i.i2645:                              ; preds = %Flow8080
  br label %Flow8081, !amdgpu.uniform !42

for.inc.i.i.i2650:                                ; preds = %Flow8081
  %inc.i.i.i2655 = add i32 %spec.store.select.i.i.i2622, 1
  br label %Flow8082, !amdgpu.uniform !42

Flow8082:                                         ; preds = %for.inc.i.i.i2650, %Flow8081
  %3826 = phi i32 [ %inc.i.i.i2655, %for.inc.i.i.i2650 ], [ poison, %Flow8081 ]
  %3827 = phi i1 [ false, %for.inc.i.i.i2650 ], [ true, %Flow8081 ]
  br i1 %3827, label %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i2656, label %for.cond.i.i1.i2615, !amdgpu.uniform !42

_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i2656: ; preds = %Flow8082
  %.lcssa9177 = phi i32 [ %3819, %Flow8082 ]
  %.lcssa9176 = phi i64 [ %3820, %Flow8082 ]
  %.lcssa9175 = phi i32 [ %3821, %Flow8082 ]
  fence acquire
  %3828 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %mul.i.i.i.i.i.i2657 = shl i32 %.lcssa9177, 6
  %idxprom.i19.i.i.i.i.i2658 = zext i32 %mul.i.i.i.i.i.i2657 to i64
  %arrayidx.i20.i.i.i.i.i2659 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %3828, i64 %idxprom.i19.i.i.i.i.i2658
  %arrayidx.i22.i.i.i.i.i2660 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %arrayidx.i20.i.i.i.i.i2659, i64 %sh_prom.i.i.i.i756
  store i64 sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.22.39 to ptr), i64 5) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.22.39 to ptr) to i64)), ptr %arrayidx.i22.i.i.i.i.i2660, align 8, !tbaa !82
  br i1 %1130, label %loop-memcpy-expansion5361.preheader, label %Flow8078, !amdgpu.uniform !42

loop-memcpy-expansion5361.preheader:              ; preds = %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i2656
  %scevgep6960 = getelementptr i8, ptr %3828, i64 8
  %scevgep6961 = getelementptr i8, ptr %scevgep6960, i64 %1474
  %3829 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i2658, 6
  %scevgep6962 = getelementptr i8, ptr %scevgep6961, i64 %3829
  br label %loop-memcpy-expansion5361, !amdgpu.uniform !42

Flow8078:                                         ; preds = %Flow8077, %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i2656
  br label %loop-memcpy-residual-header5364, !amdgpu.uniform !42

loop-memcpy-expansion5361:                        ; preds = %loop-memcpy-expansion5361.preheader, %loop-memcpy-expansion5361
  %loop-index5362 = phi i64 [ %3832, %loop-memcpy-expansion5361 ], [ 0, %loop-memcpy-expansion5361.preheader ]
  %3830 = getelementptr i8, ptr addrspace(4) @.str.22.39, i64 %loop-index5362, !amdgpu.uniform !42
  %3831 = load <4 x i32>, ptr addrspace(4) %3830, align 1
  %scevgep6963 = getelementptr i8, ptr %scevgep6962, i64 %loop-index5362
  store <4 x i32> %3831, ptr %scevgep6963, align 1
  %3832 = add i64 %loop-index5362, 16
  %3833 = icmp uge i64 %3832, %1129
  br i1 %3833, label %Flow8077, label %loop-memcpy-expansion5361, !amdgpu.uniform !42

Flow8076:                                         ; preds = %Flow8074, %Flow8075
  %3834 = phi i64 [ %6623, %Flow8074 ], [ %6621, %Flow8075 ]
  br label %post-loop-memcpy-expansion5360, !amdgpu.uniform !42

loop-memcpy-residual5363:                         ; preds = %loop-memcpy-residual5363.preheader, %loop-memcpy-residual5363
  %residual-loop-index5365 = phi i64 [ %3836, %loop-memcpy-residual5363 ], [ 0, %loop-memcpy-residual5363.preheader ]
  %scevgep6968 = getelementptr i8, ptr addrspace(4) @.str.22.39, i64 %residual-loop-index5365, !amdgpu.uniform !42
  %3835 = load i8, ptr addrspace(4) %scevgep6968, align 1
  %scevgep6967 = getelementptr i8, ptr %scevgep6966, i64 %residual-loop-index5365
  store i8 %3835, ptr %scevgep6967, align 1
  %3836 = add i64 %residual-loop-index5365, 1
  %3837 = icmp uge i64 %3836, %1128
  br i1 %3837, label %Flow8074, label %loop-memcpy-residual5363, !amdgpu.uniform !42

Flow8074:                                         ; preds = %loop-memcpy-residual5363
  br label %Flow8076, !amdgpu.uniform !42

post-loop-memcpy-expansion5360:                   ; preds = %Flow8076
  %tobool.not.i.i.i.i.i.i2663 = icmp eq i32 %.lcssa9175, 0
  %conv.i.i.i.i.i.i2664 = zext i1 %tobool.not.i.i.i.i.i.i2663 to i32
  fence release
  %3838 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %idxprom.i23.i.i.i.i.i2665 = zext i32 %.lcssa9177 to i64
  %arrayidx.i24.i.i.i.i.i2666 = getelementptr inbounds nuw i32, ptr %3838, i64 %idxprom.i23.i.i.i.i.i2665
  store atomic i32 %conv.i.i.i.i.i.i2664, ptr %arrayidx.i24.i.i.i.i.i2666 monotonic, align 4
  %3839 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !84
  %arrayidx.i.i5.i.i2667 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %3839, i64 %idxprom.i23.i.i.i.i.i2665
  %3840 = load i64, ptr %arrayidx.i.i5.i.i2667, align 8, !tbaa !80
  br label %while.cond.i.i.i.i2668, !amdgpu.uniform !42

while.cond.i.i.i.i2668:                           ; preds = %post-loop-memcpy-expansion5360, %Flow8073
  %phi.broken8731 = phi i64 [ 0, %post-loop-memcpy-expansion5360 ], [ %3862, %Flow8073 ]
  %lsr.iv6974 = phi i64 [ %3860, %Flow8073 ], [ -51, %post-loop-memcpy-expansion5360 ]
  %lsr.iv6969 = phi ptr addrspace(4) [ %3859, %Flow8073 ], [ getelementptr (i8, ptr addrspace(4) @.str.22.39, i64 56), %post-loop-memcpy-expansion5360 ]
  %port.sroa.37.0.i.i2669 = phi i32 [ %3858, %Flow8073 ], [ %conv.i.i.i.i.i.i2664, %post-loop-memcpy-expansion5360 ]
  %idx.0.i.i.i.i2670 = phi i64 [ %3857, %Flow8073 ], [ 56, %post-loop-memcpy-expansion5360 ]
  %cmp.i.i.i.i2671 = icmp ult i64 %idx.0.i.i.i.i2670, sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.22.39 to ptr), i64 5) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.22.39 to ptr) to i64))
  %3841 = call i64 @llvm.amdgcn.ballot.i64(i1 %cmp.i.i.i.i2671)
  %and.i.i.i.i6.i.i2672 = and i64 %3841, %3840
  %tobool.not.i.i.i.i2673 = icmp ne i64 %and.i.i.i.i6.i.i2672, 0
  %3842 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i22.i.i.i2674 = getelementptr inbounds nuw i32, ptr %3842, i64 %idxprom.i23.i.i.i.i.i2665
  %3843 = load atomic i32, ptr %arrayidx.i22.i.i.i2674 monotonic, align 4
  %3844 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %tobool.not.i.i.i.i2673)
  %3845 = extractvalue { i1, i64 } %3844, 0
  %3846 = extractvalue { i1, i64 } %3844, 1
  br i1 %3845, label %cond.false.i8.i.i.i.i2675, label %Flow8073

cond.false.i8.i.i.i.i2675:                        ; preds = %while.cond.i.i.i.i2668
  %3847 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %3843)
  %3848 = icmp ne i32 %3847, %port.sroa.37.0.i.i2669
  %3849 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3848)
  %3850 = extractvalue { i1, i64 } %3849, 0
  %3851 = extractvalue { i1, i64 } %3849, 1
  br i1 %3850, label %while.body.i.i21.i.i.i.i2676.preheader, label %Flow8072

while.body.i.i21.i.i.i.i2676.preheader:           ; preds = %cond.false.i8.i.i.i.i2675
  br label %while.body.i.i21.i.i.i.i2676, !amdgpu.uniform !42

Flow8072:                                         ; preds = %Flow8071, %cond.false.i8.i.i.i.i2675
  call void @llvm.amdgcn.end.cf.i64(i64 %3851)
  br label %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i2679, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i2676:                     ; preds = %while.body.i.i21.i.i.i.i2676.preheader, %while.body.i.i21.i.i.i.i2676
  %phi.broken8730 = phi i64 [ 0, %while.body.i.i21.i.i.i.i2676.preheader ], [ %3855, %while.body.i.i21.i.i.i.i2676 ]
  call void @llvm.amdgcn.s.sleep(i32 2)
  %3852 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i22.i.i.i.i2677 = getelementptr inbounds nuw i32, ptr %3852, i64 %idxprom.i23.i.i.i.i.i2665
  %3853 = load atomic i32, ptr %arrayidx.i.i.i22.i.i.i.i2677 monotonic, align 4
  %3854 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %3853)
  %cmp.i.not.i.i23.i.i.i.i2678 = icmp eq i32 %3854, %port.sroa.37.0.i.i2669
  %3855 = call i64 @llvm.amdgcn.if.break.i64(i1 %cmp.i.not.i.i23.i.i.i.i2678, i64 %phi.broken8730)
  %3856 = call i1 @llvm.amdgcn.loop.i64(i64 %3855)
  br i1 %3856, label %Flow8071, label %while.body.i.i21.i.i.i.i2676

Flow8071:                                         ; preds = %while.body.i.i21.i.i.i.i2676
  %.lcssa9178 = phi i64 [ %3855, %while.body.i.i21.i.i.i.i2676 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa9178)
  br label %Flow8072, !amdgpu.uniform !42

Flow8073:                                         ; preds = %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i2680, %while.cond.i.i.i.i2668
  %3857 = phi i64 [ %add.i.i.i.i2683, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i2680 ], [ poison, %while.cond.i.i.i.i2668 ]
  %3858 = phi i32 [ %conv.i.i26.i.i.i.i2681, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i2680 ], [ poison, %while.cond.i.i.i.i2668 ]
  %3859 = phi ptr addrspace(4) [ %scevgep6970, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i2680 ], [ poison, %while.cond.i.i.i.i2668 ]
  %3860 = phi i64 [ %lsr.iv.next6975, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i2680 ], [ poison, %while.cond.i.i.i.i2668 ]
  %3861 = phi i1 [ false, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i2680 ], [ true, %while.cond.i.i.i.i2668 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %3846)
  %3862 = call i64 @llvm.amdgcn.if.break.i64(i1 %3861, i64 %phi.broken8731)
  %3863 = call i1 @llvm.amdgcn.loop.i64(i64 %3862)
  br i1 %3863, label %cond.false.i.i.i2690, label %while.cond.i.i.i.i2668

_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i2679: ; preds = %Flow8072
  fence acquire
  br i1 %cmp.i.i.i.i2671, label %if.then.i.i.i.i.i.i.i2684, label %Flow8070, !amdgpu.uniform !42

if.then.i.i.i.i.i.i.i2684:                        ; preds = %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i2679
  %sub.i.i.i.i.i.i.i2686 = sub nuw i64 sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.22.39 to ptr), i64 5) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.22.39 to ptr) to i64)), %idx.0.i.i.i.i2670
  %spec.select.i.i.i36.i.i.i.i2687 = call i64 @llvm.umin.i64(i64 %sub.i.i.i.i.i.i.i2686, i64 64)
  %3864 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %3865 = and i64 %spec.select.i.i.i36.i.i.i.i2687, 15
  %3866 = sub i64 %spec.select.i.i.i36.i.i.i.i2687, %3865
  %3867 = icmp ne i64 %3866, 0
  br i1 %3867, label %loop-memcpy-expansion5367.preheader, label %Flow8069, !amdgpu.uniform !42

loop-memcpy-expansion5367.preheader:              ; preds = %if.then.i.i.i.i.i.i.i2684
  %arrayidx.i20.i35.i.i.i.i2688 = getelementptr %"struct.rpc::Buffer", ptr %3864, i64 %idxprom.i19.i.i.i.i.i2658
  %arrayidx.i22.i38.i.i.i.i2689 = getelementptr %"struct.rpc::Buffer", ptr %arrayidx.i20.i35.i.i.i.i2688, i64 %sh_prom.i.i.i.i756
  br label %loop-memcpy-expansion5367, !amdgpu.uniform !42

Flow8069:                                         ; preds = %Flow8068, %if.then.i.i.i.i.i.i.i2684
  br label %loop-memcpy-residual-header5370, !amdgpu.uniform !42

loop-memcpy-expansion5367:                        ; preds = %loop-memcpy-expansion5367.preheader, %loop-memcpy-expansion5367
  %loop-index5368 = phi i64 [ %3870, %loop-memcpy-expansion5367 ], [ 0, %loop-memcpy-expansion5367.preheader ]
  %scevgep6971 = getelementptr i8, ptr addrspace(4) %lsr.iv6969, i64 %loop-index5368, !amdgpu.uniform !42
  %3868 = load <4 x i32>, ptr addrspace(4) %scevgep6971, align 1
  %3869 = getelementptr i8, ptr %arrayidx.i22.i38.i.i.i.i2689, i64 %loop-index5368
  store <4 x i32> %3868, ptr %3869, align 1
  %3870 = add i64 %loop-index5368, 16
  %3871 = icmp uge i64 %3870, %3866
  br i1 %3871, label %Flow8068, label %loop-memcpy-expansion5367, !amdgpu.uniform !42

Flow8067:                                         ; preds = %Flow8066, %loop-memcpy-residual-header5370
  br label %Flow8070, !amdgpu.uniform !42

loop-memcpy-residual5369:                         ; preds = %loop-memcpy-residual5369.preheader, %loop-memcpy-residual5369
  %residual-loop-index5371 = phi i64 [ %3873, %loop-memcpy-residual5369 ], [ 0, %loop-memcpy-residual5369.preheader ]
  %scevgep6980 = getelementptr i8, ptr addrspace(4) %scevgep6979, i64 %residual-loop-index5371, !amdgpu.uniform !42
  %3872 = load i8, ptr addrspace(4) %scevgep6980, align 1
  %scevgep6978 = getelementptr i8, ptr %scevgep6977, i64 %residual-loop-index5371
  store i8 %3872, ptr %scevgep6978, align 1
  %3873 = add i64 %residual-loop-index5371, 1
  %3874 = icmp uge i64 %3873, %3865
  br i1 %3874, label %Flow8066, label %loop-memcpy-residual5369, !amdgpu.uniform !42

Flow8066:                                         ; preds = %loop-memcpy-residual5369
  br label %Flow8067, !amdgpu.uniform !42

_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i2680: ; preds = %Flow8070
  %conv.i.i26.i.i.i.i2681 = xor i32 %port.sroa.37.0.i.i2669, 1
  fence release
  %3875 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %arrayidx.i24.i29.i.i.i.i2682 = getelementptr inbounds nuw i32, ptr %3875, i64 %idxprom.i23.i.i.i.i.i2665
  store atomic i32 %conv.i.i26.i.i.i.i2681, ptr %arrayidx.i24.i29.i.i.i.i2682 monotonic, align 4
  %add.i.i.i.i2683 = add i64 %idx.0.i.i.i.i2670, 64
  %scevgep6970 = getelementptr i8, ptr addrspace(4) %lsr.iv6969, i64 64
  %lsr.iv.next6975 = add i64 %lsr.iv6974, -64
  br label %Flow8073, !amdgpu.uniform !42

cond.false.i.i.i2690:                             ; preds = %Flow8073
  %.lcssa9180 = phi i64 [ %3862, %Flow8073 ]
  %port.sroa.37.0.i.i2669.lcssa = phi i32 [ %port.sroa.37.0.i.i2669, %Flow8073 ]
  %.lcssa9179 = phi i32 [ %3843, %Flow8073 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa9180)
  %3876 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %.lcssa9179)
  %3877 = icmp ne i32 %3876, %port.sroa.37.0.i.i2669.lcssa
  %3878 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %3877)
  %3879 = extractvalue { i1, i64 } %3878, 0
  %3880 = extractvalue { i1, i64 } %3878, 1
  br i1 %3879, label %while.body.i.i.i.i2691.preheader, label %Flow8065

while.body.i.i.i.i2691.preheader:                 ; preds = %cond.false.i.i.i2690
  br label %while.body.i.i.i.i2691, !amdgpu.uniform !42

Flow8065:                                         ; preds = %Flow8064, %cond.false.i.i.i2690
  call void @llvm.amdgcn.end.cf.i64(i64 %3880)
  br label %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit2712, !amdgpu.uniform !42

while.body.i.i.i.i2691:                           ; preds = %while.body.i.i.i.i2691.preheader, %while.body.i.i.i.i2691
  %phi.broken8732 = phi i64 [ 0, %while.body.i.i.i.i2691.preheader ], [ %3884, %while.body.i.i.i.i2691 ]
  call void @llvm.amdgcn.s.sleep(i32 2)
  %3881 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i8.i.i2692 = getelementptr inbounds nuw i32, ptr %3881, i64 %idxprom.i23.i.i.i.i.i2665
  %3882 = load atomic i32, ptr %arrayidx.i.i.i8.i.i2692 monotonic, align 4
  %3883 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %3882)
  %cmp.i.not.i.i.i.i2693 = icmp eq i32 %3883, %port.sroa.37.0.i.i2669.lcssa
  %3884 = call i64 @llvm.amdgcn.if.break.i64(i1 %cmp.i.not.i.i.i.i2693, i64 %phi.broken8732)
  %3885 = call i1 @llvm.amdgcn.loop.i64(i64 %3884)
  br i1 %3885, label %Flow8064, label %while.body.i.i.i.i2691

Flow8064:                                         ; preds = %while.body.i.i.i.i2691
  %.lcssa9181 = phi i64 [ %3884, %while.body.i.i.i.i2691 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa9181)
  br label %Flow8065, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit2712: ; preds = %Flow8065
  fence acquire
  call void @llvm.amdgcn.wave.barrier()
  fence syncscope("agent") release
  %3886 = call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %.lcssa9176, i1 true)
  %iszero.i.i.i.i.i13.i.i2695 = icmp ne i64 %.lcssa9176, 0
  %cmp2.i.i.i.i14.i.i2696 = icmp eq i64 %3886, %sh_prom.i.i.i.i756
  %cmp.i.i.i.i15.i.i2697 = select i1 %iszero.i.i.i.i.i13.i.i2695, i1 %cmp2.i.i.i.i14.i.i2696, i1 false
  %3887 = and i32 %.lcssa9177, 31
  %conv4.i.i.i19.i.i2701 = zext i1 %cmp.i.i.i.i15.i.i2697 to i32
  %shl.i.i.i20.i.i2702 = shl nuw i32 %conv4.i.i.i19.i.i2701, %3887
  %xor.i.i.i21.i.i2703 = xor i32 %shl.i.i.i20.i.i2702, -1
  %3888 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop5797, !amdgpu.uniform !42

3889:                                             ; preds = %ComputeEnd5798
  %div8.i.i.i16.i.i2698 = lshr i32 %.lcssa9177, 5
  %idxprom.i.i.i17.i.i2699 = zext nneg i32 %div8.i.i.i16.i.i2698 to i64
  %3890 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i17.i.i2699
  %sunkaddr7448 = getelementptr inbounds i8, ptr addrspace(1) %3890, i64 40
  %3891 = atomicrmw and ptr addrspace(1) %sunkaddr7448, i32 %.lcssa9182 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %3892, !amdgpu.uniform !42

3892:                                             ; preds = %3889, %ComputeEnd5798
  call void @llvm.amdgcn.end.cf.i64(i64 %8092)
  call void @llvm.amdgcn.wave.barrier()
  br label %Flow8128, !amdgpu.uniform !42

if.else77:                                        ; preds = %if.else
  %mul66 = mul i64 %sub, 1000000
  %cmp78.not = icmp uge i64 %mul66, 1000000
  br i1 %cmp78.not, label %if.then79, label %Flow8062, !amdgpu.uniform !42

if.then79:                                        ; preds = %if.else77
  %div67 = udiv i64 %mul66, 1000000
  call void @llvm.lifetime.start.p5(ptr addrspace(5) %buffer.i.i.i2988) #34, !noalias !170
  br label %while.body.i.i.i.i.i.i2716, !amdgpu.uniform !42

Flow8062:                                         ; preds = %4169, %if.else77
  %3893 = phi i32 [ %.lcssa9131, %4169 ], [ poison, %if.else77 ]
  %3894 = phi i64 [ %.lcssa9132, %4169 ], [ poison, %if.else77 ]
  %3895 = phi i32 [ %.lcssa9133, %4169 ], [ poison, %if.else77 ]
  %3896 = phi i32 [ %.lcssa9120, %4169 ], [ poison, %if.else77 ]
  %3897 = phi i64 [ %.lcssa9121, %4169 ], [ poison, %if.else77 ]
  %3898 = phi i32 [ %.lcssa9122, %4169 ], [ poison, %if.else77 ]
  %3899 = phi i1 [ false, %4169 ], [ true, %if.else77 ]
  br i1 %3899, label %if.else82, label %Flow8063, !amdgpu.uniform !42

while.body.i.i.i.i.i.i2716:                       ; preds = %if.then79, %_ZN22__llvm_libc_22_0_0_git7details22StringBufferWriterImplILb0EE4pushEc.exit.i.i.i.i.i.i2730
  %writer.sroa.8.0.i.i.i.i2717 = phi i64 [ %writer.sroa.8.1.i.i.i.i2731, %_ZN22__llvm_libc_22_0_0_git7details22StringBufferWriterImplILb0EE4pushEc.exit.i.i.i.i.i.i2730 ], [ 0, %if.then79 ]
  %value.addr.09.i.i.i.i.i.i2718 = phi i64 [ %div.i.i.i.i.i.i.i2719, %_ZN22__llvm_libc_22_0_0_git7details22StringBufferWriterImplILb0EE4pushEc.exit.i.i.i.i.i.i2730 ], [ %div67, %if.then79 ]
  br label %NodeBlock7559, !amdgpu.uniform !42

NodeBlock7559:                                    ; preds = %while.body.i.i.i.i.i.i2716
  %div.i.i.i.i.i.i.i2719 = udiv i64 %value.addr.09.i.i.i.i.i.i2718, 10
  %.neg.i2720 = mul nuw nsw i64 %div.i.i.i.i.i.i.i2719, 246
  %rem.i.i.i.i.i.i.decomposed.i2721 = add nuw nsw i64 %.neg.i2720, %value.addr.09.i.i.i.i.i.i2718
  %conv.i.i.i.i.i.i.i2722 = trunc i64 %rem.i.i.i.i.i.i.decomposed.i2721 to i8
  %3900 = zext i8 %conv.i.i.i.i.i.i.i2722 to i16
  %Pivot7560 = icmp sge i16 %3900, 5
  br i1 %Pivot7560, label %NodeBlock7557, label %Flow7998, !amdgpu.uniform !42

NodeBlock7557:                                    ; preds = %NodeBlock7559
  %Pivot7558 = icmp sge i16 %3900, 7
  br i1 %Pivot7558, label %NodeBlock7555, label %Flow7989, !amdgpu.uniform !42

NodeBlock7555:                                    ; preds = %NodeBlock7557
  %Pivot7556 = icmp sge i16 %3900, 8
  br i1 %Pivot7556, label %NodeBlock7553, label %Flow7985, !amdgpu.uniform !42

NodeBlock7553:                                    ; preds = %NodeBlock7555
  %Pivot7554 = icmp sge i16 %3900, 9
  br i1 %Pivot7554, label %sw.bb9.i.i.i.i.i.i.i.i2723, label %Flow7983, !amdgpu.uniform !42

Flow7989:                                         ; preds = %Flow7986, %NodeBlock7557
  %3901 = phi i8 [ %3914, %Flow7986 ], [ poison, %NodeBlock7557 ]
  %3902 = phi i1 [ false, %Flow7986 ], [ true, %NodeBlock7557 ]
  br i1 %3902, label %NodeBlock7551, label %Flow7990, !amdgpu.uniform !42

NodeBlock7551:                                    ; preds = %Flow7989
  %Pivot7552 = icmp sge i16 %3900, 6
  br i1 %Pivot7552, label %sw.bb6.i.i.i.i.i.i.i.i2874, label %Flow7987, !amdgpu.uniform !42

Flow7998:                                         ; preds = %Flow7990, %NodeBlock7559
  %3903 = phi i8 [ %3916, %Flow7990 ], [ poison, %NodeBlock7559 ]
  %3904 = phi i1 [ false, %Flow7990 ], [ true, %NodeBlock7559 ]
  br i1 %3904, label %NodeBlock7549, label %if.then.i.i.i.i.i.i.i2724, !amdgpu.uniform !42

NodeBlock7549:                                    ; preds = %Flow7998
  %Pivot7550 = icmp sge i16 %3900, 3
  br i1 %Pivot7550, label %NodeBlock7547, label %Flow7996, !amdgpu.uniform !42

NodeBlock7547:                                    ; preds = %NodeBlock7549
  %Pivot7548 = icmp sge i16 %3900, 4
  br i1 %Pivot7548, label %sw.bb4.i.i.i.i.i.i.i.i2876, label %Flow7991, !amdgpu.uniform !42

Flow7996:                                         ; preds = %Flow7992, %NodeBlock7549
  %3905 = phi i8 [ %3917, %Flow7992 ], [ poison, %NodeBlock7549 ]
  %3906 = phi i1 [ false, %Flow7992 ], [ true, %NodeBlock7549 ]
  br i1 %3906, label %NodeBlock7545, label %Flow7997, !amdgpu.uniform !42

NodeBlock7545:                                    ; preds = %Flow7996
  %Pivot7546 = icmp sge i16 %3900, 2
  br i1 %Pivot7546, label %sw.bb2.i.i.i.i.i.i.i.i2878, label %Flow7994, !amdgpu.uniform !42

Flow7994:                                         ; preds = %sw.bb2.i.i.i.i.i.i.i.i2878, %NodeBlock7545
  %3907 = phi i1 [ false, %sw.bb2.i.i.i.i.i.i.i.i2878 ], [ true, %NodeBlock7545 ]
  br i1 %3907, label %LeafBlock7543, label %Flow7995, !amdgpu.uniform !42

LeafBlock7543:                                    ; preds = %Flow7994
  %SwitchLeaf7544 = icmp eq i16 %3900, 1
  br i1 %SwitchLeaf7544, label %sw.bb1.i.i.i.i.i.i.i.i2879, label %Flow7993, !amdgpu.uniform !42

sw.bb1.i.i.i.i.i.i.i.i2879:                       ; preds = %LeafBlock7543
  br label %Flow7993, !amdgpu.uniform !42

sw.bb2.i.i.i.i.i.i.i.i2878:                       ; preds = %NodeBlock7545
  br label %Flow7994, !amdgpu.uniform !42

Flow7991:                                         ; preds = %sw.bb4.i.i.i.i.i.i.i.i2876, %NodeBlock7547
  %3908 = phi i1 [ false, %sw.bb4.i.i.i.i.i.i.i.i2876 ], [ true, %NodeBlock7547 ]
  br i1 %3908, label %sw.bb3.i.i.i.i.i.i.i.i2877, label %Flow7992, !amdgpu.uniform !42

sw.bb3.i.i.i.i.i.i.i.i2877:                       ; preds = %Flow7991
  br label %Flow7992, !amdgpu.uniform !42

sw.bb4.i.i.i.i.i.i.i.i2876:                       ; preds = %NodeBlock7547
  br label %Flow7991, !amdgpu.uniform !42

Flow7987:                                         ; preds = %sw.bb6.i.i.i.i.i.i.i.i2874, %NodeBlock7551
  %3909 = phi i1 [ false, %sw.bb6.i.i.i.i.i.i.i.i2874 ], [ true, %NodeBlock7551 ]
  br i1 %3909, label %sw.bb5.i.i.i.i.i.i.i.i2875, label %Flow7988, !amdgpu.uniform !42

sw.bb5.i.i.i.i.i.i.i.i2875:                       ; preds = %Flow7987
  br label %Flow7988, !amdgpu.uniform !42

sw.bb6.i.i.i.i.i.i.i.i2874:                       ; preds = %NodeBlock7551
  br label %Flow7987, !amdgpu.uniform !42

Flow7985:                                         ; preds = %Flow7984, %NodeBlock7555
  %3910 = phi i8 [ %3913, %Flow7984 ], [ poison, %NodeBlock7555 ]
  %3911 = phi i1 [ false, %Flow7984 ], [ true, %NodeBlock7555 ]
  br i1 %3911, label %sw.bb7.i.i.i.i.i.i.i.i2873, label %Flow7986, !amdgpu.uniform !42

sw.bb7.i.i.i.i.i.i.i.i2873:                       ; preds = %Flow7985
  br label %Flow7986, !amdgpu.uniform !42

Flow7983:                                         ; preds = %sw.bb9.i.i.i.i.i.i.i.i2723, %NodeBlock7553
  %3912 = phi i1 [ false, %sw.bb9.i.i.i.i.i.i.i.i2723 ], [ true, %NodeBlock7553 ]
  br i1 %3912, label %sw.bb8.i.i.i.i.i.i.i.i2872, label %Flow7984, !amdgpu.uniform !42

sw.bb8.i.i.i.i.i.i.i.i2872:                       ; preds = %Flow7983
  br label %Flow7984, !amdgpu.uniform !42

sw.bb9.i.i.i.i.i.i.i.i2723:                       ; preds = %NodeBlock7553
  br label %Flow7983, !amdgpu.uniform !42

Flow7984:                                         ; preds = %sw.bb8.i.i.i.i.i.i.i.i2872, %Flow7983
  %3913 = phi i8 [ 56, %sw.bb8.i.i.i.i.i.i.i.i2872 ], [ 57, %Flow7983 ]
  br label %Flow7985, !amdgpu.uniform !42

Flow7986:                                         ; preds = %sw.bb7.i.i.i.i.i.i.i.i2873, %Flow7985
  %3914 = phi i8 [ 55, %sw.bb7.i.i.i.i.i.i.i.i2873 ], [ %3910, %Flow7985 ]
  br label %Flow7989, !amdgpu.uniform !42

Flow7988:                                         ; preds = %sw.bb5.i.i.i.i.i.i.i.i2875, %Flow7987
  %3915 = phi i8 [ 53, %sw.bb5.i.i.i.i.i.i.i.i2875 ], [ 54, %Flow7987 ]
  br label %Flow7990, !amdgpu.uniform !42

Flow7990:                                         ; preds = %Flow7988, %Flow7989
  %3916 = phi i8 [ %3915, %Flow7988 ], [ %3901, %Flow7989 ]
  br label %Flow7998, !amdgpu.uniform !42

Flow7992:                                         ; preds = %sw.bb3.i.i.i.i.i.i.i.i2877, %Flow7991
  %3917 = phi i8 [ 51, %sw.bb3.i.i.i.i.i.i.i.i2877 ], [ 52, %Flow7991 ]
  br label %Flow7996, !amdgpu.uniform !42

Flow7993:                                         ; preds = %sw.bb1.i.i.i.i.i.i.i.i2879, %LeafBlock7543
  %3918 = phi i8 [ 49, %sw.bb1.i.i.i.i.i.i.i.i2879 ], [ 48, %LeafBlock7543 ]
  br label %Flow7995, !amdgpu.uniform !42

Flow7995:                                         ; preds = %Flow7993, %Flow7994
  %3919 = phi i8 [ %3918, %Flow7993 ], [ 50, %Flow7994 ]
  br label %Flow7997, !amdgpu.uniform !42

Flow7997:                                         ; preds = %Flow7995, %Flow7996
  %3920 = phi i8 [ %3919, %Flow7995 ], [ %3905, %Flow7996 ]
  br label %if.then.i.i.i.i.i.i.i2724, !amdgpu.uniform !42

if.then.i.i.i.i.i.i.i2724:                        ; preds = %Flow7998, %Flow7997
  %retval.0.i.i.i.i.i.i.i.i2725 = phi i8 [ %3903, %Flow7998 ], [ %3920, %Flow7997 ]
  %cmp.i.i.i.i.i.i.i.i2726 = icmp eq i64 %writer.sroa.8.0.i.i.i.i2717, 21
  %cmp.i.i.i.i.i.i.i.i2726.inv = xor i1 %cmp.i.i.i.i.i.i.i.i2726, true
  br i1 %cmp.i.i.i.i.i.i.i.i2726.inv, label %if.then3.i.i.i.i.i.i.i2727, label %_ZN22__llvm_libc_22_0_0_git7details22StringBufferWriterImplILb0EE4pushEc.exit.i.i.i.i.i.i2730, !amdgpu.uniform !42

if.then3.i.i.i.i.i.i.i2727:                       ; preds = %if.then.i.i.i.i.i.i.i2724
  %3921 = trunc i64 %writer.sroa.8.0.i.i.i.i2717 to i32
  %3922 = xor i32 %3921, -1
  %3923 = getelementptr i8, ptr addrspace(5) %buffer.i.i.i2988, i32 %3922
  %arrayidx.i.i.i.i.i.i.i.i27286049 = getelementptr i8, ptr addrspace(5) %3923, i32 21
  %inc.i.i.i.i.i.i.i2729 = add i64 %writer.sroa.8.0.i.i.i.i2717, 1
  store i8 %retval.0.i.i.i.i.i.i.i.i2725, ptr addrspace(5) %arrayidx.i.i.i.i.i.i.i.i27286049, align 1, !tbaa !57, !noalias !170
  br label %_ZN22__llvm_libc_22_0_0_git7details22StringBufferWriterImplILb0EE4pushEc.exit.i.i.i.i.i.i2730, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git7details22StringBufferWriterImplILb0EE4pushEc.exit.i.i.i.i.i.i2730: ; preds = %if.then3.i.i.i.i.i.i.i2727, %if.then.i.i.i.i.i.i.i2724
  %writer.sroa.8.1.i.i.i.i2731 = phi i64 [ %inc.i.i.i.i.i.i.i2729, %if.then3.i.i.i.i.i.i.i2727 ], [ 21, %if.then.i.i.i.i.i.i.i2724 ]
  %cmp.i.i.i.i.i.i2732 = icmp samesign ult i64 %value.addr.09.i.i.i.i.i.i2718, 10
  %.not.i.i.i.i.i.i2733 = or i1 %cmp.i.i.i.i.i.i.i.i2726, %cmp.i.i.i.i.i.i2732
  br i1 %.not.i.i.i.i.i.i2733, label %_ZN22__llvm_libc_22_0_0_git15IntegerToStringIyNS_7details3FmtILh10ELb0ELb0ELb0ELm1EEEE13IntegerWriter25write_unsigned_number_decEyRNS1_22StringBufferWriterImplILb0EEE.exit.i.i.i.i.i2734, label %while.body.i.i.i.i.i.i2716, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git15IntegerToStringIyNS_7details3FmtILh10ELb0ELb0ELb0ELm1EEEE13IntegerWriter25write_unsigned_number_decEyRNS1_22StringBufferWriterImplILb0EEE.exit.i.i.i.i.i2734: ; preds = %_ZN22__llvm_libc_22_0_0_git7details22StringBufferWriterImplILb0EE4pushEc.exit.i.i.i.i.i.i2730
  %writer.sroa.8.1.i.i.i.i2731.lcssa = phi i64 [ %writer.sroa.8.1.i.i.i.i2731, %_ZN22__llvm_libc_22_0_0_git7details22StringBufferWriterImplILb0EE4pushEc.exit.i.i.i.i.i.i2730 ]
  %cmp.i.i.i.i.i.i.i.i2726.lcssa = phi i1 [ %cmp.i.i.i.i.i.i.i.i2726, %_ZN22__llvm_libc_22_0_0_git7details22StringBufferWriterImplILb0EE4pushEc.exit.i.i.i.i.i.i2730 ]
  %cmp.i15.i.i.i.i2737 = icmp ne i64 %writer.sroa.8.1.i.i.i.i2731.lcssa, 0
  %or.cond.not.i.i.i.i2738 = select i1 %cmp.i.i.i.i.i.i.i.i2726.lcssa, i1 true, i1 %cmp.i15.i.i.i.i2737
  %or.cond.not.i.i.i.i2738.inv = xor i1 %or.cond.not.i.i.i.i2738, true
  br i1 %or.cond.not.i.i.i.i2738.inv, label %if.then3.i.i.lr.ph.i.i.i.i2739, label %_ZN22__llvm_libc_22_0_0_git15IntegerToStringIyNS_7details3FmtILh10ELb0ELb0ELb0ELm1EEEEC2Ey.exit.i.i.i2741, !amdgpu.uniform !42

if.then3.i.i.lr.ph.i.i.i.i2739:                   ; preds = %_ZN22__llvm_libc_22_0_0_git15IntegerToStringIyNS_7details3FmtILh10ELb0ELb0ELb0ELm1EEEE13IntegerWriter25write_unsigned_number_decEyRNS1_22StringBufferWriterImplILb0EEE.exit.i.i.i.i.i2734
  %sunkaddr7449 = getelementptr inbounds i8, ptr addrspace(5) %buffer.i.i.i2988, i32 20
  store i8 48, ptr addrspace(5) %sunkaddr7449, align 4, !tbaa !57, !noalias !170
  br label %_ZN22__llvm_libc_22_0_0_git15IntegerToStringIyNS_7details3FmtILh10ELb0ELb0ELb0ELm1EEEEC2Ey.exit.i.i.i2741, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git15IntegerToStringIyNS_7details3FmtILh10ELb0ELb0ELb0ELm1EEEEC2Ey.exit.i.i.i2741: ; preds = %if.then3.i.i.lr.ph.i.i.i.i2739, %_ZN22__llvm_libc_22_0_0_git15IntegerToStringIyNS_7details3FmtILh10ELb0ELb0ELb0ELm1EEEE13IntegerWriter25write_unsigned_number_decEyRNS1_22StringBufferWriterImplILb0EEE.exit.i.i.i.i.i2734
  %writer.sroa.8.4.i.i.i.i2742 = phi i64 [ 1, %if.then3.i.i.lr.ph.i.i.i.i2739 ], [ %writer.sroa.8.1.i.i.i.i2731.lcssa, %_ZN22__llvm_libc_22_0_0_git15IntegerToStringIyNS_7details3FmtILh10ELb0ELb0ELb0ELm1EEEE13IntegerWriter25write_unsigned_number_decEyRNS1_22StringBufferWriterImplILb0EEE.exit.i.i.i.i.i2734 ]
  %sunkaddr7450 = getelementptr inbounds i8, ptr addrspace(5) %buffer.i.i.i2988, i32 24
  store i64 %writer.sroa.8.4.i.i.i.i2742, ptr addrspace(5) %sunkaddr7450, align 8, !tbaa !167, !noalias !170
  %cmp.i.not.i.i.i.i.i2744 = icmp ne i64 %writer.sroa.8.4.i.i.i.i2742, 0
  br i1 %cmp.i.not.i.i.i.i.i2744, label %if.then.i.i.i.i.i.i2745, label %Flow7982, !amdgpu.uniform !42

if.then.i.i.i.i.i.i2745:                          ; preds = %_ZN22__llvm_libc_22_0_0_git15IntegerToStringIyNS_7details3FmtILh10ELb0ELb0ELb0ELm1EEEEC2Ey.exit.i.i.i2741
  %3924 = call { i64, i1 } @llvm.uadd.with.overflow.i64(i64 %writer.sroa.8.4.i.i.i.i2742, i64 1)
  %ov7452 = extractvalue { i64, i1 } %3924, 1
  %ov7452.inv = xor i1 %ov7452, true
  br i1 %ov7452.inv, label %if.end.i.i.i.i.i.i.i2748, label %Flow7981, !amdgpu.uniform !42

if.end.i.i.i.i.i.i.i2748:                         ; preds = %if.then.i.i.i.i.i.i2745
  %math7451 = extractvalue { i64, i1 } %3924, 0
  %cmp2.i.i.i.i.i.i.i2749 = icmp ult i64 %math7451, 1676976733973595601
  br i1 %cmp2.i.i.i.i.i.i.i2749, label %if.then3.i.i.i.i4.i.i.i2869, label %if.end4.i.i.i.i.i.i.i2750, !amdgpu.uniform !42

if.then3.i.i.i.i4.i.i.i2869:                      ; preds = %if.end.i.i.i.i.i.i.i2748
  %mul.i.i.i.i.i.i.i2870 = mul nuw i64 %math7451, 11
  %div18.i.i.i.i.i.i.i2871 = lshr i64 %mul.i.i.i.i.i.i.i2870, 3
  br label %if.end4.i.i.i.i.i.i.i2750, !amdgpu.uniform !42

Flow7981:                                         ; preds = %if.end4.i.i.i.i.i.i.i2750, %if.then.i.i.i.i.i.i2745
  %3925 = phi ptr [ %cond.i.i.i2758, %if.end4.i.i.i.i.i.i.i2750 ], [ addrspacecast (ptr addrspace(4) @_ZN22__llvm_libc_22_0_0_git3cpp6string14NULL_CHARACTERE to ptr), %if.then.i.i.i.i.i.i2745 ]
  br label %_ZN22__llvm_libc_22_0_0_git3cpp6string7reserveEm.exit.i.i.i.i.i.i2760, !amdgpu.uniform !42

if.end4.i.i.i.i.i.i.i2750:                        ; preds = %if.then3.i.i.i.i4.i.i.i2869, %if.end.i.i.i.i.i.i.i2748
  %new_capacity.addr.0.i.i.i.i.i.i.i2751 = phi i64 [ %div18.i.i.i.i.i.i.i2871, %if.then3.i.i.i.i4.i.i.i2869 ], [ %math7451, %if.end.i.i.i.i.i.i.i2748 ]
  %sub.i.i.i2752 = add i64 %new_capacity.addr.0.i.i.i.i.i.i.i2751, 7
  %div2.i.i.i2753 = and i64 %sub.i.i.i2752, -8
  %3926 = load ptr, ptr addrspace(1) @_ZN12_GLOBAL__N_13ptrE, align 8, !tbaa !55
  %add.ptr.i.i.i2754 = getelementptr inbounds nuw i8, ptr %3926, i64 %div2.i.i.i2753
  store ptr %add.ptr.i.i.i2754, ptr addrspace(1) @_ZN12_GLOBAL__N_13ptrE, align 8, !tbaa !55
  %sub.ptr.lhs.cast.i.i.i2755 = ptrtoint ptr %add.ptr.i.i.i2754 to i64
  %sub.ptr.sub.i.i.i2756 = sub i64 %sub.ptr.lhs.cast.i.i.i2755, ptrtoint (ptr addrspacecast (ptr addrspace(1) @_ZN12_GLOBAL__N_16memoryE to ptr) to i64)
  %cmp.i.i.i2757 = icmp ugt i64 %sub.ptr.sub.i.i.i2756, 65335
  %cond.i.i.i2758 = select i1 %cmp.i.i.i2757, ptr null, ptr %3926
  br label %Flow7981, !amdgpu.uniform !42

Flow7982:                                         ; preds = %Flow7980, %_ZN22__llvm_libc_22_0_0_git15IntegerToStringIyNS_7details3FmtILh10ELb0ELb0ELb0ELm1EEEEC2Ey.exit.i.i.i2741
  %3927 = phi ptr [ %3925, %Flow7980 ], [ addrspacecast (ptr addrspace(4) @_ZN22__llvm_libc_22_0_0_git3cpp6string14NULL_CHARACTERE to ptr), %_ZN22__llvm_libc_22_0_0_git15IntegerToStringIyNS_7details3FmtILh10ELb0ELb0ELb0ELm1EEEEC2Ey.exit.i.i.i2741 ]
  br label %if.end.i.i.i.i.i.i2762, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git3cpp6string7reserveEm.exit.i.i.i.i.i.i2760: ; preds = %Flow7981
  br label %loadstoreloop4955, !amdgpu.uniform !42

loadstoreloop4955:                                ; preds = %_ZN22__llvm_libc_22_0_0_git3cpp6string7reserveEm.exit.i.i.i.i.i.i2760, %loadstoreloop4955
  %3928 = phi i64 [ %3930, %loadstoreloop4955 ], [ 0, %_ZN22__llvm_libc_22_0_0_git3cpp6string7reserveEm.exit.i.i.i.i.i.i2760 ]
  %3929 = getelementptr i8, ptr %3925, i64 %3928
  store i8 0, ptr %3929, align 1
  %3930 = add i64 %3928, 1
  %3931 = icmp uge i64 %3930, %writer.sroa.8.4.i.i.i.i2742
  br i1 %3931, label %Flow7980, label %loadstoreloop4955, !amdgpu.uniform !42

Flow7980:                                         ; preds = %loadstoreloop4955
  br label %Flow7982, !amdgpu.uniform !42

if.end.i.i.i.i.i.i2762:                           ; preds = %Flow7982
  %cmp.not.i6.i.i.i.i.i.i2764 = icmp ne ptr %3927, addrspacecast (ptr addrspace(4) @_ZN22__llvm_libc_22_0_0_git3cpp6string14NULL_CHARACTERE to ptr)
  br i1 %cmp.not.i6.i.i.i.i.i.i2764, label %if.then.i.i.i.i2.i.i.i2765, label %_ZN22__llvm_libc_22_0_0_git3cpp9to_stringEy.exit.i2767, !amdgpu.uniform !42

if.then.i.i.i.i2.i.i.i2765:                       ; preds = %if.end.i.i.i.i.i.i2762
  %arrayidx.i.i.i.i3.i.i.i2766 = getelementptr inbounds nuw i8, ptr %3927, i64 %writer.sroa.8.4.i.i.i.i2742
  store i8 0, ptr %arrayidx.i.i.i.i3.i.i.i2766, align 1, !tbaa !57, !noalias !170
  br label %_ZN22__llvm_libc_22_0_0_git3cpp9to_stringEy.exit.i2767, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git3cpp9to_stringEy.exit.i2767: ; preds = %if.then.i.i.i.i2.i.i.i2765, %if.end.i.i.i.i.i.i2762
  %3932 = trunc i64 %writer.sroa.8.4.i.i.i.i2742 to i32
  %3933 = and i64 %writer.sroa.8.4.i.i.i.i2742, 15
  %3934 = sub i64 %writer.sroa.8.4.i.i.i.i2742, %3933
  %3935 = icmp ne i64 %3934, 0
  br i1 %3935, label %loop-memcpy-expansion5127.preheader, label %Flow7979, !amdgpu.uniform !42

loop-memcpy-expansion5127.preheader:              ; preds = %_ZN22__llvm_libc_22_0_0_git3cpp9to_stringEy.exit.i2767
  %3936 = sub i32 0, %3932
  %scevgep6820 = getelementptr i8, ptr addrspace(5) %scevgep6819, i32 %3936
  br label %loop-memcpy-expansion5127, !amdgpu.uniform !42

Flow7979:                                         ; preds = %Flow7978, %_ZN22__llvm_libc_22_0_0_git3cpp9to_stringEy.exit.i2767
  br label %loop-memcpy-residual-header5130, !amdgpu.uniform !42

loop-memcpy-expansion5127:                        ; preds = %loop-memcpy-expansion5127.preheader, %loop-memcpy-expansion5127
  %lsr.iv6821 = phi ptr addrspace(5) [ %scevgep6822, %loop-memcpy-expansion5127 ], [ %scevgep6820, %loop-memcpy-expansion5127.preheader ], !amdgpu.uniform !42
  %loop-index5128 = phi i64 [ %3939, %loop-memcpy-expansion5127 ], [ 0, %loop-memcpy-expansion5127.preheader ]
  %3937 = load <4 x i32>, ptr addrspace(5) %lsr.iv6821, align 1
  %3938 = getelementptr i8, ptr %3927, i64 %loop-index5128
  store <4 x i32> %3937, ptr %3938, align 1
  %3939 = add i64 %loop-index5128, 16
  %scevgep6822 = getelementptr i8, ptr addrspace(5) %lsr.iv6821, i32 16
  %3940 = icmp uge i64 %3939, %3934
  br i1 %3940, label %Flow7978, label %loop-memcpy-expansion5127, !amdgpu.uniform !42

Flow7977:                                         ; preds = %Flow7976, %loop-memcpy-residual-header5130
  br label %post-loop-memcpy-expansion5126, !amdgpu.uniform !42

loop-memcpy-residual5129:                         ; preds = %loop-memcpy-residual5129.preheader, %loop-memcpy-residual5129
  %lsr.iv6825 = phi ptr addrspace(5) [ %scevgep6826, %loop-memcpy-residual5129 ], [ %scevgep6824, %loop-memcpy-residual5129.preheader ], !amdgpu.uniform !42
  %residual-loop-index5131 = phi i64 [ %3942, %loop-memcpy-residual5129 ], [ 0, %loop-memcpy-residual5129.preheader ]
  %3941 = load i8, ptr addrspace(5) %lsr.iv6825, align 1
  %scevgep6828 = getelementptr i8, ptr %scevgep6827, i64 %residual-loop-index5131
  store i8 %3941, ptr %scevgep6828, align 1
  %3942 = add i64 %residual-loop-index5131, 1
  %scevgep6826 = getelementptr i8, ptr addrspace(5) %lsr.iv6825, i32 1
  %3943 = icmp uge i64 %3942, %3933
  br i1 %3943, label %Flow7976, label %loop-memcpy-residual5129, !amdgpu.uniform !42

Flow7976:                                         ; preds = %loop-memcpy-residual5129
  br label %Flow7977, !amdgpu.uniform !42

post-loop-memcpy-expansion5126:                   ; preds = %Flow7977
  call void @llvm.lifetime.end.p5(ptr addrspace(5) %buffer.i.i.i2988) #34, !noalias !170
  br label %for.cond.i.i.i2771, !amdgpu.uniform !42

for.cond.i.i.i2771:                               ; preds = %post-loop-memcpy-expansion5126, %Flow7975
  %3944 = phi i32 [ %3981, %Flow7975 ], [ %1210, %post-loop-memcpy-expansion5126 ]
  %3945 = phi i64 [ %3982, %Flow7975 ], [ %1211, %post-loop-memcpy-expansion5126 ]
  %3946 = phi i32 [ %3983, %Flow7975 ], [ %1212, %post-loop-memcpy-expansion5126 ]
  %retval.sroa.6.0.i.i.i2772 = phi i32 [ %3986, %Flow7975 ], [ undef, %post-loop-memcpy-expansion5126 ]
  %retval.sroa.2.0.i.i.i2774 = phi i64 [ %3985, %Flow7975 ], [ undef, %post-loop-memcpy-expansion5126 ]
  %retval.sroa.8.0.i.i.i2775 = phi i32 [ %3984, %Flow7975 ], [ undef, %post-loop-memcpy-expansion5126 ]
  %index.0.i.i.i2776 = phi i32 [ %3988, %Flow7975 ], [ 0, %post-loop-memcpy-expansion5126 ]
  %3947 = load i32, ptr addrspace(1) @__llvm_rpc_client, align 8, !tbaa !68
  %cmp.not.i.i.i2777 = icmp ult i32 %index.0.i.i.i2776, %3947
  %spec.store.select.i.i.i2778 = select i1 %cmp.not.i.i.i2777, i32 %index.0.i.i.i2776, i32 0
  %3948 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  %3949 = and i32 %spec.store.select.i.i.i2778, 31
  %3950 = lshr i64 %3948, %sh_prom.i.i.i.i756
  %3951 = trunc i64 %3950 to i32
  %conv4.i.i.i.i.i2782 = and i32 %3951, 1
  %shl.i.i.i.i.i2783 = shl nuw i32 %conv4.i.i.i.i.i2782, %3949
  br label %ComputeLoop5801, !amdgpu.uniform !42

3952:                                             ; preds = %ComputeEnd5802
  %sunkaddr7453 = getelementptr inbounds i8, ptr addrspace(1) %8102, i64 40
  %3953 = atomicrmw or ptr addrspace(1) %sunkaddr7453, i32 %.lcssa9117 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %3954, !amdgpu.uniform !42

3954:                                             ; preds = %3952, %ComputeEnd5802
  %3955 = phi i32 [ %3953, %3952 ], [ poison, %ComputeEnd5802 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %8111)
  %3956 = call i32 @llvm.amdgcn.readfirstlane.i32(i32 %3955)
  %3957 = or i32 %3956, %.lcssa9118
  %shl5.i.i.i.i.i2784 = shl nuw i32 1, %3949
  %and.i.i.i.i.i2785 = and i32 %shl5.i.i.i.i.i2784, %3957
  %tobool3.i.i.i.i2786 = icmp ne i32 %and.i.i.i.i.i2785, 0
  %3958 = call i64 @llvm.amdgcn.ballot.i64(i1 %tobool3.i.i.i.i2786)
  %and.i.i.i.i.i.i2787 = and i64 %3958, %3948
  %cmp.i.not.i.i.i2788 = icmp ne i64 %3948, %and.i.i.i.i.i.i2787
  br i1 %cmp.i.not.i.i.i2788, label %if.then.i.i.i.i2789, label %Flow7974, !amdgpu.uniform !42

if.then.i.i.i.i2789:                              ; preds = %3954
  fence syncscope("agent") acquire
  %3959 = load <2 x i64>, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !75
  %3960 = extractelement <2 x i64> %3959, i32 0
  %3961 = inttoptr i64 %3960 to ptr
  %3962 = extractelement <2 x i64> %3959, i32 1
  %3963 = inttoptr i64 %3962 to ptr
  %idxprom.i.i.i.i2790 = zext i32 %spec.store.select.i.i.i2778 to i64
  %arrayidx.i.i.i.i2791 = getelementptr inbounds nuw i32, ptr %3961, i64 %idxprom.i.i.i.i2790, !amdgpu.uniform !42
  %3964 = load atomic i32, ptr %arrayidx.i.i.i.i2791 monotonic, align 4
  %3965 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %3964)
  %arrayidx.i45.i.i.i2792 = getelementptr inbounds nuw i32, ptr %3963, i64 %idxprom.i.i.i.i2790, !amdgpu.uniform !42
  %3966 = load atomic i32, ptr %arrayidx.i45.i.i.i2792 monotonic, align 4
  %3967 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %3966)
  %cmp.i46.not.i.i.i2793 = icmp ne i32 %3965, %3967
  br i1 %cmp.i46.not.i.i.i2793, label %if.then12.i.i.i2794, label %Flow7972, !amdgpu.uniform !42

if.then12.i.i.i2794:                              ; preds = %if.then.i.i.i.i2789
  fence syncscope("agent") release
  %3968 = call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %3948, i1 true)
  %iszero.i.i.i.i.i.i.i2795 = icmp ne i64 %3948, 0
  %cmp2.i.i.i.i.i.i2796 = icmp eq i64 %3968, %sh_prom.i.i.i.i756
  %cmp.i.i.i.i.i2.i2797 = select i1 %iszero.i.i.i.i.i.i.i2795, i1 %cmp2.i.i.i.i.i.i2796, i1 false
  %conv4.i.i51.i.i.i2798 = zext i1 %cmp.i.i.i.i.i2.i2797 to i32
  %shl.i.i52.i.i.i2799 = shl nuw i32 %conv4.i.i51.i.i.i2798, %3949
  %xor.i.i.i.i.i2800 = xor i32 %shl.i.i52.i.i.i2799, -1
  %3969 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop5806, !amdgpu.uniform !42

3970:                                             ; preds = %ComputeEnd5807
  %sunkaddr7454 = getelementptr inbounds i8, ptr addrspace(1) %8102, i64 40
  %3971 = atomicrmw and ptr addrspace(1) %sunkaddr7454, i32 %.lcssa9119 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %3972, !amdgpu.uniform !42

3972:                                             ; preds = %3970, %ComputeEnd5807
  call void @llvm.amdgcn.end.cf.i64(i64 %8129)
  call void @llvm.amdgcn.wave.barrier()
  br label %Flow7972, !amdgpu.uniform !42

if.end14.i.i.i2861:                               ; preds = %Flow7972
  %3973 = call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %3948, i1 true)
  %iszero.i.i.i.i.i.i2862 = icmp ne i64 %3948, 0
  %cmp2.i.i.i.i.i2863 = icmp eq i64 %3973, %sh_prom.i.i.i.i756
  %cmp.i.i.i.i.i2864 = select i1 %iszero.i.i.i.i.i.i2862, i1 %cmp2.i.i.i.i.i2863, i1 false
  %3974 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.i.i.i.i.i2864)
  %3975 = extractvalue { i1, i64 } %3974, 0
  %3976 = extractvalue { i1, i64 } %3974, 1
  br i1 %3975, label %if.then16.i.i.i2866, label %if.end22.i.i.i2865

if.then16.i.i.i2866:                              ; preds = %if.end14.i.i.i2861
  %3977 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !76
  %opcode.i.i.i2867 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %3977, i64 %idxprom.i.i.i.i2790, i32 1
  store i32 1660944387, ptr %opcode.i.i.i2867, align 8, !tbaa !77
  %arrayidx21.i.i.i2868 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %3977, i64 %idxprom.i.i.i.i2790
  store i64 %3948, ptr %arrayidx21.i.i.i2868, align 8, !tbaa !80
  br label %if.end22.i.i.i2865, !amdgpu.uniform !42

Flow7973:                                         ; preds = %if.end22.i.i.i2865, %Flow7972
  %3978 = phi i32 [ %3965, %if.end22.i.i.i2865 ], [ %retval.sroa.8.0.i.i.i2775, %Flow7972 ]
  %3979 = phi i64 [ %3948, %if.end22.i.i.i2865 ], [ %retval.sroa.2.0.i.i.i2774, %Flow7972 ]
  %3980 = phi i32 [ %spec.store.select.i.i.i2778, %if.end22.i.i.i2865 ], [ %retval.sroa.6.0.i.i.i2772, %Flow7972 ]
  br label %cleanup26.i.i.i2801, !amdgpu.uniform !42

if.end22.i.i.i2865:                               ; preds = %if.then16.i.i.i2866, %if.end14.i.i.i2861
  call void @llvm.amdgcn.end.cf.i64(i64 %3976)
  call void @llvm.amdgcn.wave.barrier()
  br label %Flow7973, !amdgpu.uniform !42

Flow7974:                                         ; preds = %cleanup26.i.i.i2801, %3954
  %3981 = phi i32 [ %3980, %cleanup26.i.i.i2801 ], [ %3944, %3954 ]
  %3982 = phi i64 [ %3979, %cleanup26.i.i.i2801 ], [ %3945, %3954 ]
  %3983 = phi i32 [ %3978, %cleanup26.i.i.i2801 ], [ %3946, %3954 ]
  %3984 = phi i32 [ %3978, %cleanup26.i.i.i2801 ], [ %retval.sroa.8.0.i.i.i2775, %3954 ]
  %3985 = phi i64 [ %3979, %cleanup26.i.i.i2801 ], [ %retval.sroa.2.0.i.i.i2774, %3954 ]
  %3986 = phi i32 [ %3980, %cleanup26.i.i.i2801 ], [ %retval.sroa.6.0.i.i.i2772, %3954 ]
  %3987 = phi i1 [ %cmp.i46.not.i.i.i2793, %cleanup26.i.i.i2801 ], [ true, %3954 ]
  br i1 %3987, label %for.inc.i.i.i2806, label %Flow7975, !amdgpu.uniform !42

cleanup26.i.i.i2801:                              ; preds = %Flow7973
  br label %Flow7974, !amdgpu.uniform !42

for.inc.i.i.i2806:                                ; preds = %Flow7974
  %inc.i.i.i2811 = add i32 %spec.store.select.i.i.i2778, 1
  br label %Flow7975, !amdgpu.uniform !42

Flow7975:                                         ; preds = %for.inc.i.i.i2806, %Flow7974
  %3988 = phi i32 [ %inc.i.i.i2811, %for.inc.i.i.i2806 ], [ poison, %Flow7974 ]
  %3989 = phi i1 [ false, %for.inc.i.i.i2806 ], [ true, %Flow7974 ]
  br i1 %3989, label %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i2812, label %for.cond.i.i.i2771, !amdgpu.uniform !42

_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i2812: ; preds = %Flow7975
  %.lcssa9122 = phi i32 [ %3981, %Flow7975 ]
  %.lcssa9121 = phi i64 [ %3982, %Flow7975 ]
  %.lcssa9120 = phi i32 [ %3983, %Flow7975 ]
  fence acquire
  %3990 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %mul.i.i.i.i.i.i2813 = shl i32 %.lcssa9122, 6
  %idxprom.i19.i.i.i.i.i2814 = zext i32 %mul.i.i.i.i.i.i2813 to i64
  %arrayidx.i20.i.i.i.i.i2815 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %3990, i64 %idxprom.i19.i.i.i.i.i2814
  %arrayidx.i22.i.i.i.i.i2816 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %arrayidx.i20.i.i.i.i.i2815, i64 %sh_prom.i.i.i.i756
  store i64 %writer.sroa.8.4.i.i.i.i2742, ptr %arrayidx.i22.i.i.i.i.i2816, align 8, !tbaa !82
  %spec.select.i.i.i.i.i.i.i2817 = call i64 @llvm.umin.i64(i64 %writer.sroa.8.4.i.i.i.i2742, i64 56)
  %3991 = and i64 %spec.select.i.i.i.i.i.i.i2817, 15
  %3992 = sub i64 %spec.select.i.i.i.i.i.i.i2817, %3991
  %3993 = icmp ne i64 %3992, 0
  br i1 %3993, label %loop-memcpy-expansion5025.preheader, label %Flow7971, !amdgpu.uniform !42

loop-memcpy-expansion5025.preheader:              ; preds = %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i2812
  %scevgep6829 = getelementptr i8, ptr %3990, i64 8
  %scevgep6830 = getelementptr i8, ptr %scevgep6829, i64 %1474
  %3994 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i2814, 6
  %scevgep6831 = getelementptr i8, ptr %scevgep6830, i64 %3994
  br label %loop-memcpy-expansion5025, !amdgpu.uniform !42

Flow7971:                                         ; preds = %Flow7970, %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i2812
  br label %loop-memcpy-residual-header5028, !amdgpu.uniform !42

loop-memcpy-expansion5025:                        ; preds = %loop-memcpy-expansion5025.preheader, %loop-memcpy-expansion5025
  %loop-index5026 = phi i64 [ %3997, %loop-memcpy-expansion5025 ], [ 0, %loop-memcpy-expansion5025.preheader ]
  %3995 = getelementptr i8, ptr %3927, i64 %loop-index5026, !amdgpu.uniform !42
  %3996 = load <4 x i32>, ptr %3995, align 1
  %scevgep6832 = getelementptr i8, ptr %scevgep6831, i64 %loop-index5026
  store <4 x i32> %3996, ptr %scevgep6832, align 1
  %3997 = add i64 %loop-index5026, 16
  %3998 = icmp uge i64 %3997, %3992
  br i1 %3998, label %Flow7970, label %loop-memcpy-expansion5025, !amdgpu.uniform !42

Flow7969:                                         ; preds = %Flow7967, %Flow7968
  %3999 = phi i64 [ %6273, %Flow7967 ], [ %6270, %Flow7968 ]
  br label %post-loop-memcpy-expansion5024, !amdgpu.uniform !42

loop-memcpy-residual5027:                         ; preds = %loop-memcpy-residual5027.preheader, %loop-memcpy-residual5027
  %residual-loop-index5029 = phi i64 [ %4001, %loop-memcpy-residual5027 ], [ 0, %loop-memcpy-residual5027.preheader ]
  %scevgep6838 = getelementptr i8, ptr %scevgep6837, i64 %residual-loop-index5029, !amdgpu.uniform !42
  %4000 = load i8, ptr %scevgep6838, align 1
  %scevgep6836 = getelementptr i8, ptr %scevgep6835, i64 %residual-loop-index5029
  store i8 %4000, ptr %scevgep6836, align 1
  %4001 = add i64 %residual-loop-index5029, 1
  %4002 = icmp uge i64 %4001, %3991
  br i1 %4002, label %Flow7967, label %loop-memcpy-residual5027, !amdgpu.uniform !42

Flow7967:                                         ; preds = %loop-memcpy-residual5027
  br label %Flow7969, !amdgpu.uniform !42

post-loop-memcpy-expansion5024:                   ; preds = %Flow7969
  %tobool.not.i.i.i.i.i.i2819 = icmp eq i32 %.lcssa9120, 0
  %conv.i.i.i.i.i.i2820 = zext i1 %tobool.not.i.i.i.i.i.i2819 to i32
  fence release
  %4003 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %idxprom.i23.i.i.i.i.i2821 = zext i32 %.lcssa9122 to i64
  %arrayidx.i24.i.i.i.i.i2822 = getelementptr inbounds nuw i32, ptr %4003, i64 %idxprom.i23.i.i.i.i.i2821
  store atomic i32 %conv.i.i.i.i.i.i2820, ptr %arrayidx.i24.i.i.i.i.i2822 monotonic, align 4
  %4004 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !84
  %arrayidx.i.i5.i.i2823 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %4004, i64 %idxprom.i23.i.i.i.i.i2821
  %4005 = load i64, ptr %arrayidx.i.i5.i.i2823, align 8, !tbaa !80
  %scevgep6839 = getelementptr i8, ptr %3927, i64 56
  %4006 = add i64 %writer.sroa.8.4.i.i.i.i2742, -56
  br label %while.cond.i.i.i.i2824, !amdgpu.uniform !42

while.cond.i.i.i.i2824:                           ; preds = %post-loop-memcpy-expansion5024, %Flow7966
  %phi.broken8716 = phi i64 [ 0, %post-loop-memcpy-expansion5024 ], [ %4028, %Flow7966 ]
  %lsr.iv6845 = phi i64 [ %4026, %Flow7966 ], [ %4006, %post-loop-memcpy-expansion5024 ]
  %lsr.iv6840 = phi ptr [ %4025, %Flow7966 ], [ %scevgep6839, %post-loop-memcpy-expansion5024 ]
  %port.sroa.37.0.i.i2825 = phi i32 [ %4024, %Flow7966 ], [ %conv.i.i.i.i.i.i2820, %post-loop-memcpy-expansion5024 ]
  %idx.0.i.i.i.i2826 = phi i64 [ %4023, %Flow7966 ], [ 56, %post-loop-memcpy-expansion5024 ]
  %cmp.i.i.i.i2827 = icmp ugt i64 %writer.sroa.8.4.i.i.i.i2742, %idx.0.i.i.i.i2826
  %4007 = call i64 @llvm.amdgcn.ballot.i64(i1 %cmp.i.i.i.i2827)
  %and.i.i.i.i6.i.i2828 = and i64 %4007, %4005
  %tobool.not.i.i.i.i2829 = icmp ne i64 %and.i.i.i.i6.i.i2828, 0
  %4008 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i22.i.i.i2830 = getelementptr inbounds nuw i32, ptr %4008, i64 %idxprom.i23.i.i.i.i.i2821
  %4009 = load atomic i32, ptr %arrayidx.i22.i.i.i2830 monotonic, align 4
  %4010 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %tobool.not.i.i.i.i2829)
  %4011 = extractvalue { i1, i64 } %4010, 0
  %4012 = extractvalue { i1, i64 } %4010, 1
  br i1 %4011, label %cond.false.i8.i.i.i.i2831, label %Flow7966

cond.false.i8.i.i.i.i2831:                        ; preds = %while.cond.i.i.i.i2824
  %4013 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %4009)
  %4014 = icmp ne i32 %4013, %port.sroa.37.0.i.i2825
  %4015 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %4014)
  %4016 = extractvalue { i1, i64 } %4015, 0
  %4017 = extractvalue { i1, i64 } %4015, 1
  br i1 %4016, label %while.body.i.i21.i.i.i.i2832.preheader, label %Flow7965

while.body.i.i21.i.i.i.i2832.preheader:           ; preds = %cond.false.i8.i.i.i.i2831
  br label %while.body.i.i21.i.i.i.i2832, !amdgpu.uniform !42

Flow7965:                                         ; preds = %Flow7964, %cond.false.i8.i.i.i.i2831
  call void @llvm.amdgcn.end.cf.i64(i64 %4017)
  br label %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i2835, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i2832:                     ; preds = %while.body.i.i21.i.i.i.i2832.preheader, %while.body.i.i21.i.i.i.i2832
  %phi.broken8715 = phi i64 [ 0, %while.body.i.i21.i.i.i.i2832.preheader ], [ %4021, %while.body.i.i21.i.i.i.i2832 ]
  call void @llvm.amdgcn.s.sleep(i32 2)
  %4018 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i22.i.i.i.i2833 = getelementptr inbounds nuw i32, ptr %4018, i64 %idxprom.i23.i.i.i.i.i2821
  %4019 = load atomic i32, ptr %arrayidx.i.i.i22.i.i.i.i2833 monotonic, align 4
  %4020 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %4019)
  %cmp.i.not.i.i23.i.i.i.i2834 = icmp eq i32 %4020, %port.sroa.37.0.i.i2825
  %4021 = call i64 @llvm.amdgcn.if.break.i64(i1 %cmp.i.not.i.i23.i.i.i.i2834, i64 %phi.broken8715)
  %4022 = call i1 @llvm.amdgcn.loop.i64(i64 %4021)
  br i1 %4022, label %Flow7964, label %while.body.i.i21.i.i.i.i2832

Flow7964:                                         ; preds = %while.body.i.i21.i.i.i.i2832
  %.lcssa9123 = phi i64 [ %4021, %while.body.i.i21.i.i.i.i2832 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa9123)
  br label %Flow7965, !amdgpu.uniform !42

Flow7966:                                         ; preds = %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i2836, %while.cond.i.i.i.i2824
  %4023 = phi i64 [ %add.i.i.i.i2839, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i2836 ], [ poison, %while.cond.i.i.i.i2824 ]
  %4024 = phi i32 [ %conv.i.i26.i.i.i.i2837, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i2836 ], [ poison, %while.cond.i.i.i.i2824 ]
  %4025 = phi ptr [ %scevgep6841, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i2836 ], [ poison, %while.cond.i.i.i.i2824 ]
  %4026 = phi i64 [ %lsr.iv.next6846, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i2836 ], [ poison, %while.cond.i.i.i.i2824 ]
  %4027 = phi i1 [ false, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i2836 ], [ true, %while.cond.i.i.i.i2824 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %4012)
  %4028 = call i64 @llvm.amdgcn.if.break.i64(i1 %4027, i64 %phi.broken8716)
  %4029 = call i1 @llvm.amdgcn.loop.i64(i64 %4028)
  br i1 %4029, label %cond.false.i.i.i2846, label %while.cond.i.i.i.i2824

_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i2835: ; preds = %Flow7965
  fence acquire
  br i1 %cmp.i.i.i.i2827, label %if.then.i.i.i.i.i.i3.i2840, label %Flow7963, !amdgpu.uniform !42

if.then.i.i.i.i.i.i3.i2840:                       ; preds = %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i2835
  %sub.i.i.i.i.i.i.i2842 = sub nuw i64 %writer.sroa.8.4.i.i.i.i2742, %idx.0.i.i.i.i2826
  %spec.select.i.i.i36.i.i.i.i2843 = call i64 @llvm.umin.i64(i64 %sub.i.i.i.i.i.i.i2842, i64 64)
  %4030 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %4031 = and i64 %spec.select.i.i.i36.i.i.i.i2843, 15
  %4032 = sub i64 %spec.select.i.i.i36.i.i.i.i2843, %4031
  %4033 = icmp ne i64 %4032, 0
  br i1 %4033, label %loop-memcpy-expansion5019.preheader, label %Flow7962, !amdgpu.uniform !42

loop-memcpy-expansion5019.preheader:              ; preds = %if.then.i.i.i.i.i.i3.i2840
  %arrayidx.i20.i35.i.i.i.i2844 = getelementptr %"struct.rpc::Buffer", ptr %4030, i64 %idxprom.i19.i.i.i.i.i2814
  %arrayidx.i22.i38.i.i.i.i2845 = getelementptr %"struct.rpc::Buffer", ptr %arrayidx.i20.i35.i.i.i.i2844, i64 %sh_prom.i.i.i.i756
  br label %loop-memcpy-expansion5019, !amdgpu.uniform !42

Flow7962:                                         ; preds = %Flow7961, %if.then.i.i.i.i.i.i3.i2840
  br label %loop-memcpy-residual-header5022, !amdgpu.uniform !42

loop-memcpy-expansion5019:                        ; preds = %loop-memcpy-expansion5019.preheader, %loop-memcpy-expansion5019
  %loop-index5020 = phi i64 [ %4036, %loop-memcpy-expansion5019 ], [ 0, %loop-memcpy-expansion5019.preheader ]
  %scevgep6842 = getelementptr i8, ptr %lsr.iv6840, i64 %loop-index5020, !amdgpu.uniform !42
  %4034 = load <4 x i32>, ptr %scevgep6842, align 1
  %4035 = getelementptr i8, ptr %arrayidx.i22.i38.i.i.i.i2845, i64 %loop-index5020
  store <4 x i32> %4034, ptr %4035, align 1
  %4036 = add i64 %loop-index5020, 16
  %4037 = icmp uge i64 %4036, %4032
  br i1 %4037, label %Flow7961, label %loop-memcpy-expansion5019, !amdgpu.uniform !42

Flow7960:                                         ; preds = %Flow7959, %loop-memcpy-residual-header5022
  br label %Flow7963, !amdgpu.uniform !42

loop-memcpy-residual5021:                         ; preds = %loop-memcpy-residual5021.preheader, %loop-memcpy-residual5021
  %residual-loop-index5023 = phi i64 [ %4039, %loop-memcpy-residual5021 ], [ 0, %loop-memcpy-residual5021.preheader ]
  %scevgep6851 = getelementptr i8, ptr %scevgep6850, i64 %residual-loop-index5023, !amdgpu.uniform !42
  %4038 = load i8, ptr %scevgep6851, align 1
  %scevgep6849 = getelementptr i8, ptr %scevgep6848, i64 %residual-loop-index5023
  store i8 %4038, ptr %scevgep6849, align 1
  %4039 = add i64 %residual-loop-index5023, 1
  %4040 = icmp uge i64 %4039, %4031
  br i1 %4040, label %Flow7959, label %loop-memcpy-residual5021, !amdgpu.uniform !42

Flow7959:                                         ; preds = %loop-memcpy-residual5021
  br label %Flow7960, !amdgpu.uniform !42

_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i2836: ; preds = %Flow7963
  %conv.i.i26.i.i.i.i2837 = xor i32 %port.sroa.37.0.i.i2825, 1
  fence release
  %4041 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %arrayidx.i24.i29.i.i.i.i2838 = getelementptr inbounds nuw i32, ptr %4041, i64 %idxprom.i23.i.i.i.i.i2821
  store atomic i32 %conv.i.i26.i.i.i.i2837, ptr %arrayidx.i24.i29.i.i.i.i2838 monotonic, align 4
  %add.i.i.i.i2839 = add i64 %idx.0.i.i.i.i2826, 64
  %scevgep6841 = getelementptr i8, ptr %lsr.iv6840, i64 64
  %lsr.iv.next6846 = add i64 %lsr.iv6845, -64
  br label %Flow7966, !amdgpu.uniform !42

cond.false.i.i.i2846:                             ; preds = %Flow7966
  %.lcssa9125 = phi i64 [ %4028, %Flow7966 ]
  %port.sroa.37.0.i.i2825.lcssa = phi i32 [ %port.sroa.37.0.i.i2825, %Flow7966 ]
  %.lcssa9124 = phi i32 [ %4009, %Flow7966 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa9125)
  %4042 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %.lcssa9124)
  %4043 = icmp ne i32 %4042, %port.sroa.37.0.i.i2825.lcssa
  %4044 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %4043)
  %4045 = extractvalue { i1, i64 } %4044, 0
  %4046 = extractvalue { i1, i64 } %4044, 1
  br i1 %4045, label %while.body.i.i.i.i2847.preheader, label %Flow7958

while.body.i.i.i.i2847.preheader:                 ; preds = %cond.false.i.i.i2846
  br label %while.body.i.i.i.i2847, !amdgpu.uniform !42

Flow7958:                                         ; preds = %Flow7957, %cond.false.i.i.i2846
  call void @llvm.amdgcn.end.cf.i64(i64 %4046)
  br label %_ZN22__llvm_libc_22_0_0_git15write_to_stderrENS_3cpp11string_viewE.exit.i2850, !amdgpu.uniform !42

while.body.i.i.i.i2847:                           ; preds = %while.body.i.i.i.i2847.preheader, %while.body.i.i.i.i2847
  %phi.broken8717 = phi i64 [ 0, %while.body.i.i.i.i2847.preheader ], [ %4050, %while.body.i.i.i.i2847 ]
  call void @llvm.amdgcn.s.sleep(i32 2)
  %4047 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i8.i.i2848 = getelementptr inbounds nuw i32, ptr %4047, i64 %idxprom.i23.i.i.i.i.i2821
  %4048 = load atomic i32, ptr %arrayidx.i.i.i8.i.i2848 monotonic, align 4
  %4049 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %4048)
  %cmp.i.not.i.i.i.i2849 = icmp eq i32 %4049, %port.sroa.37.0.i.i2825.lcssa
  %4050 = call i64 @llvm.amdgcn.if.break.i64(i1 %cmp.i.not.i.i.i.i2849, i64 %phi.broken8717)
  %4051 = call i1 @llvm.amdgcn.loop.i64(i64 %4050)
  br i1 %4051, label %Flow7957, label %while.body.i.i.i.i2847

Flow7957:                                         ; preds = %while.body.i.i.i.i2847
  %.lcssa9126 = phi i64 [ %4050, %while.body.i.i.i.i2847 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa9126)
  br label %Flow7958, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git15write_to_stderrENS_3cpp11string_viewE.exit.i2850: ; preds = %Flow7958
  fence acquire
  call void @llvm.amdgcn.wave.barrier()
  fence syncscope("agent") release
  %4052 = call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %.lcssa9121, i1 true)
  %iszero.i.i.i.i.i13.i.i2851 = icmp ne i64 %.lcssa9121, 0
  %cmp2.i.i.i.i14.i.i2852 = icmp eq i64 %4052, %sh_prom.i.i.i.i756
  %cmp.i.i.i.i15.i.i2853 = select i1 %iszero.i.i.i.i.i13.i.i2851, i1 %cmp2.i.i.i.i14.i.i2852, i1 false
  %4053 = and i32 %.lcssa9122, 31
  %conv4.i.i.i19.i.i2857 = zext i1 %cmp.i.i.i.i15.i.i2853 to i32
  %shl.i.i.i20.i.i2858 = shl nuw i32 %conv4.i.i.i19.i.i2857, %4053
  %xor.i.i.i21.i.i2859 = xor i32 %shl.i.i.i20.i.i2858, -1
  %4054 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop5810, !amdgpu.uniform !42

4055:                                             ; preds = %ComputeEnd5811
  %div8.i.i.i16.i.i2854 = lshr i32 %.lcssa9122, 5
  %idxprom.i.i.i17.i.i2855 = zext nneg i32 %div8.i.i.i16.i.i2854 to i64
  %4056 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i17.i.i2855
  %sunkaddr7455 = getelementptr inbounds i8, ptr addrspace(1) %4056, i64 40
  %4057 = atomicrmw and ptr addrspace(1) %sunkaddr7455, i32 %.lcssa9127 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %4058, !amdgpu.uniform !42

4058:                                             ; preds = %4055, %ComputeEnd5811
  call void @llvm.amdgcn.end.cf.i64(i64 %8146)
  call void @llvm.amdgcn.wave.barrier()
  br label %for.cond.i.i1.i2890, !amdgpu.uniform !42

for.cond.i.i1.i2890:                              ; preds = %4058, %Flow7956
  %4059 = phi i32 [ %4096, %Flow7956 ], [ %1213, %4058 ]
  %4060 = phi i64 [ %4097, %Flow7956 ], [ %1214, %4058 ]
  %4061 = phi i32 [ %4098, %Flow7956 ], [ %1215, %4058 ]
  %retval.sroa.6.0.i.i.i2891 = phi i32 [ %4101, %Flow7956 ], [ undef, %4058 ]
  %retval.sroa.2.0.i.i.i2893 = phi i64 [ %4100, %Flow7956 ], [ undef, %4058 ]
  %retval.sroa.8.0.i.i.i2894 = phi i32 [ %4099, %Flow7956 ], [ undef, %4058 ]
  %index.0.i.i.i2895 = phi i32 [ %4103, %Flow7956 ], [ 0, %4058 ]
  %4062 = load i32, ptr addrspace(1) @__llvm_rpc_client, align 8, !tbaa !68
  %cmp.not.i.i.i2896 = icmp ult i32 %index.0.i.i.i2895, %4062
  %spec.store.select.i.i.i2897 = select i1 %cmp.not.i.i.i2896, i32 %index.0.i.i.i2895, i32 0
  %4063 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  %4064 = and i32 %spec.store.select.i.i.i2897, 31
  %4065 = lshr i64 %4063, %sh_prom.i.i.i.i756
  %4066 = trunc i64 %4065 to i32
  %conv4.i.i.i.i.i2901 = and i32 %4066, 1
  %shl.i.i.i.i.i2902 = shl nuw i32 %conv4.i.i.i.i.i2901, %4064
  br label %ComputeLoop5814, !amdgpu.uniform !42

4067:                                             ; preds = %ComputeEnd5815
  %sunkaddr7456 = getelementptr inbounds i8, ptr addrspace(1) %8156, i64 40
  %4068 = atomicrmw or ptr addrspace(1) %sunkaddr7456, i32 %.lcssa9128 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %4069, !amdgpu.uniform !42

4069:                                             ; preds = %4067, %ComputeEnd5815
  %4070 = phi i32 [ %4068, %4067 ], [ poison, %ComputeEnd5815 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %8165)
  %4071 = call i32 @llvm.amdgcn.readfirstlane.i32(i32 %4070)
  %4072 = or i32 %4071, %.lcssa9129
  %shl5.i.i.i.i.i2903 = shl nuw i32 1, %4064
  %and.i.i.i.i.i2904 = and i32 %shl5.i.i.i.i.i2903, %4072
  %tobool3.i.i.i.i2905 = icmp ne i32 %and.i.i.i.i.i2904, 0
  %4073 = call i64 @llvm.amdgcn.ballot.i64(i1 %tobool3.i.i.i.i2905)
  %and.i.i.i.i.i.i2906 = and i64 %4073, %4063
  %cmp.i.not.i.i.i2907 = icmp ne i64 %4063, %and.i.i.i.i.i.i2906
  br i1 %cmp.i.not.i.i.i2907, label %if.then.i.i.i.i2908, label %Flow7955, !amdgpu.uniform !42

if.then.i.i.i.i2908:                              ; preds = %4069
  fence syncscope("agent") acquire
  %4074 = load <2 x i64>, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !75
  %4075 = extractelement <2 x i64> %4074, i32 0
  %4076 = inttoptr i64 %4075 to ptr
  %4077 = extractelement <2 x i64> %4074, i32 1
  %4078 = inttoptr i64 %4077 to ptr
  %idxprom.i.i.i.i2909 = zext i32 %spec.store.select.i.i.i2897 to i64
  %arrayidx.i.i.i.i2910 = getelementptr inbounds nuw i32, ptr %4076, i64 %idxprom.i.i.i.i2909, !amdgpu.uniform !42
  %4079 = load atomic i32, ptr %arrayidx.i.i.i.i2910 monotonic, align 4
  %4080 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %4079)
  %arrayidx.i45.i.i.i2911 = getelementptr inbounds nuw i32, ptr %4078, i64 %idxprom.i.i.i.i2909, !amdgpu.uniform !42
  %4081 = load atomic i32, ptr %arrayidx.i45.i.i.i2911 monotonic, align 4
  %4082 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %4081)
  %cmp.i46.not.i.i.i2912 = icmp ne i32 %4080, %4082
  br i1 %cmp.i46.not.i.i.i2912, label %if.then12.i.i.i2913, label %Flow7953, !amdgpu.uniform !42

if.then12.i.i.i2913:                              ; preds = %if.then.i.i.i.i2908
  fence syncscope("agent") release
  %4083 = call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %4063, i1 true)
  %iszero.i.i.i.i.i.i.i2914 = icmp ne i64 %4063, 0
  %cmp2.i.i.i.i.i.i2915 = icmp eq i64 %4083, %sh_prom.i.i.i.i756
  %cmp.i.i.i.i.i.i2916 = select i1 %iszero.i.i.i.i.i.i.i2914, i1 %cmp2.i.i.i.i.i.i2915, i1 false
  %conv4.i.i51.i.i.i2917 = zext i1 %cmp.i.i.i.i.i.i2916 to i32
  %shl.i.i52.i.i.i2918 = shl nuw i32 %conv4.i.i51.i.i.i2917, %4064
  %xor.i.i.i.i.i2919 = xor i32 %shl.i.i52.i.i.i2918, -1
  %4084 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop5819, !amdgpu.uniform !42

4085:                                             ; preds = %ComputeEnd5820
  %sunkaddr7457 = getelementptr inbounds i8, ptr addrspace(1) %8156, i64 40
  %4086 = atomicrmw and ptr addrspace(1) %sunkaddr7457, i32 %.lcssa9130 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %4087, !amdgpu.uniform !42

4087:                                             ; preds = %4085, %ComputeEnd5820
  call void @llvm.amdgcn.end.cf.i64(i64 %8183)
  call void @llvm.amdgcn.wave.barrier()
  br label %Flow7953, !amdgpu.uniform !42

if.end14.i.i.i2979:                               ; preds = %Flow7953
  %4088 = call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %4063, i1 true)
  %iszero.i.i.i.i.i.i2980 = icmp ne i64 %4063, 0
  %cmp2.i.i.i.i.i2981 = icmp eq i64 %4088, %sh_prom.i.i.i.i756
  %cmp.i.i.i.i.i2982 = select i1 %iszero.i.i.i.i.i.i2980, i1 %cmp2.i.i.i.i.i2981, i1 false
  %4089 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.i.i.i.i.i2982)
  %4090 = extractvalue { i1, i64 } %4089, 0
  %4091 = extractvalue { i1, i64 } %4089, 1
  br i1 %4090, label %if.then16.i.i.i2984, label %if.end22.i.i.i2983

if.then16.i.i.i2984:                              ; preds = %if.end14.i.i.i2979
  %4092 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !76
  %opcode.i.i.i2985 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %4092, i64 %idxprom.i.i.i.i2909, i32 1
  store i32 1660944387, ptr %opcode.i.i.i2985, align 8, !tbaa !77
  %arrayidx21.i.i.i2986 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %4092, i64 %idxprom.i.i.i.i2909
  store i64 %4063, ptr %arrayidx21.i.i.i2986, align 8, !tbaa !80
  br label %if.end22.i.i.i2983, !amdgpu.uniform !42

Flow7954:                                         ; preds = %if.end22.i.i.i2983, %Flow7953
  %4093 = phi i32 [ %4080, %if.end22.i.i.i2983 ], [ %retval.sroa.8.0.i.i.i2894, %Flow7953 ]
  %4094 = phi i64 [ %4063, %if.end22.i.i.i2983 ], [ %retval.sroa.2.0.i.i.i2893, %Flow7953 ]
  %4095 = phi i32 [ %spec.store.select.i.i.i2897, %if.end22.i.i.i2983 ], [ %retval.sroa.6.0.i.i.i2891, %Flow7953 ]
  br label %cleanup26.i.i.i2920, !amdgpu.uniform !42

if.end22.i.i.i2983:                               ; preds = %if.then16.i.i.i2984, %if.end14.i.i.i2979
  call void @llvm.amdgcn.end.cf.i64(i64 %4091)
  call void @llvm.amdgcn.wave.barrier()
  br label %Flow7954, !amdgpu.uniform !42

Flow7955:                                         ; preds = %cleanup26.i.i.i2920, %4069
  %4096 = phi i32 [ %4095, %cleanup26.i.i.i2920 ], [ %4059, %4069 ]
  %4097 = phi i64 [ %4094, %cleanup26.i.i.i2920 ], [ %4060, %4069 ]
  %4098 = phi i32 [ %4093, %cleanup26.i.i.i2920 ], [ %4061, %4069 ]
  %4099 = phi i32 [ %4093, %cleanup26.i.i.i2920 ], [ %retval.sroa.8.0.i.i.i2894, %4069 ]
  %4100 = phi i64 [ %4094, %cleanup26.i.i.i2920 ], [ %retval.sroa.2.0.i.i.i2893, %4069 ]
  %4101 = phi i32 [ %4095, %cleanup26.i.i.i2920 ], [ %retval.sroa.6.0.i.i.i2891, %4069 ]
  %4102 = phi i1 [ %cmp.i46.not.i.i.i2912, %cleanup26.i.i.i2920 ], [ true, %4069 ]
  br i1 %4102, label %for.inc.i.i.i2925, label %Flow7956, !amdgpu.uniform !42

cleanup26.i.i.i2920:                              ; preds = %Flow7954
  br label %Flow7955, !amdgpu.uniform !42

for.inc.i.i.i2925:                                ; preds = %Flow7955
  %inc.i.i.i2930 = add i32 %spec.store.select.i.i.i2897, 1
  br label %Flow7956, !amdgpu.uniform !42

Flow7956:                                         ; preds = %for.inc.i.i.i2925, %Flow7955
  %4103 = phi i32 [ %inc.i.i.i2930, %for.inc.i.i.i2925 ], [ poison, %Flow7955 ]
  %4104 = phi i1 [ false, %for.inc.i.i.i2925 ], [ true, %Flow7955 ]
  br i1 %4104, label %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i2931, label %for.cond.i.i1.i2890, !amdgpu.uniform !42

_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i2931: ; preds = %Flow7956
  %.lcssa9133 = phi i32 [ %4096, %Flow7956 ]
  %.lcssa9132 = phi i64 [ %4097, %Flow7956 ]
  %.lcssa9131 = phi i32 [ %4098, %Flow7956 ]
  fence acquire
  %4105 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %mul.i.i.i.i.i.i2932 = shl i32 %.lcssa9133, 6
  %idxprom.i19.i.i.i.i.i2933 = zext i32 %mul.i.i.i.i.i.i2932 to i64
  %arrayidx.i20.i.i.i.i.i2934 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %4105, i64 %idxprom.i19.i.i.i.i.i2933
  %arrayidx.i22.i.i.i.i.i2935 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %arrayidx.i20.i.i.i.i.i2934, i64 %sh_prom.i.i.i.i756
  store i64 sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.23 to ptr), i64 5) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.23 to ptr) to i64)), ptr %arrayidx.i22.i.i.i.i.i2935, align 8, !tbaa !82
  br i1 %1122, label %loop-memcpy-expansion5373.preheader, label %Flow7952, !amdgpu.uniform !42

loop-memcpy-expansion5373.preheader:              ; preds = %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i2931
  %scevgep6852 = getelementptr i8, ptr %4105, i64 8
  %scevgep6853 = getelementptr i8, ptr %scevgep6852, i64 %1474
  %4106 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i2933, 6
  %scevgep6854 = getelementptr i8, ptr %scevgep6853, i64 %4106
  br label %loop-memcpy-expansion5373, !amdgpu.uniform !42

Flow7952:                                         ; preds = %Flow7951, %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i2931
  br label %loop-memcpy-residual-header5376, !amdgpu.uniform !42

loop-memcpy-expansion5373:                        ; preds = %loop-memcpy-expansion5373.preheader, %loop-memcpy-expansion5373
  %loop-index5374 = phi i64 [ %4109, %loop-memcpy-expansion5373 ], [ 0, %loop-memcpy-expansion5373.preheader ]
  %4107 = getelementptr i8, ptr addrspace(4) @.str.23, i64 %loop-index5374, !amdgpu.uniform !42
  %4108 = load <4 x i32>, ptr addrspace(4) %4107, align 1
  %scevgep6855 = getelementptr i8, ptr %scevgep6854, i64 %loop-index5374
  store <4 x i32> %4108, ptr %scevgep6855, align 1
  %4109 = add i64 %loop-index5374, 16
  %4110 = icmp uge i64 %4109, %1121
  br i1 %4110, label %Flow7951, label %loop-memcpy-expansion5373, !amdgpu.uniform !42

Flow7950:                                         ; preds = %Flow7948, %Flow7949
  %4111 = phi i64 [ %6629, %Flow7948 ], [ %6627, %Flow7949 ]
  br label %post-loop-memcpy-expansion5372, !amdgpu.uniform !42

loop-memcpy-residual5375:                         ; preds = %loop-memcpy-residual5375.preheader, %loop-memcpy-residual5375
  %residual-loop-index5377 = phi i64 [ %4113, %loop-memcpy-residual5375 ], [ 0, %loop-memcpy-residual5375.preheader ]
  %scevgep6860 = getelementptr i8, ptr addrspace(4) @.str.23, i64 %residual-loop-index5377, !amdgpu.uniform !42
  %4112 = load i8, ptr addrspace(4) %scevgep6860, align 1
  %scevgep6859 = getelementptr i8, ptr %scevgep6858, i64 %residual-loop-index5377
  store i8 %4112, ptr %scevgep6859, align 1
  %4113 = add i64 %residual-loop-index5377, 1
  %4114 = icmp uge i64 %4113, %1120
  br i1 %4114, label %Flow7948, label %loop-memcpy-residual5375, !amdgpu.uniform !42

Flow7948:                                         ; preds = %loop-memcpy-residual5375
  br label %Flow7950, !amdgpu.uniform !42

post-loop-memcpy-expansion5372:                   ; preds = %Flow7950
  %tobool.not.i.i.i.i.i.i2938 = icmp eq i32 %.lcssa9131, 0
  %conv.i.i.i.i.i.i2939 = zext i1 %tobool.not.i.i.i.i.i.i2938 to i32
  fence release
  %4115 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %idxprom.i23.i.i.i.i.i2940 = zext i32 %.lcssa9133 to i64
  %arrayidx.i24.i.i.i.i.i2941 = getelementptr inbounds nuw i32, ptr %4115, i64 %idxprom.i23.i.i.i.i.i2940
  store atomic i32 %conv.i.i.i.i.i.i2939, ptr %arrayidx.i24.i.i.i.i.i2941 monotonic, align 4
  %4116 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !84
  %arrayidx.i.i5.i.i2942 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %4116, i64 %idxprom.i23.i.i.i.i.i2940
  %4117 = load i64, ptr %arrayidx.i.i5.i.i2942, align 8, !tbaa !80
  br label %while.cond.i.i.i.i2943, !amdgpu.uniform !42

while.cond.i.i.i.i2943:                           ; preds = %post-loop-memcpy-expansion5372, %Flow7947
  %phi.broken8719 = phi i64 [ 0, %post-loop-memcpy-expansion5372 ], [ %4139, %Flow7947 ]
  %lsr.iv6866 = phi i64 [ %4137, %Flow7947 ], [ -51, %post-loop-memcpy-expansion5372 ]
  %lsr.iv6861 = phi ptr addrspace(4) [ %4136, %Flow7947 ], [ getelementptr (i8, ptr addrspace(4) @.str.23, i64 56), %post-loop-memcpy-expansion5372 ]
  %port.sroa.37.0.i.i2944 = phi i32 [ %4135, %Flow7947 ], [ %conv.i.i.i.i.i.i2939, %post-loop-memcpy-expansion5372 ]
  %idx.0.i.i.i.i2945 = phi i64 [ %4134, %Flow7947 ], [ 56, %post-loop-memcpy-expansion5372 ]
  %cmp.i.i.i.i2946 = icmp ult i64 %idx.0.i.i.i.i2945, sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.23 to ptr), i64 5) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.23 to ptr) to i64))
  %4118 = call i64 @llvm.amdgcn.ballot.i64(i1 %cmp.i.i.i.i2946)
  %and.i.i.i.i6.i.i2947 = and i64 %4118, %4117
  %tobool.not.i.i.i.i2948 = icmp ne i64 %and.i.i.i.i6.i.i2947, 0
  %4119 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i22.i.i.i2949 = getelementptr inbounds nuw i32, ptr %4119, i64 %idxprom.i23.i.i.i.i.i2940
  %4120 = load atomic i32, ptr %arrayidx.i22.i.i.i2949 monotonic, align 4
  %4121 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %tobool.not.i.i.i.i2948)
  %4122 = extractvalue { i1, i64 } %4121, 0
  %4123 = extractvalue { i1, i64 } %4121, 1
  br i1 %4122, label %cond.false.i8.i.i.i.i2950, label %Flow7947

cond.false.i8.i.i.i.i2950:                        ; preds = %while.cond.i.i.i.i2943
  %4124 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %4120)
  %4125 = icmp ne i32 %4124, %port.sroa.37.0.i.i2944
  %4126 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %4125)
  %4127 = extractvalue { i1, i64 } %4126, 0
  %4128 = extractvalue { i1, i64 } %4126, 1
  br i1 %4127, label %while.body.i.i21.i.i.i.i2951.preheader, label %Flow7946

while.body.i.i21.i.i.i.i2951.preheader:           ; preds = %cond.false.i8.i.i.i.i2950
  br label %while.body.i.i21.i.i.i.i2951, !amdgpu.uniform !42

Flow7946:                                         ; preds = %Flow7945, %cond.false.i8.i.i.i.i2950
  call void @llvm.amdgcn.end.cf.i64(i64 %4128)
  br label %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i2954, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i2951:                     ; preds = %while.body.i.i21.i.i.i.i2951.preheader, %while.body.i.i21.i.i.i.i2951
  %phi.broken8718 = phi i64 [ 0, %while.body.i.i21.i.i.i.i2951.preheader ], [ %4132, %while.body.i.i21.i.i.i.i2951 ]
  call void @llvm.amdgcn.s.sleep(i32 2)
  %4129 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i22.i.i.i.i2952 = getelementptr inbounds nuw i32, ptr %4129, i64 %idxprom.i23.i.i.i.i.i2940
  %4130 = load atomic i32, ptr %arrayidx.i.i.i22.i.i.i.i2952 monotonic, align 4
  %4131 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %4130)
  %cmp.i.not.i.i23.i.i.i.i2953 = icmp eq i32 %4131, %port.sroa.37.0.i.i2944
  %4132 = call i64 @llvm.amdgcn.if.break.i64(i1 %cmp.i.not.i.i23.i.i.i.i2953, i64 %phi.broken8718)
  %4133 = call i1 @llvm.amdgcn.loop.i64(i64 %4132)
  br i1 %4133, label %Flow7945, label %while.body.i.i21.i.i.i.i2951

Flow7945:                                         ; preds = %while.body.i.i21.i.i.i.i2951
  %.lcssa9134 = phi i64 [ %4132, %while.body.i.i21.i.i.i.i2951 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa9134)
  br label %Flow7946, !amdgpu.uniform !42

Flow7947:                                         ; preds = %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i2955, %while.cond.i.i.i.i2943
  %4134 = phi i64 [ %add.i.i.i.i2958, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i2955 ], [ poison, %while.cond.i.i.i.i2943 ]
  %4135 = phi i32 [ %conv.i.i26.i.i.i.i2956, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i2955 ], [ poison, %while.cond.i.i.i.i2943 ]
  %4136 = phi ptr addrspace(4) [ %scevgep6862, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i2955 ], [ poison, %while.cond.i.i.i.i2943 ]
  %4137 = phi i64 [ %lsr.iv.next6867, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i2955 ], [ poison, %while.cond.i.i.i.i2943 ]
  %4138 = phi i1 [ false, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i2955 ], [ true, %while.cond.i.i.i.i2943 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %4123)
  %4139 = call i64 @llvm.amdgcn.if.break.i64(i1 %4138, i64 %phi.broken8719)
  %4140 = call i1 @llvm.amdgcn.loop.i64(i64 %4139)
  br i1 %4140, label %cond.false.i.i.i2965, label %while.cond.i.i.i.i2943

_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i2954: ; preds = %Flow7946
  fence acquire
  br i1 %cmp.i.i.i.i2946, label %if.then.i.i.i.i.i.i.i2959, label %Flow7944, !amdgpu.uniform !42

if.then.i.i.i.i.i.i.i2959:                        ; preds = %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i2954
  %sub.i.i.i.i.i.i.i2961 = sub nuw i64 sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.23 to ptr), i64 5) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.23 to ptr) to i64)), %idx.0.i.i.i.i2945
  %spec.select.i.i.i36.i.i.i.i2962 = call i64 @llvm.umin.i64(i64 %sub.i.i.i.i.i.i.i2961, i64 64)
  %4141 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %4142 = and i64 %spec.select.i.i.i36.i.i.i.i2962, 15
  %4143 = sub i64 %spec.select.i.i.i36.i.i.i.i2962, %4142
  %4144 = icmp ne i64 %4143, 0
  br i1 %4144, label %loop-memcpy-expansion5379.preheader, label %Flow7943, !amdgpu.uniform !42

loop-memcpy-expansion5379.preheader:              ; preds = %if.then.i.i.i.i.i.i.i2959
  %arrayidx.i20.i35.i.i.i.i2963 = getelementptr %"struct.rpc::Buffer", ptr %4141, i64 %idxprom.i19.i.i.i.i.i2933
  %arrayidx.i22.i38.i.i.i.i2964 = getelementptr %"struct.rpc::Buffer", ptr %arrayidx.i20.i35.i.i.i.i2963, i64 %sh_prom.i.i.i.i756
  br label %loop-memcpy-expansion5379, !amdgpu.uniform !42

Flow7943:                                         ; preds = %Flow7942, %if.then.i.i.i.i.i.i.i2959
  br label %loop-memcpy-residual-header5382, !amdgpu.uniform !42

loop-memcpy-expansion5379:                        ; preds = %loop-memcpy-expansion5379.preheader, %loop-memcpy-expansion5379
  %loop-index5380 = phi i64 [ %4147, %loop-memcpy-expansion5379 ], [ 0, %loop-memcpy-expansion5379.preheader ]
  %scevgep6863 = getelementptr i8, ptr addrspace(4) %lsr.iv6861, i64 %loop-index5380, !amdgpu.uniform !42
  %4145 = load <4 x i32>, ptr addrspace(4) %scevgep6863, align 1
  %4146 = getelementptr i8, ptr %arrayidx.i22.i38.i.i.i.i2964, i64 %loop-index5380
  store <4 x i32> %4145, ptr %4146, align 1
  %4147 = add i64 %loop-index5380, 16
  %4148 = icmp uge i64 %4147, %4143
  br i1 %4148, label %Flow7942, label %loop-memcpy-expansion5379, !amdgpu.uniform !42

Flow7941:                                         ; preds = %Flow7940, %loop-memcpy-residual-header5382
  br label %Flow7944, !amdgpu.uniform !42

loop-memcpy-residual5381:                         ; preds = %loop-memcpy-residual5381.preheader, %loop-memcpy-residual5381
  %residual-loop-index5383 = phi i64 [ %4150, %loop-memcpy-residual5381 ], [ 0, %loop-memcpy-residual5381.preheader ]
  %scevgep6872 = getelementptr i8, ptr addrspace(4) %scevgep6871, i64 %residual-loop-index5383, !amdgpu.uniform !42
  %4149 = load i8, ptr addrspace(4) %scevgep6872, align 1
  %scevgep6870 = getelementptr i8, ptr %scevgep6869, i64 %residual-loop-index5383
  store i8 %4149, ptr %scevgep6870, align 1
  %4150 = add i64 %residual-loop-index5383, 1
  %4151 = icmp uge i64 %4150, %4142
  br i1 %4151, label %Flow7940, label %loop-memcpy-residual5381, !amdgpu.uniform !42

Flow7940:                                         ; preds = %loop-memcpy-residual5381
  br label %Flow7941, !amdgpu.uniform !42

_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i2955: ; preds = %Flow7944
  %conv.i.i26.i.i.i.i2956 = xor i32 %port.sroa.37.0.i.i2944, 1
  fence release
  %4152 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %arrayidx.i24.i29.i.i.i.i2957 = getelementptr inbounds nuw i32, ptr %4152, i64 %idxprom.i23.i.i.i.i.i2940
  store atomic i32 %conv.i.i26.i.i.i.i2956, ptr %arrayidx.i24.i29.i.i.i.i2957 monotonic, align 4
  %add.i.i.i.i2958 = add i64 %idx.0.i.i.i.i2945, 64
  %scevgep6862 = getelementptr i8, ptr addrspace(4) %lsr.iv6861, i64 64
  %lsr.iv.next6867 = add i64 %lsr.iv6866, -64
  br label %Flow7947, !amdgpu.uniform !42

cond.false.i.i.i2965:                             ; preds = %Flow7947
  %.lcssa9136 = phi i64 [ %4139, %Flow7947 ]
  %port.sroa.37.0.i.i2944.lcssa = phi i32 [ %port.sroa.37.0.i.i2944, %Flow7947 ]
  %.lcssa9135 = phi i32 [ %4120, %Flow7947 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa9136)
  %4153 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %.lcssa9135)
  %4154 = icmp ne i32 %4153, %port.sroa.37.0.i.i2944.lcssa
  %4155 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %4154)
  %4156 = extractvalue { i1, i64 } %4155, 0
  %4157 = extractvalue { i1, i64 } %4155, 1
  br i1 %4156, label %while.body.i.i.i.i2966.preheader, label %Flow7939

while.body.i.i.i.i2966.preheader:                 ; preds = %cond.false.i.i.i2965
  br label %while.body.i.i.i.i2966, !amdgpu.uniform !42

Flow7939:                                         ; preds = %Flow7938, %cond.false.i.i.i2965
  call void @llvm.amdgcn.end.cf.i64(i64 %4157)
  br label %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit2987, !amdgpu.uniform !42

while.body.i.i.i.i2966:                           ; preds = %while.body.i.i.i.i2966.preheader, %while.body.i.i.i.i2966
  %phi.broken8720 = phi i64 [ 0, %while.body.i.i.i.i2966.preheader ], [ %4161, %while.body.i.i.i.i2966 ]
  call void @llvm.amdgcn.s.sleep(i32 2)
  %4158 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i8.i.i2967 = getelementptr inbounds nuw i32, ptr %4158, i64 %idxprom.i23.i.i.i.i.i2940
  %4159 = load atomic i32, ptr %arrayidx.i.i.i8.i.i2967 monotonic, align 4
  %4160 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %4159)
  %cmp.i.not.i.i.i.i2968 = icmp eq i32 %4160, %port.sroa.37.0.i.i2944.lcssa
  %4161 = call i64 @llvm.amdgcn.if.break.i64(i1 %cmp.i.not.i.i.i.i2968, i64 %phi.broken8720)
  %4162 = call i1 @llvm.amdgcn.loop.i64(i64 %4161)
  br i1 %4162, label %Flow7938, label %while.body.i.i.i.i2966

Flow7938:                                         ; preds = %while.body.i.i.i.i2966
  %.lcssa9137 = phi i64 [ %4161, %while.body.i.i.i.i2966 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa9137)
  br label %Flow7939, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit2987: ; preds = %Flow7939
  fence acquire
  call void @llvm.amdgcn.wave.barrier()
  fence syncscope("agent") release
  %4163 = call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %.lcssa9132, i1 true)
  %iszero.i.i.i.i.i13.i.i2970 = icmp ne i64 %.lcssa9132, 0
  %cmp2.i.i.i.i14.i.i2971 = icmp eq i64 %4163, %sh_prom.i.i.i.i756
  %cmp.i.i.i.i15.i.i2972 = select i1 %iszero.i.i.i.i.i13.i.i2970, i1 %cmp2.i.i.i.i14.i.i2971, i1 false
  %4164 = and i32 %.lcssa9133, 31
  %conv4.i.i.i19.i.i2976 = zext i1 %cmp.i.i.i.i15.i.i2972 to i32
  %shl.i.i.i20.i.i2977 = shl nuw i32 %conv4.i.i.i19.i.i2976, %4164
  %xor.i.i.i21.i.i2978 = xor i32 %shl.i.i.i20.i.i2977, -1
  %4165 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop5823, !amdgpu.uniform !42

4166:                                             ; preds = %ComputeEnd5824
  %div8.i.i.i16.i.i2973 = lshr i32 %.lcssa9133, 5
  %idxprom.i.i.i17.i.i2974 = zext nneg i32 %div8.i.i.i16.i.i2973 to i64
  %4167 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i17.i.i2974
  %sunkaddr7458 = getelementptr inbounds i8, ptr addrspace(1) %4167, i64 40
  %4168 = atomicrmw and ptr addrspace(1) %sunkaddr7458, i32 %.lcssa9138 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %4169, !amdgpu.uniform !42

4169:                                             ; preds = %4166, %ComputeEnd5824
  call void @llvm.amdgcn.end.cf.i64(i64 %8200)
  call void @llvm.amdgcn.wave.barrier()
  br label %Flow8062, !amdgpu.uniform !42

if.else82:                                        ; preds = %Flow8062
  %mul70 = mul i64 %sub, 1000000000
  call void @llvm.lifetime.start.p5(ptr addrspace(5) %buffer.i.i.i2988) #34, !noalias !175
  %cmp7.i.i.i.i.i.i2989 = icmp uge i64 %mul70, 1000000
  br i1 %cmp7.i.i.i.i.i.i2989, label %while.body.i.i.i.i.i.i2991.preheader, label %Flow8061, !amdgpu.uniform !42

while.body.i.i.i.i.i.i2991.preheader:             ; preds = %if.else82
  %div71 = udiv i64 %mul70, 1000000
  br label %while.body.i.i.i.i.i.i2991, !amdgpu.uniform !42

Flow8061:                                         ; preds = %Flow8060, %if.else82
  %4170 = phi i1 [ %cmp.i.i.i.i.i.i.i.i3001.lcssa, %Flow8060 ], [ false, %if.else82 ]
  %4171 = phi i64 [ %writer.sroa.8.1.i.i.i.i3006.lcssa, %Flow8060 ], [ 0, %if.else82 ]
  br label %_ZN22__llvm_libc_22_0_0_git15IntegerToStringIyNS_7details3FmtILh10ELb0ELb0ELb0ELm1EEEE13IntegerWriter25write_unsigned_number_decEyRNS1_22StringBufferWriterImplILb0EEE.exit.i.i.i.i.i3009, !amdgpu.uniform !42

while.body.i.i.i.i.i.i2991:                       ; preds = %while.body.i.i.i.i.i.i2991.preheader, %_ZN22__llvm_libc_22_0_0_git7details22StringBufferWriterImplILb0EE4pushEc.exit.i.i.i.i.i.i3005
  %writer.sroa.8.0.i.i.i.i2992 = phi i64 [ %writer.sroa.8.1.i.i.i.i3006, %_ZN22__llvm_libc_22_0_0_git7details22StringBufferWriterImplILb0EE4pushEc.exit.i.i.i.i.i.i3005 ], [ 0, %while.body.i.i.i.i.i.i2991.preheader ]
  %value.addr.09.i.i.i.i.i.i2993 = phi i64 [ %div.i.i.i.i.i.i.i2994, %_ZN22__llvm_libc_22_0_0_git7details22StringBufferWriterImplILb0EE4pushEc.exit.i.i.i.i.i.i3005 ], [ %div71, %while.body.i.i.i.i.i.i2991.preheader ]
  br label %NodeBlock7577, !amdgpu.uniform !42

NodeBlock7577:                                    ; preds = %while.body.i.i.i.i.i.i2991
  %div.i.i.i.i.i.i.i2994 = udiv i64 %value.addr.09.i.i.i.i.i.i2993, 10
  %.neg.i2995 = mul nuw nsw i64 %div.i.i.i.i.i.i.i2994, 246
  %rem.i.i.i.i.i.i.decomposed.i2996 = add nuw nsw i64 %.neg.i2995, %value.addr.09.i.i.i.i.i.i2993
  %conv.i.i.i.i.i.i.i2997 = trunc i64 %rem.i.i.i.i.i.i.decomposed.i2996 to i8
  %4172 = zext i8 %conv.i.i.i.i.i.i.i2997 to i16
  %Pivot7578 = icmp sge i16 %4172, 5
  br i1 %Pivot7578, label %NodeBlock7575, label %Flow8059, !amdgpu.uniform !42

NodeBlock7575:                                    ; preds = %NodeBlock7577
  %Pivot7576 = icmp sge i16 %4172, 7
  br i1 %Pivot7576, label %NodeBlock7573, label %Flow8050, !amdgpu.uniform !42

NodeBlock7573:                                    ; preds = %NodeBlock7575
  %Pivot7574 = icmp sge i16 %4172, 8
  br i1 %Pivot7574, label %NodeBlock7571, label %Flow8046, !amdgpu.uniform !42

NodeBlock7571:                                    ; preds = %NodeBlock7573
  %Pivot7572 = icmp sge i16 %4172, 9
  br i1 %Pivot7572, label %sw.bb9.i.i.i.i.i.i.i.i2998, label %Flow8044, !amdgpu.uniform !42

Flow8050:                                         ; preds = %Flow8047, %NodeBlock7575
  %4173 = phi i8 [ %4186, %Flow8047 ], [ poison, %NodeBlock7575 ]
  %4174 = phi i1 [ false, %Flow8047 ], [ true, %NodeBlock7575 ]
  br i1 %4174, label %NodeBlock7569, label %Flow8051, !amdgpu.uniform !42

NodeBlock7569:                                    ; preds = %Flow8050
  %Pivot7570 = icmp sge i16 %4172, 6
  br i1 %Pivot7570, label %sw.bb6.i.i.i.i.i.i.i.i3149, label %Flow8048, !amdgpu.uniform !42

Flow8059:                                         ; preds = %Flow8051, %NodeBlock7577
  %4175 = phi i8 [ %4188, %Flow8051 ], [ poison, %NodeBlock7577 ]
  %4176 = phi i1 [ false, %Flow8051 ], [ true, %NodeBlock7577 ]
  br i1 %4176, label %NodeBlock7567, label %if.then.i.i.i.i.i.i.i2999, !amdgpu.uniform !42

NodeBlock7567:                                    ; preds = %Flow8059
  %Pivot7568 = icmp sge i16 %4172, 3
  br i1 %Pivot7568, label %NodeBlock7565, label %Flow8057, !amdgpu.uniform !42

NodeBlock7565:                                    ; preds = %NodeBlock7567
  %Pivot7566 = icmp sge i16 %4172, 4
  br i1 %Pivot7566, label %sw.bb4.i.i.i.i.i.i.i.i3151, label %Flow8052, !amdgpu.uniform !42

Flow8057:                                         ; preds = %Flow8053, %NodeBlock7567
  %4177 = phi i8 [ %4189, %Flow8053 ], [ poison, %NodeBlock7567 ]
  %4178 = phi i1 [ false, %Flow8053 ], [ true, %NodeBlock7567 ]
  br i1 %4178, label %NodeBlock7563, label %Flow8058, !amdgpu.uniform !42

NodeBlock7563:                                    ; preds = %Flow8057
  %Pivot7564 = icmp sge i16 %4172, 2
  br i1 %Pivot7564, label %sw.bb2.i.i.i.i.i.i.i.i3153, label %Flow8055, !amdgpu.uniform !42

Flow8055:                                         ; preds = %sw.bb2.i.i.i.i.i.i.i.i3153, %NodeBlock7563
  %4179 = phi i1 [ false, %sw.bb2.i.i.i.i.i.i.i.i3153 ], [ true, %NodeBlock7563 ]
  br i1 %4179, label %LeafBlock7561, label %Flow8056, !amdgpu.uniform !42

LeafBlock7561:                                    ; preds = %Flow8055
  %SwitchLeaf7562 = icmp eq i16 %4172, 1
  br i1 %SwitchLeaf7562, label %sw.bb1.i.i.i.i.i.i.i.i3154, label %Flow8054, !amdgpu.uniform !42

sw.bb1.i.i.i.i.i.i.i.i3154:                       ; preds = %LeafBlock7561
  br label %Flow8054, !amdgpu.uniform !42

sw.bb2.i.i.i.i.i.i.i.i3153:                       ; preds = %NodeBlock7563
  br label %Flow8055, !amdgpu.uniform !42

Flow8052:                                         ; preds = %sw.bb4.i.i.i.i.i.i.i.i3151, %NodeBlock7565
  %4180 = phi i1 [ false, %sw.bb4.i.i.i.i.i.i.i.i3151 ], [ true, %NodeBlock7565 ]
  br i1 %4180, label %sw.bb3.i.i.i.i.i.i.i.i3152, label %Flow8053, !amdgpu.uniform !42

sw.bb3.i.i.i.i.i.i.i.i3152:                       ; preds = %Flow8052
  br label %Flow8053, !amdgpu.uniform !42

sw.bb4.i.i.i.i.i.i.i.i3151:                       ; preds = %NodeBlock7565
  br label %Flow8052, !amdgpu.uniform !42

Flow8048:                                         ; preds = %sw.bb6.i.i.i.i.i.i.i.i3149, %NodeBlock7569
  %4181 = phi i1 [ false, %sw.bb6.i.i.i.i.i.i.i.i3149 ], [ true, %NodeBlock7569 ]
  br i1 %4181, label %sw.bb5.i.i.i.i.i.i.i.i3150, label %Flow8049, !amdgpu.uniform !42

sw.bb5.i.i.i.i.i.i.i.i3150:                       ; preds = %Flow8048
  br label %Flow8049, !amdgpu.uniform !42

sw.bb6.i.i.i.i.i.i.i.i3149:                       ; preds = %NodeBlock7569
  br label %Flow8048, !amdgpu.uniform !42

Flow8046:                                         ; preds = %Flow8045, %NodeBlock7573
  %4182 = phi i8 [ %4185, %Flow8045 ], [ poison, %NodeBlock7573 ]
  %4183 = phi i1 [ false, %Flow8045 ], [ true, %NodeBlock7573 ]
  br i1 %4183, label %sw.bb7.i.i.i.i.i.i.i.i3148, label %Flow8047, !amdgpu.uniform !42

sw.bb7.i.i.i.i.i.i.i.i3148:                       ; preds = %Flow8046
  br label %Flow8047, !amdgpu.uniform !42

Flow8044:                                         ; preds = %sw.bb9.i.i.i.i.i.i.i.i2998, %NodeBlock7571
  %4184 = phi i1 [ false, %sw.bb9.i.i.i.i.i.i.i.i2998 ], [ true, %NodeBlock7571 ]
  br i1 %4184, label %sw.bb8.i.i.i.i.i.i.i.i3147, label %Flow8045, !amdgpu.uniform !42

sw.bb8.i.i.i.i.i.i.i.i3147:                       ; preds = %Flow8044
  br label %Flow8045, !amdgpu.uniform !42

sw.bb9.i.i.i.i.i.i.i.i2998:                       ; preds = %NodeBlock7571
  br label %Flow8044, !amdgpu.uniform !42

Flow8045:                                         ; preds = %sw.bb8.i.i.i.i.i.i.i.i3147, %Flow8044
  %4185 = phi i8 [ 56, %sw.bb8.i.i.i.i.i.i.i.i3147 ], [ 57, %Flow8044 ]
  br label %Flow8046, !amdgpu.uniform !42

Flow8047:                                         ; preds = %sw.bb7.i.i.i.i.i.i.i.i3148, %Flow8046
  %4186 = phi i8 [ 55, %sw.bb7.i.i.i.i.i.i.i.i3148 ], [ %4182, %Flow8046 ]
  br label %Flow8050, !amdgpu.uniform !42

Flow8049:                                         ; preds = %sw.bb5.i.i.i.i.i.i.i.i3150, %Flow8048
  %4187 = phi i8 [ 53, %sw.bb5.i.i.i.i.i.i.i.i3150 ], [ 54, %Flow8048 ]
  br label %Flow8051, !amdgpu.uniform !42

Flow8051:                                         ; preds = %Flow8049, %Flow8050
  %4188 = phi i8 [ %4187, %Flow8049 ], [ %4173, %Flow8050 ]
  br label %Flow8059, !amdgpu.uniform !42

Flow8053:                                         ; preds = %sw.bb3.i.i.i.i.i.i.i.i3152, %Flow8052
  %4189 = phi i8 [ 51, %sw.bb3.i.i.i.i.i.i.i.i3152 ], [ 52, %Flow8052 ]
  br label %Flow8057, !amdgpu.uniform !42

Flow8054:                                         ; preds = %sw.bb1.i.i.i.i.i.i.i.i3154, %LeafBlock7561
  %4190 = phi i8 [ 49, %sw.bb1.i.i.i.i.i.i.i.i3154 ], [ 48, %LeafBlock7561 ]
  br label %Flow8056, !amdgpu.uniform !42

Flow8056:                                         ; preds = %Flow8054, %Flow8055
  %4191 = phi i8 [ %4190, %Flow8054 ], [ 50, %Flow8055 ]
  br label %Flow8058, !amdgpu.uniform !42

Flow8058:                                         ; preds = %Flow8056, %Flow8057
  %4192 = phi i8 [ %4191, %Flow8056 ], [ %4177, %Flow8057 ]
  br label %if.then.i.i.i.i.i.i.i2999, !amdgpu.uniform !42

if.then.i.i.i.i.i.i.i2999:                        ; preds = %Flow8059, %Flow8058
  %retval.0.i.i.i.i.i.i.i.i3000 = phi i8 [ %4175, %Flow8059 ], [ %4192, %Flow8058 ]
  %cmp.i.i.i.i.i.i.i.i3001 = icmp eq i64 %writer.sroa.8.0.i.i.i.i2992, 21
  %cmp.i.i.i.i.i.i.i.i3001.inv = xor i1 %cmp.i.i.i.i.i.i.i.i3001, true
  br i1 %cmp.i.i.i.i.i.i.i.i3001.inv, label %if.then3.i.i.i.i.i.i.i3002, label %_ZN22__llvm_libc_22_0_0_git7details22StringBufferWriterImplILb0EE4pushEc.exit.i.i.i.i.i.i3005, !amdgpu.uniform !42

if.then3.i.i.i.i.i.i.i3002:                       ; preds = %if.then.i.i.i.i.i.i.i2999
  %4193 = trunc i64 %writer.sroa.8.0.i.i.i.i2992 to i32
  %4194 = xor i32 %4193, -1
  %4195 = getelementptr i8, ptr addrspace(5) %buffer.i.i.i2988, i32 %4194
  %arrayidx.i.i.i.i.i.i.i.i30036057 = getelementptr i8, ptr addrspace(5) %4195, i32 21
  %inc.i.i.i.i.i.i.i3004 = add i64 %writer.sroa.8.0.i.i.i.i2992, 1
  store i8 %retval.0.i.i.i.i.i.i.i.i3000, ptr addrspace(5) %arrayidx.i.i.i.i.i.i.i.i30036057, align 1, !tbaa !57, !noalias !175
  br label %_ZN22__llvm_libc_22_0_0_git7details22StringBufferWriterImplILb0EE4pushEc.exit.i.i.i.i.i.i3005, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git7details22StringBufferWriterImplILb0EE4pushEc.exit.i.i.i.i.i.i3005: ; preds = %if.then3.i.i.i.i.i.i.i3002, %if.then.i.i.i.i.i.i.i2999
  %writer.sroa.8.1.i.i.i.i3006 = phi i64 [ %inc.i.i.i.i.i.i.i3004, %if.then3.i.i.i.i.i.i.i3002 ], [ 21, %if.then.i.i.i.i.i.i.i2999 ]
  %cmp.i.i.i.i.i.i3007 = icmp samesign ult i64 %value.addr.09.i.i.i.i.i.i2993, 10
  %.not.i.i.i.i.i.i3008 = or i1 %cmp.i.i.i.i.i.i.i.i3001, %cmp.i.i.i.i.i.i3007
  br i1 %.not.i.i.i.i.i.i3008, label %Flow8060, label %while.body.i.i.i.i.i.i2991, !amdgpu.uniform !42

Flow8060:                                         ; preds = %_ZN22__llvm_libc_22_0_0_git7details22StringBufferWriterImplILb0EE4pushEc.exit.i.i.i.i.i.i3005
  %writer.sroa.8.1.i.i.i.i3006.lcssa = phi i64 [ %writer.sroa.8.1.i.i.i.i3006, %_ZN22__llvm_libc_22_0_0_git7details22StringBufferWriterImplILb0EE4pushEc.exit.i.i.i.i.i.i3005 ]
  %cmp.i.i.i.i.i.i.i.i3001.lcssa = phi i1 [ %cmp.i.i.i.i.i.i.i.i3001, %_ZN22__llvm_libc_22_0_0_git7details22StringBufferWriterImplILb0EE4pushEc.exit.i.i.i.i.i.i3005 ]
  br label %Flow8061, !amdgpu.uniform !42

Flow8063:                                         ; preds = %4453, %Flow8062
  %4196 = phi i32 [ %1215, %4453 ], [ %3893, %Flow8062 ]
  %4197 = phi i64 [ %1214, %4453 ], [ %3894, %Flow8062 ]
  %4198 = phi i32 [ %1213, %4453 ], [ %3895, %Flow8062 ]
  %4199 = phi i32 [ %1212, %4453 ], [ %3896, %Flow8062 ]
  %4200 = phi i64 [ %1211, %4453 ], [ %3897, %Flow8062 ]
  %4201 = phi i32 [ %1210, %4453 ], [ %3898, %Flow8062 ]
  %4202 = phi i32 [ %.lcssa9153, %4453 ], [ %1209, %Flow8062 ]
  %4203 = phi i64 [ %.lcssa9154, %4453 ], [ %1208, %Flow8062 ]
  %4204 = phi i32 [ %.lcssa9155, %4453 ], [ %1207, %Flow8062 ]
  %4205 = phi i32 [ %.lcssa9142, %4453 ], [ %1206, %Flow8062 ]
  %4206 = phi i64 [ %.lcssa9143, %4453 ], [ %1205, %Flow8062 ]
  %4207 = phi i32 [ %.lcssa9144, %4453 ], [ %1204, %Flow8062 ]
  br label %Flow8127, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git15IntegerToStringIyNS_7details3FmtILh10ELb0ELb0ELb0ELm1EEEE13IntegerWriter25write_unsigned_number_decEyRNS1_22StringBufferWriterImplILb0EEE.exit.i.i.i.i.i3009: ; preds = %Flow8061
  %cmp.i15.i.i.i.i3012 = icmp ne i64 %4171, 0
  %or.cond.not.i.i.i.i3013 = select i1 %4170, i1 true, i1 %cmp.i15.i.i.i.i3012
  %or.cond.not.i.i.i.i3013.inv = xor i1 %or.cond.not.i.i.i.i3013, true
  br i1 %or.cond.not.i.i.i.i3013.inv, label %if.then3.i.i.lr.ph.i.i.i.i3014, label %_ZN22__llvm_libc_22_0_0_git15IntegerToStringIyNS_7details3FmtILh10ELb0ELb0ELb0ELm1EEEEC2Ey.exit.i.i.i3016, !amdgpu.uniform !42

if.then3.i.i.lr.ph.i.i.i.i3014:                   ; preds = %_ZN22__llvm_libc_22_0_0_git15IntegerToStringIyNS_7details3FmtILh10ELb0ELb0ELb0ELm1EEEE13IntegerWriter25write_unsigned_number_decEyRNS1_22StringBufferWriterImplILb0EEE.exit.i.i.i.i.i3009
  %sunkaddr7459 = getelementptr inbounds i8, ptr addrspace(5) %buffer.i.i.i2988, i32 20
  store i8 48, ptr addrspace(5) %sunkaddr7459, align 4, !tbaa !57, !noalias !175
  br label %_ZN22__llvm_libc_22_0_0_git15IntegerToStringIyNS_7details3FmtILh10ELb0ELb0ELb0ELm1EEEEC2Ey.exit.i.i.i3016, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git15IntegerToStringIyNS_7details3FmtILh10ELb0ELb0ELb0ELm1EEEEC2Ey.exit.i.i.i3016: ; preds = %if.then3.i.i.lr.ph.i.i.i.i3014, %_ZN22__llvm_libc_22_0_0_git15IntegerToStringIyNS_7details3FmtILh10ELb0ELb0ELb0ELm1EEEE13IntegerWriter25write_unsigned_number_decEyRNS1_22StringBufferWriterImplILb0EEE.exit.i.i.i.i.i3009
  %writer.sroa.8.4.i.i.i.i3017 = phi i64 [ 1, %if.then3.i.i.lr.ph.i.i.i.i3014 ], [ %4171, %_ZN22__llvm_libc_22_0_0_git15IntegerToStringIyNS_7details3FmtILh10ELb0ELb0ELb0ELm1EEEE13IntegerWriter25write_unsigned_number_decEyRNS1_22StringBufferWriterImplILb0EEE.exit.i.i.i.i.i3009 ]
  %sunkaddr7460 = getelementptr inbounds i8, ptr addrspace(5) %buffer.i.i.i2988, i32 24
  store i64 %writer.sroa.8.4.i.i.i.i3017, ptr addrspace(5) %sunkaddr7460, align 8, !tbaa !167, !noalias !175
  %cmp.i.not.i.i.i.i.i3019 = icmp ne i64 %writer.sroa.8.4.i.i.i.i3017, 0
  br i1 %cmp.i.not.i.i.i.i.i3019, label %if.then.i.i.i.i.i.i3020, label %Flow8043, !amdgpu.uniform !42

if.then.i.i.i.i.i.i3020:                          ; preds = %_ZN22__llvm_libc_22_0_0_git15IntegerToStringIyNS_7details3FmtILh10ELb0ELb0ELb0ELm1EEEEC2Ey.exit.i.i.i3016
  %4208 = call { i64, i1 } @llvm.uadd.with.overflow.i64(i64 %writer.sroa.8.4.i.i.i.i3017, i64 1)
  %ov7462 = extractvalue { i64, i1 } %4208, 1
  %ov7462.inv = xor i1 %ov7462, true
  br i1 %ov7462.inv, label %if.end.i.i.i.i.i.i.i3023, label %Flow8042, !amdgpu.uniform !42

if.end.i.i.i.i.i.i.i3023:                         ; preds = %if.then.i.i.i.i.i.i3020
  %math7461 = extractvalue { i64, i1 } %4208, 0
  %cmp2.i.i.i.i.i.i.i3024 = icmp ult i64 %math7461, 1676976733973595601
  br i1 %cmp2.i.i.i.i.i.i.i3024, label %if.then3.i.i.i.i4.i.i.i3144, label %if.end4.i.i.i.i.i.i.i3025, !amdgpu.uniform !42

if.then3.i.i.i.i4.i.i.i3144:                      ; preds = %if.end.i.i.i.i.i.i.i3023
  %mul.i.i.i.i.i.i.i3145 = mul nuw i64 %math7461, 11
  %div18.i.i.i.i.i.i.i3146 = lshr i64 %mul.i.i.i.i.i.i.i3145, 3
  br label %if.end4.i.i.i.i.i.i.i3025, !amdgpu.uniform !42

Flow8042:                                         ; preds = %if.end4.i.i.i.i.i.i.i3025, %if.then.i.i.i.i.i.i3020
  %4209 = phi ptr [ %cond.i.i.i3033, %if.end4.i.i.i.i.i.i.i3025 ], [ addrspacecast (ptr addrspace(4) @_ZN22__llvm_libc_22_0_0_git3cpp6string14NULL_CHARACTERE to ptr), %if.then.i.i.i.i.i.i3020 ]
  br label %_ZN22__llvm_libc_22_0_0_git3cpp6string7reserveEm.exit.i.i.i.i.i.i3035, !amdgpu.uniform !42

if.end4.i.i.i.i.i.i.i3025:                        ; preds = %if.then3.i.i.i.i4.i.i.i3144, %if.end.i.i.i.i.i.i.i3023
  %new_capacity.addr.0.i.i.i.i.i.i.i3026 = phi i64 [ %div18.i.i.i.i.i.i.i3146, %if.then3.i.i.i.i4.i.i.i3144 ], [ %math7461, %if.end.i.i.i.i.i.i.i3023 ]
  %sub.i.i.i3027 = add i64 %new_capacity.addr.0.i.i.i.i.i.i.i3026, 7
  %div2.i.i.i3028 = and i64 %sub.i.i.i3027, -8
  %4210 = load ptr, ptr addrspace(1) @_ZN12_GLOBAL__N_13ptrE, align 8, !tbaa !55
  %add.ptr.i.i.i3029 = getelementptr inbounds nuw i8, ptr %4210, i64 %div2.i.i.i3028
  store ptr %add.ptr.i.i.i3029, ptr addrspace(1) @_ZN12_GLOBAL__N_13ptrE, align 8, !tbaa !55
  %sub.ptr.lhs.cast.i.i.i3030 = ptrtoint ptr %add.ptr.i.i.i3029 to i64
  %sub.ptr.sub.i.i.i3031 = sub i64 %sub.ptr.lhs.cast.i.i.i3030, ptrtoint (ptr addrspacecast (ptr addrspace(1) @_ZN12_GLOBAL__N_16memoryE to ptr) to i64)
  %cmp.i.i.i3032 = icmp ugt i64 %sub.ptr.sub.i.i.i3031, 65335
  %cond.i.i.i3033 = select i1 %cmp.i.i.i3032, ptr null, ptr %4210
  br label %Flow8042, !amdgpu.uniform !42

Flow8043:                                         ; preds = %Flow8041, %_ZN22__llvm_libc_22_0_0_git15IntegerToStringIyNS_7details3FmtILh10ELb0ELb0ELb0ELm1EEEEC2Ey.exit.i.i.i3016
  %4211 = phi ptr [ %4209, %Flow8041 ], [ addrspacecast (ptr addrspace(4) @_ZN22__llvm_libc_22_0_0_git3cpp6string14NULL_CHARACTERE to ptr), %_ZN22__llvm_libc_22_0_0_git15IntegerToStringIyNS_7details3FmtILh10ELb0ELb0ELb0ELm1EEEEC2Ey.exit.i.i.i3016 ]
  br label %if.end.i.i.i.i.i.i3037, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git3cpp6string7reserveEm.exit.i.i.i.i.i.i3035: ; preds = %Flow8042
  br label %loadstoreloop4953, !amdgpu.uniform !42

loadstoreloop4953:                                ; preds = %_ZN22__llvm_libc_22_0_0_git3cpp6string7reserveEm.exit.i.i.i.i.i.i3035, %loadstoreloop4953
  %4212 = phi i64 [ %4214, %loadstoreloop4953 ], [ 0, %_ZN22__llvm_libc_22_0_0_git3cpp6string7reserveEm.exit.i.i.i.i.i.i3035 ]
  %4213 = getelementptr i8, ptr %4209, i64 %4212
  store i8 0, ptr %4213, align 1
  %4214 = add i64 %4212, 1
  %4215 = icmp uge i64 %4214, %writer.sroa.8.4.i.i.i.i3017
  br i1 %4215, label %Flow8041, label %loadstoreloop4953, !amdgpu.uniform !42

Flow8041:                                         ; preds = %loadstoreloop4953
  br label %Flow8043, !amdgpu.uniform !42

if.end.i.i.i.i.i.i3037:                           ; preds = %Flow8043
  %cmp.not.i6.i.i.i.i.i.i3039 = icmp ne ptr %4211, addrspacecast (ptr addrspace(4) @_ZN22__llvm_libc_22_0_0_git3cpp6string14NULL_CHARACTERE to ptr)
  br i1 %cmp.not.i6.i.i.i.i.i.i3039, label %if.then.i.i.i.i2.i.i.i3040, label %_ZN22__llvm_libc_22_0_0_git3cpp9to_stringEy.exit.i3042, !amdgpu.uniform !42

if.then.i.i.i.i2.i.i.i3040:                       ; preds = %if.end.i.i.i.i.i.i3037
  %arrayidx.i.i.i.i3.i.i.i3041 = getelementptr inbounds nuw i8, ptr %4211, i64 %writer.sroa.8.4.i.i.i.i3017
  store i8 0, ptr %arrayidx.i.i.i.i3.i.i.i3041, align 1, !tbaa !57, !noalias !175
  br label %_ZN22__llvm_libc_22_0_0_git3cpp9to_stringEy.exit.i3042, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git3cpp9to_stringEy.exit.i3042: ; preds = %if.then.i.i.i.i2.i.i.i3040, %if.end.i.i.i.i.i.i3037
  %4216 = trunc i64 %writer.sroa.8.4.i.i.i.i3017 to i32
  %4217 = and i64 %writer.sroa.8.4.i.i.i.i3017, 15
  %4218 = sub i64 %writer.sroa.8.4.i.i.i.i3017, %4217
  %4219 = icmp ne i64 %4218, 0
  br i1 %4219, label %loop-memcpy-expansion5121.preheader, label %Flow8040, !amdgpu.uniform !42

loop-memcpy-expansion5121.preheader:              ; preds = %_ZN22__llvm_libc_22_0_0_git3cpp9to_stringEy.exit.i3042
  %4220 = sub i32 0, %4216
  %scevgep6874 = getelementptr i8, ptr addrspace(5) %scevgep6873, i32 %4220
  br label %loop-memcpy-expansion5121, !amdgpu.uniform !42

Flow8040:                                         ; preds = %Flow8039, %_ZN22__llvm_libc_22_0_0_git3cpp9to_stringEy.exit.i3042
  br label %loop-memcpy-residual-header5124, !amdgpu.uniform !42

loop-memcpy-expansion5121:                        ; preds = %loop-memcpy-expansion5121.preheader, %loop-memcpy-expansion5121
  %lsr.iv6875 = phi ptr addrspace(5) [ %scevgep6876, %loop-memcpy-expansion5121 ], [ %scevgep6874, %loop-memcpy-expansion5121.preheader ], !amdgpu.uniform !42
  %loop-index5122 = phi i64 [ %4223, %loop-memcpy-expansion5121 ], [ 0, %loop-memcpy-expansion5121.preheader ]
  %4221 = load <4 x i32>, ptr addrspace(5) %lsr.iv6875, align 1
  %4222 = getelementptr i8, ptr %4211, i64 %loop-index5122
  store <4 x i32> %4221, ptr %4222, align 1
  %4223 = add i64 %loop-index5122, 16
  %scevgep6876 = getelementptr i8, ptr addrspace(5) %lsr.iv6875, i32 16
  %4224 = icmp uge i64 %4223, %4218
  br i1 %4224, label %Flow8039, label %loop-memcpy-expansion5121, !amdgpu.uniform !42

Flow8038:                                         ; preds = %Flow8037, %loop-memcpy-residual-header5124
  br label %post-loop-memcpy-expansion5120, !amdgpu.uniform !42

loop-memcpy-residual5123:                         ; preds = %loop-memcpy-residual5123.preheader, %loop-memcpy-residual5123
  %lsr.iv6879 = phi ptr addrspace(5) [ %scevgep6880, %loop-memcpy-residual5123 ], [ %scevgep6878, %loop-memcpy-residual5123.preheader ], !amdgpu.uniform !42
  %residual-loop-index5125 = phi i64 [ %4226, %loop-memcpy-residual5123 ], [ 0, %loop-memcpy-residual5123.preheader ]
  %4225 = load i8, ptr addrspace(5) %lsr.iv6879, align 1
  %scevgep6882 = getelementptr i8, ptr %scevgep6881, i64 %residual-loop-index5125
  store i8 %4225, ptr %scevgep6882, align 1
  %4226 = add i64 %residual-loop-index5125, 1
  %scevgep6880 = getelementptr i8, ptr addrspace(5) %lsr.iv6879, i32 1
  %4227 = icmp uge i64 %4226, %4217
  br i1 %4227, label %Flow8037, label %loop-memcpy-residual5123, !amdgpu.uniform !42

Flow8037:                                         ; preds = %loop-memcpy-residual5123
  br label %Flow8038, !amdgpu.uniform !42

post-loop-memcpy-expansion5120:                   ; preds = %Flow8038
  call void @llvm.lifetime.end.p5(ptr addrspace(5) %buffer.i.i.i2988) #34, !noalias !175
  br label %for.cond.i.i.i3046, !amdgpu.uniform !42

for.cond.i.i.i3046:                               ; preds = %post-loop-memcpy-expansion5120, %Flow8036
  %4228 = phi i32 [ %4265, %Flow8036 ], [ %1204, %post-loop-memcpy-expansion5120 ]
  %4229 = phi i64 [ %4266, %Flow8036 ], [ %1205, %post-loop-memcpy-expansion5120 ]
  %4230 = phi i32 [ %4267, %Flow8036 ], [ %1206, %post-loop-memcpy-expansion5120 ]
  %retval.sroa.6.0.i.i.i3047 = phi i32 [ %4270, %Flow8036 ], [ undef, %post-loop-memcpy-expansion5120 ]
  %retval.sroa.2.0.i.i.i3049 = phi i64 [ %4269, %Flow8036 ], [ undef, %post-loop-memcpy-expansion5120 ]
  %retval.sroa.8.0.i.i.i3050 = phi i32 [ %4268, %Flow8036 ], [ undef, %post-loop-memcpy-expansion5120 ]
  %index.0.i.i.i3051 = phi i32 [ %4272, %Flow8036 ], [ 0, %post-loop-memcpy-expansion5120 ]
  %4231 = load i32, ptr addrspace(1) @__llvm_rpc_client, align 8, !tbaa !68
  %cmp.not.i.i.i3052 = icmp ult i32 %index.0.i.i.i3051, %4231
  %spec.store.select.i.i.i3053 = select i1 %cmp.not.i.i.i3052, i32 %index.0.i.i.i3051, i32 0
  %4232 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  %4233 = and i32 %spec.store.select.i.i.i3053, 31
  %4234 = lshr i64 %4232, %sh_prom.i.i.i.i756
  %4235 = trunc i64 %4234 to i32
  %conv4.i.i.i.i.i3057 = and i32 %4235, 1
  %shl.i.i.i.i.i3058 = shl nuw i32 %conv4.i.i.i.i.i3057, %4233
  br label %ComputeLoop5827, !amdgpu.uniform !42

4236:                                             ; preds = %ComputeEnd5828
  %sunkaddr7463 = getelementptr inbounds i8, ptr addrspace(1) %8210, i64 40
  %4237 = atomicrmw or ptr addrspace(1) %sunkaddr7463, i32 %.lcssa9139 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %4238, !amdgpu.uniform !42

4238:                                             ; preds = %4236, %ComputeEnd5828
  %4239 = phi i32 [ %4237, %4236 ], [ poison, %ComputeEnd5828 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %8219)
  %4240 = call i32 @llvm.amdgcn.readfirstlane.i32(i32 %4239)
  %4241 = or i32 %4240, %.lcssa9140
  %shl5.i.i.i.i.i3059 = shl nuw i32 1, %4233
  %and.i.i.i.i.i3060 = and i32 %shl5.i.i.i.i.i3059, %4241
  %tobool3.i.i.i.i3061 = icmp ne i32 %and.i.i.i.i.i3060, 0
  %4242 = call i64 @llvm.amdgcn.ballot.i64(i1 %tobool3.i.i.i.i3061)
  %and.i.i.i.i.i.i3062 = and i64 %4242, %4232
  %cmp.i.not.i.i.i3063 = icmp ne i64 %4232, %and.i.i.i.i.i.i3062
  br i1 %cmp.i.not.i.i.i3063, label %if.then.i.i.i.i3064, label %Flow8035, !amdgpu.uniform !42

if.then.i.i.i.i3064:                              ; preds = %4238
  fence syncscope("agent") acquire
  %4243 = load <2 x i64>, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !75
  %4244 = extractelement <2 x i64> %4243, i32 0
  %4245 = inttoptr i64 %4244 to ptr
  %4246 = extractelement <2 x i64> %4243, i32 1
  %4247 = inttoptr i64 %4246 to ptr
  %idxprom.i.i.i.i3065 = zext i32 %spec.store.select.i.i.i3053 to i64
  %arrayidx.i.i.i.i3066 = getelementptr inbounds nuw i32, ptr %4245, i64 %idxprom.i.i.i.i3065, !amdgpu.uniform !42
  %4248 = load atomic i32, ptr %arrayidx.i.i.i.i3066 monotonic, align 4
  %4249 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %4248)
  %arrayidx.i45.i.i.i3067 = getelementptr inbounds nuw i32, ptr %4247, i64 %idxprom.i.i.i.i3065, !amdgpu.uniform !42
  %4250 = load atomic i32, ptr %arrayidx.i45.i.i.i3067 monotonic, align 4
  %4251 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %4250)
  %cmp.i46.not.i.i.i3068 = icmp ne i32 %4249, %4251
  br i1 %cmp.i46.not.i.i.i3068, label %if.then12.i.i.i3069, label %Flow8033, !amdgpu.uniform !42

if.then12.i.i.i3069:                              ; preds = %if.then.i.i.i.i3064
  fence syncscope("agent") release
  %4252 = call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %4232, i1 true)
  %iszero.i.i.i.i.i.i.i3070 = icmp ne i64 %4232, 0
  %cmp2.i.i.i.i.i.i3071 = icmp eq i64 %4252, %sh_prom.i.i.i.i756
  %cmp.i.i.i.i.i2.i3072 = select i1 %iszero.i.i.i.i.i.i.i3070, i1 %cmp2.i.i.i.i.i.i3071, i1 false
  %conv4.i.i51.i.i.i3073 = zext i1 %cmp.i.i.i.i.i2.i3072 to i32
  %shl.i.i52.i.i.i3074 = shl nuw i32 %conv4.i.i51.i.i.i3073, %4233
  %xor.i.i.i.i.i3075 = xor i32 %shl.i.i52.i.i.i3074, -1
  %4253 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop5832, !amdgpu.uniform !42

4254:                                             ; preds = %ComputeEnd5833
  %sunkaddr7464 = getelementptr inbounds i8, ptr addrspace(1) %8210, i64 40
  %4255 = atomicrmw and ptr addrspace(1) %sunkaddr7464, i32 %.lcssa9141 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %4256, !amdgpu.uniform !42

4256:                                             ; preds = %4254, %ComputeEnd5833
  call void @llvm.amdgcn.end.cf.i64(i64 %8237)
  call void @llvm.amdgcn.wave.barrier()
  br label %Flow8033, !amdgpu.uniform !42

if.end14.i.i.i3136:                               ; preds = %Flow8033
  %4257 = call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %4232, i1 true)
  %iszero.i.i.i.i.i.i3137 = icmp ne i64 %4232, 0
  %cmp2.i.i.i.i.i3138 = icmp eq i64 %4257, %sh_prom.i.i.i.i756
  %cmp.i.i.i.i.i3139 = select i1 %iszero.i.i.i.i.i.i3137, i1 %cmp2.i.i.i.i.i3138, i1 false
  %4258 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.i.i.i.i.i3139)
  %4259 = extractvalue { i1, i64 } %4258, 0
  %4260 = extractvalue { i1, i64 } %4258, 1
  br i1 %4259, label %if.then16.i.i.i3141, label %if.end22.i.i.i3140

if.then16.i.i.i3141:                              ; preds = %if.end14.i.i.i3136
  %4261 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !76
  %opcode.i.i.i3142 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %4261, i64 %idxprom.i.i.i.i3065, i32 1
  store i32 1660944387, ptr %opcode.i.i.i3142, align 8, !tbaa !77
  %arrayidx21.i.i.i3143 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %4261, i64 %idxprom.i.i.i.i3065
  store i64 %4232, ptr %arrayidx21.i.i.i3143, align 8, !tbaa !80
  br label %if.end22.i.i.i3140, !amdgpu.uniform !42

Flow8034:                                         ; preds = %if.end22.i.i.i3140, %Flow8033
  %4262 = phi i32 [ %4249, %if.end22.i.i.i3140 ], [ %retval.sroa.8.0.i.i.i3050, %Flow8033 ]
  %4263 = phi i64 [ %4232, %if.end22.i.i.i3140 ], [ %retval.sroa.2.0.i.i.i3049, %Flow8033 ]
  %4264 = phi i32 [ %spec.store.select.i.i.i3053, %if.end22.i.i.i3140 ], [ %retval.sroa.6.0.i.i.i3047, %Flow8033 ]
  br label %cleanup26.i.i.i3076, !amdgpu.uniform !42

if.end22.i.i.i3140:                               ; preds = %if.then16.i.i.i3141, %if.end14.i.i.i3136
  call void @llvm.amdgcn.end.cf.i64(i64 %4260)
  call void @llvm.amdgcn.wave.barrier()
  br label %Flow8034, !amdgpu.uniform !42

Flow8035:                                         ; preds = %cleanup26.i.i.i3076, %4238
  %4265 = phi i32 [ %4264, %cleanup26.i.i.i3076 ], [ %4228, %4238 ]
  %4266 = phi i64 [ %4263, %cleanup26.i.i.i3076 ], [ %4229, %4238 ]
  %4267 = phi i32 [ %4262, %cleanup26.i.i.i3076 ], [ %4230, %4238 ]
  %4268 = phi i32 [ %4262, %cleanup26.i.i.i3076 ], [ %retval.sroa.8.0.i.i.i3050, %4238 ]
  %4269 = phi i64 [ %4263, %cleanup26.i.i.i3076 ], [ %retval.sroa.2.0.i.i.i3049, %4238 ]
  %4270 = phi i32 [ %4264, %cleanup26.i.i.i3076 ], [ %retval.sroa.6.0.i.i.i3047, %4238 ]
  %4271 = phi i1 [ %cmp.i46.not.i.i.i3068, %cleanup26.i.i.i3076 ], [ true, %4238 ]
  br i1 %4271, label %for.inc.i.i.i3081, label %Flow8036, !amdgpu.uniform !42

cleanup26.i.i.i3076:                              ; preds = %Flow8034
  br label %Flow8035, !amdgpu.uniform !42

for.inc.i.i.i3081:                                ; preds = %Flow8035
  %inc.i.i.i3086 = add i32 %spec.store.select.i.i.i3053, 1
  br label %Flow8036, !amdgpu.uniform !42

Flow8036:                                         ; preds = %for.inc.i.i.i3081, %Flow8035
  %4272 = phi i32 [ %inc.i.i.i3086, %for.inc.i.i.i3081 ], [ poison, %Flow8035 ]
  %4273 = phi i1 [ false, %for.inc.i.i.i3081 ], [ true, %Flow8035 ]
  br i1 %4273, label %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i3087, label %for.cond.i.i.i3046, !amdgpu.uniform !42

_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i3087: ; preds = %Flow8036
  %.lcssa9144 = phi i32 [ %4265, %Flow8036 ]
  %.lcssa9143 = phi i64 [ %4266, %Flow8036 ]
  %.lcssa9142 = phi i32 [ %4267, %Flow8036 ]
  fence acquire
  %4274 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %mul.i.i.i.i.i.i3088 = shl i32 %.lcssa9144, 6
  %idxprom.i19.i.i.i.i.i3089 = zext i32 %mul.i.i.i.i.i.i3088 to i64
  %arrayidx.i20.i.i.i.i.i3090 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %4274, i64 %idxprom.i19.i.i.i.i.i3089
  %arrayidx.i22.i.i.i.i.i3091 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %arrayidx.i20.i.i.i.i.i3090, i64 %sh_prom.i.i.i.i756
  store i64 %writer.sroa.8.4.i.i.i.i3017, ptr %arrayidx.i22.i.i.i.i.i3091, align 8, !tbaa !82
  %spec.select.i.i.i.i.i.i.i3092 = call i64 @llvm.umin.i64(i64 %writer.sroa.8.4.i.i.i.i3017, i64 56)
  %4275 = and i64 %spec.select.i.i.i.i.i.i.i3092, 15
  %4276 = sub i64 %spec.select.i.i.i.i.i.i.i3092, %4275
  %4277 = icmp ne i64 %4276, 0
  br i1 %4277, label %loop-memcpy-expansion5013.preheader, label %Flow8032, !amdgpu.uniform !42

loop-memcpy-expansion5013.preheader:              ; preds = %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i3087
  %scevgep6883 = getelementptr i8, ptr %4274, i64 8
  %scevgep6884 = getelementptr i8, ptr %scevgep6883, i64 %1474
  %4278 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i3089, 6
  %scevgep6885 = getelementptr i8, ptr %scevgep6884, i64 %4278
  br label %loop-memcpy-expansion5013, !amdgpu.uniform !42

Flow8032:                                         ; preds = %Flow8031, %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i3087
  br label %loop-memcpy-residual-header5016, !amdgpu.uniform !42

loop-memcpy-expansion5013:                        ; preds = %loop-memcpy-expansion5013.preheader, %loop-memcpy-expansion5013
  %loop-index5014 = phi i64 [ %4281, %loop-memcpy-expansion5013 ], [ 0, %loop-memcpy-expansion5013.preheader ]
  %4279 = getelementptr i8, ptr %4211, i64 %loop-index5014, !amdgpu.uniform !42
  %4280 = load <4 x i32>, ptr %4279, align 1
  %scevgep6886 = getelementptr i8, ptr %scevgep6885, i64 %loop-index5014
  store <4 x i32> %4280, ptr %scevgep6886, align 1
  %4281 = add i64 %loop-index5014, 16
  %4282 = icmp uge i64 %4281, %4276
  br i1 %4282, label %Flow8031, label %loop-memcpy-expansion5013, !amdgpu.uniform !42

Flow8030:                                         ; preds = %Flow8028, %Flow8029
  %4283 = phi i64 [ %6265, %Flow8028 ], [ %6262, %Flow8029 ]
  br label %post-loop-memcpy-expansion5012, !amdgpu.uniform !42

loop-memcpy-residual5015:                         ; preds = %loop-memcpy-residual5015.preheader, %loop-memcpy-residual5015
  %residual-loop-index5017 = phi i64 [ %4285, %loop-memcpy-residual5015 ], [ 0, %loop-memcpy-residual5015.preheader ]
  %scevgep6892 = getelementptr i8, ptr %scevgep6891, i64 %residual-loop-index5017, !amdgpu.uniform !42
  %4284 = load i8, ptr %scevgep6892, align 1
  %scevgep6890 = getelementptr i8, ptr %scevgep6889, i64 %residual-loop-index5017
  store i8 %4284, ptr %scevgep6890, align 1
  %4285 = add i64 %residual-loop-index5017, 1
  %4286 = icmp uge i64 %4285, %4275
  br i1 %4286, label %Flow8028, label %loop-memcpy-residual5015, !amdgpu.uniform !42

Flow8028:                                         ; preds = %loop-memcpy-residual5015
  br label %Flow8030, !amdgpu.uniform !42

post-loop-memcpy-expansion5012:                   ; preds = %Flow8030
  %tobool.not.i.i.i.i.i.i3094 = icmp eq i32 %.lcssa9142, 0
  %conv.i.i.i.i.i.i3095 = zext i1 %tobool.not.i.i.i.i.i.i3094 to i32
  fence release
  %4287 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %idxprom.i23.i.i.i.i.i3096 = zext i32 %.lcssa9144 to i64
  %arrayidx.i24.i.i.i.i.i3097 = getelementptr inbounds nuw i32, ptr %4287, i64 %idxprom.i23.i.i.i.i.i3096
  store atomic i32 %conv.i.i.i.i.i.i3095, ptr %arrayidx.i24.i.i.i.i.i3097 monotonic, align 4
  %4288 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !84
  %arrayidx.i.i5.i.i3098 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %4288, i64 %idxprom.i23.i.i.i.i.i3096
  %4289 = load i64, ptr %arrayidx.i.i5.i.i3098, align 8, !tbaa !80
  %scevgep6893 = getelementptr i8, ptr %4211, i64 56
  %4290 = add i64 %writer.sroa.8.4.i.i.i.i3017, -56
  br label %while.cond.i.i.i.i3099, !amdgpu.uniform !42

while.cond.i.i.i.i3099:                           ; preds = %post-loop-memcpy-expansion5012, %Flow8027
  %phi.broken8722 = phi i64 [ 0, %post-loop-memcpy-expansion5012 ], [ %4312, %Flow8027 ]
  %lsr.iv6899 = phi i64 [ %4310, %Flow8027 ], [ %4290, %post-loop-memcpy-expansion5012 ]
  %lsr.iv6894 = phi ptr [ %4309, %Flow8027 ], [ %scevgep6893, %post-loop-memcpy-expansion5012 ]
  %port.sroa.37.0.i.i3100 = phi i32 [ %4308, %Flow8027 ], [ %conv.i.i.i.i.i.i3095, %post-loop-memcpy-expansion5012 ]
  %idx.0.i.i.i.i3101 = phi i64 [ %4307, %Flow8027 ], [ 56, %post-loop-memcpy-expansion5012 ]
  %cmp.i.i.i.i3102 = icmp ugt i64 %writer.sroa.8.4.i.i.i.i3017, %idx.0.i.i.i.i3101
  %4291 = call i64 @llvm.amdgcn.ballot.i64(i1 %cmp.i.i.i.i3102)
  %and.i.i.i.i6.i.i3103 = and i64 %4291, %4289
  %tobool.not.i.i.i.i3104 = icmp ne i64 %and.i.i.i.i6.i.i3103, 0
  %4292 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i22.i.i.i3105 = getelementptr inbounds nuw i32, ptr %4292, i64 %idxprom.i23.i.i.i.i.i3096
  %4293 = load atomic i32, ptr %arrayidx.i22.i.i.i3105 monotonic, align 4
  %4294 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %tobool.not.i.i.i.i3104)
  %4295 = extractvalue { i1, i64 } %4294, 0
  %4296 = extractvalue { i1, i64 } %4294, 1
  br i1 %4295, label %cond.false.i8.i.i.i.i3106, label %Flow8027

cond.false.i8.i.i.i.i3106:                        ; preds = %while.cond.i.i.i.i3099
  %4297 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %4293)
  %4298 = icmp ne i32 %4297, %port.sroa.37.0.i.i3100
  %4299 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %4298)
  %4300 = extractvalue { i1, i64 } %4299, 0
  %4301 = extractvalue { i1, i64 } %4299, 1
  br i1 %4300, label %while.body.i.i21.i.i.i.i3107.preheader, label %Flow8026

while.body.i.i21.i.i.i.i3107.preheader:           ; preds = %cond.false.i8.i.i.i.i3106
  br label %while.body.i.i21.i.i.i.i3107, !amdgpu.uniform !42

Flow8026:                                         ; preds = %Flow8025, %cond.false.i8.i.i.i.i3106
  call void @llvm.amdgcn.end.cf.i64(i64 %4301)
  br label %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i3110, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i3107:                     ; preds = %while.body.i.i21.i.i.i.i3107.preheader, %while.body.i.i21.i.i.i.i3107
  %phi.broken8721 = phi i64 [ 0, %while.body.i.i21.i.i.i.i3107.preheader ], [ %4305, %while.body.i.i21.i.i.i.i3107 ]
  call void @llvm.amdgcn.s.sleep(i32 2)
  %4302 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i22.i.i.i.i3108 = getelementptr inbounds nuw i32, ptr %4302, i64 %idxprom.i23.i.i.i.i.i3096
  %4303 = load atomic i32, ptr %arrayidx.i.i.i22.i.i.i.i3108 monotonic, align 4
  %4304 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %4303)
  %cmp.i.not.i.i23.i.i.i.i3109 = icmp eq i32 %4304, %port.sroa.37.0.i.i3100
  %4305 = call i64 @llvm.amdgcn.if.break.i64(i1 %cmp.i.not.i.i23.i.i.i.i3109, i64 %phi.broken8721)
  %4306 = call i1 @llvm.amdgcn.loop.i64(i64 %4305)
  br i1 %4306, label %Flow8025, label %while.body.i.i21.i.i.i.i3107

Flow8025:                                         ; preds = %while.body.i.i21.i.i.i.i3107
  %.lcssa9145 = phi i64 [ %4305, %while.body.i.i21.i.i.i.i3107 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa9145)
  br label %Flow8026, !amdgpu.uniform !42

Flow8027:                                         ; preds = %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i3111, %while.cond.i.i.i.i3099
  %4307 = phi i64 [ %add.i.i.i.i3114, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i3111 ], [ poison, %while.cond.i.i.i.i3099 ]
  %4308 = phi i32 [ %conv.i.i26.i.i.i.i3112, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i3111 ], [ poison, %while.cond.i.i.i.i3099 ]
  %4309 = phi ptr [ %scevgep6895, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i3111 ], [ poison, %while.cond.i.i.i.i3099 ]
  %4310 = phi i64 [ %lsr.iv.next6900, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i3111 ], [ poison, %while.cond.i.i.i.i3099 ]
  %4311 = phi i1 [ false, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i3111 ], [ true, %while.cond.i.i.i.i3099 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %4296)
  %4312 = call i64 @llvm.amdgcn.if.break.i64(i1 %4311, i64 %phi.broken8722)
  %4313 = call i1 @llvm.amdgcn.loop.i64(i64 %4312)
  br i1 %4313, label %cond.false.i.i.i3121, label %while.cond.i.i.i.i3099

_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i3110: ; preds = %Flow8026
  fence acquire
  br i1 %cmp.i.i.i.i3102, label %if.then.i.i.i.i.i.i3.i3115, label %Flow8024, !amdgpu.uniform !42

if.then.i.i.i.i.i.i3.i3115:                       ; preds = %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i3110
  %sub.i.i.i.i.i.i.i3117 = sub nuw i64 %writer.sroa.8.4.i.i.i.i3017, %idx.0.i.i.i.i3101
  %spec.select.i.i.i36.i.i.i.i3118 = call i64 @llvm.umin.i64(i64 %sub.i.i.i.i.i.i.i3117, i64 64)
  %4314 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %4315 = and i64 %spec.select.i.i.i36.i.i.i.i3118, 15
  %4316 = sub i64 %spec.select.i.i.i36.i.i.i.i3118, %4315
  %4317 = icmp ne i64 %4316, 0
  br i1 %4317, label %loop-memcpy-expansion5007.preheader, label %Flow8023, !amdgpu.uniform !42

loop-memcpy-expansion5007.preheader:              ; preds = %if.then.i.i.i.i.i.i3.i3115
  %arrayidx.i20.i35.i.i.i.i3119 = getelementptr %"struct.rpc::Buffer", ptr %4314, i64 %idxprom.i19.i.i.i.i.i3089
  %arrayidx.i22.i38.i.i.i.i3120 = getelementptr %"struct.rpc::Buffer", ptr %arrayidx.i20.i35.i.i.i.i3119, i64 %sh_prom.i.i.i.i756
  br label %loop-memcpy-expansion5007, !amdgpu.uniform !42

Flow8023:                                         ; preds = %Flow8022, %if.then.i.i.i.i.i.i3.i3115
  br label %loop-memcpy-residual-header5010, !amdgpu.uniform !42

loop-memcpy-expansion5007:                        ; preds = %loop-memcpy-expansion5007.preheader, %loop-memcpy-expansion5007
  %loop-index5008 = phi i64 [ %4320, %loop-memcpy-expansion5007 ], [ 0, %loop-memcpy-expansion5007.preheader ]
  %scevgep6896 = getelementptr i8, ptr %lsr.iv6894, i64 %loop-index5008, !amdgpu.uniform !42
  %4318 = load <4 x i32>, ptr %scevgep6896, align 1
  %4319 = getelementptr i8, ptr %arrayidx.i22.i38.i.i.i.i3120, i64 %loop-index5008
  store <4 x i32> %4318, ptr %4319, align 1
  %4320 = add i64 %loop-index5008, 16
  %4321 = icmp uge i64 %4320, %4316
  br i1 %4321, label %Flow8022, label %loop-memcpy-expansion5007, !amdgpu.uniform !42

Flow8021:                                         ; preds = %Flow8020, %loop-memcpy-residual-header5010
  br label %Flow8024, !amdgpu.uniform !42

loop-memcpy-residual5009:                         ; preds = %loop-memcpy-residual5009.preheader, %loop-memcpy-residual5009
  %residual-loop-index5011 = phi i64 [ %4323, %loop-memcpy-residual5009 ], [ 0, %loop-memcpy-residual5009.preheader ]
  %scevgep6905 = getelementptr i8, ptr %scevgep6904, i64 %residual-loop-index5011, !amdgpu.uniform !42
  %4322 = load i8, ptr %scevgep6905, align 1
  %scevgep6903 = getelementptr i8, ptr %scevgep6902, i64 %residual-loop-index5011
  store i8 %4322, ptr %scevgep6903, align 1
  %4323 = add i64 %residual-loop-index5011, 1
  %4324 = icmp uge i64 %4323, %4315
  br i1 %4324, label %Flow8020, label %loop-memcpy-residual5009, !amdgpu.uniform !42

Flow8020:                                         ; preds = %loop-memcpy-residual5009
  br label %Flow8021, !amdgpu.uniform !42

_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i3111: ; preds = %Flow8024
  %conv.i.i26.i.i.i.i3112 = xor i32 %port.sroa.37.0.i.i3100, 1
  fence release
  %4325 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %arrayidx.i24.i29.i.i.i.i3113 = getelementptr inbounds nuw i32, ptr %4325, i64 %idxprom.i23.i.i.i.i.i3096
  store atomic i32 %conv.i.i26.i.i.i.i3112, ptr %arrayidx.i24.i29.i.i.i.i3113 monotonic, align 4
  %add.i.i.i.i3114 = add i64 %idx.0.i.i.i.i3101, 64
  %scevgep6895 = getelementptr i8, ptr %lsr.iv6894, i64 64
  %lsr.iv.next6900 = add i64 %lsr.iv6899, -64
  br label %Flow8027, !amdgpu.uniform !42

cond.false.i.i.i3121:                             ; preds = %Flow8027
  %.lcssa9147 = phi i64 [ %4312, %Flow8027 ]
  %port.sroa.37.0.i.i3100.lcssa = phi i32 [ %port.sroa.37.0.i.i3100, %Flow8027 ]
  %.lcssa9146 = phi i32 [ %4293, %Flow8027 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa9147)
  %4326 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %.lcssa9146)
  %4327 = icmp ne i32 %4326, %port.sroa.37.0.i.i3100.lcssa
  %4328 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %4327)
  %4329 = extractvalue { i1, i64 } %4328, 0
  %4330 = extractvalue { i1, i64 } %4328, 1
  br i1 %4329, label %while.body.i.i.i.i3122.preheader, label %Flow8019

while.body.i.i.i.i3122.preheader:                 ; preds = %cond.false.i.i.i3121
  br label %while.body.i.i.i.i3122, !amdgpu.uniform !42

Flow8019:                                         ; preds = %Flow8018, %cond.false.i.i.i3121
  call void @llvm.amdgcn.end.cf.i64(i64 %4330)
  br label %_ZN22__llvm_libc_22_0_0_git15write_to_stderrENS_3cpp11string_viewE.exit.i3125, !amdgpu.uniform !42

while.body.i.i.i.i3122:                           ; preds = %while.body.i.i.i.i3122.preheader, %while.body.i.i.i.i3122
  %phi.broken8723 = phi i64 [ 0, %while.body.i.i.i.i3122.preheader ], [ %4334, %while.body.i.i.i.i3122 ]
  call void @llvm.amdgcn.s.sleep(i32 2)
  %4331 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i8.i.i3123 = getelementptr inbounds nuw i32, ptr %4331, i64 %idxprom.i23.i.i.i.i.i3096
  %4332 = load atomic i32, ptr %arrayidx.i.i.i8.i.i3123 monotonic, align 4
  %4333 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %4332)
  %cmp.i.not.i.i.i.i3124 = icmp eq i32 %4333, %port.sroa.37.0.i.i3100.lcssa
  %4334 = call i64 @llvm.amdgcn.if.break.i64(i1 %cmp.i.not.i.i.i.i3124, i64 %phi.broken8723)
  %4335 = call i1 @llvm.amdgcn.loop.i64(i64 %4334)
  br i1 %4335, label %Flow8018, label %while.body.i.i.i.i3122

Flow8018:                                         ; preds = %while.body.i.i.i.i3122
  %.lcssa9148 = phi i64 [ %4334, %while.body.i.i.i.i3122 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa9148)
  br label %Flow8019, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git15write_to_stderrENS_3cpp11string_viewE.exit.i3125: ; preds = %Flow8019
  fence acquire
  call void @llvm.amdgcn.wave.barrier()
  fence syncscope("agent") release
  %4336 = call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %.lcssa9143, i1 true)
  %iszero.i.i.i.i.i13.i.i3126 = icmp ne i64 %.lcssa9143, 0
  %cmp2.i.i.i.i14.i.i3127 = icmp eq i64 %4336, %sh_prom.i.i.i.i756
  %cmp.i.i.i.i15.i.i3128 = select i1 %iszero.i.i.i.i.i13.i.i3126, i1 %cmp2.i.i.i.i14.i.i3127, i1 false
  %4337 = and i32 %.lcssa9144, 31
  %conv4.i.i.i19.i.i3132 = zext i1 %cmp.i.i.i.i15.i.i3128 to i32
  %shl.i.i.i20.i.i3133 = shl nuw i32 %conv4.i.i.i19.i.i3132, %4337
  %xor.i.i.i21.i.i3134 = xor i32 %shl.i.i.i20.i.i3133, -1
  %4338 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop5836, !amdgpu.uniform !42

4339:                                             ; preds = %ComputeEnd5837
  %div8.i.i.i16.i.i3129 = lshr i32 %.lcssa9144, 5
  %idxprom.i.i.i17.i.i3130 = zext nneg i32 %div8.i.i.i16.i.i3129 to i64
  %4340 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i17.i.i3130
  %sunkaddr7465 = getelementptr inbounds i8, ptr addrspace(1) %4340, i64 40
  %4341 = atomicrmw and ptr addrspace(1) %sunkaddr7465, i32 %.lcssa9149 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %4342, !amdgpu.uniform !42

4342:                                             ; preds = %4339, %ComputeEnd5837
  call void @llvm.amdgcn.end.cf.i64(i64 %8254)
  call void @llvm.amdgcn.wave.barrier()
  br label %for.cond.i.i1.i3165, !amdgpu.uniform !42

for.cond.i.i1.i3165:                              ; preds = %4342, %Flow8017
  %4343 = phi i32 [ %4380, %Flow8017 ], [ %1207, %4342 ]
  %4344 = phi i64 [ %4381, %Flow8017 ], [ %1208, %4342 ]
  %4345 = phi i32 [ %4382, %Flow8017 ], [ %1209, %4342 ]
  %retval.sroa.6.0.i.i.i3166 = phi i32 [ %4385, %Flow8017 ], [ undef, %4342 ]
  %retval.sroa.2.0.i.i.i3168 = phi i64 [ %4384, %Flow8017 ], [ undef, %4342 ]
  %retval.sroa.8.0.i.i.i3169 = phi i32 [ %4383, %Flow8017 ], [ undef, %4342 ]
  %index.0.i.i.i3170 = phi i32 [ %4387, %Flow8017 ], [ 0, %4342 ]
  %4346 = load i32, ptr addrspace(1) @__llvm_rpc_client, align 8, !tbaa !68
  %cmp.not.i.i.i3171 = icmp ult i32 %index.0.i.i.i3170, %4346
  %spec.store.select.i.i.i3172 = select i1 %cmp.not.i.i.i3171, i32 %index.0.i.i.i3170, i32 0
  %4347 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  %4348 = and i32 %spec.store.select.i.i.i3172, 31
  %4349 = lshr i64 %4347, %sh_prom.i.i.i.i756
  %4350 = trunc i64 %4349 to i32
  %conv4.i.i.i.i.i3176 = and i32 %4350, 1
  %shl.i.i.i.i.i3177 = shl nuw i32 %conv4.i.i.i.i.i3176, %4348
  br label %ComputeLoop5840, !amdgpu.uniform !42

4351:                                             ; preds = %ComputeEnd5841
  %sunkaddr7466 = getelementptr inbounds i8, ptr addrspace(1) %8264, i64 40
  %4352 = atomicrmw or ptr addrspace(1) %sunkaddr7466, i32 %.lcssa9150 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %4353, !amdgpu.uniform !42

4353:                                             ; preds = %4351, %ComputeEnd5841
  %4354 = phi i32 [ %4352, %4351 ], [ poison, %ComputeEnd5841 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %8273)
  %4355 = call i32 @llvm.amdgcn.readfirstlane.i32(i32 %4354)
  %4356 = or i32 %4355, %.lcssa9151
  %shl5.i.i.i.i.i3178 = shl nuw i32 1, %4348
  %and.i.i.i.i.i3179 = and i32 %shl5.i.i.i.i.i3178, %4356
  %tobool3.i.i.i.i3180 = icmp ne i32 %and.i.i.i.i.i3179, 0
  %4357 = call i64 @llvm.amdgcn.ballot.i64(i1 %tobool3.i.i.i.i3180)
  %and.i.i.i.i.i.i3181 = and i64 %4357, %4347
  %cmp.i.not.i.i.i3182 = icmp ne i64 %4347, %and.i.i.i.i.i.i3181
  br i1 %cmp.i.not.i.i.i3182, label %if.then.i.i.i.i3183, label %Flow8016, !amdgpu.uniform !42

if.then.i.i.i.i3183:                              ; preds = %4353
  fence syncscope("agent") acquire
  %4358 = load <2 x i64>, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !75
  %4359 = extractelement <2 x i64> %4358, i32 0
  %4360 = inttoptr i64 %4359 to ptr
  %4361 = extractelement <2 x i64> %4358, i32 1
  %4362 = inttoptr i64 %4361 to ptr
  %idxprom.i.i.i.i3184 = zext i32 %spec.store.select.i.i.i3172 to i64
  %arrayidx.i.i.i.i3185 = getelementptr inbounds nuw i32, ptr %4360, i64 %idxprom.i.i.i.i3184, !amdgpu.uniform !42
  %4363 = load atomic i32, ptr %arrayidx.i.i.i.i3185 monotonic, align 4
  %4364 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %4363)
  %arrayidx.i45.i.i.i3186 = getelementptr inbounds nuw i32, ptr %4362, i64 %idxprom.i.i.i.i3184, !amdgpu.uniform !42
  %4365 = load atomic i32, ptr %arrayidx.i45.i.i.i3186 monotonic, align 4
  %4366 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %4365)
  %cmp.i46.not.i.i.i3187 = icmp ne i32 %4364, %4366
  br i1 %cmp.i46.not.i.i.i3187, label %if.then12.i.i.i3188, label %Flow8014, !amdgpu.uniform !42

if.then12.i.i.i3188:                              ; preds = %if.then.i.i.i.i3183
  fence syncscope("agent") release
  %4367 = call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %4347, i1 true)
  %iszero.i.i.i.i.i.i.i3189 = icmp ne i64 %4347, 0
  %cmp2.i.i.i.i.i.i3190 = icmp eq i64 %4367, %sh_prom.i.i.i.i756
  %cmp.i.i.i.i.i.i3191 = select i1 %iszero.i.i.i.i.i.i.i3189, i1 %cmp2.i.i.i.i.i.i3190, i1 false
  %conv4.i.i51.i.i.i3192 = zext i1 %cmp.i.i.i.i.i.i3191 to i32
  %shl.i.i52.i.i.i3193 = shl nuw i32 %conv4.i.i51.i.i.i3192, %4348
  %xor.i.i.i.i.i3194 = xor i32 %shl.i.i52.i.i.i3193, -1
  %4368 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop5845, !amdgpu.uniform !42

4369:                                             ; preds = %ComputeEnd5846
  %sunkaddr7467 = getelementptr inbounds i8, ptr addrspace(1) %8264, i64 40
  %4370 = atomicrmw and ptr addrspace(1) %sunkaddr7467, i32 %.lcssa9152 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %4371, !amdgpu.uniform !42

4371:                                             ; preds = %4369, %ComputeEnd5846
  call void @llvm.amdgcn.end.cf.i64(i64 %8291)
  call void @llvm.amdgcn.wave.barrier()
  br label %Flow8014, !amdgpu.uniform !42

if.end14.i.i.i3254:                               ; preds = %Flow8014
  %4372 = call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %4347, i1 true)
  %iszero.i.i.i.i.i.i3255 = icmp ne i64 %4347, 0
  %cmp2.i.i.i.i.i3256 = icmp eq i64 %4372, %sh_prom.i.i.i.i756
  %cmp.i.i.i.i.i3257 = select i1 %iszero.i.i.i.i.i.i3255, i1 %cmp2.i.i.i.i.i3256, i1 false
  %4373 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.i.i.i.i.i3257)
  %4374 = extractvalue { i1, i64 } %4373, 0
  %4375 = extractvalue { i1, i64 } %4373, 1
  br i1 %4374, label %if.then16.i.i.i3259, label %if.end22.i.i.i3258

if.then16.i.i.i3259:                              ; preds = %if.end14.i.i.i3254
  %4376 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !76
  %opcode.i.i.i3260 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %4376, i64 %idxprom.i.i.i.i3184, i32 1
  store i32 1660944387, ptr %opcode.i.i.i3260, align 8, !tbaa !77
  %arrayidx21.i.i.i3261 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %4376, i64 %idxprom.i.i.i.i3184
  store i64 %4347, ptr %arrayidx21.i.i.i3261, align 8, !tbaa !80
  br label %if.end22.i.i.i3258, !amdgpu.uniform !42

Flow8015:                                         ; preds = %if.end22.i.i.i3258, %Flow8014
  %4377 = phi i32 [ %4364, %if.end22.i.i.i3258 ], [ %retval.sroa.8.0.i.i.i3169, %Flow8014 ]
  %4378 = phi i64 [ %4347, %if.end22.i.i.i3258 ], [ %retval.sroa.2.0.i.i.i3168, %Flow8014 ]
  %4379 = phi i32 [ %spec.store.select.i.i.i3172, %if.end22.i.i.i3258 ], [ %retval.sroa.6.0.i.i.i3166, %Flow8014 ]
  br label %cleanup26.i.i.i3195, !amdgpu.uniform !42

if.end22.i.i.i3258:                               ; preds = %if.then16.i.i.i3259, %if.end14.i.i.i3254
  call void @llvm.amdgcn.end.cf.i64(i64 %4375)
  call void @llvm.amdgcn.wave.barrier()
  br label %Flow8015, !amdgpu.uniform !42

Flow8016:                                         ; preds = %cleanup26.i.i.i3195, %4353
  %4380 = phi i32 [ %4379, %cleanup26.i.i.i3195 ], [ %4343, %4353 ]
  %4381 = phi i64 [ %4378, %cleanup26.i.i.i3195 ], [ %4344, %4353 ]
  %4382 = phi i32 [ %4377, %cleanup26.i.i.i3195 ], [ %4345, %4353 ]
  %4383 = phi i32 [ %4377, %cleanup26.i.i.i3195 ], [ %retval.sroa.8.0.i.i.i3169, %4353 ]
  %4384 = phi i64 [ %4378, %cleanup26.i.i.i3195 ], [ %retval.sroa.2.0.i.i.i3168, %4353 ]
  %4385 = phi i32 [ %4379, %cleanup26.i.i.i3195 ], [ %retval.sroa.6.0.i.i.i3166, %4353 ]
  %4386 = phi i1 [ %cmp.i46.not.i.i.i3187, %cleanup26.i.i.i3195 ], [ true, %4353 ]
  br i1 %4386, label %for.inc.i.i.i3200, label %Flow8017, !amdgpu.uniform !42

cleanup26.i.i.i3195:                              ; preds = %Flow8015
  br label %Flow8016, !amdgpu.uniform !42

for.inc.i.i.i3200:                                ; preds = %Flow8016
  %inc.i.i.i3205 = add i32 %spec.store.select.i.i.i3172, 1
  br label %Flow8017, !amdgpu.uniform !42

Flow8017:                                         ; preds = %for.inc.i.i.i3200, %Flow8016
  %4387 = phi i32 [ %inc.i.i.i3205, %for.inc.i.i.i3200 ], [ poison, %Flow8016 ]
  %4388 = phi i1 [ false, %for.inc.i.i.i3200 ], [ true, %Flow8016 ]
  br i1 %4388, label %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i3206, label %for.cond.i.i1.i3165, !amdgpu.uniform !42

_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i3206: ; preds = %Flow8017
  %.lcssa9155 = phi i32 [ %4380, %Flow8017 ]
  %.lcssa9154 = phi i64 [ %4381, %Flow8017 ]
  %.lcssa9153 = phi i32 [ %4382, %Flow8017 ]
  fence acquire
  %4389 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %mul.i.i.i.i.i.i3207 = shl i32 %.lcssa9155, 6
  %idxprom.i19.i.i.i.i.i3208 = zext i32 %mul.i.i.i.i.i.i3207 to i64
  %arrayidx.i20.i.i.i.i.i3209 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %4389, i64 %idxprom.i19.i.i.i.i.i3208
  %arrayidx.i22.i.i.i.i.i3210 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %arrayidx.i20.i.i.i.i.i3209, i64 %sh_prom.i.i.i.i756
  store i64 sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.24.40 to ptr), i64 5) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.24.40 to ptr) to i64)), ptr %arrayidx.i22.i.i.i.i.i3210, align 8, !tbaa !82
  br i1 %1126, label %loop-memcpy-expansion5385.preheader, label %Flow8013, !amdgpu.uniform !42

loop-memcpy-expansion5385.preheader:              ; preds = %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i3206
  %scevgep6906 = getelementptr i8, ptr %4389, i64 8
  %scevgep6907 = getelementptr i8, ptr %scevgep6906, i64 %1474
  %4390 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i3208, 6
  %scevgep6908 = getelementptr i8, ptr %scevgep6907, i64 %4390
  br label %loop-memcpy-expansion5385, !amdgpu.uniform !42

Flow8013:                                         ; preds = %Flow8012, %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i3206
  br label %loop-memcpy-residual-header5388, !amdgpu.uniform !42

loop-memcpy-expansion5385:                        ; preds = %loop-memcpy-expansion5385.preheader, %loop-memcpy-expansion5385
  %loop-index5386 = phi i64 [ %4393, %loop-memcpy-expansion5385 ], [ 0, %loop-memcpy-expansion5385.preheader ]
  %4391 = getelementptr i8, ptr addrspace(4) @.str.24.40, i64 %loop-index5386, !amdgpu.uniform !42
  %4392 = load <4 x i32>, ptr addrspace(4) %4391, align 1
  %scevgep6909 = getelementptr i8, ptr %scevgep6908, i64 %loop-index5386
  store <4 x i32> %4392, ptr %scevgep6909, align 1
  %4393 = add i64 %loop-index5386, 16
  %4394 = icmp uge i64 %4393, %1125
  br i1 %4394, label %Flow8012, label %loop-memcpy-expansion5385, !amdgpu.uniform !42

Flow8011:                                         ; preds = %Flow8009, %Flow8010
  %4395 = phi i64 [ %6635, %Flow8009 ], [ %6633, %Flow8010 ]
  br label %post-loop-memcpy-expansion5384, !amdgpu.uniform !42

loop-memcpy-residual5387:                         ; preds = %loop-memcpy-residual5387.preheader, %loop-memcpy-residual5387
  %residual-loop-index5389 = phi i64 [ %4397, %loop-memcpy-residual5387 ], [ 0, %loop-memcpy-residual5387.preheader ]
  %scevgep6914 = getelementptr i8, ptr addrspace(4) @.str.24.40, i64 %residual-loop-index5389, !amdgpu.uniform !42
  %4396 = load i8, ptr addrspace(4) %scevgep6914, align 1
  %scevgep6913 = getelementptr i8, ptr %scevgep6912, i64 %residual-loop-index5389
  store i8 %4396, ptr %scevgep6913, align 1
  %4397 = add i64 %residual-loop-index5389, 1
  %4398 = icmp uge i64 %4397, %1124
  br i1 %4398, label %Flow8009, label %loop-memcpy-residual5387, !amdgpu.uniform !42

Flow8009:                                         ; preds = %loop-memcpy-residual5387
  br label %Flow8011, !amdgpu.uniform !42

post-loop-memcpy-expansion5384:                   ; preds = %Flow8011
  %tobool.not.i.i.i.i.i.i3213 = icmp eq i32 %.lcssa9153, 0
  %conv.i.i.i.i.i.i3214 = zext i1 %tobool.not.i.i.i.i.i.i3213 to i32
  fence release
  %4399 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %idxprom.i23.i.i.i.i.i3215 = zext i32 %.lcssa9155 to i64
  %arrayidx.i24.i.i.i.i.i3216 = getelementptr inbounds nuw i32, ptr %4399, i64 %idxprom.i23.i.i.i.i.i3215
  store atomic i32 %conv.i.i.i.i.i.i3214, ptr %arrayidx.i24.i.i.i.i.i3216 monotonic, align 4
  %4400 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !84
  %arrayidx.i.i5.i.i3217 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %4400, i64 %idxprom.i23.i.i.i.i.i3215
  %4401 = load i64, ptr %arrayidx.i.i5.i.i3217, align 8, !tbaa !80
  br label %while.cond.i.i.i.i3218, !amdgpu.uniform !42

while.cond.i.i.i.i3218:                           ; preds = %post-loop-memcpy-expansion5384, %Flow8008
  %phi.broken8725 = phi i64 [ 0, %post-loop-memcpy-expansion5384 ], [ %4423, %Flow8008 ]
  %lsr.iv6920 = phi i64 [ %4421, %Flow8008 ], [ -51, %post-loop-memcpy-expansion5384 ]
  %lsr.iv6915 = phi ptr addrspace(4) [ %4420, %Flow8008 ], [ getelementptr (i8, ptr addrspace(4) @.str.24.40, i64 56), %post-loop-memcpy-expansion5384 ]
  %port.sroa.37.0.i.i3219 = phi i32 [ %4419, %Flow8008 ], [ %conv.i.i.i.i.i.i3214, %post-loop-memcpy-expansion5384 ]
  %idx.0.i.i.i.i3220 = phi i64 [ %4418, %Flow8008 ], [ 56, %post-loop-memcpy-expansion5384 ]
  %cmp.i.i.i.i3221 = icmp ult i64 %idx.0.i.i.i.i3220, sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.24.40 to ptr), i64 5) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.24.40 to ptr) to i64))
  %4402 = call i64 @llvm.amdgcn.ballot.i64(i1 %cmp.i.i.i.i3221)
  %and.i.i.i.i6.i.i3222 = and i64 %4402, %4401
  %tobool.not.i.i.i.i3223 = icmp ne i64 %and.i.i.i.i6.i.i3222, 0
  %4403 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i22.i.i.i3224 = getelementptr inbounds nuw i32, ptr %4403, i64 %idxprom.i23.i.i.i.i.i3215
  %4404 = load atomic i32, ptr %arrayidx.i22.i.i.i3224 monotonic, align 4
  %4405 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %tobool.not.i.i.i.i3223)
  %4406 = extractvalue { i1, i64 } %4405, 0
  %4407 = extractvalue { i1, i64 } %4405, 1
  br i1 %4406, label %cond.false.i8.i.i.i.i3225, label %Flow8008

cond.false.i8.i.i.i.i3225:                        ; preds = %while.cond.i.i.i.i3218
  %4408 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %4404)
  %4409 = icmp ne i32 %4408, %port.sroa.37.0.i.i3219
  %4410 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %4409)
  %4411 = extractvalue { i1, i64 } %4410, 0
  %4412 = extractvalue { i1, i64 } %4410, 1
  br i1 %4411, label %while.body.i.i21.i.i.i.i3226.preheader, label %Flow8007

while.body.i.i21.i.i.i.i3226.preheader:           ; preds = %cond.false.i8.i.i.i.i3225
  br label %while.body.i.i21.i.i.i.i3226, !amdgpu.uniform !42

Flow8007:                                         ; preds = %Flow8006, %cond.false.i8.i.i.i.i3225
  call void @llvm.amdgcn.end.cf.i64(i64 %4412)
  br label %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i3229, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i3226:                     ; preds = %while.body.i.i21.i.i.i.i3226.preheader, %while.body.i.i21.i.i.i.i3226
  %phi.broken8724 = phi i64 [ 0, %while.body.i.i21.i.i.i.i3226.preheader ], [ %4416, %while.body.i.i21.i.i.i.i3226 ]
  call void @llvm.amdgcn.s.sleep(i32 2)
  %4413 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i22.i.i.i.i3227 = getelementptr inbounds nuw i32, ptr %4413, i64 %idxprom.i23.i.i.i.i.i3215
  %4414 = load atomic i32, ptr %arrayidx.i.i.i22.i.i.i.i3227 monotonic, align 4
  %4415 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %4414)
  %cmp.i.not.i.i23.i.i.i.i3228 = icmp eq i32 %4415, %port.sroa.37.0.i.i3219
  %4416 = call i64 @llvm.amdgcn.if.break.i64(i1 %cmp.i.not.i.i23.i.i.i.i3228, i64 %phi.broken8724)
  %4417 = call i1 @llvm.amdgcn.loop.i64(i64 %4416)
  br i1 %4417, label %Flow8006, label %while.body.i.i21.i.i.i.i3226

Flow8006:                                         ; preds = %while.body.i.i21.i.i.i.i3226
  %.lcssa9156 = phi i64 [ %4416, %while.body.i.i21.i.i.i.i3226 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa9156)
  br label %Flow8007, !amdgpu.uniform !42

Flow8008:                                         ; preds = %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i3230, %while.cond.i.i.i.i3218
  %4418 = phi i64 [ %add.i.i.i.i3233, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i3230 ], [ poison, %while.cond.i.i.i.i3218 ]
  %4419 = phi i32 [ %conv.i.i26.i.i.i.i3231, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i3230 ], [ poison, %while.cond.i.i.i.i3218 ]
  %4420 = phi ptr addrspace(4) [ %scevgep6916, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i3230 ], [ poison, %while.cond.i.i.i.i3218 ]
  %4421 = phi i64 [ %lsr.iv.next6921, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i3230 ], [ poison, %while.cond.i.i.i.i3218 ]
  %4422 = phi i1 [ false, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i3230 ], [ true, %while.cond.i.i.i.i3218 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %4407)
  %4423 = call i64 @llvm.amdgcn.if.break.i64(i1 %4422, i64 %phi.broken8725)
  %4424 = call i1 @llvm.amdgcn.loop.i64(i64 %4423)
  br i1 %4424, label %cond.false.i.i.i3240, label %while.cond.i.i.i.i3218

_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i3229: ; preds = %Flow8007
  fence acquire
  br i1 %cmp.i.i.i.i3221, label %if.then.i.i.i.i.i.i.i3234, label %Flow8005, !amdgpu.uniform !42

if.then.i.i.i.i.i.i.i3234:                        ; preds = %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i3229
  %sub.i.i.i.i.i.i.i3236 = sub nuw i64 sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.24.40 to ptr), i64 5) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.24.40 to ptr) to i64)), %idx.0.i.i.i.i3220
  %spec.select.i.i.i36.i.i.i.i3237 = call i64 @llvm.umin.i64(i64 %sub.i.i.i.i.i.i.i3236, i64 64)
  %4425 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %4426 = and i64 %spec.select.i.i.i36.i.i.i.i3237, 15
  %4427 = sub i64 %spec.select.i.i.i36.i.i.i.i3237, %4426
  %4428 = icmp ne i64 %4427, 0
  br i1 %4428, label %loop-memcpy-expansion5391.preheader, label %Flow8004, !amdgpu.uniform !42

loop-memcpy-expansion5391.preheader:              ; preds = %if.then.i.i.i.i.i.i.i3234
  %arrayidx.i20.i35.i.i.i.i3238 = getelementptr %"struct.rpc::Buffer", ptr %4425, i64 %idxprom.i19.i.i.i.i.i3208
  %arrayidx.i22.i38.i.i.i.i3239 = getelementptr %"struct.rpc::Buffer", ptr %arrayidx.i20.i35.i.i.i.i3238, i64 %sh_prom.i.i.i.i756
  br label %loop-memcpy-expansion5391, !amdgpu.uniform !42

Flow8004:                                         ; preds = %Flow8003, %if.then.i.i.i.i.i.i.i3234
  br label %loop-memcpy-residual-header5394, !amdgpu.uniform !42

loop-memcpy-expansion5391:                        ; preds = %loop-memcpy-expansion5391.preheader, %loop-memcpy-expansion5391
  %loop-index5392 = phi i64 [ %4431, %loop-memcpy-expansion5391 ], [ 0, %loop-memcpy-expansion5391.preheader ]
  %scevgep6917 = getelementptr i8, ptr addrspace(4) %lsr.iv6915, i64 %loop-index5392, !amdgpu.uniform !42
  %4429 = load <4 x i32>, ptr addrspace(4) %scevgep6917, align 1
  %4430 = getelementptr i8, ptr %arrayidx.i22.i38.i.i.i.i3239, i64 %loop-index5392
  store <4 x i32> %4429, ptr %4430, align 1
  %4431 = add i64 %loop-index5392, 16
  %4432 = icmp uge i64 %4431, %4427
  br i1 %4432, label %Flow8003, label %loop-memcpy-expansion5391, !amdgpu.uniform !42

Flow8002:                                         ; preds = %Flow8001, %loop-memcpy-residual-header5394
  br label %Flow8005, !amdgpu.uniform !42

loop-memcpy-residual5393:                         ; preds = %loop-memcpy-residual5393.preheader, %loop-memcpy-residual5393
  %residual-loop-index5395 = phi i64 [ %4434, %loop-memcpy-residual5393 ], [ 0, %loop-memcpy-residual5393.preheader ]
  %scevgep6926 = getelementptr i8, ptr addrspace(4) %scevgep6925, i64 %residual-loop-index5395, !amdgpu.uniform !42
  %4433 = load i8, ptr addrspace(4) %scevgep6926, align 1
  %scevgep6924 = getelementptr i8, ptr %scevgep6923, i64 %residual-loop-index5395
  store i8 %4433, ptr %scevgep6924, align 1
  %4434 = add i64 %residual-loop-index5395, 1
  %4435 = icmp uge i64 %4434, %4426
  br i1 %4435, label %Flow8001, label %loop-memcpy-residual5393, !amdgpu.uniform !42

Flow8001:                                         ; preds = %loop-memcpy-residual5393
  br label %Flow8002, !amdgpu.uniform !42

_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i3230: ; preds = %Flow8005
  %conv.i.i26.i.i.i.i3231 = xor i32 %port.sroa.37.0.i.i3219, 1
  fence release
  %4436 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %arrayidx.i24.i29.i.i.i.i3232 = getelementptr inbounds nuw i32, ptr %4436, i64 %idxprom.i23.i.i.i.i.i3215
  store atomic i32 %conv.i.i26.i.i.i.i3231, ptr %arrayidx.i24.i29.i.i.i.i3232 monotonic, align 4
  %add.i.i.i.i3233 = add i64 %idx.0.i.i.i.i3220, 64
  %scevgep6916 = getelementptr i8, ptr addrspace(4) %lsr.iv6915, i64 64
  %lsr.iv.next6921 = add i64 %lsr.iv6920, -64
  br label %Flow8008, !amdgpu.uniform !42

cond.false.i.i.i3240:                             ; preds = %Flow8008
  %.lcssa9158 = phi i64 [ %4423, %Flow8008 ]
  %port.sroa.37.0.i.i3219.lcssa = phi i32 [ %port.sroa.37.0.i.i3219, %Flow8008 ]
  %.lcssa9157 = phi i32 [ %4404, %Flow8008 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa9158)
  %4437 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %.lcssa9157)
  %4438 = icmp ne i32 %4437, %port.sroa.37.0.i.i3219.lcssa
  %4439 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %4438)
  %4440 = extractvalue { i1, i64 } %4439, 0
  %4441 = extractvalue { i1, i64 } %4439, 1
  br i1 %4440, label %while.body.i.i.i.i3241.preheader, label %Flow8000

while.body.i.i.i.i3241.preheader:                 ; preds = %cond.false.i.i.i3240
  br label %while.body.i.i.i.i3241, !amdgpu.uniform !42

Flow8000:                                         ; preds = %Flow7999, %cond.false.i.i.i3240
  call void @llvm.amdgcn.end.cf.i64(i64 %4441)
  br label %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit3262, !amdgpu.uniform !42

while.body.i.i.i.i3241:                           ; preds = %while.body.i.i.i.i3241.preheader, %while.body.i.i.i.i3241
  %phi.broken8726 = phi i64 [ 0, %while.body.i.i.i.i3241.preheader ], [ %4445, %while.body.i.i.i.i3241 ]
  call void @llvm.amdgcn.s.sleep(i32 2)
  %4442 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i8.i.i3242 = getelementptr inbounds nuw i32, ptr %4442, i64 %idxprom.i23.i.i.i.i.i3215
  %4443 = load atomic i32, ptr %arrayidx.i.i.i8.i.i3242 monotonic, align 4
  %4444 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %4443)
  %cmp.i.not.i.i.i.i3243 = icmp eq i32 %4444, %port.sroa.37.0.i.i3219.lcssa
  %4445 = call i64 @llvm.amdgcn.if.break.i64(i1 %cmp.i.not.i.i.i.i3243, i64 %phi.broken8726)
  %4446 = call i1 @llvm.amdgcn.loop.i64(i64 %4445)
  br i1 %4446, label %Flow7999, label %while.body.i.i.i.i3241

Flow7999:                                         ; preds = %while.body.i.i.i.i3241
  %.lcssa9159 = phi i64 [ %4445, %while.body.i.i.i.i3241 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa9159)
  br label %Flow8000, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit3262: ; preds = %Flow8000
  fence acquire
  call void @llvm.amdgcn.wave.barrier()
  fence syncscope("agent") release
  %4447 = call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %.lcssa9154, i1 true)
  %iszero.i.i.i.i.i13.i.i3245 = icmp ne i64 %.lcssa9154, 0
  %cmp2.i.i.i.i14.i.i3246 = icmp eq i64 %4447, %sh_prom.i.i.i.i756
  %cmp.i.i.i.i15.i.i3247 = select i1 %iszero.i.i.i.i.i13.i.i3245, i1 %cmp2.i.i.i.i14.i.i3246, i1 false
  %4448 = and i32 %.lcssa9155, 31
  %conv4.i.i.i19.i.i3251 = zext i1 %cmp.i.i.i.i15.i.i3247 to i32
  %shl.i.i.i20.i.i3252 = shl nuw i32 %conv4.i.i.i19.i.i3251, %4448
  %xor.i.i.i21.i.i3253 = xor i32 %shl.i.i.i20.i.i3252, -1
  %4449 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop5849, !amdgpu.uniform !42

4450:                                             ; preds = %ComputeEnd5850
  %div8.i.i.i16.i.i3248 = lshr i32 %.lcssa9155, 5
  %idxprom.i.i.i17.i.i3249 = zext nneg i32 %div8.i.i.i16.i.i3248 to i64
  %4451 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i17.i.i3249
  %sunkaddr7468 = getelementptr inbounds i8, ptr addrspace(1) %4451, i64 40
  %4452 = atomicrmw and ptr addrspace(1) %sunkaddr7468, i32 %.lcssa9160 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %4453, !amdgpu.uniform !42

4453:                                             ; preds = %4450, %ComputeEnd5850
  call void @llvm.amdgcn.end.cf.i64(i64 %8308)
  call void @llvm.amdgcn.wave.barrier()
  br label %Flow8063, !amdgpu.uniform !42

sw.epilog:                                        ; preds = %Flow8331
  call void @llvm.lifetime.end.p5(ptr addrspace(5) %Ctx) #34
  br label %Flow8429, !amdgpu.uniform !42

cleanup:                                          ; preds = %Flow8429
  %Next = getelementptr inbounds nuw i8, ptr %T.0207, i64 8
  %T.0 = load ptr, ptr %Next, align 8, !tbaa !63
  %cmp17.not = icmp eq ptr %T.0, null
  %4454 = call i64 @llvm.amdgcn.if.break.i64(i1 %cmp17.not, i64 %phi.broken8766)
  %4455 = call i1 @llvm.amdgcn.loop.i64(i64 %4454)
  br i1 %4455, label %Flow8430, label %for.body

Flow7917:                                         ; preds = %5820, %Flow7916
  call void @llvm.amdgcn.end.cf.i64(i64 %5823)
  br label %if.end111, !amdgpu.uniform !42

for.cond.i.i1.i3271:                              ; preds = %for.cond.i.i1.i3271.preheader, %Flow7915
  %4456 = phi i32 [ %4493, %Flow7915 ], [ poison, %for.cond.i.i1.i3271.preheader ]
  %4457 = phi i64 [ %4494, %Flow7915 ], [ poison, %for.cond.i.i1.i3271.preheader ]
  %4458 = phi i32 [ %4495, %Flow7915 ], [ poison, %for.cond.i.i1.i3271.preheader ]
  %retval.sroa.6.0.i.i.i3272 = phi i32 [ %4498, %Flow7915 ], [ undef, %for.cond.i.i1.i3271.preheader ]
  %retval.sroa.2.0.i.i.i3274 = phi i64 [ %4497, %Flow7915 ], [ undef, %for.cond.i.i1.i3271.preheader ]
  %retval.sroa.8.0.i.i.i3275 = phi i32 [ %4496, %Flow7915 ], [ undef, %for.cond.i.i1.i3271.preheader ]
  %index.0.i.i.i3276 = phi i32 [ %4500, %Flow7915 ], [ 0, %for.cond.i.i1.i3271.preheader ]
  %4459 = load i32, ptr addrspace(1) @__llvm_rpc_client, align 8, !tbaa !68
  %cmp.not.i.i.i3277 = icmp ult i32 %index.0.i.i.i3276, %4459
  %spec.store.select.i.i.i3278 = select i1 %cmp.not.i.i.i3277, i32 %index.0.i.i.i3276, i32 0
  %4460 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  %4461 = and i32 %spec.store.select.i.i.i3278, 31
  %4462 = lshr i64 %4460, %sh_prom.i.i.i.i3270
  %4463 = trunc i64 %4462 to i32
  %conv4.i.i.i.i.i3282 = and i32 %4463, 1
  %shl.i.i.i.i.i3283 = shl nuw i32 %conv4.i.i.i.i.i3282, %4461
  br label %ComputeLoop5853, !amdgpu.uniform !42

4464:                                             ; preds = %ComputeEnd5854
  %sunkaddr7469 = getelementptr inbounds i8, ptr addrspace(1) %8318, i64 40
  %4465 = atomicrmw or ptr addrspace(1) %sunkaddr7469, i32 %.lcssa8927 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %4466, !amdgpu.uniform !42

4466:                                             ; preds = %4464, %ComputeEnd5854
  %4467 = phi i32 [ %4465, %4464 ], [ poison, %ComputeEnd5854 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %8327)
  %4468 = call i32 @llvm.amdgcn.readfirstlane.i32(i32 %4467)
  %4469 = or i32 %4468, %.lcssa8928
  %shl5.i.i.i.i.i3284 = shl nuw i32 1, %4461
  %and.i.i.i.i.i3285 = and i32 %shl5.i.i.i.i.i3284, %4469
  %tobool3.i.i.i.i3286 = icmp ne i32 %and.i.i.i.i.i3285, 0
  %4470 = call i64 @llvm.amdgcn.ballot.i64(i1 %tobool3.i.i.i.i3286)
  %and.i.i.i.i.i.i3287 = and i64 %4470, %4460
  %cmp.i.not.i.i.i3288 = icmp ne i64 %4460, %and.i.i.i.i.i.i3287
  br i1 %cmp.i.not.i.i.i3288, label %if.then.i.i.i.i3289, label %Flow7914, !amdgpu.uniform !42

if.then.i.i.i.i3289:                              ; preds = %4466
  fence syncscope("agent") acquire
  %4471 = load <2 x i64>, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !75
  %4472 = extractelement <2 x i64> %4471, i32 0
  %4473 = inttoptr i64 %4472 to ptr
  %4474 = extractelement <2 x i64> %4471, i32 1
  %4475 = inttoptr i64 %4474 to ptr
  %idxprom.i.i.i.i3290 = zext i32 %spec.store.select.i.i.i3278 to i64
  %arrayidx.i.i.i.i3291 = getelementptr inbounds nuw i32, ptr %4473, i64 %idxprom.i.i.i.i3290, !amdgpu.uniform !42
  %4476 = load atomic i32, ptr %arrayidx.i.i.i.i3291 monotonic, align 4
  %4477 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %4476)
  %arrayidx.i45.i.i.i3292 = getelementptr inbounds nuw i32, ptr %4475, i64 %idxprom.i.i.i.i3290, !amdgpu.uniform !42
  %4478 = load atomic i32, ptr %arrayidx.i45.i.i.i3292 monotonic, align 4
  %4479 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %4478)
  %cmp.i46.not.i.i.i3293 = icmp ne i32 %4477, %4479
  br i1 %cmp.i46.not.i.i.i3293, label %if.then12.i.i.i3294, label %Flow7912, !amdgpu.uniform !42

if.then12.i.i.i3294:                              ; preds = %if.then.i.i.i.i3289
  fence syncscope("agent") release
  %4480 = call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %4460, i1 true)
  %iszero.i.i.i.i.i.i.i3295 = icmp ne i64 %4460, 0
  %cmp2.i.i.i.i.i.i3296 = icmp eq i64 %4480, %sh_prom.i.i.i.i3270
  %cmp.i.i.i.i.i.i3297 = select i1 %iszero.i.i.i.i.i.i.i3295, i1 %cmp2.i.i.i.i.i.i3296, i1 false
  %conv4.i.i51.i.i.i3298 = zext i1 %cmp.i.i.i.i.i.i3297 to i32
  %shl.i.i52.i.i.i3299 = shl nuw i32 %conv4.i.i51.i.i.i3298, %4461
  %xor.i.i.i.i.i3300 = xor i32 %shl.i.i52.i.i.i3299, -1
  %4481 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop5858, !amdgpu.uniform !42

4482:                                             ; preds = %ComputeEnd5859
  %sunkaddr7470 = getelementptr inbounds i8, ptr addrspace(1) %8318, i64 40
  %4483 = atomicrmw and ptr addrspace(1) %sunkaddr7470, i32 %.lcssa8929 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %4484, !amdgpu.uniform !42

4484:                                             ; preds = %4482, %ComputeEnd5859
  call void @llvm.amdgcn.end.cf.i64(i64 %8345)
  call void @llvm.amdgcn.wave.barrier()
  br label %Flow7912, !amdgpu.uniform !42

if.end14.i.i.i3360:                               ; preds = %Flow7912
  %4485 = call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %4460, i1 true)
  %iszero.i.i.i.i.i.i3361 = icmp ne i64 %4460, 0
  %cmp2.i.i.i.i.i3362 = icmp eq i64 %4485, %sh_prom.i.i.i.i3270
  %cmp.i.i.i.i.i3363 = select i1 %iszero.i.i.i.i.i.i3361, i1 %cmp2.i.i.i.i.i3362, i1 false
  %4486 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.i.i.i.i.i3363)
  %4487 = extractvalue { i1, i64 } %4486, 0
  %4488 = extractvalue { i1, i64 } %4486, 1
  br i1 %4487, label %if.then16.i.i.i3365, label %if.end22.i.i.i3364

if.then16.i.i.i3365:                              ; preds = %if.end14.i.i.i3360
  %4489 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !76
  %opcode.i.i.i3366 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %4489, i64 %idxprom.i.i.i.i3290, i32 1
  store i32 1660944387, ptr %opcode.i.i.i3366, align 8, !tbaa !77
  %arrayidx21.i.i.i3367 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %4489, i64 %idxprom.i.i.i.i3290
  store i64 %4460, ptr %arrayidx21.i.i.i3367, align 8, !tbaa !80
  br label %if.end22.i.i.i3364, !amdgpu.uniform !42

Flow7913:                                         ; preds = %if.end22.i.i.i3364, %Flow7912
  %4490 = phi i32 [ %4477, %if.end22.i.i.i3364 ], [ %retval.sroa.8.0.i.i.i3275, %Flow7912 ]
  %4491 = phi i64 [ %4460, %if.end22.i.i.i3364 ], [ %retval.sroa.2.0.i.i.i3274, %Flow7912 ]
  %4492 = phi i32 [ %spec.store.select.i.i.i3278, %if.end22.i.i.i3364 ], [ %retval.sroa.6.0.i.i.i3272, %Flow7912 ]
  br label %cleanup26.i.i.i3301, !amdgpu.uniform !42

if.end22.i.i.i3364:                               ; preds = %if.then16.i.i.i3365, %if.end14.i.i.i3360
  call void @llvm.amdgcn.end.cf.i64(i64 %4488)
  call void @llvm.amdgcn.wave.barrier()
  br label %Flow7913, !amdgpu.uniform !42

Flow7914:                                         ; preds = %cleanup26.i.i.i3301, %4466
  %4493 = phi i32 [ %4492, %cleanup26.i.i.i3301 ], [ %4456, %4466 ]
  %4494 = phi i64 [ %4491, %cleanup26.i.i.i3301 ], [ %4457, %4466 ]
  %4495 = phi i32 [ %4490, %cleanup26.i.i.i3301 ], [ %4458, %4466 ]
  %4496 = phi i32 [ %4490, %cleanup26.i.i.i3301 ], [ %retval.sroa.8.0.i.i.i3275, %4466 ]
  %4497 = phi i64 [ %4491, %cleanup26.i.i.i3301 ], [ %retval.sroa.2.0.i.i.i3274, %4466 ]
  %4498 = phi i32 [ %4492, %cleanup26.i.i.i3301 ], [ %retval.sroa.6.0.i.i.i3272, %4466 ]
  %4499 = phi i1 [ %cmp.i46.not.i.i.i3293, %cleanup26.i.i.i3301 ], [ true, %4466 ]
  br i1 %4499, label %for.inc.i.i.i3306, label %Flow7915, !amdgpu.uniform !42

cleanup26.i.i.i3301:                              ; preds = %Flow7913
  br label %Flow7914, !amdgpu.uniform !42

for.inc.i.i.i3306:                                ; preds = %Flow7914
  %inc.i.i.i3311 = add i32 %spec.store.select.i.i.i3278, 1
  br label %Flow7915, !amdgpu.uniform !42

Flow7915:                                         ; preds = %for.inc.i.i.i3306, %Flow7914
  %4500 = phi i32 [ %inc.i.i.i3311, %for.inc.i.i.i3306 ], [ poison, %Flow7914 ]
  %4501 = phi i1 [ false, %for.inc.i.i.i3306 ], [ true, %Flow7914 ]
  br i1 %4501, label %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i3312, label %for.cond.i.i1.i3271, !amdgpu.uniform !42

_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i3312: ; preds = %Flow7915
  %.lcssa8932 = phi i32 [ %4493, %Flow7915 ]
  %.lcssa8931 = phi i64 [ %4494, %Flow7915 ]
  %.lcssa8930 = phi i32 [ %4495, %Flow7915 ]
  fence acquire
  %4502 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %mul.i.i.i.i.i.i3313 = shl i32 %.lcssa8932, 6
  %idxprom.i19.i.i.i.i.i3314 = zext i32 %mul.i.i.i.i.i.i3313 to i64
  %arrayidx.i20.i.i.i.i.i3315 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %4502, i64 %idxprom.i19.i.i.i.i.i3314
  %arrayidx.i22.i.i.i.i.i3316 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %arrayidx.i20.i.i.i.i.i3315, i64 %sh_prom.i.i.i.i3270
  store i64 sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.25.41 to ptr), i64 4) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.25.41 to ptr) to i64)), ptr %arrayidx.i22.i.i.i.i.i3316, align 8, !tbaa !82
  %spec.select.i.i.i.i.i.i.i3317 = call i64 @llvm.umin.i64(i64 sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.25.41 to ptr), i64 4) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.25.41 to ptr) to i64)), i64 56)
  %4503 = and i64 %spec.select.i.i.i.i.i.i.i3317, 15
  %4504 = sub i64 %spec.select.i.i.i.i.i.i.i3317, %4503
  %4505 = icmp ne i64 %4504, 0
  br i1 %4505, label %loop-memcpy-expansion5397.preheader, label %Flow7911, !amdgpu.uniform !42

loop-memcpy-expansion5397.preheader:              ; preds = %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i3312
  %4506 = shl nuw nsw i64 %sh_prom.i.i.i.i3270, 6
  %4507 = add nuw nsw i64 %4506, 8
  %scevgep6496 = getelementptr i8, ptr %4502, i64 %4507
  %4508 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i3314, 6
  %scevgep6497 = getelementptr i8, ptr %scevgep6496, i64 %4508
  br label %loop-memcpy-expansion5397, !amdgpu.uniform !42

Flow7911:                                         ; preds = %Flow7910, %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i3312
  br label %loop-memcpy-residual-header5400, !amdgpu.uniform !42

loop-memcpy-expansion5397:                        ; preds = %loop-memcpy-expansion5397.preheader, %loop-memcpy-expansion5397
  %loop-index5398 = phi i64 [ %4511, %loop-memcpy-expansion5397 ], [ 0, %loop-memcpy-expansion5397.preheader ]
  %4509 = getelementptr i8, ptr addrspace(4) @.str.25.41, i64 %loop-index5398, !amdgpu.uniform !42
  %4510 = load <4 x i32>, ptr addrspace(4) %4509, align 1
  %scevgep6498 = getelementptr i8, ptr %scevgep6497, i64 %loop-index5398
  store <4 x i32> %4510, ptr %scevgep6498, align 1
  %4511 = add i64 %loop-index5398, 16
  %4512 = icmp uge i64 %4511, %4504
  br i1 %4512, label %Flow7910, label %loop-memcpy-expansion5397, !amdgpu.uniform !42

Flow7909:                                         ; preds = %Flow7907, %Flow7908
  %4513 = phi i64 [ %6643, %Flow7907 ], [ %6640, %Flow7908 ]
  %4514 = phi i64 [ %6645, %Flow7907 ], [ %6641, %Flow7908 ]
  br label %post-loop-memcpy-expansion5396, !amdgpu.uniform !42

loop-memcpy-residual5399:                         ; preds = %loop-memcpy-residual5399.preheader, %loop-memcpy-residual5399
  %residual-loop-index5401 = phi i64 [ %4516, %loop-memcpy-residual5399 ], [ 0, %loop-memcpy-residual5399.preheader ]
  %scevgep6495 = getelementptr i8, ptr addrspace(4) @.str.25.41, i64 %residual-loop-index5401, !amdgpu.uniform !42
  %4515 = load i8, ptr addrspace(4) %scevgep6495, align 1
  %scevgep6494 = getelementptr i8, ptr %scevgep6493, i64 %residual-loop-index5401
  store i8 %4515, ptr %scevgep6494, align 1
  %4516 = add i64 %residual-loop-index5401, 1
  %4517 = icmp uge i64 %4516, %4503
  br i1 %4517, label %Flow7907, label %loop-memcpy-residual5399, !amdgpu.uniform !42

Flow7907:                                         ; preds = %loop-memcpy-residual5399
  br label %Flow7909, !amdgpu.uniform !42

post-loop-memcpy-expansion5396:                   ; preds = %Flow7909
  %tobool.not.i.i.i.i.i.i3319 = icmp eq i32 %.lcssa8930, 0
  %conv.i.i.i.i.i.i3320 = zext i1 %tobool.not.i.i.i.i.i.i3319 to i32
  fence release
  %4518 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %idxprom.i23.i.i.i.i.i3321 = zext i32 %.lcssa8932 to i64
  %arrayidx.i24.i.i.i.i.i3322 = getelementptr inbounds nuw i32, ptr %4518, i64 %idxprom.i23.i.i.i.i.i3321
  store atomic i32 %conv.i.i.i.i.i.i3320, ptr %arrayidx.i24.i.i.i.i.i3322 monotonic, align 4
  %4519 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !84
  %arrayidx.i.i5.i.i3323 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %4519, i64 %idxprom.i23.i.i.i.i.i3321, !amdgpu.uniform !42
  %4520 = load i64, ptr %arrayidx.i.i5.i.i3323, align 8, !tbaa !80
  br label %while.cond.i.i.i.i3324, !amdgpu.uniform !42

while.cond.i.i.i.i3324:                           ; preds = %post-loop-memcpy-expansion5396, %Flow7906
  %phi.broken8778 = phi i64 [ 0, %post-loop-memcpy-expansion5396 ], [ %4537, %Flow7906 ]
  %lsr.iv6485 = phi i64 [ %4535, %Flow7906 ], [ -52, %post-loop-memcpy-expansion5396 ]
  %lsr.iv6480 = phi ptr addrspace(4) [ %4534, %Flow7906 ], [ getelementptr (i8, ptr addrspace(4) @.str.25.41, i64 56), %post-loop-memcpy-expansion5396 ]
  %port.sroa.37.0.i.i3325 = phi i32 [ %4533, %Flow7906 ], [ %conv.i.i.i.i.i.i3320, %post-loop-memcpy-expansion5396 ]
  %idx.0.i.i.i.i3326 = phi i64 [ %4532, %Flow7906 ], [ 56, %post-loop-memcpy-expansion5396 ]
  %cmp.i.i.i.i3327 = icmp ult i64 %idx.0.i.i.i.i3326, sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.25.41 to ptr), i64 4) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.25.41 to ptr) to i64))
  %4521 = call i64 @llvm.amdgcn.ballot.i64(i1 %cmp.i.i.i.i3327)
  %and.i.i.i.i6.i.i3328 = and i64 %4521, %4520
  %tobool.not.i.i.i.i3329 = icmp ne i64 %and.i.i.i.i6.i.i3328, 0
  %4522 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i22.i.i.i3330 = getelementptr inbounds nuw i32, ptr %4522, i64 %idxprom.i23.i.i.i.i.i3321, !amdgpu.uniform !42
  %4523 = load atomic i32, ptr %arrayidx.i22.i.i.i3330 monotonic, align 4
  %4524 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %tobool.not.i.i.i.i3329)
  %4525 = extractvalue { i1, i64 } %4524, 0
  %4526 = extractvalue { i1, i64 } %4524, 1
  br i1 %4525, label %cond.false.i8.i.i.i.i3331, label %Flow7906

cond.false.i8.i.i.i.i3331:                        ; preds = %while.cond.i.i.i.i3324
  %4527 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %4523)
  %4528 = icmp ne i32 %4527, %port.sroa.37.0.i.i3325
  br i1 %4528, label %while.body.i.i21.i.i.i.i3332.preheader, label %Flow7905, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i3332.preheader:           ; preds = %cond.false.i8.i.i.i.i3331
  br label %while.body.i.i21.i.i.i.i3332, !amdgpu.uniform !42

Flow7905:                                         ; preds = %Flow7904, %cond.false.i8.i.i.i.i3331
  br label %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i3335, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i3332:                     ; preds = %while.body.i.i21.i.i.i.i3332.preheader, %while.body.i.i21.i.i.i.i3332
  call void @llvm.amdgcn.s.sleep(i32 2)
  %4529 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i22.i.i.i.i3333 = getelementptr inbounds nuw i32, ptr %4529, i64 %idxprom.i23.i.i.i.i.i3321, !amdgpu.uniform !42
  %4530 = load atomic i32, ptr %arrayidx.i.i.i22.i.i.i.i3333 monotonic, align 4
  %4531 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %4530)
  %cmp.i.not.i.i23.i.i.i.i3334 = icmp eq i32 %4531, %port.sroa.37.0.i.i3325
  br i1 %cmp.i.not.i.i23.i.i.i.i3334, label %Flow7904, label %while.body.i.i21.i.i.i.i3332, !amdgpu.uniform !42

Flow7904:                                         ; preds = %while.body.i.i21.i.i.i.i3332
  br label %Flow7905, !amdgpu.uniform !42

Flow7906:                                         ; preds = %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i3336, %while.cond.i.i.i.i3324
  %4532 = phi i64 [ %add.i.i.i.i3339, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i3336 ], [ poison, %while.cond.i.i.i.i3324 ]
  %4533 = phi i32 [ %conv.i.i26.i.i.i.i3337, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i3336 ], [ poison, %while.cond.i.i.i.i3324 ]
  %4534 = phi ptr addrspace(4) [ %scevgep6481, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i3336 ], [ poison, %while.cond.i.i.i.i3324 ]
  %4535 = phi i64 [ %lsr.iv.next6486, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i3336 ], [ poison, %while.cond.i.i.i.i3324 ]
  %4536 = phi i1 [ false, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i3336 ], [ true, %while.cond.i.i.i.i3324 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %4526)
  %4537 = call i64 @llvm.amdgcn.if.break.i64(i1 %4536, i64 %phi.broken8778)
  %4538 = call i1 @llvm.amdgcn.loop.i64(i64 %4537)
  br i1 %4538, label %cond.false.i.i.i3346, label %while.cond.i.i.i.i3324

_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i3335: ; preds = %Flow7905
  fence acquire
  br i1 %cmp.i.i.i.i3327, label %if.then.i.i.i.i.i.i.i3340, label %Flow7903, !amdgpu.uniform !42

if.then.i.i.i.i.i.i.i3340:                        ; preds = %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i3335
  %sub.i.i.i.i.i.i.i3342 = sub nuw i64 sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.25.41 to ptr), i64 4) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.25.41 to ptr) to i64)), %idx.0.i.i.i.i3326
  %spec.select.i.i.i36.i.i.i.i3343 = call i64 @llvm.umin.i64(i64 %sub.i.i.i.i.i.i.i3342, i64 64)
  %4539 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %4540 = and i64 %spec.select.i.i.i36.i.i.i.i3343, 15
  %4541 = sub i64 %spec.select.i.i.i36.i.i.i.i3343, %4540
  %4542 = icmp ne i64 %4541, 0
  br i1 %4542, label %loop-memcpy-expansion5403.preheader, label %Flow7902, !amdgpu.uniform !42

loop-memcpy-expansion5403.preheader:              ; preds = %if.then.i.i.i.i.i.i.i3340
  %arrayidx.i20.i35.i.i.i.i3344 = getelementptr %"struct.rpc::Buffer", ptr %4539, i64 %idxprom.i19.i.i.i.i.i3314
  %arrayidx.i22.i38.i.i.i.i3345 = getelementptr %"struct.rpc::Buffer", ptr %arrayidx.i20.i35.i.i.i.i3344, i64 %sh_prom.i.i.i.i3270
  br label %loop-memcpy-expansion5403, !amdgpu.uniform !42

Flow7902:                                         ; preds = %Flow7901, %if.then.i.i.i.i.i.i.i3340
  br label %loop-memcpy-residual-header5406, !amdgpu.uniform !42

loop-memcpy-expansion5403:                        ; preds = %loop-memcpy-expansion5403.preheader, %loop-memcpy-expansion5403
  %loop-index5404 = phi i64 [ %4545, %loop-memcpy-expansion5403 ], [ 0, %loop-memcpy-expansion5403.preheader ]
  %scevgep6482 = getelementptr i8, ptr addrspace(4) %lsr.iv6480, i64 %loop-index5404, !amdgpu.uniform !42
  %4543 = load <4 x i32>, ptr addrspace(4) %scevgep6482, align 1
  %4544 = getelementptr i8, ptr %arrayidx.i22.i38.i.i.i.i3345, i64 %loop-index5404
  store <4 x i32> %4543, ptr %4544, align 1
  %4545 = add i64 %loop-index5404, 16
  %4546 = icmp uge i64 %4545, %4541
  br i1 %4546, label %Flow7901, label %loop-memcpy-expansion5403, !amdgpu.uniform !42

Flow7900:                                         ; preds = %Flow7899, %loop-memcpy-residual-header5406
  br label %Flow7903, !amdgpu.uniform !42

loop-memcpy-residual5405:                         ; preds = %loop-memcpy-residual5405.preheader, %loop-memcpy-residual5405
  %residual-loop-index5407 = phi i64 [ %4548, %loop-memcpy-residual5405 ], [ 0, %loop-memcpy-residual5405.preheader ]
  %scevgep6491 = getelementptr i8, ptr addrspace(4) %scevgep6490, i64 %residual-loop-index5407, !amdgpu.uniform !42
  %4547 = load i8, ptr addrspace(4) %scevgep6491, align 1
  %scevgep6489 = getelementptr i8, ptr %scevgep6488, i64 %residual-loop-index5407
  store i8 %4547, ptr %scevgep6489, align 1
  %4548 = add i64 %residual-loop-index5407, 1
  %4549 = icmp uge i64 %4548, %4540
  br i1 %4549, label %Flow7899, label %loop-memcpy-residual5405, !amdgpu.uniform !42

Flow7899:                                         ; preds = %loop-memcpy-residual5405
  br label %Flow7900, !amdgpu.uniform !42

_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i3336: ; preds = %Flow7903
  %conv.i.i26.i.i.i.i3337 = xor i32 %port.sroa.37.0.i.i3325, 1
  fence release
  %4550 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %arrayidx.i24.i29.i.i.i.i3338 = getelementptr inbounds nuw i32, ptr %4550, i64 %idxprom.i23.i.i.i.i.i3321
  store atomic i32 %conv.i.i26.i.i.i.i3337, ptr %arrayidx.i24.i29.i.i.i.i3338 monotonic, align 4
  %add.i.i.i.i3339 = add i64 %idx.0.i.i.i.i3326, 64
  %scevgep6481 = getelementptr i8, ptr addrspace(4) %lsr.iv6480, i64 64
  %lsr.iv.next6486 = add i64 %lsr.iv6485, -64
  br label %Flow7906, !amdgpu.uniform !42

cond.false.i.i.i3346:                             ; preds = %Flow7906
  %.lcssa8926 = phi i64 [ %4537, %Flow7906 ]
  %port.sroa.37.0.i.i3325.lcssa = phi i32 [ %port.sroa.37.0.i.i3325, %Flow7906 ]
  %.lcssa8925 = phi i32 [ %4523, %Flow7906 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa8926)
  %4551 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %.lcssa8925)
  %4552 = icmp ne i32 %4551, %port.sroa.37.0.i.i3325.lcssa
  %4553 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %4552)
  %4554 = extractvalue { i1, i64 } %4553, 0
  %4555 = extractvalue { i1, i64 } %4553, 1
  br i1 %4554, label %while.body.i.i.i.i3347.preheader, label %Flow7898

while.body.i.i.i.i3347.preheader:                 ; preds = %cond.false.i.i.i3346
  br label %while.body.i.i.i.i3347, !amdgpu.uniform !42

Flow7898:                                         ; preds = %Flow7897, %cond.false.i.i.i3346
  call void @llvm.amdgcn.end.cf.i64(i64 %4555)
  br label %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit3368, !amdgpu.uniform !42

while.body.i.i.i.i3347:                           ; preds = %while.body.i.i.i.i3347.preheader, %while.body.i.i.i.i3347
  %phi.broken8779 = phi i64 [ 0, %while.body.i.i.i.i3347.preheader ], [ %4559, %while.body.i.i.i.i3347 ]
  call void @llvm.amdgcn.s.sleep(i32 2)
  %4556 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i8.i.i3348 = getelementptr inbounds nuw i32, ptr %4556, i64 %idxprom.i23.i.i.i.i.i3321, !amdgpu.uniform !42
  %4557 = load atomic i32, ptr %arrayidx.i.i.i8.i.i3348 monotonic, align 4
  %4558 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %4557)
  %cmp.i.not.i.i.i.i3349 = icmp eq i32 %4558, %port.sroa.37.0.i.i3325.lcssa
  %4559 = call i64 @llvm.amdgcn.if.break.i64(i1 %cmp.i.not.i.i.i.i3349, i64 %phi.broken8779)
  %4560 = call i1 @llvm.amdgcn.loop.i64(i64 %4559)
  br i1 %4560, label %Flow7897, label %while.body.i.i.i.i3347

Flow7897:                                         ; preds = %while.body.i.i.i.i3347
  %.lcssa8924 = phi i64 [ %4559, %while.body.i.i.i.i3347 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa8924)
  br label %Flow7898, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit3368: ; preds = %Flow7898
  fence acquire
  call void @llvm.amdgcn.wave.barrier()
  fence syncscope("agent") release
  %4561 = call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %.lcssa8931, i1 true)
  %iszero.i.i.i.i.i13.i.i3351 = icmp ne i64 %.lcssa8931, 0
  %cmp2.i.i.i.i14.i.i3352 = icmp eq i64 %4561, %sh_prom.i.i.i.i3270
  %cmp.i.i.i.i15.i.i3353 = select i1 %iszero.i.i.i.i.i13.i.i3351, i1 %cmp2.i.i.i.i14.i.i3352, i1 false
  %4562 = and i32 %.lcssa8932, 31
  %conv4.i.i.i19.i.i3357 = zext i1 %cmp.i.i.i.i15.i.i3353 to i32
  %shl.i.i.i20.i.i3358 = shl nuw i32 %conv4.i.i.i19.i.i3357, %4562
  %xor.i.i.i21.i.i3359 = xor i32 %shl.i.i.i20.i.i3358, -1
  %4563 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop5862, !amdgpu.uniform !42

4564:                                             ; preds = %ComputeEnd5863
  %div8.i.i.i16.i.i3354 = lshr i32 %.lcssa8932, 5
  %idxprom.i.i.i17.i.i3355 = zext nneg i32 %div8.i.i.i16.i.i3354 to i64
  %4565 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i17.i.i3355
  %sunkaddr7471 = getelementptr inbounds i8, ptr addrspace(1) %4565, i64 40
  %4566 = atomicrmw and ptr addrspace(1) %sunkaddr7471, i32 %.lcssa8923 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %4567, !amdgpu.uniform !42

4567:                                             ; preds = %4564, %ComputeEnd5863
  call void @llvm.amdgcn.end.cf.i64(i64 %8362)
  call void @llvm.amdgcn.wave.barrier()
  call void @llvm.lifetime.start.p5(ptr addrspace(5) %buffer.i.i.i2988) #34, !noalias !180
  br label %while.body.i.i.i.i.i.i3373, !amdgpu.uniform !42

while.body.i.i.i.i.i.i3373:                       ; preds = %4567, %_ZN22__llvm_libc_22_0_0_git7details22StringBufferWriterImplILb0EE4pushEc.exit.i.i.i.i.i.i3387
  %phi.broken8780 = phi i64 [ 0, %4567 ], [ %4635, %_ZN22__llvm_libc_22_0_0_git7details22StringBufferWriterImplILb0EE4pushEc.exit.i.i.i.i.i.i3387 ]
  %writer.sroa.10.0.i.i.i.i3374 = phi i64 [ %writer.sroa.10.1.i.i.i.i3388, %_ZN22__llvm_libc_22_0_0_git7details22StringBufferWriterImplILb0EE4pushEc.exit.i.i.i.i.i.i3387 ], [ 0, %4567 ]
  %value.addr.09.i.i.i.i.i.i3375 = phi i32 [ %div.i.i.i.i.i.i.i3376, %_ZN22__llvm_libc_22_0_0_git7details22StringBufferWriterImplILb0EE4pushEc.exit.i.i.i.i.i.i3387 ], [ %1148, %4567 ]
  br label %NodeBlock7595, !amdgpu.uniform !42

NodeBlock7595:                                    ; preds = %while.body.i.i.i.i.i.i3373
  %div.i.i.i.i.i.i.i3376 = udiv i32 %value.addr.09.i.i.i.i.i.i3375, 10
  %.neg.i3377 = mul i32 %div.i.i.i.i.i.i.i3376, 246
  %rem.i.i.i.i.i.i.decomposed.i3378 = add i32 %.neg.i3377, %value.addr.09.i.i.i.i.i.i3375
  %conv.i.i.i.i.i.i.i3379 = trunc i32 %rem.i.i.i.i.i.i.decomposed.i3378 to i8
  %4568 = zext i8 %conv.i.i.i.i.i.i.i3379 to i16
  %Pivot7596 = icmp sge i16 %4568, 5
  %4569 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %Pivot7596)
  %4570 = extractvalue { i1, i64 } %4569, 0
  %4571 = extractvalue { i1, i64 } %4569, 1
  br i1 %4570, label %NodeBlock7593, label %Flow7896

NodeBlock7593:                                    ; preds = %NodeBlock7595
  %Pivot7594 = icmp sge i16 %4568, 7
  %4572 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %Pivot7594)
  %4573 = extractvalue { i1, i64 } %4572, 0
  %4574 = extractvalue { i1, i64 } %4572, 1
  br i1 %4573, label %NodeBlock7591, label %Flow7887

NodeBlock7591:                                    ; preds = %NodeBlock7593
  %Pivot7592 = icmp sge i16 %4568, 8
  %4575 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %Pivot7592)
  %4576 = extractvalue { i1, i64 } %4575, 0
  %4577 = extractvalue { i1, i64 } %4575, 1
  br i1 %4576, label %NodeBlock7589, label %Flow7883

NodeBlock7589:                                    ; preds = %NodeBlock7591
  %Pivot7590 = icmp sge i16 %4568, 9
  %4578 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %Pivot7590)
  %4579 = extractvalue { i1, i64 } %4578, 0
  %4580 = extractvalue { i1, i64 } %4578, 1
  br i1 %4579, label %sw.bb9.i.i.i.i.i.i.i.i3380, label %Flow7881

Flow7887:                                         ; preds = %Flow7884, %NodeBlock7593
  %4581 = phi i8 [ %4625, %Flow7884 ], [ poison, %NodeBlock7593 ]
  %4582 = call { i1, i64 } @llvm.amdgcn.else.i64.i64(i64 %4574)
  %4583 = extractvalue { i1, i64 } %4582, 0
  %4584 = extractvalue { i1, i64 } %4582, 1
  br i1 %4583, label %NodeBlock7587, label %Flow7888

NodeBlock7587:                                    ; preds = %Flow7887
  %Pivot7588 = icmp sge i16 %4568, 6
  %4585 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %Pivot7588)
  %4586 = extractvalue { i1, i64 } %4585, 0
  %4587 = extractvalue { i1, i64 } %4585, 1
  br i1 %4586, label %sw.bb6.i.i.i.i.i.i.i.i3540, label %Flow7885

Flow7896:                                         ; preds = %Flow7888, %NodeBlock7595
  %4588 = phi i8 [ %4627, %Flow7888 ], [ poison, %NodeBlock7595 ]
  %4589 = call { i1, i64 } @llvm.amdgcn.else.i64.i64(i64 %4571)
  %4590 = extractvalue { i1, i64 } %4589, 0
  %4591 = extractvalue { i1, i64 } %4589, 1
  br i1 %4590, label %NodeBlock7585, label %if.then.i.i.i.i.i.i.i3381

NodeBlock7585:                                    ; preds = %Flow7896
  %Pivot7586 = icmp sge i16 %4568, 3
  %4592 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %Pivot7586)
  %4593 = extractvalue { i1, i64 } %4592, 0
  %4594 = extractvalue { i1, i64 } %4592, 1
  br i1 %4593, label %NodeBlock7583, label %Flow7894

NodeBlock7583:                                    ; preds = %NodeBlock7585
  %Pivot7584 = icmp sge i16 %4568, 4
  %4595 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %Pivot7584)
  %4596 = extractvalue { i1, i64 } %4595, 0
  %4597 = extractvalue { i1, i64 } %4595, 1
  br i1 %4596, label %sw.bb4.i.i.i.i.i.i.i.i3542, label %Flow7889

Flow7894:                                         ; preds = %Flow7890, %NodeBlock7585
  %4598 = phi i8 [ %4628, %Flow7890 ], [ poison, %NodeBlock7585 ]
  %4599 = call { i1, i64 } @llvm.amdgcn.else.i64.i64(i64 %4594)
  %4600 = extractvalue { i1, i64 } %4599, 0
  %4601 = extractvalue { i1, i64 } %4599, 1
  br i1 %4600, label %NodeBlock7581, label %Flow7895

NodeBlock7581:                                    ; preds = %Flow7894
  %Pivot7582 = icmp sge i16 %4568, 2
  %4602 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %Pivot7582)
  %4603 = extractvalue { i1, i64 } %4602, 0
  %4604 = extractvalue { i1, i64 } %4602, 1
  br i1 %4603, label %sw.bb2.i.i.i.i.i.i.i.i3544, label %Flow7892

Flow7892:                                         ; preds = %sw.bb2.i.i.i.i.i.i.i.i3544, %NodeBlock7581
  %4605 = call { i1, i64 } @llvm.amdgcn.else.i64.i64(i64 %4604)
  %4606 = extractvalue { i1, i64 } %4605, 0
  %4607 = extractvalue { i1, i64 } %4605, 1
  br i1 %4606, label %LeafBlock7579, label %Flow7893

LeafBlock7579:                                    ; preds = %Flow7892
  %SwitchLeaf7580 = icmp eq i16 %4568, 1
  %4608 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %SwitchLeaf7580)
  %4609 = extractvalue { i1, i64 } %4608, 0
  %4610 = extractvalue { i1, i64 } %4608, 1
  br i1 %4609, label %sw.bb1.i.i.i.i.i.i.i.i3545, label %Flow7891

sw.bb1.i.i.i.i.i.i.i.i3545:                       ; preds = %LeafBlock7579
  br label %Flow7891, !amdgpu.uniform !42

sw.bb2.i.i.i.i.i.i.i.i3544:                       ; preds = %NodeBlock7581
  br label %Flow7892, !amdgpu.uniform !42

Flow7889:                                         ; preds = %sw.bb4.i.i.i.i.i.i.i.i3542, %NodeBlock7583
  %4611 = call { i1, i64 } @llvm.amdgcn.else.i64.i64(i64 %4597)
  %4612 = extractvalue { i1, i64 } %4611, 0
  %4613 = extractvalue { i1, i64 } %4611, 1
  br i1 %4612, label %sw.bb3.i.i.i.i.i.i.i.i3543, label %Flow7890

sw.bb3.i.i.i.i.i.i.i.i3543:                       ; preds = %Flow7889
  br label %Flow7890, !amdgpu.uniform !42

sw.bb4.i.i.i.i.i.i.i.i3542:                       ; preds = %NodeBlock7583
  br label %Flow7889, !amdgpu.uniform !42

Flow7885:                                         ; preds = %sw.bb6.i.i.i.i.i.i.i.i3540, %NodeBlock7587
  %4614 = call { i1, i64 } @llvm.amdgcn.else.i64.i64(i64 %4587)
  %4615 = extractvalue { i1, i64 } %4614, 0
  %4616 = extractvalue { i1, i64 } %4614, 1
  br i1 %4615, label %sw.bb5.i.i.i.i.i.i.i.i3541, label %Flow7886

sw.bb5.i.i.i.i.i.i.i.i3541:                       ; preds = %Flow7885
  br label %Flow7886, !amdgpu.uniform !42

sw.bb6.i.i.i.i.i.i.i.i3540:                       ; preds = %NodeBlock7587
  br label %Flow7885, !amdgpu.uniform !42

Flow7883:                                         ; preds = %Flow7882, %NodeBlock7591
  %4617 = phi i8 [ %4624, %Flow7882 ], [ poison, %NodeBlock7591 ]
  %4618 = call { i1, i64 } @llvm.amdgcn.else.i64.i64(i64 %4577)
  %4619 = extractvalue { i1, i64 } %4618, 0
  %4620 = extractvalue { i1, i64 } %4618, 1
  br i1 %4619, label %sw.bb7.i.i.i.i.i.i.i.i3539, label %Flow7884

sw.bb7.i.i.i.i.i.i.i.i3539:                       ; preds = %Flow7883
  br label %Flow7884, !amdgpu.uniform !42

Flow7881:                                         ; preds = %sw.bb9.i.i.i.i.i.i.i.i3380, %NodeBlock7589
  %4621 = call { i1, i64 } @llvm.amdgcn.else.i64.i64(i64 %4580)
  %4622 = extractvalue { i1, i64 } %4621, 0
  %4623 = extractvalue { i1, i64 } %4621, 1
  br i1 %4622, label %sw.bb8.i.i.i.i.i.i.i.i3538, label %Flow7882

sw.bb8.i.i.i.i.i.i.i.i3538:                       ; preds = %Flow7881
  br label %Flow7882, !amdgpu.uniform !42

sw.bb9.i.i.i.i.i.i.i.i3380:                       ; preds = %NodeBlock7589
  br label %Flow7881, !amdgpu.uniform !42

Flow7882:                                         ; preds = %sw.bb8.i.i.i.i.i.i.i.i3538, %Flow7881
  %4624 = phi i8 [ 56, %sw.bb8.i.i.i.i.i.i.i.i3538 ], [ 57, %Flow7881 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %4623)
  br label %Flow7883, !amdgpu.uniform !42

Flow7884:                                         ; preds = %sw.bb7.i.i.i.i.i.i.i.i3539, %Flow7883
  %4625 = phi i8 [ 55, %sw.bb7.i.i.i.i.i.i.i.i3539 ], [ %4617, %Flow7883 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %4620)
  br label %Flow7887, !amdgpu.uniform !42

Flow7886:                                         ; preds = %sw.bb5.i.i.i.i.i.i.i.i3541, %Flow7885
  %4626 = phi i8 [ 53, %sw.bb5.i.i.i.i.i.i.i.i3541 ], [ 54, %Flow7885 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %4616)
  br label %Flow7888, !amdgpu.uniform !42

Flow7888:                                         ; preds = %Flow7886, %Flow7887
  %4627 = phi i8 [ %4626, %Flow7886 ], [ %4581, %Flow7887 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %4584)
  br label %Flow7896, !amdgpu.uniform !42

Flow7890:                                         ; preds = %sw.bb3.i.i.i.i.i.i.i.i3543, %Flow7889
  %4628 = phi i8 [ 51, %sw.bb3.i.i.i.i.i.i.i.i3543 ], [ 52, %Flow7889 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %4613)
  br label %Flow7894, !amdgpu.uniform !42

Flow7891:                                         ; preds = %sw.bb1.i.i.i.i.i.i.i.i3545, %LeafBlock7579
  %4629 = phi i8 [ 49, %sw.bb1.i.i.i.i.i.i.i.i3545 ], [ 48, %LeafBlock7579 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %4610)
  br label %Flow7893, !amdgpu.uniform !42

Flow7893:                                         ; preds = %Flow7891, %Flow7892
  %4630 = phi i8 [ %4629, %Flow7891 ], [ 50, %Flow7892 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %4607)
  br label %Flow7895, !amdgpu.uniform !42

Flow7895:                                         ; preds = %Flow7893, %Flow7894
  %4631 = phi i8 [ %4630, %Flow7893 ], [ %4598, %Flow7894 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %4601)
  br label %if.then.i.i.i.i.i.i.i3381, !amdgpu.uniform !42

if.then.i.i.i.i.i.i.i3381:                        ; preds = %Flow7896, %Flow7895
  %retval.0.i.i.i.i.i.i.i.i3382 = phi i8 [ %4588, %Flow7896 ], [ %4631, %Flow7895 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %4591)
  %cmp.i.i.i.i.i.i.i.i3383 = icmp eq i64 %writer.sroa.10.0.i.i.i.i3374, 11
  %cmp.i.i.i.i.i.i.i.i3383.inv = xor i1 %cmp.i.i.i.i.i.i.i.i3383, true
  br i1 %cmp.i.i.i.i.i.i.i.i3383.inv, label %if.then3.i.i.i.i.i.i.i3384, label %_ZN22__llvm_libc_22_0_0_git7details22StringBufferWriterImplILb0EE4pushEc.exit.i.i.i.i.i.i3387, !amdgpu.uniform !42

if.then3.i.i.i.i.i.i.i3384:                       ; preds = %if.then.i.i.i.i.i.i.i3381
  %4632 = trunc i64 %writer.sroa.10.0.i.i.i.i3374 to i32
  %4633 = xor i32 %4632, -1
  %4634 = getelementptr i8, ptr addrspace(5) %buffer.i.i.i2988, i32 %4633
  %arrayidx.i.i.i.i.i.i.i.i33856093 = getelementptr i8, ptr addrspace(5) %4634, i32 11
  %inc.i.i.i.i.i.i.i3386 = add i64 %writer.sroa.10.0.i.i.i.i3374, 1
  store i8 %retval.0.i.i.i.i.i.i.i.i3382, ptr addrspace(5) %arrayidx.i.i.i.i.i.i.i.i33856093, align 1, !tbaa !57, !noalias !180
  br label %_ZN22__llvm_libc_22_0_0_git7details22StringBufferWriterImplILb0EE4pushEc.exit.i.i.i.i.i.i3387, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git7details22StringBufferWriterImplILb0EE4pushEc.exit.i.i.i.i.i.i3387: ; preds = %if.then3.i.i.i.i.i.i.i3384, %if.then.i.i.i.i.i.i.i3381
  %writer.sroa.10.1.i.i.i.i3388 = phi i64 [ %inc.i.i.i.i.i.i.i3386, %if.then3.i.i.i.i.i.i.i3384 ], [ 11, %if.then.i.i.i.i.i.i.i3381 ]
  %cmp.i.i.i.i.i.i3389 = icmp ult i32 %value.addr.09.i.i.i.i.i.i3375, 10
  %.not.i.i.i.i.i.i3390 = or i1 %cmp.i.i.i.i.i.i.i.i3383, %cmp.i.i.i.i.i.i3389
  %4635 = call i64 @llvm.amdgcn.if.break.i64(i1 %.not.i.i.i.i.i.i3390, i64 %phi.broken8780)
  %4636 = call i1 @llvm.amdgcn.loop.i64(i64 %4635)
  br i1 %4636, label %_ZN22__llvm_libc_22_0_0_git15IntegerToStringIiNS_7details3FmtILh10ELb0ELb0ELb0ELm1EEEE13IntegerWriter25write_unsigned_number_decEjRNS1_22StringBufferWriterImplILb0EEE.exit.i.i.i.i.i3391, label %while.body.i.i.i.i.i.i3373

_ZN22__llvm_libc_22_0_0_git15IntegerToStringIiNS_7details3FmtILh10ELb0ELb0ELb0ELm1EEEE13IntegerWriter25write_unsigned_number_decEjRNS1_22StringBufferWriterImplILb0EEE.exit.i.i.i.i.i3391: ; preds = %_ZN22__llvm_libc_22_0_0_git7details22StringBufferWriterImplILb0EE4pushEc.exit.i.i.i.i.i.i3387
  %writer.sroa.10.1.i.i.i.i3388.lcssa = phi i64 [ %writer.sroa.10.1.i.i.i.i3388, %_ZN22__llvm_libc_22_0_0_git7details22StringBufferWriterImplILb0EE4pushEc.exit.i.i.i.i.i.i3387 ]
  %.lcssa8922 = phi i64 [ %4635, %_ZN22__llvm_libc_22_0_0_git7details22StringBufferWriterImplILb0EE4pushEc.exit.i.i.i.i.i.i3387 ]
  %cmp.i.i.i.i.i.i.i.i3383.lcssa = phi i1 [ %cmp.i.i.i.i.i.i.i.i3383, %_ZN22__llvm_libc_22_0_0_git7details22StringBufferWriterImplILb0EE4pushEc.exit.i.i.i.i.i.i3387 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa8922)
  %cmp.i19.i.i.i.i3394 = icmp ne i64 %writer.sroa.10.1.i.i.i.i3388.lcssa, 0
  %or.cond.not.i.i.i.i3395 = select i1 %cmp.i.i.i.i.i.i.i.i3383.lcssa, i1 true, i1 %cmp.i19.i.i.i.i3394
  %or.cond.not.i.i.i.i3395.inv = xor i1 %or.cond.not.i.i.i.i3395, true
  %4637 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %or.cond.not.i.i.i.i3395.inv)
  %4638 = extractvalue { i1, i64 } %4637, 0
  %4639 = extractvalue { i1, i64 } %4637, 1
  br i1 %4638, label %if.then3.i.i.lr.ph.i.i.i.i3396, label %_ZN22__llvm_libc_22_0_0_git15IntegerToStringIiNS_7details3FmtILh10ELb0ELb0ELb0ELm1EEEEC2Ei.exit.i.i.i3407

if.then3.i.i.lr.ph.i.i.i.i3396:                   ; preds = %_ZN22__llvm_libc_22_0_0_git15IntegerToStringIiNS_7details3FmtILh10ELb0ELb0ELb0ELm1EEEE13IntegerWriter25write_unsigned_number_decEjRNS1_22StringBufferWriterImplILb0EEE.exit.i.i.i.i.i3391
  %arrayidx.i.i.i.i.i.i.i3397 = getelementptr inbounds nuw i8, ptr addrspace(5) %buffer.i.i.i2988, i32 10
  store i8 48, ptr addrspace(5) %arrayidx.i.i.i.i.i.i.i3397, align 2, !tbaa !57, !noalias !180
  br label %_ZN22__llvm_libc_22_0_0_git15IntegerToStringIiNS_7details3FmtILh10ELb0ELb0ELb0ELm1EEEEC2Ei.exit.i.i.i3407, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git15IntegerToStringIiNS_7details3FmtILh10ELb0ELb0ELb0ELm1EEEEC2Ei.exit.i.i.i3407: ; preds = %if.then3.i.i.lr.ph.i.i.i.i3396, %_ZN22__llvm_libc_22_0_0_git15IntegerToStringIiNS_7details3FmtILh10ELb0ELb0ELb0ELm1EEEE13IntegerWriter25write_unsigned_number_decEjRNS1_22StringBufferWriterImplILb0EEE.exit.i.i.i.i.i3391
  %writer.sroa.10.4.i.i.i.i3399 = phi i64 [ 1, %if.then3.i.i.lr.ph.i.i.i.i3396 ], [ %writer.sroa.10.1.i.i.i.i3388.lcssa, %_ZN22__llvm_libc_22_0_0_git15IntegerToStringIiNS_7details3FmtILh10ELb0ELb0ELb0ELm1EEEE13IntegerWriter25write_unsigned_number_decEjRNS1_22StringBufferWriterImplILb0EEE.exit.i.i.i.i.i3391 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %4639)
  %written.i.i.i.i3409 = getelementptr inbounds nuw i8, ptr addrspace(5) %buffer.i.i.i2988, i32 16
  store i64 %writer.sroa.10.4.i.i.i.i3399, ptr addrspace(5) %written.i.i.i.i3409, align 8, !tbaa !91, !noalias !180
  %cmp.i.not.i.i.i.i.i3410 = icmp ne i64 %writer.sroa.10.4.i.i.i.i3399, 0
  %4640 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.i.not.i.i.i.i.i3410)
  %4641 = extractvalue { i1, i64 } %4640, 0
  %4642 = extractvalue { i1, i64 } %4640, 1
  br i1 %4641, label %if.then.i.i.i.i.i.i3411, label %Flow7880

if.then.i.i.i.i.i.i3411:                          ; preds = %_ZN22__llvm_libc_22_0_0_git15IntegerToStringIiNS_7details3FmtILh10ELb0ELb0ELb0ELm1EEEEC2Ei.exit.i.i.i3407
  %4643 = call { i64, i1 } @llvm.uadd.with.overflow.i64(i64 %writer.sroa.10.4.i.i.i.i3399, i64 1)
  %ov7473 = extractvalue { i64, i1 } %4643, 1
  %ov7473.inv = xor i1 %ov7473, true
  %4644 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %ov7473.inv)
  %4645 = extractvalue { i1, i64 } %4644, 0
  %4646 = extractvalue { i1, i64 } %4644, 1
  br i1 %4645, label %if.end.i.i.i.i.i.i.i3414, label %Flow7879

if.end.i.i.i.i.i.i.i3414:                         ; preds = %if.then.i.i.i.i.i.i3411
  %math7472 = extractvalue { i64, i1 } %4643, 0
  %cmp2.i.i.i.i.i.i.i3415 = icmp ult i64 %math7472, 1676976733973595601
  %4647 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp2.i.i.i.i.i.i.i3415)
  %4648 = extractvalue { i1, i64 } %4647, 0
  %4649 = extractvalue { i1, i64 } %4647, 1
  br i1 %4648, label %if.then3.i.i.i.i4.i.i.i3535, label %if.end4.i.i.i.i.i.i.i3416

if.then3.i.i.i.i4.i.i.i3535:                      ; preds = %if.end.i.i.i.i.i.i.i3414
  %mul.i.i.i.i.i.i.i3536 = mul nuw i64 %math7472, 11
  %div18.i.i.i.i.i.i.i3537 = lshr i64 %mul.i.i.i.i.i.i.i3536, 3
  br label %if.end4.i.i.i.i.i.i.i3416, !amdgpu.uniform !42

Flow7879:                                         ; preds = %if.end4.i.i.i.i.i.i.i3416, %if.then.i.i.i.i.i.i3411
  %4650 = phi ptr [ %cond.i.i.i3424, %if.end4.i.i.i.i.i.i.i3416 ], [ addrspacecast (ptr addrspace(4) @_ZN22__llvm_libc_22_0_0_git3cpp6string14NULL_CHARACTERE to ptr), %if.then.i.i.i.i.i.i3411 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %4646)
  br label %_ZN22__llvm_libc_22_0_0_git3cpp6string7reserveEm.exit.i.i.i.i.i.i3426, !amdgpu.uniform !42

if.end4.i.i.i.i.i.i.i3416:                        ; preds = %if.then3.i.i.i.i4.i.i.i3535, %if.end.i.i.i.i.i.i.i3414
  %new_capacity.addr.0.i.i.i.i.i.i.i3417 = phi i64 [ %div18.i.i.i.i.i.i.i3537, %if.then3.i.i.i.i4.i.i.i3535 ], [ %math7472, %if.end.i.i.i.i.i.i.i3414 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %4649)
  %sub.i.i.i3418 = add i64 %new_capacity.addr.0.i.i.i.i.i.i.i3417, 7
  %div2.i.i.i3419 = and i64 %sub.i.i.i3418, -8
  %4651 = load ptr, ptr addrspace(1) @_ZN12_GLOBAL__N_13ptrE, align 8, !tbaa !55
  %add.ptr.i.i.i3420 = getelementptr inbounds nuw i8, ptr %4651, i64 %div2.i.i.i3419
  store ptr %add.ptr.i.i.i3420, ptr addrspace(1) @_ZN12_GLOBAL__N_13ptrE, align 8, !tbaa !55
  %sub.ptr.lhs.cast.i.i.i3421 = ptrtoint ptr %add.ptr.i.i.i3420 to i64
  %sub.ptr.sub.i.i.i3422 = sub i64 %sub.ptr.lhs.cast.i.i.i3421, ptrtoint (ptr addrspacecast (ptr addrspace(1) @_ZN12_GLOBAL__N_16memoryE to ptr) to i64)
  %cmp.i.i.i3423 = icmp ugt i64 %sub.ptr.sub.i.i.i3422, 65335
  %cond.i.i.i3424 = select i1 %cmp.i.i.i3423, ptr null, ptr %4651
  br label %Flow7879, !amdgpu.uniform !42

Flow7880:                                         ; preds = %Flow7878, %_ZN22__llvm_libc_22_0_0_git15IntegerToStringIiNS_7details3FmtILh10ELb0ELb0ELb0ELm1EEEEC2Ei.exit.i.i.i3407
  %4652 = phi ptr [ %4650, %Flow7878 ], [ addrspacecast (ptr addrspace(4) @_ZN22__llvm_libc_22_0_0_git3cpp6string14NULL_CHARACTERE to ptr), %_ZN22__llvm_libc_22_0_0_git15IntegerToStringIiNS_7details3FmtILh10ELb0ELb0ELb0ELm1EEEEC2Ei.exit.i.i.i3407 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %4642)
  br label %if.end.i.i.i.i.i.i3428, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git3cpp6string7reserveEm.exit.i.i.i.i.i.i3426: ; preds = %Flow7879
  br label %loadstoreloop4951, !amdgpu.uniform !42

loadstoreloop4951:                                ; preds = %_ZN22__llvm_libc_22_0_0_git3cpp6string7reserveEm.exit.i.i.i.i.i.i3426, %loadstoreloop4951
  %phi.broken8781 = phi i64 [ 0, %_ZN22__llvm_libc_22_0_0_git3cpp6string7reserveEm.exit.i.i.i.i.i.i3426 ], [ %4657, %loadstoreloop4951 ]
  %4653 = phi i64 [ %4655, %loadstoreloop4951 ], [ 0, %_ZN22__llvm_libc_22_0_0_git3cpp6string7reserveEm.exit.i.i.i.i.i.i3426 ]
  %4654 = getelementptr i8, ptr %4650, i64 %4653
  store i8 0, ptr %4654, align 1
  %4655 = add i64 %4653, 1
  %4656 = icmp uge i64 %4655, %writer.sroa.10.4.i.i.i.i3399
  %4657 = call i64 @llvm.amdgcn.if.break.i64(i1 %4656, i64 %phi.broken8781)
  %4658 = call i1 @llvm.amdgcn.loop.i64(i64 %4657)
  br i1 %4658, label %Flow7878, label %loadstoreloop4951

Flow7878:                                         ; preds = %loadstoreloop4951
  %.lcssa8921 = phi i64 [ %4657, %loadstoreloop4951 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa8921)
  br label %Flow7880, !amdgpu.uniform !42

if.end.i.i.i.i.i.i3428:                           ; preds = %Flow7880
  %cmp.not.i6.i.i.i.i.i.i3430 = icmp ne ptr %4652, addrspacecast (ptr addrspace(4) @_ZN22__llvm_libc_22_0_0_git3cpp6string14NULL_CHARACTERE to ptr)
  %4659 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.not.i6.i.i.i.i.i.i3430)
  %4660 = extractvalue { i1, i64 } %4659, 0
  %4661 = extractvalue { i1, i64 } %4659, 1
  br i1 %4660, label %if.then.i.i.i.i2.i.i.i3431, label %_ZN22__llvm_libc_22_0_0_git3cpp9to_stringEi.exit.i3433

if.then.i.i.i.i2.i.i.i3431:                       ; preds = %if.end.i.i.i.i.i.i3428
  %arrayidx.i.i.i.i3.i.i.i3432 = getelementptr inbounds nuw i8, ptr %4652, i64 %writer.sroa.10.4.i.i.i.i3399
  store i8 0, ptr %arrayidx.i.i.i.i3.i.i.i3432, align 1, !tbaa !57, !noalias !180
  br label %_ZN22__llvm_libc_22_0_0_git3cpp9to_stringEi.exit.i3433, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git3cpp9to_stringEi.exit.i3433: ; preds = %if.then.i.i.i.i2.i.i.i3431, %if.end.i.i.i.i.i.i3428
  call void @llvm.amdgcn.end.cf.i64(i64 %4661)
  %4662 = trunc i64 %writer.sroa.10.4.i.i.i.i3399 to i32
  %4663 = and i64 %writer.sroa.10.4.i.i.i.i3399, 15
  %4664 = sub i64 %writer.sroa.10.4.i.i.i.i3399, %4663
  %4665 = icmp ne i64 %4664, 0
  %4666 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %4665)
  %4667 = extractvalue { i1, i64 } %4666, 0
  %4668 = extractvalue { i1, i64 } %4666, 1
  br i1 %4667, label %loop-memcpy-expansion5115.preheader, label %Flow7877

loop-memcpy-expansion5115.preheader:              ; preds = %_ZN22__llvm_libc_22_0_0_git3cpp9to_stringEi.exit.i3433
  %4669 = sub i32 11, %4662
  %scevgep6477 = getelementptr i8, ptr addrspace(5) %buffer.i.i.i2988, i32 %4669
  br label %loop-memcpy-expansion5115, !amdgpu.uniform !42

Flow7877:                                         ; preds = %Flow7876, %_ZN22__llvm_libc_22_0_0_git3cpp9to_stringEi.exit.i3433
  call void @llvm.amdgcn.end.cf.i64(i64 %4668)
  br label %loop-memcpy-residual-header5118, !amdgpu.uniform !42

loop-memcpy-expansion5115:                        ; preds = %loop-memcpy-expansion5115.preheader, %loop-memcpy-expansion5115
  %phi.broken8782 = phi i64 [ 0, %loop-memcpy-expansion5115.preheader ], [ %4674, %loop-memcpy-expansion5115 ]
  %lsr.iv6478 = phi ptr addrspace(5) [ %scevgep6479, %loop-memcpy-expansion5115 ], [ %scevgep6477, %loop-memcpy-expansion5115.preheader ]
  %loop-index5116 = phi i64 [ %4672, %loop-memcpy-expansion5115 ], [ 0, %loop-memcpy-expansion5115.preheader ]
  %4670 = load <4 x i32>, ptr addrspace(5) %lsr.iv6478, align 1
  %4671 = getelementptr i8, ptr %4652, i64 %loop-index5116
  store <4 x i32> %4670, ptr %4671, align 1
  %4672 = add i64 %loop-index5116, 16
  %scevgep6479 = getelementptr i8, ptr addrspace(5) %lsr.iv6478, i32 16
  %4673 = icmp uge i64 %4672, %4664
  %4674 = call i64 @llvm.amdgcn.if.break.i64(i1 %4673, i64 %phi.broken8782)
  %4675 = call i1 @llvm.amdgcn.loop.i64(i64 %4674)
  br i1 %4675, label %Flow7876, label %loop-memcpy-expansion5115

Flow7875:                                         ; preds = %Flow7874, %loop-memcpy-residual-header5118
  call void @llvm.amdgcn.end.cf.i64(i64 %6383)
  br label %post-loop-memcpy-expansion5114, !amdgpu.uniform !42

loop-memcpy-residual5117:                         ; preds = %loop-memcpy-residual5117.preheader, %loop-memcpy-residual5117
  %phi.broken8783 = phi i64 [ 0, %loop-memcpy-residual5117.preheader ], [ %4679, %loop-memcpy-residual5117 ]
  %lsr.iv6473 = phi ptr addrspace(5) [ %scevgep6474, %loop-memcpy-residual5117 ], [ %scevgep6472, %loop-memcpy-residual5117.preheader ]
  %residual-loop-index5119 = phi i64 [ %4677, %loop-memcpy-residual5117 ], [ 0, %loop-memcpy-residual5117.preheader ]
  %4676 = load i8, ptr addrspace(5) %lsr.iv6473, align 1
  %scevgep6476 = getelementptr i8, ptr %scevgep6475, i64 %residual-loop-index5119
  store i8 %4676, ptr %scevgep6476, align 1
  %4677 = add i64 %residual-loop-index5119, 1
  %scevgep6474 = getelementptr i8, ptr addrspace(5) %lsr.iv6473, i32 1
  %4678 = icmp uge i64 %4677, %4663
  %4679 = call i64 @llvm.amdgcn.if.break.i64(i1 %4678, i64 %phi.broken8783)
  %4680 = call i1 @llvm.amdgcn.loop.i64(i64 %4679)
  br i1 %4680, label %Flow7874, label %loop-memcpy-residual5117

Flow7874:                                         ; preds = %loop-memcpy-residual5117
  %.lcssa8919 = phi i64 [ %4679, %loop-memcpy-residual5117 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa8919)
  br label %Flow7875, !amdgpu.uniform !42

post-loop-memcpy-expansion5114:                   ; preds = %Flow7875
  call void @llvm.lifetime.end.p5(ptr addrspace(5) %buffer.i.i.i2988) #34, !noalias !180
  br label %for.cond.i.i.i3437, !amdgpu.uniform !42

for.cond.i.i.i3437:                               ; preds = %post-loop-memcpy-expansion5114, %Flow7873
  %4681 = phi i32 [ %4718, %Flow7873 ], [ poison, %post-loop-memcpy-expansion5114 ]
  %4682 = phi i64 [ %4719, %Flow7873 ], [ poison, %post-loop-memcpy-expansion5114 ]
  %4683 = phi i32 [ %4720, %Flow7873 ], [ poison, %post-loop-memcpy-expansion5114 ]
  %retval.sroa.6.0.i.i.i3438 = phi i32 [ %4723, %Flow7873 ], [ undef, %post-loop-memcpy-expansion5114 ]
  %retval.sroa.2.0.i.i.i3440 = phi i64 [ %4722, %Flow7873 ], [ undef, %post-loop-memcpy-expansion5114 ]
  %retval.sroa.8.0.i.i.i3441 = phi i32 [ %4721, %Flow7873 ], [ undef, %post-loop-memcpy-expansion5114 ]
  %index.0.i.i.i3442 = phi i32 [ %4725, %Flow7873 ], [ 0, %post-loop-memcpy-expansion5114 ]
  %4684 = load i32, ptr addrspace(1) @__llvm_rpc_client, align 8, !tbaa !68
  %cmp.not.i.i.i3443 = icmp ult i32 %index.0.i.i.i3442, %4684
  %spec.store.select.i.i.i3444 = select i1 %cmp.not.i.i.i3443, i32 %index.0.i.i.i3442, i32 0
  %4685 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  %4686 = and i32 %spec.store.select.i.i.i3444, 31
  %4687 = lshr i64 %4685, %sh_prom.i.i.i.i3270
  %4688 = trunc i64 %4687 to i32
  %conv4.i.i.i.i.i3448 = and i32 %4688, 1
  %shl.i.i.i.i.i3449 = shl nuw i32 %conv4.i.i.i.i.i3448, %4686
  br label %ComputeLoop5866, !amdgpu.uniform !42

4689:                                             ; preds = %ComputeEnd5867
  %sunkaddr7474 = getelementptr inbounds i8, ptr addrspace(1) %8372, i64 40
  %4690 = atomicrmw or ptr addrspace(1) %sunkaddr7474, i32 %.lcssa8913 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %4691, !amdgpu.uniform !42

4691:                                             ; preds = %4689, %ComputeEnd5867
  %4692 = phi i32 [ %4690, %4689 ], [ poison, %ComputeEnd5867 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %8381)
  %4693 = call i32 @llvm.amdgcn.readfirstlane.i32(i32 %4692)
  %4694 = or i32 %4693, %.lcssa8914
  %shl5.i.i.i.i.i3450 = shl nuw i32 1, %4686
  %and.i.i.i.i.i3451 = and i32 %shl5.i.i.i.i.i3450, %4694
  %tobool3.i.i.i.i3452 = icmp ne i32 %and.i.i.i.i.i3451, 0
  %4695 = call i64 @llvm.amdgcn.ballot.i64(i1 %tobool3.i.i.i.i3452)
  %and.i.i.i.i.i.i3453 = and i64 %4695, %4685
  %cmp.i.not.i.i.i3454 = icmp ne i64 %4685, %and.i.i.i.i.i.i3453
  br i1 %cmp.i.not.i.i.i3454, label %if.then.i.i.i.i3455, label %Flow7872, !amdgpu.uniform !42

if.then.i.i.i.i3455:                              ; preds = %4691
  fence syncscope("agent") acquire
  %4696 = load <2 x i64>, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !75
  %4697 = extractelement <2 x i64> %4696, i32 0
  %4698 = inttoptr i64 %4697 to ptr
  %4699 = extractelement <2 x i64> %4696, i32 1
  %4700 = inttoptr i64 %4699 to ptr
  %idxprom.i.i.i.i3456 = zext i32 %spec.store.select.i.i.i3444 to i64
  %arrayidx.i.i.i.i3457 = getelementptr inbounds nuw i32, ptr %4698, i64 %idxprom.i.i.i.i3456, !amdgpu.uniform !42
  %4701 = load atomic i32, ptr %arrayidx.i.i.i.i3457 monotonic, align 4
  %4702 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %4701)
  %arrayidx.i45.i.i.i3458 = getelementptr inbounds nuw i32, ptr %4700, i64 %idxprom.i.i.i.i3456, !amdgpu.uniform !42
  %4703 = load atomic i32, ptr %arrayidx.i45.i.i.i3458 monotonic, align 4
  %4704 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %4703)
  %cmp.i46.not.i.i.i3459 = icmp ne i32 %4702, %4704
  br i1 %cmp.i46.not.i.i.i3459, label %if.then12.i.i.i3460, label %Flow7870, !amdgpu.uniform !42

if.then12.i.i.i3460:                              ; preds = %if.then.i.i.i.i3455
  fence syncscope("agent") release
  %4705 = call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %4685, i1 true)
  %iszero.i.i.i.i.i.i.i3461 = icmp ne i64 %4685, 0
  %cmp2.i.i.i.i.i.i3462 = icmp eq i64 %4705, %sh_prom.i.i.i.i3270
  %cmp.i.i.i.i.i2.i3463 = select i1 %iszero.i.i.i.i.i.i.i3461, i1 %cmp2.i.i.i.i.i.i3462, i1 false
  %conv4.i.i51.i.i.i3464 = zext i1 %cmp.i.i.i.i.i2.i3463 to i32
  %shl.i.i52.i.i.i3465 = shl nuw i32 %conv4.i.i51.i.i.i3464, %4686
  %xor.i.i.i.i.i3466 = xor i32 %shl.i.i52.i.i.i3465, -1
  %4706 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop5871, !amdgpu.uniform !42

4707:                                             ; preds = %ComputeEnd5872
  %sunkaddr7475 = getelementptr inbounds i8, ptr addrspace(1) %8372, i64 40
  %4708 = atomicrmw and ptr addrspace(1) %sunkaddr7475, i32 %.lcssa8915 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %4709, !amdgpu.uniform !42

4709:                                             ; preds = %4707, %ComputeEnd5872
  call void @llvm.amdgcn.end.cf.i64(i64 %8399)
  call void @llvm.amdgcn.wave.barrier()
  br label %Flow7870, !amdgpu.uniform !42

if.end14.i.i.i3527:                               ; preds = %Flow7870
  %4710 = call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %4685, i1 true)
  %iszero.i.i.i.i.i.i3528 = icmp ne i64 %4685, 0
  %cmp2.i.i.i.i.i3529 = icmp eq i64 %4710, %sh_prom.i.i.i.i3270
  %cmp.i.i.i.i.i3530 = select i1 %iszero.i.i.i.i.i.i3528, i1 %cmp2.i.i.i.i.i3529, i1 false
  %4711 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.i.i.i.i.i3530)
  %4712 = extractvalue { i1, i64 } %4711, 0
  %4713 = extractvalue { i1, i64 } %4711, 1
  br i1 %4712, label %if.then16.i.i.i3532, label %if.end22.i.i.i3531

if.then16.i.i.i3532:                              ; preds = %if.end14.i.i.i3527
  %4714 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !76
  %opcode.i.i.i3533 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %4714, i64 %idxprom.i.i.i.i3456, i32 1
  store i32 1660944387, ptr %opcode.i.i.i3533, align 8, !tbaa !77
  %arrayidx21.i.i.i3534 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %4714, i64 %idxprom.i.i.i.i3456
  store i64 %4685, ptr %arrayidx21.i.i.i3534, align 8, !tbaa !80
  br label %if.end22.i.i.i3531, !amdgpu.uniform !42

Flow7871:                                         ; preds = %if.end22.i.i.i3531, %Flow7870
  %4715 = phi i32 [ %4702, %if.end22.i.i.i3531 ], [ %retval.sroa.8.0.i.i.i3441, %Flow7870 ]
  %4716 = phi i64 [ %4685, %if.end22.i.i.i3531 ], [ %retval.sroa.2.0.i.i.i3440, %Flow7870 ]
  %4717 = phi i32 [ %spec.store.select.i.i.i3444, %if.end22.i.i.i3531 ], [ %retval.sroa.6.0.i.i.i3438, %Flow7870 ]
  br label %cleanup26.i.i.i3467, !amdgpu.uniform !42

if.end22.i.i.i3531:                               ; preds = %if.then16.i.i.i3532, %if.end14.i.i.i3527
  call void @llvm.amdgcn.end.cf.i64(i64 %4713)
  call void @llvm.amdgcn.wave.barrier()
  br label %Flow7871, !amdgpu.uniform !42

Flow7872:                                         ; preds = %cleanup26.i.i.i3467, %4691
  %4718 = phi i32 [ %4717, %cleanup26.i.i.i3467 ], [ %4681, %4691 ]
  %4719 = phi i64 [ %4716, %cleanup26.i.i.i3467 ], [ %4682, %4691 ]
  %4720 = phi i32 [ %4715, %cleanup26.i.i.i3467 ], [ %4683, %4691 ]
  %4721 = phi i32 [ %4715, %cleanup26.i.i.i3467 ], [ %retval.sroa.8.0.i.i.i3441, %4691 ]
  %4722 = phi i64 [ %4716, %cleanup26.i.i.i3467 ], [ %retval.sroa.2.0.i.i.i3440, %4691 ]
  %4723 = phi i32 [ %4717, %cleanup26.i.i.i3467 ], [ %retval.sroa.6.0.i.i.i3438, %4691 ]
  %4724 = phi i1 [ %cmp.i46.not.i.i.i3459, %cleanup26.i.i.i3467 ], [ true, %4691 ]
  br i1 %4724, label %for.inc.i.i.i3472, label %Flow7873, !amdgpu.uniform !42

cleanup26.i.i.i3467:                              ; preds = %Flow7871
  br label %Flow7872, !amdgpu.uniform !42

for.inc.i.i.i3472:                                ; preds = %Flow7872
  %inc.i.i.i3477 = add i32 %spec.store.select.i.i.i3444, 1
  br label %Flow7873, !amdgpu.uniform !42

Flow7873:                                         ; preds = %for.inc.i.i.i3472, %Flow7872
  %4725 = phi i32 [ %inc.i.i.i3477, %for.inc.i.i.i3472 ], [ poison, %Flow7872 ]
  %4726 = phi i1 [ false, %for.inc.i.i.i3472 ], [ true, %Flow7872 ]
  br i1 %4726, label %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i3478, label %for.cond.i.i.i3437, !amdgpu.uniform !42

_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i3478: ; preds = %Flow7873
  %.lcssa8918 = phi i32 [ %4718, %Flow7873 ]
  %.lcssa8917 = phi i64 [ %4719, %Flow7873 ]
  %.lcssa8916 = phi i32 [ %4720, %Flow7873 ]
  fence acquire
  %4727 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %mul.i.i.i.i.i.i3479 = shl i32 %.lcssa8918, 6
  %idxprom.i19.i.i.i.i.i3480 = zext i32 %mul.i.i.i.i.i.i3479 to i64
  %arrayidx.i20.i.i.i.i.i3481 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %4727, i64 %idxprom.i19.i.i.i.i.i3480
  %arrayidx.i22.i.i.i.i.i3482 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %arrayidx.i20.i.i.i.i.i3481, i64 %sh_prom.i.i.i.i3270
  store i64 %writer.sroa.10.4.i.i.i.i3399, ptr %arrayidx.i22.i.i.i.i.i3482, align 8, !tbaa !82
  %spec.select.i.i.i.i.i.i.i3483 = call i64 @llvm.umin.i64(i64 %writer.sroa.10.4.i.i.i.i3399, i64 56)
  %4728 = and i64 %spec.select.i.i.i.i.i.i.i3483, 15
  %4729 = sub i64 %spec.select.i.i.i.i.i.i.i3483, %4728
  %4730 = icmp ne i64 %4729, 0
  %4731 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %4730)
  %4732 = extractvalue { i1, i64 } %4731, 0
  %4733 = extractvalue { i1, i64 } %4731, 1
  br i1 %4732, label %loop-memcpy-expansion5001.preheader, label %Flow7869

loop-memcpy-expansion5001.preheader:              ; preds = %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i3478
  %4734 = add nuw nsw i64 %4513, 8
  %scevgep6469 = getelementptr i8, ptr %4727, i64 %4734
  %4735 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i3480, 6
  %scevgep6470 = getelementptr i8, ptr %scevgep6469, i64 %4735
  br label %loop-memcpy-expansion5001, !amdgpu.uniform !42

Flow7869:                                         ; preds = %Flow7868, %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i3478
  call void @llvm.amdgcn.end.cf.i64(i64 %4733)
  br label %loop-memcpy-residual-header5004, !amdgpu.uniform !42

loop-memcpy-expansion5001:                        ; preds = %loop-memcpy-expansion5001.preheader, %loop-memcpy-expansion5001
  %phi.broken8784 = phi i64 [ 0, %loop-memcpy-expansion5001.preheader ], [ %4740, %loop-memcpy-expansion5001 ]
  %loop-index5002 = phi i64 [ %4738, %loop-memcpy-expansion5001 ], [ 0, %loop-memcpy-expansion5001.preheader ]
  %4736 = getelementptr i8, ptr %4652, i64 %loop-index5002
  %4737 = load <4 x i32>, ptr %4736, align 1
  %scevgep6471 = getelementptr i8, ptr %scevgep6470, i64 %loop-index5002
  store <4 x i32> %4737, ptr %scevgep6471, align 1
  %4738 = add i64 %loop-index5002, 16
  %4739 = icmp uge i64 %4738, %4729
  %4740 = call i64 @llvm.amdgcn.if.break.i64(i1 %4739, i64 %phi.broken8784)
  %4741 = call i1 @llvm.amdgcn.loop.i64(i64 %4740)
  br i1 %4741, label %Flow7868, label %loop-memcpy-expansion5001

Flow7867:                                         ; preds = %Flow7865, %Flow7866
  %4742 = phi i64 [ %6257, %Flow7865 ], [ %6249, %Flow7866 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %6252)
  br label %post-loop-memcpy-expansion5000, !amdgpu.uniform !42

loop-memcpy-residual5003:                         ; preds = %loop-memcpy-residual5003.preheader, %loop-memcpy-residual5003
  %phi.broken8785 = phi i64 [ 0, %loop-memcpy-residual5003.preheader ], [ %4746, %loop-memcpy-residual5003 ]
  %residual-loop-index5005 = phi i64 [ %4744, %loop-memcpy-residual5003 ], [ 0, %loop-memcpy-residual5003.preheader ]
  %scevgep6468 = getelementptr i8, ptr %scevgep6467, i64 %residual-loop-index5005
  %4743 = load i8, ptr %scevgep6468, align 1
  %scevgep6466 = getelementptr i8, ptr %scevgep6465, i64 %residual-loop-index5005
  store i8 %4743, ptr %scevgep6466, align 1
  %4744 = add i64 %residual-loop-index5005, 1
  %4745 = icmp uge i64 %4744, %4728
  %4746 = call i64 @llvm.amdgcn.if.break.i64(i1 %4745, i64 %phi.broken8785)
  %4747 = call i1 @llvm.amdgcn.loop.i64(i64 %4746)
  br i1 %4747, label %Flow7865, label %loop-memcpy-residual5003

Flow7865:                                         ; preds = %loop-memcpy-residual5003
  %.lcssa8911 = phi i64 [ %4746, %loop-memcpy-residual5003 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa8911)
  br label %Flow7867, !amdgpu.uniform !42

post-loop-memcpy-expansion5000:                   ; preds = %Flow7867
  %tobool.not.i.i.i.i.i.i3485 = icmp eq i32 %.lcssa8916, 0
  %conv.i.i.i.i.i.i3486 = zext i1 %tobool.not.i.i.i.i.i.i3485 to i32
  fence release
  %4748 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %idxprom.i23.i.i.i.i.i3487 = zext i32 %.lcssa8918 to i64
  %arrayidx.i24.i.i.i.i.i3488 = getelementptr inbounds nuw i32, ptr %4748, i64 %idxprom.i23.i.i.i.i.i3487
  store atomic i32 %conv.i.i.i.i.i.i3486, ptr %arrayidx.i24.i.i.i.i.i3488 monotonic, align 4
  %4749 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !84
  %arrayidx.i.i5.i.i3489 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %4749, i64 %idxprom.i23.i.i.i.i.i3487, !amdgpu.uniform !42
  %4750 = load i64, ptr %arrayidx.i.i5.i.i3489, align 8, !tbaa !80
  %scevgep6451 = getelementptr i8, ptr %4652, i64 56
  %4751 = add i64 %writer.sroa.10.4.i.i.i.i3399, -56
  br label %while.cond.i.i.i.i3490, !amdgpu.uniform !42

while.cond.i.i.i.i3490:                           ; preds = %post-loop-memcpy-expansion5000, %Flow7864
  %phi.broken8788 = phi i64 [ 0, %post-loop-memcpy-expansion5000 ], [ %4768, %Flow7864 ]
  %lsr.iv6457 = phi i64 [ %4766, %Flow7864 ], [ %4751, %post-loop-memcpy-expansion5000 ]
  %lsr.iv6452 = phi ptr [ %4765, %Flow7864 ], [ %scevgep6451, %post-loop-memcpy-expansion5000 ]
  %port.sroa.37.0.i.i3491 = phi i32 [ %4764, %Flow7864 ], [ %conv.i.i.i.i.i.i3486, %post-loop-memcpy-expansion5000 ]
  %idx.0.i.i.i.i3492 = phi i64 [ %4763, %Flow7864 ], [ 56, %post-loop-memcpy-expansion5000 ]
  %cmp.i.i.i.i3493 = icmp ugt i64 %writer.sroa.10.4.i.i.i.i3399, %idx.0.i.i.i.i3492
  %4752 = call i64 @llvm.amdgcn.ballot.i64(i1 %cmp.i.i.i.i3493)
  %and.i.i.i.i6.i.i3494 = and i64 %4752, %4750
  %tobool.not.i.i.i.i3495 = icmp ne i64 %and.i.i.i.i6.i.i3494, 0
  %4753 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i22.i.i.i3496 = getelementptr inbounds nuw i32, ptr %4753, i64 %idxprom.i23.i.i.i.i.i3487, !amdgpu.uniform !42
  %4754 = load atomic i32, ptr %arrayidx.i22.i.i.i3496 monotonic, align 4
  %4755 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %tobool.not.i.i.i.i3495)
  %4756 = extractvalue { i1, i64 } %4755, 0
  %4757 = extractvalue { i1, i64 } %4755, 1
  br i1 %4756, label %cond.false.i8.i.i.i.i3497, label %Flow7864

cond.false.i8.i.i.i.i3497:                        ; preds = %while.cond.i.i.i.i3490
  %4758 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %4754)
  %4759 = icmp ne i32 %4758, %port.sroa.37.0.i.i3491
  br i1 %4759, label %while.body.i.i21.i.i.i.i3498.preheader, label %Flow7863, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i3498.preheader:           ; preds = %cond.false.i8.i.i.i.i3497
  br label %while.body.i.i21.i.i.i.i3498, !amdgpu.uniform !42

Flow7863:                                         ; preds = %Flow7862, %cond.false.i8.i.i.i.i3497
  br label %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i3501, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i3498:                     ; preds = %while.body.i.i21.i.i.i.i3498.preheader, %while.body.i.i21.i.i.i.i3498
  call void @llvm.amdgcn.s.sleep(i32 2)
  %4760 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i22.i.i.i.i3499 = getelementptr inbounds nuw i32, ptr %4760, i64 %idxprom.i23.i.i.i.i.i3487, !amdgpu.uniform !42
  %4761 = load atomic i32, ptr %arrayidx.i.i.i22.i.i.i.i3499 monotonic, align 4
  %4762 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %4761)
  %cmp.i.not.i.i23.i.i.i.i3500 = icmp eq i32 %4762, %port.sroa.37.0.i.i3491
  br i1 %cmp.i.not.i.i23.i.i.i.i3500, label %Flow7862, label %while.body.i.i21.i.i.i.i3498, !amdgpu.uniform !42

Flow7862:                                         ; preds = %while.body.i.i21.i.i.i.i3498
  br label %Flow7863, !amdgpu.uniform !42

Flow7864:                                         ; preds = %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i3502, %while.cond.i.i.i.i3490
  %4763 = phi i64 [ %add.i.i.i.i3505, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i3502 ], [ poison, %while.cond.i.i.i.i3490 ]
  %4764 = phi i32 [ %conv.i.i26.i.i.i.i3503, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i3502 ], [ poison, %while.cond.i.i.i.i3490 ]
  %4765 = phi ptr [ %scevgep6453, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i3502 ], [ poison, %while.cond.i.i.i.i3490 ]
  %4766 = phi i64 [ %lsr.iv.next6458, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i3502 ], [ poison, %while.cond.i.i.i.i3490 ]
  %4767 = phi i1 [ false, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i3502 ], [ true, %while.cond.i.i.i.i3490 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %4757)
  %4768 = call i64 @llvm.amdgcn.if.break.i64(i1 %4767, i64 %phi.broken8788)
  %4769 = call i1 @llvm.amdgcn.loop.i64(i64 %4768)
  br i1 %4769, label %cond.false.i.i.i3512, label %while.cond.i.i.i.i3490

_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i3501: ; preds = %Flow7863
  fence acquire
  %4770 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.i.i.i.i3493)
  %4771 = extractvalue { i1, i64 } %4770, 0
  %4772 = extractvalue { i1, i64 } %4770, 1
  br i1 %4771, label %if.then.i.i.i.i.i.i3.i3506, label %Flow7861

if.then.i.i.i.i.i.i3.i3506:                       ; preds = %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i3501
  %sub.i.i.i.i.i.i.i3508 = sub nuw i64 %writer.sroa.10.4.i.i.i.i3399, %idx.0.i.i.i.i3492
  %spec.select.i.i.i36.i.i.i.i3509 = call i64 @llvm.umin.i64(i64 %sub.i.i.i.i.i.i.i3508, i64 64)
  %4773 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %4774 = and i64 %spec.select.i.i.i36.i.i.i.i3509, 15
  %4775 = sub i64 %spec.select.i.i.i36.i.i.i.i3509, %4774
  %4776 = icmp ne i64 %4775, 0
  %4777 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %4776)
  %4778 = extractvalue { i1, i64 } %4777, 0
  %4779 = extractvalue { i1, i64 } %4777, 1
  br i1 %4778, label %loop-memcpy-expansion4995.preheader, label %Flow7860

loop-memcpy-expansion4995.preheader:              ; preds = %if.then.i.i.i.i.i.i3.i3506
  %arrayidx.i20.i35.i.i.i.i3510 = getelementptr %"struct.rpc::Buffer", ptr %4773, i64 %idxprom.i19.i.i.i.i.i3480
  %arrayidx.i22.i38.i.i.i.i3511 = getelementptr %"struct.rpc::Buffer", ptr %arrayidx.i20.i35.i.i.i.i3510, i64 %sh_prom.i.i.i.i3270
  br label %loop-memcpy-expansion4995, !amdgpu.uniform !42

Flow7860:                                         ; preds = %Flow7859, %if.then.i.i.i.i.i.i3.i3506
  call void @llvm.amdgcn.end.cf.i64(i64 %4779)
  br label %loop-memcpy-residual-header4998, !amdgpu.uniform !42

loop-memcpy-expansion4995:                        ; preds = %loop-memcpy-expansion4995.preheader, %loop-memcpy-expansion4995
  %phi.broken8786 = phi i64 [ 0, %loop-memcpy-expansion4995.preheader ], [ %4784, %loop-memcpy-expansion4995 ]
  %loop-index4996 = phi i64 [ %4782, %loop-memcpy-expansion4995 ], [ 0, %loop-memcpy-expansion4995.preheader ]
  %scevgep6454 = getelementptr i8, ptr %lsr.iv6452, i64 %loop-index4996
  %4780 = load <4 x i32>, ptr %scevgep6454, align 1
  %4781 = getelementptr i8, ptr %arrayidx.i22.i38.i.i.i.i3511, i64 %loop-index4996
  store <4 x i32> %4780, ptr %4781, align 1
  %4782 = add i64 %loop-index4996, 16
  %4783 = icmp uge i64 %4782, %4775
  %4784 = call i64 @llvm.amdgcn.if.break.i64(i1 %4783, i64 %phi.broken8786)
  %4785 = call i1 @llvm.amdgcn.loop.i64(i64 %4784)
  br i1 %4785, label %Flow7859, label %loop-memcpy-expansion4995

Flow7858:                                         ; preds = %Flow7857, %loop-memcpy-residual-header4998
  call void @llvm.amdgcn.end.cf.i64(i64 %6242)
  br label %Flow7861, !amdgpu.uniform !42

loop-memcpy-residual4997:                         ; preds = %loop-memcpy-residual4997.preheader, %loop-memcpy-residual4997
  %phi.broken8787 = phi i64 [ 0, %loop-memcpy-residual4997.preheader ], [ %4789, %loop-memcpy-residual4997 ]
  %residual-loop-index4999 = phi i64 [ %4787, %loop-memcpy-residual4997 ], [ 0, %loop-memcpy-residual4997.preheader ]
  %scevgep6463 = getelementptr i8, ptr %scevgep6462, i64 %residual-loop-index4999
  %4786 = load i8, ptr %scevgep6463, align 1
  %scevgep6461 = getelementptr i8, ptr %scevgep6460, i64 %residual-loop-index4999
  store i8 %4786, ptr %scevgep6461, align 1
  %4787 = add i64 %residual-loop-index4999, 1
  %4788 = icmp uge i64 %4787, %4774
  %4789 = call i64 @llvm.amdgcn.if.break.i64(i1 %4788, i64 %phi.broken8787)
  %4790 = call i1 @llvm.amdgcn.loop.i64(i64 %4789)
  br i1 %4790, label %Flow7857, label %loop-memcpy-residual4997

Flow7857:                                         ; preds = %loop-memcpy-residual4997
  %.lcssa8908 = phi i64 [ %4789, %loop-memcpy-residual4997 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa8908)
  br label %Flow7858, !amdgpu.uniform !42

_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i3502: ; preds = %Flow7861
  %conv.i.i26.i.i.i.i3503 = xor i32 %port.sroa.37.0.i.i3491, 1
  fence release
  %4791 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %arrayidx.i24.i29.i.i.i.i3504 = getelementptr inbounds nuw i32, ptr %4791, i64 %idxprom.i23.i.i.i.i.i3487
  store atomic i32 %conv.i.i26.i.i.i.i3503, ptr %arrayidx.i24.i29.i.i.i.i3504 monotonic, align 4
  %add.i.i.i.i3505 = add i64 %idx.0.i.i.i.i3492, 64
  %scevgep6453 = getelementptr i8, ptr %lsr.iv6452, i64 64
  %lsr.iv.next6458 = add i64 %lsr.iv6457, -64
  br label %Flow7864, !amdgpu.uniform !42

cond.false.i.i.i3512:                             ; preds = %Flow7864
  %.lcssa8910 = phi i64 [ %4768, %Flow7864 ]
  %port.sroa.37.0.i.i3491.lcssa = phi i32 [ %port.sroa.37.0.i.i3491, %Flow7864 ]
  %.lcssa8909 = phi i32 [ %4754, %Flow7864 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa8910)
  %4792 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %.lcssa8909)
  %4793 = icmp ne i32 %4792, %port.sroa.37.0.i.i3491.lcssa
  %4794 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %4793)
  %4795 = extractvalue { i1, i64 } %4794, 0
  %4796 = extractvalue { i1, i64 } %4794, 1
  br i1 %4795, label %while.body.i.i.i.i3513.preheader, label %Flow7856

while.body.i.i.i.i3513.preheader:                 ; preds = %cond.false.i.i.i3512
  br label %while.body.i.i.i.i3513, !amdgpu.uniform !42

Flow7856:                                         ; preds = %Flow7855, %cond.false.i.i.i3512
  call void @llvm.amdgcn.end.cf.i64(i64 %4796)
  br label %_ZN22__llvm_libc_22_0_0_git15write_to_stderrENS_3cpp11string_viewE.exit.i3516, !amdgpu.uniform !42

while.body.i.i.i.i3513:                           ; preds = %while.body.i.i.i.i3513.preheader, %while.body.i.i.i.i3513
  %phi.broken8789 = phi i64 [ 0, %while.body.i.i.i.i3513.preheader ], [ %4800, %while.body.i.i.i.i3513 ]
  call void @llvm.amdgcn.s.sleep(i32 2)
  %4797 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i8.i.i3514 = getelementptr inbounds nuw i32, ptr %4797, i64 %idxprom.i23.i.i.i.i.i3487, !amdgpu.uniform !42
  %4798 = load atomic i32, ptr %arrayidx.i.i.i8.i.i3514 monotonic, align 4
  %4799 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %4798)
  %cmp.i.not.i.i.i.i3515 = icmp eq i32 %4799, %port.sroa.37.0.i.i3491.lcssa
  %4800 = call i64 @llvm.amdgcn.if.break.i64(i1 %cmp.i.not.i.i.i.i3515, i64 %phi.broken8789)
  %4801 = call i1 @llvm.amdgcn.loop.i64(i64 %4800)
  br i1 %4801, label %Flow7855, label %while.body.i.i.i.i3513

Flow7855:                                         ; preds = %while.body.i.i.i.i3513
  %.lcssa8906 = phi i64 [ %4800, %while.body.i.i.i.i3513 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa8906)
  br label %Flow7856, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git15write_to_stderrENS_3cpp11string_viewE.exit.i3516: ; preds = %Flow7856
  fence acquire
  call void @llvm.amdgcn.wave.barrier()
  fence syncscope("agent") release
  %4802 = call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %.lcssa8917, i1 true)
  %iszero.i.i.i.i.i13.i.i3517 = icmp ne i64 %.lcssa8917, 0
  %cmp2.i.i.i.i14.i.i3518 = icmp eq i64 %4802, %sh_prom.i.i.i.i3270
  %cmp.i.i.i.i15.i.i3519 = select i1 %iszero.i.i.i.i.i13.i.i3517, i1 %cmp2.i.i.i.i14.i.i3518, i1 false
  %4803 = and i32 %.lcssa8918, 31
  %conv4.i.i.i19.i.i3523 = zext i1 %cmp.i.i.i.i15.i.i3519 to i32
  %shl.i.i.i20.i.i3524 = shl nuw i32 %conv4.i.i.i19.i.i3523, %4803
  %xor.i.i.i21.i.i3525 = xor i32 %shl.i.i.i20.i.i3524, -1
  %4804 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop5875, !amdgpu.uniform !42

4805:                                             ; preds = %ComputeEnd5876
  %div8.i.i.i16.i.i3520 = lshr i32 %.lcssa8918, 5
  %idxprom.i.i.i17.i.i3521 = zext nneg i32 %div8.i.i.i16.i.i3520 to i64
  %4806 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i17.i.i3521
  %sunkaddr7476 = getelementptr inbounds i8, ptr addrspace(1) %4806, i64 40
  %4807 = atomicrmw and ptr addrspace(1) %sunkaddr7476, i32 %.lcssa8905 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %4808, !amdgpu.uniform !42

4808:                                             ; preds = %4805, %ComputeEnd5876
  call void @llvm.amdgcn.end.cf.i64(i64 %8416)
  call void @llvm.amdgcn.wave.barrier()
  br label %for.cond.i.i1.i3556, !amdgpu.uniform !42

for.cond.i.i1.i3556:                              ; preds = %4808, %Flow7854
  %4809 = phi i32 [ %4846, %Flow7854 ], [ poison, %4808 ]
  %4810 = phi i64 [ %4847, %Flow7854 ], [ poison, %4808 ]
  %4811 = phi i32 [ %4848, %Flow7854 ], [ poison, %4808 ]
  %retval.sroa.6.0.i.i.i3557 = phi i32 [ %4851, %Flow7854 ], [ undef, %4808 ]
  %retval.sroa.2.0.i.i.i3559 = phi i64 [ %4850, %Flow7854 ], [ undef, %4808 ]
  %retval.sroa.8.0.i.i.i3560 = phi i32 [ %4849, %Flow7854 ], [ undef, %4808 ]
  %index.0.i.i.i3561 = phi i32 [ %4853, %Flow7854 ], [ 0, %4808 ]
  %4812 = load i32, ptr addrspace(1) @__llvm_rpc_client, align 8, !tbaa !68
  %cmp.not.i.i.i3562 = icmp ult i32 %index.0.i.i.i3561, %4812
  %spec.store.select.i.i.i3563 = select i1 %cmp.not.i.i.i3562, i32 %index.0.i.i.i3561, i32 0
  %4813 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  %4814 = and i32 %spec.store.select.i.i.i3563, 31
  %4815 = lshr i64 %4813, %sh_prom.i.i.i.i3270
  %4816 = trunc i64 %4815 to i32
  %conv4.i.i.i.i.i3567 = and i32 %4816, 1
  %shl.i.i.i.i.i3568 = shl nuw i32 %conv4.i.i.i.i.i3567, %4814
  br label %ComputeLoop5879, !amdgpu.uniform !42

4817:                                             ; preds = %ComputeEnd5880
  %sunkaddr7477 = getelementptr inbounds i8, ptr addrspace(1) %8426, i64 40
  %4818 = atomicrmw or ptr addrspace(1) %sunkaddr7477, i32 %.lcssa8899 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %4819, !amdgpu.uniform !42

4819:                                             ; preds = %4817, %ComputeEnd5880
  %4820 = phi i32 [ %4818, %4817 ], [ poison, %ComputeEnd5880 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %8435)
  %4821 = call i32 @llvm.amdgcn.readfirstlane.i32(i32 %4820)
  %4822 = or i32 %4821, %.lcssa8900
  %shl5.i.i.i.i.i3569 = shl nuw i32 1, %4814
  %and.i.i.i.i.i3570 = and i32 %shl5.i.i.i.i.i3569, %4822
  %tobool3.i.i.i.i3571 = icmp ne i32 %and.i.i.i.i.i3570, 0
  %4823 = call i64 @llvm.amdgcn.ballot.i64(i1 %tobool3.i.i.i.i3571)
  %and.i.i.i.i.i.i3572 = and i64 %4823, %4813
  %cmp.i.not.i.i.i3573 = icmp ne i64 %4813, %and.i.i.i.i.i.i3572
  br i1 %cmp.i.not.i.i.i3573, label %if.then.i.i.i.i3574, label %Flow7853, !amdgpu.uniform !42

if.then.i.i.i.i3574:                              ; preds = %4819
  fence syncscope("agent") acquire
  %4824 = load <2 x i64>, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !75
  %4825 = extractelement <2 x i64> %4824, i32 0
  %4826 = inttoptr i64 %4825 to ptr
  %4827 = extractelement <2 x i64> %4824, i32 1
  %4828 = inttoptr i64 %4827 to ptr
  %idxprom.i.i.i.i3575 = zext i32 %spec.store.select.i.i.i3563 to i64
  %arrayidx.i.i.i.i3576 = getelementptr inbounds nuw i32, ptr %4826, i64 %idxprom.i.i.i.i3575, !amdgpu.uniform !42
  %4829 = load atomic i32, ptr %arrayidx.i.i.i.i3576 monotonic, align 4
  %4830 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %4829)
  %arrayidx.i45.i.i.i3577 = getelementptr inbounds nuw i32, ptr %4828, i64 %idxprom.i.i.i.i3575, !amdgpu.uniform !42
  %4831 = load atomic i32, ptr %arrayidx.i45.i.i.i3577 monotonic, align 4
  %4832 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %4831)
  %cmp.i46.not.i.i.i3578 = icmp ne i32 %4830, %4832
  br i1 %cmp.i46.not.i.i.i3578, label %if.then12.i.i.i3579, label %Flow7851, !amdgpu.uniform !42

if.then12.i.i.i3579:                              ; preds = %if.then.i.i.i.i3574
  fence syncscope("agent") release
  %4833 = call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %4813, i1 true)
  %iszero.i.i.i.i.i.i.i3580 = icmp ne i64 %4813, 0
  %cmp2.i.i.i.i.i.i3581 = icmp eq i64 %4833, %sh_prom.i.i.i.i3270
  %cmp.i.i.i.i.i.i3582 = select i1 %iszero.i.i.i.i.i.i.i3580, i1 %cmp2.i.i.i.i.i.i3581, i1 false
  %conv4.i.i51.i.i.i3583 = zext i1 %cmp.i.i.i.i.i.i3582 to i32
  %shl.i.i52.i.i.i3584 = shl nuw i32 %conv4.i.i51.i.i.i3583, %4814
  %xor.i.i.i.i.i3585 = xor i32 %shl.i.i52.i.i.i3584, -1
  %4834 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop5884, !amdgpu.uniform !42

4835:                                             ; preds = %ComputeEnd5885
  %sunkaddr7478 = getelementptr inbounds i8, ptr addrspace(1) %8426, i64 40
  %4836 = atomicrmw and ptr addrspace(1) %sunkaddr7478, i32 %.lcssa8901 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %4837, !amdgpu.uniform !42

4837:                                             ; preds = %4835, %ComputeEnd5885
  call void @llvm.amdgcn.end.cf.i64(i64 %8453)
  call void @llvm.amdgcn.wave.barrier()
  br label %Flow7851, !amdgpu.uniform !42

if.end14.i.i.i3645:                               ; preds = %Flow7851
  %4838 = call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %4813, i1 true)
  %iszero.i.i.i.i.i.i3646 = icmp ne i64 %4813, 0
  %cmp2.i.i.i.i.i3647 = icmp eq i64 %4838, %sh_prom.i.i.i.i3270
  %cmp.i.i.i.i.i3648 = select i1 %iszero.i.i.i.i.i.i3646, i1 %cmp2.i.i.i.i.i3647, i1 false
  %4839 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.i.i.i.i.i3648)
  %4840 = extractvalue { i1, i64 } %4839, 0
  %4841 = extractvalue { i1, i64 } %4839, 1
  br i1 %4840, label %if.then16.i.i.i3650, label %if.end22.i.i.i3649

if.then16.i.i.i3650:                              ; preds = %if.end14.i.i.i3645
  %4842 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !76
  %opcode.i.i.i3651 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %4842, i64 %idxprom.i.i.i.i3575, i32 1
  store i32 1660944387, ptr %opcode.i.i.i3651, align 8, !tbaa !77
  %arrayidx21.i.i.i3652 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %4842, i64 %idxprom.i.i.i.i3575
  store i64 %4813, ptr %arrayidx21.i.i.i3652, align 8, !tbaa !80
  br label %if.end22.i.i.i3649, !amdgpu.uniform !42

Flow7852:                                         ; preds = %if.end22.i.i.i3649, %Flow7851
  %4843 = phi i32 [ %4830, %if.end22.i.i.i3649 ], [ %retval.sroa.8.0.i.i.i3560, %Flow7851 ]
  %4844 = phi i64 [ %4813, %if.end22.i.i.i3649 ], [ %retval.sroa.2.0.i.i.i3559, %Flow7851 ]
  %4845 = phi i32 [ %spec.store.select.i.i.i3563, %if.end22.i.i.i3649 ], [ %retval.sroa.6.0.i.i.i3557, %Flow7851 ]
  br label %cleanup26.i.i.i3586, !amdgpu.uniform !42

if.end22.i.i.i3649:                               ; preds = %if.then16.i.i.i3650, %if.end14.i.i.i3645
  call void @llvm.amdgcn.end.cf.i64(i64 %4841)
  call void @llvm.amdgcn.wave.barrier()
  br label %Flow7852, !amdgpu.uniform !42

Flow7853:                                         ; preds = %cleanup26.i.i.i3586, %4819
  %4846 = phi i32 [ %4845, %cleanup26.i.i.i3586 ], [ %4809, %4819 ]
  %4847 = phi i64 [ %4844, %cleanup26.i.i.i3586 ], [ %4810, %4819 ]
  %4848 = phi i32 [ %4843, %cleanup26.i.i.i3586 ], [ %4811, %4819 ]
  %4849 = phi i32 [ %4843, %cleanup26.i.i.i3586 ], [ %retval.sroa.8.0.i.i.i3560, %4819 ]
  %4850 = phi i64 [ %4844, %cleanup26.i.i.i3586 ], [ %retval.sroa.2.0.i.i.i3559, %4819 ]
  %4851 = phi i32 [ %4845, %cleanup26.i.i.i3586 ], [ %retval.sroa.6.0.i.i.i3557, %4819 ]
  %4852 = phi i1 [ %cmp.i46.not.i.i.i3578, %cleanup26.i.i.i3586 ], [ true, %4819 ]
  br i1 %4852, label %for.inc.i.i.i3591, label %Flow7854, !amdgpu.uniform !42

cleanup26.i.i.i3586:                              ; preds = %Flow7852
  br label %Flow7853, !amdgpu.uniform !42

for.inc.i.i.i3591:                                ; preds = %Flow7853
  %inc.i.i.i3596 = add i32 %spec.store.select.i.i.i3563, 1
  br label %Flow7854, !amdgpu.uniform !42

Flow7854:                                         ; preds = %for.inc.i.i.i3591, %Flow7853
  %4853 = phi i32 [ %inc.i.i.i3596, %for.inc.i.i.i3591 ], [ poison, %Flow7853 ]
  %4854 = phi i1 [ false, %for.inc.i.i.i3591 ], [ true, %Flow7853 ]
  br i1 %4854, label %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i3597, label %for.cond.i.i1.i3556, !amdgpu.uniform !42

_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i3597: ; preds = %Flow7854
  %.lcssa8904 = phi i32 [ %4846, %Flow7854 ]
  %.lcssa8903 = phi i64 [ %4847, %Flow7854 ]
  %.lcssa8902 = phi i32 [ %4848, %Flow7854 ]
  fence acquire
  %4855 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %mul.i.i.i.i.i.i3598 = shl i32 %.lcssa8904, 6
  %idxprom.i19.i.i.i.i.i3599 = zext i32 %mul.i.i.i.i.i.i3598 to i64
  %arrayidx.i20.i.i.i.i.i3600 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %4855, i64 %idxprom.i19.i.i.i.i.i3599
  %arrayidx.i22.i.i.i.i.i3601 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %arrayidx.i20.i.i.i.i.i3600, i64 %sh_prom.i.i.i.i3270
  store i64 sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.26.42 to ptr), i64 8) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.26.42 to ptr) to i64)), ptr %arrayidx.i22.i.i.i.i.i3601, align 8, !tbaa !82
  %spec.select.i.i.i.i.i.i.i3602 = call i64 @llvm.umin.i64(i64 sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.26.42 to ptr), i64 8) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.26.42 to ptr) to i64)), i64 56)
  %4856 = and i64 %spec.select.i.i.i.i.i.i.i3602, 15
  %4857 = sub i64 %spec.select.i.i.i.i.i.i.i3602, %4856
  %4858 = icmp ne i64 %4857, 0
  br i1 %4858, label %loop-memcpy-expansion5409.preheader, label %Flow7850, !amdgpu.uniform !42

loop-memcpy-expansion5409.preheader:              ; preds = %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i3597
  %4859 = add nuw nsw i64 %4513, 8
  %scevgep6448 = getelementptr i8, ptr %4855, i64 %4859
  %4860 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i3599, 6
  %scevgep6449 = getelementptr i8, ptr %scevgep6448, i64 %4860
  br label %loop-memcpy-expansion5409, !amdgpu.uniform !42

Flow7850:                                         ; preds = %Flow7849, %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i3597
  br label %loop-memcpy-residual-header5412, !amdgpu.uniform !42

loop-memcpy-expansion5409:                        ; preds = %loop-memcpy-expansion5409.preheader, %loop-memcpy-expansion5409
  %loop-index5410 = phi i64 [ %4863, %loop-memcpy-expansion5409 ], [ 0, %loop-memcpy-expansion5409.preheader ]
  %4861 = getelementptr i8, ptr addrspace(4) @.str.26.42, i64 %loop-index5410, !amdgpu.uniform !42
  %4862 = load <4 x i32>, ptr addrspace(4) %4861, align 1
  %scevgep6450 = getelementptr i8, ptr %scevgep6449, i64 %loop-index5410
  store <4 x i32> %4862, ptr %scevgep6450, align 1
  %4863 = add i64 %loop-index5410, 16
  %4864 = icmp uge i64 %4863, %4857
  br i1 %4864, label %Flow7849, label %loop-memcpy-expansion5409, !amdgpu.uniform !42

Flow7848:                                         ; preds = %Flow7846, %Flow7847
  %4865 = phi i64 [ %6653, %Flow7846 ], [ %6650, %Flow7847 ]
  br label %post-loop-memcpy-expansion5408, !amdgpu.uniform !42

loop-memcpy-residual5411:                         ; preds = %loop-memcpy-residual5411.preheader, %loop-memcpy-residual5411
  %residual-loop-index5413 = phi i64 [ %4867, %loop-memcpy-residual5411 ], [ 0, %loop-memcpy-residual5411.preheader ]
  %scevgep6447 = getelementptr i8, ptr addrspace(4) @.str.26.42, i64 %residual-loop-index5413, !amdgpu.uniform !42
  %4866 = load i8, ptr addrspace(4) %scevgep6447, align 1
  %scevgep6446 = getelementptr i8, ptr %scevgep6445, i64 %residual-loop-index5413
  store i8 %4866, ptr %scevgep6446, align 1
  %4867 = add i64 %residual-loop-index5413, 1
  %4868 = icmp uge i64 %4867, %4856
  br i1 %4868, label %Flow7846, label %loop-memcpy-residual5411, !amdgpu.uniform !42

Flow7846:                                         ; preds = %loop-memcpy-residual5411
  br label %Flow7848, !amdgpu.uniform !42

post-loop-memcpy-expansion5408:                   ; preds = %Flow7848
  %tobool.not.i.i.i.i.i.i3604 = icmp eq i32 %.lcssa8902, 0
  %conv.i.i.i.i.i.i3605 = zext i1 %tobool.not.i.i.i.i.i.i3604 to i32
  fence release
  %4869 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %idxprom.i23.i.i.i.i.i3606 = zext i32 %.lcssa8904 to i64
  %arrayidx.i24.i.i.i.i.i3607 = getelementptr inbounds nuw i32, ptr %4869, i64 %idxprom.i23.i.i.i.i.i3606
  store atomic i32 %conv.i.i.i.i.i.i3605, ptr %arrayidx.i24.i.i.i.i.i3607 monotonic, align 4
  %4870 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !84
  %arrayidx.i.i5.i.i3608 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %4870, i64 %idxprom.i23.i.i.i.i.i3606, !amdgpu.uniform !42
  %4871 = load i64, ptr %arrayidx.i.i5.i.i3608, align 8, !tbaa !80
  br label %while.cond.i.i.i.i3609, !amdgpu.uniform !42

while.cond.i.i.i.i3609:                           ; preds = %post-loop-memcpy-expansion5408, %Flow7845
  %phi.broken8790 = phi i64 [ 0, %post-loop-memcpy-expansion5408 ], [ %4888, %Flow7845 ]
  %lsr.iv6437 = phi i64 [ %4886, %Flow7845 ], [ -48, %post-loop-memcpy-expansion5408 ]
  %lsr.iv6432 = phi ptr addrspace(4) [ %4885, %Flow7845 ], [ getelementptr (i8, ptr addrspace(4) @.str.26.42, i64 56), %post-loop-memcpy-expansion5408 ]
  %port.sroa.37.0.i.i3610 = phi i32 [ %4884, %Flow7845 ], [ %conv.i.i.i.i.i.i3605, %post-loop-memcpy-expansion5408 ]
  %idx.0.i.i.i.i3611 = phi i64 [ %4883, %Flow7845 ], [ 56, %post-loop-memcpy-expansion5408 ]
  %cmp.i.i.i.i3612 = icmp ult i64 %idx.0.i.i.i.i3611, sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.26.42 to ptr), i64 8) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.26.42 to ptr) to i64))
  %4872 = call i64 @llvm.amdgcn.ballot.i64(i1 %cmp.i.i.i.i3612)
  %and.i.i.i.i6.i.i3613 = and i64 %4872, %4871
  %tobool.not.i.i.i.i3614 = icmp ne i64 %and.i.i.i.i6.i.i3613, 0
  %4873 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i22.i.i.i3615 = getelementptr inbounds nuw i32, ptr %4873, i64 %idxprom.i23.i.i.i.i.i3606, !amdgpu.uniform !42
  %4874 = load atomic i32, ptr %arrayidx.i22.i.i.i3615 monotonic, align 4
  %4875 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %tobool.not.i.i.i.i3614)
  %4876 = extractvalue { i1, i64 } %4875, 0
  %4877 = extractvalue { i1, i64 } %4875, 1
  br i1 %4876, label %cond.false.i8.i.i.i.i3616, label %Flow7845

cond.false.i8.i.i.i.i3616:                        ; preds = %while.cond.i.i.i.i3609
  %4878 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %4874)
  %4879 = icmp ne i32 %4878, %port.sroa.37.0.i.i3610
  br i1 %4879, label %while.body.i.i21.i.i.i.i3617.preheader, label %Flow7844, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i3617.preheader:           ; preds = %cond.false.i8.i.i.i.i3616
  br label %while.body.i.i21.i.i.i.i3617, !amdgpu.uniform !42

Flow7844:                                         ; preds = %Flow7843, %cond.false.i8.i.i.i.i3616
  br label %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i3620, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i3617:                     ; preds = %while.body.i.i21.i.i.i.i3617.preheader, %while.body.i.i21.i.i.i.i3617
  call void @llvm.amdgcn.s.sleep(i32 2)
  %4880 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i22.i.i.i.i3618 = getelementptr inbounds nuw i32, ptr %4880, i64 %idxprom.i23.i.i.i.i.i3606, !amdgpu.uniform !42
  %4881 = load atomic i32, ptr %arrayidx.i.i.i22.i.i.i.i3618 monotonic, align 4
  %4882 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %4881)
  %cmp.i.not.i.i23.i.i.i.i3619 = icmp eq i32 %4882, %port.sroa.37.0.i.i3610
  br i1 %cmp.i.not.i.i23.i.i.i.i3619, label %Flow7843, label %while.body.i.i21.i.i.i.i3617, !amdgpu.uniform !42

Flow7843:                                         ; preds = %while.body.i.i21.i.i.i.i3617
  br label %Flow7844, !amdgpu.uniform !42

Flow7845:                                         ; preds = %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i3621, %while.cond.i.i.i.i3609
  %4883 = phi i64 [ %add.i.i.i.i3624, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i3621 ], [ poison, %while.cond.i.i.i.i3609 ]
  %4884 = phi i32 [ %conv.i.i26.i.i.i.i3622, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i3621 ], [ poison, %while.cond.i.i.i.i3609 ]
  %4885 = phi ptr addrspace(4) [ %scevgep6433, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i3621 ], [ poison, %while.cond.i.i.i.i3609 ]
  %4886 = phi i64 [ %lsr.iv.next6438, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i3621 ], [ poison, %while.cond.i.i.i.i3609 ]
  %4887 = phi i1 [ false, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i3621 ], [ true, %while.cond.i.i.i.i3609 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %4877)
  %4888 = call i64 @llvm.amdgcn.if.break.i64(i1 %4887, i64 %phi.broken8790)
  %4889 = call i1 @llvm.amdgcn.loop.i64(i64 %4888)
  br i1 %4889, label %cond.false.i.i.i3631, label %while.cond.i.i.i.i3609

_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i3620: ; preds = %Flow7844
  fence acquire
  br i1 %cmp.i.i.i.i3612, label %if.then.i.i.i.i.i.i.i3625, label %Flow7842, !amdgpu.uniform !42

if.then.i.i.i.i.i.i.i3625:                        ; preds = %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i3620
  %sub.i.i.i.i.i.i.i3627 = sub nuw i64 sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.26.42 to ptr), i64 8) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.26.42 to ptr) to i64)), %idx.0.i.i.i.i3611
  %spec.select.i.i.i36.i.i.i.i3628 = call i64 @llvm.umin.i64(i64 %sub.i.i.i.i.i.i.i3627, i64 64)
  %4890 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %4891 = and i64 %spec.select.i.i.i36.i.i.i.i3628, 15
  %4892 = sub i64 %spec.select.i.i.i36.i.i.i.i3628, %4891
  %4893 = icmp ne i64 %4892, 0
  br i1 %4893, label %loop-memcpy-expansion5415.preheader, label %Flow7841, !amdgpu.uniform !42

loop-memcpy-expansion5415.preheader:              ; preds = %if.then.i.i.i.i.i.i.i3625
  %arrayidx.i20.i35.i.i.i.i3629 = getelementptr %"struct.rpc::Buffer", ptr %4890, i64 %idxprom.i19.i.i.i.i.i3599
  %arrayidx.i22.i38.i.i.i.i3630 = getelementptr %"struct.rpc::Buffer", ptr %arrayidx.i20.i35.i.i.i.i3629, i64 %sh_prom.i.i.i.i3270
  br label %loop-memcpy-expansion5415, !amdgpu.uniform !42

Flow7841:                                         ; preds = %Flow7840, %if.then.i.i.i.i.i.i.i3625
  br label %loop-memcpy-residual-header5418, !amdgpu.uniform !42

loop-memcpy-expansion5415:                        ; preds = %loop-memcpy-expansion5415.preheader, %loop-memcpy-expansion5415
  %loop-index5416 = phi i64 [ %4896, %loop-memcpy-expansion5415 ], [ 0, %loop-memcpy-expansion5415.preheader ]
  %scevgep6434 = getelementptr i8, ptr addrspace(4) %lsr.iv6432, i64 %loop-index5416, !amdgpu.uniform !42
  %4894 = load <4 x i32>, ptr addrspace(4) %scevgep6434, align 1
  %4895 = getelementptr i8, ptr %arrayidx.i22.i38.i.i.i.i3630, i64 %loop-index5416
  store <4 x i32> %4894, ptr %4895, align 1
  %4896 = add i64 %loop-index5416, 16
  %4897 = icmp uge i64 %4896, %4892
  br i1 %4897, label %Flow7840, label %loop-memcpy-expansion5415, !amdgpu.uniform !42

Flow7839:                                         ; preds = %Flow7838, %loop-memcpy-residual-header5418
  br label %Flow7842, !amdgpu.uniform !42

loop-memcpy-residual5417:                         ; preds = %loop-memcpy-residual5417.preheader, %loop-memcpy-residual5417
  %residual-loop-index5419 = phi i64 [ %4899, %loop-memcpy-residual5417 ], [ 0, %loop-memcpy-residual5417.preheader ]
  %scevgep6443 = getelementptr i8, ptr addrspace(4) %scevgep6442, i64 %residual-loop-index5419, !amdgpu.uniform !42
  %4898 = load i8, ptr addrspace(4) %scevgep6443, align 1
  %scevgep6441 = getelementptr i8, ptr %scevgep6440, i64 %residual-loop-index5419
  store i8 %4898, ptr %scevgep6441, align 1
  %4899 = add i64 %residual-loop-index5419, 1
  %4900 = icmp uge i64 %4899, %4891
  br i1 %4900, label %Flow7838, label %loop-memcpy-residual5417, !amdgpu.uniform !42

Flow7838:                                         ; preds = %loop-memcpy-residual5417
  br label %Flow7839, !amdgpu.uniform !42

_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i3621: ; preds = %Flow7842
  %conv.i.i26.i.i.i.i3622 = xor i32 %port.sroa.37.0.i.i3610, 1
  fence release
  %4901 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %arrayidx.i24.i29.i.i.i.i3623 = getelementptr inbounds nuw i32, ptr %4901, i64 %idxprom.i23.i.i.i.i.i3606
  store atomic i32 %conv.i.i26.i.i.i.i3622, ptr %arrayidx.i24.i29.i.i.i.i3623 monotonic, align 4
  %add.i.i.i.i3624 = add i64 %idx.0.i.i.i.i3611, 64
  %scevgep6433 = getelementptr i8, ptr addrspace(4) %lsr.iv6432, i64 64
  %lsr.iv.next6438 = add i64 %lsr.iv6437, -64
  br label %Flow7845, !amdgpu.uniform !42

cond.false.i.i.i3631:                             ; preds = %Flow7845
  %.lcssa8898 = phi i64 [ %4888, %Flow7845 ]
  %port.sroa.37.0.i.i3610.lcssa = phi i32 [ %port.sroa.37.0.i.i3610, %Flow7845 ]
  %.lcssa8897 = phi i32 [ %4874, %Flow7845 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa8898)
  %4902 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %.lcssa8897)
  %4903 = icmp ne i32 %4902, %port.sroa.37.0.i.i3610.lcssa
  %4904 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %4903)
  %4905 = extractvalue { i1, i64 } %4904, 0
  %4906 = extractvalue { i1, i64 } %4904, 1
  br i1 %4905, label %while.body.i.i.i.i3632.preheader, label %Flow7837

while.body.i.i.i.i3632.preheader:                 ; preds = %cond.false.i.i.i3631
  br label %while.body.i.i.i.i3632, !amdgpu.uniform !42

Flow7837:                                         ; preds = %Flow7836, %cond.false.i.i.i3631
  call void @llvm.amdgcn.end.cf.i64(i64 %4906)
  br label %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit3653, !amdgpu.uniform !42

while.body.i.i.i.i3632:                           ; preds = %while.body.i.i.i.i3632.preheader, %while.body.i.i.i.i3632
  %phi.broken8791 = phi i64 [ 0, %while.body.i.i.i.i3632.preheader ], [ %4910, %while.body.i.i.i.i3632 ]
  call void @llvm.amdgcn.s.sleep(i32 2)
  %4907 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i8.i.i3633 = getelementptr inbounds nuw i32, ptr %4907, i64 %idxprom.i23.i.i.i.i.i3606, !amdgpu.uniform !42
  %4908 = load atomic i32, ptr %arrayidx.i.i.i8.i.i3633 monotonic, align 4
  %4909 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %4908)
  %cmp.i.not.i.i.i.i3634 = icmp eq i32 %4909, %port.sroa.37.0.i.i3610.lcssa
  %4910 = call i64 @llvm.amdgcn.if.break.i64(i1 %cmp.i.not.i.i.i.i3634, i64 %phi.broken8791)
  %4911 = call i1 @llvm.amdgcn.loop.i64(i64 %4910)
  br i1 %4911, label %Flow7836, label %while.body.i.i.i.i3632

Flow7836:                                         ; preds = %while.body.i.i.i.i3632
  %.lcssa8896 = phi i64 [ %4910, %while.body.i.i.i.i3632 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa8896)
  br label %Flow7837, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit3653: ; preds = %Flow7837
  fence acquire
  call void @llvm.amdgcn.wave.barrier()
  fence syncscope("agent") release
  %4912 = call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %.lcssa8903, i1 true)
  %iszero.i.i.i.i.i13.i.i3636 = icmp ne i64 %.lcssa8903, 0
  %cmp2.i.i.i.i14.i.i3637 = icmp eq i64 %4912, %sh_prom.i.i.i.i3270
  %cmp.i.i.i.i15.i.i3638 = select i1 %iszero.i.i.i.i.i13.i.i3636, i1 %cmp2.i.i.i.i14.i.i3637, i1 false
  %4913 = and i32 %.lcssa8904, 31
  %conv4.i.i.i19.i.i3642 = zext i1 %cmp.i.i.i.i15.i.i3638 to i32
  %shl.i.i.i20.i.i3643 = shl nuw i32 %conv4.i.i.i19.i.i3642, %4913
  %xor.i.i.i21.i.i3644 = xor i32 %shl.i.i.i20.i.i3643, -1
  %4914 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop5888, !amdgpu.uniform !42

4915:                                             ; preds = %ComputeEnd5889
  %div8.i.i.i16.i.i3639 = lshr i32 %.lcssa8904, 5
  %idxprom.i.i.i17.i.i3640 = zext nneg i32 %div8.i.i.i16.i.i3639 to i64
  %4916 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i17.i.i3640
  %sunkaddr7479 = getelementptr inbounds i8, ptr addrspace(1) %4916, i64 40
  %4917 = atomicrmw and ptr addrspace(1) %sunkaddr7479, i32 %.lcssa8895 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %4918, !amdgpu.uniform !42

4918:                                             ; preds = %4915, %ComputeEnd5889
  call void @llvm.amdgcn.end.cf.i64(i64 %8470)
  call void @llvm.amdgcn.wave.barrier()
  br label %for.cond.i.i1.i3662, !amdgpu.uniform !42

for.cond.i.i1.i3662:                              ; preds = %4918, %Flow7835
  %4919 = phi i32 [ %4956, %Flow7835 ], [ poison, %4918 ]
  %4920 = phi i64 [ %4957, %Flow7835 ], [ poison, %4918 ]
  %4921 = phi i32 [ %4958, %Flow7835 ], [ poison, %4918 ]
  %retval.sroa.6.0.i.i.i3663 = phi i32 [ %4961, %Flow7835 ], [ undef, %4918 ]
  %retval.sroa.2.0.i.i.i3665 = phi i64 [ %4960, %Flow7835 ], [ undef, %4918 ]
  %retval.sroa.8.0.i.i.i3666 = phi i32 [ %4959, %Flow7835 ], [ undef, %4918 ]
  %index.0.i.i.i3667 = phi i32 [ %4963, %Flow7835 ], [ 0, %4918 ]
  %4922 = load i32, ptr addrspace(1) @__llvm_rpc_client, align 8, !tbaa !68
  %cmp.not.i.i.i3668 = icmp ult i32 %index.0.i.i.i3667, %4922
  %spec.store.select.i.i.i3669 = select i1 %cmp.not.i.i.i3668, i32 %index.0.i.i.i3667, i32 0
  %4923 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  %4924 = and i32 %spec.store.select.i.i.i3669, 31
  %4925 = lshr i64 %4923, %sh_prom.i.i.i.i3270
  %4926 = trunc i64 %4925 to i32
  %conv4.i.i.i.i.i3673 = and i32 %4926, 1
  %shl.i.i.i.i.i3674 = shl nuw i32 %conv4.i.i.i.i.i3673, %4924
  br label %ComputeLoop5892, !amdgpu.uniform !42

4927:                                             ; preds = %ComputeEnd5893
  %sunkaddr7480 = getelementptr inbounds i8, ptr addrspace(1) %8480, i64 40
  %4928 = atomicrmw or ptr addrspace(1) %sunkaddr7480, i32 %.lcssa8889 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %4929, !amdgpu.uniform !42

4929:                                             ; preds = %4927, %ComputeEnd5893
  %4930 = phi i32 [ %4928, %4927 ], [ poison, %ComputeEnd5893 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %8489)
  %4931 = call i32 @llvm.amdgcn.readfirstlane.i32(i32 %4930)
  %4932 = or i32 %4931, %.lcssa8890
  %shl5.i.i.i.i.i3675 = shl nuw i32 1, %4924
  %and.i.i.i.i.i3676 = and i32 %shl5.i.i.i.i.i3675, %4932
  %tobool3.i.i.i.i3677 = icmp ne i32 %and.i.i.i.i.i3676, 0
  %4933 = call i64 @llvm.amdgcn.ballot.i64(i1 %tobool3.i.i.i.i3677)
  %and.i.i.i.i.i.i3678 = and i64 %4933, %4923
  %cmp.i.not.i.i.i3679 = icmp ne i64 %4923, %and.i.i.i.i.i.i3678
  br i1 %cmp.i.not.i.i.i3679, label %if.then.i.i.i.i3680, label %Flow7834, !amdgpu.uniform !42

if.then.i.i.i.i3680:                              ; preds = %4929
  fence syncscope("agent") acquire
  %4934 = load <2 x i64>, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !75
  %4935 = extractelement <2 x i64> %4934, i32 0
  %4936 = inttoptr i64 %4935 to ptr
  %4937 = extractelement <2 x i64> %4934, i32 1
  %4938 = inttoptr i64 %4937 to ptr
  %idxprom.i.i.i.i3681 = zext i32 %spec.store.select.i.i.i3669 to i64
  %arrayidx.i.i.i.i3682 = getelementptr inbounds nuw i32, ptr %4936, i64 %idxprom.i.i.i.i3681, !amdgpu.uniform !42
  %4939 = load atomic i32, ptr %arrayidx.i.i.i.i3682 monotonic, align 4
  %4940 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %4939)
  %arrayidx.i45.i.i.i3683 = getelementptr inbounds nuw i32, ptr %4938, i64 %idxprom.i.i.i.i3681, !amdgpu.uniform !42
  %4941 = load atomic i32, ptr %arrayidx.i45.i.i.i3683 monotonic, align 4
  %4942 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %4941)
  %cmp.i46.not.i.i.i3684 = icmp ne i32 %4940, %4942
  br i1 %cmp.i46.not.i.i.i3684, label %if.then12.i.i.i3685, label %Flow7832, !amdgpu.uniform !42

if.then12.i.i.i3685:                              ; preds = %if.then.i.i.i.i3680
  fence syncscope("agent") release
  %4943 = call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %4923, i1 true)
  %iszero.i.i.i.i.i.i.i3686 = icmp ne i64 %4923, 0
  %cmp2.i.i.i.i.i.i3687 = icmp eq i64 %4943, %sh_prom.i.i.i.i3270
  %cmp.i.i.i.i.i.i3688 = select i1 %iszero.i.i.i.i.i.i.i3686, i1 %cmp2.i.i.i.i.i.i3687, i1 false
  %conv4.i.i51.i.i.i3689 = zext i1 %cmp.i.i.i.i.i.i3688 to i32
  %shl.i.i52.i.i.i3690 = shl nuw i32 %conv4.i.i51.i.i.i3689, %4924
  %xor.i.i.i.i.i3691 = xor i32 %shl.i.i52.i.i.i3690, -1
  %4944 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop5897, !amdgpu.uniform !42

4945:                                             ; preds = %ComputeEnd5898
  %sunkaddr7481 = getelementptr inbounds i8, ptr addrspace(1) %8480, i64 40
  %4946 = atomicrmw and ptr addrspace(1) %sunkaddr7481, i32 %.lcssa8891 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %4947, !amdgpu.uniform !42

4947:                                             ; preds = %4945, %ComputeEnd5898
  call void @llvm.amdgcn.end.cf.i64(i64 %8507)
  call void @llvm.amdgcn.wave.barrier()
  br label %Flow7832, !amdgpu.uniform !42

if.end14.i.i.i3751:                               ; preds = %Flow7832
  %4948 = call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %4923, i1 true)
  %iszero.i.i.i.i.i.i3752 = icmp ne i64 %4923, 0
  %cmp2.i.i.i.i.i3753 = icmp eq i64 %4948, %sh_prom.i.i.i.i3270
  %cmp.i.i.i.i.i3754 = select i1 %iszero.i.i.i.i.i.i3752, i1 %cmp2.i.i.i.i.i3753, i1 false
  %4949 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.i.i.i.i.i3754)
  %4950 = extractvalue { i1, i64 } %4949, 0
  %4951 = extractvalue { i1, i64 } %4949, 1
  br i1 %4950, label %if.then16.i.i.i3756, label %if.end22.i.i.i3755

if.then16.i.i.i3756:                              ; preds = %if.end14.i.i.i3751
  %4952 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !76
  %opcode.i.i.i3757 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %4952, i64 %idxprom.i.i.i.i3681, i32 1
  store i32 1660944387, ptr %opcode.i.i.i3757, align 8, !tbaa !77
  %arrayidx21.i.i.i3758 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %4952, i64 %idxprom.i.i.i.i3681
  store i64 %4923, ptr %arrayidx21.i.i.i3758, align 8, !tbaa !80
  br label %if.end22.i.i.i3755, !amdgpu.uniform !42

Flow7833:                                         ; preds = %if.end22.i.i.i3755, %Flow7832
  %4953 = phi i32 [ %4940, %if.end22.i.i.i3755 ], [ %retval.sroa.8.0.i.i.i3666, %Flow7832 ]
  %4954 = phi i64 [ %4923, %if.end22.i.i.i3755 ], [ %retval.sroa.2.0.i.i.i3665, %Flow7832 ]
  %4955 = phi i32 [ %spec.store.select.i.i.i3669, %if.end22.i.i.i3755 ], [ %retval.sroa.6.0.i.i.i3663, %Flow7832 ]
  br label %cleanup26.i.i.i3692, !amdgpu.uniform !42

if.end22.i.i.i3755:                               ; preds = %if.then16.i.i.i3756, %if.end14.i.i.i3751
  call void @llvm.amdgcn.end.cf.i64(i64 %4951)
  call void @llvm.amdgcn.wave.barrier()
  br label %Flow7833, !amdgpu.uniform !42

Flow7834:                                         ; preds = %cleanup26.i.i.i3692, %4929
  %4956 = phi i32 [ %4955, %cleanup26.i.i.i3692 ], [ %4919, %4929 ]
  %4957 = phi i64 [ %4954, %cleanup26.i.i.i3692 ], [ %4920, %4929 ]
  %4958 = phi i32 [ %4953, %cleanup26.i.i.i3692 ], [ %4921, %4929 ]
  %4959 = phi i32 [ %4953, %cleanup26.i.i.i3692 ], [ %retval.sroa.8.0.i.i.i3666, %4929 ]
  %4960 = phi i64 [ %4954, %cleanup26.i.i.i3692 ], [ %retval.sroa.2.0.i.i.i3665, %4929 ]
  %4961 = phi i32 [ %4955, %cleanup26.i.i.i3692 ], [ %retval.sroa.6.0.i.i.i3663, %4929 ]
  %4962 = phi i1 [ %cmp.i46.not.i.i.i3684, %cleanup26.i.i.i3692 ], [ true, %4929 ]
  br i1 %4962, label %for.inc.i.i.i3697, label %Flow7835, !amdgpu.uniform !42

cleanup26.i.i.i3692:                              ; preds = %Flow7833
  br label %Flow7834, !amdgpu.uniform !42

for.inc.i.i.i3697:                                ; preds = %Flow7834
  %inc.i.i.i3702 = add i32 %spec.store.select.i.i.i3669, 1
  br label %Flow7835, !amdgpu.uniform !42

Flow7835:                                         ; preds = %for.inc.i.i.i3697, %Flow7834
  %4963 = phi i32 [ %inc.i.i.i3702, %for.inc.i.i.i3697 ], [ poison, %Flow7834 ]
  %4964 = phi i1 [ false, %for.inc.i.i.i3697 ], [ true, %Flow7834 ]
  br i1 %4964, label %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i3703, label %for.cond.i.i1.i3662, !amdgpu.uniform !42

_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i3703: ; preds = %Flow7835
  %.lcssa8894 = phi i32 [ %4956, %Flow7835 ]
  %.lcssa8893 = phi i64 [ %4957, %Flow7835 ]
  %.lcssa8892 = phi i32 [ %4958, %Flow7835 ]
  fence acquire
  %4965 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %mul.i.i.i.i.i.i3704 = shl i32 %.lcssa8894, 6
  %idxprom.i19.i.i.i.i.i3705 = zext i32 %mul.i.i.i.i.i.i3704 to i64
  %arrayidx.i20.i.i.i.i.i3706 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %4965, i64 %idxprom.i19.i.i.i.i.i3705
  %arrayidx.i22.i.i.i.i.i3707 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %arrayidx.i20.i.i.i.i.i3706, i64 %sh_prom.i.i.i.i3270
  store i64 sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.27.43 to ptr), i64 7) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.27.43 to ptr) to i64)), ptr %arrayidx.i22.i.i.i.i.i3707, align 8, !tbaa !82
  %spec.select.i.i.i.i.i.i.i3708 = call i64 @llvm.umin.i64(i64 sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.27.43 to ptr), i64 7) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.27.43 to ptr) to i64)), i64 56)
  %4966 = and i64 %spec.select.i.i.i.i.i.i.i3708, 15
  %4967 = sub i64 %spec.select.i.i.i.i.i.i.i3708, %4966
  %4968 = icmp ne i64 %4967, 0
  br i1 %4968, label %loop-memcpy-expansion5421.preheader, label %Flow7831, !amdgpu.uniform !42

loop-memcpy-expansion5421.preheader:              ; preds = %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i3703
  %4969 = add nuw nsw i64 %4513, 8
  %scevgep6429 = getelementptr i8, ptr %4965, i64 %4969
  %4970 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i3705, 6
  %scevgep6430 = getelementptr i8, ptr %scevgep6429, i64 %4970
  br label %loop-memcpy-expansion5421, !amdgpu.uniform !42

Flow7831:                                         ; preds = %Flow7830, %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i3703
  br label %loop-memcpy-residual-header5424, !amdgpu.uniform !42

loop-memcpy-expansion5421:                        ; preds = %loop-memcpy-expansion5421.preheader, %loop-memcpy-expansion5421
  %loop-index5422 = phi i64 [ %4973, %loop-memcpy-expansion5421 ], [ 0, %loop-memcpy-expansion5421.preheader ]
  %4971 = getelementptr i8, ptr addrspace(4) @.str.27.43, i64 %loop-index5422, !amdgpu.uniform !42
  %4972 = load <4 x i32>, ptr addrspace(4) %4971, align 1
  %scevgep6431 = getelementptr i8, ptr %scevgep6430, i64 %loop-index5422
  store <4 x i32> %4972, ptr %scevgep6431, align 1
  %4973 = add i64 %loop-index5422, 16
  %4974 = icmp uge i64 %4973, %4967
  br i1 %4974, label %Flow7830, label %loop-memcpy-expansion5421, !amdgpu.uniform !42

Flow7829:                                         ; preds = %Flow7827, %Flow7828
  %4975 = phi i64 [ %6659, %Flow7827 ], [ %6656, %Flow7828 ]
  br label %post-loop-memcpy-expansion5420, !amdgpu.uniform !42

loop-memcpy-residual5423:                         ; preds = %loop-memcpy-residual5423.preheader, %loop-memcpy-residual5423
  %residual-loop-index5425 = phi i64 [ %4977, %loop-memcpy-residual5423 ], [ 0, %loop-memcpy-residual5423.preheader ]
  %scevgep6428 = getelementptr i8, ptr addrspace(4) @.str.27.43, i64 %residual-loop-index5425, !amdgpu.uniform !42
  %4976 = load i8, ptr addrspace(4) %scevgep6428, align 1
  %scevgep6427 = getelementptr i8, ptr %scevgep6426, i64 %residual-loop-index5425
  store i8 %4976, ptr %scevgep6427, align 1
  %4977 = add i64 %residual-loop-index5425, 1
  %4978 = icmp uge i64 %4977, %4966
  br i1 %4978, label %Flow7827, label %loop-memcpy-residual5423, !amdgpu.uniform !42

Flow7827:                                         ; preds = %loop-memcpy-residual5423
  br label %Flow7829, !amdgpu.uniform !42

post-loop-memcpy-expansion5420:                   ; preds = %Flow7829
  %tobool.not.i.i.i.i.i.i3710 = icmp eq i32 %.lcssa8892, 0
  %conv.i.i.i.i.i.i3711 = zext i1 %tobool.not.i.i.i.i.i.i3710 to i32
  fence release
  %4979 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %idxprom.i23.i.i.i.i.i3712 = zext i32 %.lcssa8894 to i64
  %arrayidx.i24.i.i.i.i.i3713 = getelementptr inbounds nuw i32, ptr %4979, i64 %idxprom.i23.i.i.i.i.i3712
  store atomic i32 %conv.i.i.i.i.i.i3711, ptr %arrayidx.i24.i.i.i.i.i3713 monotonic, align 4
  %4980 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !84
  %arrayidx.i.i5.i.i3714 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %4980, i64 %idxprom.i23.i.i.i.i.i3712, !amdgpu.uniform !42
  %4981 = load i64, ptr %arrayidx.i.i5.i.i3714, align 8, !tbaa !80
  br label %while.cond.i.i.i.i3715, !amdgpu.uniform !42

while.cond.i.i.i.i3715:                           ; preds = %post-loop-memcpy-expansion5420, %Flow7826
  %phi.broken8792 = phi i64 [ 0, %post-loop-memcpy-expansion5420 ], [ %4998, %Flow7826 ]
  %lsr.iv6418 = phi i64 [ %4996, %Flow7826 ], [ -49, %post-loop-memcpy-expansion5420 ]
  %lsr.iv6413 = phi ptr addrspace(4) [ %4995, %Flow7826 ], [ getelementptr (i8, ptr addrspace(4) @.str.27.43, i64 56), %post-loop-memcpy-expansion5420 ]
  %port.sroa.37.0.i.i3716 = phi i32 [ %4994, %Flow7826 ], [ %conv.i.i.i.i.i.i3711, %post-loop-memcpy-expansion5420 ]
  %idx.0.i.i.i.i3717 = phi i64 [ %4993, %Flow7826 ], [ 56, %post-loop-memcpy-expansion5420 ]
  %cmp.i.i.i.i3718 = icmp ult i64 %idx.0.i.i.i.i3717, sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.27.43 to ptr), i64 7) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.27.43 to ptr) to i64))
  %4982 = call i64 @llvm.amdgcn.ballot.i64(i1 %cmp.i.i.i.i3718)
  %and.i.i.i.i6.i.i3719 = and i64 %4982, %4981
  %tobool.not.i.i.i.i3720 = icmp ne i64 %and.i.i.i.i6.i.i3719, 0
  %4983 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i22.i.i.i3721 = getelementptr inbounds nuw i32, ptr %4983, i64 %idxprom.i23.i.i.i.i.i3712, !amdgpu.uniform !42
  %4984 = load atomic i32, ptr %arrayidx.i22.i.i.i3721 monotonic, align 4
  %4985 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %tobool.not.i.i.i.i3720)
  %4986 = extractvalue { i1, i64 } %4985, 0
  %4987 = extractvalue { i1, i64 } %4985, 1
  br i1 %4986, label %cond.false.i8.i.i.i.i3722, label %Flow7826

cond.false.i8.i.i.i.i3722:                        ; preds = %while.cond.i.i.i.i3715
  %4988 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %4984)
  %4989 = icmp ne i32 %4988, %port.sroa.37.0.i.i3716
  br i1 %4989, label %while.body.i.i21.i.i.i.i3723.preheader, label %Flow7825, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i3723.preheader:           ; preds = %cond.false.i8.i.i.i.i3722
  br label %while.body.i.i21.i.i.i.i3723, !amdgpu.uniform !42

Flow7825:                                         ; preds = %Flow7824, %cond.false.i8.i.i.i.i3722
  br label %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i3726, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i3723:                     ; preds = %while.body.i.i21.i.i.i.i3723.preheader, %while.body.i.i21.i.i.i.i3723
  call void @llvm.amdgcn.s.sleep(i32 2)
  %4990 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i22.i.i.i.i3724 = getelementptr inbounds nuw i32, ptr %4990, i64 %idxprom.i23.i.i.i.i.i3712, !amdgpu.uniform !42
  %4991 = load atomic i32, ptr %arrayidx.i.i.i22.i.i.i.i3724 monotonic, align 4
  %4992 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %4991)
  %cmp.i.not.i.i23.i.i.i.i3725 = icmp eq i32 %4992, %port.sroa.37.0.i.i3716
  br i1 %cmp.i.not.i.i23.i.i.i.i3725, label %Flow7824, label %while.body.i.i21.i.i.i.i3723, !amdgpu.uniform !42

Flow7824:                                         ; preds = %while.body.i.i21.i.i.i.i3723
  br label %Flow7825, !amdgpu.uniform !42

Flow7826:                                         ; preds = %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i3727, %while.cond.i.i.i.i3715
  %4993 = phi i64 [ %add.i.i.i.i3730, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i3727 ], [ poison, %while.cond.i.i.i.i3715 ]
  %4994 = phi i32 [ %conv.i.i26.i.i.i.i3728, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i3727 ], [ poison, %while.cond.i.i.i.i3715 ]
  %4995 = phi ptr addrspace(4) [ %scevgep6414, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i3727 ], [ poison, %while.cond.i.i.i.i3715 ]
  %4996 = phi i64 [ %lsr.iv.next6419, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i3727 ], [ poison, %while.cond.i.i.i.i3715 ]
  %4997 = phi i1 [ false, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i3727 ], [ true, %while.cond.i.i.i.i3715 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %4987)
  %4998 = call i64 @llvm.amdgcn.if.break.i64(i1 %4997, i64 %phi.broken8792)
  %4999 = call i1 @llvm.amdgcn.loop.i64(i64 %4998)
  br i1 %4999, label %cond.false.i.i.i3737, label %while.cond.i.i.i.i3715

_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i3726: ; preds = %Flow7825
  fence acquire
  br i1 %cmp.i.i.i.i3718, label %if.then.i.i.i.i.i.i.i3731, label %Flow7823, !amdgpu.uniform !42

if.then.i.i.i.i.i.i.i3731:                        ; preds = %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i3726
  %sub.i.i.i.i.i.i.i3733 = sub nuw i64 sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.27.43 to ptr), i64 7) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.27.43 to ptr) to i64)), %idx.0.i.i.i.i3717
  %spec.select.i.i.i36.i.i.i.i3734 = call i64 @llvm.umin.i64(i64 %sub.i.i.i.i.i.i.i3733, i64 64)
  %5000 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %5001 = and i64 %spec.select.i.i.i36.i.i.i.i3734, 15
  %5002 = sub i64 %spec.select.i.i.i36.i.i.i.i3734, %5001
  %5003 = icmp ne i64 %5002, 0
  br i1 %5003, label %loop-memcpy-expansion5427.preheader, label %Flow7822, !amdgpu.uniform !42

loop-memcpy-expansion5427.preheader:              ; preds = %if.then.i.i.i.i.i.i.i3731
  %arrayidx.i20.i35.i.i.i.i3735 = getelementptr %"struct.rpc::Buffer", ptr %5000, i64 %idxprom.i19.i.i.i.i.i3705
  %arrayidx.i22.i38.i.i.i.i3736 = getelementptr %"struct.rpc::Buffer", ptr %arrayidx.i20.i35.i.i.i.i3735, i64 %sh_prom.i.i.i.i3270
  br label %loop-memcpy-expansion5427, !amdgpu.uniform !42

Flow7822:                                         ; preds = %Flow7821, %if.then.i.i.i.i.i.i.i3731
  br label %loop-memcpy-residual-header5430, !amdgpu.uniform !42

loop-memcpy-expansion5427:                        ; preds = %loop-memcpy-expansion5427.preheader, %loop-memcpy-expansion5427
  %loop-index5428 = phi i64 [ %5006, %loop-memcpy-expansion5427 ], [ 0, %loop-memcpy-expansion5427.preheader ]
  %scevgep6415 = getelementptr i8, ptr addrspace(4) %lsr.iv6413, i64 %loop-index5428, !amdgpu.uniform !42
  %5004 = load <4 x i32>, ptr addrspace(4) %scevgep6415, align 1
  %5005 = getelementptr i8, ptr %arrayidx.i22.i38.i.i.i.i3736, i64 %loop-index5428
  store <4 x i32> %5004, ptr %5005, align 1
  %5006 = add i64 %loop-index5428, 16
  %5007 = icmp uge i64 %5006, %5002
  br i1 %5007, label %Flow7821, label %loop-memcpy-expansion5427, !amdgpu.uniform !42

Flow7820:                                         ; preds = %Flow7819, %loop-memcpy-residual-header5430
  br label %Flow7823, !amdgpu.uniform !42

loop-memcpy-residual5429:                         ; preds = %loop-memcpy-residual5429.preheader, %loop-memcpy-residual5429
  %residual-loop-index5431 = phi i64 [ %5009, %loop-memcpy-residual5429 ], [ 0, %loop-memcpy-residual5429.preheader ]
  %scevgep6424 = getelementptr i8, ptr addrspace(4) %scevgep6423, i64 %residual-loop-index5431, !amdgpu.uniform !42
  %5008 = load i8, ptr addrspace(4) %scevgep6424, align 1
  %scevgep6422 = getelementptr i8, ptr %scevgep6421, i64 %residual-loop-index5431
  store i8 %5008, ptr %scevgep6422, align 1
  %5009 = add i64 %residual-loop-index5431, 1
  %5010 = icmp uge i64 %5009, %5001
  br i1 %5010, label %Flow7819, label %loop-memcpy-residual5429, !amdgpu.uniform !42

Flow7819:                                         ; preds = %loop-memcpy-residual5429
  br label %Flow7820, !amdgpu.uniform !42

_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i3727: ; preds = %Flow7823
  %conv.i.i26.i.i.i.i3728 = xor i32 %port.sroa.37.0.i.i3716, 1
  fence release
  %5011 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %arrayidx.i24.i29.i.i.i.i3729 = getelementptr inbounds nuw i32, ptr %5011, i64 %idxprom.i23.i.i.i.i.i3712
  store atomic i32 %conv.i.i26.i.i.i.i3728, ptr %arrayidx.i24.i29.i.i.i.i3729 monotonic, align 4
  %add.i.i.i.i3730 = add i64 %idx.0.i.i.i.i3717, 64
  %scevgep6414 = getelementptr i8, ptr addrspace(4) %lsr.iv6413, i64 64
  %lsr.iv.next6419 = add i64 %lsr.iv6418, -64
  br label %Flow7826, !amdgpu.uniform !42

cond.false.i.i.i3737:                             ; preds = %Flow7826
  %.lcssa8888 = phi i64 [ %4998, %Flow7826 ]
  %port.sroa.37.0.i.i3716.lcssa = phi i32 [ %port.sroa.37.0.i.i3716, %Flow7826 ]
  %.lcssa8887 = phi i32 [ %4984, %Flow7826 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa8888)
  %5012 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %.lcssa8887)
  %5013 = icmp ne i32 %5012, %port.sroa.37.0.i.i3716.lcssa
  %5014 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %5013)
  %5015 = extractvalue { i1, i64 } %5014, 0
  %5016 = extractvalue { i1, i64 } %5014, 1
  br i1 %5015, label %while.body.i.i.i.i3738.preheader, label %Flow7818

while.body.i.i.i.i3738.preheader:                 ; preds = %cond.false.i.i.i3737
  br label %while.body.i.i.i.i3738, !amdgpu.uniform !42

Flow7818:                                         ; preds = %Flow7817, %cond.false.i.i.i3737
  call void @llvm.amdgcn.end.cf.i64(i64 %5016)
  br label %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit3759, !amdgpu.uniform !42

while.body.i.i.i.i3738:                           ; preds = %while.body.i.i.i.i3738.preheader, %while.body.i.i.i.i3738
  %phi.broken8793 = phi i64 [ 0, %while.body.i.i.i.i3738.preheader ], [ %5020, %while.body.i.i.i.i3738 ]
  call void @llvm.amdgcn.s.sleep(i32 2)
  %5017 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i8.i.i3739 = getelementptr inbounds nuw i32, ptr %5017, i64 %idxprom.i23.i.i.i.i.i3712, !amdgpu.uniform !42
  %5018 = load atomic i32, ptr %arrayidx.i.i.i8.i.i3739 monotonic, align 4
  %5019 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %5018)
  %cmp.i.not.i.i.i.i3740 = icmp eq i32 %5019, %port.sroa.37.0.i.i3716.lcssa
  %5020 = call i64 @llvm.amdgcn.if.break.i64(i1 %cmp.i.not.i.i.i.i3740, i64 %phi.broken8793)
  %5021 = call i1 @llvm.amdgcn.loop.i64(i64 %5020)
  br i1 %5021, label %Flow7817, label %while.body.i.i.i.i3738

Flow7817:                                         ; preds = %while.body.i.i.i.i3738
  %.lcssa8886 = phi i64 [ %5020, %while.body.i.i.i.i3738 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa8886)
  br label %Flow7818, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit3759: ; preds = %Flow7818
  fence acquire
  call void @llvm.amdgcn.wave.barrier()
  fence syncscope("agent") release
  %5022 = call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %.lcssa8893, i1 true)
  %iszero.i.i.i.i.i13.i.i3742 = icmp ne i64 %.lcssa8893, 0
  %cmp2.i.i.i.i14.i.i3743 = icmp eq i64 %5022, %sh_prom.i.i.i.i3270
  %cmp.i.i.i.i15.i.i3744 = select i1 %iszero.i.i.i.i.i13.i.i3742, i1 %cmp2.i.i.i.i14.i.i3743, i1 false
  %5023 = and i32 %.lcssa8894, 31
  %conv4.i.i.i19.i.i3748 = zext i1 %cmp.i.i.i.i15.i.i3744 to i32
  %shl.i.i.i20.i.i3749 = shl nuw i32 %conv4.i.i.i19.i.i3748, %5023
  %xor.i.i.i21.i.i3750 = xor i32 %shl.i.i.i20.i.i3749, -1
  %5024 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop5901, !amdgpu.uniform !42

5025:                                             ; preds = %ComputeEnd5902
  %div8.i.i.i16.i.i3745 = lshr i32 %.lcssa8894, 5
  %idxprom.i.i.i17.i.i3746 = zext nneg i32 %div8.i.i.i16.i.i3745 to i64
  %5026 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i17.i.i3746
  %sunkaddr7482 = getelementptr inbounds i8, ptr addrspace(1) %5026, i64 40
  %5027 = atomicrmw and ptr addrspace(1) %sunkaddr7482, i32 %.lcssa8885 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %5028, !amdgpu.uniform !42

5028:                                             ; preds = %5025, %ComputeEnd5902
  call void @llvm.amdgcn.end.cf.i64(i64 %8524)
  call void @llvm.amdgcn.wave.barrier()
  %sub95 = sub nsw i32 %1148, %1149
  call void @llvm.lifetime.start.p5(ptr addrspace(5) %buffer.i.i.i2988) #34, !noalias !185
  %cmp7.i.i.i.i.i.i3761 = icmp ne i32 %1148, %1149
  %5029 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp7.i.i.i.i.i.i3761)
  %5030 = extractvalue { i1, i64 } %5029, 0
  %5031 = extractvalue { i1, i64 } %5029, 1
  br i1 %5030, label %while.body.lr.ph.i.i.i.i.i.i3762, label %Flow7816

while.body.lr.ph.i.i.i.i.i.i3762:                 ; preds = %5028
  %retval.0.i.i.i.i.i.i3763 = call noundef i32 @llvm.abs.i32(i32 %sub95, i1 false)
  br label %while.body.i.i.i.i.i.i3764, !amdgpu.uniform !42

Flow7816:                                         ; preds = %Flow7815, %5028
  %5032 = phi i1 [ %cmp.i.i.i.i.i.i.i.i3774.lcssa, %Flow7815 ], [ false, %5028 ]
  %5033 = phi i64 [ %writer.sroa.10.1.i.i.i.i3779.lcssa, %Flow7815 ], [ 0, %5028 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %5031)
  br label %_ZN22__llvm_libc_22_0_0_git15IntegerToStringIiNS_7details3FmtILh10ELb0ELb0ELb0ELm1EEEE13IntegerWriter25write_unsigned_number_decEjRNS1_22StringBufferWriterImplILb0EEE.exit.i.i.i.i.i3782, !amdgpu.uniform !42

while.body.i.i.i.i.i.i3764:                       ; preds = %while.body.lr.ph.i.i.i.i.i.i3762, %_ZN22__llvm_libc_22_0_0_git7details22StringBufferWriterImplILb0EE4pushEc.exit.i.i.i.i.i.i3778
  %phi.broken8794 = phi i64 [ 0, %while.body.lr.ph.i.i.i.i.i.i3762 ], [ %5101, %_ZN22__llvm_libc_22_0_0_git7details22StringBufferWriterImplILb0EE4pushEc.exit.i.i.i.i.i.i3778 ]
  %writer.sroa.10.0.i.i.i.i3765 = phi i64 [ %writer.sroa.10.1.i.i.i.i3779, %_ZN22__llvm_libc_22_0_0_git7details22StringBufferWriterImplILb0EE4pushEc.exit.i.i.i.i.i.i3778 ], [ 0, %while.body.lr.ph.i.i.i.i.i.i3762 ]
  %value.addr.09.i.i.i.i.i.i3766 = phi i32 [ %div.i.i.i.i.i.i.i3767, %_ZN22__llvm_libc_22_0_0_git7details22StringBufferWriterImplILb0EE4pushEc.exit.i.i.i.i.i.i3778 ], [ %retval.0.i.i.i.i.i.i3763, %while.body.lr.ph.i.i.i.i.i.i3762 ]
  br label %NodeBlock7613, !amdgpu.uniform !42

NodeBlock7613:                                    ; preds = %while.body.i.i.i.i.i.i3764
  %div.i.i.i.i.i.i.i3767 = udiv i32 %value.addr.09.i.i.i.i.i.i3766, 10
  %.neg.i3768 = mul i32 %div.i.i.i.i.i.i.i3767, 246
  %rem.i.i.i.i.i.i.decomposed.i3769 = add i32 %.neg.i3768, %value.addr.09.i.i.i.i.i.i3766
  %conv.i.i.i.i.i.i.i3770 = trunc i32 %rem.i.i.i.i.i.i.decomposed.i3769 to i8
  %5034 = zext i8 %conv.i.i.i.i.i.i.i3770 to i16
  %Pivot7614 = icmp sge i16 %5034, 5
  %5035 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %Pivot7614)
  %5036 = extractvalue { i1, i64 } %5035, 0
  %5037 = extractvalue { i1, i64 } %5035, 1
  br i1 %5036, label %NodeBlock7611, label %Flow7814

NodeBlock7611:                                    ; preds = %NodeBlock7613
  %Pivot7612 = icmp sge i16 %5034, 7
  %5038 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %Pivot7612)
  %5039 = extractvalue { i1, i64 } %5038, 0
  %5040 = extractvalue { i1, i64 } %5038, 1
  br i1 %5039, label %NodeBlock7609, label %Flow7805

NodeBlock7609:                                    ; preds = %NodeBlock7611
  %Pivot7610 = icmp sge i16 %5034, 8
  %5041 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %Pivot7610)
  %5042 = extractvalue { i1, i64 } %5041, 0
  %5043 = extractvalue { i1, i64 } %5041, 1
  br i1 %5042, label %NodeBlock7607, label %Flow7801

NodeBlock7607:                                    ; preds = %NodeBlock7609
  %Pivot7608 = icmp sge i16 %5034, 9
  %5044 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %Pivot7608)
  %5045 = extractvalue { i1, i64 } %5044, 0
  %5046 = extractvalue { i1, i64 } %5044, 1
  br i1 %5045, label %sw.bb9.i.i.i.i.i.i.i.i3771, label %Flow7799

Flow7805:                                         ; preds = %Flow7802, %NodeBlock7611
  %5047 = phi i8 [ %5091, %Flow7802 ], [ poison, %NodeBlock7611 ]
  %5048 = call { i1, i64 } @llvm.amdgcn.else.i64.i64(i64 %5040)
  %5049 = extractvalue { i1, i64 } %5048, 0
  %5050 = extractvalue { i1, i64 } %5048, 1
  br i1 %5049, label %NodeBlock7605, label %Flow7806

NodeBlock7605:                                    ; preds = %Flow7805
  %Pivot7606 = icmp sge i16 %5034, 6
  %5051 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %Pivot7606)
  %5052 = extractvalue { i1, i64 } %5051, 0
  %5053 = extractvalue { i1, i64 } %5051, 1
  br i1 %5052, label %sw.bb6.i.i.i.i.i.i.i.i3931, label %Flow7803

Flow7814:                                         ; preds = %Flow7806, %NodeBlock7613
  %5054 = phi i8 [ %5093, %Flow7806 ], [ poison, %NodeBlock7613 ]
  %5055 = call { i1, i64 } @llvm.amdgcn.else.i64.i64(i64 %5037)
  %5056 = extractvalue { i1, i64 } %5055, 0
  %5057 = extractvalue { i1, i64 } %5055, 1
  br i1 %5056, label %NodeBlock7603, label %if.then.i.i.i.i.i.i.i3772

NodeBlock7603:                                    ; preds = %Flow7814
  %Pivot7604 = icmp sge i16 %5034, 3
  %5058 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %Pivot7604)
  %5059 = extractvalue { i1, i64 } %5058, 0
  %5060 = extractvalue { i1, i64 } %5058, 1
  br i1 %5059, label %NodeBlock7601, label %Flow7812

NodeBlock7601:                                    ; preds = %NodeBlock7603
  %Pivot7602 = icmp sge i16 %5034, 4
  %5061 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %Pivot7602)
  %5062 = extractvalue { i1, i64 } %5061, 0
  %5063 = extractvalue { i1, i64 } %5061, 1
  br i1 %5062, label %sw.bb4.i.i.i.i.i.i.i.i3933, label %Flow7807

Flow7812:                                         ; preds = %Flow7808, %NodeBlock7603
  %5064 = phi i8 [ %5094, %Flow7808 ], [ poison, %NodeBlock7603 ]
  %5065 = call { i1, i64 } @llvm.amdgcn.else.i64.i64(i64 %5060)
  %5066 = extractvalue { i1, i64 } %5065, 0
  %5067 = extractvalue { i1, i64 } %5065, 1
  br i1 %5066, label %NodeBlock7599, label %Flow7813

NodeBlock7599:                                    ; preds = %Flow7812
  %Pivot7600 = icmp sge i16 %5034, 2
  %5068 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %Pivot7600)
  %5069 = extractvalue { i1, i64 } %5068, 0
  %5070 = extractvalue { i1, i64 } %5068, 1
  br i1 %5069, label %sw.bb2.i.i.i.i.i.i.i.i3935, label %Flow7810

Flow7810:                                         ; preds = %sw.bb2.i.i.i.i.i.i.i.i3935, %NodeBlock7599
  %5071 = call { i1, i64 } @llvm.amdgcn.else.i64.i64(i64 %5070)
  %5072 = extractvalue { i1, i64 } %5071, 0
  %5073 = extractvalue { i1, i64 } %5071, 1
  br i1 %5072, label %LeafBlock7597, label %Flow7811

LeafBlock7597:                                    ; preds = %Flow7810
  %SwitchLeaf7598 = icmp eq i16 %5034, 1
  %5074 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %SwitchLeaf7598)
  %5075 = extractvalue { i1, i64 } %5074, 0
  %5076 = extractvalue { i1, i64 } %5074, 1
  br i1 %5075, label %sw.bb1.i.i.i.i.i.i.i.i3936, label %Flow7809

sw.bb1.i.i.i.i.i.i.i.i3936:                       ; preds = %LeafBlock7597
  br label %Flow7809, !amdgpu.uniform !42

sw.bb2.i.i.i.i.i.i.i.i3935:                       ; preds = %NodeBlock7599
  br label %Flow7810, !amdgpu.uniform !42

Flow7807:                                         ; preds = %sw.bb4.i.i.i.i.i.i.i.i3933, %NodeBlock7601
  %5077 = call { i1, i64 } @llvm.amdgcn.else.i64.i64(i64 %5063)
  %5078 = extractvalue { i1, i64 } %5077, 0
  %5079 = extractvalue { i1, i64 } %5077, 1
  br i1 %5078, label %sw.bb3.i.i.i.i.i.i.i.i3934, label %Flow7808

sw.bb3.i.i.i.i.i.i.i.i3934:                       ; preds = %Flow7807
  br label %Flow7808, !amdgpu.uniform !42

sw.bb4.i.i.i.i.i.i.i.i3933:                       ; preds = %NodeBlock7601
  br label %Flow7807, !amdgpu.uniform !42

Flow7803:                                         ; preds = %sw.bb6.i.i.i.i.i.i.i.i3931, %NodeBlock7605
  %5080 = call { i1, i64 } @llvm.amdgcn.else.i64.i64(i64 %5053)
  %5081 = extractvalue { i1, i64 } %5080, 0
  %5082 = extractvalue { i1, i64 } %5080, 1
  br i1 %5081, label %sw.bb5.i.i.i.i.i.i.i.i3932, label %Flow7804

sw.bb5.i.i.i.i.i.i.i.i3932:                       ; preds = %Flow7803
  br label %Flow7804, !amdgpu.uniform !42

sw.bb6.i.i.i.i.i.i.i.i3931:                       ; preds = %NodeBlock7605
  br label %Flow7803, !amdgpu.uniform !42

Flow7801:                                         ; preds = %Flow7800, %NodeBlock7609
  %5083 = phi i8 [ %5090, %Flow7800 ], [ poison, %NodeBlock7609 ]
  %5084 = call { i1, i64 } @llvm.amdgcn.else.i64.i64(i64 %5043)
  %5085 = extractvalue { i1, i64 } %5084, 0
  %5086 = extractvalue { i1, i64 } %5084, 1
  br i1 %5085, label %sw.bb7.i.i.i.i.i.i.i.i3930, label %Flow7802

sw.bb7.i.i.i.i.i.i.i.i3930:                       ; preds = %Flow7801
  br label %Flow7802, !amdgpu.uniform !42

Flow7799:                                         ; preds = %sw.bb9.i.i.i.i.i.i.i.i3771, %NodeBlock7607
  %5087 = call { i1, i64 } @llvm.amdgcn.else.i64.i64(i64 %5046)
  %5088 = extractvalue { i1, i64 } %5087, 0
  %5089 = extractvalue { i1, i64 } %5087, 1
  br i1 %5088, label %sw.bb8.i.i.i.i.i.i.i.i3929, label %Flow7800

sw.bb8.i.i.i.i.i.i.i.i3929:                       ; preds = %Flow7799
  br label %Flow7800, !amdgpu.uniform !42

sw.bb9.i.i.i.i.i.i.i.i3771:                       ; preds = %NodeBlock7607
  br label %Flow7799, !amdgpu.uniform !42

Flow7800:                                         ; preds = %sw.bb8.i.i.i.i.i.i.i.i3929, %Flow7799
  %5090 = phi i8 [ 56, %sw.bb8.i.i.i.i.i.i.i.i3929 ], [ 57, %Flow7799 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %5089)
  br label %Flow7801, !amdgpu.uniform !42

Flow7802:                                         ; preds = %sw.bb7.i.i.i.i.i.i.i.i3930, %Flow7801
  %5091 = phi i8 [ 55, %sw.bb7.i.i.i.i.i.i.i.i3930 ], [ %5083, %Flow7801 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %5086)
  br label %Flow7805, !amdgpu.uniform !42

Flow7804:                                         ; preds = %sw.bb5.i.i.i.i.i.i.i.i3932, %Flow7803
  %5092 = phi i8 [ 53, %sw.bb5.i.i.i.i.i.i.i.i3932 ], [ 54, %Flow7803 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %5082)
  br label %Flow7806, !amdgpu.uniform !42

Flow7806:                                         ; preds = %Flow7804, %Flow7805
  %5093 = phi i8 [ %5092, %Flow7804 ], [ %5047, %Flow7805 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %5050)
  br label %Flow7814, !amdgpu.uniform !42

Flow7808:                                         ; preds = %sw.bb3.i.i.i.i.i.i.i.i3934, %Flow7807
  %5094 = phi i8 [ 51, %sw.bb3.i.i.i.i.i.i.i.i3934 ], [ 52, %Flow7807 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %5079)
  br label %Flow7812, !amdgpu.uniform !42

Flow7809:                                         ; preds = %sw.bb1.i.i.i.i.i.i.i.i3936, %LeafBlock7597
  %5095 = phi i8 [ 49, %sw.bb1.i.i.i.i.i.i.i.i3936 ], [ 48, %LeafBlock7597 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %5076)
  br label %Flow7811, !amdgpu.uniform !42

Flow7811:                                         ; preds = %Flow7809, %Flow7810
  %5096 = phi i8 [ %5095, %Flow7809 ], [ 50, %Flow7810 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %5073)
  br label %Flow7813, !amdgpu.uniform !42

Flow7813:                                         ; preds = %Flow7811, %Flow7812
  %5097 = phi i8 [ %5096, %Flow7811 ], [ %5064, %Flow7812 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %5067)
  br label %if.then.i.i.i.i.i.i.i3772, !amdgpu.uniform !42

if.then.i.i.i.i.i.i.i3772:                        ; preds = %Flow7814, %Flow7813
  %retval.0.i.i.i.i.i.i.i.i3773 = phi i8 [ %5054, %Flow7814 ], [ %5097, %Flow7813 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %5057)
  %cmp.i.i.i.i.i.i.i.i3774 = icmp eq i64 %writer.sroa.10.0.i.i.i.i3765, 11
  %cmp.i.i.i.i.i.i.i.i3774.inv = xor i1 %cmp.i.i.i.i.i.i.i.i3774, true
  br i1 %cmp.i.i.i.i.i.i.i.i3774.inv, label %if.then3.i.i.i.i.i.i.i3775, label %_ZN22__llvm_libc_22_0_0_git7details22StringBufferWriterImplILb0EE4pushEc.exit.i.i.i.i.i.i3778, !amdgpu.uniform !42

if.then3.i.i.i.i.i.i.i3775:                       ; preds = %if.then.i.i.i.i.i.i.i3772
  %5098 = trunc i64 %writer.sroa.10.0.i.i.i.i3765 to i32
  %5099 = xor i32 %5098, -1
  %5100 = getelementptr i8, ptr addrspace(5) %buffer.i.i.i2988, i32 %5099
  %arrayidx.i.i.i.i.i.i.i.i37766103 = getelementptr i8, ptr addrspace(5) %5100, i32 11
  %inc.i.i.i.i.i.i.i3777 = add i64 %writer.sroa.10.0.i.i.i.i3765, 1
  store i8 %retval.0.i.i.i.i.i.i.i.i3773, ptr addrspace(5) %arrayidx.i.i.i.i.i.i.i.i37766103, align 1, !tbaa !57, !noalias !185
  br label %_ZN22__llvm_libc_22_0_0_git7details22StringBufferWriterImplILb0EE4pushEc.exit.i.i.i.i.i.i3778, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git7details22StringBufferWriterImplILb0EE4pushEc.exit.i.i.i.i.i.i3778: ; preds = %if.then3.i.i.i.i.i.i.i3775, %if.then.i.i.i.i.i.i.i3772
  %writer.sroa.10.1.i.i.i.i3779 = phi i64 [ %inc.i.i.i.i.i.i.i3777, %if.then3.i.i.i.i.i.i.i3775 ], [ 11, %if.then.i.i.i.i.i.i.i3772 ]
  %cmp.i.i.i.i.i.i3780 = icmp ult i32 %value.addr.09.i.i.i.i.i.i3766, 10
  %.not.i.i.i.i.i.i3781 = or i1 %cmp.i.i.i.i.i.i.i.i3774, %cmp.i.i.i.i.i.i3780
  %5101 = call i64 @llvm.amdgcn.if.break.i64(i1 %.not.i.i.i.i.i.i3781, i64 %phi.broken8794)
  %5102 = call i1 @llvm.amdgcn.loop.i64(i64 %5101)
  br i1 %5102, label %Flow7815, label %while.body.i.i.i.i.i.i3764

Flow7815:                                         ; preds = %_ZN22__llvm_libc_22_0_0_git7details22StringBufferWriterImplILb0EE4pushEc.exit.i.i.i.i.i.i3778
  %writer.sroa.10.1.i.i.i.i3779.lcssa = phi i64 [ %writer.sroa.10.1.i.i.i.i3779, %_ZN22__llvm_libc_22_0_0_git7details22StringBufferWriterImplILb0EE4pushEc.exit.i.i.i.i.i.i3778 ]
  %.lcssa8884 = phi i64 [ %5101, %_ZN22__llvm_libc_22_0_0_git7details22StringBufferWriterImplILb0EE4pushEc.exit.i.i.i.i.i.i3778 ]
  %cmp.i.i.i.i.i.i.i.i3774.lcssa = phi i1 [ %cmp.i.i.i.i.i.i.i.i3774, %_ZN22__llvm_libc_22_0_0_git7details22StringBufferWriterImplILb0EE4pushEc.exit.i.i.i.i.i.i3778 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa8884)
  br label %Flow7816, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git15IntegerToStringIiNS_7details3FmtILh10ELb0ELb0ELb0ELm1EEEE13IntegerWriter25write_unsigned_number_decEjRNS1_22StringBufferWriterImplILb0EEE.exit.i.i.i.i.i3782: ; preds = %Flow7816
  %cmp.i19.i.i.i.i3785 = icmp ne i64 %5033, 0
  %or.cond.not.i.i.i.i3786 = select i1 %5032, i1 true, i1 %cmp.i19.i.i.i.i3785
  %or.cond.not.i.i.i.i3786.inv = xor i1 %or.cond.not.i.i.i.i3786, true
  %5103 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %or.cond.not.i.i.i.i3786.inv)
  %5104 = extractvalue { i1, i64 } %5103, 0
  %5105 = extractvalue { i1, i64 } %5103, 1
  br i1 %5104, label %if.then3.i.i.lr.ph.i.i.i.i3787, label %while.end.i.i.i.i.i3789

if.then3.i.i.lr.ph.i.i.i.i3787:                   ; preds = %_ZN22__llvm_libc_22_0_0_git15IntegerToStringIiNS_7details3FmtILh10ELb0ELb0ELb0ELm1EEEE13IntegerWriter25write_unsigned_number_decEjRNS1_22StringBufferWriterImplILb0EEE.exit.i.i.i.i.i3782
  %arrayidx.i.i.i.i.i.i.i3788 = getelementptr inbounds nuw i8, ptr addrspace(5) %buffer.i.i.i2988, i32 10
  store i8 48, ptr addrspace(5) %arrayidx.i.i.i.i.i.i.i3788, align 2, !tbaa !57, !noalias !185
  br label %while.end.i.i.i.i.i3789, !amdgpu.uniform !42

while.end.i.i.i.i.i3789:                          ; preds = %if.then3.i.i.lr.ph.i.i.i.i3787, %_ZN22__llvm_libc_22_0_0_git15IntegerToStringIiNS_7details3FmtILh10ELb0ELb0ELb0ELm1EEEE13IntegerWriter25write_unsigned_number_decEjRNS1_22StringBufferWriterImplILb0EEE.exit.i.i.i.i.i3782
  %writer.sroa.10.4.i.i.i.i3790 = phi i64 [ 1, %if.then3.i.i.lr.ph.i.i.i.i3787 ], [ %5033, %_ZN22__llvm_libc_22_0_0_git15IntegerToStringIiNS_7details3FmtILh10ELb0ELb0ELb0ELm1EEEE13IntegerWriter25write_unsigned_number_decEjRNS1_22StringBufferWriterImplILb0EEE.exit.i.i.i.i.i3782 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %5105)
  %cmp4.i.i.i.i.i3791 = icmp sgt i32 %sub95, -1
  %brmerge.i.i.i.i.i3792 = or i1 %cmp4.i.i.i.i.i3791, %5032
  %brmerge.i.i.i.i.i3792.inv = xor i1 %brmerge.i.i.i.i.i3792, true
  %5106 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %brmerge.i.i.i.i.i3792.inv)
  %5107 = extractvalue { i1, i64 } %5106, 0
  %5108 = extractvalue { i1, i64 } %5106, 1
  br i1 %5107, label %if.then.i14.i.i.i.i.i3793, label %_ZN22__llvm_libc_22_0_0_git15IntegerToStringIiNS_7details3FmtILh10ELb0ELb0ELb0ELm1EEEEC2Ei.exit.i.i.i3798

if.then.i14.i.i.i.i.i3793:                        ; preds = %while.end.i.i.i.i.i3789
  %cmp.i.i17.i.i.i.i.i3794 = icmp ne i64 %writer.sroa.10.4.i.i.i.i3790, 11
  %5109 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.i.i17.i.i.i.i.i3794)
  %5110 = extractvalue { i1, i64 } %5109, 0
  %5111 = extractvalue { i1, i64 } %5109, 1
  br i1 %5110, label %if.then3.i18.i.i.i.i.i3795, label %Flow7798

if.then3.i18.i.i.i.i.i3795:                       ; preds = %if.then.i14.i.i.i.i.i3793
  %5112 = trunc i64 %writer.sroa.10.4.i.i.i.i3790 to i32
  %5113 = xor i32 %5112, -1
  %5114 = getelementptr i8, ptr addrspace(5) %buffer.i.i.i2988, i32 %5113
  %arrayidx.i.i19.i.i.i.i.i37966104 = getelementptr i8, ptr addrspace(5) %5114, i32 11
  %inc.i20.i.i.i.i.i3797 = add i64 %writer.sroa.10.4.i.i.i.i3790, 1
  store i8 45, ptr addrspace(5) %arrayidx.i.i19.i.i.i.i.i37966104, align 1, !tbaa !57, !noalias !185
  br label %Flow7798, !amdgpu.uniform !42

Flow7798:                                         ; preds = %if.then3.i18.i.i.i.i.i3795, %if.then.i14.i.i.i.i.i3793
  %5115 = phi i64 [ %inc.i20.i.i.i.i.i3797, %if.then3.i18.i.i.i.i.i3795 ], [ 11, %if.then.i14.i.i.i.i.i3793 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %5111)
  br label %_ZN22__llvm_libc_22_0_0_git15IntegerToStringIiNS_7details3FmtILh10ELb0ELb0ELb0ELm1EEEEC2Ei.exit.i.i.i3798, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git15IntegerToStringIiNS_7details3FmtILh10ELb0ELb0ELb0ELm1EEEEC2Ei.exit.i.i.i3798: ; preds = %while.end.i.i.i.i.i3789, %Flow7798
  %writer.sroa.10.5.i.i.i.i3799 = phi i64 [ %writer.sroa.10.4.i.i.i.i3790, %while.end.i.i.i.i.i3789 ], [ %5115, %Flow7798 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %5108)
  %written.i.i.i.i3800 = getelementptr inbounds nuw i8, ptr addrspace(5) %buffer.i.i.i2988, i32 16
  store i64 %writer.sroa.10.5.i.i.i.i3799, ptr addrspace(5) %written.i.i.i.i3800, align 8, !tbaa !91, !noalias !185
  %cmp.i.not.i.i.i.i.i3801 = icmp ne i64 %writer.sroa.10.5.i.i.i.i3799, 0
  %5116 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.i.not.i.i.i.i.i3801)
  %5117 = extractvalue { i1, i64 } %5116, 0
  %5118 = extractvalue { i1, i64 } %5116, 1
  br i1 %5117, label %if.then.i.i.i.i.i.i3802, label %Flow7797

if.then.i.i.i.i.i.i3802:                          ; preds = %_ZN22__llvm_libc_22_0_0_git15IntegerToStringIiNS_7details3FmtILh10ELb0ELb0ELb0ELm1EEEEC2Ei.exit.i.i.i3798
  %5119 = call { i64, i1 } @llvm.uadd.with.overflow.i64(i64 %writer.sroa.10.5.i.i.i.i3799, i64 1)
  %ov7484 = extractvalue { i64, i1 } %5119, 1
  %ov7484.inv = xor i1 %ov7484, true
  %5120 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %ov7484.inv)
  %5121 = extractvalue { i1, i64 } %5120, 0
  %5122 = extractvalue { i1, i64 } %5120, 1
  br i1 %5121, label %if.end.i.i.i.i.i.i.i3805, label %Flow7796

if.end.i.i.i.i.i.i.i3805:                         ; preds = %if.then.i.i.i.i.i.i3802
  %math7483 = extractvalue { i64, i1 } %5119, 0
  %cmp2.i.i.i.i.i.i.i3806 = icmp ult i64 %math7483, 1676976733973595601
  %5123 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp2.i.i.i.i.i.i.i3806)
  %5124 = extractvalue { i1, i64 } %5123, 0
  %5125 = extractvalue { i1, i64 } %5123, 1
  br i1 %5124, label %if.then3.i.i.i.i4.i.i.i3926, label %if.end4.i.i.i.i.i.i.i3807

if.then3.i.i.i.i4.i.i.i3926:                      ; preds = %if.end.i.i.i.i.i.i.i3805
  %mul.i.i.i.i.i.i.i3927 = mul nuw i64 %math7483, 11
  %div18.i.i.i.i.i.i.i3928 = lshr i64 %mul.i.i.i.i.i.i.i3927, 3
  br label %if.end4.i.i.i.i.i.i.i3807, !amdgpu.uniform !42

Flow7796:                                         ; preds = %if.end4.i.i.i.i.i.i.i3807, %if.then.i.i.i.i.i.i3802
  %5126 = phi ptr [ %cond.i.i.i3815, %if.end4.i.i.i.i.i.i.i3807 ], [ addrspacecast (ptr addrspace(4) @_ZN22__llvm_libc_22_0_0_git3cpp6string14NULL_CHARACTERE to ptr), %if.then.i.i.i.i.i.i3802 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %5122)
  br label %_ZN22__llvm_libc_22_0_0_git3cpp6string7reserveEm.exit.i.i.i.i.i.i3817, !amdgpu.uniform !42

if.end4.i.i.i.i.i.i.i3807:                        ; preds = %if.then3.i.i.i.i4.i.i.i3926, %if.end.i.i.i.i.i.i.i3805
  %new_capacity.addr.0.i.i.i.i.i.i.i3808 = phi i64 [ %div18.i.i.i.i.i.i.i3928, %if.then3.i.i.i.i4.i.i.i3926 ], [ %math7483, %if.end.i.i.i.i.i.i.i3805 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %5125)
  %sub.i.i.i3809 = add i64 %new_capacity.addr.0.i.i.i.i.i.i.i3808, 7
  %div2.i.i.i3810 = and i64 %sub.i.i.i3809, -8
  %5127 = load ptr, ptr addrspace(1) @_ZN12_GLOBAL__N_13ptrE, align 8, !tbaa !55
  %add.ptr.i.i.i3811 = getelementptr inbounds nuw i8, ptr %5127, i64 %div2.i.i.i3810
  store ptr %add.ptr.i.i.i3811, ptr addrspace(1) @_ZN12_GLOBAL__N_13ptrE, align 8, !tbaa !55
  %sub.ptr.lhs.cast.i.i.i3812 = ptrtoint ptr %add.ptr.i.i.i3811 to i64
  %sub.ptr.sub.i.i.i3813 = sub i64 %sub.ptr.lhs.cast.i.i.i3812, ptrtoint (ptr addrspacecast (ptr addrspace(1) @_ZN12_GLOBAL__N_16memoryE to ptr) to i64)
  %cmp.i.i.i3814 = icmp ugt i64 %sub.ptr.sub.i.i.i3813, 65335
  %cond.i.i.i3815 = select i1 %cmp.i.i.i3814, ptr null, ptr %5127
  br label %Flow7796, !amdgpu.uniform !42

Flow7797:                                         ; preds = %Flow7795, %_ZN22__llvm_libc_22_0_0_git15IntegerToStringIiNS_7details3FmtILh10ELb0ELb0ELb0ELm1EEEEC2Ei.exit.i.i.i3798
  %5128 = phi ptr [ %5126, %Flow7795 ], [ addrspacecast (ptr addrspace(4) @_ZN22__llvm_libc_22_0_0_git3cpp6string14NULL_CHARACTERE to ptr), %_ZN22__llvm_libc_22_0_0_git15IntegerToStringIiNS_7details3FmtILh10ELb0ELb0ELb0ELm1EEEEC2Ei.exit.i.i.i3798 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %5118)
  br label %if.end.i.i.i.i.i.i3819, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git3cpp6string7reserveEm.exit.i.i.i.i.i.i3817: ; preds = %Flow7796
  br label %loadstoreloop4949, !amdgpu.uniform !42

loadstoreloop4949:                                ; preds = %_ZN22__llvm_libc_22_0_0_git3cpp6string7reserveEm.exit.i.i.i.i.i.i3817, %loadstoreloop4949
  %phi.broken8795 = phi i64 [ 0, %_ZN22__llvm_libc_22_0_0_git3cpp6string7reserveEm.exit.i.i.i.i.i.i3817 ], [ %5133, %loadstoreloop4949 ]
  %5129 = phi i64 [ %5131, %loadstoreloop4949 ], [ 0, %_ZN22__llvm_libc_22_0_0_git3cpp6string7reserveEm.exit.i.i.i.i.i.i3817 ]
  %5130 = getelementptr i8, ptr %5126, i64 %5129
  store i8 0, ptr %5130, align 1
  %5131 = add i64 %5129, 1
  %5132 = icmp uge i64 %5131, %writer.sroa.10.5.i.i.i.i3799
  %5133 = call i64 @llvm.amdgcn.if.break.i64(i1 %5132, i64 %phi.broken8795)
  %5134 = call i1 @llvm.amdgcn.loop.i64(i64 %5133)
  br i1 %5134, label %Flow7795, label %loadstoreloop4949

Flow7795:                                         ; preds = %loadstoreloop4949
  %.lcssa8883 = phi i64 [ %5133, %loadstoreloop4949 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa8883)
  br label %Flow7797, !amdgpu.uniform !42

if.end.i.i.i.i.i.i3819:                           ; preds = %Flow7797
  %cmp.not.i6.i.i.i.i.i.i3821 = icmp ne ptr %5128, addrspacecast (ptr addrspace(4) @_ZN22__llvm_libc_22_0_0_git3cpp6string14NULL_CHARACTERE to ptr)
  %5135 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.not.i6.i.i.i.i.i.i3821)
  %5136 = extractvalue { i1, i64 } %5135, 0
  %5137 = extractvalue { i1, i64 } %5135, 1
  br i1 %5136, label %if.then.i.i.i.i2.i.i.i3822, label %_ZN22__llvm_libc_22_0_0_git3cpp9to_stringEi.exit.i3824

if.then.i.i.i.i2.i.i.i3822:                       ; preds = %if.end.i.i.i.i.i.i3819
  %arrayidx.i.i.i.i3.i.i.i3823 = getelementptr inbounds nuw i8, ptr %5128, i64 %writer.sroa.10.5.i.i.i.i3799
  store i8 0, ptr %arrayidx.i.i.i.i3.i.i.i3823, align 1, !tbaa !57, !noalias !185
  br label %_ZN22__llvm_libc_22_0_0_git3cpp9to_stringEi.exit.i3824, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git3cpp9to_stringEi.exit.i3824: ; preds = %if.then.i.i.i.i2.i.i.i3822, %if.end.i.i.i.i.i.i3819
  call void @llvm.amdgcn.end.cf.i64(i64 %5137)
  %5138 = trunc i64 %writer.sroa.10.5.i.i.i.i3799 to i32
  %5139 = and i64 %writer.sroa.10.5.i.i.i.i3799, 15
  %5140 = sub i64 %writer.sroa.10.5.i.i.i.i3799, %5139
  %5141 = icmp ne i64 %5140, 0
  %5142 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %5141)
  %5143 = extractvalue { i1, i64 } %5142, 0
  %5144 = extractvalue { i1, i64 } %5142, 1
  br i1 %5143, label %loop-memcpy-expansion5109.preheader, label %Flow7794

loop-memcpy-expansion5109.preheader:              ; preds = %_ZN22__llvm_libc_22_0_0_git3cpp9to_stringEi.exit.i3824
  %5145 = sub i32 11, %5138
  %scevgep6410 = getelementptr i8, ptr addrspace(5) %buffer.i.i.i2988, i32 %5145
  br label %loop-memcpy-expansion5109, !amdgpu.uniform !42

Flow7794:                                         ; preds = %Flow7793, %_ZN22__llvm_libc_22_0_0_git3cpp9to_stringEi.exit.i3824
  call void @llvm.amdgcn.end.cf.i64(i64 %5144)
  br label %loop-memcpy-residual-header5112, !amdgpu.uniform !42

loop-memcpy-expansion5109:                        ; preds = %loop-memcpy-expansion5109.preheader, %loop-memcpy-expansion5109
  %phi.broken8796 = phi i64 [ 0, %loop-memcpy-expansion5109.preheader ], [ %5150, %loop-memcpy-expansion5109 ]
  %lsr.iv6411 = phi ptr addrspace(5) [ %scevgep6412, %loop-memcpy-expansion5109 ], [ %scevgep6410, %loop-memcpy-expansion5109.preheader ]
  %loop-index5110 = phi i64 [ %5148, %loop-memcpy-expansion5109 ], [ 0, %loop-memcpy-expansion5109.preheader ]
  %5146 = load <4 x i32>, ptr addrspace(5) %lsr.iv6411, align 1
  %5147 = getelementptr i8, ptr %5128, i64 %loop-index5110
  store <4 x i32> %5146, ptr %5147, align 1
  %5148 = add i64 %loop-index5110, 16
  %scevgep6412 = getelementptr i8, ptr addrspace(5) %lsr.iv6411, i32 16
  %5149 = icmp uge i64 %5148, %5140
  %5150 = call i64 @llvm.amdgcn.if.break.i64(i1 %5149, i64 %phi.broken8796)
  %5151 = call i1 @llvm.amdgcn.loop.i64(i64 %5150)
  br i1 %5151, label %Flow7793, label %loop-memcpy-expansion5109

Flow7792:                                         ; preds = %Flow7791, %loop-memcpy-residual-header5112
  call void @llvm.amdgcn.end.cf.i64(i64 %6373)
  br label %post-loop-memcpy-expansion5108, !amdgpu.uniform !42

loop-memcpy-residual5111:                         ; preds = %loop-memcpy-residual5111.preheader, %loop-memcpy-residual5111
  %phi.broken8797 = phi i64 [ 0, %loop-memcpy-residual5111.preheader ], [ %5155, %loop-memcpy-residual5111 ]
  %lsr.iv6406 = phi ptr addrspace(5) [ %scevgep6407, %loop-memcpy-residual5111 ], [ %scevgep6405, %loop-memcpy-residual5111.preheader ]
  %residual-loop-index5113 = phi i64 [ %5153, %loop-memcpy-residual5111 ], [ 0, %loop-memcpy-residual5111.preheader ]
  %5152 = load i8, ptr addrspace(5) %lsr.iv6406, align 1
  %scevgep6409 = getelementptr i8, ptr %scevgep6408, i64 %residual-loop-index5113
  store i8 %5152, ptr %scevgep6409, align 1
  %5153 = add i64 %residual-loop-index5113, 1
  %scevgep6407 = getelementptr i8, ptr addrspace(5) %lsr.iv6406, i32 1
  %5154 = icmp uge i64 %5153, %5139
  %5155 = call i64 @llvm.amdgcn.if.break.i64(i1 %5154, i64 %phi.broken8797)
  %5156 = call i1 @llvm.amdgcn.loop.i64(i64 %5155)
  br i1 %5156, label %Flow7791, label %loop-memcpy-residual5111

Flow7791:                                         ; preds = %loop-memcpy-residual5111
  %.lcssa8881 = phi i64 [ %5155, %loop-memcpy-residual5111 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa8881)
  br label %Flow7792, !amdgpu.uniform !42

post-loop-memcpy-expansion5108:                   ; preds = %Flow7792
  call void @llvm.lifetime.end.p5(ptr addrspace(5) %buffer.i.i.i2988) #34, !noalias !185
  br label %for.cond.i.i.i3828, !amdgpu.uniform !42

for.cond.i.i.i3828:                               ; preds = %post-loop-memcpy-expansion5108, %Flow7790
  %5157 = phi i32 [ %5194, %Flow7790 ], [ poison, %post-loop-memcpy-expansion5108 ]
  %5158 = phi i64 [ %5195, %Flow7790 ], [ poison, %post-loop-memcpy-expansion5108 ]
  %5159 = phi i32 [ %5196, %Flow7790 ], [ poison, %post-loop-memcpy-expansion5108 ]
  %retval.sroa.6.0.i.i.i3829 = phi i32 [ %5199, %Flow7790 ], [ undef, %post-loop-memcpy-expansion5108 ]
  %retval.sroa.2.0.i.i.i3831 = phi i64 [ %5198, %Flow7790 ], [ undef, %post-loop-memcpy-expansion5108 ]
  %retval.sroa.8.0.i.i.i3832 = phi i32 [ %5197, %Flow7790 ], [ undef, %post-loop-memcpy-expansion5108 ]
  %index.0.i.i.i3833 = phi i32 [ %5201, %Flow7790 ], [ 0, %post-loop-memcpy-expansion5108 ]
  %5160 = load i32, ptr addrspace(1) @__llvm_rpc_client, align 8, !tbaa !68
  %cmp.not.i.i.i3834 = icmp ult i32 %index.0.i.i.i3833, %5160
  %spec.store.select.i.i.i3835 = select i1 %cmp.not.i.i.i3834, i32 %index.0.i.i.i3833, i32 0
  %5161 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  %5162 = and i32 %spec.store.select.i.i.i3835, 31
  %5163 = lshr i64 %5161, %sh_prom.i.i.i.i3270
  %5164 = trunc i64 %5163 to i32
  %conv4.i.i.i.i.i3839 = and i32 %5164, 1
  %shl.i.i.i.i.i3840 = shl nuw i32 %conv4.i.i.i.i.i3839, %5162
  br label %ComputeLoop5905, !amdgpu.uniform !42

5165:                                             ; preds = %ComputeEnd5906
  %sunkaddr7485 = getelementptr inbounds i8, ptr addrspace(1) %8534, i64 40
  %5166 = atomicrmw or ptr addrspace(1) %sunkaddr7485, i32 %.lcssa8875 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %5167, !amdgpu.uniform !42

5167:                                             ; preds = %5165, %ComputeEnd5906
  %5168 = phi i32 [ %5166, %5165 ], [ poison, %ComputeEnd5906 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %8543)
  %5169 = call i32 @llvm.amdgcn.readfirstlane.i32(i32 %5168)
  %5170 = or i32 %5169, %.lcssa8876
  %shl5.i.i.i.i.i3841 = shl nuw i32 1, %5162
  %and.i.i.i.i.i3842 = and i32 %shl5.i.i.i.i.i3841, %5170
  %tobool3.i.i.i.i3843 = icmp ne i32 %and.i.i.i.i.i3842, 0
  %5171 = call i64 @llvm.amdgcn.ballot.i64(i1 %tobool3.i.i.i.i3843)
  %and.i.i.i.i.i.i3844 = and i64 %5171, %5161
  %cmp.i.not.i.i.i3845 = icmp ne i64 %5161, %and.i.i.i.i.i.i3844
  br i1 %cmp.i.not.i.i.i3845, label %if.then.i.i.i.i3846, label %Flow7789, !amdgpu.uniform !42

if.then.i.i.i.i3846:                              ; preds = %5167
  fence syncscope("agent") acquire
  %5172 = load <2 x i64>, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !75
  %5173 = extractelement <2 x i64> %5172, i32 0
  %5174 = inttoptr i64 %5173 to ptr
  %5175 = extractelement <2 x i64> %5172, i32 1
  %5176 = inttoptr i64 %5175 to ptr
  %idxprom.i.i.i.i3847 = zext i32 %spec.store.select.i.i.i3835 to i64
  %arrayidx.i.i.i.i3848 = getelementptr inbounds nuw i32, ptr %5174, i64 %idxprom.i.i.i.i3847, !amdgpu.uniform !42
  %5177 = load atomic i32, ptr %arrayidx.i.i.i.i3848 monotonic, align 4
  %5178 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %5177)
  %arrayidx.i45.i.i.i3849 = getelementptr inbounds nuw i32, ptr %5176, i64 %idxprom.i.i.i.i3847, !amdgpu.uniform !42
  %5179 = load atomic i32, ptr %arrayidx.i45.i.i.i3849 monotonic, align 4
  %5180 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %5179)
  %cmp.i46.not.i.i.i3850 = icmp ne i32 %5178, %5180
  br i1 %cmp.i46.not.i.i.i3850, label %if.then12.i.i.i3851, label %Flow7787, !amdgpu.uniform !42

if.then12.i.i.i3851:                              ; preds = %if.then.i.i.i.i3846
  fence syncscope("agent") release
  %5181 = call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %5161, i1 true)
  %iszero.i.i.i.i.i.i.i3852 = icmp ne i64 %5161, 0
  %cmp2.i.i.i.i.i.i3853 = icmp eq i64 %5181, %sh_prom.i.i.i.i3270
  %cmp.i.i.i.i.i2.i3854 = select i1 %iszero.i.i.i.i.i.i.i3852, i1 %cmp2.i.i.i.i.i.i3853, i1 false
  %conv4.i.i51.i.i.i3855 = zext i1 %cmp.i.i.i.i.i2.i3854 to i32
  %shl.i.i52.i.i.i3856 = shl nuw i32 %conv4.i.i51.i.i.i3855, %5162
  %xor.i.i.i.i.i3857 = xor i32 %shl.i.i52.i.i.i3856, -1
  %5182 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop5910, !amdgpu.uniform !42

5183:                                             ; preds = %ComputeEnd5911
  %sunkaddr7486 = getelementptr inbounds i8, ptr addrspace(1) %8534, i64 40
  %5184 = atomicrmw and ptr addrspace(1) %sunkaddr7486, i32 %.lcssa8877 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %5185, !amdgpu.uniform !42

5185:                                             ; preds = %5183, %ComputeEnd5911
  call void @llvm.amdgcn.end.cf.i64(i64 %8561)
  call void @llvm.amdgcn.wave.barrier()
  br label %Flow7787, !amdgpu.uniform !42

if.end14.i.i.i3918:                               ; preds = %Flow7787
  %5186 = call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %5161, i1 true)
  %iszero.i.i.i.i.i.i3919 = icmp ne i64 %5161, 0
  %cmp2.i.i.i.i.i3920 = icmp eq i64 %5186, %sh_prom.i.i.i.i3270
  %cmp.i.i.i.i.i3921 = select i1 %iszero.i.i.i.i.i.i3919, i1 %cmp2.i.i.i.i.i3920, i1 false
  %5187 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.i.i.i.i.i3921)
  %5188 = extractvalue { i1, i64 } %5187, 0
  %5189 = extractvalue { i1, i64 } %5187, 1
  br i1 %5188, label %if.then16.i.i.i3923, label %if.end22.i.i.i3922

if.then16.i.i.i3923:                              ; preds = %if.end14.i.i.i3918
  %5190 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !76
  %opcode.i.i.i3924 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %5190, i64 %idxprom.i.i.i.i3847, i32 1
  store i32 1660944387, ptr %opcode.i.i.i3924, align 8, !tbaa !77
  %arrayidx21.i.i.i3925 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %5190, i64 %idxprom.i.i.i.i3847
  store i64 %5161, ptr %arrayidx21.i.i.i3925, align 8, !tbaa !80
  br label %if.end22.i.i.i3922, !amdgpu.uniform !42

Flow7788:                                         ; preds = %if.end22.i.i.i3922, %Flow7787
  %5191 = phi i32 [ %5178, %if.end22.i.i.i3922 ], [ %retval.sroa.8.0.i.i.i3832, %Flow7787 ]
  %5192 = phi i64 [ %5161, %if.end22.i.i.i3922 ], [ %retval.sroa.2.0.i.i.i3831, %Flow7787 ]
  %5193 = phi i32 [ %spec.store.select.i.i.i3835, %if.end22.i.i.i3922 ], [ %retval.sroa.6.0.i.i.i3829, %Flow7787 ]
  br label %cleanup26.i.i.i3858, !amdgpu.uniform !42

if.end22.i.i.i3922:                               ; preds = %if.then16.i.i.i3923, %if.end14.i.i.i3918
  call void @llvm.amdgcn.end.cf.i64(i64 %5189)
  call void @llvm.amdgcn.wave.barrier()
  br label %Flow7788, !amdgpu.uniform !42

Flow7789:                                         ; preds = %cleanup26.i.i.i3858, %5167
  %5194 = phi i32 [ %5193, %cleanup26.i.i.i3858 ], [ %5157, %5167 ]
  %5195 = phi i64 [ %5192, %cleanup26.i.i.i3858 ], [ %5158, %5167 ]
  %5196 = phi i32 [ %5191, %cleanup26.i.i.i3858 ], [ %5159, %5167 ]
  %5197 = phi i32 [ %5191, %cleanup26.i.i.i3858 ], [ %retval.sroa.8.0.i.i.i3832, %5167 ]
  %5198 = phi i64 [ %5192, %cleanup26.i.i.i3858 ], [ %retval.sroa.2.0.i.i.i3831, %5167 ]
  %5199 = phi i32 [ %5193, %cleanup26.i.i.i3858 ], [ %retval.sroa.6.0.i.i.i3829, %5167 ]
  %5200 = phi i1 [ %cmp.i46.not.i.i.i3850, %cleanup26.i.i.i3858 ], [ true, %5167 ]
  br i1 %5200, label %for.inc.i.i.i3863, label %Flow7790, !amdgpu.uniform !42

cleanup26.i.i.i3858:                              ; preds = %Flow7788
  br label %Flow7789, !amdgpu.uniform !42

for.inc.i.i.i3863:                                ; preds = %Flow7789
  %inc.i.i.i3868 = add i32 %spec.store.select.i.i.i3835, 1
  br label %Flow7790, !amdgpu.uniform !42

Flow7790:                                         ; preds = %for.inc.i.i.i3863, %Flow7789
  %5201 = phi i32 [ %inc.i.i.i3868, %for.inc.i.i.i3863 ], [ poison, %Flow7789 ]
  %5202 = phi i1 [ false, %for.inc.i.i.i3863 ], [ true, %Flow7789 ]
  br i1 %5202, label %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i3869, label %for.cond.i.i.i3828, !amdgpu.uniform !42

_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i3869: ; preds = %Flow7790
  %.lcssa8880 = phi i32 [ %5194, %Flow7790 ]
  %.lcssa8879 = phi i64 [ %5195, %Flow7790 ]
  %.lcssa8878 = phi i32 [ %5196, %Flow7790 ]
  fence acquire
  %5203 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %mul.i.i.i.i.i.i3870 = shl i32 %.lcssa8880, 6
  %idxprom.i19.i.i.i.i.i3871 = zext i32 %mul.i.i.i.i.i.i3870 to i64
  %arrayidx.i20.i.i.i.i.i3872 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %5203, i64 %idxprom.i19.i.i.i.i.i3871
  %arrayidx.i22.i.i.i.i.i3873 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %arrayidx.i20.i.i.i.i.i3872, i64 %sh_prom.i.i.i.i3270
  store i64 %writer.sroa.10.5.i.i.i.i3799, ptr %arrayidx.i22.i.i.i.i.i3873, align 8, !tbaa !82
  %spec.select.i.i.i.i.i.i.i3874 = call i64 @llvm.umin.i64(i64 %writer.sroa.10.5.i.i.i.i3799, i64 56)
  %5204 = and i64 %spec.select.i.i.i.i.i.i.i3874, 15
  %5205 = sub i64 %spec.select.i.i.i.i.i.i.i3874, %5204
  %5206 = icmp ne i64 %5205, 0
  %5207 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %5206)
  %5208 = extractvalue { i1, i64 } %5207, 0
  %5209 = extractvalue { i1, i64 } %5207, 1
  br i1 %5208, label %loop-memcpy-expansion4989.preheader, label %Flow7786

loop-memcpy-expansion4989.preheader:              ; preds = %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i3869
  %5210 = add nuw nsw i64 %4513, 8
  %scevgep6402 = getelementptr i8, ptr %5203, i64 %5210
  %5211 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i3871, 6
  %scevgep6403 = getelementptr i8, ptr %scevgep6402, i64 %5211
  br label %loop-memcpy-expansion4989, !amdgpu.uniform !42

Flow7786:                                         ; preds = %Flow7785, %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i3869
  call void @llvm.amdgcn.end.cf.i64(i64 %5209)
  br label %loop-memcpy-residual-header4992, !amdgpu.uniform !42

loop-memcpy-expansion4989:                        ; preds = %loop-memcpy-expansion4989.preheader, %loop-memcpy-expansion4989
  %phi.broken8798 = phi i64 [ 0, %loop-memcpy-expansion4989.preheader ], [ %5216, %loop-memcpy-expansion4989 ]
  %loop-index4990 = phi i64 [ %5214, %loop-memcpy-expansion4989 ], [ 0, %loop-memcpy-expansion4989.preheader ]
  %5212 = getelementptr i8, ptr %5128, i64 %loop-index4990
  %5213 = load <4 x i32>, ptr %5212, align 1
  %scevgep6404 = getelementptr i8, ptr %scevgep6403, i64 %loop-index4990
  store <4 x i32> %5213, ptr %scevgep6404, align 1
  %5214 = add i64 %loop-index4990, 16
  %5215 = icmp uge i64 %5214, %5205
  %5216 = call i64 @llvm.amdgcn.if.break.i64(i1 %5215, i64 %phi.broken8798)
  %5217 = call i1 @llvm.amdgcn.loop.i64(i64 %5216)
  br i1 %5217, label %Flow7785, label %loop-memcpy-expansion4989

Flow7784:                                         ; preds = %Flow7782, %Flow7783
  %5218 = phi i64 [ %6238, %Flow7782 ], [ %6230, %Flow7783 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %6233)
  br label %post-loop-memcpy-expansion4988, !amdgpu.uniform !42

loop-memcpy-residual4991:                         ; preds = %loop-memcpy-residual4991.preheader, %loop-memcpy-residual4991
  %phi.broken8799 = phi i64 [ 0, %loop-memcpy-residual4991.preheader ], [ %5222, %loop-memcpy-residual4991 ]
  %residual-loop-index4993 = phi i64 [ %5220, %loop-memcpy-residual4991 ], [ 0, %loop-memcpy-residual4991.preheader ]
  %scevgep6401 = getelementptr i8, ptr %scevgep6400, i64 %residual-loop-index4993
  %5219 = load i8, ptr %scevgep6401, align 1
  %scevgep6399 = getelementptr i8, ptr %scevgep6398, i64 %residual-loop-index4993
  store i8 %5219, ptr %scevgep6399, align 1
  %5220 = add i64 %residual-loop-index4993, 1
  %5221 = icmp uge i64 %5220, %5204
  %5222 = call i64 @llvm.amdgcn.if.break.i64(i1 %5221, i64 %phi.broken8799)
  %5223 = call i1 @llvm.amdgcn.loop.i64(i64 %5222)
  br i1 %5223, label %Flow7782, label %loop-memcpy-residual4991

Flow7782:                                         ; preds = %loop-memcpy-residual4991
  %.lcssa8873 = phi i64 [ %5222, %loop-memcpy-residual4991 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa8873)
  br label %Flow7784, !amdgpu.uniform !42

post-loop-memcpy-expansion4988:                   ; preds = %Flow7784
  %tobool.not.i.i.i.i.i.i3876 = icmp eq i32 %.lcssa8878, 0
  %conv.i.i.i.i.i.i3877 = zext i1 %tobool.not.i.i.i.i.i.i3876 to i32
  fence release
  %5224 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %idxprom.i23.i.i.i.i.i3878 = zext i32 %.lcssa8880 to i64
  %arrayidx.i24.i.i.i.i.i3879 = getelementptr inbounds nuw i32, ptr %5224, i64 %idxprom.i23.i.i.i.i.i3878
  store atomic i32 %conv.i.i.i.i.i.i3877, ptr %arrayidx.i24.i.i.i.i.i3879 monotonic, align 4
  %5225 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !84
  %arrayidx.i.i5.i.i3880 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %5225, i64 %idxprom.i23.i.i.i.i.i3878, !amdgpu.uniform !42
  %5226 = load i64, ptr %arrayidx.i.i5.i.i3880, align 8, !tbaa !80
  %scevgep6384 = getelementptr i8, ptr %5128, i64 56
  %5227 = add i64 %writer.sroa.10.5.i.i.i.i3799, -56
  br label %while.cond.i.i.i.i3881, !amdgpu.uniform !42

while.cond.i.i.i.i3881:                           ; preds = %post-loop-memcpy-expansion4988, %Flow7781
  %phi.broken8802 = phi i64 [ 0, %post-loop-memcpy-expansion4988 ], [ %5244, %Flow7781 ]
  %lsr.iv6390 = phi i64 [ %5242, %Flow7781 ], [ %5227, %post-loop-memcpy-expansion4988 ]
  %lsr.iv6385 = phi ptr [ %5241, %Flow7781 ], [ %scevgep6384, %post-loop-memcpy-expansion4988 ]
  %port.sroa.37.0.i.i3882 = phi i32 [ %5240, %Flow7781 ], [ %conv.i.i.i.i.i.i3877, %post-loop-memcpy-expansion4988 ]
  %idx.0.i.i.i.i3883 = phi i64 [ %5239, %Flow7781 ], [ 56, %post-loop-memcpy-expansion4988 ]
  %cmp.i.i.i.i3884 = icmp ugt i64 %writer.sroa.10.5.i.i.i.i3799, %idx.0.i.i.i.i3883
  %5228 = call i64 @llvm.amdgcn.ballot.i64(i1 %cmp.i.i.i.i3884)
  %and.i.i.i.i6.i.i3885 = and i64 %5228, %5226
  %tobool.not.i.i.i.i3886 = icmp ne i64 %and.i.i.i.i6.i.i3885, 0
  %5229 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i22.i.i.i3887 = getelementptr inbounds nuw i32, ptr %5229, i64 %idxprom.i23.i.i.i.i.i3878, !amdgpu.uniform !42
  %5230 = load atomic i32, ptr %arrayidx.i22.i.i.i3887 monotonic, align 4
  %5231 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %tobool.not.i.i.i.i3886)
  %5232 = extractvalue { i1, i64 } %5231, 0
  %5233 = extractvalue { i1, i64 } %5231, 1
  br i1 %5232, label %cond.false.i8.i.i.i.i3888, label %Flow7781

cond.false.i8.i.i.i.i3888:                        ; preds = %while.cond.i.i.i.i3881
  %5234 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %5230)
  %5235 = icmp ne i32 %5234, %port.sroa.37.0.i.i3882
  br i1 %5235, label %while.body.i.i21.i.i.i.i3889.preheader, label %Flow7780, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i3889.preheader:           ; preds = %cond.false.i8.i.i.i.i3888
  br label %while.body.i.i21.i.i.i.i3889, !amdgpu.uniform !42

Flow7780:                                         ; preds = %Flow7779, %cond.false.i8.i.i.i.i3888
  br label %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i3892, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i3889:                     ; preds = %while.body.i.i21.i.i.i.i3889.preheader, %while.body.i.i21.i.i.i.i3889
  call void @llvm.amdgcn.s.sleep(i32 2)
  %5236 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i22.i.i.i.i3890 = getelementptr inbounds nuw i32, ptr %5236, i64 %idxprom.i23.i.i.i.i.i3878, !amdgpu.uniform !42
  %5237 = load atomic i32, ptr %arrayidx.i.i.i22.i.i.i.i3890 monotonic, align 4
  %5238 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %5237)
  %cmp.i.not.i.i23.i.i.i.i3891 = icmp eq i32 %5238, %port.sroa.37.0.i.i3882
  br i1 %cmp.i.not.i.i23.i.i.i.i3891, label %Flow7779, label %while.body.i.i21.i.i.i.i3889, !amdgpu.uniform !42

Flow7779:                                         ; preds = %while.body.i.i21.i.i.i.i3889
  br label %Flow7780, !amdgpu.uniform !42

Flow7781:                                         ; preds = %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i3893, %while.cond.i.i.i.i3881
  %5239 = phi i64 [ %add.i.i.i.i3896, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i3893 ], [ poison, %while.cond.i.i.i.i3881 ]
  %5240 = phi i32 [ %conv.i.i26.i.i.i.i3894, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i3893 ], [ poison, %while.cond.i.i.i.i3881 ]
  %5241 = phi ptr [ %scevgep6386, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i3893 ], [ poison, %while.cond.i.i.i.i3881 ]
  %5242 = phi i64 [ %lsr.iv.next6391, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i3893 ], [ poison, %while.cond.i.i.i.i3881 ]
  %5243 = phi i1 [ false, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i3893 ], [ true, %while.cond.i.i.i.i3881 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %5233)
  %5244 = call i64 @llvm.amdgcn.if.break.i64(i1 %5243, i64 %phi.broken8802)
  %5245 = call i1 @llvm.amdgcn.loop.i64(i64 %5244)
  br i1 %5245, label %cond.false.i.i.i3903, label %while.cond.i.i.i.i3881

_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i3892: ; preds = %Flow7780
  fence acquire
  %5246 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.i.i.i.i3884)
  %5247 = extractvalue { i1, i64 } %5246, 0
  %5248 = extractvalue { i1, i64 } %5246, 1
  br i1 %5247, label %if.then.i.i.i.i.i.i3.i3897, label %Flow7778

if.then.i.i.i.i.i.i3.i3897:                       ; preds = %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i3892
  %sub.i.i.i.i.i.i.i3899 = sub nuw i64 %writer.sroa.10.5.i.i.i.i3799, %idx.0.i.i.i.i3883
  %spec.select.i.i.i36.i.i.i.i3900 = call i64 @llvm.umin.i64(i64 %sub.i.i.i.i.i.i.i3899, i64 64)
  %5249 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %5250 = and i64 %spec.select.i.i.i36.i.i.i.i3900, 15
  %5251 = sub i64 %spec.select.i.i.i36.i.i.i.i3900, %5250
  %5252 = icmp ne i64 %5251, 0
  %5253 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %5252)
  %5254 = extractvalue { i1, i64 } %5253, 0
  %5255 = extractvalue { i1, i64 } %5253, 1
  br i1 %5254, label %loop-memcpy-expansion4983.preheader, label %Flow7777

loop-memcpy-expansion4983.preheader:              ; preds = %if.then.i.i.i.i.i.i3.i3897
  %arrayidx.i20.i35.i.i.i.i3901 = getelementptr %"struct.rpc::Buffer", ptr %5249, i64 %idxprom.i19.i.i.i.i.i3871
  %arrayidx.i22.i38.i.i.i.i3902 = getelementptr %"struct.rpc::Buffer", ptr %arrayidx.i20.i35.i.i.i.i3901, i64 %sh_prom.i.i.i.i3270
  br label %loop-memcpy-expansion4983, !amdgpu.uniform !42

Flow7777:                                         ; preds = %Flow7776, %if.then.i.i.i.i.i.i3.i3897
  call void @llvm.amdgcn.end.cf.i64(i64 %5255)
  br label %loop-memcpy-residual-header4986, !amdgpu.uniform !42

loop-memcpy-expansion4983:                        ; preds = %loop-memcpy-expansion4983.preheader, %loop-memcpy-expansion4983
  %phi.broken8800 = phi i64 [ 0, %loop-memcpy-expansion4983.preheader ], [ %5260, %loop-memcpy-expansion4983 ]
  %loop-index4984 = phi i64 [ %5258, %loop-memcpy-expansion4983 ], [ 0, %loop-memcpy-expansion4983.preheader ]
  %scevgep6387 = getelementptr i8, ptr %lsr.iv6385, i64 %loop-index4984
  %5256 = load <4 x i32>, ptr %scevgep6387, align 1
  %5257 = getelementptr i8, ptr %arrayidx.i22.i38.i.i.i.i3902, i64 %loop-index4984
  store <4 x i32> %5256, ptr %5257, align 1
  %5258 = add i64 %loop-index4984, 16
  %5259 = icmp uge i64 %5258, %5251
  %5260 = call i64 @llvm.amdgcn.if.break.i64(i1 %5259, i64 %phi.broken8800)
  %5261 = call i1 @llvm.amdgcn.loop.i64(i64 %5260)
  br i1 %5261, label %Flow7776, label %loop-memcpy-expansion4983

Flow7775:                                         ; preds = %Flow7774, %loop-memcpy-residual-header4986
  call void @llvm.amdgcn.end.cf.i64(i64 %6223)
  br label %Flow7778, !amdgpu.uniform !42

loop-memcpy-residual4985:                         ; preds = %loop-memcpy-residual4985.preheader, %loop-memcpy-residual4985
  %phi.broken8801 = phi i64 [ 0, %loop-memcpy-residual4985.preheader ], [ %5265, %loop-memcpy-residual4985 ]
  %residual-loop-index4987 = phi i64 [ %5263, %loop-memcpy-residual4985 ], [ 0, %loop-memcpy-residual4985.preheader ]
  %scevgep6396 = getelementptr i8, ptr %scevgep6395, i64 %residual-loop-index4987
  %5262 = load i8, ptr %scevgep6396, align 1
  %scevgep6394 = getelementptr i8, ptr %scevgep6393, i64 %residual-loop-index4987
  store i8 %5262, ptr %scevgep6394, align 1
  %5263 = add i64 %residual-loop-index4987, 1
  %5264 = icmp uge i64 %5263, %5250
  %5265 = call i64 @llvm.amdgcn.if.break.i64(i1 %5264, i64 %phi.broken8801)
  %5266 = call i1 @llvm.amdgcn.loop.i64(i64 %5265)
  br i1 %5266, label %Flow7774, label %loop-memcpy-residual4985

Flow7774:                                         ; preds = %loop-memcpy-residual4985
  %.lcssa8870 = phi i64 [ %5265, %loop-memcpy-residual4985 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa8870)
  br label %Flow7775, !amdgpu.uniform !42

_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i3893: ; preds = %Flow7778
  %conv.i.i26.i.i.i.i3894 = xor i32 %port.sroa.37.0.i.i3882, 1
  fence release
  %5267 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %arrayidx.i24.i29.i.i.i.i3895 = getelementptr inbounds nuw i32, ptr %5267, i64 %idxprom.i23.i.i.i.i.i3878
  store atomic i32 %conv.i.i26.i.i.i.i3894, ptr %arrayidx.i24.i29.i.i.i.i3895 monotonic, align 4
  %add.i.i.i.i3896 = add i64 %idx.0.i.i.i.i3883, 64
  %scevgep6386 = getelementptr i8, ptr %lsr.iv6385, i64 64
  %lsr.iv.next6391 = add i64 %lsr.iv6390, -64
  br label %Flow7781, !amdgpu.uniform !42

cond.false.i.i.i3903:                             ; preds = %Flow7781
  %.lcssa8872 = phi i64 [ %5244, %Flow7781 ]
  %port.sroa.37.0.i.i3882.lcssa = phi i32 [ %port.sroa.37.0.i.i3882, %Flow7781 ]
  %.lcssa8871 = phi i32 [ %5230, %Flow7781 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa8872)
  %5268 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %.lcssa8871)
  %5269 = icmp ne i32 %5268, %port.sroa.37.0.i.i3882.lcssa
  %5270 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %5269)
  %5271 = extractvalue { i1, i64 } %5270, 0
  %5272 = extractvalue { i1, i64 } %5270, 1
  br i1 %5271, label %while.body.i.i.i.i3904.preheader, label %Flow7773

while.body.i.i.i.i3904.preheader:                 ; preds = %cond.false.i.i.i3903
  br label %while.body.i.i.i.i3904, !amdgpu.uniform !42

Flow7773:                                         ; preds = %Flow7772, %cond.false.i.i.i3903
  call void @llvm.amdgcn.end.cf.i64(i64 %5272)
  br label %_ZN22__llvm_libc_22_0_0_git15write_to_stderrENS_3cpp11string_viewE.exit.i3907, !amdgpu.uniform !42

while.body.i.i.i.i3904:                           ; preds = %while.body.i.i.i.i3904.preheader, %while.body.i.i.i.i3904
  %phi.broken8803 = phi i64 [ 0, %while.body.i.i.i.i3904.preheader ], [ %5276, %while.body.i.i.i.i3904 ]
  call void @llvm.amdgcn.s.sleep(i32 2)
  %5273 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i8.i.i3905 = getelementptr inbounds nuw i32, ptr %5273, i64 %idxprom.i23.i.i.i.i.i3878, !amdgpu.uniform !42
  %5274 = load atomic i32, ptr %arrayidx.i.i.i8.i.i3905 monotonic, align 4
  %5275 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %5274)
  %cmp.i.not.i.i.i.i3906 = icmp eq i32 %5275, %port.sroa.37.0.i.i3882.lcssa
  %5276 = call i64 @llvm.amdgcn.if.break.i64(i1 %cmp.i.not.i.i.i.i3906, i64 %phi.broken8803)
  %5277 = call i1 @llvm.amdgcn.loop.i64(i64 %5276)
  br i1 %5277, label %Flow7772, label %while.body.i.i.i.i3904

Flow7772:                                         ; preds = %while.body.i.i.i.i3904
  %.lcssa8868 = phi i64 [ %5276, %while.body.i.i.i.i3904 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa8868)
  br label %Flow7773, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git15write_to_stderrENS_3cpp11string_viewE.exit.i3907: ; preds = %Flow7773
  fence acquire
  call void @llvm.amdgcn.wave.barrier()
  fence syncscope("agent") release
  %5278 = call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %.lcssa8879, i1 true)
  %iszero.i.i.i.i.i13.i.i3908 = icmp ne i64 %.lcssa8879, 0
  %cmp2.i.i.i.i14.i.i3909 = icmp eq i64 %5278, %sh_prom.i.i.i.i3270
  %cmp.i.i.i.i15.i.i3910 = select i1 %iszero.i.i.i.i.i13.i.i3908, i1 %cmp2.i.i.i.i14.i.i3909, i1 false
  %5279 = and i32 %.lcssa8880, 31
  %conv4.i.i.i19.i.i3914 = zext i1 %cmp.i.i.i.i15.i.i3910 to i32
  %shl.i.i.i20.i.i3915 = shl nuw i32 %conv4.i.i.i19.i.i3914, %5279
  %xor.i.i.i21.i.i3916 = xor i32 %shl.i.i.i20.i.i3915, -1
  %5280 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop5914, !amdgpu.uniform !42

5281:                                             ; preds = %ComputeEnd5915
  %div8.i.i.i16.i.i3911 = lshr i32 %.lcssa8880, 5
  %idxprom.i.i.i17.i.i3912 = zext nneg i32 %div8.i.i.i16.i.i3911 to i64
  %5282 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i17.i.i3912
  %sunkaddr7487 = getelementptr inbounds i8, ptr addrspace(1) %5282, i64 40
  %5283 = atomicrmw and ptr addrspace(1) %sunkaddr7487, i32 %.lcssa8867 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %5284, !amdgpu.uniform !42

5284:                                             ; preds = %5281, %ComputeEnd5915
  call void @llvm.amdgcn.end.cf.i64(i64 %8578)
  call void @llvm.amdgcn.wave.barrier()
  br label %for.cond.i.i.i3942, !amdgpu.uniform !42

for.cond.i.i.i3942:                               ; preds = %5284, %Flow7771
  %5285 = phi i32 [ %5322, %Flow7771 ], [ poison, %5284 ]
  %5286 = phi i64 [ %5323, %Flow7771 ], [ poison, %5284 ]
  %5287 = phi i32 [ %5324, %Flow7771 ], [ poison, %5284 ]
  %retval.sroa.6.0.i.i.i3943 = phi i32 [ %5327, %Flow7771 ], [ undef, %5284 ]
  %retval.sroa.2.0.i.i.i3945 = phi i64 [ %5326, %Flow7771 ], [ undef, %5284 ]
  %retval.sroa.8.0.i.i.i3946 = phi i32 [ %5325, %Flow7771 ], [ undef, %5284 ]
  %index.0.i.i.i3947 = phi i32 [ %5329, %Flow7771 ], [ 0, %5284 ]
  %5288 = load i32, ptr addrspace(1) @__llvm_rpc_client, align 8, !tbaa !68
  %cmp.not.i.i.i3948 = icmp ult i32 %index.0.i.i.i3947, %5288
  %spec.store.select.i.i.i3949 = select i1 %cmp.not.i.i.i3948, i32 %index.0.i.i.i3947, i32 0
  %5289 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  %5290 = and i32 %spec.store.select.i.i.i3949, 31
  %5291 = lshr i64 %5289, %sh_prom.i.i.i.i3270
  %5292 = trunc i64 %5291 to i32
  %conv4.i.i.i.i.i3953 = and i32 %5292, 1
  %shl.i.i.i.i.i3954 = shl nuw i32 %conv4.i.i.i.i.i3953, %5290
  br label %ComputeLoop5918, !amdgpu.uniform !42

5293:                                             ; preds = %ComputeEnd5919
  %sunkaddr7488 = getelementptr inbounds i8, ptr addrspace(1) %8588, i64 40
  %5294 = atomicrmw or ptr addrspace(1) %sunkaddr7488, i32 %.lcssa8861 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %5295, !amdgpu.uniform !42

5295:                                             ; preds = %5293, %ComputeEnd5919
  %5296 = phi i32 [ %5294, %5293 ], [ poison, %ComputeEnd5919 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %8597)
  %5297 = call i32 @llvm.amdgcn.readfirstlane.i32(i32 %5296)
  %5298 = or i32 %5297, %.lcssa8862
  %shl5.i.i.i.i.i3955 = shl nuw i32 1, %5290
  %and.i.i.i.i.i3956 = and i32 %shl5.i.i.i.i.i3955, %5298
  %tobool3.i.i.i.i3957 = icmp ne i32 %and.i.i.i.i.i3956, 0
  %5299 = call i64 @llvm.amdgcn.ballot.i64(i1 %tobool3.i.i.i.i3957)
  %and.i.i.i.i.i.i3958 = and i64 %5299, %5289
  %cmp.i.not.i.i.i3959 = icmp ne i64 %5289, %and.i.i.i.i.i.i3958
  br i1 %cmp.i.not.i.i.i3959, label %if.then.i.i.i.i3960, label %Flow7770, !amdgpu.uniform !42

if.then.i.i.i.i3960:                              ; preds = %5295
  fence syncscope("agent") acquire
  %5300 = load <2 x i64>, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !75
  %5301 = extractelement <2 x i64> %5300, i32 0
  %5302 = inttoptr i64 %5301 to ptr
  %5303 = extractelement <2 x i64> %5300, i32 1
  %5304 = inttoptr i64 %5303 to ptr
  %idxprom.i.i.i.i3961 = zext i32 %spec.store.select.i.i.i3949 to i64
  %arrayidx.i.i.i.i3962 = getelementptr inbounds nuw i32, ptr %5302, i64 %idxprom.i.i.i.i3961, !amdgpu.uniform !42
  %5305 = load atomic i32, ptr %arrayidx.i.i.i.i3962 monotonic, align 4
  %5306 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %5305)
  %arrayidx.i45.i.i.i3963 = getelementptr inbounds nuw i32, ptr %5304, i64 %idxprom.i.i.i.i3961, !amdgpu.uniform !42
  %5307 = load atomic i32, ptr %arrayidx.i45.i.i.i3963 monotonic, align 4
  %5308 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %5307)
  %cmp.i46.not.i.i.i3964 = icmp ne i32 %5306, %5308
  br i1 %cmp.i46.not.i.i.i3964, label %if.then12.i.i.i3965, label %Flow7768, !amdgpu.uniform !42

if.then12.i.i.i3965:                              ; preds = %if.then.i.i.i.i3960
  fence syncscope("agent") release
  %5309 = call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %5289, i1 true)
  %iszero.i.i.i.i.i.i.i3966 = icmp ne i64 %5289, 0
  %cmp2.i.i.i.i.i.i3967 = icmp eq i64 %5309, %sh_prom.i.i.i.i3270
  %cmp.i.i.i.i.i.i3968 = select i1 %iszero.i.i.i.i.i.i.i3966, i1 %cmp2.i.i.i.i.i.i3967, i1 false
  %conv4.i.i51.i.i.i3969 = zext i1 %cmp.i.i.i.i.i.i3968 to i32
  %shl.i.i52.i.i.i3970 = shl nuw i32 %conv4.i.i51.i.i.i3969, %5290
  %xor.i.i.i.i.i3971 = xor i32 %shl.i.i52.i.i.i3970, -1
  %5310 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop5923, !amdgpu.uniform !42

5311:                                             ; preds = %ComputeEnd5924
  %sunkaddr7489 = getelementptr inbounds i8, ptr addrspace(1) %8588, i64 40
  %5312 = atomicrmw and ptr addrspace(1) %sunkaddr7489, i32 %.lcssa8863 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %5313, !amdgpu.uniform !42

5313:                                             ; preds = %5311, %ComputeEnd5924
  call void @llvm.amdgcn.end.cf.i64(i64 %8615)
  call void @llvm.amdgcn.wave.barrier()
  br label %Flow7768, !amdgpu.uniform !42

if.end14.i.i.i4029:                               ; preds = %Flow7768
  %5314 = call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %5289, i1 true)
  %iszero.i.i.i.i.i.i4030 = icmp ne i64 %5289, 0
  %cmp2.i.i.i.i.i4031 = icmp eq i64 %5314, %sh_prom.i.i.i.i3270
  %cmp.i.i.i.i.i4032 = select i1 %iszero.i.i.i.i.i.i4030, i1 %cmp2.i.i.i.i.i4031, i1 false
  %5315 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.i.i.i.i.i4032)
  %5316 = extractvalue { i1, i64 } %5315, 0
  %5317 = extractvalue { i1, i64 } %5315, 1
  br i1 %5316, label %if.then16.i.i.i4034, label %if.end22.i.i.i4033

if.then16.i.i.i4034:                              ; preds = %if.end14.i.i.i4029
  %5318 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !76
  %opcode.i.i.i4035 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %5318, i64 %idxprom.i.i.i.i3961, i32 1
  store i32 1660944387, ptr %opcode.i.i.i4035, align 8, !tbaa !77
  %arrayidx21.i.i.i4036 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %5318, i64 %idxprom.i.i.i.i3961
  store i64 %5289, ptr %arrayidx21.i.i.i4036, align 8, !tbaa !80
  br label %if.end22.i.i.i4033, !amdgpu.uniform !42

Flow7769:                                         ; preds = %if.end22.i.i.i4033, %Flow7768
  %5319 = phi i32 [ %5306, %if.end22.i.i.i4033 ], [ %retval.sroa.8.0.i.i.i3946, %Flow7768 ]
  %5320 = phi i64 [ %5289, %if.end22.i.i.i4033 ], [ %retval.sroa.2.0.i.i.i3945, %Flow7768 ]
  %5321 = phi i32 [ %spec.store.select.i.i.i3949, %if.end22.i.i.i4033 ], [ %retval.sroa.6.0.i.i.i3943, %Flow7768 ]
  br label %cleanup26.i.i.i3972, !amdgpu.uniform !42

if.end22.i.i.i4033:                               ; preds = %if.then16.i.i.i4034, %if.end14.i.i.i4029
  call void @llvm.amdgcn.end.cf.i64(i64 %5317)
  call void @llvm.amdgcn.wave.barrier()
  br label %Flow7769, !amdgpu.uniform !42

Flow7770:                                         ; preds = %cleanup26.i.i.i3972, %5295
  %5322 = phi i32 [ %5321, %cleanup26.i.i.i3972 ], [ %5285, %5295 ]
  %5323 = phi i64 [ %5320, %cleanup26.i.i.i3972 ], [ %5286, %5295 ]
  %5324 = phi i32 [ %5319, %cleanup26.i.i.i3972 ], [ %5287, %5295 ]
  %5325 = phi i32 [ %5319, %cleanup26.i.i.i3972 ], [ %retval.sroa.8.0.i.i.i3946, %5295 ]
  %5326 = phi i64 [ %5320, %cleanup26.i.i.i3972 ], [ %retval.sroa.2.0.i.i.i3945, %5295 ]
  %5327 = phi i32 [ %5321, %cleanup26.i.i.i3972 ], [ %retval.sroa.6.0.i.i.i3943, %5295 ]
  %5328 = phi i1 [ %cmp.i46.not.i.i.i3964, %cleanup26.i.i.i3972 ], [ true, %5295 ]
  br i1 %5328, label %for.inc.i.i.i3977, label %Flow7771, !amdgpu.uniform !42

cleanup26.i.i.i3972:                              ; preds = %Flow7769
  br label %Flow7770, !amdgpu.uniform !42

for.inc.i.i.i3977:                                ; preds = %Flow7770
  %inc.i.i.i3982 = add i32 %spec.store.select.i.i.i3949, 1
  br label %Flow7771, !amdgpu.uniform !42

Flow7771:                                         ; preds = %for.inc.i.i.i3977, %Flow7770
  %5329 = phi i32 [ %inc.i.i.i3982, %for.inc.i.i.i3977 ], [ poison, %Flow7770 ]
  %5330 = phi i1 [ false, %for.inc.i.i.i3977 ], [ true, %Flow7770 ]
  br i1 %5330, label %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i3983, label %for.cond.i.i.i3942, !amdgpu.uniform !42

_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i3983: ; preds = %Flow7771
  %.lcssa8866 = phi i32 [ %5322, %Flow7771 ]
  %.lcssa8865 = phi i64 [ %5323, %Flow7771 ]
  %.lcssa8864 = phi i32 [ %5324, %Flow7771 ]
  fence acquire
  %5331 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %mul.i.i.i.i.i.i3984 = shl i32 %.lcssa8866, 6
  %idxprom.i19.i.i.i.i.i3985 = zext i32 %mul.i.i.i.i.i.i3984 to i64
  %arrayidx.i20.i.i.i.i.i3986 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %5331, i64 %idxprom.i19.i.i.i.i.i3985
  %arrayidx.i22.i.i.i.i.i3987 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %arrayidx.i20.i.i.i.i.i3986, i64 %sh_prom.i.i.i.i3270
  store i64 1, ptr %arrayidx.i22.i.i.i.i.i3987, align 8, !tbaa !82
  %arrayidx6.i.i.i.i.i.i.i3988 = getelementptr inbounds nuw i8, ptr %arrayidx.i22.i.i.i.i.i3987, i64 8
  store i8 32, ptr %arrayidx6.i.i.i.i.i.i.i3988, align 8
  %tobool.not.i.i.i.i.i.i3989 = icmp eq i32 %.lcssa8864, 0
  %conv.i.i.i.i.i.i3990 = zext i1 %tobool.not.i.i.i.i.i.i3989 to i32
  fence release
  %5332 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %idxprom.i23.i.i.i.i.i3991 = zext i32 %.lcssa8866 to i64
  %arrayidx.i24.i.i.i.i.i3992 = getelementptr inbounds nuw i32, ptr %5332, i64 %idxprom.i23.i.i.i.i.i3991
  store atomic i32 %conv.i.i.i.i.i.i3990, ptr %arrayidx.i24.i.i.i.i.i3992 monotonic, align 4
  %5333 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !84
  %arrayidx.i.i5.i.i3993 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %5333, i64 %idxprom.i23.i.i.i.i.i3991, !amdgpu.uniform !42
  %5334 = load i64, ptr %arrayidx.i.i5.i.i3993, align 8, !tbaa !80
  br label %while.cond.i.i.i.i3994, !amdgpu.uniform !42

while.cond.i.i.i.i3994:                           ; preds = %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i3983, %Flow7767
  %phi.broken8804 = phi i64 [ 0, %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i3983 ], [ %5349, %Flow7767 ]
  %port.sroa.37.0.i.i3995 = phi i32 [ %5347, %Flow7767 ], [ %conv.i.i.i.i.i.i3990, %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i3983 ]
  %idx.0.i.i.i.i3996 = phi i64 [ %5346, %Flow7767 ], [ 56, %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i3983 ]
  %cmp.i.i.i.i3997 = icmp eq i64 %idx.0.i.i.i.i3996, 0
  %5335 = call i64 @llvm.amdgcn.ballot.i64(i1 %cmp.i.i.i.i3997)
  %and.i.i.i.i6.i.i3998 = and i64 %5335, %5334
  %tobool.not.i.i.i.i3999 = icmp ne i64 %and.i.i.i.i6.i.i3998, 0
  %5336 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i22.i.i.i4000 = getelementptr inbounds nuw i32, ptr %5336, i64 %idxprom.i23.i.i.i.i.i3991, !amdgpu.uniform !42
  %5337 = load atomic i32, ptr %arrayidx.i22.i.i.i4000 monotonic, align 4
  %5338 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %tobool.not.i.i.i.i3999)
  %5339 = extractvalue { i1, i64 } %5338, 0
  %5340 = extractvalue { i1, i64 } %5338, 1
  br i1 %5339, label %cond.false.i8.i.i.i.i4001, label %Flow7767

cond.false.i8.i.i.i.i4001:                        ; preds = %while.cond.i.i.i.i3994
  %5341 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %5337)
  %5342 = icmp ne i32 %5341, %port.sroa.37.0.i.i3995
  br i1 %5342, label %while.body.i.i21.i.i.i.i4002.preheader, label %Flow7766, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i4002.preheader:           ; preds = %cond.false.i8.i.i.i.i4001
  br label %while.body.i.i21.i.i.i.i4002, !amdgpu.uniform !42

Flow7766:                                         ; preds = %Flow7765, %cond.false.i8.i.i.i.i4001
  br label %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i4005, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i4002:                     ; preds = %while.body.i.i21.i.i.i.i4002.preheader, %while.body.i.i21.i.i.i.i4002
  call void @llvm.amdgcn.s.sleep(i32 2)
  %5343 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i22.i.i.i.i4003 = getelementptr inbounds nuw i32, ptr %5343, i64 %idxprom.i23.i.i.i.i.i3991, !amdgpu.uniform !42
  %5344 = load atomic i32, ptr %arrayidx.i.i.i22.i.i.i.i4003 monotonic, align 4
  %5345 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %5344)
  %cmp.i.not.i.i23.i.i.i.i4004 = icmp eq i32 %5345, %port.sroa.37.0.i.i3995
  br i1 %cmp.i.not.i.i23.i.i.i.i4004, label %Flow7765, label %while.body.i.i21.i.i.i.i4002, !amdgpu.uniform !42

Flow7765:                                         ; preds = %while.body.i.i21.i.i.i.i4002
  br label %Flow7766, !amdgpu.uniform !42

Flow7767:                                         ; preds = %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i4006, %while.cond.i.i.i.i3994
  %5346 = phi i64 [ %add.i.i.i.i4009, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i4006 ], [ poison, %while.cond.i.i.i.i3994 ]
  %5347 = phi i32 [ %conv.i.i26.i.i.i.i4007, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i4006 ], [ poison, %while.cond.i.i.i.i3994 ]
  %5348 = phi i1 [ false, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i4006 ], [ true, %while.cond.i.i.i.i3994 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %5340)
  %5349 = call i64 @llvm.amdgcn.if.break.i64(i1 %5348, i64 %phi.broken8804)
  %5350 = call i1 @llvm.amdgcn.loop.i64(i64 %5349)
  br i1 %5350, label %cond.false.i.i.i4015, label %while.cond.i.i.i.i3994

_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i4005: ; preds = %Flow7766
  fence acquire
  br i1 %cmp.i.i.i.i3997, label %if.then.i.i.i.i.i.i.i4010, label %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i4006, !amdgpu.uniform !42

if.then.i.i.i.i.i.i.i4010:                        ; preds = %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i4005
  %5351 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %arrayidx.i20.i35.i.i.i.i4013 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %5351, i64 %idxprom.i19.i.i.i.i.i3985
  %arrayidx.i22.i38.i.i.i.i4014 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %arrayidx.i20.i35.i.i.i.i4013, i64 %sh_prom.i.i.i.i3270
  store i8 32, ptr %arrayidx.i22.i38.i.i.i.i4014, align 1
  br label %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i4006, !amdgpu.uniform !42

_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i4006: ; preds = %if.then.i.i.i.i.i.i.i4010, %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i4005
  %conv.i.i26.i.i.i.i4007 = xor i32 %port.sroa.37.0.i.i3995, 1
  fence release
  %5352 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %arrayidx.i24.i29.i.i.i.i4008 = getelementptr inbounds nuw i32, ptr %5352, i64 %idxprom.i23.i.i.i.i.i3991
  store atomic i32 %conv.i.i26.i.i.i.i4007, ptr %arrayidx.i24.i29.i.i.i.i4008 monotonic, align 4
  %add.i.i.i.i4009 = add i64 %idx.0.i.i.i.i3996, 64
  br label %Flow7767, !amdgpu.uniform !42

cond.false.i.i.i4015:                             ; preds = %Flow7767
  %.lcssa8860 = phi i64 [ %5349, %Flow7767 ]
  %port.sroa.37.0.i.i3995.lcssa = phi i32 [ %port.sroa.37.0.i.i3995, %Flow7767 ]
  %.lcssa8859 = phi i32 [ %5337, %Flow7767 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa8860)
  %5353 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %.lcssa8859)
  %5354 = icmp ne i32 %5353, %port.sroa.37.0.i.i3995.lcssa
  %5355 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %5354)
  %5356 = extractvalue { i1, i64 } %5355, 0
  %5357 = extractvalue { i1, i64 } %5355, 1
  br i1 %5356, label %while.body.i.i.i.i4016.preheader, label %Flow7764

while.body.i.i.i.i4016.preheader:                 ; preds = %cond.false.i.i.i4015
  br label %while.body.i.i.i.i4016, !amdgpu.uniform !42

Flow7764:                                         ; preds = %Flow7763, %cond.false.i.i.i4015
  call void @llvm.amdgcn.end.cf.i64(i64 %5357)
  br label %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIcEERS1_T_.exit4037, !amdgpu.uniform !42

while.body.i.i.i.i4016:                           ; preds = %while.body.i.i.i.i4016.preheader, %while.body.i.i.i.i4016
  %phi.broken8805 = phi i64 [ 0, %while.body.i.i.i.i4016.preheader ], [ %5361, %while.body.i.i.i.i4016 ]
  call void @llvm.amdgcn.s.sleep(i32 2)
  %5358 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i8.i.i4017 = getelementptr inbounds nuw i32, ptr %5358, i64 %idxprom.i23.i.i.i.i.i3991, !amdgpu.uniform !42
  %5359 = load atomic i32, ptr %arrayidx.i.i.i8.i.i4017 monotonic, align 4
  %5360 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %5359)
  %cmp.i.not.i.i.i.i4018 = icmp eq i32 %5360, %port.sroa.37.0.i.i3995.lcssa
  %5361 = call i64 @llvm.amdgcn.if.break.i64(i1 %cmp.i.not.i.i.i.i4018, i64 %phi.broken8805)
  %5362 = call i1 @llvm.amdgcn.loop.i64(i64 %5361)
  br i1 %5362, label %Flow7763, label %while.body.i.i.i.i4016

Flow7763:                                         ; preds = %while.body.i.i.i.i4016
  %.lcssa8858 = phi i64 [ %5361, %while.body.i.i.i.i4016 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa8858)
  br label %Flow7764, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIcEERS1_T_.exit4037: ; preds = %Flow7764
  fence acquire
  call void @llvm.amdgcn.wave.barrier()
  fence syncscope("agent") release
  %5363 = call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %.lcssa8865, i1 true)
  %iszero.i.i.i.i.i13.i.i4020 = icmp ne i64 %.lcssa8865, 0
  %cmp2.i.i.i.i14.i.i4021 = icmp eq i64 %5363, %sh_prom.i.i.i.i3270
  %cmp.i.i.i.i15.i.i4022 = select i1 %iszero.i.i.i.i.i13.i.i4020, i1 %cmp2.i.i.i.i14.i.i4021, i1 false
  %5364 = and i32 %.lcssa8866, 31
  %conv4.i.i.i19.i.i4026 = zext i1 %cmp.i.i.i.i15.i.i4022 to i32
  %shl.i.i.i20.i.i4027 = shl nuw i32 %conv4.i.i.i19.i.i4026, %5364
  %xor.i.i.i21.i.i4028 = xor i32 %shl.i.i.i20.i.i4027, -1
  %5365 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop5927, !amdgpu.uniform !42

5366:                                             ; preds = %ComputeEnd5928
  %div8.i.i.i16.i.i4023 = lshr i32 %.lcssa8866, 5
  %idxprom.i.i.i17.i.i4024 = zext nneg i32 %div8.i.i.i16.i.i4023 to i64
  %5367 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i17.i.i4024
  %sunkaddr7490 = getelementptr inbounds i8, ptr addrspace(1) %5367, i64 40
  %5368 = atomicrmw and ptr addrspace(1) %sunkaddr7490, i32 %.lcssa8857 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %5369, !amdgpu.uniform !42

5369:                                             ; preds = %5366, %ComputeEnd5928
  call void @llvm.amdgcn.end.cf.i64(i64 %8632)
  call void @llvm.amdgcn.wave.barrier()
  br label %for.cond.i.i1.i4046, !amdgpu.uniform !42

for.cond.i.i1.i4046:                              ; preds = %5369, %Flow7762
  %5370 = phi i32 [ %5407, %Flow7762 ], [ poison, %5369 ]
  %5371 = phi i64 [ %5408, %Flow7762 ], [ poison, %5369 ]
  %5372 = phi i32 [ %5409, %Flow7762 ], [ poison, %5369 ]
  %retval.sroa.6.0.i.i.i4047 = phi i32 [ %5412, %Flow7762 ], [ undef, %5369 ]
  %retval.sroa.2.0.i.i.i4049 = phi i64 [ %5411, %Flow7762 ], [ undef, %5369 ]
  %retval.sroa.8.0.i.i.i4050 = phi i32 [ %5410, %Flow7762 ], [ undef, %5369 ]
  %index.0.i.i.i4051 = phi i32 [ %5414, %Flow7762 ], [ 0, %5369 ]
  %5373 = load i32, ptr addrspace(1) @__llvm_rpc_client, align 8, !tbaa !68
  %cmp.not.i.i.i4052 = icmp ult i32 %index.0.i.i.i4051, %5373
  %spec.store.select.i.i.i4053 = select i1 %cmp.not.i.i.i4052, i32 %index.0.i.i.i4051, i32 0
  %5374 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  %5375 = and i32 %spec.store.select.i.i.i4053, 31
  %5376 = lshr i64 %5374, %sh_prom.i.i.i.i3270
  %5377 = trunc i64 %5376 to i32
  %conv4.i.i.i.i.i4057 = and i32 %5377, 1
  %shl.i.i.i.i.i4058 = shl nuw i32 %conv4.i.i.i.i.i4057, %5375
  br label %ComputeLoop5931, !amdgpu.uniform !42

5378:                                             ; preds = %ComputeEnd5932
  %sunkaddr7491 = getelementptr inbounds i8, ptr addrspace(1) %8642, i64 40
  %5379 = atomicrmw or ptr addrspace(1) %sunkaddr7491, i32 %.lcssa8851 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %5380, !amdgpu.uniform !42

5380:                                             ; preds = %5378, %ComputeEnd5932
  %5381 = phi i32 [ %5379, %5378 ], [ poison, %ComputeEnd5932 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %8651)
  %5382 = call i32 @llvm.amdgcn.readfirstlane.i32(i32 %5381)
  %5383 = or i32 %5382, %.lcssa8852
  %shl5.i.i.i.i.i4059 = shl nuw i32 1, %5375
  %and.i.i.i.i.i4060 = and i32 %shl5.i.i.i.i.i4059, %5383
  %tobool3.i.i.i.i4061 = icmp ne i32 %and.i.i.i.i.i4060, 0
  %5384 = call i64 @llvm.amdgcn.ballot.i64(i1 %tobool3.i.i.i.i4061)
  %and.i.i.i.i.i.i4062 = and i64 %5384, %5374
  %cmp.i.not.i.i.i4063 = icmp ne i64 %5374, %and.i.i.i.i.i.i4062
  br i1 %cmp.i.not.i.i.i4063, label %if.then.i.i.i.i4064, label %Flow7761, !amdgpu.uniform !42

if.then.i.i.i.i4064:                              ; preds = %5380
  fence syncscope("agent") acquire
  %5385 = load <2 x i64>, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !75
  %5386 = extractelement <2 x i64> %5385, i32 0
  %5387 = inttoptr i64 %5386 to ptr
  %5388 = extractelement <2 x i64> %5385, i32 1
  %5389 = inttoptr i64 %5388 to ptr
  %idxprom.i.i.i.i4065 = zext i32 %spec.store.select.i.i.i4053 to i64
  %arrayidx.i.i.i.i4066 = getelementptr inbounds nuw i32, ptr %5387, i64 %idxprom.i.i.i.i4065, !amdgpu.uniform !42
  %5390 = load atomic i32, ptr %arrayidx.i.i.i.i4066 monotonic, align 4
  %5391 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %5390)
  %arrayidx.i45.i.i.i4067 = getelementptr inbounds nuw i32, ptr %5389, i64 %idxprom.i.i.i.i4065, !amdgpu.uniform !42
  %5392 = load atomic i32, ptr %arrayidx.i45.i.i.i4067 monotonic, align 4
  %5393 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %5392)
  %cmp.i46.not.i.i.i4068 = icmp ne i32 %5391, %5393
  br i1 %cmp.i46.not.i.i.i4068, label %if.then12.i.i.i4069, label %Flow7759, !amdgpu.uniform !42

if.then12.i.i.i4069:                              ; preds = %if.then.i.i.i.i4064
  fence syncscope("agent") release
  %5394 = call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %5374, i1 true)
  %iszero.i.i.i.i.i.i.i4070 = icmp ne i64 %5374, 0
  %cmp2.i.i.i.i.i.i4071 = icmp eq i64 %5394, %sh_prom.i.i.i.i3270
  %cmp.i.i.i.i.i.i4072 = select i1 %iszero.i.i.i.i.i.i.i4070, i1 %cmp2.i.i.i.i.i.i4071, i1 false
  %conv4.i.i51.i.i.i4073 = zext i1 %cmp.i.i.i.i.i.i4072 to i32
  %shl.i.i52.i.i.i4074 = shl nuw i32 %conv4.i.i51.i.i.i4073, %5375
  %xor.i.i.i.i.i4075 = xor i32 %shl.i.i52.i.i.i4074, -1
  %5395 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop5936, !amdgpu.uniform !42

5396:                                             ; preds = %ComputeEnd5937
  %sunkaddr7492 = getelementptr inbounds i8, ptr addrspace(1) %8642, i64 40
  %5397 = atomicrmw and ptr addrspace(1) %sunkaddr7492, i32 %.lcssa8853 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %5398, !amdgpu.uniform !42

5398:                                             ; preds = %5396, %ComputeEnd5937
  call void @llvm.amdgcn.end.cf.i64(i64 %8669)
  call void @llvm.amdgcn.wave.barrier()
  br label %Flow7759, !amdgpu.uniform !42

if.end14.i.i.i4135:                               ; preds = %Flow7759
  %5399 = call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %5374, i1 true)
  %iszero.i.i.i.i.i.i4136 = icmp ne i64 %5374, 0
  %cmp2.i.i.i.i.i4137 = icmp eq i64 %5399, %sh_prom.i.i.i.i3270
  %cmp.i.i.i.i.i4138 = select i1 %iszero.i.i.i.i.i.i4136, i1 %cmp2.i.i.i.i.i4137, i1 false
  %5400 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.i.i.i.i.i4138)
  %5401 = extractvalue { i1, i64 } %5400, 0
  %5402 = extractvalue { i1, i64 } %5400, 1
  br i1 %5401, label %if.then16.i.i.i4140, label %if.end22.i.i.i4139

if.then16.i.i.i4140:                              ; preds = %if.end14.i.i.i4135
  %5403 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !76
  %opcode.i.i.i4141 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %5403, i64 %idxprom.i.i.i.i4065, i32 1
  store i32 1660944387, ptr %opcode.i.i.i4141, align 8, !tbaa !77
  %arrayidx21.i.i.i4142 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %5403, i64 %idxprom.i.i.i.i4065
  store i64 %5374, ptr %arrayidx21.i.i.i4142, align 8, !tbaa !80
  br label %if.end22.i.i.i4139, !amdgpu.uniform !42

Flow7760:                                         ; preds = %if.end22.i.i.i4139, %Flow7759
  %5404 = phi i32 [ %5391, %if.end22.i.i.i4139 ], [ %retval.sroa.8.0.i.i.i4050, %Flow7759 ]
  %5405 = phi i64 [ %5374, %if.end22.i.i.i4139 ], [ %retval.sroa.2.0.i.i.i4049, %Flow7759 ]
  %5406 = phi i32 [ %spec.store.select.i.i.i4053, %if.end22.i.i.i4139 ], [ %retval.sroa.6.0.i.i.i4047, %Flow7759 ]
  br label %cleanup26.i.i.i4076, !amdgpu.uniform !42

if.end22.i.i.i4139:                               ; preds = %if.then16.i.i.i4140, %if.end14.i.i.i4135
  call void @llvm.amdgcn.end.cf.i64(i64 %5402)
  call void @llvm.amdgcn.wave.barrier()
  br label %Flow7760, !amdgpu.uniform !42

Flow7761:                                         ; preds = %cleanup26.i.i.i4076, %5380
  %5407 = phi i32 [ %5406, %cleanup26.i.i.i4076 ], [ %5370, %5380 ]
  %5408 = phi i64 [ %5405, %cleanup26.i.i.i4076 ], [ %5371, %5380 ]
  %5409 = phi i32 [ %5404, %cleanup26.i.i.i4076 ], [ %5372, %5380 ]
  %5410 = phi i32 [ %5404, %cleanup26.i.i.i4076 ], [ %retval.sroa.8.0.i.i.i4050, %5380 ]
  %5411 = phi i64 [ %5405, %cleanup26.i.i.i4076 ], [ %retval.sroa.2.0.i.i.i4049, %5380 ]
  %5412 = phi i32 [ %5406, %cleanup26.i.i.i4076 ], [ %retval.sroa.6.0.i.i.i4047, %5380 ]
  %5413 = phi i1 [ %cmp.i46.not.i.i.i4068, %cleanup26.i.i.i4076 ], [ true, %5380 ]
  br i1 %5413, label %for.inc.i.i.i4081, label %Flow7762, !amdgpu.uniform !42

cleanup26.i.i.i4076:                              ; preds = %Flow7760
  br label %Flow7761, !amdgpu.uniform !42

for.inc.i.i.i4081:                                ; preds = %Flow7761
  %inc.i.i.i4086 = add i32 %spec.store.select.i.i.i4053, 1
  br label %Flow7762, !amdgpu.uniform !42

Flow7762:                                         ; preds = %for.inc.i.i.i4081, %Flow7761
  %5414 = phi i32 [ %inc.i.i.i4086, %for.inc.i.i.i4081 ], [ poison, %Flow7761 ]
  %5415 = phi i1 [ false, %for.inc.i.i.i4081 ], [ true, %Flow7761 ]
  br i1 %5415, label %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i4087, label %for.cond.i.i1.i4046, !amdgpu.uniform !42

_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i4087: ; preds = %Flow7762
  %.lcssa8856 = phi i32 [ %5407, %Flow7762 ]
  %.lcssa8855 = phi i64 [ %5408, %Flow7762 ]
  %.lcssa8854 = phi i32 [ %5409, %Flow7762 ]
  fence acquire
  %5416 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %mul.i.i.i.i.i.i4088 = shl i32 %.lcssa8856, 6
  %idxprom.i19.i.i.i.i.i4089 = zext i32 %mul.i.i.i.i.i.i4088 to i64
  %arrayidx.i20.i.i.i.i.i4090 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %5416, i64 %idxprom.i19.i.i.i.i.i4089
  %arrayidx.i22.i.i.i.i.i4091 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %arrayidx.i20.i.i.i.i.i4090, i64 %sh_prom.i.i.i.i3270
  store i64 sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.28.44 to ptr), i64 7) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.28.44 to ptr) to i64)), ptr %arrayidx.i22.i.i.i.i.i4091, align 8, !tbaa !82
  %spec.select.i.i.i.i.i.i.i4092 = call i64 @llvm.umin.i64(i64 sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.28.44 to ptr), i64 7) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.28.44 to ptr) to i64)), i64 56)
  %5417 = and i64 %spec.select.i.i.i.i.i.i.i4092, 15
  %5418 = sub i64 %spec.select.i.i.i.i.i.i.i4092, %5417
  %5419 = icmp ne i64 %5418, 0
  br i1 %5419, label %loop-memcpy-expansion5433.preheader, label %Flow7758, !amdgpu.uniform !42

loop-memcpy-expansion5433.preheader:              ; preds = %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i4087
  %5420 = add nuw nsw i64 %4513, 8
  %scevgep6381 = getelementptr i8, ptr %5416, i64 %5420
  %5421 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i4089, 6
  %scevgep6382 = getelementptr i8, ptr %scevgep6381, i64 %5421
  br label %loop-memcpy-expansion5433, !amdgpu.uniform !42

Flow7758:                                         ; preds = %Flow7757, %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i4087
  br label %loop-memcpy-residual-header5436, !amdgpu.uniform !42

loop-memcpy-expansion5433:                        ; preds = %loop-memcpy-expansion5433.preheader, %loop-memcpy-expansion5433
  %loop-index5434 = phi i64 [ %5424, %loop-memcpy-expansion5433 ], [ 0, %loop-memcpy-expansion5433.preheader ]
  %5422 = getelementptr i8, ptr addrspace(4) @.str.28.44, i64 %loop-index5434, !amdgpu.uniform !42
  %5423 = load <4 x i32>, ptr addrspace(4) %5422, align 1
  %scevgep6383 = getelementptr i8, ptr %scevgep6382, i64 %loop-index5434
  store <4 x i32> %5423, ptr %scevgep6383, align 1
  %5424 = add i64 %loop-index5434, 16
  %5425 = icmp uge i64 %5424, %5418
  br i1 %5425, label %Flow7757, label %loop-memcpy-expansion5433, !amdgpu.uniform !42

Flow7756:                                         ; preds = %Flow7754, %Flow7755
  %5426 = phi i64 [ %6667, %Flow7754 ], [ %6664, %Flow7755 ]
  br label %post-loop-memcpy-expansion5432, !amdgpu.uniform !42

loop-memcpy-residual5435:                         ; preds = %loop-memcpy-residual5435.preheader, %loop-memcpy-residual5435
  %residual-loop-index5437 = phi i64 [ %5428, %loop-memcpy-residual5435 ], [ 0, %loop-memcpy-residual5435.preheader ]
  %scevgep6380 = getelementptr i8, ptr addrspace(4) @.str.28.44, i64 %residual-loop-index5437, !amdgpu.uniform !42
  %5427 = load i8, ptr addrspace(4) %scevgep6380, align 1
  %scevgep6379 = getelementptr i8, ptr %scevgep6378, i64 %residual-loop-index5437
  store i8 %5427, ptr %scevgep6379, align 1
  %5428 = add i64 %residual-loop-index5437, 1
  %5429 = icmp uge i64 %5428, %5417
  br i1 %5429, label %Flow7754, label %loop-memcpy-residual5435, !amdgpu.uniform !42

Flow7754:                                         ; preds = %loop-memcpy-residual5435
  br label %Flow7756, !amdgpu.uniform !42

post-loop-memcpy-expansion5432:                   ; preds = %Flow7756
  %tobool.not.i.i.i.i.i.i4094 = icmp eq i32 %.lcssa8854, 0
  %conv.i.i.i.i.i.i4095 = zext i1 %tobool.not.i.i.i.i.i.i4094 to i32
  fence release
  %5430 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %idxprom.i23.i.i.i.i.i4096 = zext i32 %.lcssa8856 to i64
  %arrayidx.i24.i.i.i.i.i4097 = getelementptr inbounds nuw i32, ptr %5430, i64 %idxprom.i23.i.i.i.i.i4096
  store atomic i32 %conv.i.i.i.i.i.i4095, ptr %arrayidx.i24.i.i.i.i.i4097 monotonic, align 4
  %5431 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !84
  %arrayidx.i.i5.i.i4098 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %5431, i64 %idxprom.i23.i.i.i.i.i4096, !amdgpu.uniform !42
  %5432 = load i64, ptr %arrayidx.i.i5.i.i4098, align 8, !tbaa !80
  br label %while.cond.i.i.i.i4099, !amdgpu.uniform !42

while.cond.i.i.i.i4099:                           ; preds = %post-loop-memcpy-expansion5432, %Flow7753
  %phi.broken8806 = phi i64 [ 0, %post-loop-memcpy-expansion5432 ], [ %5449, %Flow7753 ]
  %lsr.iv6370 = phi i64 [ %5447, %Flow7753 ], [ -49, %post-loop-memcpy-expansion5432 ]
  %lsr.iv6365 = phi ptr addrspace(4) [ %5446, %Flow7753 ], [ getelementptr (i8, ptr addrspace(4) @.str.28.44, i64 56), %post-loop-memcpy-expansion5432 ]
  %port.sroa.37.0.i.i4100 = phi i32 [ %5445, %Flow7753 ], [ %conv.i.i.i.i.i.i4095, %post-loop-memcpy-expansion5432 ]
  %idx.0.i.i.i.i4101 = phi i64 [ %5444, %Flow7753 ], [ 56, %post-loop-memcpy-expansion5432 ]
  %cmp.i.i.i.i4102 = icmp ult i64 %idx.0.i.i.i.i4101, sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.28.44 to ptr), i64 7) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.28.44 to ptr) to i64))
  %5433 = call i64 @llvm.amdgcn.ballot.i64(i1 %cmp.i.i.i.i4102)
  %and.i.i.i.i6.i.i4103 = and i64 %5433, %5432
  %tobool.not.i.i.i.i4104 = icmp ne i64 %and.i.i.i.i6.i.i4103, 0
  %5434 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i22.i.i.i4105 = getelementptr inbounds nuw i32, ptr %5434, i64 %idxprom.i23.i.i.i.i.i4096, !amdgpu.uniform !42
  %5435 = load atomic i32, ptr %arrayidx.i22.i.i.i4105 monotonic, align 4
  %5436 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %tobool.not.i.i.i.i4104)
  %5437 = extractvalue { i1, i64 } %5436, 0
  %5438 = extractvalue { i1, i64 } %5436, 1
  br i1 %5437, label %cond.false.i8.i.i.i.i4106, label %Flow7753

cond.false.i8.i.i.i.i4106:                        ; preds = %while.cond.i.i.i.i4099
  %5439 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %5435)
  %5440 = icmp ne i32 %5439, %port.sroa.37.0.i.i4100
  br i1 %5440, label %while.body.i.i21.i.i.i.i4107.preheader, label %Flow7752, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i4107.preheader:           ; preds = %cond.false.i8.i.i.i.i4106
  br label %while.body.i.i21.i.i.i.i4107, !amdgpu.uniform !42

Flow7752:                                         ; preds = %Flow7751, %cond.false.i8.i.i.i.i4106
  br label %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i4110, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i4107:                     ; preds = %while.body.i.i21.i.i.i.i4107.preheader, %while.body.i.i21.i.i.i.i4107
  call void @llvm.amdgcn.s.sleep(i32 2)
  %5441 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i22.i.i.i.i4108 = getelementptr inbounds nuw i32, ptr %5441, i64 %idxprom.i23.i.i.i.i.i4096, !amdgpu.uniform !42
  %5442 = load atomic i32, ptr %arrayidx.i.i.i22.i.i.i.i4108 monotonic, align 4
  %5443 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %5442)
  %cmp.i.not.i.i23.i.i.i.i4109 = icmp eq i32 %5443, %port.sroa.37.0.i.i4100
  br i1 %cmp.i.not.i.i23.i.i.i.i4109, label %Flow7751, label %while.body.i.i21.i.i.i.i4107, !amdgpu.uniform !42

Flow7751:                                         ; preds = %while.body.i.i21.i.i.i.i4107
  br label %Flow7752, !amdgpu.uniform !42

Flow7753:                                         ; preds = %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i4111, %while.cond.i.i.i.i4099
  %5444 = phi i64 [ %add.i.i.i.i4114, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i4111 ], [ poison, %while.cond.i.i.i.i4099 ]
  %5445 = phi i32 [ %conv.i.i26.i.i.i.i4112, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i4111 ], [ poison, %while.cond.i.i.i.i4099 ]
  %5446 = phi ptr addrspace(4) [ %scevgep6366, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i4111 ], [ poison, %while.cond.i.i.i.i4099 ]
  %5447 = phi i64 [ %lsr.iv.next6371, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i4111 ], [ poison, %while.cond.i.i.i.i4099 ]
  %5448 = phi i1 [ false, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i4111 ], [ true, %while.cond.i.i.i.i4099 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %5438)
  %5449 = call i64 @llvm.amdgcn.if.break.i64(i1 %5448, i64 %phi.broken8806)
  %5450 = call i1 @llvm.amdgcn.loop.i64(i64 %5449)
  br i1 %5450, label %cond.false.i.i.i4121, label %while.cond.i.i.i.i4099

_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i4110: ; preds = %Flow7752
  fence acquire
  br i1 %cmp.i.i.i.i4102, label %if.then.i.i.i.i.i.i.i4115, label %Flow7750, !amdgpu.uniform !42

if.then.i.i.i.i.i.i.i4115:                        ; preds = %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i4110
  %sub.i.i.i.i.i.i.i4117 = sub nuw i64 sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.28.44 to ptr), i64 7) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.28.44 to ptr) to i64)), %idx.0.i.i.i.i4101
  %spec.select.i.i.i36.i.i.i.i4118 = call i64 @llvm.umin.i64(i64 %sub.i.i.i.i.i.i.i4117, i64 64)
  %5451 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %5452 = and i64 %spec.select.i.i.i36.i.i.i.i4118, 15
  %5453 = sub i64 %spec.select.i.i.i36.i.i.i.i4118, %5452
  %5454 = icmp ne i64 %5453, 0
  br i1 %5454, label %loop-memcpy-expansion5439.preheader, label %Flow7749, !amdgpu.uniform !42

loop-memcpy-expansion5439.preheader:              ; preds = %if.then.i.i.i.i.i.i.i4115
  %arrayidx.i20.i35.i.i.i.i4119 = getelementptr %"struct.rpc::Buffer", ptr %5451, i64 %idxprom.i19.i.i.i.i.i4089
  %arrayidx.i22.i38.i.i.i.i4120 = getelementptr %"struct.rpc::Buffer", ptr %arrayidx.i20.i35.i.i.i.i4119, i64 %sh_prom.i.i.i.i3270
  br label %loop-memcpy-expansion5439, !amdgpu.uniform !42

Flow7749:                                         ; preds = %Flow7748, %if.then.i.i.i.i.i.i.i4115
  br label %loop-memcpy-residual-header5442, !amdgpu.uniform !42

loop-memcpy-expansion5439:                        ; preds = %loop-memcpy-expansion5439.preheader, %loop-memcpy-expansion5439
  %loop-index5440 = phi i64 [ %5457, %loop-memcpy-expansion5439 ], [ 0, %loop-memcpy-expansion5439.preheader ]
  %scevgep6367 = getelementptr i8, ptr addrspace(4) %lsr.iv6365, i64 %loop-index5440, !amdgpu.uniform !42
  %5455 = load <4 x i32>, ptr addrspace(4) %scevgep6367, align 1
  %5456 = getelementptr i8, ptr %arrayidx.i22.i38.i.i.i.i4120, i64 %loop-index5440
  store <4 x i32> %5455, ptr %5456, align 1
  %5457 = add i64 %loop-index5440, 16
  %5458 = icmp uge i64 %5457, %5453
  br i1 %5458, label %Flow7748, label %loop-memcpy-expansion5439, !amdgpu.uniform !42

Flow7747:                                         ; preds = %Flow7746, %loop-memcpy-residual-header5442
  br label %Flow7750, !amdgpu.uniform !42

loop-memcpy-residual5441:                         ; preds = %loop-memcpy-residual5441.preheader, %loop-memcpy-residual5441
  %residual-loop-index5443 = phi i64 [ %5460, %loop-memcpy-residual5441 ], [ 0, %loop-memcpy-residual5441.preheader ]
  %scevgep6376 = getelementptr i8, ptr addrspace(4) %scevgep6375, i64 %residual-loop-index5443, !amdgpu.uniform !42
  %5459 = load i8, ptr addrspace(4) %scevgep6376, align 1
  %scevgep6374 = getelementptr i8, ptr %scevgep6373, i64 %residual-loop-index5443
  store i8 %5459, ptr %scevgep6374, align 1
  %5460 = add i64 %residual-loop-index5443, 1
  %5461 = icmp uge i64 %5460, %5452
  br i1 %5461, label %Flow7746, label %loop-memcpy-residual5441, !amdgpu.uniform !42

Flow7746:                                         ; preds = %loop-memcpy-residual5441
  br label %Flow7747, !amdgpu.uniform !42

_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i4111: ; preds = %Flow7750
  %conv.i.i26.i.i.i.i4112 = xor i32 %port.sroa.37.0.i.i4100, 1
  fence release
  %5462 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %arrayidx.i24.i29.i.i.i.i4113 = getelementptr inbounds nuw i32, ptr %5462, i64 %idxprom.i23.i.i.i.i.i4096
  store atomic i32 %conv.i.i26.i.i.i.i4112, ptr %arrayidx.i24.i29.i.i.i.i4113 monotonic, align 4
  %add.i.i.i.i4114 = add i64 %idx.0.i.i.i.i4101, 64
  %scevgep6366 = getelementptr i8, ptr addrspace(4) %lsr.iv6365, i64 64
  %lsr.iv.next6371 = add i64 %lsr.iv6370, -64
  br label %Flow7753, !amdgpu.uniform !42

cond.false.i.i.i4121:                             ; preds = %Flow7753
  %.lcssa8850 = phi i64 [ %5449, %Flow7753 ]
  %port.sroa.37.0.i.i4100.lcssa = phi i32 [ %port.sroa.37.0.i.i4100, %Flow7753 ]
  %.lcssa8849 = phi i32 [ %5435, %Flow7753 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa8850)
  %5463 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %.lcssa8849)
  %5464 = icmp ne i32 %5463, %port.sroa.37.0.i.i4100.lcssa
  %5465 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %5464)
  %5466 = extractvalue { i1, i64 } %5465, 0
  %5467 = extractvalue { i1, i64 } %5465, 1
  br i1 %5466, label %while.body.i.i.i.i4122.preheader, label %Flow7745

while.body.i.i.i.i4122.preheader:                 ; preds = %cond.false.i.i.i4121
  br label %while.body.i.i.i.i4122, !amdgpu.uniform !42

Flow7745:                                         ; preds = %Flow7744, %cond.false.i.i.i4121
  call void @llvm.amdgcn.end.cf.i64(i64 %5467)
  br label %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit4143, !amdgpu.uniform !42

while.body.i.i.i.i4122:                           ; preds = %while.body.i.i.i.i4122.preheader, %while.body.i.i.i.i4122
  %phi.broken8807 = phi i64 [ 0, %while.body.i.i.i.i4122.preheader ], [ %5471, %while.body.i.i.i.i4122 ]
  call void @llvm.amdgcn.s.sleep(i32 2)
  %5468 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i8.i.i4123 = getelementptr inbounds nuw i32, ptr %5468, i64 %idxprom.i23.i.i.i.i.i4096, !amdgpu.uniform !42
  %5469 = load atomic i32, ptr %arrayidx.i.i.i8.i.i4123 monotonic, align 4
  %5470 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %5469)
  %cmp.i.not.i.i.i.i4124 = icmp eq i32 %5470, %port.sroa.37.0.i.i4100.lcssa
  %5471 = call i64 @llvm.amdgcn.if.break.i64(i1 %cmp.i.not.i.i.i.i4124, i64 %phi.broken8807)
  %5472 = call i1 @llvm.amdgcn.loop.i64(i64 %5471)
  br i1 %5472, label %Flow7744, label %while.body.i.i.i.i4122

Flow7744:                                         ; preds = %while.body.i.i.i.i4122
  %.lcssa8848 = phi i64 [ %5471, %while.body.i.i.i.i4122 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa8848)
  br label %Flow7745, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit4143: ; preds = %Flow7745
  fence acquire
  call void @llvm.amdgcn.wave.barrier()
  fence syncscope("agent") release
  %5473 = call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %.lcssa8855, i1 true)
  %iszero.i.i.i.i.i13.i.i4126 = icmp ne i64 %.lcssa8855, 0
  %cmp2.i.i.i.i14.i.i4127 = icmp eq i64 %5473, %sh_prom.i.i.i.i3270
  %cmp.i.i.i.i15.i.i4128 = select i1 %iszero.i.i.i.i.i13.i.i4126, i1 %cmp2.i.i.i.i14.i.i4127, i1 false
  %5474 = and i32 %.lcssa8856, 31
  %conv4.i.i.i19.i.i4132 = zext i1 %cmp.i.i.i.i15.i.i4128 to i32
  %shl.i.i.i20.i.i4133 = shl nuw i32 %conv4.i.i.i19.i.i4132, %5474
  %xor.i.i.i21.i.i4134 = xor i32 %shl.i.i.i20.i.i4133, -1
  %5475 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop5940, !amdgpu.uniform !42

5476:                                             ; preds = %ComputeEnd5941
  %div8.i.i.i16.i.i4129 = lshr i32 %.lcssa8856, 5
  %idxprom.i.i.i17.i.i4130 = zext nneg i32 %div8.i.i.i16.i.i4129 to i64
  %5477 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i17.i.i4130
  %sunkaddr7493 = getelementptr inbounds i8, ptr addrspace(1) %5477, i64 40
  %5478 = atomicrmw and ptr addrspace(1) %sunkaddr7493, i32 %.lcssa8847 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %5479, !amdgpu.uniform !42

5479:                                             ; preds = %5476, %ComputeEnd5941
  call void @llvm.amdgcn.end.cf.i64(i64 %8686)
  call void @llvm.amdgcn.wave.barrier()
  call void @llvm.lifetime.start.p5(ptr addrspace(5) %buffer.i.i.i2988) #34, !noalias !190
  %cmp7.i.i.i.i.i.i4145 = icmp ne i32 %1149, 0
  %5480 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp7.i.i.i.i.i.i4145)
  %5481 = extractvalue { i1, i64 } %5480, 0
  %5482 = extractvalue { i1, i64 } %5480, 1
  br i1 %5481, label %while.body.lr.ph.i.i.i.i.i.i4146, label %Flow7743

while.body.lr.ph.i.i.i.i.i.i4146:                 ; preds = %5479
  %retval.0.i.i.i.i.i.i4147 = call noundef i32 @llvm.abs.i32(i32 %1149, i1 false)
  br label %while.body.i.i.i.i.i.i4148, !amdgpu.uniform !42

Flow7743:                                         ; preds = %Flow7742, %5479
  %5483 = phi i1 [ %cmp.i.i.i.i.i.i.i.i4158.lcssa, %Flow7742 ], [ false, %5479 ]
  %5484 = phi i64 [ %writer.sroa.10.1.i.i.i.i4163.lcssa, %Flow7742 ], [ 0, %5479 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %5482)
  br label %_ZN22__llvm_libc_22_0_0_git15IntegerToStringIiNS_7details3FmtILh10ELb0ELb0ELb0ELm1EEEE13IntegerWriter25write_unsigned_number_decEjRNS1_22StringBufferWriterImplILb0EEE.exit.i.i.i.i.i4166, !amdgpu.uniform !42

while.body.i.i.i.i.i.i4148:                       ; preds = %while.body.lr.ph.i.i.i.i.i.i4146, %_ZN22__llvm_libc_22_0_0_git7details22StringBufferWriterImplILb0EE4pushEc.exit.i.i.i.i.i.i4162
  %phi.broken8808 = phi i64 [ 0, %while.body.lr.ph.i.i.i.i.i.i4146 ], [ %5552, %_ZN22__llvm_libc_22_0_0_git7details22StringBufferWriterImplILb0EE4pushEc.exit.i.i.i.i.i.i4162 ]
  %writer.sroa.10.0.i.i.i.i4149 = phi i64 [ %writer.sroa.10.1.i.i.i.i4163, %_ZN22__llvm_libc_22_0_0_git7details22StringBufferWriterImplILb0EE4pushEc.exit.i.i.i.i.i.i4162 ], [ 0, %while.body.lr.ph.i.i.i.i.i.i4146 ]
  %value.addr.09.i.i.i.i.i.i4150 = phi i32 [ %div.i.i.i.i.i.i.i4151, %_ZN22__llvm_libc_22_0_0_git7details22StringBufferWriterImplILb0EE4pushEc.exit.i.i.i.i.i.i4162 ], [ %retval.0.i.i.i.i.i.i4147, %while.body.lr.ph.i.i.i.i.i.i4146 ]
  br label %NodeBlock7631, !amdgpu.uniform !42

NodeBlock7631:                                    ; preds = %while.body.i.i.i.i.i.i4148
  %div.i.i.i.i.i.i.i4151 = udiv i32 %value.addr.09.i.i.i.i.i.i4150, 10
  %.neg.i4152 = mul i32 %div.i.i.i.i.i.i.i4151, 246
  %rem.i.i.i.i.i.i.decomposed.i4153 = add i32 %.neg.i4152, %value.addr.09.i.i.i.i.i.i4150
  %conv.i.i.i.i.i.i.i4154 = trunc i32 %rem.i.i.i.i.i.i.decomposed.i4153 to i8
  %5485 = zext i8 %conv.i.i.i.i.i.i.i4154 to i16
  %Pivot7632 = icmp sge i16 %5485, 5
  %5486 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %Pivot7632)
  %5487 = extractvalue { i1, i64 } %5486, 0
  %5488 = extractvalue { i1, i64 } %5486, 1
  br i1 %5487, label %NodeBlock7629, label %Flow7741

NodeBlock7629:                                    ; preds = %NodeBlock7631
  %Pivot7630 = icmp sge i16 %5485, 7
  %5489 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %Pivot7630)
  %5490 = extractvalue { i1, i64 } %5489, 0
  %5491 = extractvalue { i1, i64 } %5489, 1
  br i1 %5490, label %NodeBlock7627, label %Flow7732

NodeBlock7627:                                    ; preds = %NodeBlock7629
  %Pivot7628 = icmp sge i16 %5485, 8
  %5492 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %Pivot7628)
  %5493 = extractvalue { i1, i64 } %5492, 0
  %5494 = extractvalue { i1, i64 } %5492, 1
  br i1 %5493, label %NodeBlock7625, label %Flow7728

NodeBlock7625:                                    ; preds = %NodeBlock7627
  %Pivot7626 = icmp sge i16 %5485, 9
  %5495 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %Pivot7626)
  %5496 = extractvalue { i1, i64 } %5495, 0
  %5497 = extractvalue { i1, i64 } %5495, 1
  br i1 %5496, label %sw.bb9.i.i.i.i.i.i.i.i4155, label %Flow7726

Flow7732:                                         ; preds = %Flow7729, %NodeBlock7629
  %5498 = phi i8 [ %5542, %Flow7729 ], [ poison, %NodeBlock7629 ]
  %5499 = call { i1, i64 } @llvm.amdgcn.else.i64.i64(i64 %5491)
  %5500 = extractvalue { i1, i64 } %5499, 0
  %5501 = extractvalue { i1, i64 } %5499, 1
  br i1 %5500, label %NodeBlock7623, label %Flow7733

NodeBlock7623:                                    ; preds = %Flow7732
  %Pivot7624 = icmp sge i16 %5485, 6
  %5502 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %Pivot7624)
  %5503 = extractvalue { i1, i64 } %5502, 0
  %5504 = extractvalue { i1, i64 } %5502, 1
  br i1 %5503, label %sw.bb6.i.i.i.i.i.i.i.i4315, label %Flow7730

Flow7741:                                         ; preds = %Flow7733, %NodeBlock7631
  %5505 = phi i8 [ %5544, %Flow7733 ], [ poison, %NodeBlock7631 ]
  %5506 = call { i1, i64 } @llvm.amdgcn.else.i64.i64(i64 %5488)
  %5507 = extractvalue { i1, i64 } %5506, 0
  %5508 = extractvalue { i1, i64 } %5506, 1
  br i1 %5507, label %NodeBlock7621, label %if.then.i.i.i.i.i.i.i4156

NodeBlock7621:                                    ; preds = %Flow7741
  %Pivot7622 = icmp sge i16 %5485, 3
  %5509 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %Pivot7622)
  %5510 = extractvalue { i1, i64 } %5509, 0
  %5511 = extractvalue { i1, i64 } %5509, 1
  br i1 %5510, label %NodeBlock7619, label %Flow7739

NodeBlock7619:                                    ; preds = %NodeBlock7621
  %Pivot7620 = icmp sge i16 %5485, 4
  %5512 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %Pivot7620)
  %5513 = extractvalue { i1, i64 } %5512, 0
  %5514 = extractvalue { i1, i64 } %5512, 1
  br i1 %5513, label %sw.bb4.i.i.i.i.i.i.i.i4317, label %Flow7734

Flow7739:                                         ; preds = %Flow7735, %NodeBlock7621
  %5515 = phi i8 [ %5545, %Flow7735 ], [ poison, %NodeBlock7621 ]
  %5516 = call { i1, i64 } @llvm.amdgcn.else.i64.i64(i64 %5511)
  %5517 = extractvalue { i1, i64 } %5516, 0
  %5518 = extractvalue { i1, i64 } %5516, 1
  br i1 %5517, label %NodeBlock7617, label %Flow7740

NodeBlock7617:                                    ; preds = %Flow7739
  %Pivot7618 = icmp sge i16 %5485, 2
  %5519 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %Pivot7618)
  %5520 = extractvalue { i1, i64 } %5519, 0
  %5521 = extractvalue { i1, i64 } %5519, 1
  br i1 %5520, label %sw.bb2.i.i.i.i.i.i.i.i4319, label %Flow7737

Flow7737:                                         ; preds = %sw.bb2.i.i.i.i.i.i.i.i4319, %NodeBlock7617
  %5522 = call { i1, i64 } @llvm.amdgcn.else.i64.i64(i64 %5521)
  %5523 = extractvalue { i1, i64 } %5522, 0
  %5524 = extractvalue { i1, i64 } %5522, 1
  br i1 %5523, label %LeafBlock7615, label %Flow7738

LeafBlock7615:                                    ; preds = %Flow7737
  %SwitchLeaf7616 = icmp eq i16 %5485, 1
  %5525 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %SwitchLeaf7616)
  %5526 = extractvalue { i1, i64 } %5525, 0
  %5527 = extractvalue { i1, i64 } %5525, 1
  br i1 %5526, label %sw.bb1.i.i.i.i.i.i.i.i4320, label %Flow7736

sw.bb1.i.i.i.i.i.i.i.i4320:                       ; preds = %LeafBlock7615
  br label %Flow7736, !amdgpu.uniform !42

sw.bb2.i.i.i.i.i.i.i.i4319:                       ; preds = %NodeBlock7617
  br label %Flow7737, !amdgpu.uniform !42

Flow7734:                                         ; preds = %sw.bb4.i.i.i.i.i.i.i.i4317, %NodeBlock7619
  %5528 = call { i1, i64 } @llvm.amdgcn.else.i64.i64(i64 %5514)
  %5529 = extractvalue { i1, i64 } %5528, 0
  %5530 = extractvalue { i1, i64 } %5528, 1
  br i1 %5529, label %sw.bb3.i.i.i.i.i.i.i.i4318, label %Flow7735

sw.bb3.i.i.i.i.i.i.i.i4318:                       ; preds = %Flow7734
  br label %Flow7735, !amdgpu.uniform !42

sw.bb4.i.i.i.i.i.i.i.i4317:                       ; preds = %NodeBlock7619
  br label %Flow7734, !amdgpu.uniform !42

Flow7730:                                         ; preds = %sw.bb6.i.i.i.i.i.i.i.i4315, %NodeBlock7623
  %5531 = call { i1, i64 } @llvm.amdgcn.else.i64.i64(i64 %5504)
  %5532 = extractvalue { i1, i64 } %5531, 0
  %5533 = extractvalue { i1, i64 } %5531, 1
  br i1 %5532, label %sw.bb5.i.i.i.i.i.i.i.i4316, label %Flow7731

sw.bb5.i.i.i.i.i.i.i.i4316:                       ; preds = %Flow7730
  br label %Flow7731, !amdgpu.uniform !42

sw.bb6.i.i.i.i.i.i.i.i4315:                       ; preds = %NodeBlock7623
  br label %Flow7730, !amdgpu.uniform !42

Flow7728:                                         ; preds = %Flow7727, %NodeBlock7627
  %5534 = phi i8 [ %5541, %Flow7727 ], [ poison, %NodeBlock7627 ]
  %5535 = call { i1, i64 } @llvm.amdgcn.else.i64.i64(i64 %5494)
  %5536 = extractvalue { i1, i64 } %5535, 0
  %5537 = extractvalue { i1, i64 } %5535, 1
  br i1 %5536, label %sw.bb7.i.i.i.i.i.i.i.i4314, label %Flow7729

sw.bb7.i.i.i.i.i.i.i.i4314:                       ; preds = %Flow7728
  br label %Flow7729, !amdgpu.uniform !42

Flow7726:                                         ; preds = %sw.bb9.i.i.i.i.i.i.i.i4155, %NodeBlock7625
  %5538 = call { i1, i64 } @llvm.amdgcn.else.i64.i64(i64 %5497)
  %5539 = extractvalue { i1, i64 } %5538, 0
  %5540 = extractvalue { i1, i64 } %5538, 1
  br i1 %5539, label %sw.bb8.i.i.i.i.i.i.i.i4313, label %Flow7727

sw.bb8.i.i.i.i.i.i.i.i4313:                       ; preds = %Flow7726
  br label %Flow7727, !amdgpu.uniform !42

sw.bb9.i.i.i.i.i.i.i.i4155:                       ; preds = %NodeBlock7625
  br label %Flow7726, !amdgpu.uniform !42

Flow7727:                                         ; preds = %sw.bb8.i.i.i.i.i.i.i.i4313, %Flow7726
  %5541 = phi i8 [ 56, %sw.bb8.i.i.i.i.i.i.i.i4313 ], [ 57, %Flow7726 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %5540)
  br label %Flow7728, !amdgpu.uniform !42

Flow7729:                                         ; preds = %sw.bb7.i.i.i.i.i.i.i.i4314, %Flow7728
  %5542 = phi i8 [ 55, %sw.bb7.i.i.i.i.i.i.i.i4314 ], [ %5534, %Flow7728 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %5537)
  br label %Flow7732, !amdgpu.uniform !42

Flow7731:                                         ; preds = %sw.bb5.i.i.i.i.i.i.i.i4316, %Flow7730
  %5543 = phi i8 [ 53, %sw.bb5.i.i.i.i.i.i.i.i4316 ], [ 54, %Flow7730 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %5533)
  br label %Flow7733, !amdgpu.uniform !42

Flow7733:                                         ; preds = %Flow7731, %Flow7732
  %5544 = phi i8 [ %5543, %Flow7731 ], [ %5498, %Flow7732 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %5501)
  br label %Flow7741, !amdgpu.uniform !42

Flow7735:                                         ; preds = %sw.bb3.i.i.i.i.i.i.i.i4318, %Flow7734
  %5545 = phi i8 [ 51, %sw.bb3.i.i.i.i.i.i.i.i4318 ], [ 52, %Flow7734 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %5530)
  br label %Flow7739, !amdgpu.uniform !42

Flow7736:                                         ; preds = %sw.bb1.i.i.i.i.i.i.i.i4320, %LeafBlock7615
  %5546 = phi i8 [ 49, %sw.bb1.i.i.i.i.i.i.i.i4320 ], [ 48, %LeafBlock7615 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %5527)
  br label %Flow7738, !amdgpu.uniform !42

Flow7738:                                         ; preds = %Flow7736, %Flow7737
  %5547 = phi i8 [ %5546, %Flow7736 ], [ 50, %Flow7737 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %5524)
  br label %Flow7740, !amdgpu.uniform !42

Flow7740:                                         ; preds = %Flow7738, %Flow7739
  %5548 = phi i8 [ %5547, %Flow7738 ], [ %5515, %Flow7739 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %5518)
  br label %if.then.i.i.i.i.i.i.i4156, !amdgpu.uniform !42

if.then.i.i.i.i.i.i.i4156:                        ; preds = %Flow7741, %Flow7740
  %retval.0.i.i.i.i.i.i.i.i4157 = phi i8 [ %5505, %Flow7741 ], [ %5548, %Flow7740 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %5508)
  %cmp.i.i.i.i.i.i.i.i4158 = icmp eq i64 %writer.sroa.10.0.i.i.i.i4149, 11
  %cmp.i.i.i.i.i.i.i.i4158.inv = xor i1 %cmp.i.i.i.i.i.i.i.i4158, true
  br i1 %cmp.i.i.i.i.i.i.i.i4158.inv, label %if.then3.i.i.i.i.i.i.i4159, label %_ZN22__llvm_libc_22_0_0_git7details22StringBufferWriterImplILb0EE4pushEc.exit.i.i.i.i.i.i4162, !amdgpu.uniform !42

if.then3.i.i.i.i.i.i.i4159:                       ; preds = %if.then.i.i.i.i.i.i.i4156
  %5549 = trunc i64 %writer.sroa.10.0.i.i.i.i4149 to i32
  %5550 = xor i32 %5549, -1
  %5551 = getelementptr i8, ptr addrspace(5) %buffer.i.i.i2988, i32 %5550
  %arrayidx.i.i.i.i.i.i.i.i41606114 = getelementptr i8, ptr addrspace(5) %5551, i32 11
  %inc.i.i.i.i.i.i.i4161 = add i64 %writer.sroa.10.0.i.i.i.i4149, 1
  store i8 %retval.0.i.i.i.i.i.i.i.i4157, ptr addrspace(5) %arrayidx.i.i.i.i.i.i.i.i41606114, align 1, !tbaa !57, !noalias !190
  br label %_ZN22__llvm_libc_22_0_0_git7details22StringBufferWriterImplILb0EE4pushEc.exit.i.i.i.i.i.i4162, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git7details22StringBufferWriterImplILb0EE4pushEc.exit.i.i.i.i.i.i4162: ; preds = %if.then3.i.i.i.i.i.i.i4159, %if.then.i.i.i.i.i.i.i4156
  %writer.sroa.10.1.i.i.i.i4163 = phi i64 [ %inc.i.i.i.i.i.i.i4161, %if.then3.i.i.i.i.i.i.i4159 ], [ 11, %if.then.i.i.i.i.i.i.i4156 ]
  %cmp.i.i.i.i.i.i4164 = icmp ult i32 %value.addr.09.i.i.i.i.i.i4150, 10
  %.not.i.i.i.i.i.i4165 = or i1 %cmp.i.i.i.i.i.i.i.i4158, %cmp.i.i.i.i.i.i4164
  %5552 = call i64 @llvm.amdgcn.if.break.i64(i1 %.not.i.i.i.i.i.i4165, i64 %phi.broken8808)
  %5553 = call i1 @llvm.amdgcn.loop.i64(i64 %5552)
  br i1 %5553, label %Flow7742, label %while.body.i.i.i.i.i.i4148

Flow7742:                                         ; preds = %_ZN22__llvm_libc_22_0_0_git7details22StringBufferWriterImplILb0EE4pushEc.exit.i.i.i.i.i.i4162
  %writer.sroa.10.1.i.i.i.i4163.lcssa = phi i64 [ %writer.sroa.10.1.i.i.i.i4163, %_ZN22__llvm_libc_22_0_0_git7details22StringBufferWriterImplILb0EE4pushEc.exit.i.i.i.i.i.i4162 ]
  %.lcssa8846 = phi i64 [ %5552, %_ZN22__llvm_libc_22_0_0_git7details22StringBufferWriterImplILb0EE4pushEc.exit.i.i.i.i.i.i4162 ]
  %cmp.i.i.i.i.i.i.i.i4158.lcssa = phi i1 [ %cmp.i.i.i.i.i.i.i.i4158, %_ZN22__llvm_libc_22_0_0_git7details22StringBufferWriterImplILb0EE4pushEc.exit.i.i.i.i.i.i4162 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa8846)
  br label %Flow7743, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git15IntegerToStringIiNS_7details3FmtILh10ELb0ELb0ELb0ELm1EEEE13IntegerWriter25write_unsigned_number_decEjRNS1_22StringBufferWriterImplILb0EEE.exit.i.i.i.i.i4166: ; preds = %Flow7743
  %cmp.i19.i.i.i.i4169 = icmp ne i64 %5484, 0
  %or.cond.not.i.i.i.i4170 = select i1 %5483, i1 true, i1 %cmp.i19.i.i.i.i4169
  %or.cond.not.i.i.i.i4170.inv = xor i1 %or.cond.not.i.i.i.i4170, true
  %5554 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %or.cond.not.i.i.i.i4170.inv)
  %5555 = extractvalue { i1, i64 } %5554, 0
  %5556 = extractvalue { i1, i64 } %5554, 1
  br i1 %5555, label %if.then3.i.i.lr.ph.i.i.i.i4171, label %while.end.i.i.i.i.i4173

if.then3.i.i.lr.ph.i.i.i.i4171:                   ; preds = %_ZN22__llvm_libc_22_0_0_git15IntegerToStringIiNS_7details3FmtILh10ELb0ELb0ELb0ELm1EEEE13IntegerWriter25write_unsigned_number_decEjRNS1_22StringBufferWriterImplILb0EEE.exit.i.i.i.i.i4166
  %arrayidx.i.i.i.i.i.i.i4172 = getelementptr inbounds nuw i8, ptr addrspace(5) %buffer.i.i.i2988, i32 10
  store i8 48, ptr addrspace(5) %arrayidx.i.i.i.i.i.i.i4172, align 2, !tbaa !57, !noalias !190
  br label %while.end.i.i.i.i.i4173, !amdgpu.uniform !42

while.end.i.i.i.i.i4173:                          ; preds = %if.then3.i.i.lr.ph.i.i.i.i4171, %_ZN22__llvm_libc_22_0_0_git15IntegerToStringIiNS_7details3FmtILh10ELb0ELb0ELb0ELm1EEEE13IntegerWriter25write_unsigned_number_decEjRNS1_22StringBufferWriterImplILb0EEE.exit.i.i.i.i.i4166
  %writer.sroa.10.4.i.i.i.i4174 = phi i64 [ 1, %if.then3.i.i.lr.ph.i.i.i.i4171 ], [ %5484, %_ZN22__llvm_libc_22_0_0_git15IntegerToStringIiNS_7details3FmtILh10ELb0ELb0ELb0ELm1EEEE13IntegerWriter25write_unsigned_number_decEjRNS1_22StringBufferWriterImplILb0EEE.exit.i.i.i.i.i4166 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %5556)
  %cmp4.i.i.i.i.i4175 = icmp sgt i32 %1149, -1
  %brmerge.i.i.i.i.i4176 = or i1 %cmp4.i.i.i.i.i4175, %5483
  %brmerge.i.i.i.i.i4176.inv = xor i1 %brmerge.i.i.i.i.i4176, true
  %5557 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %brmerge.i.i.i.i.i4176.inv)
  %5558 = extractvalue { i1, i64 } %5557, 0
  %5559 = extractvalue { i1, i64 } %5557, 1
  br i1 %5558, label %if.then.i14.i.i.i.i.i4177, label %_ZN22__llvm_libc_22_0_0_git15IntegerToStringIiNS_7details3FmtILh10ELb0ELb0ELb0ELm1EEEEC2Ei.exit.i.i.i4182

if.then.i14.i.i.i.i.i4177:                        ; preds = %while.end.i.i.i.i.i4173
  %cmp.i.i17.i.i.i.i.i4178 = icmp ne i64 %writer.sroa.10.4.i.i.i.i4174, 11
  %5560 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.i.i17.i.i.i.i.i4178)
  %5561 = extractvalue { i1, i64 } %5560, 0
  %5562 = extractvalue { i1, i64 } %5560, 1
  br i1 %5561, label %if.then3.i18.i.i.i.i.i4179, label %Flow7725

if.then3.i18.i.i.i.i.i4179:                       ; preds = %if.then.i14.i.i.i.i.i4177
  %5563 = trunc i64 %writer.sroa.10.4.i.i.i.i4174 to i32
  %5564 = xor i32 %5563, -1
  %5565 = getelementptr i8, ptr addrspace(5) %buffer.i.i.i2988, i32 %5564
  %arrayidx.i.i19.i.i.i.i.i41806115 = getelementptr i8, ptr addrspace(5) %5565, i32 11
  %inc.i20.i.i.i.i.i4181 = add i64 %writer.sroa.10.4.i.i.i.i4174, 1
  store i8 45, ptr addrspace(5) %arrayidx.i.i19.i.i.i.i.i41806115, align 1, !tbaa !57, !noalias !190
  br label %Flow7725, !amdgpu.uniform !42

Flow7725:                                         ; preds = %if.then3.i18.i.i.i.i.i4179, %if.then.i14.i.i.i.i.i4177
  %5566 = phi i64 [ %inc.i20.i.i.i.i.i4181, %if.then3.i18.i.i.i.i.i4179 ], [ 11, %if.then.i14.i.i.i.i.i4177 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %5562)
  br label %_ZN22__llvm_libc_22_0_0_git15IntegerToStringIiNS_7details3FmtILh10ELb0ELb0ELb0ELm1EEEEC2Ei.exit.i.i.i4182, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git15IntegerToStringIiNS_7details3FmtILh10ELb0ELb0ELb0ELm1EEEEC2Ei.exit.i.i.i4182: ; preds = %while.end.i.i.i.i.i4173, %Flow7725
  %writer.sroa.10.5.i.i.i.i4183 = phi i64 [ %writer.sroa.10.4.i.i.i.i4174, %while.end.i.i.i.i.i4173 ], [ %5566, %Flow7725 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %5559)
  %written.i.i.i.i4184 = getelementptr inbounds nuw i8, ptr addrspace(5) %buffer.i.i.i2988, i32 16
  store i64 %writer.sroa.10.5.i.i.i.i4183, ptr addrspace(5) %written.i.i.i.i4184, align 8, !tbaa !91, !noalias !190
  %cmp.i.not.i.i.i.i.i4185 = icmp ne i64 %writer.sroa.10.5.i.i.i.i4183, 0
  %5567 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.i.not.i.i.i.i.i4185)
  %5568 = extractvalue { i1, i64 } %5567, 0
  %5569 = extractvalue { i1, i64 } %5567, 1
  br i1 %5568, label %if.then.i.i.i.i.i.i4186, label %Flow7724

if.then.i.i.i.i.i.i4186:                          ; preds = %_ZN22__llvm_libc_22_0_0_git15IntegerToStringIiNS_7details3FmtILh10ELb0ELb0ELb0ELm1EEEEC2Ei.exit.i.i.i4182
  %5570 = call { i64, i1 } @llvm.uadd.with.overflow.i64(i64 %writer.sroa.10.5.i.i.i.i4183, i64 1)
  %ov7495 = extractvalue { i64, i1 } %5570, 1
  %ov7495.inv = xor i1 %ov7495, true
  %5571 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %ov7495.inv)
  %5572 = extractvalue { i1, i64 } %5571, 0
  %5573 = extractvalue { i1, i64 } %5571, 1
  br i1 %5572, label %if.end.i.i.i.i.i.i.i4189, label %Flow7723

if.end.i.i.i.i.i.i.i4189:                         ; preds = %if.then.i.i.i.i.i.i4186
  %math7494 = extractvalue { i64, i1 } %5570, 0
  %cmp2.i.i.i.i.i.i.i4190 = icmp ult i64 %math7494, 1676976733973595601
  %5574 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp2.i.i.i.i.i.i.i4190)
  %5575 = extractvalue { i1, i64 } %5574, 0
  %5576 = extractvalue { i1, i64 } %5574, 1
  br i1 %5575, label %if.then3.i.i.i.i4.i.i.i4310, label %if.end4.i.i.i.i.i.i.i4191

if.then3.i.i.i.i4.i.i.i4310:                      ; preds = %if.end.i.i.i.i.i.i.i4189
  %mul.i.i.i.i.i.i.i4311 = mul nuw i64 %math7494, 11
  %div18.i.i.i.i.i.i.i4312 = lshr i64 %mul.i.i.i.i.i.i.i4311, 3
  br label %if.end4.i.i.i.i.i.i.i4191, !amdgpu.uniform !42

Flow7723:                                         ; preds = %if.end4.i.i.i.i.i.i.i4191, %if.then.i.i.i.i.i.i4186
  %5577 = phi ptr [ %cond.i.i.i4199, %if.end4.i.i.i.i.i.i.i4191 ], [ addrspacecast (ptr addrspace(4) @_ZN22__llvm_libc_22_0_0_git3cpp6string14NULL_CHARACTERE to ptr), %if.then.i.i.i.i.i.i4186 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %5573)
  br label %_ZN22__llvm_libc_22_0_0_git3cpp6string7reserveEm.exit.i.i.i.i.i.i4201, !amdgpu.uniform !42

if.end4.i.i.i.i.i.i.i4191:                        ; preds = %if.then3.i.i.i.i4.i.i.i4310, %if.end.i.i.i.i.i.i.i4189
  %new_capacity.addr.0.i.i.i.i.i.i.i4192 = phi i64 [ %div18.i.i.i.i.i.i.i4312, %if.then3.i.i.i.i4.i.i.i4310 ], [ %math7494, %if.end.i.i.i.i.i.i.i4189 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %5576)
  %sub.i.i.i4193 = add i64 %new_capacity.addr.0.i.i.i.i.i.i.i4192, 7
  %div2.i.i.i4194 = and i64 %sub.i.i.i4193, -8
  %5578 = load ptr, ptr addrspace(1) @_ZN12_GLOBAL__N_13ptrE, align 8, !tbaa !55
  %add.ptr.i.i.i4195 = getelementptr inbounds nuw i8, ptr %5578, i64 %div2.i.i.i4194
  store ptr %add.ptr.i.i.i4195, ptr addrspace(1) @_ZN12_GLOBAL__N_13ptrE, align 8, !tbaa !55
  %sub.ptr.lhs.cast.i.i.i4196 = ptrtoint ptr %add.ptr.i.i.i4195 to i64
  %sub.ptr.sub.i.i.i4197 = sub i64 %sub.ptr.lhs.cast.i.i.i4196, ptrtoint (ptr addrspacecast (ptr addrspace(1) @_ZN12_GLOBAL__N_16memoryE to ptr) to i64)
  %cmp.i.i.i4198 = icmp ugt i64 %sub.ptr.sub.i.i.i4197, 65335
  %cond.i.i.i4199 = select i1 %cmp.i.i.i4198, ptr null, ptr %5578
  br label %Flow7723, !amdgpu.uniform !42

Flow7724:                                         ; preds = %Flow7722, %_ZN22__llvm_libc_22_0_0_git15IntegerToStringIiNS_7details3FmtILh10ELb0ELb0ELb0ELm1EEEEC2Ei.exit.i.i.i4182
  %5579 = phi ptr [ %5577, %Flow7722 ], [ addrspacecast (ptr addrspace(4) @_ZN22__llvm_libc_22_0_0_git3cpp6string14NULL_CHARACTERE to ptr), %_ZN22__llvm_libc_22_0_0_git15IntegerToStringIiNS_7details3FmtILh10ELb0ELb0ELb0ELm1EEEEC2Ei.exit.i.i.i4182 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %5569)
  br label %if.end.i.i.i.i.i.i4203, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git3cpp6string7reserveEm.exit.i.i.i.i.i.i4201: ; preds = %Flow7723
  br label %loadstoreloop, !amdgpu.uniform !42

loadstoreloop:                                    ; preds = %_ZN22__llvm_libc_22_0_0_git3cpp6string7reserveEm.exit.i.i.i.i.i.i4201, %loadstoreloop
  %phi.broken8809 = phi i64 [ 0, %_ZN22__llvm_libc_22_0_0_git3cpp6string7reserveEm.exit.i.i.i.i.i.i4201 ], [ %5584, %loadstoreloop ]
  %5580 = phi i64 [ %5582, %loadstoreloop ], [ 0, %_ZN22__llvm_libc_22_0_0_git3cpp6string7reserveEm.exit.i.i.i.i.i.i4201 ]
  %5581 = getelementptr i8, ptr %5577, i64 %5580
  store i8 0, ptr %5581, align 1
  %5582 = add i64 %5580, 1
  %5583 = icmp uge i64 %5582, %writer.sroa.10.5.i.i.i.i4183
  %5584 = call i64 @llvm.amdgcn.if.break.i64(i1 %5583, i64 %phi.broken8809)
  %5585 = call i1 @llvm.amdgcn.loop.i64(i64 %5584)
  br i1 %5585, label %Flow7722, label %loadstoreloop

Flow7722:                                         ; preds = %loadstoreloop
  %.lcssa8845 = phi i64 [ %5584, %loadstoreloop ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa8845)
  br label %Flow7724, !amdgpu.uniform !42

if.end.i.i.i.i.i.i4203:                           ; preds = %Flow7724
  %cmp.not.i6.i.i.i.i.i.i4205 = icmp ne ptr %5579, addrspacecast (ptr addrspace(4) @_ZN22__llvm_libc_22_0_0_git3cpp6string14NULL_CHARACTERE to ptr)
  %5586 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.not.i6.i.i.i.i.i.i4205)
  %5587 = extractvalue { i1, i64 } %5586, 0
  %5588 = extractvalue { i1, i64 } %5586, 1
  br i1 %5587, label %if.then.i.i.i.i2.i.i.i4206, label %_ZN22__llvm_libc_22_0_0_git3cpp9to_stringEi.exit.i4208

if.then.i.i.i.i2.i.i.i4206:                       ; preds = %if.end.i.i.i.i.i.i4203
  %arrayidx.i.i.i.i3.i.i.i4207 = getelementptr inbounds nuw i8, ptr %5579, i64 %writer.sroa.10.5.i.i.i.i4183
  store i8 0, ptr %arrayidx.i.i.i.i3.i.i.i4207, align 1, !tbaa !57, !noalias !190
  br label %_ZN22__llvm_libc_22_0_0_git3cpp9to_stringEi.exit.i4208, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git3cpp9to_stringEi.exit.i4208: ; preds = %if.then.i.i.i.i2.i.i.i4206, %if.end.i.i.i.i.i.i4203
  call void @llvm.amdgcn.end.cf.i64(i64 %5588)
  %5589 = trunc i64 %writer.sroa.10.5.i.i.i.i4183 to i32
  %5590 = and i64 %writer.sroa.10.5.i.i.i.i4183, 15
  %5591 = sub i64 %writer.sroa.10.5.i.i.i.i4183, %5590
  %5592 = icmp ne i64 %5591, 0
  %5593 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %5592)
  %5594 = extractvalue { i1, i64 } %5593, 0
  %5595 = extractvalue { i1, i64 } %5593, 1
  br i1 %5594, label %loop-memcpy-expansion5103.preheader, label %Flow7721

loop-memcpy-expansion5103.preheader:              ; preds = %_ZN22__llvm_libc_22_0_0_git3cpp9to_stringEi.exit.i4208
  %5596 = sub i32 11, %5589
  %scevgep6362 = getelementptr i8, ptr addrspace(5) %buffer.i.i.i2988, i32 %5596
  br label %loop-memcpy-expansion5103, !amdgpu.uniform !42

Flow7721:                                         ; preds = %Flow7720, %_ZN22__llvm_libc_22_0_0_git3cpp9to_stringEi.exit.i4208
  call void @llvm.amdgcn.end.cf.i64(i64 %5595)
  br label %loop-memcpy-residual-header5106, !amdgpu.uniform !42

loop-memcpy-expansion5103:                        ; preds = %loop-memcpy-expansion5103.preheader, %loop-memcpy-expansion5103
  %phi.broken8810 = phi i64 [ 0, %loop-memcpy-expansion5103.preheader ], [ %5601, %loop-memcpy-expansion5103 ]
  %lsr.iv6363 = phi ptr addrspace(5) [ %scevgep6364, %loop-memcpy-expansion5103 ], [ %scevgep6362, %loop-memcpy-expansion5103.preheader ]
  %loop-index5104 = phi i64 [ %5599, %loop-memcpy-expansion5103 ], [ 0, %loop-memcpy-expansion5103.preheader ]
  %5597 = load <4 x i32>, ptr addrspace(5) %lsr.iv6363, align 1
  %5598 = getelementptr i8, ptr %5579, i64 %loop-index5104
  store <4 x i32> %5597, ptr %5598, align 1
  %5599 = add i64 %loop-index5104, 16
  %scevgep6364 = getelementptr i8, ptr addrspace(5) %lsr.iv6363, i32 16
  %5600 = icmp uge i64 %5599, %5591
  %5601 = call i64 @llvm.amdgcn.if.break.i64(i1 %5600, i64 %phi.broken8810)
  %5602 = call i1 @llvm.amdgcn.loop.i64(i64 %5601)
  br i1 %5602, label %Flow7720, label %loop-memcpy-expansion5103

Flow7719:                                         ; preds = %Flow7718, %loop-memcpy-residual-header5106
  call void @llvm.amdgcn.end.cf.i64(i64 %6363)
  br label %post-loop-memcpy-expansion5102, !amdgpu.uniform !42

loop-memcpy-residual5105:                         ; preds = %loop-memcpy-residual5105.preheader, %loop-memcpy-residual5105
  %phi.broken8811 = phi i64 [ 0, %loop-memcpy-residual5105.preheader ], [ %5606, %loop-memcpy-residual5105 ]
  %lsr.iv6358 = phi ptr addrspace(5) [ %scevgep6359, %loop-memcpy-residual5105 ], [ %scevgep6357, %loop-memcpy-residual5105.preheader ]
  %residual-loop-index5107 = phi i64 [ %5604, %loop-memcpy-residual5105 ], [ 0, %loop-memcpy-residual5105.preheader ]
  %5603 = load i8, ptr addrspace(5) %lsr.iv6358, align 1
  %scevgep6361 = getelementptr i8, ptr %scevgep6360, i64 %residual-loop-index5107
  store i8 %5603, ptr %scevgep6361, align 1
  %5604 = add i64 %residual-loop-index5107, 1
  %scevgep6359 = getelementptr i8, ptr addrspace(5) %lsr.iv6358, i32 1
  %5605 = icmp uge i64 %5604, %5590
  %5606 = call i64 @llvm.amdgcn.if.break.i64(i1 %5605, i64 %phi.broken8811)
  %5607 = call i1 @llvm.amdgcn.loop.i64(i64 %5606)
  br i1 %5607, label %Flow7718, label %loop-memcpy-residual5105

Flow7718:                                         ; preds = %loop-memcpy-residual5105
  %.lcssa8843 = phi i64 [ %5606, %loop-memcpy-residual5105 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa8843)
  br label %Flow7719, !amdgpu.uniform !42

post-loop-memcpy-expansion5102:                   ; preds = %Flow7719
  call void @llvm.lifetime.end.p5(ptr addrspace(5) %buffer.i.i.i2988) #34, !noalias !190
  br label %for.cond.i.i.i4212, !amdgpu.uniform !42

for.cond.i.i.i4212:                               ; preds = %post-loop-memcpy-expansion5102, %Flow7717
  %5608 = phi i32 [ %5645, %Flow7717 ], [ poison, %post-loop-memcpy-expansion5102 ]
  %5609 = phi i64 [ %5646, %Flow7717 ], [ poison, %post-loop-memcpy-expansion5102 ]
  %5610 = phi i32 [ %5647, %Flow7717 ], [ poison, %post-loop-memcpy-expansion5102 ]
  %retval.sroa.6.0.i.i.i4213 = phi i32 [ %5650, %Flow7717 ], [ undef, %post-loop-memcpy-expansion5102 ]
  %retval.sroa.2.0.i.i.i4215 = phi i64 [ %5649, %Flow7717 ], [ undef, %post-loop-memcpy-expansion5102 ]
  %retval.sroa.8.0.i.i.i4216 = phi i32 [ %5648, %Flow7717 ], [ undef, %post-loop-memcpy-expansion5102 ]
  %index.0.i.i.i4217 = phi i32 [ %5652, %Flow7717 ], [ 0, %post-loop-memcpy-expansion5102 ]
  %5611 = load i32, ptr addrspace(1) @__llvm_rpc_client, align 8, !tbaa !68
  %cmp.not.i.i.i4218 = icmp ult i32 %index.0.i.i.i4217, %5611
  %spec.store.select.i.i.i4219 = select i1 %cmp.not.i.i.i4218, i32 %index.0.i.i.i4217, i32 0
  %5612 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  %5613 = and i32 %spec.store.select.i.i.i4219, 31
  %5614 = lshr i64 %5612, %sh_prom.i.i.i.i3270
  %5615 = trunc i64 %5614 to i32
  %conv4.i.i.i.i.i4223 = and i32 %5615, 1
  %shl.i.i.i.i.i4224 = shl nuw i32 %conv4.i.i.i.i.i4223, %5613
  br label %ComputeLoop5944, !amdgpu.uniform !42

5616:                                             ; preds = %ComputeEnd5945
  %sunkaddr7496 = getelementptr inbounds i8, ptr addrspace(1) %8696, i64 40
  %5617 = atomicrmw or ptr addrspace(1) %sunkaddr7496, i32 %.lcssa8837 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %5618, !amdgpu.uniform !42

5618:                                             ; preds = %5616, %ComputeEnd5945
  %5619 = phi i32 [ %5617, %5616 ], [ poison, %ComputeEnd5945 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %8705)
  %5620 = call i32 @llvm.amdgcn.readfirstlane.i32(i32 %5619)
  %5621 = or i32 %5620, %.lcssa8838
  %shl5.i.i.i.i.i4225 = shl nuw i32 1, %5613
  %and.i.i.i.i.i4226 = and i32 %shl5.i.i.i.i.i4225, %5621
  %tobool3.i.i.i.i4227 = icmp ne i32 %and.i.i.i.i.i4226, 0
  %5622 = call i64 @llvm.amdgcn.ballot.i64(i1 %tobool3.i.i.i.i4227)
  %and.i.i.i.i.i.i4228 = and i64 %5622, %5612
  %cmp.i.not.i.i.i4229 = icmp ne i64 %5612, %and.i.i.i.i.i.i4228
  br i1 %cmp.i.not.i.i.i4229, label %if.then.i.i.i.i4230, label %Flow7716, !amdgpu.uniform !42

if.then.i.i.i.i4230:                              ; preds = %5618
  fence syncscope("agent") acquire
  %5623 = load <2 x i64>, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !75
  %5624 = extractelement <2 x i64> %5623, i32 0
  %5625 = inttoptr i64 %5624 to ptr
  %5626 = extractelement <2 x i64> %5623, i32 1
  %5627 = inttoptr i64 %5626 to ptr
  %idxprom.i.i.i.i4231 = zext i32 %spec.store.select.i.i.i4219 to i64
  %arrayidx.i.i.i.i4232 = getelementptr inbounds nuw i32, ptr %5625, i64 %idxprom.i.i.i.i4231, !amdgpu.uniform !42
  %5628 = load atomic i32, ptr %arrayidx.i.i.i.i4232 monotonic, align 4
  %5629 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %5628)
  %arrayidx.i45.i.i.i4233 = getelementptr inbounds nuw i32, ptr %5627, i64 %idxprom.i.i.i.i4231, !amdgpu.uniform !42
  %5630 = load atomic i32, ptr %arrayidx.i45.i.i.i4233 monotonic, align 4
  %5631 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %5630)
  %cmp.i46.not.i.i.i4234 = icmp ne i32 %5629, %5631
  br i1 %cmp.i46.not.i.i.i4234, label %if.then12.i.i.i4235, label %Flow7714, !amdgpu.uniform !42

if.then12.i.i.i4235:                              ; preds = %if.then.i.i.i.i4230
  fence syncscope("agent") release
  %5632 = call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %5612, i1 true)
  %iszero.i.i.i.i.i.i.i4236 = icmp ne i64 %5612, 0
  %cmp2.i.i.i.i.i.i4237 = icmp eq i64 %5632, %sh_prom.i.i.i.i3270
  %cmp.i.i.i.i.i2.i4238 = select i1 %iszero.i.i.i.i.i.i.i4236, i1 %cmp2.i.i.i.i.i.i4237, i1 false
  %conv4.i.i51.i.i.i4239 = zext i1 %cmp.i.i.i.i.i2.i4238 to i32
  %shl.i.i52.i.i.i4240 = shl nuw i32 %conv4.i.i51.i.i.i4239, %5613
  %xor.i.i.i.i.i4241 = xor i32 %shl.i.i52.i.i.i4240, -1
  %5633 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop5949, !amdgpu.uniform !42

5634:                                             ; preds = %ComputeEnd5950
  %sunkaddr7497 = getelementptr inbounds i8, ptr addrspace(1) %8696, i64 40
  %5635 = atomicrmw and ptr addrspace(1) %sunkaddr7497, i32 %.lcssa8839 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %5636, !amdgpu.uniform !42

5636:                                             ; preds = %5634, %ComputeEnd5950
  call void @llvm.amdgcn.end.cf.i64(i64 %8723)
  call void @llvm.amdgcn.wave.barrier()
  br label %Flow7714, !amdgpu.uniform !42

if.end14.i.i.i4302:                               ; preds = %Flow7714
  %5637 = call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %5612, i1 true)
  %iszero.i.i.i.i.i.i4303 = icmp ne i64 %5612, 0
  %cmp2.i.i.i.i.i4304 = icmp eq i64 %5637, %sh_prom.i.i.i.i3270
  %cmp.i.i.i.i.i4305 = select i1 %iszero.i.i.i.i.i.i4303, i1 %cmp2.i.i.i.i.i4304, i1 false
  %5638 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.i.i.i.i.i4305)
  %5639 = extractvalue { i1, i64 } %5638, 0
  %5640 = extractvalue { i1, i64 } %5638, 1
  br i1 %5639, label %if.then16.i.i.i4307, label %if.end22.i.i.i4306

if.then16.i.i.i4307:                              ; preds = %if.end14.i.i.i4302
  %5641 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !76
  %opcode.i.i.i4308 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %5641, i64 %idxprom.i.i.i.i4231, i32 1
  store i32 1660944387, ptr %opcode.i.i.i4308, align 8, !tbaa !77
  %arrayidx21.i.i.i4309 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %5641, i64 %idxprom.i.i.i.i4231
  store i64 %5612, ptr %arrayidx21.i.i.i4309, align 8, !tbaa !80
  br label %if.end22.i.i.i4306, !amdgpu.uniform !42

Flow7715:                                         ; preds = %if.end22.i.i.i4306, %Flow7714
  %5642 = phi i32 [ %5629, %if.end22.i.i.i4306 ], [ %retval.sroa.8.0.i.i.i4216, %Flow7714 ]
  %5643 = phi i64 [ %5612, %if.end22.i.i.i4306 ], [ %retval.sroa.2.0.i.i.i4215, %Flow7714 ]
  %5644 = phi i32 [ %spec.store.select.i.i.i4219, %if.end22.i.i.i4306 ], [ %retval.sroa.6.0.i.i.i4213, %Flow7714 ]
  br label %cleanup26.i.i.i4242, !amdgpu.uniform !42

if.end22.i.i.i4306:                               ; preds = %if.then16.i.i.i4307, %if.end14.i.i.i4302
  call void @llvm.amdgcn.end.cf.i64(i64 %5640)
  call void @llvm.amdgcn.wave.barrier()
  br label %Flow7715, !amdgpu.uniform !42

Flow7716:                                         ; preds = %cleanup26.i.i.i4242, %5618
  %5645 = phi i32 [ %5644, %cleanup26.i.i.i4242 ], [ %5608, %5618 ]
  %5646 = phi i64 [ %5643, %cleanup26.i.i.i4242 ], [ %5609, %5618 ]
  %5647 = phi i32 [ %5642, %cleanup26.i.i.i4242 ], [ %5610, %5618 ]
  %5648 = phi i32 [ %5642, %cleanup26.i.i.i4242 ], [ %retval.sroa.8.0.i.i.i4216, %5618 ]
  %5649 = phi i64 [ %5643, %cleanup26.i.i.i4242 ], [ %retval.sroa.2.0.i.i.i4215, %5618 ]
  %5650 = phi i32 [ %5644, %cleanup26.i.i.i4242 ], [ %retval.sroa.6.0.i.i.i4213, %5618 ]
  %5651 = phi i1 [ %cmp.i46.not.i.i.i4234, %cleanup26.i.i.i4242 ], [ true, %5618 ]
  br i1 %5651, label %for.inc.i.i.i4247, label %Flow7717, !amdgpu.uniform !42

cleanup26.i.i.i4242:                              ; preds = %Flow7715
  br label %Flow7716, !amdgpu.uniform !42

for.inc.i.i.i4247:                                ; preds = %Flow7716
  %inc.i.i.i4252 = add i32 %spec.store.select.i.i.i4219, 1
  br label %Flow7717, !amdgpu.uniform !42

Flow7717:                                         ; preds = %for.inc.i.i.i4247, %Flow7716
  %5652 = phi i32 [ %inc.i.i.i4252, %for.inc.i.i.i4247 ], [ poison, %Flow7716 ]
  %5653 = phi i1 [ false, %for.inc.i.i.i4247 ], [ true, %Flow7716 ]
  br i1 %5653, label %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i4253, label %for.cond.i.i.i4212, !amdgpu.uniform !42

_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i4253: ; preds = %Flow7717
  %.lcssa8842 = phi i32 [ %5645, %Flow7717 ]
  %.lcssa8841 = phi i64 [ %5646, %Flow7717 ]
  %.lcssa8840 = phi i32 [ %5647, %Flow7717 ]
  fence acquire
  %5654 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %mul.i.i.i.i.i.i4254 = shl i32 %.lcssa8842, 6
  %idxprom.i19.i.i.i.i.i4255 = zext i32 %mul.i.i.i.i.i.i4254 to i64
  %arrayidx.i20.i.i.i.i.i4256 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %5654, i64 %idxprom.i19.i.i.i.i.i4255
  %arrayidx.i22.i.i.i.i.i4257 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %arrayidx.i20.i.i.i.i.i4256, i64 %sh_prom.i.i.i.i3270
  store i64 %writer.sroa.10.5.i.i.i.i4183, ptr %arrayidx.i22.i.i.i.i.i4257, align 8, !tbaa !82
  %spec.select.i.i.i.i.i.i.i4258 = call i64 @llvm.umin.i64(i64 %writer.sroa.10.5.i.i.i.i4183, i64 56)
  %5655 = and i64 %spec.select.i.i.i.i.i.i.i4258, 15
  %5656 = sub i64 %spec.select.i.i.i.i.i.i.i4258, %5655
  %5657 = icmp ne i64 %5656, 0
  %5658 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %5657)
  %5659 = extractvalue { i1, i64 } %5658, 0
  %5660 = extractvalue { i1, i64 } %5658, 1
  br i1 %5659, label %loop-memcpy-expansion4977.preheader, label %Flow7713

loop-memcpy-expansion4977.preheader:              ; preds = %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i4253
  %5661 = add nuw nsw i64 %4513, 8
  %scevgep6354 = getelementptr i8, ptr %5654, i64 %5661
  %5662 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i4255, 6
  %scevgep6355 = getelementptr i8, ptr %scevgep6354, i64 %5662
  br label %loop-memcpy-expansion4977, !amdgpu.uniform !42

Flow7713:                                         ; preds = %Flow7712, %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i4253
  call void @llvm.amdgcn.end.cf.i64(i64 %5660)
  br label %loop-memcpy-residual-header4980, !amdgpu.uniform !42

loop-memcpy-expansion4977:                        ; preds = %loop-memcpy-expansion4977.preheader, %loop-memcpy-expansion4977
  %phi.broken8812 = phi i64 [ 0, %loop-memcpy-expansion4977.preheader ], [ %5667, %loop-memcpy-expansion4977 ]
  %loop-index4978 = phi i64 [ %5665, %loop-memcpy-expansion4977 ], [ 0, %loop-memcpy-expansion4977.preheader ]
  %5663 = getelementptr i8, ptr %5579, i64 %loop-index4978
  %5664 = load <4 x i32>, ptr %5663, align 1
  %scevgep6356 = getelementptr i8, ptr %scevgep6355, i64 %loop-index4978
  store <4 x i32> %5664, ptr %scevgep6356, align 1
  %5665 = add i64 %loop-index4978, 16
  %5666 = icmp uge i64 %5665, %5656
  %5667 = call i64 @llvm.amdgcn.if.break.i64(i1 %5666, i64 %phi.broken8812)
  %5668 = call i1 @llvm.amdgcn.loop.i64(i64 %5667)
  br i1 %5668, label %Flow7712, label %loop-memcpy-expansion4977

Flow7711:                                         ; preds = %Flow7709, %Flow7710
  %5669 = phi i64 [ %6219, %Flow7709 ], [ %6211, %Flow7710 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %6214)
  br label %post-loop-memcpy-expansion4976, !amdgpu.uniform !42

loop-memcpy-residual4979:                         ; preds = %loop-memcpy-residual4979.preheader, %loop-memcpy-residual4979
  %phi.broken8813 = phi i64 [ 0, %loop-memcpy-residual4979.preheader ], [ %5673, %loop-memcpy-residual4979 ]
  %residual-loop-index4981 = phi i64 [ %5671, %loop-memcpy-residual4979 ], [ 0, %loop-memcpy-residual4979.preheader ]
  %scevgep6353 = getelementptr i8, ptr %scevgep6352, i64 %residual-loop-index4981
  %5670 = load i8, ptr %scevgep6353, align 1
  %scevgep6351 = getelementptr i8, ptr %scevgep6350, i64 %residual-loop-index4981
  store i8 %5670, ptr %scevgep6351, align 1
  %5671 = add i64 %residual-loop-index4981, 1
  %5672 = icmp uge i64 %5671, %5655
  %5673 = call i64 @llvm.amdgcn.if.break.i64(i1 %5672, i64 %phi.broken8813)
  %5674 = call i1 @llvm.amdgcn.loop.i64(i64 %5673)
  br i1 %5674, label %Flow7709, label %loop-memcpy-residual4979

Flow7709:                                         ; preds = %loop-memcpy-residual4979
  %.lcssa8835 = phi i64 [ %5673, %loop-memcpy-residual4979 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa8835)
  br label %Flow7711, !amdgpu.uniform !42

post-loop-memcpy-expansion4976:                   ; preds = %Flow7711
  %tobool.not.i.i.i.i.i.i4260 = icmp eq i32 %.lcssa8840, 0
  %conv.i.i.i.i.i.i4261 = zext i1 %tobool.not.i.i.i.i.i.i4260 to i32
  fence release
  %5675 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %idxprom.i23.i.i.i.i.i4262 = zext i32 %.lcssa8842 to i64
  %arrayidx.i24.i.i.i.i.i4263 = getelementptr inbounds nuw i32, ptr %5675, i64 %idxprom.i23.i.i.i.i.i4262
  store atomic i32 %conv.i.i.i.i.i.i4261, ptr %arrayidx.i24.i.i.i.i.i4263 monotonic, align 4
  %5676 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !84
  %arrayidx.i.i5.i.i4264 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %5676, i64 %idxprom.i23.i.i.i.i.i4262, !amdgpu.uniform !42
  %5677 = load i64, ptr %arrayidx.i.i5.i.i4264, align 8, !tbaa !80
  %scevgep = getelementptr i8, ptr %5579, i64 56
  %5678 = add i64 %writer.sroa.10.5.i.i.i.i4183, -56
  br label %while.cond.i.i.i.i4265, !amdgpu.uniform !42

while.cond.i.i.i.i4265:                           ; preds = %post-loop-memcpy-expansion4976, %Flow7708
  %phi.broken8816 = phi i64 [ 0, %post-loop-memcpy-expansion4976 ], [ %5695, %Flow7708 ]
  %lsr.iv6344 = phi i64 [ %5693, %Flow7708 ], [ %5678, %post-loop-memcpy-expansion4976 ]
  %lsr.iv = phi ptr [ %5692, %Flow7708 ], [ %scevgep, %post-loop-memcpy-expansion4976 ]
  %port.sroa.37.0.i.i4266 = phi i32 [ %5691, %Flow7708 ], [ %conv.i.i.i.i.i.i4261, %post-loop-memcpy-expansion4976 ]
  %idx.0.i.i.i.i4267 = phi i64 [ %5690, %Flow7708 ], [ 56, %post-loop-memcpy-expansion4976 ]
  %cmp.i.i.i.i4268 = icmp ugt i64 %writer.sroa.10.5.i.i.i.i4183, %idx.0.i.i.i.i4267
  %5679 = call i64 @llvm.amdgcn.ballot.i64(i1 %cmp.i.i.i.i4268)
  %and.i.i.i.i6.i.i4269 = and i64 %5679, %5677
  %tobool.not.i.i.i.i4270 = icmp ne i64 %and.i.i.i.i6.i.i4269, 0
  %5680 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i22.i.i.i4271 = getelementptr inbounds nuw i32, ptr %5680, i64 %idxprom.i23.i.i.i.i.i4262, !amdgpu.uniform !42
  %5681 = load atomic i32, ptr %arrayidx.i22.i.i.i4271 monotonic, align 4
  %5682 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %tobool.not.i.i.i.i4270)
  %5683 = extractvalue { i1, i64 } %5682, 0
  %5684 = extractvalue { i1, i64 } %5682, 1
  br i1 %5683, label %cond.false.i8.i.i.i.i4272, label %Flow7708

cond.false.i8.i.i.i.i4272:                        ; preds = %while.cond.i.i.i.i4265
  %5685 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %5681)
  %5686 = icmp ne i32 %5685, %port.sroa.37.0.i.i4266
  br i1 %5686, label %while.body.i.i21.i.i.i.i4273.preheader, label %Flow7707, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i4273.preheader:           ; preds = %cond.false.i8.i.i.i.i4272
  br label %while.body.i.i21.i.i.i.i4273, !amdgpu.uniform !42

Flow7707:                                         ; preds = %Flow7706, %cond.false.i8.i.i.i.i4272
  br label %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i4276, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i4273:                     ; preds = %while.body.i.i21.i.i.i.i4273.preheader, %while.body.i.i21.i.i.i.i4273
  call void @llvm.amdgcn.s.sleep(i32 2)
  %5687 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i22.i.i.i.i4274 = getelementptr inbounds nuw i32, ptr %5687, i64 %idxprom.i23.i.i.i.i.i4262, !amdgpu.uniform !42
  %5688 = load atomic i32, ptr %arrayidx.i.i.i22.i.i.i.i4274 monotonic, align 4
  %5689 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %5688)
  %cmp.i.not.i.i23.i.i.i.i4275 = icmp eq i32 %5689, %port.sroa.37.0.i.i4266
  br i1 %cmp.i.not.i.i23.i.i.i.i4275, label %Flow7706, label %while.body.i.i21.i.i.i.i4273, !amdgpu.uniform !42

Flow7706:                                         ; preds = %while.body.i.i21.i.i.i.i4273
  br label %Flow7707, !amdgpu.uniform !42

Flow7708:                                         ; preds = %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i4277, %while.cond.i.i.i.i4265
  %5690 = phi i64 [ %add.i.i.i.i4280, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i4277 ], [ poison, %while.cond.i.i.i.i4265 ]
  %5691 = phi i32 [ %conv.i.i26.i.i.i.i4278, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i4277 ], [ poison, %while.cond.i.i.i.i4265 ]
  %5692 = phi ptr [ %scevgep6340, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i4277 ], [ poison, %while.cond.i.i.i.i4265 ]
  %5693 = phi i64 [ %lsr.iv.next, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i4277 ], [ poison, %while.cond.i.i.i.i4265 ]
  %5694 = phi i1 [ false, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i4277 ], [ true, %while.cond.i.i.i.i4265 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %5684)
  %5695 = call i64 @llvm.amdgcn.if.break.i64(i1 %5694, i64 %phi.broken8816)
  %5696 = call i1 @llvm.amdgcn.loop.i64(i64 %5695)
  br i1 %5696, label %cond.false.i.i.i4287, label %while.cond.i.i.i.i4265

_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i4276: ; preds = %Flow7707
  fence acquire
  %5697 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.i.i.i.i4268)
  %5698 = extractvalue { i1, i64 } %5697, 0
  %5699 = extractvalue { i1, i64 } %5697, 1
  br i1 %5698, label %if.then.i.i.i.i.i.i3.i4281, label %Flow7705

if.then.i.i.i.i.i.i3.i4281:                       ; preds = %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i4276
  %sub.i.i.i.i.i.i.i4283 = sub nuw i64 %writer.sroa.10.5.i.i.i.i4183, %idx.0.i.i.i.i4267
  %spec.select.i.i.i36.i.i.i.i4284 = call i64 @llvm.umin.i64(i64 %sub.i.i.i.i.i.i.i4283, i64 64)
  %5700 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %5701 = and i64 %spec.select.i.i.i36.i.i.i.i4284, 15
  %5702 = sub i64 %spec.select.i.i.i36.i.i.i.i4284, %5701
  %5703 = icmp ne i64 %5702, 0
  %5704 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %5703)
  %5705 = extractvalue { i1, i64 } %5704, 0
  %5706 = extractvalue { i1, i64 } %5704, 1
  br i1 %5705, label %loop-memcpy-expansion4971.preheader, label %Flow7704

loop-memcpy-expansion4971.preheader:              ; preds = %if.then.i.i.i.i.i.i3.i4281
  %arrayidx.i20.i35.i.i.i.i4285 = getelementptr %"struct.rpc::Buffer", ptr %5700, i64 %idxprom.i19.i.i.i.i.i4255
  %arrayidx.i22.i38.i.i.i.i4286 = getelementptr %"struct.rpc::Buffer", ptr %arrayidx.i20.i35.i.i.i.i4285, i64 %sh_prom.i.i.i.i3270
  br label %loop-memcpy-expansion4971, !amdgpu.uniform !42

Flow7704:                                         ; preds = %Flow7703, %if.then.i.i.i.i.i.i3.i4281
  call void @llvm.amdgcn.end.cf.i64(i64 %5706)
  br label %loop-memcpy-residual-header4974, !amdgpu.uniform !42

loop-memcpy-expansion4971:                        ; preds = %loop-memcpy-expansion4971.preheader, %loop-memcpy-expansion4971
  %phi.broken8814 = phi i64 [ 0, %loop-memcpy-expansion4971.preheader ], [ %5711, %loop-memcpy-expansion4971 ]
  %loop-index4972 = phi i64 [ %5709, %loop-memcpy-expansion4971 ], [ 0, %loop-memcpy-expansion4971.preheader ]
  %scevgep6341 = getelementptr i8, ptr %lsr.iv, i64 %loop-index4972
  %5707 = load <4 x i32>, ptr %scevgep6341, align 1
  %5708 = getelementptr i8, ptr %arrayidx.i22.i38.i.i.i.i4286, i64 %loop-index4972
  store <4 x i32> %5707, ptr %5708, align 1
  %5709 = add i64 %loop-index4972, 16
  %5710 = icmp uge i64 %5709, %5702
  %5711 = call i64 @llvm.amdgcn.if.break.i64(i1 %5710, i64 %phi.broken8814)
  %5712 = call i1 @llvm.amdgcn.loop.i64(i64 %5711)
  br i1 %5712, label %Flow7703, label %loop-memcpy-expansion4971

Flow7702:                                         ; preds = %Flow7701, %loop-memcpy-residual-header4974
  call void @llvm.amdgcn.end.cf.i64(i64 %6204)
  br label %Flow7705, !amdgpu.uniform !42

loop-memcpy-residual4973:                         ; preds = %loop-memcpy-residual4973.preheader, %loop-memcpy-residual4973
  %phi.broken8815 = phi i64 [ 0, %loop-memcpy-residual4973.preheader ], [ %5716, %loop-memcpy-residual4973 ]
  %residual-loop-index4975 = phi i64 [ %5714, %loop-memcpy-residual4973 ], [ 0, %loop-memcpy-residual4973.preheader ]
  %scevgep6348 = getelementptr i8, ptr %scevgep6347, i64 %residual-loop-index4975
  %5713 = load i8, ptr %scevgep6348, align 1
  %scevgep6346 = getelementptr i8, ptr %scevgep6345, i64 %residual-loop-index4975
  store i8 %5713, ptr %scevgep6346, align 1
  %5714 = add i64 %residual-loop-index4975, 1
  %5715 = icmp uge i64 %5714, %5701
  %5716 = call i64 @llvm.amdgcn.if.break.i64(i1 %5715, i64 %phi.broken8815)
  %5717 = call i1 @llvm.amdgcn.loop.i64(i64 %5716)
  br i1 %5717, label %Flow7701, label %loop-memcpy-residual4973

Flow7701:                                         ; preds = %loop-memcpy-residual4973
  %.lcssa8832 = phi i64 [ %5716, %loop-memcpy-residual4973 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa8832)
  br label %Flow7702, !amdgpu.uniform !42

_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i4277: ; preds = %Flow7705
  %conv.i.i26.i.i.i.i4278 = xor i32 %port.sroa.37.0.i.i4266, 1
  fence release
  %5718 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %arrayidx.i24.i29.i.i.i.i4279 = getelementptr inbounds nuw i32, ptr %5718, i64 %idxprom.i23.i.i.i.i.i4262
  store atomic i32 %conv.i.i26.i.i.i.i4278, ptr %arrayidx.i24.i29.i.i.i.i4279 monotonic, align 4
  %add.i.i.i.i4280 = add i64 %idx.0.i.i.i.i4267, 64
  %scevgep6340 = getelementptr i8, ptr %lsr.iv, i64 64
  %lsr.iv.next = add i64 %lsr.iv6344, -64
  br label %Flow7708, !amdgpu.uniform !42

cond.false.i.i.i4287:                             ; preds = %Flow7708
  %.lcssa8834 = phi i64 [ %5695, %Flow7708 ]
  %port.sroa.37.0.i.i4266.lcssa = phi i32 [ %port.sroa.37.0.i.i4266, %Flow7708 ]
  %.lcssa8833 = phi i32 [ %5681, %Flow7708 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa8834)
  %5719 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %.lcssa8833)
  %5720 = icmp ne i32 %5719, %port.sroa.37.0.i.i4266.lcssa
  %5721 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %5720)
  %5722 = extractvalue { i1, i64 } %5721, 0
  %5723 = extractvalue { i1, i64 } %5721, 1
  br i1 %5722, label %while.body.i.i.i.i4288.preheader, label %Flow7700

while.body.i.i.i.i4288.preheader:                 ; preds = %cond.false.i.i.i4287
  br label %while.body.i.i.i.i4288, !amdgpu.uniform !42

Flow7700:                                         ; preds = %Flow7699, %cond.false.i.i.i4287
  call void @llvm.amdgcn.end.cf.i64(i64 %5723)
  br label %_ZN22__llvm_libc_22_0_0_git15write_to_stderrENS_3cpp11string_viewE.exit.i4291, !amdgpu.uniform !42

while.body.i.i.i.i4288:                           ; preds = %while.body.i.i.i.i4288.preheader, %while.body.i.i.i.i4288
  %phi.broken8817 = phi i64 [ 0, %while.body.i.i.i.i4288.preheader ], [ %5727, %while.body.i.i.i.i4288 ]
  call void @llvm.amdgcn.s.sleep(i32 2)
  %5724 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i8.i.i4289 = getelementptr inbounds nuw i32, ptr %5724, i64 %idxprom.i23.i.i.i.i.i4262, !amdgpu.uniform !42
  %5725 = load atomic i32, ptr %arrayidx.i.i.i8.i.i4289 monotonic, align 4
  %5726 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %5725)
  %cmp.i.not.i.i.i.i4290 = icmp eq i32 %5726, %port.sroa.37.0.i.i4266.lcssa
  %5727 = call i64 @llvm.amdgcn.if.break.i64(i1 %cmp.i.not.i.i.i.i4290, i64 %phi.broken8817)
  %5728 = call i1 @llvm.amdgcn.loop.i64(i64 %5727)
  br i1 %5728, label %Flow7699, label %while.body.i.i.i.i4288

Flow7699:                                         ; preds = %while.body.i.i.i.i4288
  %.lcssa8830 = phi i64 [ %5727, %while.body.i.i.i.i4288 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa8830)
  br label %Flow7700, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git15write_to_stderrENS_3cpp11string_viewE.exit.i4291: ; preds = %Flow7700
  fence acquire
  call void @llvm.amdgcn.wave.barrier()
  fence syncscope("agent") release
  %5729 = call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %.lcssa8841, i1 true)
  %iszero.i.i.i.i.i13.i.i4292 = icmp ne i64 %.lcssa8841, 0
  %cmp2.i.i.i.i14.i.i4293 = icmp eq i64 %5729, %sh_prom.i.i.i.i3270
  %cmp.i.i.i.i15.i.i4294 = select i1 %iszero.i.i.i.i.i13.i.i4292, i1 %cmp2.i.i.i.i14.i.i4293, i1 false
  %5730 = and i32 %.lcssa8842, 31
  %conv4.i.i.i19.i.i4298 = zext i1 %cmp.i.i.i.i15.i.i4294 to i32
  %shl.i.i.i20.i.i4299 = shl nuw i32 %conv4.i.i.i19.i.i4298, %5730
  %xor.i.i.i21.i.i4300 = xor i32 %shl.i.i.i20.i.i4299, -1
  %5731 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop5953, !amdgpu.uniform !42

5732:                                             ; preds = %ComputeEnd5954
  %div8.i.i.i16.i.i4295 = lshr i32 %.lcssa8842, 5
  %idxprom.i.i.i17.i.i4296 = zext nneg i32 %div8.i.i.i16.i.i4295 to i64
  %5733 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i17.i.i4296
  %sunkaddr7498 = getelementptr inbounds i8, ptr addrspace(1) %5733, i64 40
  %5734 = atomicrmw and ptr addrspace(1) %sunkaddr7498, i32 %.lcssa8829 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %5735, !amdgpu.uniform !42

5735:                                             ; preds = %5732, %ComputeEnd5954
  call void @llvm.amdgcn.end.cf.i64(i64 %8740)
  call void @llvm.amdgcn.wave.barrier()
  br label %for.cond.i.i.i4326, !amdgpu.uniform !42

for.cond.i.i.i4326:                               ; preds = %5735, %Flow7698
  %5736 = phi i32 [ %5773, %Flow7698 ], [ poison, %5735 ]
  %5737 = phi i64 [ %5774, %Flow7698 ], [ poison, %5735 ]
  %5738 = phi i32 [ %5775, %Flow7698 ], [ poison, %5735 ]
  %retval.sroa.6.0.i.i.i4327 = phi i32 [ %5778, %Flow7698 ], [ undef, %5735 ]
  %retval.sroa.2.0.i.i.i4329 = phi i64 [ %5777, %Flow7698 ], [ undef, %5735 ]
  %retval.sroa.8.0.i.i.i4330 = phi i32 [ %5776, %Flow7698 ], [ undef, %5735 ]
  %index.0.i.i.i4331 = phi i32 [ %5780, %Flow7698 ], [ 0, %5735 ]
  %5739 = load i32, ptr addrspace(1) @__llvm_rpc_client, align 8, !tbaa !68
  %cmp.not.i.i.i4332 = icmp ult i32 %index.0.i.i.i4331, %5739
  %spec.store.select.i.i.i4333 = select i1 %cmp.not.i.i.i4332, i32 %index.0.i.i.i4331, i32 0
  %5740 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  %5741 = and i32 %spec.store.select.i.i.i4333, 31
  %5742 = lshr i64 %5740, %sh_prom.i.i.i.i3270
  %5743 = trunc i64 %5742 to i32
  %conv4.i.i.i.i.i4337 = and i32 %5743, 1
  %shl.i.i.i.i.i4338 = shl nuw i32 %conv4.i.i.i.i.i4337, %5741
  br label %ComputeLoop5957, !amdgpu.uniform !42

5744:                                             ; preds = %ComputeEnd5958
  %sunkaddr7499 = getelementptr inbounds i8, ptr addrspace(1) %8750, i64 40
  %5745 = atomicrmw or ptr addrspace(1) %sunkaddr7499, i32 %.lcssa8823 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %5746, !amdgpu.uniform !42

5746:                                             ; preds = %5744, %ComputeEnd5958
  %5747 = phi i32 [ %5745, %5744 ], [ poison, %ComputeEnd5958 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %8759)
  %5748 = call i32 @llvm.amdgcn.readfirstlane.i32(i32 %5747)
  %5749 = or i32 %5748, %.lcssa8824
  %shl5.i.i.i.i.i4339 = shl nuw i32 1, %5741
  %and.i.i.i.i.i4340 = and i32 %shl5.i.i.i.i.i4339, %5749
  %tobool3.i.i.i.i4341 = icmp ne i32 %and.i.i.i.i.i4340, 0
  %5750 = call i64 @llvm.amdgcn.ballot.i64(i1 %tobool3.i.i.i.i4341)
  %and.i.i.i.i.i.i4342 = and i64 %5750, %5740
  %cmp.i.not.i.i.i4343 = icmp ne i64 %5740, %and.i.i.i.i.i.i4342
  br i1 %cmp.i.not.i.i.i4343, label %if.then.i.i.i.i4344, label %Flow7697, !amdgpu.uniform !42

if.then.i.i.i.i4344:                              ; preds = %5746
  fence syncscope("agent") acquire
  %5751 = load <2 x i64>, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !75
  %5752 = extractelement <2 x i64> %5751, i32 0
  %5753 = inttoptr i64 %5752 to ptr
  %5754 = extractelement <2 x i64> %5751, i32 1
  %5755 = inttoptr i64 %5754 to ptr
  %idxprom.i.i.i.i4345 = zext i32 %spec.store.select.i.i.i4333 to i64
  %arrayidx.i.i.i.i4346 = getelementptr inbounds nuw i32, ptr %5753, i64 %idxprom.i.i.i.i4345, !amdgpu.uniform !42
  %5756 = load atomic i32, ptr %arrayidx.i.i.i.i4346 monotonic, align 4
  %5757 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %5756)
  %arrayidx.i45.i.i.i4347 = getelementptr inbounds nuw i32, ptr %5755, i64 %idxprom.i.i.i.i4345, !amdgpu.uniform !42
  %5758 = load atomic i32, ptr %arrayidx.i45.i.i.i4347 monotonic, align 4
  %5759 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %5758)
  %cmp.i46.not.i.i.i4348 = icmp ne i32 %5757, %5759
  br i1 %cmp.i46.not.i.i.i4348, label %if.then12.i.i.i4349, label %Flow7695, !amdgpu.uniform !42

if.then12.i.i.i4349:                              ; preds = %if.then.i.i.i.i4344
  fence syncscope("agent") release
  %5760 = call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %5740, i1 true)
  %iszero.i.i.i.i.i.i.i4350 = icmp ne i64 %5740, 0
  %cmp2.i.i.i.i.i.i4351 = icmp eq i64 %5760, %sh_prom.i.i.i.i3270
  %cmp.i.i.i.i.i.i4352 = select i1 %iszero.i.i.i.i.i.i.i4350, i1 %cmp2.i.i.i.i.i.i4351, i1 false
  %conv4.i.i51.i.i.i4353 = zext i1 %cmp.i.i.i.i.i.i4352 to i32
  %shl.i.i52.i.i.i4354 = shl nuw i32 %conv4.i.i51.i.i.i4353, %5741
  %xor.i.i.i.i.i4355 = xor i32 %shl.i.i52.i.i.i4354, -1
  %5761 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop5962, !amdgpu.uniform !42

5762:                                             ; preds = %ComputeEnd5963
  %sunkaddr7500 = getelementptr inbounds i8, ptr addrspace(1) %8750, i64 40
  %5763 = atomicrmw and ptr addrspace(1) %sunkaddr7500, i32 %.lcssa8825 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %5764, !amdgpu.uniform !42

5764:                                             ; preds = %5762, %ComputeEnd5963
  call void @llvm.amdgcn.end.cf.i64(i64 %8777)
  call void @llvm.amdgcn.wave.barrier()
  br label %Flow7695, !amdgpu.uniform !42

if.end14.i.i.i4413:                               ; preds = %Flow7695
  %5765 = call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %5740, i1 true)
  %iszero.i.i.i.i.i.i4414 = icmp ne i64 %5740, 0
  %cmp2.i.i.i.i.i4415 = icmp eq i64 %5765, %sh_prom.i.i.i.i3270
  %cmp.i.i.i.i.i4416 = select i1 %iszero.i.i.i.i.i.i4414, i1 %cmp2.i.i.i.i.i4415, i1 false
  %5766 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.i.i.i.i.i4416)
  %5767 = extractvalue { i1, i64 } %5766, 0
  %5768 = extractvalue { i1, i64 } %5766, 1
  br i1 %5767, label %if.then16.i.i.i4418, label %if.end22.i.i.i4417

if.then16.i.i.i4418:                              ; preds = %if.end14.i.i.i4413
  %5769 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !76
  %opcode.i.i.i4419 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %5769, i64 %idxprom.i.i.i.i4345, i32 1
  store i32 1660944387, ptr %opcode.i.i.i4419, align 8, !tbaa !77
  %arrayidx21.i.i.i4420 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %5769, i64 %idxprom.i.i.i.i4345
  store i64 %5740, ptr %arrayidx21.i.i.i4420, align 8, !tbaa !80
  br label %if.end22.i.i.i4417, !amdgpu.uniform !42

Flow7696:                                         ; preds = %if.end22.i.i.i4417, %Flow7695
  %5770 = phi i32 [ %5757, %if.end22.i.i.i4417 ], [ %retval.sroa.8.0.i.i.i4330, %Flow7695 ]
  %5771 = phi i64 [ %5740, %if.end22.i.i.i4417 ], [ %retval.sroa.2.0.i.i.i4329, %Flow7695 ]
  %5772 = phi i32 [ %spec.store.select.i.i.i4333, %if.end22.i.i.i4417 ], [ %retval.sroa.6.0.i.i.i4327, %Flow7695 ]
  br label %cleanup26.i.i.i4356, !amdgpu.uniform !42

if.end22.i.i.i4417:                               ; preds = %if.then16.i.i.i4418, %if.end14.i.i.i4413
  call void @llvm.amdgcn.end.cf.i64(i64 %5768)
  call void @llvm.amdgcn.wave.barrier()
  br label %Flow7696, !amdgpu.uniform !42

Flow7697:                                         ; preds = %cleanup26.i.i.i4356, %5746
  %5773 = phi i32 [ %5772, %cleanup26.i.i.i4356 ], [ %5736, %5746 ]
  %5774 = phi i64 [ %5771, %cleanup26.i.i.i4356 ], [ %5737, %5746 ]
  %5775 = phi i32 [ %5770, %cleanup26.i.i.i4356 ], [ %5738, %5746 ]
  %5776 = phi i32 [ %5770, %cleanup26.i.i.i4356 ], [ %retval.sroa.8.0.i.i.i4330, %5746 ]
  %5777 = phi i64 [ %5771, %cleanup26.i.i.i4356 ], [ %retval.sroa.2.0.i.i.i4329, %5746 ]
  %5778 = phi i32 [ %5772, %cleanup26.i.i.i4356 ], [ %retval.sroa.6.0.i.i.i4327, %5746 ]
  %5779 = phi i1 [ %cmp.i46.not.i.i.i4348, %cleanup26.i.i.i4356 ], [ true, %5746 ]
  br i1 %5779, label %for.inc.i.i.i4361, label %Flow7698, !amdgpu.uniform !42

cleanup26.i.i.i4356:                              ; preds = %Flow7696
  br label %Flow7697, !amdgpu.uniform !42

for.inc.i.i.i4361:                                ; preds = %Flow7697
  %inc.i.i.i4366 = add i32 %spec.store.select.i.i.i4333, 1
  br label %Flow7698, !amdgpu.uniform !42

Flow7698:                                         ; preds = %for.inc.i.i.i4361, %Flow7697
  %5780 = phi i32 [ %inc.i.i.i4366, %for.inc.i.i.i4361 ], [ poison, %Flow7697 ]
  %5781 = phi i1 [ false, %for.inc.i.i.i4361 ], [ true, %Flow7697 ]
  br i1 %5781, label %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i4367, label %for.cond.i.i.i4326, !amdgpu.uniform !42

_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i4367: ; preds = %Flow7698
  %.lcssa8828 = phi i32 [ %5773, %Flow7698 ]
  %.lcssa8827 = phi i64 [ %5774, %Flow7698 ]
  %.lcssa8826 = phi i32 [ %5775, %Flow7698 ]
  fence acquire
  %5782 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %mul.i.i.i.i.i.i4368 = shl i32 %.lcssa8828, 6
  %idxprom.i19.i.i.i.i.i4369 = zext i32 %mul.i.i.i.i.i.i4368 to i64
  %arrayidx.i20.i.i.i.i.i4370 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %5782, i64 %idxprom.i19.i.i.i.i.i4369
  %arrayidx.i22.i.i.i.i.i4371 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %arrayidx.i20.i.i.i.i.i4370, i64 %sh_prom.i.i.i.i3270
  store i64 1, ptr %arrayidx.i22.i.i.i.i.i4371, align 8, !tbaa !82
  %arrayidx6.i.i.i.i.i.i.i4372 = getelementptr inbounds nuw i8, ptr %arrayidx.i22.i.i.i.i.i4371, i64 8
  store i8 10, ptr %arrayidx6.i.i.i.i.i.i.i4372, align 8
  %tobool.not.i.i.i.i.i.i4373 = icmp eq i32 %.lcssa8826, 0
  %conv.i.i.i.i.i.i4374 = zext i1 %tobool.not.i.i.i.i.i.i4373 to i32
  fence release
  %5783 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %idxprom.i23.i.i.i.i.i4375 = zext i32 %.lcssa8828 to i64
  %arrayidx.i24.i.i.i.i.i4376 = getelementptr inbounds nuw i32, ptr %5783, i64 %idxprom.i23.i.i.i.i.i4375
  store atomic i32 %conv.i.i.i.i.i.i4374, ptr %arrayidx.i24.i.i.i.i.i4376 monotonic, align 4
  %5784 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !84
  %arrayidx.i.i5.i.i4377 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %5784, i64 %idxprom.i23.i.i.i.i.i4375, !amdgpu.uniform !42
  %5785 = load i64, ptr %arrayidx.i.i5.i.i4377, align 8, !tbaa !80
  br label %while.cond.i.i.i.i4378, !amdgpu.uniform !42

while.cond.i.i.i.i4378:                           ; preds = %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i4367, %Flow7694
  %phi.broken8818 = phi i64 [ 0, %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i4367 ], [ %5800, %Flow7694 ]
  %port.sroa.37.0.i.i4379 = phi i32 [ %5798, %Flow7694 ], [ %conv.i.i.i.i.i.i4374, %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i4367 ]
  %idx.0.i.i.i.i4380 = phi i64 [ %5797, %Flow7694 ], [ 56, %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i4367 ]
  %cmp.i.i.i.i4381 = icmp eq i64 %idx.0.i.i.i.i4380, 0
  %5786 = call i64 @llvm.amdgcn.ballot.i64(i1 %cmp.i.i.i.i4381)
  %and.i.i.i.i6.i.i4382 = and i64 %5786, %5785
  %tobool.not.i.i.i.i4383 = icmp ne i64 %and.i.i.i.i6.i.i4382, 0
  %5787 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i22.i.i.i4384 = getelementptr inbounds nuw i32, ptr %5787, i64 %idxprom.i23.i.i.i.i.i4375, !amdgpu.uniform !42
  %5788 = load atomic i32, ptr %arrayidx.i22.i.i.i4384 monotonic, align 4
  %5789 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %tobool.not.i.i.i.i4383)
  %5790 = extractvalue { i1, i64 } %5789, 0
  %5791 = extractvalue { i1, i64 } %5789, 1
  br i1 %5790, label %cond.false.i8.i.i.i.i4385, label %Flow7694

cond.false.i8.i.i.i.i4385:                        ; preds = %while.cond.i.i.i.i4378
  %5792 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %5788)
  %5793 = icmp ne i32 %5792, %port.sroa.37.0.i.i4379
  br i1 %5793, label %while.body.i.i21.i.i.i.i4386.preheader, label %Flow7693, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i4386.preheader:           ; preds = %cond.false.i8.i.i.i.i4385
  br label %while.body.i.i21.i.i.i.i4386, !amdgpu.uniform !42

Flow7693:                                         ; preds = %Flow7692, %cond.false.i8.i.i.i.i4385
  br label %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i4389, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i4386:                     ; preds = %while.body.i.i21.i.i.i.i4386.preheader, %while.body.i.i21.i.i.i.i4386
  call void @llvm.amdgcn.s.sleep(i32 2)
  %5794 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i22.i.i.i.i4387 = getelementptr inbounds nuw i32, ptr %5794, i64 %idxprom.i23.i.i.i.i.i4375, !amdgpu.uniform !42
  %5795 = load atomic i32, ptr %arrayidx.i.i.i22.i.i.i.i4387 monotonic, align 4
  %5796 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %5795)
  %cmp.i.not.i.i23.i.i.i.i4388 = icmp eq i32 %5796, %port.sroa.37.0.i.i4379
  br i1 %cmp.i.not.i.i23.i.i.i.i4388, label %Flow7692, label %while.body.i.i21.i.i.i.i4386, !amdgpu.uniform !42

Flow7692:                                         ; preds = %while.body.i.i21.i.i.i.i4386
  br label %Flow7693, !amdgpu.uniform !42

Flow7694:                                         ; preds = %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i4390, %while.cond.i.i.i.i4378
  %5797 = phi i64 [ %add.i.i.i.i4393, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i4390 ], [ poison, %while.cond.i.i.i.i4378 ]
  %5798 = phi i32 [ %conv.i.i26.i.i.i.i4391, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i4390 ], [ poison, %while.cond.i.i.i.i4378 ]
  %5799 = phi i1 [ false, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i4390 ], [ true, %while.cond.i.i.i.i4378 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %5791)
  %5800 = call i64 @llvm.amdgcn.if.break.i64(i1 %5799, i64 %phi.broken8818)
  %5801 = call i1 @llvm.amdgcn.loop.i64(i64 %5800)
  br i1 %5801, label %cond.false.i.i.i4399, label %while.cond.i.i.i.i4378

_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i4389: ; preds = %Flow7693
  fence acquire
  br i1 %cmp.i.i.i.i4381, label %if.then.i.i.i.i.i.i.i4394, label %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i4390, !amdgpu.uniform !42

if.then.i.i.i.i.i.i.i4394:                        ; preds = %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i4389
  %5802 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %arrayidx.i20.i35.i.i.i.i4397 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %5802, i64 %idxprom.i19.i.i.i.i.i4369
  %arrayidx.i22.i38.i.i.i.i4398 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %arrayidx.i20.i35.i.i.i.i4397, i64 %sh_prom.i.i.i.i3270
  store i8 10, ptr %arrayidx.i22.i38.i.i.i.i4398, align 1
  br label %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i4390, !amdgpu.uniform !42

_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i4390: ; preds = %if.then.i.i.i.i.i.i.i4394, %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i4389
  %conv.i.i26.i.i.i.i4391 = xor i32 %port.sroa.37.0.i.i4379, 1
  fence release
  %5803 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %arrayidx.i24.i29.i.i.i.i4392 = getelementptr inbounds nuw i32, ptr %5803, i64 %idxprom.i23.i.i.i.i.i4375
  store atomic i32 %conv.i.i26.i.i.i.i4391, ptr %arrayidx.i24.i29.i.i.i.i4392 monotonic, align 4
  %add.i.i.i.i4393 = add i64 %idx.0.i.i.i.i4380, 64
  br label %Flow7694, !amdgpu.uniform !42

cond.false.i.i.i4399:                             ; preds = %Flow7694
  %.lcssa8822 = phi i64 [ %5800, %Flow7694 ]
  %port.sroa.37.0.i.i4379.lcssa = phi i32 [ %port.sroa.37.0.i.i4379, %Flow7694 ]
  %.lcssa8821 = phi i32 [ %5788, %Flow7694 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa8822)
  %5804 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %.lcssa8821)
  %5805 = icmp ne i32 %5804, %port.sroa.37.0.i.i4379.lcssa
  %5806 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %5805)
  %5807 = extractvalue { i1, i64 } %5806, 0
  %5808 = extractvalue { i1, i64 } %5806, 1
  br i1 %5807, label %while.body.i.i.i.i4400.preheader, label %Flow7691

while.body.i.i.i.i4400.preheader:                 ; preds = %cond.false.i.i.i4399
  br label %while.body.i.i.i.i4400, !amdgpu.uniform !42

Flow7691:                                         ; preds = %Flow7690, %cond.false.i.i.i4399
  call void @llvm.amdgcn.end.cf.i64(i64 %5808)
  br label %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIcEERS1_T_.exit4421, !amdgpu.uniform !42

while.body.i.i.i.i4400:                           ; preds = %while.body.i.i.i.i4400.preheader, %while.body.i.i.i.i4400
  %phi.broken8819 = phi i64 [ 0, %while.body.i.i.i.i4400.preheader ], [ %5812, %while.body.i.i.i.i4400 ]
  call void @llvm.amdgcn.s.sleep(i32 2)
  %5809 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i8.i.i4401 = getelementptr inbounds nuw i32, ptr %5809, i64 %idxprom.i23.i.i.i.i.i4375, !amdgpu.uniform !42
  %5810 = load atomic i32, ptr %arrayidx.i.i.i8.i.i4401 monotonic, align 4
  %5811 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %5810)
  %cmp.i.not.i.i.i.i4402 = icmp eq i32 %5811, %port.sroa.37.0.i.i4379.lcssa
  %5812 = call i64 @llvm.amdgcn.if.break.i64(i1 %cmp.i.not.i.i.i.i4402, i64 %phi.broken8819)
  %5813 = call i1 @llvm.amdgcn.loop.i64(i64 %5812)
  br i1 %5813, label %Flow7690, label %while.body.i.i.i.i4400

Flow7690:                                         ; preds = %while.body.i.i.i.i4400
  %.lcssa8820 = phi i64 [ %5812, %while.body.i.i.i.i4400 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa8820)
  br label %Flow7691, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIcEERS1_T_.exit4421: ; preds = %Flow7691
  fence acquire
  call void @llvm.amdgcn.wave.barrier()
  fence syncscope("agent") release
  %5814 = call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %.lcssa8827, i1 true)
  %iszero.i.i.i.i.i13.i.i4404 = icmp ne i64 %.lcssa8827, 0
  %cmp2.i.i.i.i14.i.i4405 = icmp eq i64 %5814, %sh_prom.i.i.i.i3270
  %cmp.i.i.i.i15.i.i4406 = select i1 %iszero.i.i.i.i.i13.i.i4404, i1 %cmp2.i.i.i.i14.i.i4405, i1 false
  %5815 = and i32 %.lcssa8828, 31
  %conv4.i.i.i19.i.i4410 = zext i1 %cmp.i.i.i.i15.i.i4406 to i32
  %shl.i.i.i20.i.i4411 = shl nuw i32 %conv4.i.i.i19.i.i4410, %5815
  %xor.i.i.i21.i.i4412 = xor i32 %shl.i.i.i20.i.i4411, -1
  %5816 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop5966, !amdgpu.uniform !42

5817:                                             ; preds = %ComputeEnd5967
  %div8.i.i.i16.i.i4407 = lshr i32 %.lcssa8828, 5
  %idxprom.i.i.i17.i.i4408 = zext nneg i32 %div8.i.i.i16.i.i4407 to i64
  %5818 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i17.i.i4408
  %sunkaddr7501 = getelementptr inbounds i8, ptr addrspace(1) %5818, i64 40
  %5819 = atomicrmw and ptr addrspace(1) %sunkaddr7501, i32 %.lcssa syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %5820, !amdgpu.uniform !42

5820:                                             ; preds = %5817, %ComputeEnd5967
  call void @llvm.amdgcn.end.cf.i64(i64 %8794)
  call void @llvm.amdgcn.wave.barrier()
  br label %Flow7917, !amdgpu.uniform !42

Flow7916:                                         ; preds = %Flow7670, %for.cond.cleanup
  %5821 = call { i1, i64 } @llvm.amdgcn.else.i64.i64(i64 %1147)
  %5822 = extractvalue { i1, i64 } %5821, 0
  %5823 = extractvalue { i1, i64 } %5821, 1
  br i1 %5822, label %for.cond.i.i1.i3271.preheader, label %Flow7917

for.cond.i.i1.i4430:                              ; preds = %for.cond.i.i1.i4430.preheader, %Flow7689
  %5824 = phi i32 [ %5861, %Flow7689 ], [ poison, %for.cond.i.i1.i4430.preheader ]
  %5825 = phi i64 [ %5862, %Flow7689 ], [ poison, %for.cond.i.i1.i4430.preheader ]
  %5826 = phi i32 [ %5863, %Flow7689 ], [ poison, %for.cond.i.i1.i4430.preheader ]
  %retval.sroa.6.0.i.i.i4431 = phi i32 [ %5866, %Flow7689 ], [ undef, %for.cond.i.i1.i4430.preheader ]
  %retval.sroa.2.0.i.i.i4433 = phi i64 [ %5865, %Flow7689 ], [ undef, %for.cond.i.i1.i4430.preheader ]
  %retval.sroa.8.0.i.i.i4434 = phi i32 [ %5864, %Flow7689 ], [ undef, %for.cond.i.i1.i4430.preheader ]
  %index.0.i.i.i4435 = phi i32 [ %5868, %Flow7689 ], [ 0, %for.cond.i.i1.i4430.preheader ]
  %5827 = load i32, ptr addrspace(1) @__llvm_rpc_client, align 8, !tbaa !68
  %cmp.not.i.i.i4436 = icmp ult i32 %index.0.i.i.i4435, %5827
  %spec.store.select.i.i.i4437 = select i1 %cmp.not.i.i.i4436, i32 %index.0.i.i.i4435, i32 0
  %5828 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  %5829 = and i32 %spec.store.select.i.i.i4437, 31
  %5830 = lshr i64 %5828, %sh_prom.i.i.i.i3270
  %5831 = trunc i64 %5830 to i32
  %conv4.i.i.i.i.i4441 = and i32 %5831, 1
  %shl.i.i.i.i.i4442 = shl nuw i32 %conv4.i.i.i.i.i4441, %5829
  br label %ComputeLoop5970, !amdgpu.uniform !42

5832:                                             ; preds = %ComputeEnd5971
  %sunkaddr7502 = getelementptr inbounds i8, ptr addrspace(1) %8804, i64 40
  %5833 = atomicrmw or ptr addrspace(1) %sunkaddr7502, i32 %.lcssa8962 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %5834, !amdgpu.uniform !42

5834:                                             ; preds = %5832, %ComputeEnd5971
  %5835 = phi i32 [ %5833, %5832 ], [ poison, %ComputeEnd5971 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %8813)
  %5836 = call i32 @llvm.amdgcn.readfirstlane.i32(i32 %5835)
  %5837 = or i32 %5836, %.lcssa8963
  %shl5.i.i.i.i.i4443 = shl nuw i32 1, %5829
  %and.i.i.i.i.i4444 = and i32 %shl5.i.i.i.i.i4443, %5837
  %tobool3.i.i.i.i4445 = icmp ne i32 %and.i.i.i.i.i4444, 0
  %5838 = call i64 @llvm.amdgcn.ballot.i64(i1 %tobool3.i.i.i.i4445)
  %and.i.i.i.i.i.i4446 = and i64 %5838, %5828
  %cmp.i.not.i.i.i4447 = icmp ne i64 %5828, %and.i.i.i.i.i.i4446
  br i1 %cmp.i.not.i.i.i4447, label %if.then.i.i.i.i4448, label %Flow7688, !amdgpu.uniform !42

if.then.i.i.i.i4448:                              ; preds = %5834
  fence syncscope("agent") acquire
  %5839 = load <2 x i64>, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !75
  %5840 = extractelement <2 x i64> %5839, i32 0
  %5841 = inttoptr i64 %5840 to ptr
  %5842 = extractelement <2 x i64> %5839, i32 1
  %5843 = inttoptr i64 %5842 to ptr
  %idxprom.i.i.i.i4449 = zext i32 %spec.store.select.i.i.i4437 to i64
  %arrayidx.i.i.i.i4450 = getelementptr inbounds nuw i32, ptr %5841, i64 %idxprom.i.i.i.i4449, !amdgpu.uniform !42
  %5844 = load atomic i32, ptr %arrayidx.i.i.i.i4450 monotonic, align 4
  %5845 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %5844)
  %arrayidx.i45.i.i.i4451 = getelementptr inbounds nuw i32, ptr %5843, i64 %idxprom.i.i.i.i4449, !amdgpu.uniform !42
  %5846 = load atomic i32, ptr %arrayidx.i45.i.i.i4451 monotonic, align 4
  %5847 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %5846)
  %cmp.i46.not.i.i.i4452 = icmp ne i32 %5845, %5847
  br i1 %cmp.i46.not.i.i.i4452, label %if.then12.i.i.i4453, label %Flow7686, !amdgpu.uniform !42

if.then12.i.i.i4453:                              ; preds = %if.then.i.i.i.i4448
  fence syncscope("agent") release
  %5848 = call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %5828, i1 true)
  %iszero.i.i.i.i.i.i.i4454 = icmp ne i64 %5828, 0
  %cmp2.i.i.i.i.i.i4455 = icmp eq i64 %5848, %sh_prom.i.i.i.i3270
  %cmp.i.i.i.i.i.i4456 = select i1 %iszero.i.i.i.i.i.i.i4454, i1 %cmp2.i.i.i.i.i.i4455, i1 false
  %conv4.i.i51.i.i.i4457 = zext i1 %cmp.i.i.i.i.i.i4456 to i32
  %shl.i.i52.i.i.i4458 = shl nuw i32 %conv4.i.i51.i.i.i4457, %5829
  %xor.i.i.i.i.i4459 = xor i32 %shl.i.i52.i.i.i4458, -1
  %5849 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop5975, !amdgpu.uniform !42

5850:                                             ; preds = %ComputeEnd5976
  %sunkaddr7503 = getelementptr inbounds i8, ptr addrspace(1) %8804, i64 40
  %5851 = atomicrmw and ptr addrspace(1) %sunkaddr7503, i32 %.lcssa8964 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %5852, !amdgpu.uniform !42

5852:                                             ; preds = %5850, %ComputeEnd5976
  call void @llvm.amdgcn.end.cf.i64(i64 %8831)
  call void @llvm.amdgcn.wave.barrier()
  br label %Flow7686, !amdgpu.uniform !42

if.end14.i.i.i4519:                               ; preds = %Flow7686
  %5853 = call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %5828, i1 true)
  %iszero.i.i.i.i.i.i4520 = icmp ne i64 %5828, 0
  %cmp2.i.i.i.i.i4521 = icmp eq i64 %5853, %sh_prom.i.i.i.i3270
  %cmp.i.i.i.i.i4522 = select i1 %iszero.i.i.i.i.i.i4520, i1 %cmp2.i.i.i.i.i4521, i1 false
  %5854 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.i.i.i.i.i4522)
  %5855 = extractvalue { i1, i64 } %5854, 0
  %5856 = extractvalue { i1, i64 } %5854, 1
  br i1 %5855, label %if.then16.i.i.i4524, label %if.end22.i.i.i4523

if.then16.i.i.i4524:                              ; preds = %if.end14.i.i.i4519
  %5857 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !76
  %opcode.i.i.i4525 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %5857, i64 %idxprom.i.i.i.i4449, i32 1
  store i32 1660944387, ptr %opcode.i.i.i4525, align 8, !tbaa !77
  %arrayidx21.i.i.i4526 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %5857, i64 %idxprom.i.i.i.i4449
  store i64 %5828, ptr %arrayidx21.i.i.i4526, align 8, !tbaa !80
  br label %if.end22.i.i.i4523, !amdgpu.uniform !42

Flow7687:                                         ; preds = %if.end22.i.i.i4523, %Flow7686
  %5858 = phi i32 [ %5845, %if.end22.i.i.i4523 ], [ %retval.sroa.8.0.i.i.i4434, %Flow7686 ]
  %5859 = phi i64 [ %5828, %if.end22.i.i.i4523 ], [ %retval.sroa.2.0.i.i.i4433, %Flow7686 ]
  %5860 = phi i32 [ %spec.store.select.i.i.i4437, %if.end22.i.i.i4523 ], [ %retval.sroa.6.0.i.i.i4431, %Flow7686 ]
  br label %cleanup26.i.i.i4460, !amdgpu.uniform !42

if.end22.i.i.i4523:                               ; preds = %if.then16.i.i.i4524, %if.end14.i.i.i4519
  call void @llvm.amdgcn.end.cf.i64(i64 %5856)
  call void @llvm.amdgcn.wave.barrier()
  br label %Flow7687, !amdgpu.uniform !42

Flow7688:                                         ; preds = %cleanup26.i.i.i4460, %5834
  %5861 = phi i32 [ %5860, %cleanup26.i.i.i4460 ], [ %5824, %5834 ]
  %5862 = phi i64 [ %5859, %cleanup26.i.i.i4460 ], [ %5825, %5834 ]
  %5863 = phi i32 [ %5858, %cleanup26.i.i.i4460 ], [ %5826, %5834 ]
  %5864 = phi i32 [ %5858, %cleanup26.i.i.i4460 ], [ %retval.sroa.8.0.i.i.i4434, %5834 ]
  %5865 = phi i64 [ %5859, %cleanup26.i.i.i4460 ], [ %retval.sroa.2.0.i.i.i4433, %5834 ]
  %5866 = phi i32 [ %5860, %cleanup26.i.i.i4460 ], [ %retval.sroa.6.0.i.i.i4431, %5834 ]
  %5867 = phi i1 [ %cmp.i46.not.i.i.i4452, %cleanup26.i.i.i4460 ], [ true, %5834 ]
  br i1 %5867, label %for.inc.i.i.i4465, label %Flow7689, !amdgpu.uniform !42

cleanup26.i.i.i4460:                              ; preds = %Flow7687
  br label %Flow7688, !amdgpu.uniform !42

for.inc.i.i.i4465:                                ; preds = %Flow7688
  %inc.i.i.i4470 = add i32 %spec.store.select.i.i.i4437, 1
  br label %Flow7689, !amdgpu.uniform !42

Flow7689:                                         ; preds = %for.inc.i.i.i4465, %Flow7688
  %5868 = phi i32 [ %inc.i.i.i4470, %for.inc.i.i.i4465 ], [ poison, %Flow7688 ]
  %5869 = phi i1 [ false, %for.inc.i.i.i4465 ], [ true, %Flow7688 ]
  br i1 %5869, label %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i4471, label %for.cond.i.i1.i4430, !amdgpu.uniform !42

_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i4471: ; preds = %Flow7689
  %.lcssa8967 = phi i32 [ %5861, %Flow7689 ]
  %.lcssa8966 = phi i64 [ %5862, %Flow7689 ]
  %.lcssa8965 = phi i32 [ %5863, %Flow7689 ]
  fence acquire
  %5870 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %mul.i.i.i.i.i.i4472 = shl i32 %.lcssa8967, 6
  %idxprom.i19.i.i.i.i.i4473 = zext i32 %mul.i.i.i.i.i.i4472 to i64
  %arrayidx.i20.i.i.i.i.i4474 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %5870, i64 %idxprom.i19.i.i.i.i.i4473
  %arrayidx.i22.i.i.i.i.i4475 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %arrayidx.i20.i.i.i.i.i4474, i64 %sh_prom.i.i.i.i3270
  store i64 sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.29.45 to ptr), i64 14) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.29.45 to ptr) to i64)), ptr %arrayidx.i22.i.i.i.i.i4475, align 8, !tbaa !82
  %spec.select.i.i.i.i.i.i.i4476 = call i64 @llvm.umin.i64(i64 sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.29.45 to ptr), i64 14) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.29.45 to ptr) to i64)), i64 56)
  %5871 = and i64 %spec.select.i.i.i.i.i.i.i4476, 15
  %5872 = sub i64 %spec.select.i.i.i.i.i.i.i4476, %5871
  %5873 = icmp ne i64 %5872, 0
  br i1 %5873, label %loop-memcpy-expansion5445.preheader, label %Flow7685, !amdgpu.uniform !42

loop-memcpy-expansion5445.preheader:              ; preds = %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i4471
  %5874 = shl nuw nsw i64 %sh_prom.i.i.i.i3270, 6
  %5875 = add nuw nsw i64 %5874, 8
  %scevgep6565 = getelementptr i8, ptr %5870, i64 %5875
  %5876 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i4473, 6
  %scevgep6566 = getelementptr i8, ptr %scevgep6565, i64 %5876
  br label %loop-memcpy-expansion5445, !amdgpu.uniform !42

Flow7685:                                         ; preds = %Flow7684, %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i4471
  br label %loop-memcpy-residual-header5448, !amdgpu.uniform !42

loop-memcpy-expansion5445:                        ; preds = %loop-memcpy-expansion5445.preheader, %loop-memcpy-expansion5445
  %loop-index5446 = phi i64 [ %5879, %loop-memcpy-expansion5445 ], [ 0, %loop-memcpy-expansion5445.preheader ]
  %5877 = getelementptr i8, ptr addrspace(4) @.str.29.45, i64 %loop-index5446, !amdgpu.uniform !42
  %5878 = load <4 x i32>, ptr addrspace(4) %5877, align 1
  %scevgep6567 = getelementptr i8, ptr %scevgep6566, i64 %loop-index5446
  store <4 x i32> %5878, ptr %scevgep6567, align 1
  %5879 = add i64 %loop-index5446, 16
  %5880 = icmp uge i64 %5879, %5872
  br i1 %5880, label %Flow7684, label %loop-memcpy-expansion5445, !amdgpu.uniform !42

Flow7683:                                         ; preds = %Flow7681, %Flow7682
  %5881 = phi i64 [ %6675, %Flow7681 ], [ %6672, %Flow7682 ]
  %5882 = phi i64 [ %6677, %Flow7681 ], [ %6673, %Flow7682 ]
  br label %post-loop-memcpy-expansion5444, !amdgpu.uniform !42

loop-memcpy-residual5447:                         ; preds = %loop-memcpy-residual5447.preheader, %loop-memcpy-residual5447
  %residual-loop-index5449 = phi i64 [ %5884, %loop-memcpy-residual5447 ], [ 0, %loop-memcpy-residual5447.preheader ]
  %scevgep6564 = getelementptr i8, ptr addrspace(4) @.str.29.45, i64 %residual-loop-index5449, !amdgpu.uniform !42
  %5883 = load i8, ptr addrspace(4) %scevgep6564, align 1
  %scevgep6563 = getelementptr i8, ptr %scevgep6562, i64 %residual-loop-index5449
  store i8 %5883, ptr %scevgep6563, align 1
  %5884 = add i64 %residual-loop-index5449, 1
  %5885 = icmp uge i64 %5884, %5871
  br i1 %5885, label %Flow7681, label %loop-memcpy-residual5447, !amdgpu.uniform !42

Flow7681:                                         ; preds = %loop-memcpy-residual5447
  br label %Flow7683, !amdgpu.uniform !42

post-loop-memcpy-expansion5444:                   ; preds = %Flow7683
  %tobool.not.i.i.i.i.i.i4478 = icmp eq i32 %.lcssa8965, 0
  %conv.i.i.i.i.i.i4479 = zext i1 %tobool.not.i.i.i.i.i.i4478 to i32
  fence release
  %5886 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %idxprom.i23.i.i.i.i.i4480 = zext i32 %.lcssa8967 to i64
  %arrayidx.i24.i.i.i.i.i4481 = getelementptr inbounds nuw i32, ptr %5886, i64 %idxprom.i23.i.i.i.i.i4480
  store atomic i32 %conv.i.i.i.i.i.i4479, ptr %arrayidx.i24.i.i.i.i.i4481 monotonic, align 4
  %5887 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !84
  %arrayidx.i.i5.i.i4482 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %5887, i64 %idxprom.i23.i.i.i.i.i4480, !amdgpu.uniform !42
  %5888 = load i64, ptr %arrayidx.i.i5.i.i4482, align 8, !tbaa !80
  br label %while.cond.i.i.i.i4483, !amdgpu.uniform !42

while.cond.i.i.i.i4483:                           ; preds = %post-loop-memcpy-expansion5444, %Flow7680
  %phi.broken8767 = phi i64 [ 0, %post-loop-memcpy-expansion5444 ], [ %5905, %Flow7680 ]
  %lsr.iv6554 = phi i64 [ %5903, %Flow7680 ], [ -42, %post-loop-memcpy-expansion5444 ]
  %lsr.iv6549 = phi ptr addrspace(4) [ %5902, %Flow7680 ], [ getelementptr (i8, ptr addrspace(4) @.str.29.45, i64 56), %post-loop-memcpy-expansion5444 ]
  %port.sroa.37.0.i.i4484 = phi i32 [ %5901, %Flow7680 ], [ %conv.i.i.i.i.i.i4479, %post-loop-memcpy-expansion5444 ]
  %idx.0.i.i.i.i4485 = phi i64 [ %5900, %Flow7680 ], [ 56, %post-loop-memcpy-expansion5444 ]
  %cmp.i.i.i.i4486 = icmp ult i64 %idx.0.i.i.i.i4485, sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.29.45 to ptr), i64 14) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.29.45 to ptr) to i64))
  %5889 = call i64 @llvm.amdgcn.ballot.i64(i1 %cmp.i.i.i.i4486)
  %and.i.i.i.i6.i.i4487 = and i64 %5889, %5888
  %tobool.not.i.i.i.i4488 = icmp ne i64 %and.i.i.i.i6.i.i4487, 0
  %5890 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i22.i.i.i4489 = getelementptr inbounds nuw i32, ptr %5890, i64 %idxprom.i23.i.i.i.i.i4480, !amdgpu.uniform !42
  %5891 = load atomic i32, ptr %arrayidx.i22.i.i.i4489 monotonic, align 4
  %5892 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %tobool.not.i.i.i.i4488)
  %5893 = extractvalue { i1, i64 } %5892, 0
  %5894 = extractvalue { i1, i64 } %5892, 1
  br i1 %5893, label %cond.false.i8.i.i.i.i4490, label %Flow7680

cond.false.i8.i.i.i.i4490:                        ; preds = %while.cond.i.i.i.i4483
  %5895 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %5891)
  %5896 = icmp ne i32 %5895, %port.sroa.37.0.i.i4484
  br i1 %5896, label %while.body.i.i21.i.i.i.i4491.preheader, label %Flow7679, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i4491.preheader:           ; preds = %cond.false.i8.i.i.i.i4490
  br label %while.body.i.i21.i.i.i.i4491, !amdgpu.uniform !42

Flow7679:                                         ; preds = %Flow7678, %cond.false.i8.i.i.i.i4490
  br label %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i4494, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i4491:                     ; preds = %while.body.i.i21.i.i.i.i4491.preheader, %while.body.i.i21.i.i.i.i4491
  call void @llvm.amdgcn.s.sleep(i32 2)
  %5897 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i22.i.i.i.i4492 = getelementptr inbounds nuw i32, ptr %5897, i64 %idxprom.i23.i.i.i.i.i4480, !amdgpu.uniform !42
  %5898 = load atomic i32, ptr %arrayidx.i.i.i22.i.i.i.i4492 monotonic, align 4
  %5899 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %5898)
  %cmp.i.not.i.i23.i.i.i.i4493 = icmp eq i32 %5899, %port.sroa.37.0.i.i4484
  br i1 %cmp.i.not.i.i23.i.i.i.i4493, label %Flow7678, label %while.body.i.i21.i.i.i.i4491, !amdgpu.uniform !42

Flow7678:                                         ; preds = %while.body.i.i21.i.i.i.i4491
  br label %Flow7679, !amdgpu.uniform !42

Flow7680:                                         ; preds = %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i4495, %while.cond.i.i.i.i4483
  %5900 = phi i64 [ %add.i.i.i.i4498, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i4495 ], [ poison, %while.cond.i.i.i.i4483 ]
  %5901 = phi i32 [ %conv.i.i26.i.i.i.i4496, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i4495 ], [ poison, %while.cond.i.i.i.i4483 ]
  %5902 = phi ptr addrspace(4) [ %scevgep6550, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i4495 ], [ poison, %while.cond.i.i.i.i4483 ]
  %5903 = phi i64 [ %lsr.iv.next6555, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i4495 ], [ poison, %while.cond.i.i.i.i4483 ]
  %5904 = phi i1 [ false, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i4495 ], [ true, %while.cond.i.i.i.i4483 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %5894)
  %5905 = call i64 @llvm.amdgcn.if.break.i64(i1 %5904, i64 %phi.broken8767)
  %5906 = call i1 @llvm.amdgcn.loop.i64(i64 %5905)
  br i1 %5906, label %cond.false.i.i.i4505, label %while.cond.i.i.i.i4483

_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i4494: ; preds = %Flow7679
  fence acquire
  br i1 %cmp.i.i.i.i4486, label %if.then.i.i.i.i.i.i.i4499, label %Flow7677, !amdgpu.uniform !42

if.then.i.i.i.i.i.i.i4499:                        ; preds = %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i4494
  %sub.i.i.i.i.i.i.i4501 = sub nuw i64 sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.29.45 to ptr), i64 14) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.29.45 to ptr) to i64)), %idx.0.i.i.i.i4485
  %spec.select.i.i.i36.i.i.i.i4502 = call i64 @llvm.umin.i64(i64 %sub.i.i.i.i.i.i.i4501, i64 64)
  %5907 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %5908 = and i64 %spec.select.i.i.i36.i.i.i.i4502, 15
  %5909 = sub i64 %spec.select.i.i.i36.i.i.i.i4502, %5908
  %5910 = icmp ne i64 %5909, 0
  br i1 %5910, label %loop-memcpy-expansion5451.preheader, label %Flow7676, !amdgpu.uniform !42

loop-memcpy-expansion5451.preheader:              ; preds = %if.then.i.i.i.i.i.i.i4499
  %arrayidx.i20.i35.i.i.i.i4503 = getelementptr %"struct.rpc::Buffer", ptr %5907, i64 %idxprom.i19.i.i.i.i.i4473
  %arrayidx.i22.i38.i.i.i.i4504 = getelementptr %"struct.rpc::Buffer", ptr %arrayidx.i20.i35.i.i.i.i4503, i64 %sh_prom.i.i.i.i3270
  br label %loop-memcpy-expansion5451, !amdgpu.uniform !42

Flow7676:                                         ; preds = %Flow7675, %if.then.i.i.i.i.i.i.i4499
  br label %loop-memcpy-residual-header5454, !amdgpu.uniform !42

loop-memcpy-expansion5451:                        ; preds = %loop-memcpy-expansion5451.preheader, %loop-memcpy-expansion5451
  %loop-index5452 = phi i64 [ %5913, %loop-memcpy-expansion5451 ], [ 0, %loop-memcpy-expansion5451.preheader ]
  %scevgep6551 = getelementptr i8, ptr addrspace(4) %lsr.iv6549, i64 %loop-index5452, !amdgpu.uniform !42
  %5911 = load <4 x i32>, ptr addrspace(4) %scevgep6551, align 1
  %5912 = getelementptr i8, ptr %arrayidx.i22.i38.i.i.i.i4504, i64 %loop-index5452
  store <4 x i32> %5911, ptr %5912, align 1
  %5913 = add i64 %loop-index5452, 16
  %5914 = icmp uge i64 %5913, %5909
  br i1 %5914, label %Flow7675, label %loop-memcpy-expansion5451, !amdgpu.uniform !42

Flow7674:                                         ; preds = %Flow7673, %loop-memcpy-residual-header5454
  br label %Flow7677, !amdgpu.uniform !42

loop-memcpy-residual5453:                         ; preds = %loop-memcpy-residual5453.preheader, %loop-memcpy-residual5453
  %residual-loop-index5455 = phi i64 [ %5916, %loop-memcpy-residual5453 ], [ 0, %loop-memcpy-residual5453.preheader ]
  %scevgep6560 = getelementptr i8, ptr addrspace(4) %scevgep6559, i64 %residual-loop-index5455, !amdgpu.uniform !42
  %5915 = load i8, ptr addrspace(4) %scevgep6560, align 1
  %scevgep6558 = getelementptr i8, ptr %scevgep6557, i64 %residual-loop-index5455
  store i8 %5915, ptr %scevgep6558, align 1
  %5916 = add i64 %residual-loop-index5455, 1
  %5917 = icmp uge i64 %5916, %5908
  br i1 %5917, label %Flow7673, label %loop-memcpy-residual5453, !amdgpu.uniform !42

Flow7673:                                         ; preds = %loop-memcpy-residual5453
  br label %Flow7674, !amdgpu.uniform !42

_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i4495: ; preds = %Flow7677
  %conv.i.i26.i.i.i.i4496 = xor i32 %port.sroa.37.0.i.i4484, 1
  fence release
  %5918 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %arrayidx.i24.i29.i.i.i.i4497 = getelementptr inbounds nuw i32, ptr %5918, i64 %idxprom.i23.i.i.i.i.i4480
  store atomic i32 %conv.i.i26.i.i.i.i4496, ptr %arrayidx.i24.i29.i.i.i.i4497 monotonic, align 4
  %add.i.i.i.i4498 = add i64 %idx.0.i.i.i.i4485, 64
  %scevgep6550 = getelementptr i8, ptr addrspace(4) %lsr.iv6549, i64 64
  %lsr.iv.next6555 = add i64 %lsr.iv6554, -64
  br label %Flow7680, !amdgpu.uniform !42

cond.false.i.i.i4505:                             ; preds = %Flow7680
  %.lcssa8961 = phi i64 [ %5905, %Flow7680 ]
  %port.sroa.37.0.i.i4484.lcssa = phi i32 [ %port.sroa.37.0.i.i4484, %Flow7680 ]
  %.lcssa8960 = phi i32 [ %5891, %Flow7680 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa8961)
  %5919 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %.lcssa8960)
  %5920 = icmp ne i32 %5919, %port.sroa.37.0.i.i4484.lcssa
  %5921 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %5920)
  %5922 = extractvalue { i1, i64 } %5921, 0
  %5923 = extractvalue { i1, i64 } %5921, 1
  br i1 %5922, label %while.body.i.i.i.i4506.preheader, label %Flow7672

while.body.i.i.i.i4506.preheader:                 ; preds = %cond.false.i.i.i4505
  br label %while.body.i.i.i.i4506, !amdgpu.uniform !42

Flow7672:                                         ; preds = %Flow7671, %cond.false.i.i.i4505
  call void @llvm.amdgcn.end.cf.i64(i64 %5923)
  br label %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit4527, !amdgpu.uniform !42

while.body.i.i.i.i4506:                           ; preds = %while.body.i.i.i.i4506.preheader, %while.body.i.i.i.i4506
  %phi.broken8768 = phi i64 [ 0, %while.body.i.i.i.i4506.preheader ], [ %5927, %while.body.i.i.i.i4506 ]
  call void @llvm.amdgcn.s.sleep(i32 2)
  %5924 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i8.i.i4507 = getelementptr inbounds nuw i32, ptr %5924, i64 %idxprom.i23.i.i.i.i.i4480, !amdgpu.uniform !42
  %5925 = load atomic i32, ptr %arrayidx.i.i.i8.i.i4507 monotonic, align 4
  %5926 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %5925)
  %cmp.i.not.i.i.i.i4508 = icmp eq i32 %5926, %port.sroa.37.0.i.i4484.lcssa
  %5927 = call i64 @llvm.amdgcn.if.break.i64(i1 %cmp.i.not.i.i.i.i4508, i64 %phi.broken8768)
  %5928 = call i1 @llvm.amdgcn.loop.i64(i64 %5927)
  br i1 %5928, label %Flow7671, label %while.body.i.i.i.i4506

Flow7671:                                         ; preds = %while.body.i.i.i.i4506
  %.lcssa8959 = phi i64 [ %5927, %while.body.i.i.i.i4506 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa8959)
  br label %Flow7672, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit4527: ; preds = %Flow7672
  fence acquire
  call void @llvm.amdgcn.wave.barrier()
  fence syncscope("agent") release
  %5929 = call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %.lcssa8966, i1 true)
  %iszero.i.i.i.i.i13.i.i4510 = icmp ne i64 %.lcssa8966, 0
  %cmp2.i.i.i.i14.i.i4511 = icmp eq i64 %5929, %sh_prom.i.i.i.i3270
  %cmp.i.i.i.i15.i.i4512 = select i1 %iszero.i.i.i.i.i13.i.i4510, i1 %cmp2.i.i.i.i14.i.i4511, i1 false
  %5930 = and i32 %.lcssa8967, 31
  %conv4.i.i.i19.i.i4516 = zext i1 %cmp.i.i.i.i15.i.i4512 to i32
  %shl.i.i.i20.i.i4517 = shl nuw i32 %conv4.i.i.i19.i.i4516, %5930
  %xor.i.i.i21.i.i4518 = xor i32 %shl.i.i.i20.i.i4517, -1
  %5931 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop5979, !amdgpu.uniform !42

5932:                                             ; preds = %ComputeEnd5980
  %div8.i.i.i16.i.i4513 = lshr i32 %.lcssa8967, 5
  %idxprom.i.i.i17.i.i4514 = zext nneg i32 %div8.i.i.i16.i.i4513 to i64
  %5933 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i17.i.i4514
  %sunkaddr7504 = getelementptr inbounds i8, ptr addrspace(1) %5933, i64 40
  %5934 = atomicrmw and ptr addrspace(1) %sunkaddr7504, i32 %.lcssa8958 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %5935, !amdgpu.uniform !42

5935:                                             ; preds = %5932, %ComputeEnd5980
  call void @llvm.amdgcn.end.cf.i64(i64 %8848)
  call void @llvm.amdgcn.wave.barrier()
  %5936 = load ptr, ptr %Options, align 8, !tbaa !160
  %tobool104.not = icmp ne ptr %5936, null
  %5937 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %tobool104.not)
  %5938 = extractvalue { i1, i64 } %5937, 0
  %5939 = extractvalue { i1, i64 } %5937, 1
  br i1 %5938, label %for.cond.i.i1.i4536.preheader, label %Flow7670

for.cond.i.i1.i4536.preheader:                    ; preds = %5935
  br label %for.cond.i.i1.i4536, !amdgpu.uniform !42

Flow7670:                                         ; preds = %6181, %5935
  call void @llvm.amdgcn.end.cf.i64(i64 %5939)
  br label %Flow7916, !amdgpu.uniform !42

for.cond.i.i1.i4536:                              ; preds = %for.cond.i.i1.i4536.preheader, %Flow7669
  %5940 = phi i32 [ %5977, %Flow7669 ], [ poison, %for.cond.i.i1.i4536.preheader ]
  %5941 = phi i64 [ %5978, %Flow7669 ], [ poison, %for.cond.i.i1.i4536.preheader ]
  %5942 = phi i32 [ %5979, %Flow7669 ], [ poison, %for.cond.i.i1.i4536.preheader ]
  %retval.sroa.6.0.i.i.i4537 = phi i32 [ %5982, %Flow7669 ], [ undef, %for.cond.i.i1.i4536.preheader ]
  %retval.sroa.2.0.i.i.i4539 = phi i64 [ %5981, %Flow7669 ], [ undef, %for.cond.i.i1.i4536.preheader ]
  %retval.sroa.8.0.i.i.i4540 = phi i32 [ %5980, %Flow7669 ], [ undef, %for.cond.i.i1.i4536.preheader ]
  %index.0.i.i.i4541 = phi i32 [ %5984, %Flow7669 ], [ 0, %for.cond.i.i1.i4536.preheader ]
  %5943 = load i32, ptr addrspace(1) @__llvm_rpc_client, align 8, !tbaa !68
  %cmp.not.i.i.i4542 = icmp ult i32 %index.0.i.i.i4541, %5943
  %spec.store.select.i.i.i4543 = select i1 %cmp.not.i.i.i4542, i32 %index.0.i.i.i4541, i32 0
  %5944 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  %5945 = and i32 %spec.store.select.i.i.i4543, 31
  %5946 = lshr i64 %5944, %sh_prom.i.i.i.i3270
  %5947 = trunc i64 %5946 to i32
  %conv4.i.i.i.i.i4547 = and i32 %5947, 1
  %shl.i.i.i.i.i4548 = shl nuw i32 %conv4.i.i.i.i.i4547, %5945
  br label %ComputeLoop5983, !amdgpu.uniform !42

5948:                                             ; preds = %ComputeEnd5984
  %sunkaddr7505 = getelementptr inbounds i8, ptr addrspace(1) %8858, i64 40
  %5949 = atomicrmw or ptr addrspace(1) %sunkaddr7505, i32 %.lcssa8952 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %5950, !amdgpu.uniform !42

5950:                                             ; preds = %5948, %ComputeEnd5984
  %5951 = phi i32 [ %5949, %5948 ], [ poison, %ComputeEnd5984 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %8867)
  %5952 = call i32 @llvm.amdgcn.readfirstlane.i32(i32 %5951)
  %5953 = or i32 %5952, %.lcssa8953
  %shl5.i.i.i.i.i4549 = shl nuw i32 1, %5945
  %and.i.i.i.i.i4550 = and i32 %shl5.i.i.i.i.i4549, %5953
  %tobool3.i.i.i.i4551 = icmp ne i32 %and.i.i.i.i.i4550, 0
  %5954 = call i64 @llvm.amdgcn.ballot.i64(i1 %tobool3.i.i.i.i4551)
  %and.i.i.i.i.i.i4552 = and i64 %5954, %5944
  %cmp.i.not.i.i.i4553 = icmp ne i64 %5944, %and.i.i.i.i.i.i4552
  br i1 %cmp.i.not.i.i.i4553, label %if.then.i.i.i.i4554, label %Flow7668, !amdgpu.uniform !42

if.then.i.i.i.i4554:                              ; preds = %5950
  fence syncscope("agent") acquire
  %5955 = load <2 x i64>, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !75
  %5956 = extractelement <2 x i64> %5955, i32 0
  %5957 = inttoptr i64 %5956 to ptr
  %5958 = extractelement <2 x i64> %5955, i32 1
  %5959 = inttoptr i64 %5958 to ptr
  %idxprom.i.i.i.i4555 = zext i32 %spec.store.select.i.i.i4543 to i64
  %arrayidx.i.i.i.i4556 = getelementptr inbounds nuw i32, ptr %5957, i64 %idxprom.i.i.i.i4555, !amdgpu.uniform !42
  %5960 = load atomic i32, ptr %arrayidx.i.i.i.i4556 monotonic, align 4
  %5961 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %5960)
  %arrayidx.i45.i.i.i4557 = getelementptr inbounds nuw i32, ptr %5959, i64 %idxprom.i.i.i.i4555, !amdgpu.uniform !42
  %5962 = load atomic i32, ptr %arrayidx.i45.i.i.i4557 monotonic, align 4
  %5963 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %5962)
  %cmp.i46.not.i.i.i4558 = icmp ne i32 %5961, %5963
  br i1 %cmp.i46.not.i.i.i4558, label %if.then12.i.i.i4559, label %Flow7666, !amdgpu.uniform !42

if.then12.i.i.i4559:                              ; preds = %if.then.i.i.i.i4554
  fence syncscope("agent") release
  %5964 = call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %5944, i1 true)
  %iszero.i.i.i.i.i.i.i4560 = icmp ne i64 %5944, 0
  %cmp2.i.i.i.i.i.i4561 = icmp eq i64 %5964, %sh_prom.i.i.i.i3270
  %cmp.i.i.i.i.i.i4562 = select i1 %iszero.i.i.i.i.i.i.i4560, i1 %cmp2.i.i.i.i.i.i4561, i1 false
  %conv4.i.i51.i.i.i4563 = zext i1 %cmp.i.i.i.i.i.i4562 to i32
  %shl.i.i52.i.i.i4564 = shl nuw i32 %conv4.i.i51.i.i.i4563, %5945
  %xor.i.i.i.i.i4565 = xor i32 %shl.i.i52.i.i.i4564, -1
  %5965 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop5988, !amdgpu.uniform !42

5966:                                             ; preds = %ComputeEnd5989
  %sunkaddr7506 = getelementptr inbounds i8, ptr addrspace(1) %8858, i64 40
  %5967 = atomicrmw and ptr addrspace(1) %sunkaddr7506, i32 %.lcssa8954 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %5968, !amdgpu.uniform !42

5968:                                             ; preds = %5966, %ComputeEnd5989
  call void @llvm.amdgcn.end.cf.i64(i64 %8885)
  call void @llvm.amdgcn.wave.barrier()
  br label %Flow7666, !amdgpu.uniform !42

if.end14.i.i.i4625:                               ; preds = %Flow7666
  %5969 = call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %5944, i1 true)
  %iszero.i.i.i.i.i.i4626 = icmp ne i64 %5944, 0
  %cmp2.i.i.i.i.i4627 = icmp eq i64 %5969, %sh_prom.i.i.i.i3270
  %cmp.i.i.i.i.i4628 = select i1 %iszero.i.i.i.i.i.i4626, i1 %cmp2.i.i.i.i.i4627, i1 false
  %5970 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.i.i.i.i.i4628)
  %5971 = extractvalue { i1, i64 } %5970, 0
  %5972 = extractvalue { i1, i64 } %5970, 1
  br i1 %5971, label %if.then16.i.i.i4630, label %if.end22.i.i.i4629

if.then16.i.i.i4630:                              ; preds = %if.end14.i.i.i4625
  %5973 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !76
  %opcode.i.i.i4631 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %5973, i64 %idxprom.i.i.i.i4555, i32 1
  store i32 1660944387, ptr %opcode.i.i.i4631, align 8, !tbaa !77
  %arrayidx21.i.i.i4632 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %5973, i64 %idxprom.i.i.i.i4555
  store i64 %5944, ptr %arrayidx21.i.i.i4632, align 8, !tbaa !80
  br label %if.end22.i.i.i4629, !amdgpu.uniform !42

Flow7667:                                         ; preds = %if.end22.i.i.i4629, %Flow7666
  %5974 = phi i32 [ %5961, %if.end22.i.i.i4629 ], [ %retval.sroa.8.0.i.i.i4540, %Flow7666 ]
  %5975 = phi i64 [ %5944, %if.end22.i.i.i4629 ], [ %retval.sroa.2.0.i.i.i4539, %Flow7666 ]
  %5976 = phi i32 [ %spec.store.select.i.i.i4543, %if.end22.i.i.i4629 ], [ %retval.sroa.6.0.i.i.i4537, %Flow7666 ]
  br label %cleanup26.i.i.i4566, !amdgpu.uniform !42

if.end22.i.i.i4629:                               ; preds = %if.then16.i.i.i4630, %if.end14.i.i.i4625
  call void @llvm.amdgcn.end.cf.i64(i64 %5972)
  call void @llvm.amdgcn.wave.barrier()
  br label %Flow7667, !amdgpu.uniform !42

Flow7668:                                         ; preds = %cleanup26.i.i.i4566, %5950
  %5977 = phi i32 [ %5976, %cleanup26.i.i.i4566 ], [ %5940, %5950 ]
  %5978 = phi i64 [ %5975, %cleanup26.i.i.i4566 ], [ %5941, %5950 ]
  %5979 = phi i32 [ %5974, %cleanup26.i.i.i4566 ], [ %5942, %5950 ]
  %5980 = phi i32 [ %5974, %cleanup26.i.i.i4566 ], [ %retval.sroa.8.0.i.i.i4540, %5950 ]
  %5981 = phi i64 [ %5975, %cleanup26.i.i.i4566 ], [ %retval.sroa.2.0.i.i.i4539, %5950 ]
  %5982 = phi i32 [ %5976, %cleanup26.i.i.i4566 ], [ %retval.sroa.6.0.i.i.i4537, %5950 ]
  %5983 = phi i1 [ %cmp.i46.not.i.i.i4558, %cleanup26.i.i.i4566 ], [ true, %5950 ]
  br i1 %5983, label %for.inc.i.i.i4571, label %Flow7669, !amdgpu.uniform !42

cleanup26.i.i.i4566:                              ; preds = %Flow7667
  br label %Flow7668, !amdgpu.uniform !42

for.inc.i.i.i4571:                                ; preds = %Flow7668
  %inc.i.i.i4576 = add i32 %spec.store.select.i.i.i4543, 1
  br label %Flow7669, !amdgpu.uniform !42

Flow7669:                                         ; preds = %for.inc.i.i.i4571, %Flow7668
  %5984 = phi i32 [ %inc.i.i.i4576, %for.inc.i.i.i4571 ], [ poison, %Flow7668 ]
  %5985 = phi i1 [ false, %for.inc.i.i.i4571 ], [ true, %Flow7668 ]
  br i1 %5985, label %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i4577, label %for.cond.i.i1.i4536, !amdgpu.uniform !42

_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i4577: ; preds = %Flow7669
  %.lcssa8957 = phi i32 [ %5977, %Flow7669 ]
  %.lcssa8956 = phi i64 [ %5978, %Flow7669 ]
  %.lcssa8955 = phi i32 [ %5979, %Flow7669 ]
  fence acquire
  %5986 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %mul.i.i.i.i.i.i4578 = shl i32 %.lcssa8957, 6
  %idxprom.i19.i.i.i.i.i4579 = zext i32 %mul.i.i.i.i.i.i4578 to i64
  %arrayidx.i20.i.i.i.i.i4580 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %5986, i64 %idxprom.i19.i.i.i.i.i4579
  %arrayidx.i22.i.i.i.i.i4581 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %arrayidx.i20.i.i.i.i.i4580, i64 %sh_prom.i.i.i.i3270
  store i64 sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.30 to ptr), i64 21) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.30 to ptr) to i64)), ptr %arrayidx.i22.i.i.i.i.i4581, align 8, !tbaa !82
  %spec.select.i.i.i.i.i.i.i4582 = call i64 @llvm.umin.i64(i64 sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.30 to ptr), i64 21) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.30 to ptr) to i64)), i64 56)
  %5987 = and i64 %spec.select.i.i.i.i.i.i.i4582, 15
  %5988 = sub i64 %spec.select.i.i.i.i.i.i.i4582, %5987
  %5989 = icmp ne i64 %5988, 0
  br i1 %5989, label %loop-memcpy-expansion5457.preheader, label %Flow7665, !amdgpu.uniform !42

loop-memcpy-expansion5457.preheader:              ; preds = %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i4577
  %5990 = add nuw nsw i64 %5881, 8
  %scevgep6546 = getelementptr i8, ptr %5986, i64 %5990
  %5991 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i4579, 6
  %scevgep6547 = getelementptr i8, ptr %scevgep6546, i64 %5991
  br label %loop-memcpy-expansion5457, !amdgpu.uniform !42

Flow7665:                                         ; preds = %Flow7664, %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i4577
  br label %loop-memcpy-residual-header5460, !amdgpu.uniform !42

loop-memcpy-expansion5457:                        ; preds = %loop-memcpy-expansion5457.preheader, %loop-memcpy-expansion5457
  %loop-index5458 = phi i64 [ %5994, %loop-memcpy-expansion5457 ], [ 0, %loop-memcpy-expansion5457.preheader ]
  %5992 = getelementptr i8, ptr addrspace(4) @.str.30, i64 %loop-index5458, !amdgpu.uniform !42
  %5993 = load <4 x i32>, ptr addrspace(4) %5992, align 1
  %scevgep6548 = getelementptr i8, ptr %scevgep6547, i64 %loop-index5458
  store <4 x i32> %5993, ptr %scevgep6548, align 1
  %5994 = add i64 %loop-index5458, 16
  %5995 = icmp uge i64 %5994, %5988
  br i1 %5995, label %Flow7664, label %loop-memcpy-expansion5457, !amdgpu.uniform !42

Flow7663:                                         ; preds = %Flow7661, %Flow7662
  %5996 = phi i64 [ %6685, %Flow7661 ], [ %6682, %Flow7662 ]
  br label %post-loop-memcpy-expansion5456, !amdgpu.uniform !42

loop-memcpy-residual5459:                         ; preds = %loop-memcpy-residual5459.preheader, %loop-memcpy-residual5459
  %residual-loop-index5461 = phi i64 [ %5998, %loop-memcpy-residual5459 ], [ 0, %loop-memcpy-residual5459.preheader ]
  %scevgep6545 = getelementptr i8, ptr addrspace(4) getelementptr inbounds nuw (i8, ptr addrspace(4) @.str.30, i64 16), i64 %residual-loop-index5461, !amdgpu.uniform !42
  %5997 = load i8, ptr addrspace(4) %scevgep6545, align 1
  %scevgep6544 = getelementptr i8, ptr %scevgep6543, i64 %residual-loop-index5461
  store i8 %5997, ptr %scevgep6544, align 1
  %5998 = add i64 %residual-loop-index5461, 1
  %5999 = icmp uge i64 %5998, %5987
  br i1 %5999, label %Flow7661, label %loop-memcpy-residual5459, !amdgpu.uniform !42

Flow7661:                                         ; preds = %loop-memcpy-residual5459
  br label %Flow7663, !amdgpu.uniform !42

post-loop-memcpy-expansion5456:                   ; preds = %Flow7663
  %tobool.not.i.i.i.i.i.i4584 = icmp eq i32 %.lcssa8955, 0
  %conv.i.i.i.i.i.i4585 = zext i1 %tobool.not.i.i.i.i.i.i4584 to i32
  fence release
  %6000 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %idxprom.i23.i.i.i.i.i4586 = zext i32 %.lcssa8957 to i64
  %arrayidx.i24.i.i.i.i.i4587 = getelementptr inbounds nuw i32, ptr %6000, i64 %idxprom.i23.i.i.i.i.i4586
  store atomic i32 %conv.i.i.i.i.i.i4585, ptr %arrayidx.i24.i.i.i.i.i4587 monotonic, align 4
  %6001 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !84
  %arrayidx.i.i5.i.i4588 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %6001, i64 %idxprom.i23.i.i.i.i.i4586, !amdgpu.uniform !42
  %6002 = load i64, ptr %arrayidx.i.i5.i.i4588, align 8, !tbaa !80
  br label %while.cond.i.i.i.i4589, !amdgpu.uniform !42

while.cond.i.i.i.i4589:                           ; preds = %post-loop-memcpy-expansion5456, %Flow7660
  %phi.broken8769 = phi i64 [ 0, %post-loop-memcpy-expansion5456 ], [ %6019, %Flow7660 ]
  %lsr.iv6535 = phi i64 [ %6017, %Flow7660 ], [ -35, %post-loop-memcpy-expansion5456 ]
  %lsr.iv6530 = phi ptr addrspace(4) [ %6016, %Flow7660 ], [ getelementptr (i8, ptr addrspace(4) @.str.30, i64 56), %post-loop-memcpy-expansion5456 ]
  %port.sroa.37.0.i.i4590 = phi i32 [ %6015, %Flow7660 ], [ %conv.i.i.i.i.i.i4585, %post-loop-memcpy-expansion5456 ]
  %idx.0.i.i.i.i4591 = phi i64 [ %6014, %Flow7660 ], [ 56, %post-loop-memcpy-expansion5456 ]
  %cmp.i.i.i.i4592 = icmp ult i64 %idx.0.i.i.i.i4591, sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.30 to ptr), i64 21) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.30 to ptr) to i64))
  %6003 = call i64 @llvm.amdgcn.ballot.i64(i1 %cmp.i.i.i.i4592)
  %and.i.i.i.i6.i.i4593 = and i64 %6003, %6002
  %tobool.not.i.i.i.i4594 = icmp ne i64 %and.i.i.i.i6.i.i4593, 0
  %6004 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i22.i.i.i4595 = getelementptr inbounds nuw i32, ptr %6004, i64 %idxprom.i23.i.i.i.i.i4586, !amdgpu.uniform !42
  %6005 = load atomic i32, ptr %arrayidx.i22.i.i.i4595 monotonic, align 4
  %6006 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %tobool.not.i.i.i.i4594)
  %6007 = extractvalue { i1, i64 } %6006, 0
  %6008 = extractvalue { i1, i64 } %6006, 1
  br i1 %6007, label %cond.false.i8.i.i.i.i4596, label %Flow7660

cond.false.i8.i.i.i.i4596:                        ; preds = %while.cond.i.i.i.i4589
  %6009 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %6005)
  %6010 = icmp ne i32 %6009, %port.sroa.37.0.i.i4590
  br i1 %6010, label %while.body.i.i21.i.i.i.i4597.preheader, label %Flow7659, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i4597.preheader:           ; preds = %cond.false.i8.i.i.i.i4596
  br label %while.body.i.i21.i.i.i.i4597, !amdgpu.uniform !42

Flow7659:                                         ; preds = %Flow7658, %cond.false.i8.i.i.i.i4596
  br label %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i4600, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i4597:                     ; preds = %while.body.i.i21.i.i.i.i4597.preheader, %while.body.i.i21.i.i.i.i4597
  call void @llvm.amdgcn.s.sleep(i32 2)
  %6011 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i22.i.i.i.i4598 = getelementptr inbounds nuw i32, ptr %6011, i64 %idxprom.i23.i.i.i.i.i4586, !amdgpu.uniform !42
  %6012 = load atomic i32, ptr %arrayidx.i.i.i22.i.i.i.i4598 monotonic, align 4
  %6013 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %6012)
  %cmp.i.not.i.i23.i.i.i.i4599 = icmp eq i32 %6013, %port.sroa.37.0.i.i4590
  br i1 %cmp.i.not.i.i23.i.i.i.i4599, label %Flow7658, label %while.body.i.i21.i.i.i.i4597, !amdgpu.uniform !42

Flow7658:                                         ; preds = %while.body.i.i21.i.i.i.i4597
  br label %Flow7659, !amdgpu.uniform !42

Flow7660:                                         ; preds = %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i4601, %while.cond.i.i.i.i4589
  %6014 = phi i64 [ %add.i.i.i.i4604, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i4601 ], [ poison, %while.cond.i.i.i.i4589 ]
  %6015 = phi i32 [ %conv.i.i26.i.i.i.i4602, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i4601 ], [ poison, %while.cond.i.i.i.i4589 ]
  %6016 = phi ptr addrspace(4) [ %scevgep6531, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i4601 ], [ poison, %while.cond.i.i.i.i4589 ]
  %6017 = phi i64 [ %lsr.iv.next6536, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i4601 ], [ poison, %while.cond.i.i.i.i4589 ]
  %6018 = phi i1 [ false, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i4601 ], [ true, %while.cond.i.i.i.i4589 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %6008)
  %6019 = call i64 @llvm.amdgcn.if.break.i64(i1 %6018, i64 %phi.broken8769)
  %6020 = call i1 @llvm.amdgcn.loop.i64(i64 %6019)
  br i1 %6020, label %cond.false.i.i.i4611, label %while.cond.i.i.i.i4589

_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i4600: ; preds = %Flow7659
  fence acquire
  br i1 %cmp.i.i.i.i4592, label %if.then.i.i.i.i.i.i.i4605, label %Flow7657, !amdgpu.uniform !42

if.then.i.i.i.i.i.i.i4605:                        ; preds = %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i4600
  %sub.i.i.i.i.i.i.i4607 = sub nuw i64 sub (i64 ptrtoint (ptr getelementptr (i8, ptr addrspacecast (ptr addrspace(4) @.str.30 to ptr), i64 21) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(4) @.str.30 to ptr) to i64)), %idx.0.i.i.i.i4591
  %spec.select.i.i.i36.i.i.i.i4608 = call i64 @llvm.umin.i64(i64 %sub.i.i.i.i.i.i.i4607, i64 64)
  %6021 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %6022 = and i64 %spec.select.i.i.i36.i.i.i.i4608, 15
  %6023 = sub i64 %spec.select.i.i.i36.i.i.i.i4608, %6022
  %6024 = icmp ne i64 %6023, 0
  br i1 %6024, label %loop-memcpy-expansion5463.preheader, label %Flow7656, !amdgpu.uniform !42

loop-memcpy-expansion5463.preheader:              ; preds = %if.then.i.i.i.i.i.i.i4605
  %arrayidx.i20.i35.i.i.i.i4609 = getelementptr %"struct.rpc::Buffer", ptr %6021, i64 %idxprom.i19.i.i.i.i.i4579
  %arrayidx.i22.i38.i.i.i.i4610 = getelementptr %"struct.rpc::Buffer", ptr %arrayidx.i20.i35.i.i.i.i4609, i64 %sh_prom.i.i.i.i3270
  br label %loop-memcpy-expansion5463, !amdgpu.uniform !42

Flow7656:                                         ; preds = %Flow7655, %if.then.i.i.i.i.i.i.i4605
  br label %loop-memcpy-residual-header5466, !amdgpu.uniform !42

loop-memcpy-expansion5463:                        ; preds = %loop-memcpy-expansion5463.preheader, %loop-memcpy-expansion5463
  %loop-index5464 = phi i64 [ %6027, %loop-memcpy-expansion5463 ], [ 0, %loop-memcpy-expansion5463.preheader ]
  %scevgep6532 = getelementptr i8, ptr addrspace(4) %lsr.iv6530, i64 %loop-index5464, !amdgpu.uniform !42
  %6025 = load <4 x i32>, ptr addrspace(4) %scevgep6532, align 1
  %6026 = getelementptr i8, ptr %arrayidx.i22.i38.i.i.i.i4610, i64 %loop-index5464
  store <4 x i32> %6025, ptr %6026, align 1
  %6027 = add i64 %loop-index5464, 16
  %6028 = icmp uge i64 %6027, %6023
  br i1 %6028, label %Flow7655, label %loop-memcpy-expansion5463, !amdgpu.uniform !42

Flow7654:                                         ; preds = %Flow7653, %loop-memcpy-residual-header5466
  br label %Flow7657, !amdgpu.uniform !42

loop-memcpy-residual5465:                         ; preds = %loop-memcpy-residual5465.preheader, %loop-memcpy-residual5465
  %residual-loop-index5467 = phi i64 [ %6030, %loop-memcpy-residual5465 ], [ 0, %loop-memcpy-residual5465.preheader ]
  %scevgep6541 = getelementptr i8, ptr addrspace(4) %scevgep6540, i64 %residual-loop-index5467, !amdgpu.uniform !42
  %6029 = load i8, ptr addrspace(4) %scevgep6541, align 1
  %scevgep6539 = getelementptr i8, ptr %scevgep6538, i64 %residual-loop-index5467
  store i8 %6029, ptr %scevgep6539, align 1
  %6030 = add i64 %residual-loop-index5467, 1
  %6031 = icmp uge i64 %6030, %6022
  br i1 %6031, label %Flow7653, label %loop-memcpy-residual5465, !amdgpu.uniform !42

Flow7653:                                         ; preds = %loop-memcpy-residual5465
  br label %Flow7654, !amdgpu.uniform !42

_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i4601: ; preds = %Flow7657
  %conv.i.i26.i.i.i.i4602 = xor i32 %port.sroa.37.0.i.i4590, 1
  fence release
  %6032 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %arrayidx.i24.i29.i.i.i.i4603 = getelementptr inbounds nuw i32, ptr %6032, i64 %idxprom.i23.i.i.i.i.i4586
  store atomic i32 %conv.i.i26.i.i.i.i4602, ptr %arrayidx.i24.i29.i.i.i.i4603 monotonic, align 4
  %add.i.i.i.i4604 = add i64 %idx.0.i.i.i.i4591, 64
  %scevgep6531 = getelementptr i8, ptr addrspace(4) %lsr.iv6530, i64 64
  %lsr.iv.next6536 = add i64 %lsr.iv6535, -64
  br label %Flow7660, !amdgpu.uniform !42

cond.false.i.i.i4611:                             ; preds = %Flow7660
  %.lcssa8951 = phi i64 [ %6019, %Flow7660 ]
  %port.sroa.37.0.i.i4590.lcssa = phi i32 [ %port.sroa.37.0.i.i4590, %Flow7660 ]
  %.lcssa8950 = phi i32 [ %6005, %Flow7660 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa8951)
  %6033 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %.lcssa8950)
  %6034 = icmp ne i32 %6033, %port.sroa.37.0.i.i4590.lcssa
  %6035 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %6034)
  %6036 = extractvalue { i1, i64 } %6035, 0
  %6037 = extractvalue { i1, i64 } %6035, 1
  br i1 %6036, label %while.body.i.i.i.i4612.preheader, label %Flow7652

while.body.i.i.i.i4612.preheader:                 ; preds = %cond.false.i.i.i4611
  br label %while.body.i.i.i.i4612, !amdgpu.uniform !42

Flow7652:                                         ; preds = %Flow7651, %cond.false.i.i.i4611
  call void @llvm.amdgcn.end.cf.i64(i64 %6037)
  br label %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit4633, !amdgpu.uniform !42

while.body.i.i.i.i4612:                           ; preds = %while.body.i.i.i.i4612.preheader, %while.body.i.i.i.i4612
  %phi.broken8770 = phi i64 [ 0, %while.body.i.i.i.i4612.preheader ], [ %6041, %while.body.i.i.i.i4612 ]
  call void @llvm.amdgcn.s.sleep(i32 2)
  %6038 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i8.i.i4613 = getelementptr inbounds nuw i32, ptr %6038, i64 %idxprom.i23.i.i.i.i.i4586, !amdgpu.uniform !42
  %6039 = load atomic i32, ptr %arrayidx.i.i.i8.i.i4613 monotonic, align 4
  %6040 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %6039)
  %cmp.i.not.i.i.i.i4614 = icmp eq i32 %6040, %port.sroa.37.0.i.i4590.lcssa
  %6041 = call i64 @llvm.amdgcn.if.break.i64(i1 %cmp.i.not.i.i.i.i4614, i64 %phi.broken8770)
  %6042 = call i1 @llvm.amdgcn.loop.i64(i64 %6041)
  br i1 %6042, label %Flow7651, label %while.body.i.i.i.i4612

Flow7651:                                         ; preds = %while.body.i.i.i.i4612
  %.lcssa8949 = phi i64 [ %6041, %while.body.i.i.i.i4612 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa8949)
  br label %Flow7652, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit4633: ; preds = %Flow7652
  fence acquire
  call void @llvm.amdgcn.wave.barrier()
  fence syncscope("agent") release
  %6043 = call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %.lcssa8956, i1 true)
  %iszero.i.i.i.i.i13.i.i4616 = icmp ne i64 %.lcssa8956, 0
  %cmp2.i.i.i.i14.i.i4617 = icmp eq i64 %6043, %sh_prom.i.i.i.i3270
  %cmp.i.i.i.i15.i.i4618 = select i1 %iszero.i.i.i.i.i13.i.i4616, i1 %cmp2.i.i.i.i14.i.i4617, i1 false
  %6044 = and i32 %.lcssa8957, 31
  %conv4.i.i.i19.i.i4622 = zext i1 %cmp.i.i.i.i15.i.i4618 to i32
  %shl.i.i.i20.i.i4623 = shl nuw i32 %conv4.i.i.i19.i.i4622, %6044
  %xor.i.i.i21.i.i4624 = xor i32 %shl.i.i.i20.i.i4623, -1
  %6045 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop5992, !amdgpu.uniform !42

6046:                                             ; preds = %ComputeEnd5993
  %div8.i.i.i16.i.i4619 = lshr i32 %.lcssa8957, 5
  %idxprom.i.i.i17.i.i4620 = zext nneg i32 %div8.i.i.i16.i.i4619 to i64
  %6047 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i17.i.i4620
  %sunkaddr7507 = getelementptr inbounds i8, ptr addrspace(1) %6047, i64 40
  %6048 = atomicrmw and ptr addrspace(1) %sunkaddr7507, i32 %.lcssa8948 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %6049, !amdgpu.uniform !42

6049:                                             ; preds = %6046, %ComputeEnd5993
  call void @llvm.amdgcn.end.cf.i64(i64 %8902)
  call void @llvm.amdgcn.wave.barrier()
  %6050 = load ptr, ptr %Options, align 8, !tbaa !160
  br label %for.cond.i.i.i4634, !amdgpu.uniform !42

for.cond.i.i.i4634:                               ; preds = %6049, %for.cond.i.i.i4634
  %phi.broken8771 = phi i64 [ 0, %6049 ], [ %6052, %for.cond.i.i.i4634 ]
  %lsr.iv6516 = phi i64 [ %lsr.iv.next6517, %for.cond.i.i.i4634 ], [ -1, %6049 ]
  %scevgep6528 = getelementptr i8, ptr %6050, i64 %lsr.iv6516
  %scevgep6529 = getelementptr i8, ptr %scevgep6528, i64 1
  %6051 = load i8, ptr %scevgep6529, align 1, !tbaa !57
  %cmp.i.i.i4636 = icmp eq i8 %6051, 0
  %lsr.iv.next6517 = add i64 %lsr.iv6516, 1
  %6052 = call i64 @llvm.amdgcn.if.break.i64(i1 %cmp.i.i.i4636, i64 %phi.broken8771)
  %6053 = call i1 @llvm.amdgcn.loop.i64(i64 %6052)
  br i1 %6053, label %for.cond.i.i1.i4643.preheader, label %for.cond.i.i.i4634

for.cond.i.i1.i4643.preheader:                    ; preds = %for.cond.i.i.i4634
  %lsr.iv.next6517.lcssa = phi i64 [ %lsr.iv.next6517, %for.cond.i.i.i4634 ]
  %.lcssa8947 = phi i64 [ %6052, %for.cond.i.i.i4634 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa8947)
  br label %for.cond.i.i1.i4643, !amdgpu.uniform !42

for.cond.i.i1.i4643:                              ; preds = %for.cond.i.i1.i4643.preheader, %Flow7650
  %6054 = phi i32 [ %6091, %Flow7650 ], [ poison, %for.cond.i.i1.i4643.preheader ]
  %6055 = phi i64 [ %6092, %Flow7650 ], [ poison, %for.cond.i.i1.i4643.preheader ]
  %6056 = phi i32 [ %6093, %Flow7650 ], [ poison, %for.cond.i.i1.i4643.preheader ]
  %retval.sroa.6.0.i.i.i4644 = phi i32 [ %6096, %Flow7650 ], [ undef, %for.cond.i.i1.i4643.preheader ]
  %retval.sroa.2.0.i.i.i4646 = phi i64 [ %6095, %Flow7650 ], [ undef, %for.cond.i.i1.i4643.preheader ]
  %retval.sroa.8.0.i.i.i4647 = phi i32 [ %6094, %Flow7650 ], [ undef, %for.cond.i.i1.i4643.preheader ]
  %index.0.i.i.i4648 = phi i32 [ %6098, %Flow7650 ], [ 0, %for.cond.i.i1.i4643.preheader ]
  %6057 = load i32, ptr addrspace(1) @__llvm_rpc_client, align 8, !tbaa !68
  %cmp.not.i.i.i4649 = icmp ult i32 %index.0.i.i.i4648, %6057
  %spec.store.select.i.i.i4650 = select i1 %cmp.not.i.i.i4649, i32 %index.0.i.i.i4648, i32 0
  %6058 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  %6059 = and i32 %spec.store.select.i.i.i4650, 31
  %6060 = lshr i64 %6058, %sh_prom.i.i.i.i3270
  %6061 = trunc i64 %6060 to i32
  %conv4.i.i.i.i.i4654 = and i32 %6061, 1
  %shl.i.i.i.i.i4655 = shl nuw i32 %conv4.i.i.i.i.i4654, %6059
  br label %ComputeLoop5996, !amdgpu.uniform !42

6062:                                             ; preds = %ComputeEnd5997
  %sunkaddr7508 = getelementptr inbounds i8, ptr addrspace(1) %8912, i64 40
  %6063 = atomicrmw or ptr addrspace(1) %sunkaddr7508, i32 %.lcssa8941 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %6064, !amdgpu.uniform !42

6064:                                             ; preds = %6062, %ComputeEnd5997
  %6065 = phi i32 [ %6063, %6062 ], [ poison, %ComputeEnd5997 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %8921)
  %6066 = call i32 @llvm.amdgcn.readfirstlane.i32(i32 %6065)
  %6067 = or i32 %6066, %.lcssa8942
  %shl5.i.i.i.i.i4656 = shl nuw i32 1, %6059
  %and.i.i.i.i.i4657 = and i32 %shl5.i.i.i.i.i4656, %6067
  %tobool3.i.i.i.i4658 = icmp ne i32 %and.i.i.i.i.i4657, 0
  %6068 = call i64 @llvm.amdgcn.ballot.i64(i1 %tobool3.i.i.i.i4658)
  %and.i.i.i.i.i.i4659 = and i64 %6068, %6058
  %cmp.i.not.i.i.i4660 = icmp ne i64 %6058, %and.i.i.i.i.i.i4659
  br i1 %cmp.i.not.i.i.i4660, label %if.then.i.i.i.i4661, label %Flow7649, !amdgpu.uniform !42

if.then.i.i.i.i4661:                              ; preds = %6064
  fence syncscope("agent") acquire
  %6069 = load <2 x i64>, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !75
  %6070 = extractelement <2 x i64> %6069, i32 0
  %6071 = inttoptr i64 %6070 to ptr
  %6072 = extractelement <2 x i64> %6069, i32 1
  %6073 = inttoptr i64 %6072 to ptr
  %idxprom.i.i.i.i4662 = zext i32 %spec.store.select.i.i.i4650 to i64
  %arrayidx.i.i.i.i4663 = getelementptr inbounds nuw i32, ptr %6071, i64 %idxprom.i.i.i.i4662, !amdgpu.uniform !42
  %6074 = load atomic i32, ptr %arrayidx.i.i.i.i4663 monotonic, align 4
  %6075 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %6074)
  %arrayidx.i45.i.i.i4664 = getelementptr inbounds nuw i32, ptr %6073, i64 %idxprom.i.i.i.i4662, !amdgpu.uniform !42
  %6076 = load atomic i32, ptr %arrayidx.i45.i.i.i4664 monotonic, align 4
  %6077 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %6076)
  %cmp.i46.not.i.i.i4665 = icmp ne i32 %6075, %6077
  br i1 %cmp.i46.not.i.i.i4665, label %if.then12.i.i.i4666, label %Flow7647, !amdgpu.uniform !42

if.then12.i.i.i4666:                              ; preds = %if.then.i.i.i.i4661
  fence syncscope("agent") release
  %6078 = call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %6058, i1 true)
  %iszero.i.i.i.i.i.i.i4667 = icmp ne i64 %6058, 0
  %cmp2.i.i.i.i.i.i4668 = icmp eq i64 %6078, %sh_prom.i.i.i.i3270
  %cmp.i.i.i.i.i.i4669 = select i1 %iszero.i.i.i.i.i.i.i4667, i1 %cmp2.i.i.i.i.i.i4668, i1 false
  %conv4.i.i51.i.i.i4670 = zext i1 %cmp.i.i.i.i.i.i4669 to i32
  %shl.i.i52.i.i.i4671 = shl nuw i32 %conv4.i.i51.i.i.i4670, %6059
  %xor.i.i.i.i.i4672 = xor i32 %shl.i.i52.i.i.i4671, -1
  %6079 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop6001, !amdgpu.uniform !42

6080:                                             ; preds = %ComputeEnd6002
  %sunkaddr7509 = getelementptr inbounds i8, ptr addrspace(1) %8912, i64 40
  %6081 = atomicrmw and ptr addrspace(1) %sunkaddr7509, i32 %.lcssa8943 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %6082, !amdgpu.uniform !42

6082:                                             ; preds = %6080, %ComputeEnd6002
  call void @llvm.amdgcn.end.cf.i64(i64 %8939)
  call void @llvm.amdgcn.wave.barrier()
  br label %Flow7647, !amdgpu.uniform !42

if.end14.i.i.i4732:                               ; preds = %Flow7647
  %6083 = call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %6058, i1 true)
  %iszero.i.i.i.i.i.i4733 = icmp ne i64 %6058, 0
  %cmp2.i.i.i.i.i4734 = icmp eq i64 %6083, %sh_prom.i.i.i.i3270
  %cmp.i.i.i.i.i4735 = select i1 %iszero.i.i.i.i.i.i4733, i1 %cmp2.i.i.i.i.i4734, i1 false
  %6084 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.i.i.i.i.i4735)
  %6085 = extractvalue { i1, i64 } %6084, 0
  %6086 = extractvalue { i1, i64 } %6084, 1
  br i1 %6085, label %if.then16.i.i.i4737, label %if.end22.i.i.i4736

if.then16.i.i.i4737:                              ; preds = %if.end14.i.i.i4732
  %6087 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !76
  %opcode.i.i.i4738 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %6087, i64 %idxprom.i.i.i.i4662, i32 1
  store i32 1660944387, ptr %opcode.i.i.i4738, align 8, !tbaa !77
  %arrayidx21.i.i.i4739 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %6087, i64 %idxprom.i.i.i.i4662
  store i64 %6058, ptr %arrayidx21.i.i.i4739, align 8, !tbaa !80
  br label %if.end22.i.i.i4736, !amdgpu.uniform !42

Flow7648:                                         ; preds = %if.end22.i.i.i4736, %Flow7647
  %6088 = phi i32 [ %6075, %if.end22.i.i.i4736 ], [ %retval.sroa.8.0.i.i.i4647, %Flow7647 ]
  %6089 = phi i64 [ %6058, %if.end22.i.i.i4736 ], [ %retval.sroa.2.0.i.i.i4646, %Flow7647 ]
  %6090 = phi i32 [ %spec.store.select.i.i.i4650, %if.end22.i.i.i4736 ], [ %retval.sroa.6.0.i.i.i4644, %Flow7647 ]
  br label %cleanup26.i.i.i4673, !amdgpu.uniform !42

if.end22.i.i.i4736:                               ; preds = %if.then16.i.i.i4737, %if.end14.i.i.i4732
  call void @llvm.amdgcn.end.cf.i64(i64 %6086)
  call void @llvm.amdgcn.wave.barrier()
  br label %Flow7648, !amdgpu.uniform !42

Flow7649:                                         ; preds = %cleanup26.i.i.i4673, %6064
  %6091 = phi i32 [ %6090, %cleanup26.i.i.i4673 ], [ %6054, %6064 ]
  %6092 = phi i64 [ %6089, %cleanup26.i.i.i4673 ], [ %6055, %6064 ]
  %6093 = phi i32 [ %6088, %cleanup26.i.i.i4673 ], [ %6056, %6064 ]
  %6094 = phi i32 [ %6088, %cleanup26.i.i.i4673 ], [ %retval.sroa.8.0.i.i.i4647, %6064 ]
  %6095 = phi i64 [ %6089, %cleanup26.i.i.i4673 ], [ %retval.sroa.2.0.i.i.i4646, %6064 ]
  %6096 = phi i32 [ %6090, %cleanup26.i.i.i4673 ], [ %retval.sroa.6.0.i.i.i4644, %6064 ]
  %6097 = phi i1 [ %cmp.i46.not.i.i.i4665, %cleanup26.i.i.i4673 ], [ true, %6064 ]
  br i1 %6097, label %for.inc.i.i.i4678, label %Flow7650, !amdgpu.uniform !42

cleanup26.i.i.i4673:                              ; preds = %Flow7648
  br label %Flow7649, !amdgpu.uniform !42

for.inc.i.i.i4678:                                ; preds = %Flow7649
  %inc.i.i.i4683 = add i32 %spec.store.select.i.i.i4650, 1
  br label %Flow7650, !amdgpu.uniform !42

Flow7650:                                         ; preds = %for.inc.i.i.i4678, %Flow7649
  %6098 = phi i32 [ %inc.i.i.i4683, %for.inc.i.i.i4678 ], [ poison, %Flow7649 ]
  %6099 = phi i1 [ false, %for.inc.i.i.i4678 ], [ true, %Flow7649 ]
  br i1 %6099, label %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i4684, label %for.cond.i.i1.i4643, !amdgpu.uniform !42

_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i4684: ; preds = %Flow7650
  %.lcssa8946 = phi i32 [ %6091, %Flow7650 ]
  %.lcssa8945 = phi i64 [ %6092, %Flow7650 ]
  %.lcssa8944 = phi i32 [ %6093, %Flow7650 ]
  %scevgep6527 = getelementptr i8, ptr %6050, i64 %lsr.iv.next6517.lcssa
  %sub.ptr.lhs.cast.i.i.i4639 = ptrtoint ptr %scevgep6527 to i64
  %sub.ptr.rhs.cast.i.i.i4640 = ptrtoint ptr %6050 to i64
  %sub.ptr.sub.i.i.i4641 = sub i64 %sub.ptr.lhs.cast.i.i.i4639, %sub.ptr.rhs.cast.i.i.i4640
  fence acquire
  %6100 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %mul.i.i.i.i.i.i4685 = shl i32 %.lcssa8946, 6
  %idxprom.i19.i.i.i.i.i4686 = zext i32 %mul.i.i.i.i.i.i4685 to i64
  %arrayidx.i20.i.i.i.i.i4687 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %6100, i64 %idxprom.i19.i.i.i.i.i4686
  %arrayidx.i22.i.i.i.i.i4688 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %arrayidx.i20.i.i.i.i.i4687, i64 %sh_prom.i.i.i.i3270
  store i64 %sub.ptr.sub.i.i.i4641, ptr %arrayidx.i22.i.i.i.i.i4688, align 8, !tbaa !82
  %spec.select.i.i.i.i.i.i.i4689 = call i64 @llvm.umin.i64(i64 %sub.ptr.sub.i.i.i4641, i64 56)
  %6101 = and i64 %spec.select.i.i.i.i.i.i.i4689, 15
  %6102 = sub i64 %spec.select.i.i.i.i.i.i.i4689, %6101
  %6103 = icmp ne i64 %6102, 0
  %6104 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %6103)
  %6105 = extractvalue { i1, i64 } %6104, 0
  %6106 = extractvalue { i1, i64 } %6104, 1
  br i1 %6105, label %loop-memcpy-expansion4965.preheader, label %Flow7646

loop-memcpy-expansion4965.preheader:              ; preds = %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i4684
  %6107 = add nuw nsw i64 %5881, 8
  %scevgep6524 = getelementptr i8, ptr %6100, i64 %6107
  %6108 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i4686, 6
  %scevgep6525 = getelementptr i8, ptr %scevgep6524, i64 %6108
  br label %loop-memcpy-expansion4965, !amdgpu.uniform !42

Flow7646:                                         ; preds = %Flow7645, %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i.i4684
  call void @llvm.amdgcn.end.cf.i64(i64 %6106)
  br label %loop-memcpy-residual-header4968, !amdgpu.uniform !42

loop-memcpy-expansion4965:                        ; preds = %loop-memcpy-expansion4965.preheader, %loop-memcpy-expansion4965
  %phi.broken8772 = phi i64 [ 0, %loop-memcpy-expansion4965.preheader ], [ %6113, %loop-memcpy-expansion4965 ]
  %loop-index4966 = phi i64 [ %6111, %loop-memcpy-expansion4965 ], [ 0, %loop-memcpy-expansion4965.preheader ]
  %6109 = getelementptr i8, ptr %6050, i64 %loop-index4966
  %6110 = load <4 x i32>, ptr %6109, align 1
  %scevgep6526 = getelementptr i8, ptr %scevgep6525, i64 %loop-index4966
  store <4 x i32> %6110, ptr %scevgep6526, align 1
  %6111 = add i64 %loop-index4966, 16
  %6112 = icmp uge i64 %6111, %6102
  %6113 = call i64 @llvm.amdgcn.if.break.i64(i1 %6112, i64 %phi.broken8772)
  %6114 = call i1 @llvm.amdgcn.loop.i64(i64 %6113)
  br i1 %6114, label %Flow7645, label %loop-memcpy-expansion4965

Flow7644:                                         ; preds = %Flow7642, %Flow7643
  %6115 = phi i64 [ %6200, %Flow7642 ], [ %6193, %Flow7643 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %6196)
  br label %post-loop-memcpy-expansion4964, !amdgpu.uniform !42

loop-memcpy-residual4967:                         ; preds = %loop-memcpy-residual4967.preheader, %loop-memcpy-residual4967
  %phi.broken8773 = phi i64 [ 0, %loop-memcpy-residual4967.preheader ], [ %6119, %loop-memcpy-residual4967 ]
  %residual-loop-index4969 = phi i64 [ %6117, %loop-memcpy-residual4967 ], [ 0, %loop-memcpy-residual4967.preheader ]
  %scevgep6523 = getelementptr i8, ptr %scevgep6522, i64 %residual-loop-index4969
  %6116 = load i8, ptr %scevgep6523, align 1
  %scevgep6521 = getelementptr i8, ptr %scevgep6520, i64 %residual-loop-index4969
  store i8 %6116, ptr %scevgep6521, align 1
  %6117 = add i64 %residual-loop-index4969, 1
  %6118 = icmp uge i64 %6117, %6101
  %6119 = call i64 @llvm.amdgcn.if.break.i64(i1 %6118, i64 %phi.broken8773)
  %6120 = call i1 @llvm.amdgcn.loop.i64(i64 %6119)
  br i1 %6120, label %Flow7642, label %loop-memcpy-residual4967

Flow7642:                                         ; preds = %loop-memcpy-residual4967
  %.lcssa8939 = phi i64 [ %6119, %loop-memcpy-residual4967 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa8939)
  br label %Flow7644, !amdgpu.uniform !42

post-loop-memcpy-expansion4964:                   ; preds = %Flow7644
  %6121 = add i64 %lsr.iv.next6517.lcssa, -56
  %tobool.not.i.i.i.i.i.i4691 = icmp eq i32 %.lcssa8944, 0
  %conv.i.i.i.i.i.i4692 = zext i1 %tobool.not.i.i.i.i.i.i4691 to i32
  fence release
  %6122 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %idxprom.i23.i.i.i.i.i4693 = zext i32 %.lcssa8946 to i64
  %arrayidx.i24.i.i.i.i.i4694 = getelementptr inbounds nuw i32, ptr %6122, i64 %idxprom.i23.i.i.i.i.i4693
  store atomic i32 %conv.i.i.i.i.i.i4692, ptr %arrayidx.i24.i.i.i.i.i4694 monotonic, align 4
  %6123 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !84
  %arrayidx.i.i5.i.i4695 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %6123, i64 %idxprom.i23.i.i.i.i.i4693, !amdgpu.uniform !42
  %6124 = load i64, ptr %arrayidx.i.i5.i.i4695, align 8, !tbaa !80
  %scevgep6499 = getelementptr i8, ptr %6050, i64 56
  br label %while.cond.i.i.i.i4696, !amdgpu.uniform !42

while.cond.i.i.i.i4696:                           ; preds = %post-loop-memcpy-expansion4964, %Flow7641
  %phi.broken8776 = phi i64 [ 0, %post-loop-memcpy-expansion4964 ], [ %6141, %Flow7641 ]
  %lsr.iv6507 = phi i64 [ %6139, %Flow7641 ], [ %6121, %post-loop-memcpy-expansion4964 ]
  %lsr.iv6500 = phi ptr [ %6138, %Flow7641 ], [ %scevgep6499, %post-loop-memcpy-expansion4964 ]
  %port.sroa.37.0.i.i4697 = phi i32 [ %6137, %Flow7641 ], [ %conv.i.i.i.i.i.i4692, %post-loop-memcpy-expansion4964 ]
  %idx.0.i.i.i.i4698 = phi i64 [ %6136, %Flow7641 ], [ 56, %post-loop-memcpy-expansion4964 ]
  %cmp.i.i.i.i4699 = icmp ugt i64 %sub.ptr.sub.i.i.i4641, %idx.0.i.i.i.i4698
  %6125 = call i64 @llvm.amdgcn.ballot.i64(i1 %cmp.i.i.i.i4699)
  %and.i.i.i.i6.i.i4700 = and i64 %6125, %6124
  %tobool.not.i.i.i.i4701 = icmp ne i64 %and.i.i.i.i6.i.i4700, 0
  %6126 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i22.i.i.i4702 = getelementptr inbounds nuw i32, ptr %6126, i64 %idxprom.i23.i.i.i.i.i4693, !amdgpu.uniform !42
  %6127 = load atomic i32, ptr %arrayidx.i22.i.i.i4702 monotonic, align 4
  %6128 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %tobool.not.i.i.i.i4701)
  %6129 = extractvalue { i1, i64 } %6128, 0
  %6130 = extractvalue { i1, i64 } %6128, 1
  br i1 %6129, label %cond.false.i8.i.i.i.i4703, label %Flow7641

cond.false.i8.i.i.i.i4703:                        ; preds = %while.cond.i.i.i.i4696
  %6131 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %6127)
  %6132 = icmp ne i32 %6131, %port.sroa.37.0.i.i4697
  br i1 %6132, label %while.body.i.i21.i.i.i.i4704.preheader, label %Flow7640, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i4704.preheader:           ; preds = %cond.false.i8.i.i.i.i4703
  br label %while.body.i.i21.i.i.i.i4704, !amdgpu.uniform !42

Flow7640:                                         ; preds = %Flow7639, %cond.false.i8.i.i.i.i4703
  br label %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i4707, !amdgpu.uniform !42

while.body.i.i21.i.i.i.i4704:                     ; preds = %while.body.i.i21.i.i.i.i4704.preheader, %while.body.i.i21.i.i.i.i4704
  call void @llvm.amdgcn.s.sleep(i32 2)
  %6133 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i22.i.i.i.i4705 = getelementptr inbounds nuw i32, ptr %6133, i64 %idxprom.i23.i.i.i.i.i4693, !amdgpu.uniform !42
  %6134 = load atomic i32, ptr %arrayidx.i.i.i22.i.i.i.i4705 monotonic, align 4
  %6135 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %6134)
  %cmp.i.not.i.i23.i.i.i.i4706 = icmp eq i32 %6135, %port.sroa.37.0.i.i4697
  br i1 %cmp.i.not.i.i23.i.i.i.i4706, label %Flow7639, label %while.body.i.i21.i.i.i.i4704, !amdgpu.uniform !42

Flow7639:                                         ; preds = %while.body.i.i21.i.i.i.i4704
  br label %Flow7640, !amdgpu.uniform !42

Flow7641:                                         ; preds = %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i4708, %while.cond.i.i.i.i4696
  %6136 = phi i64 [ %add.i.i.i.i4711, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i4708 ], [ poison, %while.cond.i.i.i.i4696 ]
  %6137 = phi i32 [ %conv.i.i26.i.i.i.i4709, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i4708 ], [ poison, %while.cond.i.i.i.i4696 ]
  %6138 = phi ptr [ %scevgep6501, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i4708 ], [ poison, %while.cond.i.i.i.i4696 ]
  %6139 = phi i64 [ %lsr.iv.next6508, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i4708 ], [ poison, %while.cond.i.i.i.i4696 ]
  %6140 = phi i1 [ false, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i4708 ], [ true, %while.cond.i.i.i.i4696 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %6130)
  %6141 = call i64 @llvm.amdgcn.if.break.i64(i1 %6140, i64 %phi.broken8776)
  %6142 = call i1 @llvm.amdgcn.loop.i64(i64 %6141)
  br i1 %6142, label %cond.false.i.i.i4718, label %while.cond.i.i.i.i4696

_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i4707: ; preds = %Flow7640
  fence acquire
  %6143 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.i.i.i.i4699)
  %6144 = extractvalue { i1, i64 } %6143, 0
  %6145 = extractvalue { i1, i64 } %6143, 1
  br i1 %6144, label %if.then.i.i.i.i.i.i.i4712, label %Flow7638

if.then.i.i.i.i.i.i.i4712:                        ; preds = %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i4707
  %sub.i.i.i.i.i.i.i4714 = sub nuw i64 %sub.ptr.sub.i.i.i4641, %idx.0.i.i.i.i4698
  %spec.select.i.i.i36.i.i.i.i4715 = call i64 @llvm.umin.i64(i64 %sub.i.i.i.i.i.i.i4714, i64 64)
  %6146 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %6147 = and i64 %spec.select.i.i.i36.i.i.i.i4715, 15
  %6148 = sub i64 %spec.select.i.i.i36.i.i.i.i4715, %6147
  %6149 = icmp ne i64 %6148, 0
  %6150 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %6149)
  %6151 = extractvalue { i1, i64 } %6150, 0
  %6152 = extractvalue { i1, i64 } %6150, 1
  br i1 %6151, label %loop-memcpy-expansion.preheader, label %Flow7637

loop-memcpy-expansion.preheader:                  ; preds = %if.then.i.i.i.i.i.i.i4712
  %arrayidx.i20.i35.i.i.i.i4716 = getelementptr %"struct.rpc::Buffer", ptr %6146, i64 %idxprom.i19.i.i.i.i.i4686
  %arrayidx.i22.i38.i.i.i.i4717 = getelementptr %"struct.rpc::Buffer", ptr %arrayidx.i20.i35.i.i.i.i4716, i64 %sh_prom.i.i.i.i3270
  br label %loop-memcpy-expansion, !amdgpu.uniform !42

Flow7637:                                         ; preds = %Flow7636, %if.then.i.i.i.i.i.i.i4712
  call void @llvm.amdgcn.end.cf.i64(i64 %6152)
  br label %loop-memcpy-residual-header, !amdgpu.uniform !42

loop-memcpy-expansion:                            ; preds = %loop-memcpy-expansion.preheader, %loop-memcpy-expansion
  %phi.broken8774 = phi i64 [ 0, %loop-memcpy-expansion.preheader ], [ %6157, %loop-memcpy-expansion ]
  %loop-index = phi i64 [ %6155, %loop-memcpy-expansion ], [ 0, %loop-memcpy-expansion.preheader ]
  %scevgep6502 = getelementptr i8, ptr %lsr.iv6500, i64 %loop-index
  %6153 = load <4 x i32>, ptr %scevgep6502, align 1
  %6154 = getelementptr i8, ptr %arrayidx.i22.i38.i.i.i.i4717, i64 %loop-index
  store <4 x i32> %6153, ptr %6154, align 1
  %6155 = add i64 %loop-index, 16
  %6156 = icmp uge i64 %6155, %6148
  %6157 = call i64 @llvm.amdgcn.if.break.i64(i1 %6156, i64 %phi.broken8774)
  %6158 = call i1 @llvm.amdgcn.loop.i64(i64 %6157)
  br i1 %6158, label %Flow7636, label %loop-memcpy-expansion

Flow7635:                                         ; preds = %Flow7634, %loop-memcpy-residual-header
  call void @llvm.amdgcn.end.cf.i64(i64 %6186)
  br label %Flow7638, !amdgpu.uniform !42

loop-memcpy-residual:                             ; preds = %loop-memcpy-residual.preheader, %loop-memcpy-residual
  %phi.broken8775 = phi i64 [ 0, %loop-memcpy-residual.preheader ], [ %6162, %loop-memcpy-residual ]
  %residual-loop-index = phi i64 [ %6160, %loop-memcpy-residual ], [ 0, %loop-memcpy-residual.preheader ]
  %scevgep6514 = getelementptr i8, ptr %scevgep6513, i64 %residual-loop-index
  %6159 = load i8, ptr %scevgep6514, align 1
  %scevgep6511 = getelementptr i8, ptr %scevgep6510, i64 %residual-loop-index
  store i8 %6159, ptr %scevgep6511, align 1
  %6160 = add i64 %residual-loop-index, 1
  %6161 = icmp uge i64 %6160, %6147
  %6162 = call i64 @llvm.amdgcn.if.break.i64(i1 %6161, i64 %phi.broken8775)
  %6163 = call i1 @llvm.amdgcn.loop.i64(i64 %6162)
  br i1 %6163, label %Flow7634, label %loop-memcpy-residual

Flow7634:                                         ; preds = %loop-memcpy-residual
  %.lcssa8936 = phi i64 [ %6162, %loop-memcpy-residual ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa8936)
  br label %Flow7635, !amdgpu.uniform !42

_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i4708: ; preds = %Flow7638
  %conv.i.i26.i.i.i.i4709 = xor i32 %port.sroa.37.0.i.i4697, 1
  fence release
  %6164 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %arrayidx.i24.i29.i.i.i.i4710 = getelementptr inbounds nuw i32, ptr %6164, i64 %idxprom.i23.i.i.i.i.i4693
  store atomic i32 %conv.i.i26.i.i.i.i4709, ptr %arrayidx.i24.i29.i.i.i.i4710 monotonic, align 4
  %add.i.i.i.i4711 = add i64 %idx.0.i.i.i.i4698, 64
  %scevgep6501 = getelementptr i8, ptr %lsr.iv6500, i64 64
  %lsr.iv.next6508 = add i64 %lsr.iv6507, -64
  br label %Flow7641, !amdgpu.uniform !42

cond.false.i.i.i4718:                             ; preds = %Flow7641
  %.lcssa8938 = phi i64 [ %6141, %Flow7641 ]
  %port.sroa.37.0.i.i4697.lcssa = phi i32 [ %port.sroa.37.0.i.i4697, %Flow7641 ]
  %.lcssa8937 = phi i32 [ %6127, %Flow7641 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa8938)
  %6165 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %.lcssa8937)
  %6166 = icmp ne i32 %6165, %port.sroa.37.0.i.i4697.lcssa
  %6167 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %6166)
  %6168 = extractvalue { i1, i64 } %6167, 0
  %6169 = extractvalue { i1, i64 } %6167, 1
  br i1 %6168, label %while.body.i.i.i.i4719.preheader, label %Flow7633

while.body.i.i.i.i4719.preheader:                 ; preds = %cond.false.i.i.i4718
  br label %while.body.i.i.i.i4719, !amdgpu.uniform !42

Flow7633:                                         ; preds = %Flow, %cond.false.i.i.i4718
  call void @llvm.amdgcn.end.cf.i64(i64 %6169)
  br label %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit4740, !amdgpu.uniform !42

while.body.i.i.i.i4719:                           ; preds = %while.body.i.i.i.i4719.preheader, %while.body.i.i.i.i4719
  %phi.broken8777 = phi i64 [ 0, %while.body.i.i.i.i4719.preheader ], [ %6173, %while.body.i.i.i.i4719 ]
  call void @llvm.amdgcn.s.sleep(i32 2)
  %6170 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i8.i.i4720 = getelementptr inbounds nuw i32, ptr %6170, i64 %idxprom.i23.i.i.i.i.i4693, !amdgpu.uniform !42
  %6171 = load atomic i32, ptr %arrayidx.i.i.i8.i.i4720 monotonic, align 4
  %6172 = call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %6171)
  %cmp.i.not.i.i.i.i4721 = icmp eq i32 %6172, %port.sroa.37.0.i.i4697.lcssa
  %6173 = call i64 @llvm.amdgcn.if.break.i64(i1 %cmp.i.not.i.i.i.i4721, i64 %phi.broken8777)
  %6174 = call i1 @llvm.amdgcn.loop.i64(i64 %6173)
  br i1 %6174, label %Flow, label %while.body.i.i.i.i4719

Flow:                                             ; preds = %while.body.i.i.i.i4719
  %.lcssa8934 = phi i64 [ %6173, %while.body.i.i.i.i4719 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa8934)
  br label %Flow7633, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit4740: ; preds = %Flow7633
  fence acquire
  call void @llvm.amdgcn.wave.barrier()
  fence syncscope("agent") release
  %6175 = call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %.lcssa8945, i1 true)
  %iszero.i.i.i.i.i13.i.i4723 = icmp ne i64 %.lcssa8945, 0
  %cmp2.i.i.i.i14.i.i4724 = icmp eq i64 %6175, %sh_prom.i.i.i.i3270
  %cmp.i.i.i.i15.i.i4725 = select i1 %iszero.i.i.i.i.i13.i.i4723, i1 %cmp2.i.i.i.i14.i.i4724, i1 false
  %6176 = and i32 %.lcssa8946, 31
  %conv4.i.i.i19.i.i4729 = zext i1 %cmp.i.i.i.i15.i.i4725 to i32
  %shl.i.i.i20.i.i4730 = shl nuw i32 %conv4.i.i.i19.i.i4729, %6176
  %xor.i.i.i21.i.i4731 = xor i32 %shl.i.i.i20.i.i4730, -1
  %6177 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop6005, !amdgpu.uniform !42

6178:                                             ; preds = %ComputeEnd6006
  %div8.i.i.i16.i.i4726 = lshr i32 %.lcssa8946, 5
  %idxprom.i.i.i17.i.i4727 = zext nneg i32 %div8.i.i.i16.i.i4726 to i64
  %6179 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i17.i.i4727
  %sunkaddr7510 = getelementptr inbounds i8, ptr addrspace(1) %6179, i64 40
  %6180 = atomicrmw and ptr addrspace(1) %sunkaddr7510, i32 %.lcssa8933 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %6181, !amdgpu.uniform !42

6181:                                             ; preds = %6178, %ComputeEnd6006
  call void @llvm.amdgcn.end.cf.i64(i64 %8956)
  call void @llvm.amdgcn.wave.barrier()
  call fastcc void @_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIcEERS1_T_(i8 noundef signext 10) #33
  br label %Flow7670, !amdgpu.uniform !42

if.end111:                                        ; preds = %Flow7917
  %cmp112 = icmp sgt i32 %1149, 0
  %cmp113 = icmp eq i32 %1148, 0
  %6182 = or i1 %cmp112, %cmp113
  %cond114 = zext i1 %6182 to i32
  ret i32 %cond114

Flow7636:                                         ; preds = %loop-memcpy-expansion
  %.lcssa8935 = phi i64 [ %6157, %loop-memcpy-expansion ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa8935)
  br label %Flow7637, !amdgpu.uniform !42

Flow7638:                                         ; preds = %Flow7635, %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i4707
  call void @llvm.amdgcn.end.cf.i64(i64 %6145)
  br label %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i4708, !amdgpu.uniform !42

loop-memcpy-residual-header:                      ; preds = %Flow7637
  %6183 = icmp ne i64 %6147, 0
  %6184 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %6183)
  %6185 = extractvalue { i1, i64 } %6184, 0
  %6186 = extractvalue { i1, i64 } %6184, 1
  br i1 %6185, label %loop-memcpy-residual.preheader, label %Flow7635

loop-memcpy-residual.preheader:                   ; preds = %loop-memcpy-residual-header
  %umin6512 = call i64 @llvm.umin.i64(i64 %lsr.iv6507, i64 64)
  %6187 = lshr i64 %umin6512, 4
  %6188 = shl nuw nsw i64 %6187, 4
  %scevgep6513 = getelementptr i8, ptr %lsr.iv6500, i64 %6188
  %scevgep6503 = getelementptr i8, ptr %6146, i64 %5881
  %scevgep6504 = getelementptr i8, ptr %scevgep6503, i64 %6115
  %scevgep6510 = getelementptr i8, ptr %scevgep6504, i64 %6188
  br label %loop-memcpy-residual, !amdgpu.uniform !42

Flow7645:                                         ; preds = %loop-memcpy-expansion4965
  %.lcssa8940 = phi i64 [ %6113, %loop-memcpy-expansion4965 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa8940)
  br label %Flow7646, !amdgpu.uniform !42

loop-memcpy-residual-header4968:                  ; preds = %Flow7646
  %6189 = icmp eq i64 %6101, 0
  %6190 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %6189)
  %6191 = extractvalue { i1, i64 } %6190, 0
  %6192 = extractvalue { i1, i64 } %6190, 1
  br i1 %6191, label %loop-memcpy-residual-header4968.post-loop-memcpy-expansion4964_crit_edge, label %Flow7643

loop-memcpy-residual-header4968.post-loop-memcpy-expansion4964_crit_edge: ; preds = %loop-memcpy-residual-header4968
  %.pre7301 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i4686, 6
  br label %Flow7643, !amdgpu.uniform !42

Flow7643:                                         ; preds = %loop-memcpy-residual-header4968.post-loop-memcpy-expansion4964_crit_edge, %loop-memcpy-residual-header4968
  %6193 = phi i64 [ %.pre7301, %loop-memcpy-residual-header4968.post-loop-memcpy-expansion4964_crit_edge ], [ poison, %loop-memcpy-residual-header4968 ]
  %6194 = call { i1, i64 } @llvm.amdgcn.else.i64.i64(i64 %6192)
  %6195 = extractvalue { i1, i64 } %6194, 0
  %6196 = extractvalue { i1, i64 } %6194, 1
  br i1 %6195, label %loop-memcpy-residual4967.preheader, label %Flow7644

loop-memcpy-residual4967.preheader:               ; preds = %Flow7643
  %6197 = add nuw nsw i64 %5881, 8
  %scevgep6515 = getelementptr i8, ptr %6100, i64 %6197
  %umin6518 = call i64 @llvm.umin.i64(i64 %lsr.iv.next6517.lcssa, i64 56)
  %6198 = lshr i64 %umin6518, 4
  %6199 = shl nuw nsw i64 %6198, 4
  %scevgep6519 = getelementptr i8, ptr %scevgep6515, i64 %6199
  %6200 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i4686, 6
  %scevgep6520 = getelementptr i8, ptr %scevgep6519, i64 %6200
  %scevgep6522 = getelementptr i8, ptr %6050, i64 %6199
  br label %loop-memcpy-residual4967, !amdgpu.uniform !42

Flow7703:                                         ; preds = %loop-memcpy-expansion4971
  %.lcssa8831 = phi i64 [ %5711, %loop-memcpy-expansion4971 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa8831)
  br label %Flow7704, !amdgpu.uniform !42

Flow7705:                                         ; preds = %Flow7702, %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i4276
  call void @llvm.amdgcn.end.cf.i64(i64 %5699)
  br label %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i4277, !amdgpu.uniform !42

loop-memcpy-residual-header4974:                  ; preds = %Flow7704
  %6201 = icmp ne i64 %5701, 0
  %6202 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %6201)
  %6203 = extractvalue { i1, i64 } %6202, 0
  %6204 = extractvalue { i1, i64 } %6202, 1
  br i1 %6203, label %loop-memcpy-residual4973.preheader, label %Flow7702

loop-memcpy-residual4973.preheader:               ; preds = %loop-memcpy-residual-header4974
  %umin = call i64 @llvm.umin.i64(i64 %lsr.iv6344, i64 64)
  %6205 = lshr i64 %umin, 4
  %6206 = shl nuw nsw i64 %6205, 4
  %scevgep6347 = getelementptr i8, ptr %lsr.iv, i64 %6206
  %scevgep6342 = getelementptr i8, ptr %5700, i64 %4513
  %scevgep6343 = getelementptr i8, ptr %scevgep6342, i64 %5669
  %scevgep6345 = getelementptr i8, ptr %scevgep6343, i64 %6206
  br label %loop-memcpy-residual4973, !amdgpu.uniform !42

Flow7712:                                         ; preds = %loop-memcpy-expansion4977
  %.lcssa8836 = phi i64 [ %5667, %loop-memcpy-expansion4977 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa8836)
  br label %Flow7713, !amdgpu.uniform !42

loop-memcpy-residual-header4980:                  ; preds = %Flow7713
  %6207 = icmp eq i64 %5655, 0
  %6208 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %6207)
  %6209 = extractvalue { i1, i64 } %6208, 0
  %6210 = extractvalue { i1, i64 } %6208, 1
  br i1 %6209, label %loop-memcpy-residual-header4980.post-loop-memcpy-expansion4976_crit_edge, label %Flow7710

loop-memcpy-residual-header4980.post-loop-memcpy-expansion4976_crit_edge: ; preds = %loop-memcpy-residual-header4980
  %.pre7293 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i4255, 6
  br label %Flow7710, !amdgpu.uniform !42

Flow7710:                                         ; preds = %loop-memcpy-residual-header4980.post-loop-memcpy-expansion4976_crit_edge, %loop-memcpy-residual-header4980
  %6211 = phi i64 [ %.pre7293, %loop-memcpy-residual-header4980.post-loop-memcpy-expansion4976_crit_edge ], [ poison, %loop-memcpy-residual-header4980 ]
  %6212 = call { i1, i64 } @llvm.amdgcn.else.i64.i64(i64 %6210)
  %6213 = extractvalue { i1, i64 } %6212, 0
  %6214 = extractvalue { i1, i64 } %6212, 1
  br i1 %6213, label %loop-memcpy-residual4979.preheader, label %Flow7711

loop-memcpy-residual4979.preheader:               ; preds = %Flow7710
  %6215 = lshr i64 %spec.select.i.i.i.i.i.i.i4258, 4
  %6216 = shl nuw nsw i64 %6215, 4
  %6217 = add i64 %4513, %6216
  %6218 = add i64 %6217, 8
  %scevgep6349 = getelementptr i8, ptr %5654, i64 %6218
  %6219 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i4255, 6
  %scevgep6350 = getelementptr i8, ptr %scevgep6349, i64 %6219
  %scevgep6352 = getelementptr i8, ptr %5579, i64 %6216
  br label %loop-memcpy-residual4979, !amdgpu.uniform !42

Flow7776:                                         ; preds = %loop-memcpy-expansion4983
  %.lcssa8869 = phi i64 [ %5260, %loop-memcpy-expansion4983 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa8869)
  br label %Flow7777, !amdgpu.uniform !42

Flow7778:                                         ; preds = %Flow7775, %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i3892
  call void @llvm.amdgcn.end.cf.i64(i64 %5248)
  br label %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i3893, !amdgpu.uniform !42

loop-memcpy-residual-header4986:                  ; preds = %Flow7777
  %6220 = icmp ne i64 %5250, 0
  %6221 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %6220)
  %6222 = extractvalue { i1, i64 } %6221, 0
  %6223 = extractvalue { i1, i64 } %6221, 1
  br i1 %6222, label %loop-memcpy-residual4985.preheader, label %Flow7775

loop-memcpy-residual4985.preheader:               ; preds = %loop-memcpy-residual-header4986
  %umin6392 = call i64 @llvm.umin.i64(i64 %lsr.iv6390, i64 64)
  %6224 = lshr i64 %umin6392, 4
  %6225 = shl nuw nsw i64 %6224, 4
  %scevgep6395 = getelementptr i8, ptr %lsr.iv6385, i64 %6225
  %scevgep6388 = getelementptr i8, ptr %5249, i64 %4513
  %scevgep6389 = getelementptr i8, ptr %scevgep6388, i64 %5218
  %scevgep6393 = getelementptr i8, ptr %scevgep6389, i64 %6225
  br label %loop-memcpy-residual4985, !amdgpu.uniform !42

Flow7785:                                         ; preds = %loop-memcpy-expansion4989
  %.lcssa8874 = phi i64 [ %5216, %loop-memcpy-expansion4989 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa8874)
  br label %Flow7786, !amdgpu.uniform !42

loop-memcpy-residual-header4992:                  ; preds = %Flow7786
  %6226 = icmp eq i64 %5204, 0
  %6227 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %6226)
  %6228 = extractvalue { i1, i64 } %6227, 0
  %6229 = extractvalue { i1, i64 } %6227, 1
  br i1 %6228, label %loop-memcpy-residual-header4992.post-loop-memcpy-expansion4988_crit_edge, label %Flow7783

loop-memcpy-residual-header4992.post-loop-memcpy-expansion4988_crit_edge: ; preds = %loop-memcpy-residual-header4992
  %.pre7289 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i3871, 6
  br label %Flow7783, !amdgpu.uniform !42

Flow7783:                                         ; preds = %loop-memcpy-residual-header4992.post-loop-memcpy-expansion4988_crit_edge, %loop-memcpy-residual-header4992
  %6230 = phi i64 [ %.pre7289, %loop-memcpy-residual-header4992.post-loop-memcpy-expansion4988_crit_edge ], [ poison, %loop-memcpy-residual-header4992 ]
  %6231 = call { i1, i64 } @llvm.amdgcn.else.i64.i64(i64 %6229)
  %6232 = extractvalue { i1, i64 } %6231, 0
  %6233 = extractvalue { i1, i64 } %6231, 1
  br i1 %6232, label %loop-memcpy-residual4991.preheader, label %Flow7784

loop-memcpy-residual4991.preheader:               ; preds = %Flow7783
  %6234 = lshr i64 %spec.select.i.i.i.i.i.i.i3874, 4
  %6235 = shl nuw nsw i64 %6234, 4
  %6236 = add i64 %4513, %6235
  %6237 = add i64 %6236, 8
  %scevgep6397 = getelementptr i8, ptr %5203, i64 %6237
  %6238 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i3871, 6
  %scevgep6398 = getelementptr i8, ptr %scevgep6397, i64 %6238
  %scevgep6400 = getelementptr i8, ptr %5128, i64 %6235
  br label %loop-memcpy-residual4991, !amdgpu.uniform !42

Flow7859:                                         ; preds = %loop-memcpy-expansion4995
  %.lcssa8907 = phi i64 [ %4784, %loop-memcpy-expansion4995 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa8907)
  br label %Flow7860, !amdgpu.uniform !42

Flow7861:                                         ; preds = %Flow7858, %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i3501
  call void @llvm.amdgcn.end.cf.i64(i64 %4772)
  br label %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i3502, !amdgpu.uniform !42

loop-memcpy-residual-header4998:                  ; preds = %Flow7860
  %6239 = icmp ne i64 %4774, 0
  %6240 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %6239)
  %6241 = extractvalue { i1, i64 } %6240, 0
  %6242 = extractvalue { i1, i64 } %6240, 1
  br i1 %6241, label %loop-memcpy-residual4997.preheader, label %Flow7858

loop-memcpy-residual4997.preheader:               ; preds = %loop-memcpy-residual-header4998
  %umin6459 = call i64 @llvm.umin.i64(i64 %lsr.iv6457, i64 64)
  %6243 = lshr i64 %umin6459, 4
  %6244 = shl nuw nsw i64 %6243, 4
  %scevgep6462 = getelementptr i8, ptr %lsr.iv6452, i64 %6244
  %scevgep6455 = getelementptr i8, ptr %4773, i64 %4513
  %scevgep6456 = getelementptr i8, ptr %scevgep6455, i64 %4742
  %scevgep6460 = getelementptr i8, ptr %scevgep6456, i64 %6244
  br label %loop-memcpy-residual4997, !amdgpu.uniform !42

Flow7868:                                         ; preds = %loop-memcpy-expansion5001
  %.lcssa8912 = phi i64 [ %4740, %loop-memcpy-expansion5001 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa8912)
  br label %Flow7869, !amdgpu.uniform !42

loop-memcpy-residual-header5004:                  ; preds = %Flow7869
  %6245 = icmp eq i64 %4728, 0
  %6246 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %6245)
  %6247 = extractvalue { i1, i64 } %6246, 0
  %6248 = extractvalue { i1, i64 } %6246, 1
  br i1 %6247, label %loop-memcpy-residual-header5004.post-loop-memcpy-expansion5000_crit_edge, label %Flow7866

loop-memcpy-residual-header5004.post-loop-memcpy-expansion5000_crit_edge: ; preds = %loop-memcpy-residual-header5004
  %.pre7283 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i3480, 6
  br label %Flow7866, !amdgpu.uniform !42

Flow7866:                                         ; preds = %loop-memcpy-residual-header5004.post-loop-memcpy-expansion5000_crit_edge, %loop-memcpy-residual-header5004
  %6249 = phi i64 [ %.pre7283, %loop-memcpy-residual-header5004.post-loop-memcpy-expansion5000_crit_edge ], [ poison, %loop-memcpy-residual-header5004 ]
  %6250 = call { i1, i64 } @llvm.amdgcn.else.i64.i64(i64 %6248)
  %6251 = extractvalue { i1, i64 } %6250, 0
  %6252 = extractvalue { i1, i64 } %6250, 1
  br i1 %6251, label %loop-memcpy-residual5003.preheader, label %Flow7867

loop-memcpy-residual5003.preheader:               ; preds = %Flow7866
  %6253 = lshr i64 %spec.select.i.i.i.i.i.i.i3483, 4
  %6254 = shl nuw nsw i64 %6253, 4
  %6255 = add i64 %4513, %6254
  %6256 = add i64 %6255, 8
  %scevgep6464 = getelementptr i8, ptr %4727, i64 %6256
  %6257 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i3480, 6
  %scevgep6465 = getelementptr i8, ptr %scevgep6464, i64 %6257
  %scevgep6467 = getelementptr i8, ptr %4652, i64 %6254
  br label %loop-memcpy-residual5003, !amdgpu.uniform !42

Flow8022:                                         ; preds = %loop-memcpy-expansion5007
  br label %Flow8023, !amdgpu.uniform !42

Flow8024:                                         ; preds = %Flow8021, %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i3110
  br label %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i3111, !amdgpu.uniform !42

loop-memcpy-residual-header5010:                  ; preds = %Flow8023
  %6258 = icmp ne i64 %4315, 0
  br i1 %6258, label %loop-memcpy-residual5009.preheader, label %Flow8021, !amdgpu.uniform !42

loop-memcpy-residual5009.preheader:               ; preds = %loop-memcpy-residual-header5010
  %umin6901 = call i64 @llvm.umin.i64(i64 %lsr.iv6899, i64 64)
  %6259 = lshr i64 %umin6901, 4
  %6260 = shl nuw nsw i64 %6259, 4
  %scevgep6904 = getelementptr i8, ptr %lsr.iv6894, i64 %6260
  %scevgep6897 = getelementptr i8, ptr %4314, i64 %1474
  %scevgep6898 = getelementptr i8, ptr %scevgep6897, i64 %4283
  %scevgep6902 = getelementptr i8, ptr %scevgep6898, i64 %6260
  br label %loop-memcpy-residual5009, !amdgpu.uniform !42

Flow8031:                                         ; preds = %loop-memcpy-expansion5013
  br label %Flow8032, !amdgpu.uniform !42

loop-memcpy-residual-header5016:                  ; preds = %Flow8032
  %6261 = icmp eq i64 %4275, 0
  br i1 %6261, label %loop-memcpy-residual-header5016.post-loop-memcpy-expansion5012_crit_edge, label %Flow8029, !amdgpu.uniform !42

loop-memcpy-residual-header5016.post-loop-memcpy-expansion5012_crit_edge: ; preds = %loop-memcpy-residual-header5016
  %.pre7337 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i3089, 6
  br label %Flow8029, !amdgpu.uniform !42

Flow8029:                                         ; preds = %loop-memcpy-residual-header5016.post-loop-memcpy-expansion5012_crit_edge, %loop-memcpy-residual-header5016
  %6262 = phi i64 [ %.pre7337, %loop-memcpy-residual-header5016.post-loop-memcpy-expansion5012_crit_edge ], [ poison, %loop-memcpy-residual-header5016 ]
  %6263 = phi i1 [ false, %loop-memcpy-residual-header5016.post-loop-memcpy-expansion5012_crit_edge ], [ true, %loop-memcpy-residual-header5016 ]
  br i1 %6263, label %loop-memcpy-residual5015.preheader, label %Flow8030, !amdgpu.uniform !42

loop-memcpy-residual5015.preheader:               ; preds = %Flow8029
  %scevgep6887 = getelementptr i8, ptr %4274, i64 8
  %6264 = add i64 %1474, %4276
  %scevgep6888 = getelementptr i8, ptr %scevgep6887, i64 %6264
  %6265 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i3089, 6
  %scevgep6889 = getelementptr i8, ptr %scevgep6888, i64 %6265
  %scevgep6891 = getelementptr i8, ptr %4211, i64 %4276
  br label %loop-memcpy-residual5015, !amdgpu.uniform !42

Flow7961:                                         ; preds = %loop-memcpy-expansion5019
  br label %Flow7962, !amdgpu.uniform !42

Flow7963:                                         ; preds = %Flow7960, %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i2835
  br label %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i2836, !amdgpu.uniform !42

loop-memcpy-residual-header5022:                  ; preds = %Flow7962
  %6266 = icmp ne i64 %4031, 0
  br i1 %6266, label %loop-memcpy-residual5021.preheader, label %Flow7960, !amdgpu.uniform !42

loop-memcpy-residual5021.preheader:               ; preds = %loop-memcpy-residual-header5022
  %umin6847 = call i64 @llvm.umin.i64(i64 %lsr.iv6845, i64 64)
  %6267 = lshr i64 %umin6847, 4
  %6268 = shl nuw nsw i64 %6267, 4
  %scevgep6850 = getelementptr i8, ptr %lsr.iv6840, i64 %6268
  %scevgep6843 = getelementptr i8, ptr %4030, i64 %1474
  %scevgep6844 = getelementptr i8, ptr %scevgep6843, i64 %3999
  %scevgep6848 = getelementptr i8, ptr %scevgep6844, i64 %6268
  br label %loop-memcpy-residual5021, !amdgpu.uniform !42

Flow7970:                                         ; preds = %loop-memcpy-expansion5025
  br label %Flow7971, !amdgpu.uniform !42

loop-memcpy-residual-header5028:                  ; preds = %Flow7971
  %6269 = icmp eq i64 %3991, 0
  br i1 %6269, label %loop-memcpy-residual-header5028.post-loop-memcpy-expansion5024_crit_edge, label %Flow7968, !amdgpu.uniform !42

loop-memcpy-residual-header5028.post-loop-memcpy-expansion5024_crit_edge: ; preds = %loop-memcpy-residual-header5028
  %.pre7341 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i2814, 6
  br label %Flow7968, !amdgpu.uniform !42

Flow7968:                                         ; preds = %loop-memcpy-residual-header5028.post-loop-memcpy-expansion5024_crit_edge, %loop-memcpy-residual-header5028
  %6270 = phi i64 [ %.pre7341, %loop-memcpy-residual-header5028.post-loop-memcpy-expansion5024_crit_edge ], [ poison, %loop-memcpy-residual-header5028 ]
  %6271 = phi i1 [ false, %loop-memcpy-residual-header5028.post-loop-memcpy-expansion5024_crit_edge ], [ true, %loop-memcpy-residual-header5028 ]
  br i1 %6271, label %loop-memcpy-residual5027.preheader, label %Flow7969, !amdgpu.uniform !42

loop-memcpy-residual5027.preheader:               ; preds = %Flow7968
  %scevgep6833 = getelementptr i8, ptr %3990, i64 8
  %6272 = add i64 %1474, %3992
  %scevgep6834 = getelementptr i8, ptr %scevgep6833, i64 %6272
  %6273 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i2814, 6
  %scevgep6835 = getelementptr i8, ptr %scevgep6834, i64 %6273
  %scevgep6837 = getelementptr i8, ptr %3927, i64 %3992
  br label %loop-memcpy-residual5027, !amdgpu.uniform !42

Flow8087:                                         ; preds = %loop-memcpy-expansion5031
  br label %Flow8088, !amdgpu.uniform !42

Flow8089:                                         ; preds = %Flow8086, %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i2561
  br label %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i2562, !amdgpu.uniform !42

loop-memcpy-residual-header5034:                  ; preds = %Flow8088
  %6274 = icmp ne i64 %3754, 0
  br i1 %6274, label %loop-memcpy-residual5033.preheader, label %Flow8086, !amdgpu.uniform !42

loop-memcpy-residual5033.preheader:               ; preds = %loop-memcpy-residual-header5034
  %umin6955 = call i64 @llvm.umin.i64(i64 %lsr.iv6953, i64 64)
  %6275 = lshr i64 %umin6955, 4
  %6276 = shl nuw nsw i64 %6275, 4
  %scevgep6958 = getelementptr i8, ptr %lsr.iv6948, i64 %6276
  %scevgep6951 = getelementptr i8, ptr %3753, i64 %1474
  %scevgep6952 = getelementptr i8, ptr %scevgep6951, i64 %3722
  %scevgep6956 = getelementptr i8, ptr %scevgep6952, i64 %6276
  br label %loop-memcpy-residual5033, !amdgpu.uniform !42

Flow8096:                                         ; preds = %loop-memcpy-expansion5037
  br label %Flow8097, !amdgpu.uniform !42

loop-memcpy-residual-header5040:                  ; preds = %Flow8097
  %6277 = icmp eq i64 %3714, 0
  br i1 %6277, label %loop-memcpy-residual-header5040.post-loop-memcpy-expansion5036_crit_edge, label %Flow8094, !amdgpu.uniform !42

loop-memcpy-residual-header5040.post-loop-memcpy-expansion5036_crit_edge: ; preds = %loop-memcpy-residual-header5040
  %.pre7333 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i2540, 6
  br label %Flow8094, !amdgpu.uniform !42

Flow8094:                                         ; preds = %loop-memcpy-residual-header5040.post-loop-memcpy-expansion5036_crit_edge, %loop-memcpy-residual-header5040
  %6278 = phi i64 [ %.pre7333, %loop-memcpy-residual-header5040.post-loop-memcpy-expansion5036_crit_edge ], [ poison, %loop-memcpy-residual-header5040 ]
  %6279 = phi i1 [ false, %loop-memcpy-residual-header5040.post-loop-memcpy-expansion5036_crit_edge ], [ true, %loop-memcpy-residual-header5040 ]
  br i1 %6279, label %loop-memcpy-residual5039.preheader, label %Flow8095, !amdgpu.uniform !42

loop-memcpy-residual5039.preheader:               ; preds = %Flow8094
  %scevgep6941 = getelementptr i8, ptr %3713, i64 8
  %6280 = add i64 %1474, %3715
  %scevgep6942 = getelementptr i8, ptr %scevgep6941, i64 %6280
  %6281 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i2540, 6
  %scevgep6943 = getelementptr i8, ptr %scevgep6942, i64 %6281
  %scevgep6945 = getelementptr i8, ptr %3650, i64 %3715
  br label %loop-memcpy-residual5039, !amdgpu.uniform !42

Flow8173:                                         ; preds = %loop-memcpy-expansion5043
  %.lcssa9100 = phi i64 [ %3297, %loop-memcpy-expansion5043 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa9100)
  br label %Flow8174, !amdgpu.uniform !42

Flow8175:                                         ; preds = %Flow8172, %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i2202
  call void @llvm.amdgcn.end.cf.i64(i64 %3285)
  br label %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i2203, !amdgpu.uniform !42

loop-memcpy-residual-header5046:                  ; preds = %Flow8174
  %6282 = icmp ne i64 %3287, 0
  %6283 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %6282)
  %6284 = extractvalue { i1, i64 } %6283, 0
  %6285 = extractvalue { i1, i64 } %6283, 1
  br i1 %6284, label %loop-memcpy-residual5045.preheader, label %Flow8172

loop-memcpy-residual5045.preheader:               ; preds = %loop-memcpy-residual-header5046
  %umin6795 = call i64 @llvm.umin.i64(i64 %lsr.iv6790, i64 64)
  %6286 = lshr i64 %umin6795, 4
  %6287 = shl nuw nsw i64 %6286, 4
  %scevgep6796 = getelementptr i8, ptr %lsr.iv6783, i64 %6287
  %scevgep6786 = getelementptr i8, ptr %3286, i64 %1474
  %scevgep6787 = getelementptr i8, ptr %scevgep6786, i64 %3251
  %scevgep6793 = getelementptr i8, ptr %scevgep6787, i64 %6287
  br label %loop-memcpy-residual5045, !amdgpu.uniform !42

Flow8182:                                         ; preds = %loop-memcpy-expansion5049
  %.lcssa9097 = phi i64 [ %3249, %loop-memcpy-expansion5049 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa9097)
  br label %Flow8183, !amdgpu.uniform !42

loop-memcpy-residual-header5052:                  ; preds = %Flow8183
  %6288 = icmp eq i64 %3238, 0
  %6289 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %6288)
  %6290 = extractvalue { i1, i64 } %6289, 0
  %6291 = extractvalue { i1, i64 } %6289, 1
  br i1 %6290, label %loop-memcpy-residual-header5052.post-loop-memcpy-expansion5048_crit_edge, label %Flow8180

loop-memcpy-residual-header5052.post-loop-memcpy-expansion5048_crit_edge: ; preds = %loop-memcpy-residual-header5052
  %.pre7327 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i2181, 6
  br label %Flow8180, !amdgpu.uniform !42

Flow8180:                                         ; preds = %loop-memcpy-residual-header5052.post-loop-memcpy-expansion5048_crit_edge, %loop-memcpy-residual-header5052
  %6292 = phi i64 [ %.pre7327, %loop-memcpy-residual-header5052.post-loop-memcpy-expansion5048_crit_edge ], [ poison, %loop-memcpy-residual-header5052 ]
  %6293 = call { i1, i64 } @llvm.amdgcn.else.i64.i64(i64 %6291)
  %6294 = extractvalue { i1, i64 } %6293, 0
  %6295 = extractvalue { i1, i64 } %6293, 1
  br i1 %6294, label %loop-memcpy-residual5051.preheader, label %Flow8181

loop-memcpy-residual5051.preheader:               ; preds = %Flow8180
  %scevgep6773 = getelementptr i8, ptr %3237, i64 8
  %scevgep6774 = getelementptr i8, ptr %scevgep6773, i64 %1474
  %umin6777 = call i64 @llvm.umin.i64(i64 %lsr.iv.next6776.lcssa, i64 56)
  %6296 = lshr i64 %umin6777, 4
  %6297 = shl nuw nsw i64 %6296, 4
  %scevgep6778 = getelementptr i8, ptr %scevgep6774, i64 %6297
  %6298 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i2181, 6
  %scevgep6779 = getelementptr i8, ptr %scevgep6778, i64 %6298
  %scevgep6781 = getelementptr i8, ptr %call18, i64 %6297
  br label %loop-memcpy-residual5051, !amdgpu.uniform !42

Flow8258:                                         ; preds = %loop-memcpy-expansion5055
  %.lcssa9247 = phi i64 [ %2642, %loop-memcpy-expansion5055 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa9247)
  br label %Flow8259, !amdgpu.uniform !42

Flow8260:                                         ; preds = %Flow8257, %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i1676
  call void @llvm.amdgcn.end.cf.i64(i64 %2630)
  br label %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1677, !amdgpu.uniform !42

loop-memcpy-residual-header5058:                  ; preds = %Flow8259
  %6299 = icmp ne i64 %2632, 0
  %6300 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %6299)
  %6301 = extractvalue { i1, i64 } %6300, 0
  %6302 = extractvalue { i1, i64 } %6300, 1
  br i1 %6301, label %loop-memcpy-residual5057.preheader, label %Flow8257

loop-memcpy-residual5057.preheader:               ; preds = %loop-memcpy-residual-header5058
  %umin7104 = call i64 @llvm.umin.i64(i64 %lsr.iv7099, i64 64)
  %6303 = lshr i64 %umin7104, 4
  %6304 = shl nuw nsw i64 %6303, 4
  %scevgep7105 = getelementptr i8, ptr %lsr.iv7092, i64 %6304
  %scevgep7095 = getelementptr i8, ptr %2631, i64 %1474
  %scevgep7096 = getelementptr i8, ptr %scevgep7095, i64 %2596
  %scevgep7102 = getelementptr i8, ptr %scevgep7096, i64 %6304
  br label %loop-memcpy-residual5057, !amdgpu.uniform !42

Flow8267:                                         ; preds = %loop-memcpy-expansion5061
  %.lcssa9244 = phi i64 [ %2594, %loop-memcpy-expansion5061 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa9244)
  br label %Flow8268, !amdgpu.uniform !42

loop-memcpy-residual-header5064:                  ; preds = %Flow8268
  %6305 = icmp eq i64 %2583, 0
  %6306 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %6305)
  %6307 = extractvalue { i1, i64 } %6306, 0
  %6308 = extractvalue { i1, i64 } %6306, 1
  br i1 %6307, label %loop-memcpy-residual-header5064.post-loop-memcpy-expansion5060_crit_edge, label %Flow8265

loop-memcpy-residual-header5064.post-loop-memcpy-expansion5060_crit_edge: ; preds = %loop-memcpy-residual-header5064
  %.pre7319 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i1655, 6
  br label %Flow8265, !amdgpu.uniform !42

Flow8265:                                         ; preds = %loop-memcpy-residual-header5064.post-loop-memcpy-expansion5060_crit_edge, %loop-memcpy-residual-header5064
  %6309 = phi i64 [ %.pre7319, %loop-memcpy-residual-header5064.post-loop-memcpy-expansion5060_crit_edge ], [ poison, %loop-memcpy-residual-header5064 ]
  %6310 = call { i1, i64 } @llvm.amdgcn.else.i64.i64(i64 %6308)
  %6311 = extractvalue { i1, i64 } %6310, 0
  %6312 = extractvalue { i1, i64 } %6310, 1
  br i1 %6311, label %loop-memcpy-residual5063.preheader, label %Flow8266

loop-memcpy-residual5063.preheader:               ; preds = %Flow8265
  %scevgep7082 = getelementptr i8, ptr %2582, i64 8
  %scevgep7083 = getelementptr i8, ptr %scevgep7082, i64 %1474
  %umin7086 = call i64 @llvm.umin.i64(i64 %lsr.iv.next7085.lcssa, i64 56)
  %6313 = lshr i64 %umin7086, 4
  %6314 = shl nuw nsw i64 %6313, 4
  %scevgep7087 = getelementptr i8, ptr %scevgep7083, i64 %6314
  %6315 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i1655, 6
  %scevgep7088 = getelementptr i8, ptr %scevgep7087, i64 %6315
  %scevgep7090 = getelementptr i8, ptr %call18, i64 %6314
  br label %loop-memcpy-residual5063, !amdgpu.uniform !42

Flow8355:                                         ; preds = %loop-memcpy-expansion5067
  %.lcssa9030 = phi i64 [ %1909, %loop-memcpy-expansion5067 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa9030)
  br label %Flow8356, !amdgpu.uniform !42

Flow8357:                                         ; preds = %Flow8354, %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i1141
  call void @llvm.amdgcn.end.cf.i64(i64 %1897)
  br label %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1142, !amdgpu.uniform !42

loop-memcpy-residual-header5070:                  ; preds = %Flow8356
  %6316 = icmp ne i64 %1899, 0
  %6317 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %6316)
  %6318 = extractvalue { i1, i64 } %6317, 0
  %6319 = extractvalue { i1, i64 } %6317, 1
  br i1 %6318, label %loop-memcpy-residual5069.preheader, label %Flow8354

loop-memcpy-residual5069.preheader:               ; preds = %loop-memcpy-residual-header5070
  %umin6692 = call i64 @llvm.umin.i64(i64 %lsr.iv6687, i64 64)
  %6320 = lshr i64 %umin6692, 4
  %6321 = shl nuw nsw i64 %6320, 4
  %scevgep6693 = getelementptr i8, ptr %lsr.iv6680, i64 %6321
  %scevgep6683 = getelementptr i8, ptr %1898, i64 %1474
  %scevgep6684 = getelementptr i8, ptr %scevgep6683, i64 %1863
  %scevgep6690 = getelementptr i8, ptr %scevgep6684, i64 %6321
  br label %loop-memcpy-residual5069, !amdgpu.uniform !42

Flow8364:                                         ; preds = %loop-memcpy-expansion5073
  %.lcssa9027 = phi i64 [ %1861, %loop-memcpy-expansion5073 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa9027)
  br label %Flow8365, !amdgpu.uniform !42

loop-memcpy-residual-header5076:                  ; preds = %Flow8365
  %6322 = icmp eq i64 %1850, 0
  %6323 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %6322)
  %6324 = extractvalue { i1, i64 } %6323, 0
  %6325 = extractvalue { i1, i64 } %6323, 1
  br i1 %6324, label %loop-memcpy-residual-header5076.post-loop-memcpy-expansion5072_crit_edge, label %Flow8362

loop-memcpy-residual-header5076.post-loop-memcpy-expansion5072_crit_edge: ; preds = %loop-memcpy-residual-header5076
  %.pre7311 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i1120, 6
  br label %Flow8362, !amdgpu.uniform !42

Flow8362:                                         ; preds = %loop-memcpy-residual-header5076.post-loop-memcpy-expansion5072_crit_edge, %loop-memcpy-residual-header5076
  %6326 = phi i64 [ %.pre7311, %loop-memcpy-residual-header5076.post-loop-memcpy-expansion5072_crit_edge ], [ poison, %loop-memcpy-residual-header5076 ]
  %6327 = call { i1, i64 } @llvm.amdgcn.else.i64.i64(i64 %6325)
  %6328 = extractvalue { i1, i64 } %6327, 0
  %6329 = extractvalue { i1, i64 } %6327, 1
  br i1 %6328, label %loop-memcpy-residual5075.preheader, label %Flow8363

loop-memcpy-residual5075.preheader:               ; preds = %Flow8362
  %scevgep6669 = getelementptr i8, ptr %1849, i64 8
  %scevgep6670 = getelementptr i8, ptr %scevgep6669, i64 %1474
  %umin6673 = call i64 @llvm.umin.i64(i64 %lsr.iv.next6672.lcssa, i64 56)
  %6330 = lshr i64 %umin6673, 4
  %6331 = shl nuw nsw i64 %6330, 4
  %scevgep6674 = getelementptr i8, ptr %scevgep6670, i64 %6331
  %6332 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i1120, 6
  %scevgep6675 = getelementptr i8, ptr %scevgep6674, i64 %6332
  %scevgep6677 = getelementptr i8, ptr %call18, i64 %6331
  br label %loop-memcpy-residual5075, !amdgpu.uniform !42

Flow8494:                                         ; preds = %loop-memcpy-expansion5079
  %.lcssa9299 = phi i64 [ %739, %loop-memcpy-expansion5079 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa9299)
  br label %Flow8495, !amdgpu.uniform !42

Flow8496:                                         ; preds = %Flow8493, %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i390
  call void @llvm.amdgcn.end.cf.i64(i64 %727)
  br label %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i391, !amdgpu.uniform !42

loop-memcpy-residual-header5082:                  ; preds = %Flow8495
  %6333 = icmp ne i64 %729, 0
  %6334 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %6333)
  %6335 = extractvalue { i1, i64 } %6334, 0
  %6336 = extractvalue { i1, i64 } %6334, 1
  br i1 %6335, label %loop-memcpy-residual5081.preheader, label %Flow8493

loop-memcpy-residual5081.preheader:               ; preds = %loop-memcpy-residual-header5082
  %umin7172 = call i64 @llvm.umin.i64(i64 %lsr.iv7170, i64 64)
  %6337 = lshr i64 %umin7172, 4
  %6338 = shl nuw nsw i64 %6337, 4
  %scevgep7175 = getelementptr i8, ptr %lsr.iv7165, i64 %6338
  %scevgep7168 = getelementptr i8, ptr %728, i64 %78
  %scevgep7169 = getelementptr i8, ptr %scevgep7168, i64 %697
  %scevgep7173 = getelementptr i8, ptr %scevgep7169, i64 %6338
  br label %loop-memcpy-residual5081, !amdgpu.uniform !42

Flow8503:                                         ; preds = %loop-memcpy-expansion5085
  %.lcssa9304 = phi i64 [ %695, %loop-memcpy-expansion5085 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa9304)
  br label %Flow8504, !amdgpu.uniform !42

loop-memcpy-residual-header5088:                  ; preds = %Flow8504
  %6339 = icmp eq i64 %683, 0
  %6340 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %6339)
  %6341 = extractvalue { i1, i64 } %6340, 0
  %6342 = extractvalue { i1, i64 } %6340, 1
  br i1 %6341, label %loop-memcpy-residual-header5088.post-loop-memcpy-expansion5084_crit_edge, label %Flow8501

loop-memcpy-residual-header5088.post-loop-memcpy-expansion5084_crit_edge: ; preds = %loop-memcpy-residual-header5088
  %.pre7355 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i369, 6
  br label %Flow8501, !amdgpu.uniform !42

Flow8501:                                         ; preds = %loop-memcpy-residual-header5088.post-loop-memcpy-expansion5084_crit_edge, %loop-memcpy-residual-header5088
  %6343 = phi i64 [ %.pre7355, %loop-memcpy-residual-header5088.post-loop-memcpy-expansion5084_crit_edge ], [ poison, %loop-memcpy-residual-header5088 ]
  %6344 = call { i1, i64 } @llvm.amdgcn.else.i64.i64(i64 %6342)
  %6345 = extractvalue { i1, i64 } %6344, 0
  %6346 = extractvalue { i1, i64 } %6344, 1
  br i1 %6345, label %loop-memcpy-residual5087.preheader, label %Flow8502

loop-memcpy-residual5087.preheader:               ; preds = %Flow8501
  %6347 = lshr i64 %spec.select.i.i.i.i.i.i.i372, 4
  %6348 = shl nuw nsw i64 %6347, 4
  %6349 = add i64 %78, %6348
  %6350 = add i64 %6349, 8
  %scevgep7177 = getelementptr i8, ptr %682, i64 %6350
  %6351 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i369, 6
  %scevgep7178 = getelementptr i8, ptr %scevgep7177, i64 %6351
  %scevgep7180 = getelementptr i8, ptr %606, i64 %6348
  br label %loop-memcpy-residual5087, !amdgpu.uniform !42

Flow7922:                                         ; preds = %loop-memcpy-expansion5091
  %.lcssa8974 = phi i64 [ %1318, %loop-memcpy-expansion5091 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa8974)
  br label %Flow7923, !amdgpu.uniform !42

loop-memcpy-residual-header5094:                  ; preds = %Flow7923
  %6352 = icmp ne i64 %1307, 0
  %6353 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %6352)
  %6354 = extractvalue { i1, i64 } %6353, 0
  %6355 = extractvalue { i1, i64 } %6353, 1
  br i1 %6354, label %loop-memcpy-residual5093.preheader, label %Flow7921

loop-memcpy-residual5093.preheader:               ; preds = %loop-memcpy-residual-header5094
  %scevgep6576 = getelementptr i8, ptr %1297, i64 %1308
  %scevgep6578 = getelementptr i8, ptr %1276, i64 %1308
  br label %loop-memcpy-residual5093, !amdgpu.uniform !42

Flow7931:                                         ; preds = %loop-memcpy-expansion5097
  %.lcssa8970 = phi i64 [ %1269, %loop-memcpy-expansion5097 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa8970)
  br label %Flow7932, !amdgpu.uniform !42

loop-memcpy-residual-header5100:                  ; preds = %Flow7932
  %6356 = icmp ne i64 %1258, 0
  %6357 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %6356)
  %6358 = extractvalue { i1, i64 } %6357, 0
  %6359 = extractvalue { i1, i64 } %6357, 1
  br i1 %6358, label %loop-memcpy-residual5099.preheader, label %Flow7930

loop-memcpy-residual5099.preheader:               ; preds = %loop-memcpy-residual-header5100
  %scevgep6570 = getelementptr i8, ptr %1248, i64 %1259
  %scevgep6572 = getelementptr i8, ptr %call18, i64 %1259
  br label %loop-memcpy-residual5099, !amdgpu.uniform !42

Flow7720:                                         ; preds = %loop-memcpy-expansion5103
  %.lcssa8844 = phi i64 [ %5601, %loop-memcpy-expansion5103 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa8844)
  br label %Flow7721, !amdgpu.uniform !42

loop-memcpy-residual-header5106:                  ; preds = %Flow7721
  %6360 = icmp ne i64 %5590, 0
  %6361 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %6360)
  %6362 = extractvalue { i1, i64 } %6361, 0
  %6363 = extractvalue { i1, i64 } %6361, 1
  br i1 %6362, label %loop-memcpy-residual5105.preheader, label %Flow7719

loop-memcpy-residual5105.preheader:               ; preds = %loop-memcpy-residual-header5106
  %6364 = lshr i64 %writer.sroa.10.5.i.i.i.i4183, 4
  %6365 = trunc i64 %6364 to i32
  %6366 = shl i32 %6365, 4
  %6367 = add i32 %6366, 11
  %6368 = sub i32 %6367, %5589
  %scevgep6357 = getelementptr i8, ptr addrspace(5) %buffer.i.i.i2988, i32 %6368
  %6369 = shl nuw i64 %6364, 4
  %scevgep6360 = getelementptr i8, ptr %5579, i64 %6369
  br label %loop-memcpy-residual5105, !amdgpu.uniform !42

Flow7793:                                         ; preds = %loop-memcpy-expansion5109
  %.lcssa8882 = phi i64 [ %5150, %loop-memcpy-expansion5109 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa8882)
  br label %Flow7794, !amdgpu.uniform !42

loop-memcpy-residual-header5112:                  ; preds = %Flow7794
  %6370 = icmp ne i64 %5139, 0
  %6371 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %6370)
  %6372 = extractvalue { i1, i64 } %6371, 0
  %6373 = extractvalue { i1, i64 } %6371, 1
  br i1 %6372, label %loop-memcpy-residual5111.preheader, label %Flow7792

loop-memcpy-residual5111.preheader:               ; preds = %loop-memcpy-residual-header5112
  %6374 = lshr i64 %writer.sroa.10.5.i.i.i.i3799, 4
  %6375 = trunc i64 %6374 to i32
  %6376 = shl i32 %6375, 4
  %6377 = add i32 %6376, 11
  %6378 = sub i32 %6377, %5138
  %scevgep6405 = getelementptr i8, ptr addrspace(5) %buffer.i.i.i2988, i32 %6378
  %6379 = shl nuw i64 %6374, 4
  %scevgep6408 = getelementptr i8, ptr %5128, i64 %6379
  br label %loop-memcpy-residual5111, !amdgpu.uniform !42

Flow7876:                                         ; preds = %loop-memcpy-expansion5115
  %.lcssa8920 = phi i64 [ %4674, %loop-memcpy-expansion5115 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa8920)
  br label %Flow7877, !amdgpu.uniform !42

loop-memcpy-residual-header5118:                  ; preds = %Flow7877
  %6380 = icmp ne i64 %4663, 0
  %6381 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %6380)
  %6382 = extractvalue { i1, i64 } %6381, 0
  %6383 = extractvalue { i1, i64 } %6381, 1
  br i1 %6382, label %loop-memcpy-residual5117.preheader, label %Flow7875

loop-memcpy-residual5117.preheader:               ; preds = %loop-memcpy-residual-header5118
  %6384 = lshr i64 %writer.sroa.10.4.i.i.i.i3399, 4
  %6385 = trunc i64 %6384 to i32
  %6386 = shl i32 %6385, 4
  %6387 = add i32 %6386, 11
  %6388 = sub i32 %6387, %4662
  %scevgep6472 = getelementptr i8, ptr addrspace(5) %buffer.i.i.i2988, i32 %6388
  %6389 = shl nuw i64 %6384, 4
  %scevgep6475 = getelementptr i8, ptr %4652, i64 %6389
  br label %loop-memcpy-residual5117, !amdgpu.uniform !42

Flow8039:                                         ; preds = %loop-memcpy-expansion5121
  br label %Flow8040, !amdgpu.uniform !42

loop-memcpy-residual-header5124:                  ; preds = %Flow8040
  %6390 = icmp ne i64 %4217, 0
  br i1 %6390, label %loop-memcpy-residual5123.preheader, label %Flow8038, !amdgpu.uniform !42

loop-memcpy-residual5123.preheader:               ; preds = %loop-memcpy-residual-header5124
  %6391 = lshr i64 %writer.sroa.8.4.i.i.i.i3017, 4
  %6392 = trunc i64 %6391 to i32
  %6393 = shl i32 %6392, 4
  %6394 = sub i32 %6393, %4216
  %scevgep6878 = getelementptr i8, ptr addrspace(5) %scevgep6873, i32 %6394
  %scevgep6881 = getelementptr i8, ptr %4211, i64 %4218
  br label %loop-memcpy-residual5123, !amdgpu.uniform !42

Flow7978:                                         ; preds = %loop-memcpy-expansion5127
  br label %Flow7979, !amdgpu.uniform !42

loop-memcpy-residual-header5130:                  ; preds = %Flow7979
  %6395 = icmp ne i64 %3933, 0
  br i1 %6395, label %loop-memcpy-residual5129.preheader, label %Flow7977, !amdgpu.uniform !42

loop-memcpy-residual5129.preheader:               ; preds = %loop-memcpy-residual-header5130
  %6396 = lshr i64 %writer.sroa.8.4.i.i.i.i2742, 4
  %6397 = trunc i64 %6396 to i32
  %6398 = shl i32 %6397, 4
  %6399 = sub i32 %6398, %3932
  %scevgep6824 = getelementptr i8, ptr addrspace(5) %scevgep6819, i32 %6399
  %scevgep6827 = getelementptr i8, ptr %3927, i64 %3934
  br label %loop-memcpy-residual5129, !amdgpu.uniform !42

Flow8104:                                         ; preds = %loop-memcpy-expansion5133
  br label %Flow8105, !amdgpu.uniform !42

loop-memcpy-residual-header5136:                  ; preds = %Flow8105
  %6400 = icmp ne i64 %3656, 0
  br i1 %6400, label %loop-memcpy-residual5135.preheader, label %Flow8103, !amdgpu.uniform !42

loop-memcpy-residual5135.preheader:               ; preds = %loop-memcpy-residual-header5136
  %6401 = lshr i64 %writer.sroa.8.4.i.i.i.i, 4
  %6402 = trunc i64 %6401 to i32
  %6403 = shl i32 %6402, 4
  %6404 = sub i32 %6403, %3655
  %scevgep6932 = getelementptr i8, ptr addrspace(5) %scevgep6927, i32 %6404
  %scevgep6935 = getelementptr i8, ptr %3650, i64 %3657
  br label %loop-memcpy-residual5135, !amdgpu.uniform !42

Flow8511:                                         ; preds = %loop-memcpy-expansion5139
  %.lcssa9312 = phi i64 [ %628, %loop-memcpy-expansion5139 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa9312)
  br label %Flow8512, !amdgpu.uniform !42

loop-memcpy-residual-header5142:                  ; preds = %Flow8512
  %6405 = icmp ne i64 %617, 0
  %6406 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %6405)
  %6407 = extractvalue { i1, i64 } %6406, 0
  %6408 = extractvalue { i1, i64 } %6406, 1
  br i1 %6407, label %loop-memcpy-residual5141.preheader, label %Flow8510

loop-memcpy-residual5141.preheader:               ; preds = %loop-memcpy-residual-header5142
  %6409 = lshr i64 %writer.sroa.10.5.i.i.i.i, 4
  %6410 = trunc i64 %6409 to i32
  %6411 = shl i32 %6410, 4
  %6412 = add i32 %6411, 11
  %6413 = sub i32 %6412, %616
  %scevgep7185 = getelementptr i8, ptr addrspace(5) %buffer.i.i.i2988, i32 %6413
  %6414 = shl nuw i64 %6409, 4
  %scevgep7188 = getelementptr i8, ptr %606, i64 %6414
  br label %loop-memcpy-residual5141, !amdgpu.uniform !42

Flow8603:                                         ; preds = %loop-memcpy-expansion5145
  %.lcssa9357 = phi i64 [ %76, %loop-memcpy-expansion5145 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa9357)
  br label %Flow8604, !amdgpu.uniform !42

loop-memcpy-residual-header5148:                  ; preds = %Flow8604
  %6415 = icmp eq i64 %63, 0
  %6416 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %6415)
  %6417 = extractvalue { i1, i64 } %6416, 0
  %6418 = extractvalue { i1, i64 } %6416, 1
  br i1 %6417, label %loop-memcpy-residual-header5148.post-loop-memcpy-expansion5144_crit_edge, label %Flow8601

loop-memcpy-residual-header5148.post-loop-memcpy-expansion5144_crit_edge: ; preds = %loop-memcpy-residual-header5148
  %.pre7345 = shl nuw nsw i64 %sh_prom.i.i.i.i, 6
  %.pre7347 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i, 6
  br label %Flow8601, !amdgpu.uniform !42

Flow8601:                                         ; preds = %loop-memcpy-residual-header5148.post-loop-memcpy-expansion5144_crit_edge, %loop-memcpy-residual-header5148
  %6419 = phi i64 [ %.pre7345, %loop-memcpy-residual-header5148.post-loop-memcpy-expansion5144_crit_edge ], [ poison, %loop-memcpy-residual-header5148 ]
  %6420 = phi i64 [ %.pre7347, %loop-memcpy-residual-header5148.post-loop-memcpy-expansion5144_crit_edge ], [ poison, %loop-memcpy-residual-header5148 ]
  %6421 = call { i1, i64 } @llvm.amdgcn.else.i64.i64(i64 %6418)
  %6422 = extractvalue { i1, i64 } %6421, 0
  %6423 = extractvalue { i1, i64 } %6421, 1
  br i1 %6422, label %loop-memcpy-residual5147.preheader, label %Flow8602

loop-memcpy-residual5147.preheader:               ; preds = %Flow8601
  %6424 = shl nuw nsw i64 %sh_prom.i.i.i.i, 6
  %6425 = lshr i64 %spec.select.i.i.i.i.i.i.i, 4
  %6426 = shl nuw nsw i64 %6425, 4
  %6427 = add i64 %6424, %6426
  %6428 = add i64 %6427, 8
  %scevgep7269 = getelementptr i8, ptr %62, i64 %6428
  %6429 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i, 6
  %scevgep7270 = getelementptr i8, ptr %scevgep7269, i64 %6429
  %scevgep7272 = getelementptr i8, ptr addrspace(4) %cond, i64 %6426
  br label %loop-memcpy-residual5147, !amdgpu.uniform !42

Flow8421:                                         ; preds = %loop-memcpy-expansion5151
  %.lcssa8984 = phi i64 [ %1472, %loop-memcpy-expansion5151 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa8984)
  br label %Flow8422, !amdgpu.uniform !42

loop-memcpy-residual-header5154:                  ; preds = %Flow8422
  %6430 = icmp eq i64 %1460, 0
  %6431 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %6430)
  %6432 = extractvalue { i1, i64 } %6431, 0
  %6433 = extractvalue { i1, i64 } %6431, 1
  br i1 %6432, label %loop-memcpy-residual-header5154.post-loop-memcpy-expansion5150_crit_edge, label %Flow8419

loop-memcpy-residual-header5154.post-loop-memcpy-expansion5150_crit_edge: ; preds = %loop-memcpy-residual-header5154
  %.pre7303 = shl nuw nsw i64 %sh_prom.i.i.i.i756, 6
  %.pre7305 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i800, 6
  br label %Flow8419, !amdgpu.uniform !42

Flow8419:                                         ; preds = %loop-memcpy-residual-header5154.post-loop-memcpy-expansion5150_crit_edge, %loop-memcpy-residual-header5154
  %6434 = phi i64 [ %.pre7303, %loop-memcpy-residual-header5154.post-loop-memcpy-expansion5150_crit_edge ], [ poison, %loop-memcpy-residual-header5154 ]
  %6435 = phi i64 [ %.pre7305, %loop-memcpy-residual-header5154.post-loop-memcpy-expansion5150_crit_edge ], [ poison, %loop-memcpy-residual-header5154 ]
  %6436 = call { i1, i64 } @llvm.amdgcn.else.i64.i64(i64 %6433)
  %6437 = extractvalue { i1, i64 } %6436, 0
  %6438 = extractvalue { i1, i64 } %6436, 1
  br i1 %6437, label %loop-memcpy-residual5153.preheader, label %Flow8420

loop-memcpy-residual5153.preheader:               ; preds = %Flow8419
  %scevgep6594 = getelementptr i8, ptr %1459, i64 8
  %6439 = shl nuw nsw i64 %sh_prom.i.i.i.i756, 6
  %6440 = add i64 %6439, %1461
  %scevgep6595 = getelementptr i8, ptr %scevgep6594, i64 %6440
  %6441 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i800, 6
  %scevgep6596 = getelementptr i8, ptr %scevgep6595, i64 %6441
  %scevgep6598 = getelementptr i8, ptr addrspace(4) %cond, i64 %1461
  br label %loop-memcpy-residual5153, !amdgpu.uniform !42

Flow8239:                                         ; preds = %loop-memcpy-expansion5157
  %.lcssa9054 = phi i64 [ %2860, %loop-memcpy-expansion5157 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa9054)
  br label %Flow8240, !amdgpu.uniform !42

loop-memcpy-residual-header5160:                  ; preds = %Flow8240
  %6442 = icmp eq i64 %2849, 0
  %6443 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %6442)
  %6444 = extractvalue { i1, i64 } %6443, 0
  %6445 = extractvalue { i1, i64 } %6443, 1
  br i1 %6444, label %loop-memcpy-residual-header5160.post-loop-memcpy-expansion5156_crit_edge, label %Flow8237

loop-memcpy-residual-header5160.post-loop-memcpy-expansion5156_crit_edge: ; preds = %loop-memcpy-residual-header5160
  %.pre7321 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i1861, 6
  br label %Flow8237, !amdgpu.uniform !42

Flow8237:                                         ; preds = %loop-memcpy-residual-header5160.post-loop-memcpy-expansion5156_crit_edge, %loop-memcpy-residual-header5160
  %6446 = phi i64 [ %.pre7321, %loop-memcpy-residual-header5160.post-loop-memcpy-expansion5156_crit_edge ], [ poison, %loop-memcpy-residual-header5160 ]
  %6447 = call { i1, i64 } @llvm.amdgcn.else.i64.i64(i64 %6445)
  %6448 = extractvalue { i1, i64 } %6447, 0
  %6449 = extractvalue { i1, i64 } %6447, 1
  br i1 %6448, label %loop-memcpy-residual5159.preheader, label %Flow8238

loop-memcpy-residual5159.preheader:               ; preds = %Flow8237
  %scevgep6702 = getelementptr i8, ptr %2848, i64 8
  %6450 = add i64 %1474, %2850
  %scevgep6703 = getelementptr i8, ptr %scevgep6702, i64 %6450
  %6451 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i1861, 6
  %scevgep6704 = getelementptr i8, ptr %scevgep6703, i64 %6451
  %scevgep6706 = getelementptr i8, ptr addrspace(4) %cond, i64 %2850
  br label %loop-memcpy-residual5159, !amdgpu.uniform !42

Flow8594:                                         ; preds = %loop-memcpy-expansion5163
  %.lcssa9352 = phi i64 [ %126, %loop-memcpy-expansion5163 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa9352)
  br label %Flow8595, !amdgpu.uniform !42

Flow8596:                                         ; preds = %Flow8593, %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i
  call void @llvm.amdgcn.end.cf.i64(i64 %114)
  br label %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i, !amdgpu.uniform !42

loop-memcpy-residual-header5166:                  ; preds = %Flow8595
  %6452 = icmp ne i64 %116, 0
  %6453 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %6452)
  %6454 = extractvalue { i1, i64 } %6453, 0
  %6455 = extractvalue { i1, i64 } %6453, 1
  br i1 %6454, label %loop-memcpy-residual5165.preheader, label %Flow8593

loop-memcpy-residual5165.preheader:               ; preds = %loop-memcpy-residual-header5166
  %umin7264 = call i64 @llvm.umin.i64(i64 %lsr.iv7262, i64 64)
  %6456 = lshr i64 %umin7264, 4
  %6457 = shl nuw nsw i64 %6456, 4
  %scevgep7267 = getelementptr i8, ptr addrspace(4) %lsr.iv7257, i64 %6457
  %scevgep7260 = getelementptr i8, ptr %115, i64 %78
  %scevgep7261 = getelementptr i8, ptr %scevgep7260, i64 %79
  %scevgep7265 = getelementptr i8, ptr %scevgep7261, i64 %6457
  br label %loop-memcpy-residual5165, !amdgpu.uniform !42

Flow8584:                                         ; preds = %loop-memcpy-expansion5169
  br label %Flow8585, !amdgpu.uniform !42

loop-memcpy-residual-header5172:                  ; preds = %Flow8585
  %6458 = icmp eq i64 %199, 0
  br i1 %6458, label %loop-memcpy-residual-header5172.post-loop-memcpy-expansion5168_crit_edge, label %Flow8582, !amdgpu.uniform !42

loop-memcpy-residual-header5172.post-loop-memcpy-expansion5168_crit_edge: ; preds = %loop-memcpy-residual-header5172
  %.pre7349 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i56, 6
  br label %Flow8582, !amdgpu.uniform !42

Flow8582:                                         ; preds = %loop-memcpy-residual-header5172.post-loop-memcpy-expansion5168_crit_edge, %loop-memcpy-residual-header5172
  %6459 = phi i64 [ %.pre7349, %loop-memcpy-residual-header5172.post-loop-memcpy-expansion5168_crit_edge ], [ poison, %loop-memcpy-residual-header5172 ]
  %6460 = phi i1 [ false, %loop-memcpy-residual-header5172.post-loop-memcpy-expansion5168_crit_edge ], [ true, %loop-memcpy-residual-header5172 ]
  br i1 %6460, label %loop-memcpy-residual5171.preheader, label %Flow8583, !amdgpu.uniform !42

loop-memcpy-residual5171.preheader:               ; preds = %Flow8582
  %6461 = add nuw nsw i64 %78, 8
  %scevgep7249 = getelementptr i8, ptr %198, i64 %6461
  %6462 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i56, 6
  %scevgep7250 = getelementptr i8, ptr %scevgep7249, i64 %6462
  br label %loop-memcpy-residual5171, !amdgpu.uniform !42

Flow8575:                                         ; preds = %loop-memcpy-expansion5175
  br label %Flow8576, !amdgpu.uniform !42

Flow8577:                                         ; preds = %Flow8574, %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i77
  br label %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i78, !amdgpu.uniform !42

loop-memcpy-residual-header5178:                  ; preds = %Flow8576
  %6463 = icmp ne i64 %234, 0
  br i1 %6463, label %loop-memcpy-residual5177.preheader, label %Flow8574, !amdgpu.uniform !42

loop-memcpy-residual5177.preheader:               ; preds = %loop-memcpy-residual-header5178
  %umin7244 = call i64 @llvm.umin.i64(i64 %lsr.iv7242, i64 64)
  %6464 = lshr i64 %umin7244, 4
  %6465 = shl nuw nsw i64 %6464, 4
  %scevgep7247 = getelementptr i8, ptr addrspace(4) %lsr.iv7237, i64 %6465
  %scevgep7240 = getelementptr i8, ptr %233, i64 %78
  %scevgep7241 = getelementptr i8, ptr %scevgep7240, i64 %208
  %scevgep7245 = getelementptr i8, ptr %scevgep7241, i64 %6465
  br label %loop-memcpy-residual5177, !amdgpu.uniform !42

Flow8565:                                         ; preds = %loop-memcpy-expansion5181
  %.lcssa9332 = phi i64 [ %327, %loop-memcpy-expansion5181 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa9332)
  br label %Flow8566, !amdgpu.uniform !42

loop-memcpy-residual-header5184:                  ; preds = %Flow8566
  %6466 = icmp eq i64 %315, 0
  %6467 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %6466)
  %6468 = extractvalue { i1, i64 } %6467, 0
  %6469 = extractvalue { i1, i64 } %6467, 1
  br i1 %6468, label %loop-memcpy-residual-header5184.post-loop-memcpy-expansion5180_crit_edge, label %Flow8563

loop-memcpy-residual-header5184.post-loop-memcpy-expansion5180_crit_edge: ; preds = %loop-memcpy-residual-header5184
  %.pre7351 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i162, 6
  br label %Flow8563, !amdgpu.uniform !42

Flow8563:                                         ; preds = %loop-memcpy-residual-header5184.post-loop-memcpy-expansion5180_crit_edge, %loop-memcpy-residual-header5184
  %6470 = phi i64 [ %.pre7351, %loop-memcpy-residual-header5184.post-loop-memcpy-expansion5180_crit_edge ], [ poison, %loop-memcpy-residual-header5184 ]
  %6471 = call { i1, i64 } @llvm.amdgcn.else.i64.i64(i64 %6469)
  %6472 = extractvalue { i1, i64 } %6471, 0
  %6473 = extractvalue { i1, i64 } %6471, 1
  br i1 %6472, label %loop-memcpy-residual5183.preheader, label %Flow8564

loop-memcpy-residual5183.preheader:               ; preds = %Flow8563
  %6474 = lshr i64 %spec.select.i.i.i.i.i.i.i165, 4
  %6475 = shl nuw nsw i64 %6474, 4
  %6476 = add i64 %78, %6475
  %6477 = add i64 %6476, 8
  %scevgep7225 = getelementptr i8, ptr %314, i64 %6477
  %6478 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i162, 6
  %scevgep7226 = getelementptr i8, ptr %scevgep7225, i64 %6478
  %scevgep7228 = getelementptr i8, ptr addrspace(4) %cond6, i64 %6475
  br label %loop-memcpy-residual5183, !amdgpu.uniform !42

Flow8383:                                         ; preds = %loop-memcpy-expansion5187
  %.lcssa9011 = phi i64 [ %1724, %loop-memcpy-expansion5187 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa9011)
  br label %Flow8384, !amdgpu.uniform !42

loop-memcpy-residual-header5190:                  ; preds = %Flow8384
  %6479 = icmp eq i64 %1713, 0
  %6480 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %6479)
  %6481 = extractvalue { i1, i64 } %6480, 0
  %6482 = extractvalue { i1, i64 } %6480, 1
  br i1 %6481, label %loop-memcpy-residual-header5190.post-loop-memcpy-expansion5186_crit_edge, label %Flow8381

loop-memcpy-residual-header5190.post-loop-memcpy-expansion5186_crit_edge: ; preds = %loop-memcpy-residual-header5190
  %.pre7309 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i1013, 6
  br label %Flow8381, !amdgpu.uniform !42

Flow8381:                                         ; preds = %loop-memcpy-residual-header5190.post-loop-memcpy-expansion5186_crit_edge, %loop-memcpy-residual-header5190
  %6483 = phi i64 [ %.pre7309, %loop-memcpy-residual-header5190.post-loop-memcpy-expansion5186_crit_edge ], [ poison, %loop-memcpy-residual-header5190 ]
  %6484 = call { i1, i64 } @llvm.amdgcn.else.i64.i64(i64 %6482)
  %6485 = extractvalue { i1, i64 } %6484, 0
  %6486 = extractvalue { i1, i64 } %6484, 1
  br i1 %6485, label %loop-memcpy-residual5189.preheader, label %Flow8382

loop-memcpy-residual5189.preheader:               ; preds = %Flow8381
  %scevgep6642 = getelementptr i8, ptr %1712, i64 8
  %6487 = add i64 %1474, %1714
  %scevgep6643 = getelementptr i8, ptr %scevgep6642, i64 %6487
  %6488 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i1013, 6
  %scevgep6644 = getelementptr i8, ptr %scevgep6643, i64 %6488
  %scevgep6646 = getelementptr i8, ptr addrspace(4) %cond6, i64 %1714
  br label %loop-memcpy-residual5189, !amdgpu.uniform !42

Flow8286:                                         ; preds = %loop-memcpy-expansion5193
  %.lcssa9228 = phi i64 [ %2456, %loop-memcpy-expansion5193 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa9228)
  br label %Flow8287, !amdgpu.uniform !42

loop-memcpy-residual-header5196:                  ; preds = %Flow8287
  %6489 = icmp eq i64 %2445, 0
  %6490 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %6489)
  %6491 = extractvalue { i1, i64 } %6490, 0
  %6492 = extractvalue { i1, i64 } %6490, 1
  br i1 %6491, label %loop-memcpy-residual-header5196.post-loop-memcpy-expansion5192_crit_edge, label %Flow8284

loop-memcpy-residual-header5196.post-loop-memcpy-expansion5192_crit_edge: ; preds = %loop-memcpy-residual-header5196
  %.pre7317 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i1548, 6
  br label %Flow8284, !amdgpu.uniform !42

Flow8284:                                         ; preds = %loop-memcpy-residual-header5196.post-loop-memcpy-expansion5192_crit_edge, %loop-memcpy-residual-header5196
  %6493 = phi i64 [ %.pre7317, %loop-memcpy-residual-header5196.post-loop-memcpy-expansion5192_crit_edge ], [ poison, %loop-memcpy-residual-header5196 ]
  %6494 = call { i1, i64 } @llvm.amdgcn.else.i64.i64(i64 %6492)
  %6495 = extractvalue { i1, i64 } %6494, 0
  %6496 = extractvalue { i1, i64 } %6494, 1
  br i1 %6495, label %loop-memcpy-residual5195.preheader, label %Flow8285

loop-memcpy-residual5195.preheader:               ; preds = %Flow8284
  %scevgep7057 = getelementptr i8, ptr %2444, i64 8
  %6497 = add i64 %1474, %2446
  %scevgep7058 = getelementptr i8, ptr %scevgep7057, i64 %6497
  %6498 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i1548, 6
  %scevgep7059 = getelementptr i8, ptr %scevgep7058, i64 %6498
  %scevgep7061 = getelementptr i8, ptr addrspace(4) %cond6, i64 %2446
  br label %loop-memcpy-residual5195, !amdgpu.uniform !42

Flow8201:                                         ; preds = %loop-memcpy-expansion5199
  %.lcssa9081 = phi i64 [ %3111, %loop-memcpy-expansion5199 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa9081)
  br label %Flow8202, !amdgpu.uniform !42

loop-memcpy-residual-header5202:                  ; preds = %Flow8202
  %6499 = icmp eq i64 %3100, 0
  %6500 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %6499)
  %6501 = extractvalue { i1, i64 } %6500, 0
  %6502 = extractvalue { i1, i64 } %6500, 1
  br i1 %6501, label %loop-memcpy-residual-header5202.post-loop-memcpy-expansion5198_crit_edge, label %Flow8199

loop-memcpy-residual-header5202.post-loop-memcpy-expansion5198_crit_edge: ; preds = %loop-memcpy-residual-header5202
  %.pre7325 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i2074, 6
  br label %Flow8199, !amdgpu.uniform !42

Flow8199:                                         ; preds = %loop-memcpy-residual-header5202.post-loop-memcpy-expansion5198_crit_edge, %loop-memcpy-residual-header5202
  %6503 = phi i64 [ %.pre7325, %loop-memcpy-residual-header5202.post-loop-memcpy-expansion5198_crit_edge ], [ poison, %loop-memcpy-residual-header5202 ]
  %6504 = call { i1, i64 } @llvm.amdgcn.else.i64.i64(i64 %6502)
  %6505 = extractvalue { i1, i64 } %6504, 0
  %6506 = extractvalue { i1, i64 } %6504, 1
  br i1 %6505, label %loop-memcpy-residual5201.preheader, label %Flow8200

loop-memcpy-residual5201.preheader:               ; preds = %Flow8199
  %scevgep6748 = getelementptr i8, ptr %3099, i64 8
  %6507 = add i64 %1474, %3101
  %scevgep6749 = getelementptr i8, ptr %scevgep6748, i64 %6507
  %6508 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i2074, 6
  %scevgep6750 = getelementptr i8, ptr %scevgep6749, i64 %6508
  %scevgep6752 = getelementptr i8, ptr addrspace(4) %cond6, i64 %3101
  br label %loop-memcpy-residual5201, !amdgpu.uniform !42

Flow8556:                                         ; preds = %loop-memcpy-expansion5205
  %.lcssa9327 = phi i64 [ %376, %loop-memcpy-expansion5205 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa9327)
  br label %Flow8557, !amdgpu.uniform !42

Flow8558:                                         ; preds = %Flow8555, %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i183
  call void @llvm.amdgcn.end.cf.i64(i64 %364)
  br label %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i184, !amdgpu.uniform !42

loop-memcpy-residual-header5208:                  ; preds = %Flow8557
  %6509 = icmp ne i64 %366, 0
  %6510 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %6509)
  %6511 = extractvalue { i1, i64 } %6510, 0
  %6512 = extractvalue { i1, i64 } %6510, 1
  br i1 %6511, label %loop-memcpy-residual5207.preheader, label %Flow8555

loop-memcpy-residual5207.preheader:               ; preds = %loop-memcpy-residual-header5208
  %umin7220 = call i64 @llvm.umin.i64(i64 %lsr.iv7218, i64 64)
  %6513 = lshr i64 %umin7220, 4
  %6514 = shl nuw nsw i64 %6513, 4
  %scevgep7223 = getelementptr i8, ptr addrspace(4) %lsr.iv7213, i64 %6514
  %scevgep7216 = getelementptr i8, ptr %365, i64 %78
  %scevgep7217 = getelementptr i8, ptr %scevgep7216, i64 %329
  %scevgep7221 = getelementptr i8, ptr %scevgep7217, i64 %6514
  br label %loop-memcpy-residual5207, !amdgpu.uniform !42

Flow8546:                                         ; preds = %loop-memcpy-expansion5211
  br label %Flow8547, !amdgpu.uniform !42

loop-memcpy-residual-header5214:                  ; preds = %Flow8547
  %6515 = icmp eq i64 %449, 0
  br i1 %6515, label %loop-memcpy-residual-header5214.post-loop-memcpy-expansion5210_crit_edge, label %Flow8544, !amdgpu.uniform !42

loop-memcpy-residual-header5214.post-loop-memcpy-expansion5210_crit_edge: ; preds = %loop-memcpy-residual-header5214
  %.pre7353 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i267, 6
  br label %Flow8544, !amdgpu.uniform !42

Flow8544:                                         ; preds = %loop-memcpy-residual-header5214.post-loop-memcpy-expansion5210_crit_edge, %loop-memcpy-residual-header5214
  %6516 = phi i64 [ %.pre7353, %loop-memcpy-residual-header5214.post-loop-memcpy-expansion5210_crit_edge ], [ poison, %loop-memcpy-residual-header5214 ]
  %6517 = phi i1 [ false, %loop-memcpy-residual-header5214.post-loop-memcpy-expansion5210_crit_edge ], [ true, %loop-memcpy-residual-header5214 ]
  br i1 %6517, label %loop-memcpy-residual5213.preheader, label %Flow8545, !amdgpu.uniform !42

loop-memcpy-residual5213.preheader:               ; preds = %Flow8544
  %6518 = add nuw nsw i64 %78, 8
  %scevgep7205 = getelementptr i8, ptr %448, i64 %6518
  %6519 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i267, 6
  %scevgep7206 = getelementptr i8, ptr %scevgep7205, i64 %6519
  br label %loop-memcpy-residual5213, !amdgpu.uniform !42

Flow8537:                                         ; preds = %loop-memcpy-expansion5217
  br label %Flow8538, !amdgpu.uniform !42

Flow8539:                                         ; preds = %Flow8536, %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i288
  br label %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i289, !amdgpu.uniform !42

loop-memcpy-residual-header5220:                  ; preds = %Flow8538
  %6520 = icmp ne i64 %484, 0
  br i1 %6520, label %loop-memcpy-residual5219.preheader, label %Flow8536, !amdgpu.uniform !42

loop-memcpy-residual5219.preheader:               ; preds = %loop-memcpy-residual-header5220
  %umin7200 = call i64 @llvm.umin.i64(i64 %lsr.iv7198, i64 64)
  %scevgep7203 = getelementptr i8, ptr addrspace(4) %lsr.iv7193, i64 %umin7200
  %scevgep7196 = getelementptr i8, ptr %483, i64 %78
  %scevgep7197 = getelementptr i8, ptr %scevgep7196, i64 %458
  %scevgep7201 = getelementptr i8, ptr %scevgep7197, i64 %umin7200
  br label %loop-memcpy-residual5219, !amdgpu.uniform !42

Flow8484:                                         ; preds = %loop-memcpy-expansion5223
  br label %Flow8485, !amdgpu.uniform !42

loop-memcpy-residual-header5226:                  ; preds = %Flow8485
  %6521 = icmp eq i64 %812, 0
  br i1 %6521, label %loop-memcpy-residual-header5226.post-loop-memcpy-expansion5222_crit_edge, label %Flow8482, !amdgpu.uniform !42

loop-memcpy-residual-header5226.post-loop-memcpy-expansion5222_crit_edge: ; preds = %loop-memcpy-residual-header5226
  %.pre7357 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i472, 6
  br label %Flow8482, !amdgpu.uniform !42

Flow8482:                                         ; preds = %loop-memcpy-residual-header5226.post-loop-memcpy-expansion5222_crit_edge, %loop-memcpy-residual-header5226
  %6522 = phi i64 [ %.pre7357, %loop-memcpy-residual-header5226.post-loop-memcpy-expansion5222_crit_edge ], [ poison, %loop-memcpy-residual-header5226 ]
  %6523 = phi i1 [ false, %loop-memcpy-residual-header5226.post-loop-memcpy-expansion5222_crit_edge ], [ true, %loop-memcpy-residual-header5226 ]
  br i1 %6523, label %loop-memcpy-residual5225.preheader, label %Flow8483, !amdgpu.uniform !42

loop-memcpy-residual5225.preheader:               ; preds = %Flow8482
  %6524 = add nuw nsw i64 %78, 8
  %scevgep7157 = getelementptr i8, ptr %811, i64 %6524
  %6525 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i472, 6
  %scevgep7158 = getelementptr i8, ptr %scevgep7157, i64 %6525
  br label %loop-memcpy-residual5225, !amdgpu.uniform !42

Flow8475:                                         ; preds = %loop-memcpy-expansion5229
  br label %Flow8476, !amdgpu.uniform !42

Flow8477:                                         ; preds = %Flow8474, %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i493
  br label %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i494, !amdgpu.uniform !42

loop-memcpy-residual-header5232:                  ; preds = %Flow8476
  %6526 = icmp ne i64 %847, 0
  br i1 %6526, label %loop-memcpy-residual5231.preheader, label %Flow8474, !amdgpu.uniform !42

loop-memcpy-residual5231.preheader:               ; preds = %loop-memcpy-residual-header5232
  %umin7152 = call i64 @llvm.umin.i64(i64 %lsr.iv7150, i64 64)
  %6527 = lshr i64 %umin7152, 4
  %6528 = shl nuw nsw i64 %6527, 4
  %scevgep7155 = getelementptr i8, ptr addrspace(4) %lsr.iv7145, i64 %6528
  %scevgep7148 = getelementptr i8, ptr %846, i64 %78
  %scevgep7149 = getelementptr i8, ptr %scevgep7148, i64 %821
  %scevgep7153 = getelementptr i8, ptr %scevgep7149, i64 %6528
  br label %loop-memcpy-residual5231, !amdgpu.uniform !42

Flow8464:                                         ; preds = %loop-memcpy-expansion5235
  br label %Flow8465, !amdgpu.uniform !42

loop-memcpy-residual-header5238:                  ; preds = %Flow8465
  %6529 = icmp eq i64 %926, 0
  br i1 %6529, label %loop-memcpy-residual-header5238.post-loop-memcpy-expansion5234_crit_edge, label %Flow8462, !amdgpu.uniform !42

loop-memcpy-residual-header5238.post-loop-memcpy-expansion5234_crit_edge: ; preds = %loop-memcpy-residual-header5238
  %.pre7359 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i578, 6
  br label %Flow8462, !amdgpu.uniform !42

Flow8462:                                         ; preds = %loop-memcpy-residual-header5238.post-loop-memcpy-expansion5234_crit_edge, %loop-memcpy-residual-header5238
  %6530 = phi i64 [ %.pre7359, %loop-memcpy-residual-header5238.post-loop-memcpy-expansion5234_crit_edge ], [ poison, %loop-memcpy-residual-header5238 ]
  %6531 = phi i1 [ false, %loop-memcpy-residual-header5238.post-loop-memcpy-expansion5234_crit_edge ], [ true, %loop-memcpy-residual-header5238 ]
  br i1 %6531, label %loop-memcpy-residual5237.preheader, label %Flow8463, !amdgpu.uniform !42

loop-memcpy-residual5237.preheader:               ; preds = %Flow8462
  %6532 = add nuw nsw i64 %78, 8
  %scevgep7138 = getelementptr i8, ptr %925, i64 %6532
  %6533 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i578, 6
  %scevgep7139 = getelementptr i8, ptr %scevgep7138, i64 %6533
  br label %loop-memcpy-residual5237, !amdgpu.uniform !42

Flow8455:                                         ; preds = %loop-memcpy-expansion5241
  br label %Flow8456, !amdgpu.uniform !42

Flow8457:                                         ; preds = %Flow8454, %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i599
  br label %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i600, !amdgpu.uniform !42

loop-memcpy-residual-header5244:                  ; preds = %Flow8456
  %6534 = icmp ne i64 %961, 0
  br i1 %6534, label %loop-memcpy-residual5243.preheader, label %Flow8454, !amdgpu.uniform !42

loop-memcpy-residual5243.preheader:               ; preds = %loop-memcpy-residual-header5244
  %umin7133 = call i64 @llvm.umin.i64(i64 %lsr.iv7131, i64 64)
  %6535 = lshr i64 %umin7133, 4
  %6536 = shl nuw nsw i64 %6535, 4
  %scevgep7136 = getelementptr i8, ptr addrspace(4) %lsr.iv7126, i64 %6536
  %scevgep7129 = getelementptr i8, ptr %960, i64 %78
  %scevgep7130 = getelementptr i8, ptr %scevgep7129, i64 %935
  %scevgep7134 = getelementptr i8, ptr %scevgep7130, i64 %6536
  br label %loop-memcpy-residual5243, !amdgpu.uniform !42

Flow8445:                                         ; preds = %loop-memcpy-expansion5247
  br label %Flow8446, !amdgpu.uniform !42

loop-memcpy-residual-header5250:                  ; preds = %Flow8446
  %6537 = icmp eq i64 %1037, 0
  br i1 %6537, label %loop-memcpy-residual-header5250.post-loop-memcpy-expansion5246_crit_edge, label %Flow8443, !amdgpu.uniform !42

loop-memcpy-residual-header5250.post-loop-memcpy-expansion5246_crit_edge: ; preds = %loop-memcpy-residual-header5250
  %.pre7361 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i684, 6
  br label %Flow8443, !amdgpu.uniform !42

Flow8443:                                         ; preds = %loop-memcpy-residual-header5250.post-loop-memcpy-expansion5246_crit_edge, %loop-memcpy-residual-header5250
  %6538 = phi i64 [ %.pre7361, %loop-memcpy-residual-header5250.post-loop-memcpy-expansion5246_crit_edge ], [ poison, %loop-memcpy-residual-header5250 ]
  %6539 = phi i1 [ false, %loop-memcpy-residual-header5250.post-loop-memcpy-expansion5246_crit_edge ], [ true, %loop-memcpy-residual-header5250 ]
  br i1 %6539, label %loop-memcpy-residual5249.preheader, label %Flow8444, !amdgpu.uniform !42

loop-memcpy-residual5249.preheader:               ; preds = %Flow8443
  %6540 = add nuw nsw i64 %78, 24
  %scevgep7119 = getelementptr i8, ptr %1036, i64 %6540
  %6541 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i684, 6
  %scevgep7120 = getelementptr i8, ptr %scevgep7119, i64 %6541
  br label %loop-memcpy-residual5249, !amdgpu.uniform !42

Flow8436:                                         ; preds = %loop-memcpy-expansion5253
  br label %Flow8437, !amdgpu.uniform !42

Flow8438:                                         ; preds = %Flow8435, %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i705
  br label %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i706, !amdgpu.uniform !42

loop-memcpy-residual-header5256:                  ; preds = %Flow8437
  %6542 = icmp ne i64 %1072, 0
  br i1 %6542, label %loop-memcpy-residual5255.preheader, label %Flow8435, !amdgpu.uniform !42

loop-memcpy-residual5255.preheader:               ; preds = %loop-memcpy-residual-header5256
  %umin7114 = call i64 @llvm.umin.i64(i64 %lsr.iv7112, i64 64)
  %6543 = lshr i64 %umin7114, 4
  %6544 = shl nuw nsw i64 %6543, 4
  %scevgep7117 = getelementptr i8, ptr addrspace(4) %lsr.iv7107, i64 %6544
  %scevgep7110 = getelementptr i8, ptr %1071, i64 %78
  %scevgep7111 = getelementptr i8, ptr %scevgep7110, i64 %1046
  %scevgep7115 = getelementptr i8, ptr %scevgep7111, i64 %6544
  br label %loop-memcpy-residual5255, !amdgpu.uniform !42

Flow8412:                                         ; preds = %loop-memcpy-expansion5259
  %.lcssa8987 = phi i64 [ %1524, %loop-memcpy-expansion5259 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa8987)
  br label %Flow8413, !amdgpu.uniform !42

Flow8414:                                         ; preds = %Flow8411, %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i821
  call void @llvm.amdgcn.end.cf.i64(i64 %1512)
  br label %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i822, !amdgpu.uniform !42

loop-memcpy-residual-header5262:                  ; preds = %Flow8413
  %6545 = icmp ne i64 %1514, 0
  %6546 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %6545)
  %6547 = extractvalue { i1, i64 } %6546, 0
  %6548 = extractvalue { i1, i64 } %6546, 1
  br i1 %6547, label %loop-memcpy-residual5261.preheader, label %Flow8411

loop-memcpy-residual5261.preheader:               ; preds = %loop-memcpy-residual-header5262
  %umin6608 = call i64 @llvm.umin.i64(i64 %lsr.iv6606, i64 64)
  %6549 = lshr i64 %umin6608, 4
  %6550 = shl nuw nsw i64 %6549, 4
  %scevgep6611 = getelementptr i8, ptr addrspace(4) %lsr.iv6601, i64 %6550
  %scevgep6604 = getelementptr i8, ptr %1513, i64 %1474
  %scevgep6605 = getelementptr i8, ptr %scevgep6604, i64 %1475
  %scevgep6609 = getelementptr i8, ptr %scevgep6605, i64 %6550
  br label %loop-memcpy-residual5261, !amdgpu.uniform !42

Flow8402:                                         ; preds = %loop-memcpy-expansion5265
  br label %Flow8403, !amdgpu.uniform !42

loop-memcpy-residual-header5268:                  ; preds = %Flow8403
  br i1 %1111, label %loop-memcpy-residual-header5268.post-loop-memcpy-expansion5264_crit_edge, label %Flow8400, !amdgpu.uniform !42

loop-memcpy-residual-header5268.post-loop-memcpy-expansion5264_crit_edge: ; preds = %loop-memcpy-residual-header5268
  %.pre7307 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i906, 6
  br label %Flow8400, !amdgpu.uniform !42

Flow8400:                                         ; preds = %loop-memcpy-residual-header5268.post-loop-memcpy-expansion5264_crit_edge, %loop-memcpy-residual-header5268
  %6551 = phi i64 [ %.pre7307, %loop-memcpy-residual-header5268.post-loop-memcpy-expansion5264_crit_edge ], [ poison, %loop-memcpy-residual-header5268 ]
  %6552 = phi i1 [ false, %loop-memcpy-residual-header5268.post-loop-memcpy-expansion5264_crit_edge ], [ true, %loop-memcpy-residual-header5268 ]
  br i1 %6552, label %loop-memcpy-residual5267.preheader, label %Flow8401, !amdgpu.uniform !42

loop-memcpy-residual5267.preheader:               ; preds = %Flow8400
  %scevgep6617 = getelementptr i8, ptr %1595, i64 8
  %scevgep6618 = getelementptr i8, ptr %scevgep6617, i64 %1474
  %6553 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i906, 6
  %scevgep6619 = getelementptr i8, ptr %scevgep6618, i64 %6553
  br label %loop-memcpy-residual5267, !amdgpu.uniform !42

Flow8393:                                         ; preds = %loop-memcpy-expansion5271
  br label %Flow8394, !amdgpu.uniform !42

Flow8395:                                         ; preds = %Flow8392, %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i927
  br label %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i928, !amdgpu.uniform !42

loop-memcpy-residual-header5274:                  ; preds = %Flow8394
  %6554 = icmp ne i64 %1632, 0
  br i1 %6554, label %loop-memcpy-residual5273.preheader, label %Flow8392, !amdgpu.uniform !42

loop-memcpy-residual5273.preheader:               ; preds = %loop-memcpy-residual-header5274
  %umin6629 = call i64 @llvm.umin.i64(i64 %lsr.iv6627, i64 64)
  %6555 = lshr i64 %umin6629, 4
  %6556 = shl nuw nsw i64 %6555, 4
  %scevgep6632 = getelementptr i8, ptr addrspace(4) %lsr.iv6622, i64 %6556
  %scevgep6625 = getelementptr i8, ptr %1631, i64 %1474
  %scevgep6626 = getelementptr i8, ptr %scevgep6625, i64 %1601
  %scevgep6630 = getelementptr i8, ptr %scevgep6626, i64 %6556
  br label %loop-memcpy-residual5273, !amdgpu.uniform !42

Flow8374:                                         ; preds = %loop-memcpy-expansion5277
  %.lcssa9014 = phi i64 [ %1775, %loop-memcpy-expansion5277 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa9014)
  br label %Flow8375, !amdgpu.uniform !42

Flow8376:                                         ; preds = %Flow8373, %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i1034
  call void @llvm.amdgcn.end.cf.i64(i64 %1763)
  br label %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1035, !amdgpu.uniform !42

loop-memcpy-residual-header5280:                  ; preds = %Flow8375
  %6557 = icmp ne i64 %1765, 0
  %6558 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %6557)
  %6559 = extractvalue { i1, i64 } %6558, 0
  %6560 = extractvalue { i1, i64 } %6558, 1
  br i1 %6559, label %loop-memcpy-residual5279.preheader, label %Flow8373

loop-memcpy-residual5279.preheader:               ; preds = %loop-memcpy-residual-header5280
  %umin6656 = call i64 @llvm.umin.i64(i64 %lsr.iv6654, i64 64)
  %6561 = lshr i64 %umin6656, 4
  %6562 = shl nuw nsw i64 %6561, 4
  %scevgep6659 = getelementptr i8, ptr addrspace(4) %lsr.iv6649, i64 %6562
  %scevgep6652 = getelementptr i8, ptr %1764, i64 %1474
  %scevgep6653 = getelementptr i8, ptr %scevgep6652, i64 %1726
  %scevgep6657 = getelementptr i8, ptr %scevgep6653, i64 %6562
  br label %loop-memcpy-residual5279, !amdgpu.uniform !42

Flow8324:                                         ; preds = %loop-memcpy-expansion5283
  %.lcssa9201 = phi i64 [ %2205, %loop-memcpy-expansion5283 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa9201)
  br label %Flow8325, !amdgpu.uniform !42

loop-memcpy-residual-header5286:                  ; preds = %Flow8325
  %6563 = icmp eq i64 %2194, 0
  %6564 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %6563)
  %6565 = extractvalue { i1, i64 } %6564, 0
  %6566 = extractvalue { i1, i64 } %6564, 1
  br i1 %6565, label %loop-memcpy-residual-header5286.post-loop-memcpy-expansion5282_crit_edge, label %Flow8322

loop-memcpy-residual-header5286.post-loop-memcpy-expansion5282_crit_edge: ; preds = %loop-memcpy-residual-header5286
  %.pre7313 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i1335, 6
  br label %Flow8322, !amdgpu.uniform !42

Flow8322:                                         ; preds = %loop-memcpy-residual-header5286.post-loop-memcpy-expansion5282_crit_edge, %loop-memcpy-residual-header5286
  %6567 = phi i64 [ %.pre7313, %loop-memcpy-residual-header5286.post-loop-memcpy-expansion5282_crit_edge ], [ poison, %loop-memcpy-residual-header5286 ]
  %6568 = call { i1, i64 } @llvm.amdgcn.else.i64.i64(i64 %6566)
  %6569 = extractvalue { i1, i64 } %6568, 0
  %6570 = extractvalue { i1, i64 } %6568, 1
  br i1 %6569, label %loop-memcpy-residual5285.preheader, label %Flow8323

loop-memcpy-residual5285.preheader:               ; preds = %Flow8322
  %scevgep7010 = getelementptr i8, ptr %2193, i64 8
  %6571 = add i64 %1474, %2195
  %scevgep7011 = getelementptr i8, ptr %scevgep7010, i64 %6571
  %6572 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i1335, 6
  %scevgep7012 = getelementptr i8, ptr %scevgep7011, i64 %6572
  %scevgep7014 = getelementptr i8, ptr addrspace(4) %cond3, i64 %2195
  br label %loop-memcpy-residual5285, !amdgpu.uniform !42

Flow8315:                                         ; preds = %loop-memcpy-expansion5289
  %.lcssa9204 = phi i64 [ %2256, %loop-memcpy-expansion5289 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa9204)
  br label %Flow8316, !amdgpu.uniform !42

Flow8317:                                         ; preds = %Flow8314, %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i1356
  call void @llvm.amdgcn.end.cf.i64(i64 %2244)
  br label %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1357, !amdgpu.uniform !42

loop-memcpy-residual-header5292:                  ; preds = %Flow8316
  %6573 = icmp ne i64 %2246, 0
  %6574 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %6573)
  %6575 = extractvalue { i1, i64 } %6574, 0
  %6576 = extractvalue { i1, i64 } %6574, 1
  br i1 %6575, label %loop-memcpy-residual5291.preheader, label %Flow8314

loop-memcpy-residual5291.preheader:               ; preds = %loop-memcpy-residual-header5292
  %umin7024 = call i64 @llvm.umin.i64(i64 %lsr.iv7022, i64 64)
  %6577 = lshr i64 %umin7024, 4
  %6578 = shl nuw nsw i64 %6577, 4
  %scevgep7027 = getelementptr i8, ptr addrspace(4) %lsr.iv7017, i64 %6578
  %scevgep7020 = getelementptr i8, ptr %2245, i64 %1474
  %scevgep7021 = getelementptr i8, ptr %scevgep7020, i64 %2207
  %scevgep7025 = getelementptr i8, ptr %scevgep7021, i64 %6578
  br label %loop-memcpy-residual5291, !amdgpu.uniform !42

Flow8305:                                         ; preds = %loop-memcpy-expansion5295
  br label %Flow8306, !amdgpu.uniform !42

loop-memcpy-residual-header5298:                  ; preds = %Flow8306
  br i1 %1139, label %loop-memcpy-residual-header5298.post-loop-memcpy-expansion5294_crit_edge, label %Flow8303, !amdgpu.uniform !42

loop-memcpy-residual-header5298.post-loop-memcpy-expansion5294_crit_edge: ; preds = %loop-memcpy-residual-header5298
  %.pre7315 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i1441, 6
  br label %Flow8303, !amdgpu.uniform !42

Flow8303:                                         ; preds = %loop-memcpy-residual-header5298.post-loop-memcpy-expansion5294_crit_edge, %loop-memcpy-residual-header5298
  %6579 = phi i64 [ %.pre7315, %loop-memcpy-residual-header5298.post-loop-memcpy-expansion5294_crit_edge ], [ poison, %loop-memcpy-residual-header5298 ]
  %6580 = phi i1 [ false, %loop-memcpy-residual-header5298.post-loop-memcpy-expansion5294_crit_edge ], [ true, %loop-memcpy-residual-header5298 ]
  br i1 %6580, label %loop-memcpy-residual5297.preheader, label %Flow8304, !amdgpu.uniform !42

loop-memcpy-residual5297.preheader:               ; preds = %Flow8303
  %scevgep7033 = getelementptr i8, ptr %2327, i64 8
  %scevgep7034 = getelementptr i8, ptr %scevgep7033, i64 %1474
  %6581 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i1441, 6
  %scevgep7035 = getelementptr i8, ptr %scevgep7034, i64 %6581
  br label %loop-memcpy-residual5297, !amdgpu.uniform !42

Flow8296:                                         ; preds = %loop-memcpy-expansion5301
  br label %Flow8297, !amdgpu.uniform !42

Flow8298:                                         ; preds = %Flow8295, %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i1462
  br label %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1463, !amdgpu.uniform !42

loop-memcpy-residual-header5304:                  ; preds = %Flow8297
  %6582 = icmp ne i64 %2364, 0
  br i1 %6582, label %loop-memcpy-residual5303.preheader, label %Flow8295, !amdgpu.uniform !42

loop-memcpy-residual5303.preheader:               ; preds = %loop-memcpy-residual-header5304
  %umin7045 = call i64 @llvm.umin.i64(i64 %lsr.iv7043, i64 64)
  %6583 = lshr i64 %umin7045, 4
  %6584 = shl nuw nsw i64 %6583, 4
  %scevgep7048 = getelementptr i8, ptr addrspace(4) %lsr.iv7038, i64 %6584
  %scevgep7041 = getelementptr i8, ptr %2363, i64 %1474
  %scevgep7042 = getelementptr i8, ptr %scevgep7041, i64 %2333
  %scevgep7046 = getelementptr i8, ptr %scevgep7042, i64 %6584
  br label %loop-memcpy-residual5303, !amdgpu.uniform !42

Flow8277:                                         ; preds = %loop-memcpy-expansion5307
  %.lcssa9231 = phi i64 [ %2507, %loop-memcpy-expansion5307 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa9231)
  br label %Flow8278, !amdgpu.uniform !42

Flow8279:                                         ; preds = %Flow8276, %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i1569
  call void @llvm.amdgcn.end.cf.i64(i64 %2495)
  br label %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1570, !amdgpu.uniform !42

loop-memcpy-residual-header5310:                  ; preds = %Flow8278
  %6585 = icmp ne i64 %2497, 0
  %6586 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %6585)
  %6587 = extractvalue { i1, i64 } %6586, 0
  %6588 = extractvalue { i1, i64 } %6586, 1
  br i1 %6587, label %loop-memcpy-residual5309.preheader, label %Flow8276

loop-memcpy-residual5309.preheader:               ; preds = %loop-memcpy-residual-header5310
  %umin7070 = call i64 @llvm.umin.i64(i64 %lsr.iv7068, i64 64)
  %6589 = lshr i64 %umin7070, 4
  %6590 = shl nuw nsw i64 %6589, 4
  %scevgep7073 = getelementptr i8, ptr addrspace(4) %lsr.iv7063, i64 %6590
  %scevgep7066 = getelementptr i8, ptr %2496, i64 %1474
  %scevgep7067 = getelementptr i8, ptr %scevgep7066, i64 %2458
  %scevgep7071 = getelementptr i8, ptr %scevgep7067, i64 %6590
  br label %loop-memcpy-residual5309, !amdgpu.uniform !42

Flow8230:                                         ; preds = %loop-memcpy-expansion5313
  %.lcssa9057 = phi i64 [ %2911, %loop-memcpy-expansion5313 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa9057)
  br label %Flow8231, !amdgpu.uniform !42

Flow8232:                                         ; preds = %Flow8229, %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i1882
  call void @llvm.amdgcn.end.cf.i64(i64 %2899)
  br label %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1883, !amdgpu.uniform !42

loop-memcpy-residual-header5316:                  ; preds = %Flow8231
  %6591 = icmp ne i64 %2901, 0
  %6592 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %6591)
  %6593 = extractvalue { i1, i64 } %6592, 0
  %6594 = extractvalue { i1, i64 } %6592, 1
  br i1 %6593, label %loop-memcpy-residual5315.preheader, label %Flow8229

loop-memcpy-residual5315.preheader:               ; preds = %loop-memcpy-residual-header5316
  %umin6715 = call i64 @llvm.umin.i64(i64 %lsr.iv6713, i64 64)
  %6595 = lshr i64 %umin6715, 4
  %6596 = shl nuw nsw i64 %6595, 4
  %scevgep6718 = getelementptr i8, ptr addrspace(4) %lsr.iv6708, i64 %6596
  %scevgep6711 = getelementptr i8, ptr %2900, i64 %1474
  %scevgep6712 = getelementptr i8, ptr %scevgep6711, i64 %2862
  %scevgep6716 = getelementptr i8, ptr %scevgep6712, i64 %6596
  br label %loop-memcpy-residual5315, !amdgpu.uniform !42

Flow8220:                                         ; preds = %loop-memcpy-expansion5319
  br label %Flow8221, !amdgpu.uniform !42

loop-memcpy-residual-header5322:                  ; preds = %Flow8221
  br i1 %1115, label %loop-memcpy-residual-header5322.post-loop-memcpy-expansion5318_crit_edge, label %Flow8218, !amdgpu.uniform !42

loop-memcpy-residual-header5322.post-loop-memcpy-expansion5318_crit_edge: ; preds = %loop-memcpy-residual-header5322
  %.pre7323 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i1967, 6
  br label %Flow8218, !amdgpu.uniform !42

Flow8218:                                         ; preds = %loop-memcpy-residual-header5322.post-loop-memcpy-expansion5318_crit_edge, %loop-memcpy-residual-header5322
  %6597 = phi i64 [ %.pre7323, %loop-memcpy-residual-header5322.post-loop-memcpy-expansion5318_crit_edge ], [ poison, %loop-memcpy-residual-header5322 ]
  %6598 = phi i1 [ false, %loop-memcpy-residual-header5322.post-loop-memcpy-expansion5318_crit_edge ], [ true, %loop-memcpy-residual-header5322 ]
  br i1 %6598, label %loop-memcpy-residual5321.preheader, label %Flow8219, !amdgpu.uniform !42

loop-memcpy-residual5321.preheader:               ; preds = %Flow8218
  %scevgep6724 = getelementptr i8, ptr %2982, i64 8
  %scevgep6725 = getelementptr i8, ptr %scevgep6724, i64 %1474
  %6599 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i1967, 6
  %scevgep6726 = getelementptr i8, ptr %scevgep6725, i64 %6599
  br label %loop-memcpy-residual5321, !amdgpu.uniform !42

Flow8211:                                         ; preds = %loop-memcpy-expansion5325
  br label %Flow8212, !amdgpu.uniform !42

Flow8213:                                         ; preds = %Flow8210, %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i1988
  br label %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i1989, !amdgpu.uniform !42

loop-memcpy-residual-header5328:                  ; preds = %Flow8212
  %6600 = icmp ne i64 %3019, 0
  br i1 %6600, label %loop-memcpy-residual5327.preheader, label %Flow8210, !amdgpu.uniform !42

loop-memcpy-residual5327.preheader:               ; preds = %loop-memcpy-residual-header5328
  %umin6736 = call i64 @llvm.umin.i64(i64 %lsr.iv6734, i64 64)
  %6601 = lshr i64 %umin6736, 4
  %6602 = shl nuw nsw i64 %6601, 4
  %scevgep6739 = getelementptr i8, ptr addrspace(4) %lsr.iv6729, i64 %6602
  %scevgep6732 = getelementptr i8, ptr %3018, i64 %1474
  %scevgep6733 = getelementptr i8, ptr %scevgep6732, i64 %2988
  %scevgep6737 = getelementptr i8, ptr %scevgep6733, i64 %6602
  br label %loop-memcpy-residual5327, !amdgpu.uniform !42

Flow8192:                                         ; preds = %loop-memcpy-expansion5331
  %.lcssa9084 = phi i64 [ %3162, %loop-memcpy-expansion5331 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa9084)
  br label %Flow8193, !amdgpu.uniform !42

Flow8194:                                         ; preds = %Flow8191, %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i2095
  call void @llvm.amdgcn.end.cf.i64(i64 %3150)
  br label %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i2096, !amdgpu.uniform !42

loop-memcpy-residual-header5334:                  ; preds = %Flow8193
  %6603 = icmp ne i64 %3152, 0
  %6604 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %6603)
  %6605 = extractvalue { i1, i64 } %6604, 0
  %6606 = extractvalue { i1, i64 } %6604, 1
  br i1 %6605, label %loop-memcpy-residual5333.preheader, label %Flow8191

loop-memcpy-residual5333.preheader:               ; preds = %loop-memcpy-residual-header5334
  %umin6761 = call i64 @llvm.umin.i64(i64 %lsr.iv6759, i64 64)
  %6607 = lshr i64 %umin6761, 4
  %6608 = shl nuw nsw i64 %6607, 4
  %scevgep6764 = getelementptr i8, ptr addrspace(4) %lsr.iv6754, i64 %6608
  %scevgep6757 = getelementptr i8, ptr %3151, i64 %1474
  %scevgep6758 = getelementptr i8, ptr %scevgep6757, i64 %3113
  %scevgep6762 = getelementptr i8, ptr %scevgep6758, i64 %6608
  br label %loop-memcpy-residual5333, !amdgpu.uniform !42

Flow8163:                                         ; preds = %loop-memcpy-expansion5337
  br label %Flow8164, !amdgpu.uniform !42

loop-memcpy-residual-header5340:                  ; preds = %Flow8164
  br i1 %1119, label %loop-memcpy-residual-header5340.post-loop-memcpy-expansion5336_crit_edge, label %Flow8161, !amdgpu.uniform !42

loop-memcpy-residual-header5340.post-loop-memcpy-expansion5336_crit_edge: ; preds = %loop-memcpy-residual-header5340
  %.pre7329 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i2287, 6
  br label %Flow8161, !amdgpu.uniform !42

Flow8161:                                         ; preds = %loop-memcpy-residual-header5340.post-loop-memcpy-expansion5336_crit_edge, %loop-memcpy-residual-header5340
  %6609 = phi i64 [ %.pre7329, %loop-memcpy-residual-header5340.post-loop-memcpy-expansion5336_crit_edge ], [ poison, %loop-memcpy-residual-header5340 ]
  %6610 = phi i1 [ false, %loop-memcpy-residual-header5340.post-loop-memcpy-expansion5336_crit_edge ], [ true, %loop-memcpy-residual-header5340 ]
  br i1 %6610, label %loop-memcpy-residual5339.preheader, label %Flow8162, !amdgpu.uniform !42

loop-memcpy-residual5339.preheader:               ; preds = %Flow8161
  %scevgep6802 = getelementptr i8, ptr %3368, i64 8
  %scevgep6803 = getelementptr i8, ptr %scevgep6802, i64 %1474
  %6611 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i2287, 6
  %scevgep6804 = getelementptr i8, ptr %scevgep6803, i64 %6611
  br label %loop-memcpy-residual5339, !amdgpu.uniform !42

Flow8154:                                         ; preds = %loop-memcpy-expansion5343
  br label %Flow8155, !amdgpu.uniform !42

Flow8156:                                         ; preds = %Flow8153, %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i2308
  br label %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i2309, !amdgpu.uniform !42

loop-memcpy-residual-header5346:                  ; preds = %Flow8155
  %6612 = icmp ne i64 %3405, 0
  br i1 %6612, label %loop-memcpy-residual5345.preheader, label %Flow8153, !amdgpu.uniform !42

loop-memcpy-residual5345.preheader:               ; preds = %loop-memcpy-residual-header5346
  %umin6814 = call i64 @llvm.umin.i64(i64 %lsr.iv6812, i64 64)
  %6613 = lshr i64 %umin6814, 4
  %6614 = shl nuw nsw i64 %6613, 4
  %scevgep6817 = getelementptr i8, ptr addrspace(4) %lsr.iv6807, i64 %6614
  %scevgep6810 = getelementptr i8, ptr %3404, i64 %1474
  %scevgep6811 = getelementptr i8, ptr %scevgep6810, i64 %3374
  %scevgep6815 = getelementptr i8, ptr %scevgep6811, i64 %6614
  br label %loop-memcpy-residual5345, !amdgpu.uniform !42

Flow8142:                                         ; preds = %loop-memcpy-expansion5349
  br label %Flow8143, !amdgpu.uniform !42

loop-memcpy-residual-header5352:                  ; preds = %Flow8143
  br i1 %1135, label %loop-memcpy-residual-header5352.post-loop-memcpy-expansion5348_crit_edge, label %Flow8140, !amdgpu.uniform !42

loop-memcpy-residual-header5352.post-loop-memcpy-expansion5348_crit_edge: ; preds = %loop-memcpy-residual-header5352
  %.pre7331 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i2393, 6
  br label %Flow8140, !amdgpu.uniform !42

Flow8140:                                         ; preds = %loop-memcpy-residual-header5352.post-loop-memcpy-expansion5348_crit_edge, %loop-memcpy-residual-header5352
  %6615 = phi i64 [ %.pre7331, %loop-memcpy-residual-header5352.post-loop-memcpy-expansion5348_crit_edge ], [ poison, %loop-memcpy-residual-header5352 ]
  %6616 = phi i1 [ false, %loop-memcpy-residual-header5352.post-loop-memcpy-expansion5348_crit_edge ], [ true, %loop-memcpy-residual-header5352 ]
  br i1 %6616, label %loop-memcpy-residual5351.preheader, label %Flow8141, !amdgpu.uniform !42

loop-memcpy-residual5351.preheader:               ; preds = %Flow8140
  %scevgep6985 = getelementptr i8, ptr %3519, i64 24
  %scevgep6986 = getelementptr i8, ptr %scevgep6985, i64 %1474
  %6617 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i2393, 6
  %scevgep6987 = getelementptr i8, ptr %scevgep6986, i64 %6617
  br label %loop-memcpy-residual5351, !amdgpu.uniform !42

Flow8133:                                         ; preds = %loop-memcpy-expansion5355
  br label %Flow8134, !amdgpu.uniform !42

Flow8135:                                         ; preds = %Flow8132, %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i2414
  br label %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i2415, !amdgpu.uniform !42

loop-memcpy-residual-header5358:                  ; preds = %Flow8134
  %6618 = icmp ne i64 %3556, 0
  br i1 %6618, label %loop-memcpy-residual5357.preheader, label %Flow8132, !amdgpu.uniform !42

loop-memcpy-residual5357.preheader:               ; preds = %loop-memcpy-residual-header5358
  %umin6997 = call i64 @llvm.umin.i64(i64 %lsr.iv6995, i64 64)
  %6619 = lshr i64 %umin6997, 4
  %6620 = shl nuw nsw i64 %6619, 4
  %scevgep7000 = getelementptr i8, ptr addrspace(4) %lsr.iv6990, i64 %6620
  %scevgep6993 = getelementptr i8, ptr %3555, i64 %1474
  %scevgep6994 = getelementptr i8, ptr %scevgep6993, i64 %3525
  %scevgep6998 = getelementptr i8, ptr %scevgep6994, i64 %6620
  br label %loop-memcpy-residual5357, !amdgpu.uniform !42

Flow8077:                                         ; preds = %loop-memcpy-expansion5361
  br label %Flow8078, !amdgpu.uniform !42

loop-memcpy-residual-header5364:                  ; preds = %Flow8078
  br i1 %1131, label %loop-memcpy-residual-header5364.post-loop-memcpy-expansion5360_crit_edge, label %Flow8075, !amdgpu.uniform !42

loop-memcpy-residual-header5364.post-loop-memcpy-expansion5360_crit_edge: ; preds = %loop-memcpy-residual-header5364
  %.pre7335 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i2658, 6
  br label %Flow8075, !amdgpu.uniform !42

Flow8075:                                         ; preds = %loop-memcpy-residual-header5364.post-loop-memcpy-expansion5360_crit_edge, %loop-memcpy-residual-header5364
  %6621 = phi i64 [ %.pre7335, %loop-memcpy-residual-header5364.post-loop-memcpy-expansion5360_crit_edge ], [ poison, %loop-memcpy-residual-header5364 ]
  %6622 = phi i1 [ false, %loop-memcpy-residual-header5364.post-loop-memcpy-expansion5360_crit_edge ], [ true, %loop-memcpy-residual-header5364 ]
  br i1 %6622, label %loop-memcpy-residual5363.preheader, label %Flow8076, !amdgpu.uniform !42

loop-memcpy-residual5363.preheader:               ; preds = %Flow8075
  %scevgep6964 = getelementptr i8, ptr %3828, i64 8
  %scevgep6965 = getelementptr i8, ptr %scevgep6964, i64 %1474
  %6623 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i2658, 6
  %scevgep6966 = getelementptr i8, ptr %scevgep6965, i64 %6623
  br label %loop-memcpy-residual5363, !amdgpu.uniform !42

Flow8068:                                         ; preds = %loop-memcpy-expansion5367
  br label %Flow8069, !amdgpu.uniform !42

Flow8070:                                         ; preds = %Flow8067, %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i2679
  br label %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i2680, !amdgpu.uniform !42

loop-memcpy-residual-header5370:                  ; preds = %Flow8069
  %6624 = icmp ne i64 %3865, 0
  br i1 %6624, label %loop-memcpy-residual5369.preheader, label %Flow8067, !amdgpu.uniform !42

loop-memcpy-residual5369.preheader:               ; preds = %loop-memcpy-residual-header5370
  %umin6976 = call i64 @llvm.umin.i64(i64 %lsr.iv6974, i64 64)
  %6625 = lshr i64 %umin6976, 4
  %6626 = shl nuw nsw i64 %6625, 4
  %scevgep6979 = getelementptr i8, ptr addrspace(4) %lsr.iv6969, i64 %6626
  %scevgep6972 = getelementptr i8, ptr %3864, i64 %1474
  %scevgep6973 = getelementptr i8, ptr %scevgep6972, i64 %3834
  %scevgep6977 = getelementptr i8, ptr %scevgep6973, i64 %6626
  br label %loop-memcpy-residual5369, !amdgpu.uniform !42

Flow7951:                                         ; preds = %loop-memcpy-expansion5373
  br label %Flow7952, !amdgpu.uniform !42

loop-memcpy-residual-header5376:                  ; preds = %Flow7952
  br i1 %1123, label %loop-memcpy-residual-header5376.post-loop-memcpy-expansion5372_crit_edge, label %Flow7949, !amdgpu.uniform !42

loop-memcpy-residual-header5376.post-loop-memcpy-expansion5372_crit_edge: ; preds = %loop-memcpy-residual-header5376
  %.pre7343 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i2933, 6
  br label %Flow7949, !amdgpu.uniform !42

Flow7949:                                         ; preds = %loop-memcpy-residual-header5376.post-loop-memcpy-expansion5372_crit_edge, %loop-memcpy-residual-header5376
  %6627 = phi i64 [ %.pre7343, %loop-memcpy-residual-header5376.post-loop-memcpy-expansion5372_crit_edge ], [ poison, %loop-memcpy-residual-header5376 ]
  %6628 = phi i1 [ false, %loop-memcpy-residual-header5376.post-loop-memcpy-expansion5372_crit_edge ], [ true, %loop-memcpy-residual-header5376 ]
  br i1 %6628, label %loop-memcpy-residual5375.preheader, label %Flow7950, !amdgpu.uniform !42

loop-memcpy-residual5375.preheader:               ; preds = %Flow7949
  %scevgep6856 = getelementptr i8, ptr %4105, i64 8
  %scevgep6857 = getelementptr i8, ptr %scevgep6856, i64 %1474
  %6629 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i2933, 6
  %scevgep6858 = getelementptr i8, ptr %scevgep6857, i64 %6629
  br label %loop-memcpy-residual5375, !amdgpu.uniform !42

Flow7942:                                         ; preds = %loop-memcpy-expansion5379
  br label %Flow7943, !amdgpu.uniform !42

Flow7944:                                         ; preds = %Flow7941, %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i2954
  br label %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i2955, !amdgpu.uniform !42

loop-memcpy-residual-header5382:                  ; preds = %Flow7943
  %6630 = icmp ne i64 %4142, 0
  br i1 %6630, label %loop-memcpy-residual5381.preheader, label %Flow7941, !amdgpu.uniform !42

loop-memcpy-residual5381.preheader:               ; preds = %loop-memcpy-residual-header5382
  %umin6868 = call i64 @llvm.umin.i64(i64 %lsr.iv6866, i64 64)
  %6631 = lshr i64 %umin6868, 4
  %6632 = shl nuw nsw i64 %6631, 4
  %scevgep6871 = getelementptr i8, ptr addrspace(4) %lsr.iv6861, i64 %6632
  %scevgep6864 = getelementptr i8, ptr %4141, i64 %1474
  %scevgep6865 = getelementptr i8, ptr %scevgep6864, i64 %4111
  %scevgep6869 = getelementptr i8, ptr %scevgep6865, i64 %6632
  br label %loop-memcpy-residual5381, !amdgpu.uniform !42

Flow8012:                                         ; preds = %loop-memcpy-expansion5385
  br label %Flow8013, !amdgpu.uniform !42

loop-memcpy-residual-header5388:                  ; preds = %Flow8013
  br i1 %1127, label %loop-memcpy-residual-header5388.post-loop-memcpy-expansion5384_crit_edge, label %Flow8010, !amdgpu.uniform !42

loop-memcpy-residual-header5388.post-loop-memcpy-expansion5384_crit_edge: ; preds = %loop-memcpy-residual-header5388
  %.pre7339 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i3208, 6
  br label %Flow8010, !amdgpu.uniform !42

Flow8010:                                         ; preds = %loop-memcpy-residual-header5388.post-loop-memcpy-expansion5384_crit_edge, %loop-memcpy-residual-header5388
  %6633 = phi i64 [ %.pre7339, %loop-memcpy-residual-header5388.post-loop-memcpy-expansion5384_crit_edge ], [ poison, %loop-memcpy-residual-header5388 ]
  %6634 = phi i1 [ false, %loop-memcpy-residual-header5388.post-loop-memcpy-expansion5384_crit_edge ], [ true, %loop-memcpy-residual-header5388 ]
  br i1 %6634, label %loop-memcpy-residual5387.preheader, label %Flow8011, !amdgpu.uniform !42

loop-memcpy-residual5387.preheader:               ; preds = %Flow8010
  %scevgep6910 = getelementptr i8, ptr %4389, i64 8
  %scevgep6911 = getelementptr i8, ptr %scevgep6910, i64 %1474
  %6635 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i3208, 6
  %scevgep6912 = getelementptr i8, ptr %scevgep6911, i64 %6635
  br label %loop-memcpy-residual5387, !amdgpu.uniform !42

Flow8003:                                         ; preds = %loop-memcpy-expansion5391
  br label %Flow8004, !amdgpu.uniform !42

Flow8005:                                         ; preds = %Flow8002, %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i3229
  br label %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i3230, !amdgpu.uniform !42

loop-memcpy-residual-header5394:                  ; preds = %Flow8004
  %6636 = icmp ne i64 %4426, 0
  br i1 %6636, label %loop-memcpy-residual5393.preheader, label %Flow8002, !amdgpu.uniform !42

loop-memcpy-residual5393.preheader:               ; preds = %loop-memcpy-residual-header5394
  %umin6922 = call i64 @llvm.umin.i64(i64 %lsr.iv6920, i64 64)
  %6637 = lshr i64 %umin6922, 4
  %6638 = shl nuw nsw i64 %6637, 4
  %scevgep6925 = getelementptr i8, ptr addrspace(4) %lsr.iv6915, i64 %6638
  %scevgep6918 = getelementptr i8, ptr %4425, i64 %1474
  %scevgep6919 = getelementptr i8, ptr %scevgep6918, i64 %4395
  %scevgep6923 = getelementptr i8, ptr %scevgep6919, i64 %6638
  br label %loop-memcpy-residual5393, !amdgpu.uniform !42

Flow7910:                                         ; preds = %loop-memcpy-expansion5397
  br label %Flow7911, !amdgpu.uniform !42

loop-memcpy-residual-header5400:                  ; preds = %Flow7911
  %6639 = icmp eq i64 %4503, 0
  br i1 %6639, label %loop-memcpy-residual-header5400.post-loop-memcpy-expansion5396_crit_edge, label %Flow7908, !amdgpu.uniform !42

loop-memcpy-residual-header5400.post-loop-memcpy-expansion5396_crit_edge: ; preds = %loop-memcpy-residual-header5400
  %.pre = shl nuw nsw i64 %sh_prom.i.i.i.i3270, 6
  %.pre7281 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i3314, 6
  br label %Flow7908, !amdgpu.uniform !42

Flow7908:                                         ; preds = %loop-memcpy-residual-header5400.post-loop-memcpy-expansion5396_crit_edge, %loop-memcpy-residual-header5400
  %6640 = phi i64 [ %.pre, %loop-memcpy-residual-header5400.post-loop-memcpy-expansion5396_crit_edge ], [ poison, %loop-memcpy-residual-header5400 ]
  %6641 = phi i64 [ %.pre7281, %loop-memcpy-residual-header5400.post-loop-memcpy-expansion5396_crit_edge ], [ poison, %loop-memcpy-residual-header5400 ]
  %6642 = phi i1 [ false, %loop-memcpy-residual-header5400.post-loop-memcpy-expansion5396_crit_edge ], [ true, %loop-memcpy-residual-header5400 ]
  br i1 %6642, label %loop-memcpy-residual5399.preheader, label %Flow7909, !amdgpu.uniform !42

loop-memcpy-residual5399.preheader:               ; preds = %Flow7908
  %6643 = shl nuw nsw i64 %sh_prom.i.i.i.i3270, 6
  %6644 = add nuw nsw i64 %6643, 8
  %scevgep6492 = getelementptr i8, ptr %4502, i64 %6644
  %6645 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i3314, 6
  %scevgep6493 = getelementptr i8, ptr %scevgep6492, i64 %6645
  br label %loop-memcpy-residual5399, !amdgpu.uniform !42

Flow7901:                                         ; preds = %loop-memcpy-expansion5403
  br label %Flow7902, !amdgpu.uniform !42

Flow7903:                                         ; preds = %Flow7900, %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i3335
  br label %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i3336, !amdgpu.uniform !42

loop-memcpy-residual-header5406:                  ; preds = %Flow7902
  %6646 = icmp ne i64 %4540, 0
  br i1 %6646, label %loop-memcpy-residual5405.preheader, label %Flow7900, !amdgpu.uniform !42

loop-memcpy-residual5405.preheader:               ; preds = %loop-memcpy-residual-header5406
  %umin6487 = call i64 @llvm.umin.i64(i64 %lsr.iv6485, i64 64)
  %6647 = lshr i64 %umin6487, 4
  %6648 = shl nuw nsw i64 %6647, 4
  %scevgep6490 = getelementptr i8, ptr addrspace(4) %lsr.iv6480, i64 %6648
  %scevgep6483 = getelementptr i8, ptr %4539, i64 %4513
  %scevgep6484 = getelementptr i8, ptr %scevgep6483, i64 %4514
  %scevgep6488 = getelementptr i8, ptr %scevgep6484, i64 %6648
  br label %loop-memcpy-residual5405, !amdgpu.uniform !42

Flow7849:                                         ; preds = %loop-memcpy-expansion5409
  br label %Flow7850, !amdgpu.uniform !42

loop-memcpy-residual-header5412:                  ; preds = %Flow7850
  %6649 = icmp eq i64 %4856, 0
  br i1 %6649, label %loop-memcpy-residual-header5412.post-loop-memcpy-expansion5408_crit_edge, label %Flow7847, !amdgpu.uniform !42

loop-memcpy-residual-header5412.post-loop-memcpy-expansion5408_crit_edge: ; preds = %loop-memcpy-residual-header5412
  %.pre7285 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i3599, 6
  br label %Flow7847, !amdgpu.uniform !42

Flow7847:                                         ; preds = %loop-memcpy-residual-header5412.post-loop-memcpy-expansion5408_crit_edge, %loop-memcpy-residual-header5412
  %6650 = phi i64 [ %.pre7285, %loop-memcpy-residual-header5412.post-loop-memcpy-expansion5408_crit_edge ], [ poison, %loop-memcpy-residual-header5412 ]
  %6651 = phi i1 [ false, %loop-memcpy-residual-header5412.post-loop-memcpy-expansion5408_crit_edge ], [ true, %loop-memcpy-residual-header5412 ]
  br i1 %6651, label %loop-memcpy-residual5411.preheader, label %Flow7848, !amdgpu.uniform !42

loop-memcpy-residual5411.preheader:               ; preds = %Flow7847
  %6652 = add nuw nsw i64 %4513, 8
  %scevgep6444 = getelementptr i8, ptr %4855, i64 %6652
  %6653 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i3599, 6
  %scevgep6445 = getelementptr i8, ptr %scevgep6444, i64 %6653
  br label %loop-memcpy-residual5411, !amdgpu.uniform !42

Flow7840:                                         ; preds = %loop-memcpy-expansion5415
  br label %Flow7841, !amdgpu.uniform !42

Flow7842:                                         ; preds = %Flow7839, %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i3620
  br label %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i3621, !amdgpu.uniform !42

loop-memcpy-residual-header5418:                  ; preds = %Flow7841
  %6654 = icmp ne i64 %4891, 0
  br i1 %6654, label %loop-memcpy-residual5417.preheader, label %Flow7839, !amdgpu.uniform !42

loop-memcpy-residual5417.preheader:               ; preds = %loop-memcpy-residual-header5418
  %umin6439 = call i64 @llvm.umin.i64(i64 %lsr.iv6437, i64 64)
  %scevgep6442 = getelementptr i8, ptr addrspace(4) %lsr.iv6432, i64 %umin6439
  %scevgep6435 = getelementptr i8, ptr %4890, i64 %4513
  %scevgep6436 = getelementptr i8, ptr %scevgep6435, i64 %4865
  %scevgep6440 = getelementptr i8, ptr %scevgep6436, i64 %umin6439
  br label %loop-memcpy-residual5417, !amdgpu.uniform !42

Flow7830:                                         ; preds = %loop-memcpy-expansion5421
  br label %Flow7831, !amdgpu.uniform !42

loop-memcpy-residual-header5424:                  ; preds = %Flow7831
  %6655 = icmp eq i64 %4966, 0
  br i1 %6655, label %loop-memcpy-residual-header5424.post-loop-memcpy-expansion5420_crit_edge, label %Flow7828, !amdgpu.uniform !42

loop-memcpy-residual-header5424.post-loop-memcpy-expansion5420_crit_edge: ; preds = %loop-memcpy-residual-header5424
  %.pre7287 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i3705, 6
  br label %Flow7828, !amdgpu.uniform !42

Flow7828:                                         ; preds = %loop-memcpy-residual-header5424.post-loop-memcpy-expansion5420_crit_edge, %loop-memcpy-residual-header5424
  %6656 = phi i64 [ %.pre7287, %loop-memcpy-residual-header5424.post-loop-memcpy-expansion5420_crit_edge ], [ poison, %loop-memcpy-residual-header5424 ]
  %6657 = phi i1 [ false, %loop-memcpy-residual-header5424.post-loop-memcpy-expansion5420_crit_edge ], [ true, %loop-memcpy-residual-header5424 ]
  br i1 %6657, label %loop-memcpy-residual5423.preheader, label %Flow7829, !amdgpu.uniform !42

loop-memcpy-residual5423.preheader:               ; preds = %Flow7828
  %6658 = add nuw nsw i64 %4513, 8
  %scevgep6425 = getelementptr i8, ptr %4965, i64 %6658
  %6659 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i3705, 6
  %scevgep6426 = getelementptr i8, ptr %scevgep6425, i64 %6659
  br label %loop-memcpy-residual5423, !amdgpu.uniform !42

Flow7821:                                         ; preds = %loop-memcpy-expansion5427
  br label %Flow7822, !amdgpu.uniform !42

Flow7823:                                         ; preds = %Flow7820, %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i3726
  br label %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i3727, !amdgpu.uniform !42

loop-memcpy-residual-header5430:                  ; preds = %Flow7822
  %6660 = icmp ne i64 %5001, 0
  br i1 %6660, label %loop-memcpy-residual5429.preheader, label %Flow7820, !amdgpu.uniform !42

loop-memcpy-residual5429.preheader:               ; preds = %loop-memcpy-residual-header5430
  %umin6420 = call i64 @llvm.umin.i64(i64 %lsr.iv6418, i64 64)
  %6661 = lshr i64 %umin6420, 4
  %6662 = shl nuw nsw i64 %6661, 4
  %scevgep6423 = getelementptr i8, ptr addrspace(4) %lsr.iv6413, i64 %6662
  %scevgep6416 = getelementptr i8, ptr %5000, i64 %4513
  %scevgep6417 = getelementptr i8, ptr %scevgep6416, i64 %4975
  %scevgep6421 = getelementptr i8, ptr %scevgep6417, i64 %6662
  br label %loop-memcpy-residual5429, !amdgpu.uniform !42

Flow7757:                                         ; preds = %loop-memcpy-expansion5433
  br label %Flow7758, !amdgpu.uniform !42

loop-memcpy-residual-header5436:                  ; preds = %Flow7758
  %6663 = icmp eq i64 %5417, 0
  br i1 %6663, label %loop-memcpy-residual-header5436.post-loop-memcpy-expansion5432_crit_edge, label %Flow7755, !amdgpu.uniform !42

loop-memcpy-residual-header5436.post-loop-memcpy-expansion5432_crit_edge: ; preds = %loop-memcpy-residual-header5436
  %.pre7291 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i4089, 6
  br label %Flow7755, !amdgpu.uniform !42

Flow7755:                                         ; preds = %loop-memcpy-residual-header5436.post-loop-memcpy-expansion5432_crit_edge, %loop-memcpy-residual-header5436
  %6664 = phi i64 [ %.pre7291, %loop-memcpy-residual-header5436.post-loop-memcpy-expansion5432_crit_edge ], [ poison, %loop-memcpy-residual-header5436 ]
  %6665 = phi i1 [ false, %loop-memcpy-residual-header5436.post-loop-memcpy-expansion5432_crit_edge ], [ true, %loop-memcpy-residual-header5436 ]
  br i1 %6665, label %loop-memcpy-residual5435.preheader, label %Flow7756, !amdgpu.uniform !42

loop-memcpy-residual5435.preheader:               ; preds = %Flow7755
  %6666 = add nuw nsw i64 %4513, 8
  %scevgep6377 = getelementptr i8, ptr %5416, i64 %6666
  %6667 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i4089, 6
  %scevgep6378 = getelementptr i8, ptr %scevgep6377, i64 %6667
  br label %loop-memcpy-residual5435, !amdgpu.uniform !42

Flow7748:                                         ; preds = %loop-memcpy-expansion5439
  br label %Flow7749, !amdgpu.uniform !42

Flow7750:                                         ; preds = %Flow7747, %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i4110
  br label %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i4111, !amdgpu.uniform !42

loop-memcpy-residual-header5442:                  ; preds = %Flow7749
  %6668 = icmp ne i64 %5452, 0
  br i1 %6668, label %loop-memcpy-residual5441.preheader, label %Flow7747, !amdgpu.uniform !42

loop-memcpy-residual5441.preheader:               ; preds = %loop-memcpy-residual-header5442
  %umin6372 = call i64 @llvm.umin.i64(i64 %lsr.iv6370, i64 64)
  %6669 = lshr i64 %umin6372, 4
  %6670 = shl nuw nsw i64 %6669, 4
  %scevgep6375 = getelementptr i8, ptr addrspace(4) %lsr.iv6365, i64 %6670
  %scevgep6368 = getelementptr i8, ptr %5451, i64 %4513
  %scevgep6369 = getelementptr i8, ptr %scevgep6368, i64 %5426
  %scevgep6373 = getelementptr i8, ptr %scevgep6369, i64 %6670
  br label %loop-memcpy-residual5441, !amdgpu.uniform !42

Flow7684:                                         ; preds = %loop-memcpy-expansion5445
  br label %Flow7685, !amdgpu.uniform !42

loop-memcpy-residual-header5448:                  ; preds = %Flow7685
  %6671 = icmp eq i64 %5871, 0
  br i1 %6671, label %loop-memcpy-residual-header5448.post-loop-memcpy-expansion5444_crit_edge, label %Flow7682, !amdgpu.uniform !42

loop-memcpy-residual-header5448.post-loop-memcpy-expansion5444_crit_edge: ; preds = %loop-memcpy-residual-header5448
  %.pre7295 = shl nuw nsw i64 %sh_prom.i.i.i.i3270, 6
  %.pre7297 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i4473, 6
  br label %Flow7682, !amdgpu.uniform !42

Flow7682:                                         ; preds = %loop-memcpy-residual-header5448.post-loop-memcpy-expansion5444_crit_edge, %loop-memcpy-residual-header5448
  %6672 = phi i64 [ %.pre7295, %loop-memcpy-residual-header5448.post-loop-memcpy-expansion5444_crit_edge ], [ poison, %loop-memcpy-residual-header5448 ]
  %6673 = phi i64 [ %.pre7297, %loop-memcpy-residual-header5448.post-loop-memcpy-expansion5444_crit_edge ], [ poison, %loop-memcpy-residual-header5448 ]
  %6674 = phi i1 [ false, %loop-memcpy-residual-header5448.post-loop-memcpy-expansion5444_crit_edge ], [ true, %loop-memcpy-residual-header5448 ]
  br i1 %6674, label %loop-memcpy-residual5447.preheader, label %Flow7683, !amdgpu.uniform !42

loop-memcpy-residual5447.preheader:               ; preds = %Flow7682
  %6675 = shl nuw nsw i64 %sh_prom.i.i.i.i3270, 6
  %6676 = add nuw nsw i64 %6675, 8
  %scevgep6561 = getelementptr i8, ptr %5870, i64 %6676
  %6677 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i4473, 6
  %scevgep6562 = getelementptr i8, ptr %scevgep6561, i64 %6677
  br label %loop-memcpy-residual5447, !amdgpu.uniform !42

Flow7675:                                         ; preds = %loop-memcpy-expansion5451
  br label %Flow7676, !amdgpu.uniform !42

Flow7677:                                         ; preds = %Flow7674, %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i4494
  br label %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i4495, !amdgpu.uniform !42

loop-memcpy-residual-header5454:                  ; preds = %Flow7676
  %6678 = icmp ne i64 %5908, 0
  br i1 %6678, label %loop-memcpy-residual5453.preheader, label %Flow7674, !amdgpu.uniform !42

loop-memcpy-residual5453.preheader:               ; preds = %loop-memcpy-residual-header5454
  %umin6556 = call i64 @llvm.umin.i64(i64 %lsr.iv6554, i64 64)
  %6679 = lshr i64 %umin6556, 4
  %6680 = shl nuw nsw i64 %6679, 4
  %scevgep6559 = getelementptr i8, ptr addrspace(4) %lsr.iv6549, i64 %6680
  %scevgep6552 = getelementptr i8, ptr %5907, i64 %5881
  %scevgep6553 = getelementptr i8, ptr %scevgep6552, i64 %5882
  %scevgep6557 = getelementptr i8, ptr %scevgep6553, i64 %6680
  br label %loop-memcpy-residual5453, !amdgpu.uniform !42

Flow7664:                                         ; preds = %loop-memcpy-expansion5457
  br label %Flow7665, !amdgpu.uniform !42

loop-memcpy-residual-header5460:                  ; preds = %Flow7665
  %6681 = icmp eq i64 %5987, 0
  br i1 %6681, label %loop-memcpy-residual-header5460.post-loop-memcpy-expansion5456_crit_edge, label %Flow7662, !amdgpu.uniform !42

loop-memcpy-residual-header5460.post-loop-memcpy-expansion5456_crit_edge: ; preds = %loop-memcpy-residual-header5460
  %.pre7299 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i4579, 6
  br label %Flow7662, !amdgpu.uniform !42

Flow7662:                                         ; preds = %loop-memcpy-residual-header5460.post-loop-memcpy-expansion5456_crit_edge, %loop-memcpy-residual-header5460
  %6682 = phi i64 [ %.pre7299, %loop-memcpy-residual-header5460.post-loop-memcpy-expansion5456_crit_edge ], [ poison, %loop-memcpy-residual-header5460 ]
  %6683 = phi i1 [ false, %loop-memcpy-residual-header5460.post-loop-memcpy-expansion5456_crit_edge ], [ true, %loop-memcpy-residual-header5460 ]
  br i1 %6683, label %loop-memcpy-residual5459.preheader, label %Flow7663, !amdgpu.uniform !42

loop-memcpy-residual5459.preheader:               ; preds = %Flow7662
  %6684 = add nuw nsw i64 %5881, 24
  %scevgep6542 = getelementptr i8, ptr %5986, i64 %6684
  %6685 = shl nuw nsw i64 %idxprom.i19.i.i.i.i.i4579, 6
  %scevgep6543 = getelementptr i8, ptr %scevgep6542, i64 %6685
  br label %loop-memcpy-residual5459, !amdgpu.uniform !42

Flow7655:                                         ; preds = %loop-memcpy-expansion5463
  br label %Flow7656, !amdgpu.uniform !42

Flow7657:                                         ; preds = %Flow7654, %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i.i4600
  br label %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i.i4601, !amdgpu.uniform !42

loop-memcpy-residual-header5466:                  ; preds = %Flow7656
  %6686 = icmp ne i64 %6022, 0
  br i1 %6686, label %loop-memcpy-residual5465.preheader, label %Flow7654, !amdgpu.uniform !42

loop-memcpy-residual5465.preheader:               ; preds = %loop-memcpy-residual-header5466
  %umin6537 = call i64 @llvm.umin.i64(i64 %lsr.iv6535, i64 64)
  %6687 = lshr i64 %umin6537, 4
  %6688 = shl nuw nsw i64 %6687, 4
  %scevgep6540 = getelementptr i8, ptr addrspace(4) %lsr.iv6530, i64 %6688
  %scevgep6533 = getelementptr i8, ptr %6021, i64 %5881
  %scevgep6534 = getelementptr i8, ptr %scevgep6533, i64 %5996
  %scevgep6538 = getelementptr i8, ptr %scevgep6534, i64 %6688
  br label %loop-memcpy-residual5465, !amdgpu.uniform !42

ComputeLoop:                                      ; preds = %for.cond.i.i1.i, %ComputeLoop
  %Accumulator = phi i32 [ %6693, %ComputeLoop ], [ 0, %for.cond.i.i1.i ]
  %OldValuePhi = phi i32 [ %6692, %ComputeLoop ], [ poison, %for.cond.i.i1.i ]
  %ActiveBits = phi i64 [ %6696, %ComputeLoop ], [ %21, %for.cond.i.i1.i ]
  %6689 = call i64 @llvm.cttz.i64(i64 %ActiveBits, i1 true)
  %6690 = trunc i64 %6689 to i32
  %6691 = call i32 @llvm.amdgcn.readlane.i32(i32 %shl.i.i.i.i.i, i32 %6690)
  %6692 = call i32 @llvm.amdgcn.writelane.i32(i32 %Accumulator, i32 %6690, i32 %OldValuePhi)
  %6693 = or i32 %Accumulator, %6691
  %6694 = shl i64 1, %6689
  %6695 = xor i64 %6694, -1
  %6696 = and i64 %ActiveBits, %6695
  %6697 = icmp eq i64 %6696, 0
  br i1 %6697, label %ComputeEnd, label %ComputeLoop, !amdgpu.uniform !42

ComputeEnd:                                       ; preds = %ComputeLoop
  %.lcssa9359 = phi i32 [ %6692, %ComputeLoop ]
  %.lcssa9358 = phi i32 [ %6693, %ComputeLoop ]
  %div8.i.i.i.i.i = lshr i32 %spec.store.select.i.i.i, 5
  %idxprom.i.i.i.i.i = zext nneg i32 %div8.i.i.i.i.i to i64
  %6698 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i.i.i
  %6699 = trunc i64 %21 to i32
  %6700 = lshr i64 %21, 32
  %6701 = trunc i64 %6700 to i32
  %6702 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %6699, i32 0)
  %6703 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %6701, i32 %6702)
  %6704 = icmp eq i32 %6703, 0
  %6705 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %6704)
  %6706 = extractvalue { i1, i64 } %6705, 0
  %6707 = extractvalue { i1, i64 } %6705, 1
  br i1 %6706, label %22, label %24

Flow8605:                                         ; preds = %42, %if.then.i.i.i.i2
  %6708 = phi i1 [ false, %42 ], [ true, %if.then.i.i.i.i2 ]
  br i1 %6708, label %if.end14.i.i.i, label %Flow8606, !amdgpu.uniform !42

ComputeLoop5468:                                  ; preds = %if.then12.i.i.i, %ComputeLoop5468
  %Accumulator5470 = phi i32 [ %6712, %ComputeLoop5468 ], [ -1, %if.then12.i.i.i ]
  %ActiveBits5471 = phi i64 [ %6715, %ComputeLoop5468 ], [ %39, %if.then12.i.i.i ]
  %6709 = call i64 @llvm.cttz.i64(i64 %ActiveBits5471, i1 true)
  %6710 = trunc i64 %6709 to i32
  %6711 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i.i.i, i32 %6710)
  %6712 = and i32 %Accumulator5470, %6711
  %6713 = shl i64 1, %6709
  %6714 = xor i64 %6713, -1
  %6715 = and i64 %ActiveBits5471, %6714
  %6716 = icmp eq i64 %6715, 0
  br i1 %6716, label %ComputeEnd5469, label %ComputeLoop5468, !amdgpu.uniform !42

ComputeEnd5469:                                   ; preds = %ComputeLoop5468
  %.lcssa9360 = phi i32 [ %6712, %ComputeLoop5468 ]
  %6717 = trunc i64 %39 to i32
  %6718 = lshr i64 %39, 32
  %6719 = trunc i64 %6718 to i32
  %6720 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %6717, i32 0)
  %6721 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %6719, i32 %6720)
  %6722 = icmp eq i32 %6721, 0
  %6723 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %6722)
  %6724 = extractvalue { i1, i64 } %6723, 0
  %6725 = extractvalue { i1, i64 } %6723, 1
  br i1 %6724, label %40, label %42

ComputeLoop5472:                                  ; preds = %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit, %ComputeLoop5472
  %Accumulator5474 = phi i32 [ %6729, %ComputeLoop5472 ], [ -1, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit ]
  %ActiveBits5475 = phi i64 [ %6732, %ComputeLoop5472 ], [ %146, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit ]
  %6726 = call i64 @llvm.cttz.i64(i64 %ActiveBits5475, i1 true)
  %6727 = trunc i64 %6726 to i32
  %6728 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i21.i.i, i32 %6727)
  %6729 = and i32 %Accumulator5474, %6728
  %6730 = shl i64 1, %6726
  %6731 = xor i64 %6730, -1
  %6732 = and i64 %ActiveBits5475, %6731
  %6733 = icmp eq i64 %6732, 0
  br i1 %6733, label %ComputeEnd5473, label %ComputeLoop5472, !amdgpu.uniform !42

ComputeEnd5473:                                   ; preds = %ComputeLoop5472
  %.lcssa9350 = phi i32 [ %6729, %ComputeLoop5472 ]
  %6734 = trunc i64 %146 to i32
  %6735 = lshr i64 %146, 32
  %6736 = trunc i64 %6735 to i32
  %6737 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %6734, i32 0)
  %6738 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %6736, i32 %6737)
  %6739 = icmp eq i32 %6738, 0
  %6740 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %6739)
  %6741 = extractvalue { i1, i64 } %6740, 0
  %6742 = extractvalue { i1, i64 } %6740, 1
  br i1 %6741, label %147, label %150

ComputeLoop5476:                                  ; preds = %for.cond.i.i1.i13, %ComputeLoop5476
  %Accumulator5478 = phi i32 [ %6747, %ComputeLoop5476 ], [ 0, %for.cond.i.i1.i13 ]
  %OldValuePhi5479 = phi i32 [ %6746, %ComputeLoop5476 ], [ poison, %for.cond.i.i1.i13 ]
  %ActiveBits5480 = phi i64 [ %6750, %ComputeLoop5476 ], [ %159, %for.cond.i.i1.i13 ]
  %6743 = call i64 @llvm.cttz.i64(i64 %ActiveBits5480, i1 true)
  %6744 = trunc i64 %6743 to i32
  %6745 = call i32 @llvm.amdgcn.readlane.i32(i32 %shl.i.i.i.i.i25, i32 %6744)
  %6746 = call i32 @llvm.amdgcn.writelane.i32(i32 %Accumulator5478, i32 %6744, i32 %OldValuePhi5479)
  %6747 = or i32 %Accumulator5478, %6745
  %6748 = shl i64 1, %6743
  %6749 = xor i64 %6748, -1
  %6750 = and i64 %ActiveBits5480, %6749
  %6751 = icmp eq i64 %6750, 0
  br i1 %6751, label %ComputeEnd5477, label %ComputeLoop5476, !amdgpu.uniform !42

ComputeEnd5477:                                   ; preds = %ComputeLoop5476
  %.lcssa9345 = phi i32 [ %6746, %ComputeLoop5476 ]
  %.lcssa9344 = phi i32 [ %6747, %ComputeLoop5476 ]
  %div8.i.i.i.i.i21 = lshr i32 %spec.store.select.i.i.i20, 5
  %idxprom.i.i.i.i.i22 = zext nneg i32 %div8.i.i.i.i.i21 to i64
  %6752 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i.i.i22
  %6753 = trunc i64 %159 to i32
  %6754 = lshr i64 %159, 32
  %6755 = trunc i64 %6754 to i32
  %6756 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %6753, i32 0)
  %6757 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %6755, i32 %6756)
  %6758 = icmp eq i32 %6757, 0
  %6759 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %6758)
  %6760 = extractvalue { i1, i64 } %6759, 0
  %6761 = extractvalue { i1, i64 } %6759, 1
  br i1 %6760, label %160, label %162

Flow8586:                                         ; preds = %180, %if.then.i.i.i.i31
  %6762 = phi i1 [ false, %180 ], [ true, %if.then.i.i.i.i31 ]
  br i1 %6762, label %if.end14.i.i.i101, label %Flow8587, !amdgpu.uniform !42

ComputeLoop5481:                                  ; preds = %if.then12.i.i.i36, %ComputeLoop5481
  %Accumulator5483 = phi i32 [ %6766, %ComputeLoop5481 ], [ -1, %if.then12.i.i.i36 ]
  %ActiveBits5484 = phi i64 [ %6769, %ComputeLoop5481 ], [ %177, %if.then12.i.i.i36 ]
  %6763 = call i64 @llvm.cttz.i64(i64 %ActiveBits5484, i1 true)
  %6764 = trunc i64 %6763 to i32
  %6765 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i.i.i42, i32 %6764)
  %6766 = and i32 %Accumulator5483, %6765
  %6767 = shl i64 1, %6763
  %6768 = xor i64 %6767, -1
  %6769 = and i64 %ActiveBits5484, %6768
  %6770 = icmp eq i64 %6769, 0
  br i1 %6770, label %ComputeEnd5482, label %ComputeLoop5481, !amdgpu.uniform !42

ComputeEnd5482:                                   ; preds = %ComputeLoop5481
  %.lcssa9346 = phi i32 [ %6766, %ComputeLoop5481 ]
  %6771 = trunc i64 %177 to i32
  %6772 = lshr i64 %177, 32
  %6773 = trunc i64 %6772 to i32
  %6774 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %6771, i32 0)
  %6775 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %6773, i32 %6774)
  %6776 = icmp eq i32 %6775, 0
  %6777 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %6776)
  %6778 = extractvalue { i1, i64 } %6777, 0
  %6779 = extractvalue { i1, i64 } %6777, 1
  br i1 %6778, label %178, label %180

ComputeLoop5485:                                  ; preds = %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit109, %ComputeLoop5485
  %Accumulator5487 = phi i32 [ %6783, %ComputeLoop5485 ], [ -1, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit109 ]
  %ActiveBits5488 = phi i64 [ %6786, %ComputeLoop5485 ], [ %257, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit109 ]
  %6780 = call i64 @llvm.cttz.i64(i64 %ActiveBits5488, i1 true)
  %6781 = trunc i64 %6780 to i32
  %6782 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i21.i.i100, i32 %6781)
  %6783 = and i32 %Accumulator5487, %6782
  %6784 = shl i64 1, %6780
  %6785 = xor i64 %6784, -1
  %6786 = and i64 %ActiveBits5488, %6785
  %6787 = icmp eq i64 %6786, 0
  br i1 %6787, label %ComputeEnd5486, label %ComputeLoop5485, !amdgpu.uniform !42

ComputeEnd5486:                                   ; preds = %ComputeLoop5485
  %.lcssa9340 = phi i32 [ %6783, %ComputeLoop5485 ]
  %6788 = trunc i64 %257 to i32
  %6789 = lshr i64 %257, 32
  %6790 = trunc i64 %6789 to i32
  %6791 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %6788, i32 0)
  %6792 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %6790, i32 %6791)
  %6793 = icmp eq i32 %6792, 0
  %6794 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %6793)
  %6795 = extractvalue { i1, i64 } %6794, 0
  %6796 = extractvalue { i1, i64 } %6794, 1
  br i1 %6795, label %258, label %261

ComputeLoop5489:                                  ; preds = %for.cond.i.i1.i119, %ComputeLoop5489
  %Accumulator5491 = phi i32 [ %6801, %ComputeLoop5489 ], [ 0, %for.cond.i.i1.i119 ]
  %OldValuePhi5492 = phi i32 [ %6800, %ComputeLoop5489 ], [ poison, %for.cond.i.i1.i119 ]
  %ActiveBits5493 = phi i64 [ %6804, %ComputeLoop5489 ], [ %273, %for.cond.i.i1.i119 ]
  %6797 = call i64 @llvm.cttz.i64(i64 %ActiveBits5493, i1 true)
  %6798 = trunc i64 %6797 to i32
  %6799 = call i32 @llvm.amdgcn.readlane.i32(i32 %shl.i.i.i.i.i131, i32 %6798)
  %6800 = call i32 @llvm.amdgcn.writelane.i32(i32 %Accumulator5491, i32 %6798, i32 %OldValuePhi5492)
  %6801 = or i32 %Accumulator5491, %6799
  %6802 = shl i64 1, %6797
  %6803 = xor i64 %6802, -1
  %6804 = and i64 %ActiveBits5493, %6803
  %6805 = icmp eq i64 %6804, 0
  br i1 %6805, label %ComputeEnd5490, label %ComputeLoop5489, !amdgpu.uniform !42

ComputeEnd5490:                                   ; preds = %ComputeLoop5489
  %.lcssa9334 = phi i32 [ %6800, %ComputeLoop5489 ]
  %.lcssa9333 = phi i32 [ %6801, %ComputeLoop5489 ]
  %div8.i.i.i.i.i127 = lshr i32 %spec.store.select.i.i.i126, 5
  %idxprom.i.i.i.i.i128 = zext nneg i32 %div8.i.i.i.i.i127 to i64
  %6806 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i.i.i128
  %6807 = trunc i64 %273 to i32
  %6808 = lshr i64 %273, 32
  %6809 = trunc i64 %6808 to i32
  %6810 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %6807, i32 0)
  %6811 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %6809, i32 %6810)
  %6812 = icmp eq i32 %6811, 0
  %6813 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %6812)
  %6814 = extractvalue { i1, i64 } %6813, 0
  %6815 = extractvalue { i1, i64 } %6813, 1
  br i1 %6814, label %274, label %276

Flow8567:                                         ; preds = %294, %if.then.i.i.i.i137
  %6816 = phi i1 [ false, %294 ], [ true, %if.then.i.i.i.i137 ]
  br i1 %6816, label %if.end14.i.i.i207, label %Flow8568, !amdgpu.uniform !42

ComputeLoop5494:                                  ; preds = %if.then12.i.i.i142, %ComputeLoop5494
  %Accumulator5496 = phi i32 [ %6820, %ComputeLoop5494 ], [ -1, %if.then12.i.i.i142 ]
  %ActiveBits5497 = phi i64 [ %6823, %ComputeLoop5494 ], [ %291, %if.then12.i.i.i142 ]
  %6817 = call i64 @llvm.cttz.i64(i64 %ActiveBits5497, i1 true)
  %6818 = trunc i64 %6817 to i32
  %6819 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i.i.i148, i32 %6818)
  %6820 = and i32 %Accumulator5496, %6819
  %6821 = shl i64 1, %6817
  %6822 = xor i64 %6821, -1
  %6823 = and i64 %ActiveBits5497, %6822
  %6824 = icmp eq i64 %6823, 0
  br i1 %6824, label %ComputeEnd5495, label %ComputeLoop5494, !amdgpu.uniform !42

ComputeEnd5495:                                   ; preds = %ComputeLoop5494
  %.lcssa9335 = phi i32 [ %6820, %ComputeLoop5494 ]
  %6825 = trunc i64 %291 to i32
  %6826 = lshr i64 %291, 32
  %6827 = trunc i64 %6826 to i32
  %6828 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %6825, i32 0)
  %6829 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %6827, i32 %6828)
  %6830 = icmp eq i32 %6829, 0
  %6831 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %6830)
  %6832 = extractvalue { i1, i64 } %6831, 0
  %6833 = extractvalue { i1, i64 } %6831, 1
  br i1 %6832, label %292, label %294

ComputeLoop5498:                                  ; preds = %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit215, %ComputeLoop5498
  %Accumulator5500 = phi i32 [ %6837, %ComputeLoop5498 ], [ -1, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit215 ]
  %ActiveBits5501 = phi i64 [ %6840, %ComputeLoop5498 ], [ %396, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit215 ]
  %6834 = call i64 @llvm.cttz.i64(i64 %ActiveBits5501, i1 true)
  %6835 = trunc i64 %6834 to i32
  %6836 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i21.i.i206, i32 %6835)
  %6837 = and i32 %Accumulator5500, %6836
  %6838 = shl i64 1, %6834
  %6839 = xor i64 %6838, -1
  %6840 = and i64 %ActiveBits5501, %6839
  %6841 = icmp eq i64 %6840, 0
  br i1 %6841, label %ComputeEnd5499, label %ComputeLoop5498, !amdgpu.uniform !42

ComputeEnd5499:                                   ; preds = %ComputeLoop5498
  %.lcssa9325 = phi i32 [ %6837, %ComputeLoop5498 ]
  %6842 = trunc i64 %396 to i32
  %6843 = lshr i64 %396, 32
  %6844 = trunc i64 %6843 to i32
  %6845 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %6842, i32 0)
  %6846 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %6844, i32 %6845)
  %6847 = icmp eq i32 %6846, 0
  %6848 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %6847)
  %6849 = extractvalue { i1, i64 } %6848, 0
  %6850 = extractvalue { i1, i64 } %6848, 1
  br i1 %6849, label %397, label %400

ComputeLoop5502:                                  ; preds = %for.cond.i.i1.i224, %ComputeLoop5502
  %Accumulator5504 = phi i32 [ %6855, %ComputeLoop5502 ], [ 0, %for.cond.i.i1.i224 ]
  %OldValuePhi5505 = phi i32 [ %6854, %ComputeLoop5502 ], [ poison, %for.cond.i.i1.i224 ]
  %ActiveBits5506 = phi i64 [ %6858, %ComputeLoop5502 ], [ %409, %for.cond.i.i1.i224 ]
  %6851 = call i64 @llvm.cttz.i64(i64 %ActiveBits5506, i1 true)
  %6852 = trunc i64 %6851 to i32
  %6853 = call i32 @llvm.amdgcn.readlane.i32(i32 %shl.i.i.i.i.i236, i32 %6852)
  %6854 = call i32 @llvm.amdgcn.writelane.i32(i32 %Accumulator5504, i32 %6852, i32 %OldValuePhi5505)
  %6855 = or i32 %Accumulator5504, %6853
  %6856 = shl i64 1, %6851
  %6857 = xor i64 %6856, -1
  %6858 = and i64 %ActiveBits5506, %6857
  %6859 = icmp eq i64 %6858, 0
  br i1 %6859, label %ComputeEnd5503, label %ComputeLoop5502, !amdgpu.uniform !42

ComputeEnd5503:                                   ; preds = %ComputeLoop5502
  %.lcssa9320 = phi i32 [ %6854, %ComputeLoop5502 ]
  %.lcssa9319 = phi i32 [ %6855, %ComputeLoop5502 ]
  %div8.i.i.i.i.i232 = lshr i32 %spec.store.select.i.i.i231, 5
  %idxprom.i.i.i.i.i233 = zext nneg i32 %div8.i.i.i.i.i232 to i64
  %6860 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i.i.i233
  %6861 = trunc i64 %409 to i32
  %6862 = lshr i64 %409, 32
  %6863 = trunc i64 %6862 to i32
  %6864 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %6861, i32 0)
  %6865 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %6863, i32 %6864)
  %6866 = icmp eq i32 %6865, 0
  %6867 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %6866)
  %6868 = extractvalue { i1, i64 } %6867, 0
  %6869 = extractvalue { i1, i64 } %6867, 1
  br i1 %6868, label %410, label %412

Flow8548:                                         ; preds = %430, %if.then.i.i.i.i242
  %6870 = phi i1 [ false, %430 ], [ true, %if.then.i.i.i.i242 ]
  br i1 %6870, label %if.end14.i.i.i312, label %Flow8549, !amdgpu.uniform !42

ComputeLoop5507:                                  ; preds = %if.then12.i.i.i247, %ComputeLoop5507
  %Accumulator5509 = phi i32 [ %6874, %ComputeLoop5507 ], [ -1, %if.then12.i.i.i247 ]
  %ActiveBits5510 = phi i64 [ %6877, %ComputeLoop5507 ], [ %427, %if.then12.i.i.i247 ]
  %6871 = call i64 @llvm.cttz.i64(i64 %ActiveBits5510, i1 true)
  %6872 = trunc i64 %6871 to i32
  %6873 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i.i.i253, i32 %6872)
  %6874 = and i32 %Accumulator5509, %6873
  %6875 = shl i64 1, %6871
  %6876 = xor i64 %6875, -1
  %6877 = and i64 %ActiveBits5510, %6876
  %6878 = icmp eq i64 %6877, 0
  br i1 %6878, label %ComputeEnd5508, label %ComputeLoop5507, !amdgpu.uniform !42

ComputeEnd5508:                                   ; preds = %ComputeLoop5507
  %.lcssa9321 = phi i32 [ %6874, %ComputeLoop5507 ]
  %6879 = trunc i64 %427 to i32
  %6880 = lshr i64 %427, 32
  %6881 = trunc i64 %6880 to i32
  %6882 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %6879, i32 0)
  %6883 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %6881, i32 %6882)
  %6884 = icmp eq i32 %6883, 0
  %6885 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %6884)
  %6886 = extractvalue { i1, i64 } %6885, 0
  %6887 = extractvalue { i1, i64 } %6885, 1
  br i1 %6886, label %428, label %430

ComputeLoop5511:                                  ; preds = %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit320, %ComputeLoop5511
  %Accumulator5513 = phi i32 [ %6891, %ComputeLoop5511 ], [ -1, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit320 ]
  %ActiveBits5514 = phi i64 [ %6894, %ComputeLoop5511 ], [ %507, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit320 ]
  %6888 = call i64 @llvm.cttz.i64(i64 %ActiveBits5514, i1 true)
  %6889 = trunc i64 %6888 to i32
  %6890 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i21.i.i311, i32 %6889)
  %6891 = and i32 %Accumulator5513, %6890
  %6892 = shl i64 1, %6888
  %6893 = xor i64 %6892, -1
  %6894 = and i64 %ActiveBits5514, %6893
  %6895 = icmp eq i64 %6894, 0
  br i1 %6895, label %ComputeEnd5512, label %ComputeLoop5511, !amdgpu.uniform !42

ComputeEnd5512:                                   ; preds = %ComputeLoop5511
  %.lcssa9315 = phi i32 [ %6891, %ComputeLoop5511 ]
  %6896 = trunc i64 %507 to i32
  %6897 = lshr i64 %507, 32
  %6898 = trunc i64 %6897 to i32
  %6899 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %6896, i32 0)
  %6900 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %6898, i32 %6899)
  %6901 = icmp eq i32 %6900, 0
  %6902 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %6901)
  %6903 = extractvalue { i1, i64 } %6902, 0
  %6904 = extractvalue { i1, i64 } %6902, 1
  br i1 %6903, label %508, label %511

ComputeLoop5515:                                  ; preds = %for.cond.i.i.i327, %ComputeLoop5515
  %Accumulator5517 = phi i32 [ %6909, %ComputeLoop5515 ], [ 0, %for.cond.i.i.i327 ]
  %OldValuePhi5518 = phi i32 [ %6908, %ComputeLoop5515 ], [ poison, %for.cond.i.i.i327 ]
  %ActiveBits5519 = phi i64 [ %6912, %ComputeLoop5515 ], [ %643, %for.cond.i.i.i327 ]
  %6905 = call i64 @llvm.cttz.i64(i64 %ActiveBits5519, i1 true)
  %6906 = trunc i64 %6905 to i32
  %6907 = call i32 @llvm.amdgcn.readlane.i32(i32 %shl.i.i.i.i.i339, i32 %6906)
  %6908 = call i32 @llvm.amdgcn.writelane.i32(i32 %Accumulator5517, i32 %6906, i32 %OldValuePhi5518)
  %6909 = or i32 %Accumulator5517, %6907
  %6910 = shl i64 1, %6905
  %6911 = xor i64 %6910, -1
  %6912 = and i64 %ActiveBits5519, %6911
  %6913 = icmp eq i64 %6912, 0
  br i1 %6913, label %ComputeEnd5516, label %ComputeLoop5515, !amdgpu.uniform !42

ComputeEnd5516:                                   ; preds = %ComputeLoop5515
  %.lcssa9306 = phi i32 [ %6908, %ComputeLoop5515 ]
  %.lcssa9305 = phi i32 [ %6909, %ComputeLoop5515 ]
  %div8.i.i.i.i.i335 = lshr i32 %spec.store.select.i.i.i334, 5
  %idxprom.i.i.i.i.i336 = zext nneg i32 %div8.i.i.i.i.i335 to i64
  %6914 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i.i.i336
  %6915 = trunc i64 %643 to i32
  %6916 = lshr i64 %643, 32
  %6917 = trunc i64 %6916 to i32
  %6918 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %6915, i32 0)
  %6919 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %6917, i32 %6918)
  %6920 = icmp eq i32 %6919, 0
  %6921 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %6920)
  %6922 = extractvalue { i1, i64 } %6921, 0
  %6923 = extractvalue { i1, i64 } %6921, 1
  br i1 %6922, label %644, label %646

Flow8505:                                         ; preds = %664, %if.then.i.i.i.i345
  %6924 = phi i1 [ false, %664 ], [ true, %if.then.i.i.i.i345 ]
  br i1 %6924, label %if.end14.i.i.i413, label %Flow8506, !amdgpu.uniform !42

ComputeLoop5520:                                  ; preds = %if.then12.i.i.i350, %ComputeLoop5520
  %Accumulator5522 = phi i32 [ %6928, %ComputeLoop5520 ], [ -1, %if.then12.i.i.i350 ]
  %ActiveBits5523 = phi i64 [ %6931, %ComputeLoop5520 ], [ %661, %if.then12.i.i.i350 ]
  %6925 = call i64 @llvm.cttz.i64(i64 %ActiveBits5523, i1 true)
  %6926 = trunc i64 %6925 to i32
  %6927 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i.i.i355, i32 %6926)
  %6928 = and i32 %Accumulator5522, %6927
  %6929 = shl i64 1, %6925
  %6930 = xor i64 %6929, -1
  %6931 = and i64 %ActiveBits5523, %6930
  %6932 = icmp eq i64 %6931, 0
  br i1 %6932, label %ComputeEnd5521, label %ComputeLoop5520, !amdgpu.uniform !42

ComputeEnd5521:                                   ; preds = %ComputeLoop5520
  %.lcssa9307 = phi i32 [ %6928, %ComputeLoop5520 ]
  %6933 = trunc i64 %661 to i32
  %6934 = lshr i64 %661, 32
  %6935 = trunc i64 %6934 to i32
  %6936 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %6933, i32 0)
  %6937 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %6935, i32 %6936)
  %6938 = icmp eq i32 %6937, 0
  %6939 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %6938)
  %6940 = extractvalue { i1, i64 } %6939, 0
  %6941 = extractvalue { i1, i64 } %6939, 1
  br i1 %6940, label %662, label %664

ComputeLoop5524:                                  ; preds = %_ZN22__llvm_libc_22_0_0_git15write_to_stderrENS_3cpp11string_viewE.exit.i, %ComputeLoop5524
  %Accumulator5526 = phi i32 [ %6945, %ComputeLoop5524 ], [ -1, %_ZN22__llvm_libc_22_0_0_git15write_to_stderrENS_3cpp11string_viewE.exit.i ]
  %ActiveBits5527 = phi i64 [ %6948, %ComputeLoop5524 ], [ %759, %_ZN22__llvm_libc_22_0_0_git15write_to_stderrENS_3cpp11string_viewE.exit.i ]
  %6942 = call i64 @llvm.cttz.i64(i64 %ActiveBits5527, i1 true)
  %6943 = trunc i64 %6942 to i32
  %6944 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i21.i.i412, i32 %6943)
  %6945 = and i32 %Accumulator5526, %6944
  %6946 = shl i64 1, %6942
  %6947 = xor i64 %6946, -1
  %6948 = and i64 %ActiveBits5527, %6947
  %6949 = icmp eq i64 %6948, 0
  br i1 %6949, label %ComputeEnd5525, label %ComputeLoop5524, !amdgpu.uniform !42

ComputeEnd5525:                                   ; preds = %ComputeLoop5524
  %.lcssa9297 = phi i32 [ %6945, %ComputeLoop5524 ]
  %6950 = trunc i64 %759 to i32
  %6951 = lshr i64 %759, 32
  %6952 = trunc i64 %6951 to i32
  %6953 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %6950, i32 0)
  %6954 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %6952, i32 %6953)
  %6955 = icmp eq i32 %6954, 0
  %6956 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %6955)
  %6957 = extractvalue { i1, i64 } %6956, 0
  %6958 = extractvalue { i1, i64 } %6956, 1
  br i1 %6957, label %760, label %763

ComputeLoop5528:                                  ; preds = %for.cond.i.i1.i429, %ComputeLoop5528
  %Accumulator5530 = phi i32 [ %6963, %ComputeLoop5528 ], [ 0, %for.cond.i.i1.i429 ]
  %OldValuePhi5531 = phi i32 [ %6962, %ComputeLoop5528 ], [ poison, %for.cond.i.i1.i429 ]
  %ActiveBits5532 = phi i64 [ %6966, %ComputeLoop5528 ], [ %772, %for.cond.i.i1.i429 ]
  %6959 = call i64 @llvm.cttz.i64(i64 %ActiveBits5532, i1 true)
  %6960 = trunc i64 %6959 to i32
  %6961 = call i32 @llvm.amdgcn.readlane.i32(i32 %shl.i.i.i.i.i441, i32 %6960)
  %6962 = call i32 @llvm.amdgcn.writelane.i32(i32 %Accumulator5530, i32 %6960, i32 %OldValuePhi5531)
  %6963 = or i32 %Accumulator5530, %6961
  %6964 = shl i64 1, %6959
  %6965 = xor i64 %6964, -1
  %6966 = and i64 %ActiveBits5532, %6965
  %6967 = icmp eq i64 %6966, 0
  br i1 %6967, label %ComputeEnd5529, label %ComputeLoop5528, !amdgpu.uniform !42

ComputeEnd5529:                                   ; preds = %ComputeLoop5528
  %.lcssa9292 = phi i32 [ %6962, %ComputeLoop5528 ]
  %.lcssa9291 = phi i32 [ %6963, %ComputeLoop5528 ]
  %div8.i.i.i.i.i437 = lshr i32 %spec.store.select.i.i.i436, 5
  %idxprom.i.i.i.i.i438 = zext nneg i32 %div8.i.i.i.i.i437 to i64
  %6968 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i.i.i438
  %6969 = trunc i64 %772 to i32
  %6970 = lshr i64 %772, 32
  %6971 = trunc i64 %6970 to i32
  %6972 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %6969, i32 0)
  %6973 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %6971, i32 %6972)
  %6974 = icmp eq i32 %6973, 0
  %6975 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %6974)
  %6976 = extractvalue { i1, i64 } %6975, 0
  %6977 = extractvalue { i1, i64 } %6975, 1
  br i1 %6976, label %773, label %775

Flow8486:                                         ; preds = %793, %if.then.i.i.i.i447
  %6978 = phi i1 [ false, %793 ], [ true, %if.then.i.i.i.i447 ]
  br i1 %6978, label %if.end14.i.i.i518, label %Flow8487, !amdgpu.uniform !42

ComputeLoop5533:                                  ; preds = %if.then12.i.i.i452, %ComputeLoop5533
  %Accumulator5535 = phi i32 [ %6982, %ComputeLoop5533 ], [ -1, %if.then12.i.i.i452 ]
  %ActiveBits5536 = phi i64 [ %6985, %ComputeLoop5533 ], [ %790, %if.then12.i.i.i452 ]
  %6979 = call i64 @llvm.cttz.i64(i64 %ActiveBits5536, i1 true)
  %6980 = trunc i64 %6979 to i32
  %6981 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i.i.i458, i32 %6980)
  %6982 = and i32 %Accumulator5535, %6981
  %6983 = shl i64 1, %6979
  %6984 = xor i64 %6983, -1
  %6985 = and i64 %ActiveBits5536, %6984
  %6986 = icmp eq i64 %6985, 0
  br i1 %6986, label %ComputeEnd5534, label %ComputeLoop5533, !amdgpu.uniform !42

ComputeEnd5534:                                   ; preds = %ComputeLoop5533
  %.lcssa9293 = phi i32 [ %6982, %ComputeLoop5533 ]
  %6987 = trunc i64 %790 to i32
  %6988 = lshr i64 %790, 32
  %6989 = trunc i64 %6988 to i32
  %6990 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %6987, i32 0)
  %6991 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %6989, i32 %6990)
  %6992 = icmp eq i32 %6991, 0
  %6993 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %6992)
  %6994 = extractvalue { i1, i64 } %6993, 0
  %6995 = extractvalue { i1, i64 } %6993, 1
  br i1 %6994, label %791, label %793

ComputeLoop5537:                                  ; preds = %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit526, %ComputeLoop5537
  %Accumulator5539 = phi i32 [ %6999, %ComputeLoop5537 ], [ -1, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit526 ]
  %ActiveBits5540 = phi i64 [ %7002, %ComputeLoop5537 ], [ %870, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit526 ]
  %6996 = call i64 @llvm.cttz.i64(i64 %ActiveBits5540, i1 true)
  %6997 = trunc i64 %6996 to i32
  %6998 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i21.i.i517, i32 %6997)
  %6999 = and i32 %Accumulator5539, %6998
  %7000 = shl i64 1, %6996
  %7001 = xor i64 %7000, -1
  %7002 = and i64 %ActiveBits5540, %7001
  %7003 = icmp eq i64 %7002, 0
  br i1 %7003, label %ComputeEnd5538, label %ComputeLoop5537, !amdgpu.uniform !42

ComputeEnd5538:                                   ; preds = %ComputeLoop5537
  %.lcssa9287 = phi i32 [ %6999, %ComputeLoop5537 ]
  %7004 = trunc i64 %870 to i32
  %7005 = lshr i64 %870, 32
  %7006 = trunc i64 %7005 to i32
  %7007 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %7004, i32 0)
  %7008 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %7006, i32 %7007)
  %7009 = icmp eq i32 %7008, 0
  %7010 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %7009)
  %7011 = extractvalue { i1, i64 } %7010, 0
  %7012 = extractvalue { i1, i64 } %7010, 1
  br i1 %7011, label %871, label %874

ComputeLoop5541:                                  ; preds = %for.cond.i.i1.i535, %ComputeLoop5541
  %Accumulator5543 = phi i32 [ %7017, %ComputeLoop5541 ], [ 0, %for.cond.i.i1.i535 ]
  %OldValuePhi5544 = phi i32 [ %7016, %ComputeLoop5541 ], [ poison, %for.cond.i.i1.i535 ]
  %ActiveBits5545 = phi i64 [ %7020, %ComputeLoop5541 ], [ %886, %for.cond.i.i1.i535 ]
  %7013 = call i64 @llvm.cttz.i64(i64 %ActiveBits5545, i1 true)
  %7014 = trunc i64 %7013 to i32
  %7015 = call i32 @llvm.amdgcn.readlane.i32(i32 %shl.i.i.i.i.i547, i32 %7014)
  %7016 = call i32 @llvm.amdgcn.writelane.i32(i32 %Accumulator5543, i32 %7014, i32 %OldValuePhi5544)
  %7017 = or i32 %Accumulator5543, %7015
  %7018 = shl i64 1, %7013
  %7019 = xor i64 %7018, -1
  %7020 = and i64 %ActiveBits5545, %7019
  %7021 = icmp eq i64 %7020, 0
  br i1 %7021, label %ComputeEnd5542, label %ComputeLoop5541, !amdgpu.uniform !42

ComputeEnd5542:                                   ; preds = %ComputeLoop5541
  %.lcssa9282 = phi i32 [ %7016, %ComputeLoop5541 ]
  %.lcssa9281 = phi i32 [ %7017, %ComputeLoop5541 ]
  %div8.i.i.i.i.i543 = lshr i32 %spec.store.select.i.i.i542, 5
  %idxprom.i.i.i.i.i544 = zext nneg i32 %div8.i.i.i.i.i543 to i64
  %7022 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i.i.i544
  %7023 = trunc i64 %886 to i32
  %7024 = lshr i64 %886, 32
  %7025 = trunc i64 %7024 to i32
  %7026 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %7023, i32 0)
  %7027 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %7025, i32 %7026)
  %7028 = icmp eq i32 %7027, 0
  %7029 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %7028)
  %7030 = extractvalue { i1, i64 } %7029, 0
  %7031 = extractvalue { i1, i64 } %7029, 1
  br i1 %7030, label %887, label %889

Flow8466:                                         ; preds = %907, %if.then.i.i.i.i553
  %7032 = phi i1 [ false, %907 ], [ true, %if.then.i.i.i.i553 ]
  br i1 %7032, label %if.end14.i.i.i624, label %Flow8467, !amdgpu.uniform !42

ComputeLoop5546:                                  ; preds = %if.then12.i.i.i558, %ComputeLoop5546
  %Accumulator5548 = phi i32 [ %7036, %ComputeLoop5546 ], [ -1, %if.then12.i.i.i558 ]
  %ActiveBits5549 = phi i64 [ %7039, %ComputeLoop5546 ], [ %904, %if.then12.i.i.i558 ]
  %7033 = call i64 @llvm.cttz.i64(i64 %ActiveBits5549, i1 true)
  %7034 = trunc i64 %7033 to i32
  %7035 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i.i.i564, i32 %7034)
  %7036 = and i32 %Accumulator5548, %7035
  %7037 = shl i64 1, %7033
  %7038 = xor i64 %7037, -1
  %7039 = and i64 %ActiveBits5549, %7038
  %7040 = icmp eq i64 %7039, 0
  br i1 %7040, label %ComputeEnd5547, label %ComputeLoop5546, !amdgpu.uniform !42

ComputeEnd5547:                                   ; preds = %ComputeLoop5546
  %.lcssa9283 = phi i32 [ %7036, %ComputeLoop5546 ]
  %7041 = trunc i64 %904 to i32
  %7042 = lshr i64 %904, 32
  %7043 = trunc i64 %7042 to i32
  %7044 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %7041, i32 0)
  %7045 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %7043, i32 %7044)
  %7046 = icmp eq i32 %7045, 0
  %7047 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %7046)
  %7048 = extractvalue { i1, i64 } %7047, 0
  %7049 = extractvalue { i1, i64 } %7047, 1
  br i1 %7048, label %905, label %907

ComputeLoop5550:                                  ; preds = %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit632, %ComputeLoop5550
  %Accumulator5552 = phi i32 [ %7053, %ComputeLoop5550 ], [ -1, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit632 ]
  %ActiveBits5553 = phi i64 [ %7056, %ComputeLoop5550 ], [ %984, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit632 ]
  %7050 = call i64 @llvm.cttz.i64(i64 %ActiveBits5553, i1 true)
  %7051 = trunc i64 %7050 to i32
  %7052 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i21.i.i623, i32 %7051)
  %7053 = and i32 %Accumulator5552, %7052
  %7054 = shl i64 1, %7050
  %7055 = xor i64 %7054, -1
  %7056 = and i64 %ActiveBits5553, %7055
  %7057 = icmp eq i64 %7056, 0
  br i1 %7057, label %ComputeEnd5551, label %ComputeLoop5550, !amdgpu.uniform !42

ComputeEnd5551:                                   ; preds = %ComputeLoop5550
  %.lcssa9277 = phi i32 [ %7053, %ComputeLoop5550 ]
  %7058 = trunc i64 %984 to i32
  %7059 = lshr i64 %984, 32
  %7060 = trunc i64 %7059 to i32
  %7061 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %7058, i32 0)
  %7062 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %7060, i32 %7061)
  %7063 = icmp eq i32 %7062, 0
  %7064 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %7063)
  %7065 = extractvalue { i1, i64 } %7064, 0
  %7066 = extractvalue { i1, i64 } %7064, 1
  br i1 %7065, label %985, label %988

ComputeLoop5554:                                  ; preds = %for.cond.i.i1.i641, %ComputeLoop5554
  %Accumulator5556 = phi i32 [ %7071, %ComputeLoop5554 ], [ 0, %for.cond.i.i1.i641 ]
  %OldValuePhi5557 = phi i32 [ %7070, %ComputeLoop5554 ], [ poison, %for.cond.i.i1.i641 ]
  %ActiveBits5558 = phi i64 [ %7074, %ComputeLoop5554 ], [ %997, %for.cond.i.i1.i641 ]
  %7067 = call i64 @llvm.cttz.i64(i64 %ActiveBits5558, i1 true)
  %7068 = trunc i64 %7067 to i32
  %7069 = call i32 @llvm.amdgcn.readlane.i32(i32 %shl.i.i.i.i.i653, i32 %7068)
  %7070 = call i32 @llvm.amdgcn.writelane.i32(i32 %Accumulator5556, i32 %7068, i32 %OldValuePhi5557)
  %7071 = or i32 %Accumulator5556, %7069
  %7072 = shl i64 1, %7067
  %7073 = xor i64 %7072, -1
  %7074 = and i64 %ActiveBits5558, %7073
  %7075 = icmp eq i64 %7074, 0
  br i1 %7075, label %ComputeEnd5555, label %ComputeLoop5554, !amdgpu.uniform !42

ComputeEnd5555:                                   ; preds = %ComputeLoop5554
  %.lcssa9272 = phi i32 [ %7070, %ComputeLoop5554 ]
  %.lcssa9271 = phi i32 [ %7071, %ComputeLoop5554 ]
  %div8.i.i.i.i.i649 = lshr i32 %spec.store.select.i.i.i648, 5
  %idxprom.i.i.i.i.i650 = zext nneg i32 %div8.i.i.i.i.i649 to i64
  %7076 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i.i.i650
  %7077 = trunc i64 %997 to i32
  %7078 = lshr i64 %997, 32
  %7079 = trunc i64 %7078 to i32
  %7080 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %7077, i32 0)
  %7081 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %7079, i32 %7080)
  %7082 = icmp eq i32 %7081, 0
  %7083 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %7082)
  %7084 = extractvalue { i1, i64 } %7083, 0
  %7085 = extractvalue { i1, i64 } %7083, 1
  br i1 %7084, label %998, label %1000

Flow8447:                                         ; preds = %1018, %if.then.i.i.i.i659
  %7086 = phi i1 [ false, %1018 ], [ true, %if.then.i.i.i.i659 ]
  br i1 %7086, label %if.end14.i.i.i730, label %Flow8448, !amdgpu.uniform !42

ComputeLoop5559:                                  ; preds = %if.then12.i.i.i664, %ComputeLoop5559
  %Accumulator5561 = phi i32 [ %7090, %ComputeLoop5559 ], [ -1, %if.then12.i.i.i664 ]
  %ActiveBits5562 = phi i64 [ %7093, %ComputeLoop5559 ], [ %1015, %if.then12.i.i.i664 ]
  %7087 = call i64 @llvm.cttz.i64(i64 %ActiveBits5562, i1 true)
  %7088 = trunc i64 %7087 to i32
  %7089 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i.i.i670, i32 %7088)
  %7090 = and i32 %Accumulator5561, %7089
  %7091 = shl i64 1, %7087
  %7092 = xor i64 %7091, -1
  %7093 = and i64 %ActiveBits5562, %7092
  %7094 = icmp eq i64 %7093, 0
  br i1 %7094, label %ComputeEnd5560, label %ComputeLoop5559, !amdgpu.uniform !42

ComputeEnd5560:                                   ; preds = %ComputeLoop5559
  %.lcssa9273 = phi i32 [ %7090, %ComputeLoop5559 ]
  %7095 = trunc i64 %1015 to i32
  %7096 = lshr i64 %1015, 32
  %7097 = trunc i64 %7096 to i32
  %7098 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %7095, i32 0)
  %7099 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %7097, i32 %7098)
  %7100 = icmp eq i32 %7099, 0
  %7101 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %7100)
  %7102 = extractvalue { i1, i64 } %7101, 0
  %7103 = extractvalue { i1, i64 } %7101, 1
  br i1 %7102, label %1016, label %1018

ComputeLoop5563:                                  ; preds = %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit738, %ComputeLoop5563
  %Accumulator5565 = phi i32 [ %7107, %ComputeLoop5563 ], [ -1, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit738 ]
  %ActiveBits5566 = phi i64 [ %7110, %ComputeLoop5563 ], [ %1095, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit738 ]
  %7104 = call i64 @llvm.cttz.i64(i64 %ActiveBits5566, i1 true)
  %7105 = trunc i64 %7104 to i32
  %7106 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i21.i.i729, i32 %7105)
  %7107 = and i32 %Accumulator5565, %7106
  %7108 = shl i64 1, %7104
  %7109 = xor i64 %7108, -1
  %7110 = and i64 %ActiveBits5566, %7109
  %7111 = icmp eq i64 %7110, 0
  br i1 %7111, label %ComputeEnd5564, label %ComputeLoop5563, !amdgpu.uniform !42

ComputeEnd5564:                                   ; preds = %ComputeLoop5563
  %.lcssa9267 = phi i32 [ %7107, %ComputeLoop5563 ]
  %7112 = trunc i64 %1095 to i32
  %7113 = lshr i64 %1095, 32
  %7114 = trunc i64 %7113 to i32
  %7115 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %7112, i32 0)
  %7116 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %7114, i32 %7115)
  %7117 = icmp eq i32 %7116, 0
  %7118 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %7117)
  %7119 = extractvalue { i1, i64 } %7118, 0
  %7120 = extractvalue { i1, i64 } %7118, 1
  br i1 %7119, label %1096, label %1099

ComputeLoop5567:                                  ; preds = %for.cond.i.i1.i757, %ComputeLoop5567
  %Accumulator5569 = phi i32 [ %7125, %ComputeLoop5567 ], [ 0, %for.cond.i.i1.i757 ]
  %OldValuePhi5570 = phi i32 [ %7124, %ComputeLoop5567 ], [ poison, %for.cond.i.i1.i757 ]
  %ActiveBits5571 = phi i64 [ %7128, %ComputeLoop5567 ], [ %1414, %for.cond.i.i1.i757 ]
  %7121 = call i64 @llvm.cttz.i64(i64 %ActiveBits5571, i1 true)
  %7122 = trunc i64 %7121 to i32
  %7123 = call i32 @llvm.amdgcn.readlane.i32(i32 %shl.i.i.i.i.i769, i32 %7122)
  %7124 = call i32 @llvm.amdgcn.writelane.i32(i32 %Accumulator5569, i32 %7122, i32 %OldValuePhi5570)
  %7125 = or i32 %Accumulator5569, %7123
  %7126 = shl i64 1, %7121
  %7127 = xor i64 %7126, -1
  %7128 = and i64 %ActiveBits5571, %7127
  %7129 = icmp eq i64 %7128, 0
  br i1 %7129, label %ComputeEnd5568, label %ComputeLoop5567, !amdgpu.uniform !42

ComputeEnd5568:                                   ; preds = %ComputeLoop5567
  %.lcssa8979 = phi i32 [ %7124, %ComputeLoop5567 ]
  %.lcssa8978 = phi i32 [ %7125, %ComputeLoop5567 ]
  %div8.i.i.i.i.i765 = lshr i32 %spec.store.select.i.i.i764, 5
  %idxprom.i.i.i.i.i766 = zext nneg i32 %div8.i.i.i.i.i765 to i64
  %7130 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i.i.i766
  %7131 = trunc i64 %1414 to i32
  %7132 = lshr i64 %1414, 32
  %7133 = trunc i64 %7132 to i32
  %7134 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %7131, i32 0)
  %7135 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %7133, i32 %7134)
  %7136 = icmp eq i32 %7135, 0
  %7137 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %7136)
  %7138 = extractvalue { i1, i64 } %7137, 0
  %7139 = extractvalue { i1, i64 } %7137, 1
  br i1 %7138, label %1418, label %1420

Flow8423:                                         ; preds = %1438, %if.then.i.i.i.i775
  %7140 = phi i1 [ false, %1438 ], [ true, %if.then.i.i.i.i775 ]
  br i1 %7140, label %if.end14.i.i.i846, label %Flow8424, !amdgpu.uniform !42

ComputeLoop5572:                                  ; preds = %if.then12.i.i.i780, %ComputeLoop5572
  %Accumulator5574 = phi i32 [ %7144, %ComputeLoop5572 ], [ -1, %if.then12.i.i.i780 ]
  %ActiveBits5575 = phi i64 [ %7147, %ComputeLoop5572 ], [ %1435, %if.then12.i.i.i780 ]
  %7141 = call i64 @llvm.cttz.i64(i64 %ActiveBits5575, i1 true)
  %7142 = trunc i64 %7141 to i32
  %7143 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i.i.i786, i32 %7142)
  %7144 = and i32 %Accumulator5574, %7143
  %7145 = shl i64 1, %7141
  %7146 = xor i64 %7145, -1
  %7147 = and i64 %ActiveBits5575, %7146
  %7148 = icmp eq i64 %7147, 0
  br i1 %7148, label %ComputeEnd5573, label %ComputeLoop5572, !amdgpu.uniform !42

ComputeEnd5573:                                   ; preds = %ComputeLoop5572
  %.lcssa8980 = phi i32 [ %7144, %ComputeLoop5572 ]
  %7149 = trunc i64 %1435 to i32
  %7150 = lshr i64 %1435, 32
  %7151 = trunc i64 %7150 to i32
  %7152 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %7149, i32 0)
  %7153 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %7151, i32 %7152)
  %7154 = icmp eq i32 %7153, 0
  %7155 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %7154)
  %7156 = extractvalue { i1, i64 } %7155, 0
  %7157 = extractvalue { i1, i64 } %7155, 1
  br i1 %7156, label %1436, label %1438

ComputeLoop5576:                                  ; preds = %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit854, %ComputeLoop5576
  %Accumulator5578 = phi i32 [ %7161, %ComputeLoop5576 ], [ -1, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit854 ]
  %ActiveBits5579 = phi i64 [ %7164, %ComputeLoop5576 ], [ %1544, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit854 ]
  %7158 = call i64 @llvm.cttz.i64(i64 %ActiveBits5579, i1 true)
  %7159 = trunc i64 %7158 to i32
  %7160 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i21.i.i845, i32 %7159)
  %7161 = and i32 %Accumulator5578, %7160
  %7162 = shl i64 1, %7158
  %7163 = xor i64 %7162, -1
  %7164 = and i64 %ActiveBits5579, %7163
  %7165 = icmp eq i64 %7164, 0
  br i1 %7165, label %ComputeEnd5577, label %ComputeLoop5576, !amdgpu.uniform !42

ComputeEnd5577:                                   ; preds = %ComputeLoop5576
  %.lcssa8992 = phi i32 [ %7161, %ComputeLoop5576 ]
  %7166 = trunc i64 %1544 to i32
  %7167 = lshr i64 %1544, 32
  %7168 = trunc i64 %7167 to i32
  %7169 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %7166, i32 0)
  %7170 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %7168, i32 %7169)
  %7171 = icmp eq i32 %7170, 0
  %7172 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %7171)
  %7173 = extractvalue { i1, i64 } %7172, 0
  %7174 = extractvalue { i1, i64 } %7172, 1
  br i1 %7173, label %1545, label %1548

ComputeLoop5580:                                  ; preds = %for.cond.i.i1.i863, %ComputeLoop5580
  %Accumulator5582 = phi i32 [ %7179, %ComputeLoop5580 ], [ 0, %for.cond.i.i1.i863 ]
  %OldValuePhi5583 = phi i32 [ %7178, %ComputeLoop5580 ], [ poison, %for.cond.i.i1.i863 ]
  %ActiveBits5584 = phi i64 [ %7182, %ComputeLoop5580 ], [ %1553, %for.cond.i.i1.i863 ]
  %7175 = call i64 @llvm.cttz.i64(i64 %ActiveBits5584, i1 true)
  %7176 = trunc i64 %7175 to i32
  %7177 = call i32 @llvm.amdgcn.readlane.i32(i32 %shl.i.i.i.i.i875, i32 %7176)
  %7178 = call i32 @llvm.amdgcn.writelane.i32(i32 %Accumulator5582, i32 %7176, i32 %OldValuePhi5583)
  %7179 = or i32 %Accumulator5582, %7177
  %7180 = shl i64 1, %7175
  %7181 = xor i64 %7180, -1
  %7182 = and i64 %ActiveBits5584, %7181
  %7183 = icmp eq i64 %7182, 0
  br i1 %7183, label %ComputeEnd5581, label %ComputeLoop5580, !amdgpu.uniform !42

ComputeEnd5581:                                   ; preds = %ComputeLoop5580
  %.lcssa8994 = phi i32 [ %7178, %ComputeLoop5580 ]
  %.lcssa8993 = phi i32 [ %7179, %ComputeLoop5580 ]
  %div8.i.i.i.i.i871 = lshr i32 %spec.store.select.i.i.i870, 5
  %idxprom.i.i.i.i.i872 = zext nneg i32 %div8.i.i.i.i.i871 to i64
  %7184 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i.i.i872
  %7185 = trunc i64 %1553 to i32
  %7186 = lshr i64 %1553, 32
  %7187 = trunc i64 %7186 to i32
  %7188 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %7185, i32 0)
  %7189 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %7187, i32 %7188)
  %7190 = icmp eq i32 %7189, 0
  %7191 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %7190)
  %7192 = extractvalue { i1, i64 } %7191, 0
  %7193 = extractvalue { i1, i64 } %7191, 1
  br i1 %7192, label %1557, label %1559

Flow8404:                                         ; preds = %1577, %if.then.i.i.i.i881
  %7194 = phi i1 [ false, %1577 ], [ true, %if.then.i.i.i.i881 ]
  br i1 %7194, label %if.end14.i.i.i952, label %Flow8405, !amdgpu.uniform !42

ComputeLoop5585:                                  ; preds = %if.then12.i.i.i886, %ComputeLoop5585
  %Accumulator5587 = phi i32 [ %7198, %ComputeLoop5585 ], [ -1, %if.then12.i.i.i886 ]
  %ActiveBits5588 = phi i64 [ %7201, %ComputeLoop5585 ], [ %1574, %if.then12.i.i.i886 ]
  %7195 = call i64 @llvm.cttz.i64(i64 %ActiveBits5588, i1 true)
  %7196 = trunc i64 %7195 to i32
  %7197 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i.i.i892, i32 %7196)
  %7198 = and i32 %Accumulator5587, %7197
  %7199 = shl i64 1, %7195
  %7200 = xor i64 %7199, -1
  %7201 = and i64 %ActiveBits5588, %7200
  %7202 = icmp eq i64 %7201, 0
  br i1 %7202, label %ComputeEnd5586, label %ComputeLoop5585, !amdgpu.uniform !42

ComputeEnd5586:                                   ; preds = %ComputeLoop5585
  %.lcssa8995 = phi i32 [ %7198, %ComputeLoop5585 ]
  %7203 = trunc i64 %1574 to i32
  %7204 = lshr i64 %1574, 32
  %7205 = trunc i64 %7204 to i32
  %7206 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %7203, i32 0)
  %7207 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %7205, i32 %7206)
  %7208 = icmp eq i32 %7207, 0
  %7209 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %7208)
  %7210 = extractvalue { i1, i64 } %7209, 0
  %7211 = extractvalue { i1, i64 } %7209, 1
  br i1 %7210, label %1575, label %1577

ComputeLoop5589:                                  ; preds = %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit960, %ComputeLoop5589
  %Accumulator5591 = phi i32 [ %7215, %ComputeLoop5589 ], [ -1, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit960 ]
  %ActiveBits5592 = phi i64 [ %7218, %ComputeLoop5589 ], [ %1655, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit960 ]
  %7212 = call i64 @llvm.cttz.i64(i64 %ActiveBits5592, i1 true)
  %7213 = trunc i64 %7212 to i32
  %7214 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i21.i.i951, i32 %7213)
  %7215 = and i32 %Accumulator5591, %7214
  %7216 = shl i64 1, %7212
  %7217 = xor i64 %7216, -1
  %7218 = and i64 %ActiveBits5592, %7217
  %7219 = icmp eq i64 %7218, 0
  br i1 %7219, label %ComputeEnd5590, label %ComputeLoop5589, !amdgpu.uniform !42

ComputeEnd5590:                                   ; preds = %ComputeLoop5589
  %.lcssa9003 = phi i32 [ %7215, %ComputeLoop5589 ]
  %7220 = trunc i64 %1655 to i32
  %7221 = lshr i64 %1655, 32
  %7222 = trunc i64 %7221 to i32
  %7223 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %7220, i32 0)
  %7224 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %7222, i32 %7223)
  %7225 = icmp eq i32 %7224, 0
  %7226 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %7225)
  %7227 = extractvalue { i1, i64 } %7226, 0
  %7228 = extractvalue { i1, i64 } %7226, 1
  br i1 %7227, label %1656, label %1659

ComputeLoop5593:                                  ; preds = %for.cond.i.i1.i970, %ComputeLoop5593
  %Accumulator5595 = phi i32 [ %7233, %ComputeLoop5593 ], [ 0, %for.cond.i.i1.i970 ]
  %OldValuePhi5596 = phi i32 [ %7232, %ComputeLoop5593 ], [ poison, %for.cond.i.i1.i970 ]
  %ActiveBits5597 = phi i64 [ %7236, %ComputeLoop5593 ], [ %1667, %for.cond.i.i1.i970 ]
  %7229 = call i64 @llvm.cttz.i64(i64 %ActiveBits5597, i1 true)
  %7230 = trunc i64 %7229 to i32
  %7231 = call i32 @llvm.amdgcn.readlane.i32(i32 %shl.i.i.i.i.i982, i32 %7230)
  %7232 = call i32 @llvm.amdgcn.writelane.i32(i32 %Accumulator5595, i32 %7230, i32 %OldValuePhi5596)
  %7233 = or i32 %Accumulator5595, %7231
  %7234 = shl i64 1, %7229
  %7235 = xor i64 %7234, -1
  %7236 = and i64 %ActiveBits5597, %7235
  %7237 = icmp eq i64 %7236, 0
  br i1 %7237, label %ComputeEnd5594, label %ComputeLoop5593, !amdgpu.uniform !42

ComputeEnd5594:                                   ; preds = %ComputeLoop5593
  %.lcssa9006 = phi i32 [ %7232, %ComputeLoop5593 ]
  %.lcssa9005 = phi i32 [ %7233, %ComputeLoop5593 ]
  %div8.i.i.i.i.i978 = lshr i32 %spec.store.select.i.i.i977, 5
  %idxprom.i.i.i.i.i979 = zext nneg i32 %div8.i.i.i.i.i978 to i64
  %7238 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i.i.i979
  %7239 = trunc i64 %1667 to i32
  %7240 = lshr i64 %1667, 32
  %7241 = trunc i64 %7240 to i32
  %7242 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %7239, i32 0)
  %7243 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %7241, i32 %7242)
  %7244 = icmp eq i32 %7243, 0
  %7245 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %7244)
  %7246 = extractvalue { i1, i64 } %7245, 0
  %7247 = extractvalue { i1, i64 } %7245, 1
  br i1 %7246, label %1671, label %1673

Flow8385:                                         ; preds = %1691, %if.then.i.i.i.i988
  %7248 = phi i1 [ false, %1691 ], [ true, %if.then.i.i.i.i988 ]
  br i1 %7248, label %if.end14.i.i.i1059, label %Flow8386, !amdgpu.uniform !42

ComputeLoop5598:                                  ; preds = %if.then12.i.i.i993, %ComputeLoop5598
  %Accumulator5600 = phi i32 [ %7252, %ComputeLoop5598 ], [ -1, %if.then12.i.i.i993 ]
  %ActiveBits5601 = phi i64 [ %7255, %ComputeLoop5598 ], [ %1688, %if.then12.i.i.i993 ]
  %7249 = call i64 @llvm.cttz.i64(i64 %ActiveBits5601, i1 true)
  %7250 = trunc i64 %7249 to i32
  %7251 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i.i.i999, i32 %7250)
  %7252 = and i32 %Accumulator5600, %7251
  %7253 = shl i64 1, %7249
  %7254 = xor i64 %7253, -1
  %7255 = and i64 %ActiveBits5601, %7254
  %7256 = icmp eq i64 %7255, 0
  br i1 %7256, label %ComputeEnd5599, label %ComputeLoop5598, !amdgpu.uniform !42

ComputeEnd5599:                                   ; preds = %ComputeLoop5598
  %.lcssa9007 = phi i32 [ %7252, %ComputeLoop5598 ]
  %7257 = trunc i64 %1688 to i32
  %7258 = lshr i64 %1688, 32
  %7259 = trunc i64 %7258 to i32
  %7260 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %7257, i32 0)
  %7261 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %7259, i32 %7260)
  %7262 = icmp eq i32 %7261, 0
  %7263 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %7262)
  %7264 = extractvalue { i1, i64 } %7263, 0
  %7265 = extractvalue { i1, i64 } %7263, 1
  br i1 %7264, label %1689, label %1691

ComputeLoop5602:                                  ; preds = %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit1067, %ComputeLoop5602
  %Accumulator5604 = phi i32 [ %7269, %ComputeLoop5602 ], [ -1, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit1067 ]
  %ActiveBits5605 = phi i64 [ %7272, %ComputeLoop5602 ], [ %1795, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit1067 ]
  %7266 = call i64 @llvm.cttz.i64(i64 %ActiveBits5605, i1 true)
  %7267 = trunc i64 %7266 to i32
  %7268 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i21.i.i1058, i32 %7267)
  %7269 = and i32 %Accumulator5604, %7268
  %7270 = shl i64 1, %7266
  %7271 = xor i64 %7270, -1
  %7272 = and i64 %ActiveBits5605, %7271
  %7273 = icmp eq i64 %7272, 0
  br i1 %7273, label %ComputeEnd5603, label %ComputeLoop5602, !amdgpu.uniform !42

ComputeEnd5603:                                   ; preds = %ComputeLoop5602
  %.lcssa9019 = phi i32 [ %7269, %ComputeLoop5602 ]
  %7274 = trunc i64 %1795 to i32
  %7275 = lshr i64 %1795, 32
  %7276 = trunc i64 %7275 to i32
  %7277 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %7274, i32 0)
  %7278 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %7276, i32 %7277)
  %7279 = icmp eq i32 %7278, 0
  %7280 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %7279)
  %7281 = extractvalue { i1, i64 } %7280, 0
  %7282 = extractvalue { i1, i64 } %7280, 1
  br i1 %7281, label %1796, label %1799

ComputeLoop5606:                                  ; preds = %for.cond.i.i1.i1077, %ComputeLoop5606
  %Accumulator5608 = phi i32 [ %7287, %ComputeLoop5606 ], [ 0, %for.cond.i.i1.i1077 ]
  %OldValuePhi5609 = phi i32 [ %7286, %ComputeLoop5606 ], [ poison, %for.cond.i.i1.i1077 ]
  %ActiveBits5610 = phi i64 [ %7290, %ComputeLoop5606 ], [ %1807, %for.cond.i.i1.i1077 ]
  %7283 = call i64 @llvm.cttz.i64(i64 %ActiveBits5610, i1 true)
  %7284 = trunc i64 %7283 to i32
  %7285 = call i32 @llvm.amdgcn.readlane.i32(i32 %shl.i.i.i.i.i1089, i32 %7284)
  %7286 = call i32 @llvm.amdgcn.writelane.i32(i32 %Accumulator5608, i32 %7284, i32 %OldValuePhi5609)
  %7287 = or i32 %Accumulator5608, %7285
  %7288 = shl i64 1, %7283
  %7289 = xor i64 %7288, -1
  %7290 = and i64 %ActiveBits5610, %7289
  %7291 = icmp eq i64 %7290, 0
  br i1 %7291, label %ComputeEnd5607, label %ComputeLoop5606, !amdgpu.uniform !42

ComputeEnd5607:                                   ; preds = %ComputeLoop5606
  %.lcssa9022 = phi i32 [ %7286, %ComputeLoop5606 ]
  %.lcssa9021 = phi i32 [ %7287, %ComputeLoop5606 ]
  %div8.i.i.i.i.i1085 = lshr i32 %spec.store.select.i.i.i1084, 5
  %idxprom.i.i.i.i.i1086 = zext nneg i32 %div8.i.i.i.i.i1085 to i64
  %7292 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i.i.i1086
  %7293 = trunc i64 %1807 to i32
  %7294 = lshr i64 %1807, 32
  %7295 = trunc i64 %7294 to i32
  %7296 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %7293, i32 0)
  %7297 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %7295, i32 %7296)
  %7298 = icmp eq i32 %7297, 0
  %7299 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %7298)
  %7300 = extractvalue { i1, i64 } %7299, 0
  %7301 = extractvalue { i1, i64 } %7299, 1
  br i1 %7300, label %1811, label %1813

Flow8366:                                         ; preds = %1831, %if.then.i.i.i.i1095
  %7302 = phi i1 [ false, %1831 ], [ true, %if.then.i.i.i.i1095 ]
  br i1 %7302, label %if.end14.i.i.i1166, label %Flow8367, !amdgpu.uniform !42

ComputeLoop5611:                                  ; preds = %if.then12.i.i.i1100, %ComputeLoop5611
  %Accumulator5613 = phi i32 [ %7306, %ComputeLoop5611 ], [ -1, %if.then12.i.i.i1100 ]
  %ActiveBits5614 = phi i64 [ %7309, %ComputeLoop5611 ], [ %1828, %if.then12.i.i.i1100 ]
  %7303 = call i64 @llvm.cttz.i64(i64 %ActiveBits5614, i1 true)
  %7304 = trunc i64 %7303 to i32
  %7305 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i.i.i1106, i32 %7304)
  %7306 = and i32 %Accumulator5613, %7305
  %7307 = shl i64 1, %7303
  %7308 = xor i64 %7307, -1
  %7309 = and i64 %ActiveBits5614, %7308
  %7310 = icmp eq i64 %7309, 0
  br i1 %7310, label %ComputeEnd5612, label %ComputeLoop5611, !amdgpu.uniform !42

ComputeEnd5612:                                   ; preds = %ComputeLoop5611
  %.lcssa9023 = phi i32 [ %7306, %ComputeLoop5611 ]
  %7311 = trunc i64 %1828 to i32
  %7312 = lshr i64 %1828, 32
  %7313 = trunc i64 %7312 to i32
  %7314 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %7311, i32 0)
  %7315 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %7313, i32 %7314)
  %7316 = icmp eq i32 %7315, 0
  %7317 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %7316)
  %7318 = extractvalue { i1, i64 } %7317, 0
  %7319 = extractvalue { i1, i64 } %7317, 1
  br i1 %7318, label %1829, label %1831

ComputeLoop5615:                                  ; preds = %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit1174, %ComputeLoop5615
  %Accumulator5617 = phi i32 [ %7323, %ComputeLoop5615 ], [ -1, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit1174 ]
  %ActiveBits5618 = phi i64 [ %7326, %ComputeLoop5615 ], [ %1929, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit1174 ]
  %7320 = call i64 @llvm.cttz.i64(i64 %ActiveBits5618, i1 true)
  %7321 = trunc i64 %7320 to i32
  %7322 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i21.i.i1165, i32 %7321)
  %7323 = and i32 %Accumulator5617, %7322
  %7324 = shl i64 1, %7320
  %7325 = xor i64 %7324, -1
  %7326 = and i64 %ActiveBits5618, %7325
  %7327 = icmp eq i64 %7326, 0
  br i1 %7327, label %ComputeEnd5616, label %ComputeLoop5615, !amdgpu.uniform !42

ComputeEnd5616:                                   ; preds = %ComputeLoop5615
  %.lcssa9035 = phi i32 [ %7323, %ComputeLoop5615 ]
  %7328 = trunc i64 %1929 to i32
  %7329 = lshr i64 %1929, 32
  %7330 = trunc i64 %7329 to i32
  %7331 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %7328, i32 0)
  %7332 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %7330, i32 %7331)
  %7333 = icmp eq i32 %7332, 0
  %7334 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %7333)
  %7335 = extractvalue { i1, i64 } %7334, 0
  %7336 = extractvalue { i1, i64 } %7334, 1
  br i1 %7335, label %1930, label %1933

ComputeLoop5619:                                  ; preds = %for.cond.i.i.i1176, %ComputeLoop5619
  %Accumulator5621 = phi i32 [ %7341, %ComputeLoop5619 ], [ 0, %for.cond.i.i.i1176 ]
  %OldValuePhi5622 = phi i32 [ %7340, %ComputeLoop5619 ], [ poison, %for.cond.i.i.i1176 ]
  %ActiveBits5623 = phi i64 [ %7344, %ComputeLoop5619 ], [ %1938, %for.cond.i.i.i1176 ]
  %7337 = call i64 @llvm.cttz.i64(i64 %ActiveBits5623, i1 true)
  %7338 = trunc i64 %7337 to i32
  %7339 = call i32 @llvm.amdgcn.readlane.i32(i32 %shl.i.i.i.i.i1188, i32 %7338)
  %7340 = call i32 @llvm.amdgcn.writelane.i32(i32 %Accumulator5621, i32 %7338, i32 %OldValuePhi5622)
  %7341 = or i32 %Accumulator5621, %7339
  %7342 = shl i64 1, %7337
  %7343 = xor i64 %7342, -1
  %7344 = and i64 %ActiveBits5623, %7343
  %7345 = icmp eq i64 %7344, 0
  br i1 %7345, label %ComputeEnd5620, label %ComputeLoop5619, !amdgpu.uniform !42

ComputeEnd5620:                                   ; preds = %ComputeLoop5619
  %.lcssa9037 = phi i32 [ %7340, %ComputeLoop5619 ]
  %.lcssa9036 = phi i32 [ %7341, %ComputeLoop5619 ]
  %div8.i.i.i.i.i1184 = lshr i32 %spec.store.select.i.i.i1183, 5
  %idxprom.i.i.i.i.i1185 = zext nneg i32 %div8.i.i.i.i.i1184 to i64
  %7346 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i.i.i1185
  %7347 = trunc i64 %1938 to i32
  %7348 = lshr i64 %1938, 32
  %7349 = trunc i64 %7348 to i32
  %7350 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %7347, i32 0)
  %7351 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %7349, i32 %7350)
  %7352 = icmp eq i32 %7351, 0
  %7353 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %7352)
  %7354 = extractvalue { i1, i64 } %7353, 0
  %7355 = extractvalue { i1, i64 } %7353, 1
  br i1 %7354, label %1942, label %1944

Flow8347:                                         ; preds = %1962, %if.then.i.i.i.i1194
  %7356 = phi i1 [ false, %1962 ], [ true, %if.then.i.i.i.i1194 ]
  br i1 %7356, label %if.end14.i.i.i1263, label %Flow8348, !amdgpu.uniform !42

ComputeLoop5624:                                  ; preds = %if.then12.i.i.i1199, %ComputeLoop5624
  %Accumulator5626 = phi i32 [ %7360, %ComputeLoop5624 ], [ -1, %if.then12.i.i.i1199 ]
  %ActiveBits5627 = phi i64 [ %7363, %ComputeLoop5624 ], [ %1959, %if.then12.i.i.i1199 ]
  %7357 = call i64 @llvm.cttz.i64(i64 %ActiveBits5627, i1 true)
  %7358 = trunc i64 %7357 to i32
  %7359 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i.i.i1205, i32 %7358)
  %7360 = and i32 %Accumulator5626, %7359
  %7361 = shl i64 1, %7357
  %7362 = xor i64 %7361, -1
  %7363 = and i64 %ActiveBits5627, %7362
  %7364 = icmp eq i64 %7363, 0
  br i1 %7364, label %ComputeEnd5625, label %ComputeLoop5624, !amdgpu.uniform !42

ComputeEnd5625:                                   ; preds = %ComputeLoop5624
  %.lcssa9038 = phi i32 [ %7360, %ComputeLoop5624 ]
  %7365 = trunc i64 %1959 to i32
  %7366 = lshr i64 %1959, 32
  %7367 = trunc i64 %7366 to i32
  %7368 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %7365, i32 0)
  %7369 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %7367, i32 %7368)
  %7370 = icmp eq i32 %7369, 0
  %7371 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %7370)
  %7372 = extractvalue { i1, i64 } %7371, 0
  %7373 = extractvalue { i1, i64 } %7371, 1
  br i1 %7372, label %1960, label %1962

ComputeLoop5628:                                  ; preds = %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIcEERS1_T_.exit, %ComputeLoop5628
  %Accumulator5630 = phi i32 [ %7377, %ComputeLoop5628 ], [ -1, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIcEERS1_T_.exit ]
  %ActiveBits5631 = phi i64 [ %7380, %ComputeLoop5628 ], [ %2019, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIcEERS1_T_.exit ]
  %7374 = call i64 @llvm.cttz.i64(i64 %ActiveBits5631, i1 true)
  %7375 = trunc i64 %7374 to i32
  %7376 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i21.i.i1262, i32 %7375)
  %7377 = and i32 %Accumulator5630, %7376
  %7378 = shl i64 1, %7374
  %7379 = xor i64 %7378, -1
  %7380 = and i64 %ActiveBits5631, %7379
  %7381 = icmp eq i64 %7380, 0
  br i1 %7381, label %ComputeEnd5629, label %ComputeLoop5628, !amdgpu.uniform !42

ComputeEnd5629:                                   ; preds = %ComputeLoop5628
  %.lcssa9046 = phi i32 [ %7377, %ComputeLoop5628 ]
  %7382 = trunc i64 %2019 to i32
  %7383 = lshr i64 %2019, 32
  %7384 = trunc i64 %7383 to i32
  %7385 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %7382, i32 0)
  %7386 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %7384, i32 %7385)
  %7387 = icmp eq i32 %7386, 0
  %7388 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %7387)
  %7389 = extractvalue { i1, i64 } %7388, 0
  %7390 = extractvalue { i1, i64 } %7388, 1
  br i1 %7389, label %2020, label %2023

ComputeLoop5632:                                  ; preds = %for.cond.i.i1.i1292, %ComputeLoop5632
  %Accumulator5634 = phi i32 [ %7395, %ComputeLoop5632 ], [ 0, %for.cond.i.i1.i1292 ]
  %OldValuePhi5635 = phi i32 [ %7394, %ComputeLoop5632 ], [ poison, %for.cond.i.i1.i1292 ]
  %ActiveBits5636 = phi i64 [ %7398, %ComputeLoop5632 ], [ %2148, %for.cond.i.i1.i1292 ]
  %7391 = call i64 @llvm.cttz.i64(i64 %ActiveBits5636, i1 true)
  %7392 = trunc i64 %7391 to i32
  %7393 = call i32 @llvm.amdgcn.readlane.i32(i32 %shl.i.i.i.i.i1304, i32 %7392)
  %7394 = call i32 @llvm.amdgcn.writelane.i32(i32 %Accumulator5634, i32 %7392, i32 %OldValuePhi5635)
  %7395 = or i32 %Accumulator5634, %7393
  %7396 = shl i64 1, %7391
  %7397 = xor i64 %7396, -1
  %7398 = and i64 %ActiveBits5636, %7397
  %7399 = icmp eq i64 %7398, 0
  br i1 %7399, label %ComputeEnd5633, label %ComputeLoop5632, !amdgpu.uniform !42

ComputeEnd5633:                                   ; preds = %ComputeLoop5632
  %.lcssa9196 = phi i32 [ %7394, %ComputeLoop5632 ]
  %.lcssa9195 = phi i32 [ %7395, %ComputeLoop5632 ]
  %div8.i.i.i.i.i1300 = lshr i32 %spec.store.select.i.i.i1299, 5
  %idxprom.i.i.i.i.i1301 = zext nneg i32 %div8.i.i.i.i.i1300 to i64
  %7400 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i.i.i1301
  %7401 = trunc i64 %2148 to i32
  %7402 = lshr i64 %2148, 32
  %7403 = trunc i64 %7402 to i32
  %7404 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %7401, i32 0)
  %7405 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %7403, i32 %7404)
  %7406 = icmp eq i32 %7405, 0
  %7407 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %7406)
  %7408 = extractvalue { i1, i64 } %7407, 0
  %7409 = extractvalue { i1, i64 } %7407, 1
  br i1 %7408, label %2152, label %2154

Flow8326:                                         ; preds = %2172, %if.then.i.i.i.i1310
  %7410 = phi i1 [ false, %2172 ], [ true, %if.then.i.i.i.i1310 ]
  br i1 %7410, label %if.end14.i.i.i1381, label %Flow8327, !amdgpu.uniform !42

ComputeLoop5637:                                  ; preds = %if.then12.i.i.i1315, %ComputeLoop5637
  %Accumulator5639 = phi i32 [ %7414, %ComputeLoop5637 ], [ -1, %if.then12.i.i.i1315 ]
  %ActiveBits5640 = phi i64 [ %7417, %ComputeLoop5637 ], [ %2169, %if.then12.i.i.i1315 ]
  %7411 = call i64 @llvm.cttz.i64(i64 %ActiveBits5640, i1 true)
  %7412 = trunc i64 %7411 to i32
  %7413 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i.i.i1321, i32 %7412)
  %7414 = and i32 %Accumulator5639, %7413
  %7415 = shl i64 1, %7411
  %7416 = xor i64 %7415, -1
  %7417 = and i64 %ActiveBits5640, %7416
  %7418 = icmp eq i64 %7417, 0
  br i1 %7418, label %ComputeEnd5638, label %ComputeLoop5637, !amdgpu.uniform !42

ComputeEnd5638:                                   ; preds = %ComputeLoop5637
  %.lcssa9197 = phi i32 [ %7414, %ComputeLoop5637 ]
  %7419 = trunc i64 %2169 to i32
  %7420 = lshr i64 %2169, 32
  %7421 = trunc i64 %7420 to i32
  %7422 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %7419, i32 0)
  %7423 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %7421, i32 %7422)
  %7424 = icmp eq i32 %7423, 0
  %7425 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %7424)
  %7426 = extractvalue { i1, i64 } %7425, 0
  %7427 = extractvalue { i1, i64 } %7425, 1
  br i1 %7426, label %2170, label %2172

ComputeLoop5641:                                  ; preds = %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit1389, %ComputeLoop5641
  %Accumulator5643 = phi i32 [ %7431, %ComputeLoop5641 ], [ -1, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit1389 ]
  %ActiveBits5644 = phi i64 [ %7434, %ComputeLoop5641 ], [ %2276, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit1389 ]
  %7428 = call i64 @llvm.cttz.i64(i64 %ActiveBits5644, i1 true)
  %7429 = trunc i64 %7428 to i32
  %7430 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i21.i.i1380, i32 %7429)
  %7431 = and i32 %Accumulator5643, %7430
  %7432 = shl i64 1, %7428
  %7433 = xor i64 %7432, -1
  %7434 = and i64 %ActiveBits5644, %7433
  %7435 = icmp eq i64 %7434, 0
  br i1 %7435, label %ComputeEnd5642, label %ComputeLoop5641, !amdgpu.uniform !42

ComputeEnd5642:                                   ; preds = %ComputeLoop5641
  %.lcssa9209 = phi i32 [ %7431, %ComputeLoop5641 ]
  %7436 = trunc i64 %2276 to i32
  %7437 = lshr i64 %2276, 32
  %7438 = trunc i64 %7437 to i32
  %7439 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %7436, i32 0)
  %7440 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %7438, i32 %7439)
  %7441 = icmp eq i32 %7440, 0
  %7442 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %7441)
  %7443 = extractvalue { i1, i64 } %7442, 0
  %7444 = extractvalue { i1, i64 } %7442, 1
  br i1 %7443, label %2277, label %2280

ComputeLoop5645:                                  ; preds = %for.cond.i.i1.i1398, %ComputeLoop5645
  %Accumulator5647 = phi i32 [ %7449, %ComputeLoop5645 ], [ 0, %for.cond.i.i1.i1398 ]
  %OldValuePhi5648 = phi i32 [ %7448, %ComputeLoop5645 ], [ poison, %for.cond.i.i1.i1398 ]
  %ActiveBits5649 = phi i64 [ %7452, %ComputeLoop5645 ], [ %2285, %for.cond.i.i1.i1398 ]
  %7445 = call i64 @llvm.cttz.i64(i64 %ActiveBits5649, i1 true)
  %7446 = trunc i64 %7445 to i32
  %7447 = call i32 @llvm.amdgcn.readlane.i32(i32 %shl.i.i.i.i.i1410, i32 %7446)
  %7448 = call i32 @llvm.amdgcn.writelane.i32(i32 %Accumulator5647, i32 %7446, i32 %OldValuePhi5648)
  %7449 = or i32 %Accumulator5647, %7447
  %7450 = shl i64 1, %7445
  %7451 = xor i64 %7450, -1
  %7452 = and i64 %ActiveBits5649, %7451
  %7453 = icmp eq i64 %7452, 0
  br i1 %7453, label %ComputeEnd5646, label %ComputeLoop5645, !amdgpu.uniform !42

ComputeEnd5646:                                   ; preds = %ComputeLoop5645
  %.lcssa9211 = phi i32 [ %7448, %ComputeLoop5645 ]
  %.lcssa9210 = phi i32 [ %7449, %ComputeLoop5645 ]
  %div8.i.i.i.i.i1406 = lshr i32 %spec.store.select.i.i.i1405, 5
  %idxprom.i.i.i.i.i1407 = zext nneg i32 %div8.i.i.i.i.i1406 to i64
  %7454 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i.i.i1407
  %7455 = trunc i64 %2285 to i32
  %7456 = lshr i64 %2285, 32
  %7457 = trunc i64 %7456 to i32
  %7458 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %7455, i32 0)
  %7459 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %7457, i32 %7458)
  %7460 = icmp eq i32 %7459, 0
  %7461 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %7460)
  %7462 = extractvalue { i1, i64 } %7461, 0
  %7463 = extractvalue { i1, i64 } %7461, 1
  br i1 %7462, label %2289, label %2291

Flow8307:                                         ; preds = %2309, %if.then.i.i.i.i1416
  %7464 = phi i1 [ false, %2309 ], [ true, %if.then.i.i.i.i1416 ]
  br i1 %7464, label %if.end14.i.i.i1487, label %Flow8308, !amdgpu.uniform !42

ComputeLoop5650:                                  ; preds = %if.then12.i.i.i1421, %ComputeLoop5650
  %Accumulator5652 = phi i32 [ %7468, %ComputeLoop5650 ], [ -1, %if.then12.i.i.i1421 ]
  %ActiveBits5653 = phi i64 [ %7471, %ComputeLoop5650 ], [ %2306, %if.then12.i.i.i1421 ]
  %7465 = call i64 @llvm.cttz.i64(i64 %ActiveBits5653, i1 true)
  %7466 = trunc i64 %7465 to i32
  %7467 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i.i.i1427, i32 %7466)
  %7468 = and i32 %Accumulator5652, %7467
  %7469 = shl i64 1, %7465
  %7470 = xor i64 %7469, -1
  %7471 = and i64 %ActiveBits5653, %7470
  %7472 = icmp eq i64 %7471, 0
  br i1 %7472, label %ComputeEnd5651, label %ComputeLoop5650, !amdgpu.uniform !42

ComputeEnd5651:                                   ; preds = %ComputeLoop5650
  %.lcssa9212 = phi i32 [ %7468, %ComputeLoop5650 ]
  %7473 = trunc i64 %2306 to i32
  %7474 = lshr i64 %2306, 32
  %7475 = trunc i64 %7474 to i32
  %7476 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %7473, i32 0)
  %7477 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %7475, i32 %7476)
  %7478 = icmp eq i32 %7477, 0
  %7479 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %7478)
  %7480 = extractvalue { i1, i64 } %7479, 0
  %7481 = extractvalue { i1, i64 } %7479, 1
  br i1 %7480, label %2307, label %2309

ComputeLoop5654:                                  ; preds = %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit1495, %ComputeLoop5654
  %Accumulator5656 = phi i32 [ %7485, %ComputeLoop5654 ], [ -1, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit1495 ]
  %ActiveBits5657 = phi i64 [ %7488, %ComputeLoop5654 ], [ %2387, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit1495 ]
  %7482 = call i64 @llvm.cttz.i64(i64 %ActiveBits5657, i1 true)
  %7483 = trunc i64 %7482 to i32
  %7484 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i21.i.i1486, i32 %7483)
  %7485 = and i32 %Accumulator5656, %7484
  %7486 = shl i64 1, %7482
  %7487 = xor i64 %7486, -1
  %7488 = and i64 %ActiveBits5657, %7487
  %7489 = icmp eq i64 %7488, 0
  br i1 %7489, label %ComputeEnd5655, label %ComputeLoop5654, !amdgpu.uniform !42

ComputeEnd5655:                                   ; preds = %ComputeLoop5654
  %.lcssa9220 = phi i32 [ %7485, %ComputeLoop5654 ]
  %7490 = trunc i64 %2387 to i32
  %7491 = lshr i64 %2387, 32
  %7492 = trunc i64 %7491 to i32
  %7493 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %7490, i32 0)
  %7494 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %7492, i32 %7493)
  %7495 = icmp eq i32 %7494, 0
  %7496 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %7495)
  %7497 = extractvalue { i1, i64 } %7496, 0
  %7498 = extractvalue { i1, i64 } %7496, 1
  br i1 %7497, label %2388, label %2391

ComputeLoop5658:                                  ; preds = %for.cond.i.i1.i1505, %ComputeLoop5658
  %Accumulator5660 = phi i32 [ %7503, %ComputeLoop5658 ], [ 0, %for.cond.i.i1.i1505 ]
  %OldValuePhi5661 = phi i32 [ %7502, %ComputeLoop5658 ], [ poison, %for.cond.i.i1.i1505 ]
  %ActiveBits5662 = phi i64 [ %7506, %ComputeLoop5658 ], [ %2399, %for.cond.i.i1.i1505 ]
  %7499 = call i64 @llvm.cttz.i64(i64 %ActiveBits5662, i1 true)
  %7500 = trunc i64 %7499 to i32
  %7501 = call i32 @llvm.amdgcn.readlane.i32(i32 %shl.i.i.i.i.i1517, i32 %7500)
  %7502 = call i32 @llvm.amdgcn.writelane.i32(i32 %Accumulator5660, i32 %7500, i32 %OldValuePhi5661)
  %7503 = or i32 %Accumulator5660, %7501
  %7504 = shl i64 1, %7499
  %7505 = xor i64 %7504, -1
  %7506 = and i64 %ActiveBits5662, %7505
  %7507 = icmp eq i64 %7506, 0
  br i1 %7507, label %ComputeEnd5659, label %ComputeLoop5658, !amdgpu.uniform !42

ComputeEnd5659:                                   ; preds = %ComputeLoop5658
  %.lcssa9223 = phi i32 [ %7502, %ComputeLoop5658 ]
  %.lcssa9222 = phi i32 [ %7503, %ComputeLoop5658 ]
  %div8.i.i.i.i.i1513 = lshr i32 %spec.store.select.i.i.i1512, 5
  %idxprom.i.i.i.i.i1514 = zext nneg i32 %div8.i.i.i.i.i1513 to i64
  %7508 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i.i.i1514
  %7509 = trunc i64 %2399 to i32
  %7510 = lshr i64 %2399, 32
  %7511 = trunc i64 %7510 to i32
  %7512 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %7509, i32 0)
  %7513 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %7511, i32 %7512)
  %7514 = icmp eq i32 %7513, 0
  %7515 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %7514)
  %7516 = extractvalue { i1, i64 } %7515, 0
  %7517 = extractvalue { i1, i64 } %7515, 1
  br i1 %7516, label %2403, label %2405

Flow8288:                                         ; preds = %2423, %if.then.i.i.i.i1523
  %7518 = phi i1 [ false, %2423 ], [ true, %if.then.i.i.i.i1523 ]
  br i1 %7518, label %if.end14.i.i.i1594, label %Flow8289, !amdgpu.uniform !42

ComputeLoop5663:                                  ; preds = %if.then12.i.i.i1528, %ComputeLoop5663
  %Accumulator5665 = phi i32 [ %7522, %ComputeLoop5663 ], [ -1, %if.then12.i.i.i1528 ]
  %ActiveBits5666 = phi i64 [ %7525, %ComputeLoop5663 ], [ %2420, %if.then12.i.i.i1528 ]
  %7519 = call i64 @llvm.cttz.i64(i64 %ActiveBits5666, i1 true)
  %7520 = trunc i64 %7519 to i32
  %7521 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i.i.i1534, i32 %7520)
  %7522 = and i32 %Accumulator5665, %7521
  %7523 = shl i64 1, %7519
  %7524 = xor i64 %7523, -1
  %7525 = and i64 %ActiveBits5666, %7524
  %7526 = icmp eq i64 %7525, 0
  br i1 %7526, label %ComputeEnd5664, label %ComputeLoop5663, !amdgpu.uniform !42

ComputeEnd5664:                                   ; preds = %ComputeLoop5663
  %.lcssa9224 = phi i32 [ %7522, %ComputeLoop5663 ]
  %7527 = trunc i64 %2420 to i32
  %7528 = lshr i64 %2420, 32
  %7529 = trunc i64 %7528 to i32
  %7530 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %7527, i32 0)
  %7531 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %7529, i32 %7530)
  %7532 = icmp eq i32 %7531, 0
  %7533 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %7532)
  %7534 = extractvalue { i1, i64 } %7533, 0
  %7535 = extractvalue { i1, i64 } %7533, 1
  br i1 %7534, label %2421, label %2423

ComputeLoop5667:                                  ; preds = %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit1602, %ComputeLoop5667
  %Accumulator5669 = phi i32 [ %7539, %ComputeLoop5667 ], [ -1, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit1602 ]
  %ActiveBits5670 = phi i64 [ %7542, %ComputeLoop5667 ], [ %2527, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit1602 ]
  %7536 = call i64 @llvm.cttz.i64(i64 %ActiveBits5670, i1 true)
  %7537 = trunc i64 %7536 to i32
  %7538 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i21.i.i1593, i32 %7537)
  %7539 = and i32 %Accumulator5669, %7538
  %7540 = shl i64 1, %7536
  %7541 = xor i64 %7540, -1
  %7542 = and i64 %ActiveBits5670, %7541
  %7543 = icmp eq i64 %7542, 0
  br i1 %7543, label %ComputeEnd5668, label %ComputeLoop5667, !amdgpu.uniform !42

ComputeEnd5668:                                   ; preds = %ComputeLoop5667
  %.lcssa9236 = phi i32 [ %7539, %ComputeLoop5667 ]
  %7544 = trunc i64 %2527 to i32
  %7545 = lshr i64 %2527, 32
  %7546 = trunc i64 %7545 to i32
  %7547 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %7544, i32 0)
  %7548 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %7546, i32 %7547)
  %7549 = icmp eq i32 %7548, 0
  %7550 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %7549)
  %7551 = extractvalue { i1, i64 } %7550, 0
  %7552 = extractvalue { i1, i64 } %7550, 1
  br i1 %7551, label %2528, label %2531

ComputeLoop5671:                                  ; preds = %for.cond.i.i1.i1612, %ComputeLoop5671
  %Accumulator5673 = phi i32 [ %7557, %ComputeLoop5671 ], [ 0, %for.cond.i.i1.i1612 ]
  %OldValuePhi5674 = phi i32 [ %7556, %ComputeLoop5671 ], [ poison, %for.cond.i.i1.i1612 ]
  %ActiveBits5675 = phi i64 [ %7560, %ComputeLoop5671 ], [ %2539, %for.cond.i.i1.i1612 ]
  %7553 = call i64 @llvm.cttz.i64(i64 %ActiveBits5675, i1 true)
  %7554 = trunc i64 %7553 to i32
  %7555 = call i32 @llvm.amdgcn.readlane.i32(i32 %shl.i.i.i.i.i1624, i32 %7554)
  %7556 = call i32 @llvm.amdgcn.writelane.i32(i32 %Accumulator5673, i32 %7554, i32 %OldValuePhi5674)
  %7557 = or i32 %Accumulator5673, %7555
  %7558 = shl i64 1, %7553
  %7559 = xor i64 %7558, -1
  %7560 = and i64 %ActiveBits5675, %7559
  %7561 = icmp eq i64 %7560, 0
  br i1 %7561, label %ComputeEnd5672, label %ComputeLoop5671, !amdgpu.uniform !42

ComputeEnd5672:                                   ; preds = %ComputeLoop5671
  %.lcssa9239 = phi i32 [ %7556, %ComputeLoop5671 ]
  %.lcssa9238 = phi i32 [ %7557, %ComputeLoop5671 ]
  %div8.i.i.i.i.i1620 = lshr i32 %spec.store.select.i.i.i1619, 5
  %idxprom.i.i.i.i.i1621 = zext nneg i32 %div8.i.i.i.i.i1620 to i64
  %7562 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i.i.i1621
  %7563 = trunc i64 %2539 to i32
  %7564 = lshr i64 %2539, 32
  %7565 = trunc i64 %7564 to i32
  %7566 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %7563, i32 0)
  %7567 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %7565, i32 %7566)
  %7568 = icmp eq i32 %7567, 0
  %7569 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %7568)
  %7570 = extractvalue { i1, i64 } %7569, 0
  %7571 = extractvalue { i1, i64 } %7569, 1
  br i1 %7570, label %2543, label %2545

Flow8269:                                         ; preds = %2563, %if.then.i.i.i.i1630
  %7572 = phi i1 [ false, %2563 ], [ true, %if.then.i.i.i.i1630 ]
  br i1 %7572, label %if.end14.i.i.i1701, label %Flow8270, !amdgpu.uniform !42

ComputeLoop5676:                                  ; preds = %if.then12.i.i.i1635, %ComputeLoop5676
  %Accumulator5678 = phi i32 [ %7576, %ComputeLoop5676 ], [ -1, %if.then12.i.i.i1635 ]
  %ActiveBits5679 = phi i64 [ %7579, %ComputeLoop5676 ], [ %2560, %if.then12.i.i.i1635 ]
  %7573 = call i64 @llvm.cttz.i64(i64 %ActiveBits5679, i1 true)
  %7574 = trunc i64 %7573 to i32
  %7575 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i.i.i1641, i32 %7574)
  %7576 = and i32 %Accumulator5678, %7575
  %7577 = shl i64 1, %7573
  %7578 = xor i64 %7577, -1
  %7579 = and i64 %ActiveBits5679, %7578
  %7580 = icmp eq i64 %7579, 0
  br i1 %7580, label %ComputeEnd5677, label %ComputeLoop5676, !amdgpu.uniform !42

ComputeEnd5677:                                   ; preds = %ComputeLoop5676
  %.lcssa9240 = phi i32 [ %7576, %ComputeLoop5676 ]
  %7581 = trunc i64 %2560 to i32
  %7582 = lshr i64 %2560, 32
  %7583 = trunc i64 %7582 to i32
  %7584 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %7581, i32 0)
  %7585 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %7583, i32 %7584)
  %7586 = icmp eq i32 %7585, 0
  %7587 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %7586)
  %7588 = extractvalue { i1, i64 } %7587, 0
  %7589 = extractvalue { i1, i64 } %7587, 1
  br i1 %7588, label %2561, label %2563

ComputeLoop5680:                                  ; preds = %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit1709, %ComputeLoop5680
  %Accumulator5682 = phi i32 [ %7593, %ComputeLoop5680 ], [ -1, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit1709 ]
  %ActiveBits5683 = phi i64 [ %7596, %ComputeLoop5680 ], [ %2662, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit1709 ]
  %7590 = call i64 @llvm.cttz.i64(i64 %ActiveBits5683, i1 true)
  %7591 = trunc i64 %7590 to i32
  %7592 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i21.i.i1700, i32 %7591)
  %7593 = and i32 %Accumulator5682, %7592
  %7594 = shl i64 1, %7590
  %7595 = xor i64 %7594, -1
  %7596 = and i64 %ActiveBits5683, %7595
  %7597 = icmp eq i64 %7596, 0
  br i1 %7597, label %ComputeEnd5681, label %ComputeLoop5680, !amdgpu.uniform !42

ComputeEnd5681:                                   ; preds = %ComputeLoop5680
  %.lcssa9252 = phi i32 [ %7593, %ComputeLoop5680 ]
  %7598 = trunc i64 %2662 to i32
  %7599 = lshr i64 %2662, 32
  %7600 = trunc i64 %7599 to i32
  %7601 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %7598, i32 0)
  %7602 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %7600, i32 %7601)
  %7603 = icmp eq i32 %7602, 0
  %7604 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %7603)
  %7605 = extractvalue { i1, i64 } %7604, 0
  %7606 = extractvalue { i1, i64 } %7604, 1
  br i1 %7605, label %2663, label %2666

ComputeLoop5684:                                  ; preds = %for.cond.i.i.i1713, %ComputeLoop5684
  %Accumulator5686 = phi i32 [ %7611, %ComputeLoop5684 ], [ 0, %for.cond.i.i.i1713 ]
  %OldValuePhi5687 = phi i32 [ %7610, %ComputeLoop5684 ], [ poison, %for.cond.i.i.i1713 ]
  %ActiveBits5688 = phi i64 [ %7614, %ComputeLoop5684 ], [ %2671, %for.cond.i.i.i1713 ]
  %7607 = call i64 @llvm.cttz.i64(i64 %ActiveBits5688, i1 true)
  %7608 = trunc i64 %7607 to i32
  %7609 = call i32 @llvm.amdgcn.readlane.i32(i32 %shl.i.i.i.i.i1725, i32 %7608)
  %7610 = call i32 @llvm.amdgcn.writelane.i32(i32 %Accumulator5686, i32 %7608, i32 %OldValuePhi5687)
  %7611 = or i32 %Accumulator5686, %7609
  %7612 = shl i64 1, %7607
  %7613 = xor i64 %7612, -1
  %7614 = and i64 %ActiveBits5688, %7613
  %7615 = icmp eq i64 %7614, 0
  br i1 %7615, label %ComputeEnd5685, label %ComputeLoop5684, !amdgpu.uniform !42

ComputeEnd5685:                                   ; preds = %ComputeLoop5684
  %.lcssa9254 = phi i32 [ %7610, %ComputeLoop5684 ]
  %.lcssa9253 = phi i32 [ %7611, %ComputeLoop5684 ]
  %div8.i.i.i.i.i1721 = lshr i32 %spec.store.select.i.i.i1720, 5
  %idxprom.i.i.i.i.i1722 = zext nneg i32 %div8.i.i.i.i.i1721 to i64
  %7616 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i.i.i1722
  %7617 = trunc i64 %2671 to i32
  %7618 = lshr i64 %2671, 32
  %7619 = trunc i64 %7618 to i32
  %7620 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %7617, i32 0)
  %7621 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %7619, i32 %7620)
  %7622 = icmp eq i32 %7621, 0
  %7623 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %7622)
  %7624 = extractvalue { i1, i64 } %7623, 0
  %7625 = extractvalue { i1, i64 } %7623, 1
  br i1 %7624, label %2675, label %2677

Flow8250:                                         ; preds = %2695, %if.then.i.i.i.i1731
  %7626 = phi i1 [ false, %2695 ], [ true, %if.then.i.i.i.i1731 ]
  br i1 %7626, label %if.end14.i.i.i1800, label %Flow8251, !amdgpu.uniform !42

ComputeLoop5689:                                  ; preds = %if.then12.i.i.i1736, %ComputeLoop5689
  %Accumulator5691 = phi i32 [ %7630, %ComputeLoop5689 ], [ -1, %if.then12.i.i.i1736 ]
  %ActiveBits5692 = phi i64 [ %7633, %ComputeLoop5689 ], [ %2692, %if.then12.i.i.i1736 ]
  %7627 = call i64 @llvm.cttz.i64(i64 %ActiveBits5692, i1 true)
  %7628 = trunc i64 %7627 to i32
  %7629 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i.i.i1742, i32 %7628)
  %7630 = and i32 %Accumulator5691, %7629
  %7631 = shl i64 1, %7627
  %7632 = xor i64 %7631, -1
  %7633 = and i64 %ActiveBits5692, %7632
  %7634 = icmp eq i64 %7633, 0
  br i1 %7634, label %ComputeEnd5690, label %ComputeLoop5689, !amdgpu.uniform !42

ComputeEnd5690:                                   ; preds = %ComputeLoop5689
  %.lcssa9255 = phi i32 [ %7630, %ComputeLoop5689 ]
  %7635 = trunc i64 %2692 to i32
  %7636 = lshr i64 %2692, 32
  %7637 = trunc i64 %7636 to i32
  %7638 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %7635, i32 0)
  %7639 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %7637, i32 %7638)
  %7640 = icmp eq i32 %7639, 0
  %7641 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %7640)
  %7642 = extractvalue { i1, i64 } %7641, 0
  %7643 = extractvalue { i1, i64 } %7641, 1
  br i1 %7642, label %2693, label %2695

ComputeLoop5693:                                  ; preds = %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIcEERS1_T_.exit1808, %ComputeLoop5693
  %Accumulator5695 = phi i32 [ %7647, %ComputeLoop5693 ], [ -1, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIcEERS1_T_.exit1808 ]
  %ActiveBits5696 = phi i64 [ %7650, %ComputeLoop5693 ], [ %2752, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIcEERS1_T_.exit1808 ]
  %7644 = call i64 @llvm.cttz.i64(i64 %ActiveBits5696, i1 true)
  %7645 = trunc i64 %7644 to i32
  %7646 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i21.i.i1799, i32 %7645)
  %7647 = and i32 %Accumulator5695, %7646
  %7648 = shl i64 1, %7644
  %7649 = xor i64 %7648, -1
  %7650 = and i64 %ActiveBits5696, %7649
  %7651 = icmp eq i64 %7650, 0
  br i1 %7651, label %ComputeEnd5694, label %ComputeLoop5693, !amdgpu.uniform !42

ComputeEnd5694:                                   ; preds = %ComputeLoop5693
  %.lcssa9263 = phi i32 [ %7647, %ComputeLoop5693 ]
  %7652 = trunc i64 %2752 to i32
  %7653 = lshr i64 %2752, 32
  %7654 = trunc i64 %7653 to i32
  %7655 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %7652, i32 0)
  %7656 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %7654, i32 %7655)
  %7657 = icmp eq i32 %7656, 0
  %7658 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %7657)
  %7659 = extractvalue { i1, i64 } %7658, 0
  %7660 = extractvalue { i1, i64 } %7658, 1
  br i1 %7659, label %2753, label %2756

ComputeLoop5697:                                  ; preds = %for.cond.i.i1.i1818, %ComputeLoop5697
  %Accumulator5699 = phi i32 [ %7665, %ComputeLoop5697 ], [ 0, %for.cond.i.i1.i1818 ]
  %OldValuePhi5700 = phi i32 [ %7664, %ComputeLoop5697 ], [ poison, %for.cond.i.i1.i1818 ]
  %ActiveBits5701 = phi i64 [ %7668, %ComputeLoop5697 ], [ %2803, %for.cond.i.i1.i1818 ]
  %7661 = call i64 @llvm.cttz.i64(i64 %ActiveBits5701, i1 true)
  %7662 = trunc i64 %7661 to i32
  %7663 = call i32 @llvm.amdgcn.readlane.i32(i32 %shl.i.i.i.i.i1830, i32 %7662)
  %7664 = call i32 @llvm.amdgcn.writelane.i32(i32 %Accumulator5699, i32 %7662, i32 %OldValuePhi5700)
  %7665 = or i32 %Accumulator5699, %7663
  %7666 = shl i64 1, %7661
  %7667 = xor i64 %7666, -1
  %7668 = and i64 %ActiveBits5701, %7667
  %7669 = icmp eq i64 %7668, 0
  br i1 %7669, label %ComputeEnd5698, label %ComputeLoop5697, !amdgpu.uniform !42

ComputeEnd5698:                                   ; preds = %ComputeLoop5697
  %.lcssa9049 = phi i32 [ %7664, %ComputeLoop5697 ]
  %.lcssa9048 = phi i32 [ %7665, %ComputeLoop5697 ]
  %div8.i.i.i.i.i1826 = lshr i32 %spec.store.select.i.i.i1825, 5
  %idxprom.i.i.i.i.i1827 = zext nneg i32 %div8.i.i.i.i.i1826 to i64
  %7670 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i.i.i1827
  %7671 = trunc i64 %2803 to i32
  %7672 = lshr i64 %2803, 32
  %7673 = trunc i64 %7672 to i32
  %7674 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %7671, i32 0)
  %7675 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %7673, i32 %7674)
  %7676 = icmp eq i32 %7675, 0
  %7677 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %7676)
  %7678 = extractvalue { i1, i64 } %7677, 0
  %7679 = extractvalue { i1, i64 } %7677, 1
  br i1 %7678, label %2807, label %2809

Flow8241:                                         ; preds = %2827, %if.then.i.i.i.i1836
  %7680 = phi i1 [ false, %2827 ], [ true, %if.then.i.i.i.i1836 ]
  br i1 %7680, label %if.end14.i.i.i1907, label %Flow8242, !amdgpu.uniform !42

ComputeLoop5702:                                  ; preds = %if.then12.i.i.i1841, %ComputeLoop5702
  %Accumulator5704 = phi i32 [ %7684, %ComputeLoop5702 ], [ -1, %if.then12.i.i.i1841 ]
  %ActiveBits5705 = phi i64 [ %7687, %ComputeLoop5702 ], [ %2824, %if.then12.i.i.i1841 ]
  %7681 = call i64 @llvm.cttz.i64(i64 %ActiveBits5705, i1 true)
  %7682 = trunc i64 %7681 to i32
  %7683 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i.i.i1847, i32 %7682)
  %7684 = and i32 %Accumulator5704, %7683
  %7685 = shl i64 1, %7681
  %7686 = xor i64 %7685, -1
  %7687 = and i64 %ActiveBits5705, %7686
  %7688 = icmp eq i64 %7687, 0
  br i1 %7688, label %ComputeEnd5703, label %ComputeLoop5702, !amdgpu.uniform !42

ComputeEnd5703:                                   ; preds = %ComputeLoop5702
  %.lcssa9050 = phi i32 [ %7684, %ComputeLoop5702 ]
  %7689 = trunc i64 %2824 to i32
  %7690 = lshr i64 %2824, 32
  %7691 = trunc i64 %7690 to i32
  %7692 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %7689, i32 0)
  %7693 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %7691, i32 %7692)
  %7694 = icmp eq i32 %7693, 0
  %7695 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %7694)
  %7696 = extractvalue { i1, i64 } %7695, 0
  %7697 = extractvalue { i1, i64 } %7695, 1
  br i1 %7696, label %2825, label %2827

ComputeLoop5706:                                  ; preds = %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit1915, %ComputeLoop5706
  %Accumulator5708 = phi i32 [ %7701, %ComputeLoop5706 ], [ -1, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit1915 ]
  %ActiveBits5709 = phi i64 [ %7704, %ComputeLoop5706 ], [ %2931, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit1915 ]
  %7698 = call i64 @llvm.cttz.i64(i64 %ActiveBits5709, i1 true)
  %7699 = trunc i64 %7698 to i32
  %7700 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i21.i.i1906, i32 %7699)
  %7701 = and i32 %Accumulator5708, %7700
  %7702 = shl i64 1, %7698
  %7703 = xor i64 %7702, -1
  %7704 = and i64 %ActiveBits5709, %7703
  %7705 = icmp eq i64 %7704, 0
  br i1 %7705, label %ComputeEnd5707, label %ComputeLoop5706, !amdgpu.uniform !42

ComputeEnd5707:                                   ; preds = %ComputeLoop5706
  %.lcssa9062 = phi i32 [ %7701, %ComputeLoop5706 ]
  %7706 = trunc i64 %2931 to i32
  %7707 = lshr i64 %2931, 32
  %7708 = trunc i64 %7707 to i32
  %7709 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %7706, i32 0)
  %7710 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %7708, i32 %7709)
  %7711 = icmp eq i32 %7710, 0
  %7712 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %7711)
  %7713 = extractvalue { i1, i64 } %7712, 0
  %7714 = extractvalue { i1, i64 } %7712, 1
  br i1 %7713, label %2932, label %2935

ComputeLoop5710:                                  ; preds = %for.cond.i.i1.i1924, %ComputeLoop5710
  %Accumulator5712 = phi i32 [ %7719, %ComputeLoop5710 ], [ 0, %for.cond.i.i1.i1924 ]
  %OldValuePhi5713 = phi i32 [ %7718, %ComputeLoop5710 ], [ poison, %for.cond.i.i1.i1924 ]
  %ActiveBits5714 = phi i64 [ %7722, %ComputeLoop5710 ], [ %2940, %for.cond.i.i1.i1924 ]
  %7715 = call i64 @llvm.cttz.i64(i64 %ActiveBits5714, i1 true)
  %7716 = trunc i64 %7715 to i32
  %7717 = call i32 @llvm.amdgcn.readlane.i32(i32 %shl.i.i.i.i.i1936, i32 %7716)
  %7718 = call i32 @llvm.amdgcn.writelane.i32(i32 %Accumulator5712, i32 %7716, i32 %OldValuePhi5713)
  %7719 = or i32 %Accumulator5712, %7717
  %7720 = shl i64 1, %7715
  %7721 = xor i64 %7720, -1
  %7722 = and i64 %ActiveBits5714, %7721
  %7723 = icmp eq i64 %7722, 0
  br i1 %7723, label %ComputeEnd5711, label %ComputeLoop5710, !amdgpu.uniform !42

ComputeEnd5711:                                   ; preds = %ComputeLoop5710
  %.lcssa9064 = phi i32 [ %7718, %ComputeLoop5710 ]
  %.lcssa9063 = phi i32 [ %7719, %ComputeLoop5710 ]
  %div8.i.i.i.i.i1932 = lshr i32 %spec.store.select.i.i.i1931, 5
  %idxprom.i.i.i.i.i1933 = zext nneg i32 %div8.i.i.i.i.i1932 to i64
  %7724 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i.i.i1933
  %7725 = trunc i64 %2940 to i32
  %7726 = lshr i64 %2940, 32
  %7727 = trunc i64 %7726 to i32
  %7728 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %7725, i32 0)
  %7729 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %7727, i32 %7728)
  %7730 = icmp eq i32 %7729, 0
  %7731 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %7730)
  %7732 = extractvalue { i1, i64 } %7731, 0
  %7733 = extractvalue { i1, i64 } %7731, 1
  br i1 %7732, label %2944, label %2946

Flow8222:                                         ; preds = %2964, %if.then.i.i.i.i1942
  %7734 = phi i1 [ false, %2964 ], [ true, %if.then.i.i.i.i1942 ]
  br i1 %7734, label %if.end14.i.i.i2013, label %Flow8223, !amdgpu.uniform !42

ComputeLoop5715:                                  ; preds = %if.then12.i.i.i1947, %ComputeLoop5715
  %Accumulator5717 = phi i32 [ %7738, %ComputeLoop5715 ], [ -1, %if.then12.i.i.i1947 ]
  %ActiveBits5718 = phi i64 [ %7741, %ComputeLoop5715 ], [ %2961, %if.then12.i.i.i1947 ]
  %7735 = call i64 @llvm.cttz.i64(i64 %ActiveBits5718, i1 true)
  %7736 = trunc i64 %7735 to i32
  %7737 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i.i.i1953, i32 %7736)
  %7738 = and i32 %Accumulator5717, %7737
  %7739 = shl i64 1, %7735
  %7740 = xor i64 %7739, -1
  %7741 = and i64 %ActiveBits5718, %7740
  %7742 = icmp eq i64 %7741, 0
  br i1 %7742, label %ComputeEnd5716, label %ComputeLoop5715, !amdgpu.uniform !42

ComputeEnd5716:                                   ; preds = %ComputeLoop5715
  %.lcssa9065 = phi i32 [ %7738, %ComputeLoop5715 ]
  %7743 = trunc i64 %2961 to i32
  %7744 = lshr i64 %2961, 32
  %7745 = trunc i64 %7744 to i32
  %7746 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %7743, i32 0)
  %7747 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %7745, i32 %7746)
  %7748 = icmp eq i32 %7747, 0
  %7749 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %7748)
  %7750 = extractvalue { i1, i64 } %7749, 0
  %7751 = extractvalue { i1, i64 } %7749, 1
  br i1 %7750, label %2962, label %2964

ComputeLoop5719:                                  ; preds = %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit2021, %ComputeLoop5719
  %Accumulator5721 = phi i32 [ %7755, %ComputeLoop5719 ], [ -1, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit2021 ]
  %ActiveBits5722 = phi i64 [ %7758, %ComputeLoop5719 ], [ %3042, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit2021 ]
  %7752 = call i64 @llvm.cttz.i64(i64 %ActiveBits5722, i1 true)
  %7753 = trunc i64 %7752 to i32
  %7754 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i21.i.i2012, i32 %7753)
  %7755 = and i32 %Accumulator5721, %7754
  %7756 = shl i64 1, %7752
  %7757 = xor i64 %7756, -1
  %7758 = and i64 %ActiveBits5722, %7757
  %7759 = icmp eq i64 %7758, 0
  br i1 %7759, label %ComputeEnd5720, label %ComputeLoop5719, !amdgpu.uniform !42

ComputeEnd5720:                                   ; preds = %ComputeLoop5719
  %.lcssa9073 = phi i32 [ %7755, %ComputeLoop5719 ]
  %7760 = trunc i64 %3042 to i32
  %7761 = lshr i64 %3042, 32
  %7762 = trunc i64 %7761 to i32
  %7763 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %7760, i32 0)
  %7764 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %7762, i32 %7763)
  %7765 = icmp eq i32 %7764, 0
  %7766 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %7765)
  %7767 = extractvalue { i1, i64 } %7766, 0
  %7768 = extractvalue { i1, i64 } %7766, 1
  br i1 %7767, label %3043, label %3046

ComputeLoop5723:                                  ; preds = %for.cond.i.i1.i2031, %ComputeLoop5723
  %Accumulator5725 = phi i32 [ %7773, %ComputeLoop5723 ], [ 0, %for.cond.i.i1.i2031 ]
  %OldValuePhi5726 = phi i32 [ %7772, %ComputeLoop5723 ], [ poison, %for.cond.i.i1.i2031 ]
  %ActiveBits5727 = phi i64 [ %7776, %ComputeLoop5723 ], [ %3054, %for.cond.i.i1.i2031 ]
  %7769 = call i64 @llvm.cttz.i64(i64 %ActiveBits5727, i1 true)
  %7770 = trunc i64 %7769 to i32
  %7771 = call i32 @llvm.amdgcn.readlane.i32(i32 %shl.i.i.i.i.i2043, i32 %7770)
  %7772 = call i32 @llvm.amdgcn.writelane.i32(i32 %Accumulator5725, i32 %7770, i32 %OldValuePhi5726)
  %7773 = or i32 %Accumulator5725, %7771
  %7774 = shl i64 1, %7769
  %7775 = xor i64 %7774, -1
  %7776 = and i64 %ActiveBits5727, %7775
  %7777 = icmp eq i64 %7776, 0
  br i1 %7777, label %ComputeEnd5724, label %ComputeLoop5723, !amdgpu.uniform !42

ComputeEnd5724:                                   ; preds = %ComputeLoop5723
  %.lcssa9076 = phi i32 [ %7772, %ComputeLoop5723 ]
  %.lcssa9075 = phi i32 [ %7773, %ComputeLoop5723 ]
  %div8.i.i.i.i.i2039 = lshr i32 %spec.store.select.i.i.i2038, 5
  %idxprom.i.i.i.i.i2040 = zext nneg i32 %div8.i.i.i.i.i2039 to i64
  %7778 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i.i.i2040
  %7779 = trunc i64 %3054 to i32
  %7780 = lshr i64 %3054, 32
  %7781 = trunc i64 %7780 to i32
  %7782 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %7779, i32 0)
  %7783 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %7781, i32 %7782)
  %7784 = icmp eq i32 %7783, 0
  %7785 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %7784)
  %7786 = extractvalue { i1, i64 } %7785, 0
  %7787 = extractvalue { i1, i64 } %7785, 1
  br i1 %7786, label %3058, label %3060

Flow8203:                                         ; preds = %3078, %if.then.i.i.i.i2049
  %7788 = phi i1 [ false, %3078 ], [ true, %if.then.i.i.i.i2049 ]
  br i1 %7788, label %if.end14.i.i.i2120, label %Flow8204, !amdgpu.uniform !42

ComputeLoop5728:                                  ; preds = %if.then12.i.i.i2054, %ComputeLoop5728
  %Accumulator5730 = phi i32 [ %7792, %ComputeLoop5728 ], [ -1, %if.then12.i.i.i2054 ]
  %ActiveBits5731 = phi i64 [ %7795, %ComputeLoop5728 ], [ %3075, %if.then12.i.i.i2054 ]
  %7789 = call i64 @llvm.cttz.i64(i64 %ActiveBits5731, i1 true)
  %7790 = trunc i64 %7789 to i32
  %7791 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i.i.i2060, i32 %7790)
  %7792 = and i32 %Accumulator5730, %7791
  %7793 = shl i64 1, %7789
  %7794 = xor i64 %7793, -1
  %7795 = and i64 %ActiveBits5731, %7794
  %7796 = icmp eq i64 %7795, 0
  br i1 %7796, label %ComputeEnd5729, label %ComputeLoop5728, !amdgpu.uniform !42

ComputeEnd5729:                                   ; preds = %ComputeLoop5728
  %.lcssa9077 = phi i32 [ %7792, %ComputeLoop5728 ]
  %7797 = trunc i64 %3075 to i32
  %7798 = lshr i64 %3075, 32
  %7799 = trunc i64 %7798 to i32
  %7800 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %7797, i32 0)
  %7801 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %7799, i32 %7800)
  %7802 = icmp eq i32 %7801, 0
  %7803 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %7802)
  %7804 = extractvalue { i1, i64 } %7803, 0
  %7805 = extractvalue { i1, i64 } %7803, 1
  br i1 %7804, label %3076, label %3078

ComputeLoop5732:                                  ; preds = %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit2128, %ComputeLoop5732
  %Accumulator5734 = phi i32 [ %7809, %ComputeLoop5732 ], [ -1, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit2128 ]
  %ActiveBits5735 = phi i64 [ %7812, %ComputeLoop5732 ], [ %3182, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit2128 ]
  %7806 = call i64 @llvm.cttz.i64(i64 %ActiveBits5735, i1 true)
  %7807 = trunc i64 %7806 to i32
  %7808 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i21.i.i2119, i32 %7807)
  %7809 = and i32 %Accumulator5734, %7808
  %7810 = shl i64 1, %7806
  %7811 = xor i64 %7810, -1
  %7812 = and i64 %ActiveBits5735, %7811
  %7813 = icmp eq i64 %7812, 0
  br i1 %7813, label %ComputeEnd5733, label %ComputeLoop5732, !amdgpu.uniform !42

ComputeEnd5733:                                   ; preds = %ComputeLoop5732
  %.lcssa9089 = phi i32 [ %7809, %ComputeLoop5732 ]
  %7814 = trunc i64 %3182 to i32
  %7815 = lshr i64 %3182, 32
  %7816 = trunc i64 %7815 to i32
  %7817 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %7814, i32 0)
  %7818 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %7816, i32 %7817)
  %7819 = icmp eq i32 %7818, 0
  %7820 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %7819)
  %7821 = extractvalue { i1, i64 } %7820, 0
  %7822 = extractvalue { i1, i64 } %7820, 1
  br i1 %7821, label %3183, label %3186

ComputeLoop5736:                                  ; preds = %for.cond.i.i1.i2138, %ComputeLoop5736
  %Accumulator5738 = phi i32 [ %7827, %ComputeLoop5736 ], [ 0, %for.cond.i.i1.i2138 ]
  %OldValuePhi5739 = phi i32 [ %7826, %ComputeLoop5736 ], [ poison, %for.cond.i.i1.i2138 ]
  %ActiveBits5740 = phi i64 [ %7830, %ComputeLoop5736 ], [ %3194, %for.cond.i.i1.i2138 ]
  %7823 = call i64 @llvm.cttz.i64(i64 %ActiveBits5740, i1 true)
  %7824 = trunc i64 %7823 to i32
  %7825 = call i32 @llvm.amdgcn.readlane.i32(i32 %shl.i.i.i.i.i2150, i32 %7824)
  %7826 = call i32 @llvm.amdgcn.writelane.i32(i32 %Accumulator5738, i32 %7824, i32 %OldValuePhi5739)
  %7827 = or i32 %Accumulator5738, %7825
  %7828 = shl i64 1, %7823
  %7829 = xor i64 %7828, -1
  %7830 = and i64 %ActiveBits5740, %7829
  %7831 = icmp eq i64 %7830, 0
  br i1 %7831, label %ComputeEnd5737, label %ComputeLoop5736, !amdgpu.uniform !42

ComputeEnd5737:                                   ; preds = %ComputeLoop5736
  %.lcssa9092 = phi i32 [ %7826, %ComputeLoop5736 ]
  %.lcssa9091 = phi i32 [ %7827, %ComputeLoop5736 ]
  %div8.i.i.i.i.i2146 = lshr i32 %spec.store.select.i.i.i2145, 5
  %idxprom.i.i.i.i.i2147 = zext nneg i32 %div8.i.i.i.i.i2146 to i64
  %7832 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i.i.i2147
  %7833 = trunc i64 %3194 to i32
  %7834 = lshr i64 %3194, 32
  %7835 = trunc i64 %7834 to i32
  %7836 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %7833, i32 0)
  %7837 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %7835, i32 %7836)
  %7838 = icmp eq i32 %7837, 0
  %7839 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %7838)
  %7840 = extractvalue { i1, i64 } %7839, 0
  %7841 = extractvalue { i1, i64 } %7839, 1
  br i1 %7840, label %3198, label %3200

Flow8184:                                         ; preds = %3218, %if.then.i.i.i.i2156
  %7842 = phi i1 [ false, %3218 ], [ true, %if.then.i.i.i.i2156 ]
  br i1 %7842, label %if.end14.i.i.i2227, label %Flow8185, !amdgpu.uniform !42

ComputeLoop5741:                                  ; preds = %if.then12.i.i.i2161, %ComputeLoop5741
  %Accumulator5743 = phi i32 [ %7846, %ComputeLoop5741 ], [ -1, %if.then12.i.i.i2161 ]
  %ActiveBits5744 = phi i64 [ %7849, %ComputeLoop5741 ], [ %3215, %if.then12.i.i.i2161 ]
  %7843 = call i64 @llvm.cttz.i64(i64 %ActiveBits5744, i1 true)
  %7844 = trunc i64 %7843 to i32
  %7845 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i.i.i2167, i32 %7844)
  %7846 = and i32 %Accumulator5743, %7845
  %7847 = shl i64 1, %7843
  %7848 = xor i64 %7847, -1
  %7849 = and i64 %ActiveBits5744, %7848
  %7850 = icmp eq i64 %7849, 0
  br i1 %7850, label %ComputeEnd5742, label %ComputeLoop5741, !amdgpu.uniform !42

ComputeEnd5742:                                   ; preds = %ComputeLoop5741
  %.lcssa9093 = phi i32 [ %7846, %ComputeLoop5741 ]
  %7851 = trunc i64 %3215 to i32
  %7852 = lshr i64 %3215, 32
  %7853 = trunc i64 %7852 to i32
  %7854 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %7851, i32 0)
  %7855 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %7853, i32 %7854)
  %7856 = icmp eq i32 %7855, 0
  %7857 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %7856)
  %7858 = extractvalue { i1, i64 } %7857, 0
  %7859 = extractvalue { i1, i64 } %7857, 1
  br i1 %7858, label %3216, label %3218

ComputeLoop5745:                                  ; preds = %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit2235, %ComputeLoop5745
  %Accumulator5747 = phi i32 [ %7863, %ComputeLoop5745 ], [ -1, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit2235 ]
  %ActiveBits5748 = phi i64 [ %7866, %ComputeLoop5745 ], [ %3317, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit2235 ]
  %7860 = call i64 @llvm.cttz.i64(i64 %ActiveBits5748, i1 true)
  %7861 = trunc i64 %7860 to i32
  %7862 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i21.i.i2226, i32 %7861)
  %7863 = and i32 %Accumulator5747, %7862
  %7864 = shl i64 1, %7860
  %7865 = xor i64 %7864, -1
  %7866 = and i64 %ActiveBits5748, %7865
  %7867 = icmp eq i64 %7866, 0
  br i1 %7867, label %ComputeEnd5746, label %ComputeLoop5745, !amdgpu.uniform !42

ComputeEnd5746:                                   ; preds = %ComputeLoop5745
  %.lcssa9105 = phi i32 [ %7863, %ComputeLoop5745 ]
  %7868 = trunc i64 %3317 to i32
  %7869 = lshr i64 %3317, 32
  %7870 = trunc i64 %7869 to i32
  %7871 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %7868, i32 0)
  %7872 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %7870, i32 %7871)
  %7873 = icmp eq i32 %7872, 0
  %7874 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %7873)
  %7875 = extractvalue { i1, i64 } %7874, 0
  %7876 = extractvalue { i1, i64 } %7874, 1
  br i1 %7875, label %3318, label %3321

ComputeLoop5749:                                  ; preds = %for.cond.i.i1.i2244, %ComputeLoop5749
  %Accumulator5751 = phi i32 [ %7881, %ComputeLoop5749 ], [ 0, %for.cond.i.i1.i2244 ]
  %OldValuePhi5752 = phi i32 [ %7880, %ComputeLoop5749 ], [ poison, %for.cond.i.i1.i2244 ]
  %ActiveBits5753 = phi i64 [ %7884, %ComputeLoop5749 ], [ %3326, %for.cond.i.i1.i2244 ]
  %7877 = call i64 @llvm.cttz.i64(i64 %ActiveBits5753, i1 true)
  %7878 = trunc i64 %7877 to i32
  %7879 = call i32 @llvm.amdgcn.readlane.i32(i32 %shl.i.i.i.i.i2256, i32 %7878)
  %7880 = call i32 @llvm.amdgcn.writelane.i32(i32 %Accumulator5751, i32 %7878, i32 %OldValuePhi5752)
  %7881 = or i32 %Accumulator5751, %7879
  %7882 = shl i64 1, %7877
  %7883 = xor i64 %7882, -1
  %7884 = and i64 %ActiveBits5753, %7883
  %7885 = icmp eq i64 %7884, 0
  br i1 %7885, label %ComputeEnd5750, label %ComputeLoop5749, !amdgpu.uniform !42

ComputeEnd5750:                                   ; preds = %ComputeLoop5749
  %.lcssa9107 = phi i32 [ %7880, %ComputeLoop5749 ]
  %.lcssa9106 = phi i32 [ %7881, %ComputeLoop5749 ]
  %div8.i.i.i.i.i2252 = lshr i32 %spec.store.select.i.i.i2251, 5
  %idxprom.i.i.i.i.i2253 = zext nneg i32 %div8.i.i.i.i.i2252 to i64
  %7886 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i.i.i2253
  %7887 = trunc i64 %3326 to i32
  %7888 = lshr i64 %3326, 32
  %7889 = trunc i64 %7888 to i32
  %7890 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %7887, i32 0)
  %7891 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %7889, i32 %7890)
  %7892 = icmp eq i32 %7891, 0
  %7893 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %7892)
  %7894 = extractvalue { i1, i64 } %7893, 0
  %7895 = extractvalue { i1, i64 } %7893, 1
  br i1 %7894, label %3330, label %3332

Flow8165:                                         ; preds = %3350, %if.then.i.i.i.i2262
  %7896 = phi i1 [ false, %3350 ], [ true, %if.then.i.i.i.i2262 ]
  br i1 %7896, label %if.end14.i.i.i2333, label %Flow8166, !amdgpu.uniform !42

ComputeLoop5754:                                  ; preds = %if.then12.i.i.i2267, %ComputeLoop5754
  %Accumulator5756 = phi i32 [ %7900, %ComputeLoop5754 ], [ -1, %if.then12.i.i.i2267 ]
  %ActiveBits5757 = phi i64 [ %7903, %ComputeLoop5754 ], [ %3347, %if.then12.i.i.i2267 ]
  %7897 = call i64 @llvm.cttz.i64(i64 %ActiveBits5757, i1 true)
  %7898 = trunc i64 %7897 to i32
  %7899 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i.i.i2273, i32 %7898)
  %7900 = and i32 %Accumulator5756, %7899
  %7901 = shl i64 1, %7897
  %7902 = xor i64 %7901, -1
  %7903 = and i64 %ActiveBits5757, %7902
  %7904 = icmp eq i64 %7903, 0
  br i1 %7904, label %ComputeEnd5755, label %ComputeLoop5754, !amdgpu.uniform !42

ComputeEnd5755:                                   ; preds = %ComputeLoop5754
  %.lcssa9108 = phi i32 [ %7900, %ComputeLoop5754 ]
  %7905 = trunc i64 %3347 to i32
  %7906 = lshr i64 %3347, 32
  %7907 = trunc i64 %7906 to i32
  %7908 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %7905, i32 0)
  %7909 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %7907, i32 %7908)
  %7910 = icmp eq i32 %7909, 0
  %7911 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %7910)
  %7912 = extractvalue { i1, i64 } %7911, 0
  %7913 = extractvalue { i1, i64 } %7911, 1
  br i1 %7912, label %3348, label %3350

ComputeLoop5758:                                  ; preds = %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit2341, %ComputeLoop5758
  %Accumulator5760 = phi i32 [ %7917, %ComputeLoop5758 ], [ -1, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit2341 ]
  %ActiveBits5761 = phi i64 [ %7920, %ComputeLoop5758 ], [ %3428, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit2341 ]
  %7914 = call i64 @llvm.cttz.i64(i64 %ActiveBits5761, i1 true)
  %7915 = trunc i64 %7914 to i32
  %7916 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i21.i.i2332, i32 %7915)
  %7917 = and i32 %Accumulator5760, %7916
  %7918 = shl i64 1, %7914
  %7919 = xor i64 %7918, -1
  %7920 = and i64 %ActiveBits5761, %7919
  %7921 = icmp eq i64 %7920, 0
  br i1 %7921, label %ComputeEnd5759, label %ComputeLoop5758, !amdgpu.uniform !42

ComputeEnd5759:                                   ; preds = %ComputeLoop5758
  %.lcssa9116 = phi i32 [ %7917, %ComputeLoop5758 ]
  %7922 = trunc i64 %3428 to i32
  %7923 = lshr i64 %3428, 32
  %7924 = trunc i64 %7923 to i32
  %7925 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %7922, i32 0)
  %7926 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %7924, i32 %7925)
  %7927 = icmp eq i32 %7926, 0
  %7928 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %7927)
  %7929 = extractvalue { i1, i64 } %7928, 0
  %7930 = extractvalue { i1, i64 } %7928, 1
  br i1 %7929, label %3429, label %3432

ComputeLoop5762:                                  ; preds = %for.cond.i.i1.i2350, %ComputeLoop5762
  %Accumulator5764 = phi i32 [ %7935, %ComputeLoop5762 ], [ 0, %for.cond.i.i1.i2350 ]
  %OldValuePhi5765 = phi i32 [ %7934, %ComputeLoop5762 ], [ poison, %for.cond.i.i1.i2350 ]
  %ActiveBits5766 = phi i64 [ %7938, %ComputeLoop5762 ], [ %3477, %for.cond.i.i1.i2350 ]
  %7931 = call i64 @llvm.cttz.i64(i64 %ActiveBits5766, i1 true)
  %7932 = trunc i64 %7931 to i32
  %7933 = call i32 @llvm.amdgcn.readlane.i32(i32 %shl.i.i.i.i.i2362, i32 %7932)
  %7934 = call i32 @llvm.amdgcn.writelane.i32(i32 %Accumulator5764, i32 %7932, i32 %OldValuePhi5765)
  %7935 = or i32 %Accumulator5764, %7933
  %7936 = shl i64 1, %7931
  %7937 = xor i64 %7936, -1
  %7938 = and i64 %ActiveBits5766, %7937
  %7939 = icmp eq i64 %7938, 0
  br i1 %7939, label %ComputeEnd5763, label %ComputeLoop5762, !amdgpu.uniform !42

ComputeEnd5763:                                   ; preds = %ComputeLoop5762
  %.lcssa9184 = phi i32 [ %7934, %ComputeLoop5762 ]
  %.lcssa9183 = phi i32 [ %7935, %ComputeLoop5762 ]
  %div8.i.i.i.i.i2358 = lshr i32 %spec.store.select.i.i.i2357, 5
  %idxprom.i.i.i.i.i2359 = zext nneg i32 %div8.i.i.i.i.i2358 to i64
  %7940 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i.i.i2359
  %7941 = trunc i64 %3477 to i32
  %7942 = lshr i64 %3477, 32
  %7943 = trunc i64 %7942 to i32
  %7944 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %7941, i32 0)
  %7945 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %7943, i32 %7944)
  %7946 = icmp eq i32 %7945, 0
  %7947 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %7946)
  %7948 = extractvalue { i1, i64 } %7947, 0
  %7949 = extractvalue { i1, i64 } %7947, 1
  br i1 %7948, label %3481, label %3483

Flow8144:                                         ; preds = %3501, %if.then.i.i.i.i2368
  %7950 = phi i1 [ false, %3501 ], [ true, %if.then.i.i.i.i2368 ]
  br i1 %7950, label %if.end14.i.i.i2439, label %Flow8145, !amdgpu.uniform !42

ComputeLoop5767:                                  ; preds = %if.then12.i.i.i2373, %ComputeLoop5767
  %Accumulator5769 = phi i32 [ %7954, %ComputeLoop5767 ], [ -1, %if.then12.i.i.i2373 ]
  %ActiveBits5770 = phi i64 [ %7957, %ComputeLoop5767 ], [ %3498, %if.then12.i.i.i2373 ]
  %7951 = call i64 @llvm.cttz.i64(i64 %ActiveBits5770, i1 true)
  %7952 = trunc i64 %7951 to i32
  %7953 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i.i.i2379, i32 %7952)
  %7954 = and i32 %Accumulator5769, %7953
  %7955 = shl i64 1, %7951
  %7956 = xor i64 %7955, -1
  %7957 = and i64 %ActiveBits5770, %7956
  %7958 = icmp eq i64 %7957, 0
  br i1 %7958, label %ComputeEnd5768, label %ComputeLoop5767, !amdgpu.uniform !42

ComputeEnd5768:                                   ; preds = %ComputeLoop5767
  %.lcssa9185 = phi i32 [ %7954, %ComputeLoop5767 ]
  %7959 = trunc i64 %3498 to i32
  %7960 = lshr i64 %3498, 32
  %7961 = trunc i64 %7960 to i32
  %7962 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %7959, i32 0)
  %7963 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %7961, i32 %7962)
  %7964 = icmp eq i32 %7963, 0
  %7965 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %7964)
  %7966 = extractvalue { i1, i64 } %7965, 0
  %7967 = extractvalue { i1, i64 } %7965, 1
  br i1 %7966, label %3499, label %3501

ComputeLoop5771:                                  ; preds = %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit2447, %ComputeLoop5771
  %Accumulator5773 = phi i32 [ %7971, %ComputeLoop5771 ], [ -1, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit2447 ]
  %ActiveBits5774 = phi i64 [ %7974, %ComputeLoop5771 ], [ %3579, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit2447 ]
  %7968 = call i64 @llvm.cttz.i64(i64 %ActiveBits5774, i1 true)
  %7969 = trunc i64 %7968 to i32
  %7970 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i21.i.i2438, i32 %7969)
  %7971 = and i32 %Accumulator5773, %7970
  %7972 = shl i64 1, %7968
  %7973 = xor i64 %7972, -1
  %7974 = and i64 %ActiveBits5774, %7973
  %7975 = icmp eq i64 %7974, 0
  br i1 %7975, label %ComputeEnd5772, label %ComputeLoop5771, !amdgpu.uniform !42

ComputeEnd5772:                                   ; preds = %ComputeLoop5771
  %.lcssa9193 = phi i32 [ %7971, %ComputeLoop5771 ]
  %7976 = trunc i64 %3579 to i32
  %7977 = lshr i64 %3579, 32
  %7978 = trunc i64 %7977 to i32
  %7979 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %7976, i32 0)
  %7980 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %7978, i32 %7979)
  %7981 = icmp eq i32 %7980, 0
  %7982 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %7981)
  %7983 = extractvalue { i1, i64 } %7982, 0
  %7984 = extractvalue { i1, i64 } %7982, 1
  br i1 %7983, label %3580, label %3583

ComputeLoop5775:                                  ; preds = %for.cond.i.i.i2497, %ComputeLoop5775
  %Accumulator5777 = phi i32 [ %7989, %ComputeLoop5775 ], [ 0, %for.cond.i.i.i2497 ]
  %OldValuePhi5778 = phi i32 [ %7988, %ComputeLoop5775 ], [ poison, %for.cond.i.i.i2497 ]
  %ActiveBits5779 = phi i64 [ %7992, %ComputeLoop5775 ], [ %3671, %for.cond.i.i.i2497 ]
  %7985 = call i64 @llvm.cttz.i64(i64 %ActiveBits5779, i1 true)
  %7986 = trunc i64 %7985 to i32
  %7987 = call i32 @llvm.amdgcn.readlane.i32(i32 %shl.i.i.i.i.i2509, i32 %7986)
  %7988 = call i32 @llvm.amdgcn.writelane.i32(i32 %Accumulator5777, i32 %7986, i32 %OldValuePhi5778)
  %7989 = or i32 %Accumulator5777, %7987
  %7990 = shl i64 1, %7985
  %7991 = xor i64 %7990, -1
  %7992 = and i64 %ActiveBits5779, %7991
  %7993 = icmp eq i64 %7992, 0
  br i1 %7993, label %ComputeEnd5776, label %ComputeLoop5775, !amdgpu.uniform !42

ComputeEnd5776:                                   ; preds = %ComputeLoop5775
  %.lcssa9162 = phi i32 [ %7988, %ComputeLoop5775 ]
  %.lcssa9161 = phi i32 [ %7989, %ComputeLoop5775 ]
  %div8.i.i.i.i.i2505 = lshr i32 %spec.store.select.i.i.i2504, 5
  %idxprom.i.i.i.i.i2506 = zext nneg i32 %div8.i.i.i.i.i2505 to i64
  %7994 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i.i.i2506
  %7995 = trunc i64 %3671 to i32
  %7996 = lshr i64 %3671, 32
  %7997 = trunc i64 %7996 to i32
  %7998 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %7995, i32 0)
  %7999 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %7997, i32 %7998)
  %8000 = icmp eq i32 %7999, 0
  %8001 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %8000)
  %8002 = extractvalue { i1, i64 } %8001, 0
  %8003 = extractvalue { i1, i64 } %8001, 1
  br i1 %8002, label %3675, label %3677

Flow8098:                                         ; preds = %3695, %if.then.i.i.i.i2515
  %8004 = phi i1 [ false, %3695 ], [ true, %if.then.i.i.i.i2515 ]
  br i1 %8004, label %if.end14.i.i.i2587, label %Flow8099, !amdgpu.uniform !42

ComputeLoop5780:                                  ; preds = %if.then12.i.i.i2520, %ComputeLoop5780
  %Accumulator5782 = phi i32 [ %8008, %ComputeLoop5780 ], [ -1, %if.then12.i.i.i2520 ]
  %ActiveBits5783 = phi i64 [ %8011, %ComputeLoop5780 ], [ %3692, %if.then12.i.i.i2520 ]
  %8005 = call i64 @llvm.cttz.i64(i64 %ActiveBits5783, i1 true)
  %8006 = trunc i64 %8005 to i32
  %8007 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i.i.i2526, i32 %8006)
  %8008 = and i32 %Accumulator5782, %8007
  %8009 = shl i64 1, %8005
  %8010 = xor i64 %8009, -1
  %8011 = and i64 %ActiveBits5783, %8010
  %8012 = icmp eq i64 %8011, 0
  br i1 %8012, label %ComputeEnd5781, label %ComputeLoop5780, !amdgpu.uniform !42

ComputeEnd5781:                                   ; preds = %ComputeLoop5780
  %.lcssa9163 = phi i32 [ %8008, %ComputeLoop5780 ]
  %8013 = trunc i64 %3692 to i32
  %8014 = lshr i64 %3692, 32
  %8015 = trunc i64 %8014 to i32
  %8016 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %8013, i32 0)
  %8017 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %8015, i32 %8016)
  %8018 = icmp eq i32 %8017, 0
  %8019 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %8018)
  %8020 = extractvalue { i1, i64 } %8019, 0
  %8021 = extractvalue { i1, i64 } %8019, 1
  br i1 %8020, label %3693, label %3695

ComputeLoop5784:                                  ; preds = %_ZN22__llvm_libc_22_0_0_git15write_to_stderrENS_3cpp11string_viewE.exit.i2576, %ComputeLoop5784
  %Accumulator5786 = phi i32 [ %8025, %ComputeLoop5784 ], [ -1, %_ZN22__llvm_libc_22_0_0_git15write_to_stderrENS_3cpp11string_viewE.exit.i2576 ]
  %ActiveBits5787 = phi i64 [ %8028, %ComputeLoop5784 ], [ %3777, %_ZN22__llvm_libc_22_0_0_git15write_to_stderrENS_3cpp11string_viewE.exit.i2576 ]
  %8022 = call i64 @llvm.cttz.i64(i64 %ActiveBits5787, i1 true)
  %8023 = trunc i64 %8022 to i32
  %8024 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i21.i.i2585, i32 %8023)
  %8025 = and i32 %Accumulator5786, %8024
  %8026 = shl i64 1, %8022
  %8027 = xor i64 %8026, -1
  %8028 = and i64 %ActiveBits5787, %8027
  %8029 = icmp eq i64 %8028, 0
  br i1 %8029, label %ComputeEnd5785, label %ComputeLoop5784, !amdgpu.uniform !42

ComputeEnd5785:                                   ; preds = %ComputeLoop5784
  %.lcssa9171 = phi i32 [ %8025, %ComputeLoop5784 ]
  %8030 = trunc i64 %3777 to i32
  %8031 = lshr i64 %3777, 32
  %8032 = trunc i64 %8031 to i32
  %8033 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %8030, i32 0)
  %8034 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %8032, i32 %8033)
  %8035 = icmp eq i32 %8034, 0
  %8036 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %8035)
  %8037 = extractvalue { i1, i64 } %8036, 0
  %8038 = extractvalue { i1, i64 } %8036, 1
  br i1 %8037, label %3778, label %3781

ComputeLoop5788:                                  ; preds = %for.cond.i.i1.i2615, %ComputeLoop5788
  %Accumulator5790 = phi i32 [ %8043, %ComputeLoop5788 ], [ 0, %for.cond.i.i1.i2615 ]
  %OldValuePhi5791 = phi i32 [ %8042, %ComputeLoop5788 ], [ poison, %for.cond.i.i1.i2615 ]
  %ActiveBits5792 = phi i64 [ %8046, %ComputeLoop5788 ], [ %3786, %for.cond.i.i1.i2615 ]
  %8039 = call i64 @llvm.cttz.i64(i64 %ActiveBits5792, i1 true)
  %8040 = trunc i64 %8039 to i32
  %8041 = call i32 @llvm.amdgcn.readlane.i32(i32 %shl.i.i.i.i.i2627, i32 %8040)
  %8042 = call i32 @llvm.amdgcn.writelane.i32(i32 %Accumulator5790, i32 %8040, i32 %OldValuePhi5791)
  %8043 = or i32 %Accumulator5790, %8041
  %8044 = shl i64 1, %8039
  %8045 = xor i64 %8044, -1
  %8046 = and i64 %ActiveBits5792, %8045
  %8047 = icmp eq i64 %8046, 0
  br i1 %8047, label %ComputeEnd5789, label %ComputeLoop5788, !amdgpu.uniform !42

ComputeEnd5789:                                   ; preds = %ComputeLoop5788
  %.lcssa9173 = phi i32 [ %8042, %ComputeLoop5788 ]
  %.lcssa9172 = phi i32 [ %8043, %ComputeLoop5788 ]
  %div8.i.i.i.i.i2623 = lshr i32 %spec.store.select.i.i.i2622, 5
  %idxprom.i.i.i.i.i2624 = zext nneg i32 %div8.i.i.i.i.i2623 to i64
  %8048 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i.i.i2624
  %8049 = trunc i64 %3786 to i32
  %8050 = lshr i64 %3786, 32
  %8051 = trunc i64 %8050 to i32
  %8052 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %8049, i32 0)
  %8053 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %8051, i32 %8052)
  %8054 = icmp eq i32 %8053, 0
  %8055 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %8054)
  %8056 = extractvalue { i1, i64 } %8055, 0
  %8057 = extractvalue { i1, i64 } %8055, 1
  br i1 %8056, label %3790, label %3792

Flow8079:                                         ; preds = %3810, %if.then.i.i.i.i2633
  %8058 = phi i1 [ false, %3810 ], [ true, %if.then.i.i.i.i2633 ]
  br i1 %8058, label %if.end14.i.i.i2704, label %Flow8080, !amdgpu.uniform !42

ComputeLoop5793:                                  ; preds = %if.then12.i.i.i2638, %ComputeLoop5793
  %Accumulator5795 = phi i32 [ %8062, %ComputeLoop5793 ], [ -1, %if.then12.i.i.i2638 ]
  %ActiveBits5796 = phi i64 [ %8065, %ComputeLoop5793 ], [ %3807, %if.then12.i.i.i2638 ]
  %8059 = call i64 @llvm.cttz.i64(i64 %ActiveBits5796, i1 true)
  %8060 = trunc i64 %8059 to i32
  %8061 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i.i.i2644, i32 %8060)
  %8062 = and i32 %Accumulator5795, %8061
  %8063 = shl i64 1, %8059
  %8064 = xor i64 %8063, -1
  %8065 = and i64 %ActiveBits5796, %8064
  %8066 = icmp eq i64 %8065, 0
  br i1 %8066, label %ComputeEnd5794, label %ComputeLoop5793, !amdgpu.uniform !42

ComputeEnd5794:                                   ; preds = %ComputeLoop5793
  %.lcssa9174 = phi i32 [ %8062, %ComputeLoop5793 ]
  %8067 = trunc i64 %3807 to i32
  %8068 = lshr i64 %3807, 32
  %8069 = trunc i64 %8068 to i32
  %8070 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %8067, i32 0)
  %8071 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %8069, i32 %8070)
  %8072 = icmp eq i32 %8071, 0
  %8073 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %8072)
  %8074 = extractvalue { i1, i64 } %8073, 0
  %8075 = extractvalue { i1, i64 } %8073, 1
  br i1 %8074, label %3808, label %3810

ComputeLoop5797:                                  ; preds = %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit2712, %ComputeLoop5797
  %Accumulator5799 = phi i32 [ %8079, %ComputeLoop5797 ], [ -1, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit2712 ]
  %ActiveBits5800 = phi i64 [ %8082, %ComputeLoop5797 ], [ %3888, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit2712 ]
  %8076 = call i64 @llvm.cttz.i64(i64 %ActiveBits5800, i1 true)
  %8077 = trunc i64 %8076 to i32
  %8078 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i21.i.i2703, i32 %8077)
  %8079 = and i32 %Accumulator5799, %8078
  %8080 = shl i64 1, %8076
  %8081 = xor i64 %8080, -1
  %8082 = and i64 %ActiveBits5800, %8081
  %8083 = icmp eq i64 %8082, 0
  br i1 %8083, label %ComputeEnd5798, label %ComputeLoop5797, !amdgpu.uniform !42

ComputeEnd5798:                                   ; preds = %ComputeLoop5797
  %.lcssa9182 = phi i32 [ %8079, %ComputeLoop5797 ]
  %8084 = trunc i64 %3888 to i32
  %8085 = lshr i64 %3888, 32
  %8086 = trunc i64 %8085 to i32
  %8087 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %8084, i32 0)
  %8088 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %8086, i32 %8087)
  %8089 = icmp eq i32 %8088, 0
  %8090 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %8089)
  %8091 = extractvalue { i1, i64 } %8090, 0
  %8092 = extractvalue { i1, i64 } %8090, 1
  br i1 %8091, label %3889, label %3892

ComputeLoop5801:                                  ; preds = %for.cond.i.i.i2771, %ComputeLoop5801
  %Accumulator5803 = phi i32 [ %8097, %ComputeLoop5801 ], [ 0, %for.cond.i.i.i2771 ]
  %OldValuePhi5804 = phi i32 [ %8096, %ComputeLoop5801 ], [ poison, %for.cond.i.i.i2771 ]
  %ActiveBits5805 = phi i64 [ %8100, %ComputeLoop5801 ], [ %3948, %for.cond.i.i.i2771 ]
  %8093 = call i64 @llvm.cttz.i64(i64 %ActiveBits5805, i1 true)
  %8094 = trunc i64 %8093 to i32
  %8095 = call i32 @llvm.amdgcn.readlane.i32(i32 %shl.i.i.i.i.i2783, i32 %8094)
  %8096 = call i32 @llvm.amdgcn.writelane.i32(i32 %Accumulator5803, i32 %8094, i32 %OldValuePhi5804)
  %8097 = or i32 %Accumulator5803, %8095
  %8098 = shl i64 1, %8093
  %8099 = xor i64 %8098, -1
  %8100 = and i64 %ActiveBits5805, %8099
  %8101 = icmp eq i64 %8100, 0
  br i1 %8101, label %ComputeEnd5802, label %ComputeLoop5801, !amdgpu.uniform !42

ComputeEnd5802:                                   ; preds = %ComputeLoop5801
  %.lcssa9118 = phi i32 [ %8096, %ComputeLoop5801 ]
  %.lcssa9117 = phi i32 [ %8097, %ComputeLoop5801 ]
  %div8.i.i.i.i.i2779 = lshr i32 %spec.store.select.i.i.i2778, 5
  %idxprom.i.i.i.i.i2780 = zext nneg i32 %div8.i.i.i.i.i2779 to i64
  %8102 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i.i.i2780
  %8103 = trunc i64 %3948 to i32
  %8104 = lshr i64 %3948, 32
  %8105 = trunc i64 %8104 to i32
  %8106 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %8103, i32 0)
  %8107 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %8105, i32 %8106)
  %8108 = icmp eq i32 %8107, 0
  %8109 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %8108)
  %8110 = extractvalue { i1, i64 } %8109, 0
  %8111 = extractvalue { i1, i64 } %8109, 1
  br i1 %8110, label %3952, label %3954

Flow7972:                                         ; preds = %3972, %if.then.i.i.i.i2789
  %8112 = phi i1 [ false, %3972 ], [ true, %if.then.i.i.i.i2789 ]
  br i1 %8112, label %if.end14.i.i.i2861, label %Flow7973, !amdgpu.uniform !42

ComputeLoop5806:                                  ; preds = %if.then12.i.i.i2794, %ComputeLoop5806
  %Accumulator5808 = phi i32 [ %8116, %ComputeLoop5806 ], [ -1, %if.then12.i.i.i2794 ]
  %ActiveBits5809 = phi i64 [ %8119, %ComputeLoop5806 ], [ %3969, %if.then12.i.i.i2794 ]
  %8113 = call i64 @llvm.cttz.i64(i64 %ActiveBits5809, i1 true)
  %8114 = trunc i64 %8113 to i32
  %8115 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i.i.i2800, i32 %8114)
  %8116 = and i32 %Accumulator5808, %8115
  %8117 = shl i64 1, %8113
  %8118 = xor i64 %8117, -1
  %8119 = and i64 %ActiveBits5809, %8118
  %8120 = icmp eq i64 %8119, 0
  br i1 %8120, label %ComputeEnd5807, label %ComputeLoop5806, !amdgpu.uniform !42

ComputeEnd5807:                                   ; preds = %ComputeLoop5806
  %.lcssa9119 = phi i32 [ %8116, %ComputeLoop5806 ]
  %8121 = trunc i64 %3969 to i32
  %8122 = lshr i64 %3969, 32
  %8123 = trunc i64 %8122 to i32
  %8124 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %8121, i32 0)
  %8125 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %8123, i32 %8124)
  %8126 = icmp eq i32 %8125, 0
  %8127 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %8126)
  %8128 = extractvalue { i1, i64 } %8127, 0
  %8129 = extractvalue { i1, i64 } %8127, 1
  br i1 %8128, label %3970, label %3972

ComputeLoop5810:                                  ; preds = %_ZN22__llvm_libc_22_0_0_git15write_to_stderrENS_3cpp11string_viewE.exit.i2850, %ComputeLoop5810
  %Accumulator5812 = phi i32 [ %8133, %ComputeLoop5810 ], [ -1, %_ZN22__llvm_libc_22_0_0_git15write_to_stderrENS_3cpp11string_viewE.exit.i2850 ]
  %ActiveBits5813 = phi i64 [ %8136, %ComputeLoop5810 ], [ %4054, %_ZN22__llvm_libc_22_0_0_git15write_to_stderrENS_3cpp11string_viewE.exit.i2850 ]
  %8130 = call i64 @llvm.cttz.i64(i64 %ActiveBits5813, i1 true)
  %8131 = trunc i64 %8130 to i32
  %8132 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i21.i.i2859, i32 %8131)
  %8133 = and i32 %Accumulator5812, %8132
  %8134 = shl i64 1, %8130
  %8135 = xor i64 %8134, -1
  %8136 = and i64 %ActiveBits5813, %8135
  %8137 = icmp eq i64 %8136, 0
  br i1 %8137, label %ComputeEnd5811, label %ComputeLoop5810, !amdgpu.uniform !42

ComputeEnd5811:                                   ; preds = %ComputeLoop5810
  %.lcssa9127 = phi i32 [ %8133, %ComputeLoop5810 ]
  %8138 = trunc i64 %4054 to i32
  %8139 = lshr i64 %4054, 32
  %8140 = trunc i64 %8139 to i32
  %8141 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %8138, i32 0)
  %8142 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %8140, i32 %8141)
  %8143 = icmp eq i32 %8142, 0
  %8144 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %8143)
  %8145 = extractvalue { i1, i64 } %8144, 0
  %8146 = extractvalue { i1, i64 } %8144, 1
  br i1 %8145, label %4055, label %4058

ComputeLoop5814:                                  ; preds = %for.cond.i.i1.i2890, %ComputeLoop5814
  %Accumulator5816 = phi i32 [ %8151, %ComputeLoop5814 ], [ 0, %for.cond.i.i1.i2890 ]
  %OldValuePhi5817 = phi i32 [ %8150, %ComputeLoop5814 ], [ poison, %for.cond.i.i1.i2890 ]
  %ActiveBits5818 = phi i64 [ %8154, %ComputeLoop5814 ], [ %4063, %for.cond.i.i1.i2890 ]
  %8147 = call i64 @llvm.cttz.i64(i64 %ActiveBits5818, i1 true)
  %8148 = trunc i64 %8147 to i32
  %8149 = call i32 @llvm.amdgcn.readlane.i32(i32 %shl.i.i.i.i.i2902, i32 %8148)
  %8150 = call i32 @llvm.amdgcn.writelane.i32(i32 %Accumulator5816, i32 %8148, i32 %OldValuePhi5817)
  %8151 = or i32 %Accumulator5816, %8149
  %8152 = shl i64 1, %8147
  %8153 = xor i64 %8152, -1
  %8154 = and i64 %ActiveBits5818, %8153
  %8155 = icmp eq i64 %8154, 0
  br i1 %8155, label %ComputeEnd5815, label %ComputeLoop5814, !amdgpu.uniform !42

ComputeEnd5815:                                   ; preds = %ComputeLoop5814
  %.lcssa9129 = phi i32 [ %8150, %ComputeLoop5814 ]
  %.lcssa9128 = phi i32 [ %8151, %ComputeLoop5814 ]
  %div8.i.i.i.i.i2898 = lshr i32 %spec.store.select.i.i.i2897, 5
  %idxprom.i.i.i.i.i2899 = zext nneg i32 %div8.i.i.i.i.i2898 to i64
  %8156 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i.i.i2899
  %8157 = trunc i64 %4063 to i32
  %8158 = lshr i64 %4063, 32
  %8159 = trunc i64 %8158 to i32
  %8160 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %8157, i32 0)
  %8161 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %8159, i32 %8160)
  %8162 = icmp eq i32 %8161, 0
  %8163 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %8162)
  %8164 = extractvalue { i1, i64 } %8163, 0
  %8165 = extractvalue { i1, i64 } %8163, 1
  br i1 %8164, label %4067, label %4069

Flow7953:                                         ; preds = %4087, %if.then.i.i.i.i2908
  %8166 = phi i1 [ false, %4087 ], [ true, %if.then.i.i.i.i2908 ]
  br i1 %8166, label %if.end14.i.i.i2979, label %Flow7954, !amdgpu.uniform !42

ComputeLoop5819:                                  ; preds = %if.then12.i.i.i2913, %ComputeLoop5819
  %Accumulator5821 = phi i32 [ %8170, %ComputeLoop5819 ], [ -1, %if.then12.i.i.i2913 ]
  %ActiveBits5822 = phi i64 [ %8173, %ComputeLoop5819 ], [ %4084, %if.then12.i.i.i2913 ]
  %8167 = call i64 @llvm.cttz.i64(i64 %ActiveBits5822, i1 true)
  %8168 = trunc i64 %8167 to i32
  %8169 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i.i.i2919, i32 %8168)
  %8170 = and i32 %Accumulator5821, %8169
  %8171 = shl i64 1, %8167
  %8172 = xor i64 %8171, -1
  %8173 = and i64 %ActiveBits5822, %8172
  %8174 = icmp eq i64 %8173, 0
  br i1 %8174, label %ComputeEnd5820, label %ComputeLoop5819, !amdgpu.uniform !42

ComputeEnd5820:                                   ; preds = %ComputeLoop5819
  %.lcssa9130 = phi i32 [ %8170, %ComputeLoop5819 ]
  %8175 = trunc i64 %4084 to i32
  %8176 = lshr i64 %4084, 32
  %8177 = trunc i64 %8176 to i32
  %8178 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %8175, i32 0)
  %8179 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %8177, i32 %8178)
  %8180 = icmp eq i32 %8179, 0
  %8181 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %8180)
  %8182 = extractvalue { i1, i64 } %8181, 0
  %8183 = extractvalue { i1, i64 } %8181, 1
  br i1 %8182, label %4085, label %4087

ComputeLoop5823:                                  ; preds = %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit2987, %ComputeLoop5823
  %Accumulator5825 = phi i32 [ %8187, %ComputeLoop5823 ], [ -1, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit2987 ]
  %ActiveBits5826 = phi i64 [ %8190, %ComputeLoop5823 ], [ %4165, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit2987 ]
  %8184 = call i64 @llvm.cttz.i64(i64 %ActiveBits5826, i1 true)
  %8185 = trunc i64 %8184 to i32
  %8186 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i21.i.i2978, i32 %8185)
  %8187 = and i32 %Accumulator5825, %8186
  %8188 = shl i64 1, %8184
  %8189 = xor i64 %8188, -1
  %8190 = and i64 %ActiveBits5826, %8189
  %8191 = icmp eq i64 %8190, 0
  br i1 %8191, label %ComputeEnd5824, label %ComputeLoop5823, !amdgpu.uniform !42

ComputeEnd5824:                                   ; preds = %ComputeLoop5823
  %.lcssa9138 = phi i32 [ %8187, %ComputeLoop5823 ]
  %8192 = trunc i64 %4165 to i32
  %8193 = lshr i64 %4165, 32
  %8194 = trunc i64 %8193 to i32
  %8195 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %8192, i32 0)
  %8196 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %8194, i32 %8195)
  %8197 = icmp eq i32 %8196, 0
  %8198 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %8197)
  %8199 = extractvalue { i1, i64 } %8198, 0
  %8200 = extractvalue { i1, i64 } %8198, 1
  br i1 %8199, label %4166, label %4169

ComputeLoop5827:                                  ; preds = %for.cond.i.i.i3046, %ComputeLoop5827
  %Accumulator5829 = phi i32 [ %8205, %ComputeLoop5827 ], [ 0, %for.cond.i.i.i3046 ]
  %OldValuePhi5830 = phi i32 [ %8204, %ComputeLoop5827 ], [ poison, %for.cond.i.i.i3046 ]
  %ActiveBits5831 = phi i64 [ %8208, %ComputeLoop5827 ], [ %4232, %for.cond.i.i.i3046 ]
  %8201 = call i64 @llvm.cttz.i64(i64 %ActiveBits5831, i1 true)
  %8202 = trunc i64 %8201 to i32
  %8203 = call i32 @llvm.amdgcn.readlane.i32(i32 %shl.i.i.i.i.i3058, i32 %8202)
  %8204 = call i32 @llvm.amdgcn.writelane.i32(i32 %Accumulator5829, i32 %8202, i32 %OldValuePhi5830)
  %8205 = or i32 %Accumulator5829, %8203
  %8206 = shl i64 1, %8201
  %8207 = xor i64 %8206, -1
  %8208 = and i64 %ActiveBits5831, %8207
  %8209 = icmp eq i64 %8208, 0
  br i1 %8209, label %ComputeEnd5828, label %ComputeLoop5827, !amdgpu.uniform !42

ComputeEnd5828:                                   ; preds = %ComputeLoop5827
  %.lcssa9140 = phi i32 [ %8204, %ComputeLoop5827 ]
  %.lcssa9139 = phi i32 [ %8205, %ComputeLoop5827 ]
  %div8.i.i.i.i.i3054 = lshr i32 %spec.store.select.i.i.i3053, 5
  %idxprom.i.i.i.i.i3055 = zext nneg i32 %div8.i.i.i.i.i3054 to i64
  %8210 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i.i.i3055
  %8211 = trunc i64 %4232 to i32
  %8212 = lshr i64 %4232, 32
  %8213 = trunc i64 %8212 to i32
  %8214 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %8211, i32 0)
  %8215 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %8213, i32 %8214)
  %8216 = icmp eq i32 %8215, 0
  %8217 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %8216)
  %8218 = extractvalue { i1, i64 } %8217, 0
  %8219 = extractvalue { i1, i64 } %8217, 1
  br i1 %8218, label %4236, label %4238

Flow8033:                                         ; preds = %4256, %if.then.i.i.i.i3064
  %8220 = phi i1 [ false, %4256 ], [ true, %if.then.i.i.i.i3064 ]
  br i1 %8220, label %if.end14.i.i.i3136, label %Flow8034, !amdgpu.uniform !42

ComputeLoop5832:                                  ; preds = %if.then12.i.i.i3069, %ComputeLoop5832
  %Accumulator5834 = phi i32 [ %8224, %ComputeLoop5832 ], [ -1, %if.then12.i.i.i3069 ]
  %ActiveBits5835 = phi i64 [ %8227, %ComputeLoop5832 ], [ %4253, %if.then12.i.i.i3069 ]
  %8221 = call i64 @llvm.cttz.i64(i64 %ActiveBits5835, i1 true)
  %8222 = trunc i64 %8221 to i32
  %8223 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i.i.i3075, i32 %8222)
  %8224 = and i32 %Accumulator5834, %8223
  %8225 = shl i64 1, %8221
  %8226 = xor i64 %8225, -1
  %8227 = and i64 %ActiveBits5835, %8226
  %8228 = icmp eq i64 %8227, 0
  br i1 %8228, label %ComputeEnd5833, label %ComputeLoop5832, !amdgpu.uniform !42

ComputeEnd5833:                                   ; preds = %ComputeLoop5832
  %.lcssa9141 = phi i32 [ %8224, %ComputeLoop5832 ]
  %8229 = trunc i64 %4253 to i32
  %8230 = lshr i64 %4253, 32
  %8231 = trunc i64 %8230 to i32
  %8232 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %8229, i32 0)
  %8233 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %8231, i32 %8232)
  %8234 = icmp eq i32 %8233, 0
  %8235 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %8234)
  %8236 = extractvalue { i1, i64 } %8235, 0
  %8237 = extractvalue { i1, i64 } %8235, 1
  br i1 %8236, label %4254, label %4256

ComputeLoop5836:                                  ; preds = %_ZN22__llvm_libc_22_0_0_git15write_to_stderrENS_3cpp11string_viewE.exit.i3125, %ComputeLoop5836
  %Accumulator5838 = phi i32 [ %8241, %ComputeLoop5836 ], [ -1, %_ZN22__llvm_libc_22_0_0_git15write_to_stderrENS_3cpp11string_viewE.exit.i3125 ]
  %ActiveBits5839 = phi i64 [ %8244, %ComputeLoop5836 ], [ %4338, %_ZN22__llvm_libc_22_0_0_git15write_to_stderrENS_3cpp11string_viewE.exit.i3125 ]
  %8238 = call i64 @llvm.cttz.i64(i64 %ActiveBits5839, i1 true)
  %8239 = trunc i64 %8238 to i32
  %8240 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i21.i.i3134, i32 %8239)
  %8241 = and i32 %Accumulator5838, %8240
  %8242 = shl i64 1, %8238
  %8243 = xor i64 %8242, -1
  %8244 = and i64 %ActiveBits5839, %8243
  %8245 = icmp eq i64 %8244, 0
  br i1 %8245, label %ComputeEnd5837, label %ComputeLoop5836, !amdgpu.uniform !42

ComputeEnd5837:                                   ; preds = %ComputeLoop5836
  %.lcssa9149 = phi i32 [ %8241, %ComputeLoop5836 ]
  %8246 = trunc i64 %4338 to i32
  %8247 = lshr i64 %4338, 32
  %8248 = trunc i64 %8247 to i32
  %8249 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %8246, i32 0)
  %8250 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %8248, i32 %8249)
  %8251 = icmp eq i32 %8250, 0
  %8252 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %8251)
  %8253 = extractvalue { i1, i64 } %8252, 0
  %8254 = extractvalue { i1, i64 } %8252, 1
  br i1 %8253, label %4339, label %4342

ComputeLoop5840:                                  ; preds = %for.cond.i.i1.i3165, %ComputeLoop5840
  %Accumulator5842 = phi i32 [ %8259, %ComputeLoop5840 ], [ 0, %for.cond.i.i1.i3165 ]
  %OldValuePhi5843 = phi i32 [ %8258, %ComputeLoop5840 ], [ poison, %for.cond.i.i1.i3165 ]
  %ActiveBits5844 = phi i64 [ %8262, %ComputeLoop5840 ], [ %4347, %for.cond.i.i1.i3165 ]
  %8255 = call i64 @llvm.cttz.i64(i64 %ActiveBits5844, i1 true)
  %8256 = trunc i64 %8255 to i32
  %8257 = call i32 @llvm.amdgcn.readlane.i32(i32 %shl.i.i.i.i.i3177, i32 %8256)
  %8258 = call i32 @llvm.amdgcn.writelane.i32(i32 %Accumulator5842, i32 %8256, i32 %OldValuePhi5843)
  %8259 = or i32 %Accumulator5842, %8257
  %8260 = shl i64 1, %8255
  %8261 = xor i64 %8260, -1
  %8262 = and i64 %ActiveBits5844, %8261
  %8263 = icmp eq i64 %8262, 0
  br i1 %8263, label %ComputeEnd5841, label %ComputeLoop5840, !amdgpu.uniform !42

ComputeEnd5841:                                   ; preds = %ComputeLoop5840
  %.lcssa9151 = phi i32 [ %8258, %ComputeLoop5840 ]
  %.lcssa9150 = phi i32 [ %8259, %ComputeLoop5840 ]
  %div8.i.i.i.i.i3173 = lshr i32 %spec.store.select.i.i.i3172, 5
  %idxprom.i.i.i.i.i3174 = zext nneg i32 %div8.i.i.i.i.i3173 to i64
  %8264 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i.i.i3174
  %8265 = trunc i64 %4347 to i32
  %8266 = lshr i64 %4347, 32
  %8267 = trunc i64 %8266 to i32
  %8268 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %8265, i32 0)
  %8269 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %8267, i32 %8268)
  %8270 = icmp eq i32 %8269, 0
  %8271 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %8270)
  %8272 = extractvalue { i1, i64 } %8271, 0
  %8273 = extractvalue { i1, i64 } %8271, 1
  br i1 %8272, label %4351, label %4353

Flow8014:                                         ; preds = %4371, %if.then.i.i.i.i3183
  %8274 = phi i1 [ false, %4371 ], [ true, %if.then.i.i.i.i3183 ]
  br i1 %8274, label %if.end14.i.i.i3254, label %Flow8015, !amdgpu.uniform !42

ComputeLoop5845:                                  ; preds = %if.then12.i.i.i3188, %ComputeLoop5845
  %Accumulator5847 = phi i32 [ %8278, %ComputeLoop5845 ], [ -1, %if.then12.i.i.i3188 ]
  %ActiveBits5848 = phi i64 [ %8281, %ComputeLoop5845 ], [ %4368, %if.then12.i.i.i3188 ]
  %8275 = call i64 @llvm.cttz.i64(i64 %ActiveBits5848, i1 true)
  %8276 = trunc i64 %8275 to i32
  %8277 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i.i.i3194, i32 %8276)
  %8278 = and i32 %Accumulator5847, %8277
  %8279 = shl i64 1, %8275
  %8280 = xor i64 %8279, -1
  %8281 = and i64 %ActiveBits5848, %8280
  %8282 = icmp eq i64 %8281, 0
  br i1 %8282, label %ComputeEnd5846, label %ComputeLoop5845, !amdgpu.uniform !42

ComputeEnd5846:                                   ; preds = %ComputeLoop5845
  %.lcssa9152 = phi i32 [ %8278, %ComputeLoop5845 ]
  %8283 = trunc i64 %4368 to i32
  %8284 = lshr i64 %4368, 32
  %8285 = trunc i64 %8284 to i32
  %8286 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %8283, i32 0)
  %8287 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %8285, i32 %8286)
  %8288 = icmp eq i32 %8287, 0
  %8289 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %8288)
  %8290 = extractvalue { i1, i64 } %8289, 0
  %8291 = extractvalue { i1, i64 } %8289, 1
  br i1 %8290, label %4369, label %4371

ComputeLoop5849:                                  ; preds = %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit3262, %ComputeLoop5849
  %Accumulator5851 = phi i32 [ %8295, %ComputeLoop5849 ], [ -1, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit3262 ]
  %ActiveBits5852 = phi i64 [ %8298, %ComputeLoop5849 ], [ %4449, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit3262 ]
  %8292 = call i64 @llvm.cttz.i64(i64 %ActiveBits5852, i1 true)
  %8293 = trunc i64 %8292 to i32
  %8294 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i21.i.i3253, i32 %8293)
  %8295 = and i32 %Accumulator5851, %8294
  %8296 = shl i64 1, %8292
  %8297 = xor i64 %8296, -1
  %8298 = and i64 %ActiveBits5852, %8297
  %8299 = icmp eq i64 %8298, 0
  br i1 %8299, label %ComputeEnd5850, label %ComputeLoop5849, !amdgpu.uniform !42

ComputeEnd5850:                                   ; preds = %ComputeLoop5849
  %.lcssa9160 = phi i32 [ %8295, %ComputeLoop5849 ]
  %8300 = trunc i64 %4449 to i32
  %8301 = lshr i64 %4449, 32
  %8302 = trunc i64 %8301 to i32
  %8303 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %8300, i32 0)
  %8304 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %8302, i32 %8303)
  %8305 = icmp eq i32 %8304, 0
  %8306 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %8305)
  %8307 = extractvalue { i1, i64 } %8306, 0
  %8308 = extractvalue { i1, i64 } %8306, 1
  br i1 %8307, label %4450, label %4453

ComputeLoop5853:                                  ; preds = %for.cond.i.i1.i3271, %ComputeLoop5853
  %Accumulator5855 = phi i32 [ %8313, %ComputeLoop5853 ], [ 0, %for.cond.i.i1.i3271 ]
  %OldValuePhi5856 = phi i32 [ %8312, %ComputeLoop5853 ], [ poison, %for.cond.i.i1.i3271 ]
  %ActiveBits5857 = phi i64 [ %8316, %ComputeLoop5853 ], [ %4460, %for.cond.i.i1.i3271 ]
  %8309 = call i64 @llvm.cttz.i64(i64 %ActiveBits5857, i1 true)
  %8310 = trunc i64 %8309 to i32
  %8311 = call i32 @llvm.amdgcn.readlane.i32(i32 %shl.i.i.i.i.i3283, i32 %8310)
  %8312 = call i32 @llvm.amdgcn.writelane.i32(i32 %Accumulator5855, i32 %8310, i32 %OldValuePhi5856)
  %8313 = or i32 %Accumulator5855, %8311
  %8314 = shl i64 1, %8309
  %8315 = xor i64 %8314, -1
  %8316 = and i64 %ActiveBits5857, %8315
  %8317 = icmp eq i64 %8316, 0
  br i1 %8317, label %ComputeEnd5854, label %ComputeLoop5853, !amdgpu.uniform !42

ComputeEnd5854:                                   ; preds = %ComputeLoop5853
  %.lcssa8928 = phi i32 [ %8312, %ComputeLoop5853 ]
  %.lcssa8927 = phi i32 [ %8313, %ComputeLoop5853 ]
  %div8.i.i.i.i.i3279 = lshr i32 %spec.store.select.i.i.i3278, 5
  %idxprom.i.i.i.i.i3280 = zext nneg i32 %div8.i.i.i.i.i3279 to i64
  %8318 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i.i.i3280
  %8319 = trunc i64 %4460 to i32
  %8320 = lshr i64 %4460, 32
  %8321 = trunc i64 %8320 to i32
  %8322 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %8319, i32 0)
  %8323 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %8321, i32 %8322)
  %8324 = icmp eq i32 %8323, 0
  %8325 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %8324)
  %8326 = extractvalue { i1, i64 } %8325, 0
  %8327 = extractvalue { i1, i64 } %8325, 1
  br i1 %8326, label %4464, label %4466

Flow7912:                                         ; preds = %4484, %if.then.i.i.i.i3289
  %8328 = phi i1 [ false, %4484 ], [ true, %if.then.i.i.i.i3289 ]
  br i1 %8328, label %if.end14.i.i.i3360, label %Flow7913, !amdgpu.uniform !42

ComputeLoop5858:                                  ; preds = %if.then12.i.i.i3294, %ComputeLoop5858
  %Accumulator5860 = phi i32 [ %8332, %ComputeLoop5858 ], [ -1, %if.then12.i.i.i3294 ]
  %ActiveBits5861 = phi i64 [ %8335, %ComputeLoop5858 ], [ %4481, %if.then12.i.i.i3294 ]
  %8329 = call i64 @llvm.cttz.i64(i64 %ActiveBits5861, i1 true)
  %8330 = trunc i64 %8329 to i32
  %8331 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i.i.i3300, i32 %8330)
  %8332 = and i32 %Accumulator5860, %8331
  %8333 = shl i64 1, %8329
  %8334 = xor i64 %8333, -1
  %8335 = and i64 %ActiveBits5861, %8334
  %8336 = icmp eq i64 %8335, 0
  br i1 %8336, label %ComputeEnd5859, label %ComputeLoop5858, !amdgpu.uniform !42

ComputeEnd5859:                                   ; preds = %ComputeLoop5858
  %.lcssa8929 = phi i32 [ %8332, %ComputeLoop5858 ]
  %8337 = trunc i64 %4481 to i32
  %8338 = lshr i64 %4481, 32
  %8339 = trunc i64 %8338 to i32
  %8340 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %8337, i32 0)
  %8341 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %8339, i32 %8340)
  %8342 = icmp eq i32 %8341, 0
  %8343 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %8342)
  %8344 = extractvalue { i1, i64 } %8343, 0
  %8345 = extractvalue { i1, i64 } %8343, 1
  br i1 %8344, label %4482, label %4484

ComputeLoop5862:                                  ; preds = %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit3368, %ComputeLoop5862
  %Accumulator5864 = phi i32 [ %8349, %ComputeLoop5862 ], [ -1, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit3368 ]
  %ActiveBits5865 = phi i64 [ %8352, %ComputeLoop5862 ], [ %4563, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit3368 ]
  %8346 = call i64 @llvm.cttz.i64(i64 %ActiveBits5865, i1 true)
  %8347 = trunc i64 %8346 to i32
  %8348 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i21.i.i3359, i32 %8347)
  %8349 = and i32 %Accumulator5864, %8348
  %8350 = shl i64 1, %8346
  %8351 = xor i64 %8350, -1
  %8352 = and i64 %ActiveBits5865, %8351
  %8353 = icmp eq i64 %8352, 0
  br i1 %8353, label %ComputeEnd5863, label %ComputeLoop5862, !amdgpu.uniform !42

ComputeEnd5863:                                   ; preds = %ComputeLoop5862
  %.lcssa8923 = phi i32 [ %8349, %ComputeLoop5862 ]
  %8354 = trunc i64 %4563 to i32
  %8355 = lshr i64 %4563, 32
  %8356 = trunc i64 %8355 to i32
  %8357 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %8354, i32 0)
  %8358 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %8356, i32 %8357)
  %8359 = icmp eq i32 %8358, 0
  %8360 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %8359)
  %8361 = extractvalue { i1, i64 } %8360, 0
  %8362 = extractvalue { i1, i64 } %8360, 1
  br i1 %8361, label %4564, label %4567

ComputeLoop5866:                                  ; preds = %for.cond.i.i.i3437, %ComputeLoop5866
  %Accumulator5868 = phi i32 [ %8367, %ComputeLoop5866 ], [ 0, %for.cond.i.i.i3437 ]
  %OldValuePhi5869 = phi i32 [ %8366, %ComputeLoop5866 ], [ poison, %for.cond.i.i.i3437 ]
  %ActiveBits5870 = phi i64 [ %8370, %ComputeLoop5866 ], [ %4685, %for.cond.i.i.i3437 ]
  %8363 = call i64 @llvm.cttz.i64(i64 %ActiveBits5870, i1 true)
  %8364 = trunc i64 %8363 to i32
  %8365 = call i32 @llvm.amdgcn.readlane.i32(i32 %shl.i.i.i.i.i3449, i32 %8364)
  %8366 = call i32 @llvm.amdgcn.writelane.i32(i32 %Accumulator5868, i32 %8364, i32 %OldValuePhi5869)
  %8367 = or i32 %Accumulator5868, %8365
  %8368 = shl i64 1, %8363
  %8369 = xor i64 %8368, -1
  %8370 = and i64 %ActiveBits5870, %8369
  %8371 = icmp eq i64 %8370, 0
  br i1 %8371, label %ComputeEnd5867, label %ComputeLoop5866, !amdgpu.uniform !42

ComputeEnd5867:                                   ; preds = %ComputeLoop5866
  %.lcssa8914 = phi i32 [ %8366, %ComputeLoop5866 ]
  %.lcssa8913 = phi i32 [ %8367, %ComputeLoop5866 ]
  %div8.i.i.i.i.i3445 = lshr i32 %spec.store.select.i.i.i3444, 5
  %idxprom.i.i.i.i.i3446 = zext nneg i32 %div8.i.i.i.i.i3445 to i64
  %8372 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i.i.i3446
  %8373 = trunc i64 %4685 to i32
  %8374 = lshr i64 %4685, 32
  %8375 = trunc i64 %8374 to i32
  %8376 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %8373, i32 0)
  %8377 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %8375, i32 %8376)
  %8378 = icmp eq i32 %8377, 0
  %8379 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %8378)
  %8380 = extractvalue { i1, i64 } %8379, 0
  %8381 = extractvalue { i1, i64 } %8379, 1
  br i1 %8380, label %4689, label %4691

Flow7870:                                         ; preds = %4709, %if.then.i.i.i.i3455
  %8382 = phi i1 [ false, %4709 ], [ true, %if.then.i.i.i.i3455 ]
  br i1 %8382, label %if.end14.i.i.i3527, label %Flow7871, !amdgpu.uniform !42

ComputeLoop5871:                                  ; preds = %if.then12.i.i.i3460, %ComputeLoop5871
  %Accumulator5873 = phi i32 [ %8386, %ComputeLoop5871 ], [ -1, %if.then12.i.i.i3460 ]
  %ActiveBits5874 = phi i64 [ %8389, %ComputeLoop5871 ], [ %4706, %if.then12.i.i.i3460 ]
  %8383 = call i64 @llvm.cttz.i64(i64 %ActiveBits5874, i1 true)
  %8384 = trunc i64 %8383 to i32
  %8385 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i.i.i3466, i32 %8384)
  %8386 = and i32 %Accumulator5873, %8385
  %8387 = shl i64 1, %8383
  %8388 = xor i64 %8387, -1
  %8389 = and i64 %ActiveBits5874, %8388
  %8390 = icmp eq i64 %8389, 0
  br i1 %8390, label %ComputeEnd5872, label %ComputeLoop5871, !amdgpu.uniform !42

ComputeEnd5872:                                   ; preds = %ComputeLoop5871
  %.lcssa8915 = phi i32 [ %8386, %ComputeLoop5871 ]
  %8391 = trunc i64 %4706 to i32
  %8392 = lshr i64 %4706, 32
  %8393 = trunc i64 %8392 to i32
  %8394 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %8391, i32 0)
  %8395 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %8393, i32 %8394)
  %8396 = icmp eq i32 %8395, 0
  %8397 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %8396)
  %8398 = extractvalue { i1, i64 } %8397, 0
  %8399 = extractvalue { i1, i64 } %8397, 1
  br i1 %8398, label %4707, label %4709

ComputeLoop5875:                                  ; preds = %_ZN22__llvm_libc_22_0_0_git15write_to_stderrENS_3cpp11string_viewE.exit.i3516, %ComputeLoop5875
  %Accumulator5877 = phi i32 [ %8403, %ComputeLoop5875 ], [ -1, %_ZN22__llvm_libc_22_0_0_git15write_to_stderrENS_3cpp11string_viewE.exit.i3516 ]
  %ActiveBits5878 = phi i64 [ %8406, %ComputeLoop5875 ], [ %4804, %_ZN22__llvm_libc_22_0_0_git15write_to_stderrENS_3cpp11string_viewE.exit.i3516 ]
  %8400 = call i64 @llvm.cttz.i64(i64 %ActiveBits5878, i1 true)
  %8401 = trunc i64 %8400 to i32
  %8402 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i21.i.i3525, i32 %8401)
  %8403 = and i32 %Accumulator5877, %8402
  %8404 = shl i64 1, %8400
  %8405 = xor i64 %8404, -1
  %8406 = and i64 %ActiveBits5878, %8405
  %8407 = icmp eq i64 %8406, 0
  br i1 %8407, label %ComputeEnd5876, label %ComputeLoop5875, !amdgpu.uniform !42

ComputeEnd5876:                                   ; preds = %ComputeLoop5875
  %.lcssa8905 = phi i32 [ %8403, %ComputeLoop5875 ]
  %8408 = trunc i64 %4804 to i32
  %8409 = lshr i64 %4804, 32
  %8410 = trunc i64 %8409 to i32
  %8411 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %8408, i32 0)
  %8412 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %8410, i32 %8411)
  %8413 = icmp eq i32 %8412, 0
  %8414 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %8413)
  %8415 = extractvalue { i1, i64 } %8414, 0
  %8416 = extractvalue { i1, i64 } %8414, 1
  br i1 %8415, label %4805, label %4808

ComputeLoop5879:                                  ; preds = %for.cond.i.i1.i3556, %ComputeLoop5879
  %Accumulator5881 = phi i32 [ %8421, %ComputeLoop5879 ], [ 0, %for.cond.i.i1.i3556 ]
  %OldValuePhi5882 = phi i32 [ %8420, %ComputeLoop5879 ], [ poison, %for.cond.i.i1.i3556 ]
  %ActiveBits5883 = phi i64 [ %8424, %ComputeLoop5879 ], [ %4813, %for.cond.i.i1.i3556 ]
  %8417 = call i64 @llvm.cttz.i64(i64 %ActiveBits5883, i1 true)
  %8418 = trunc i64 %8417 to i32
  %8419 = call i32 @llvm.amdgcn.readlane.i32(i32 %shl.i.i.i.i.i3568, i32 %8418)
  %8420 = call i32 @llvm.amdgcn.writelane.i32(i32 %Accumulator5881, i32 %8418, i32 %OldValuePhi5882)
  %8421 = or i32 %Accumulator5881, %8419
  %8422 = shl i64 1, %8417
  %8423 = xor i64 %8422, -1
  %8424 = and i64 %ActiveBits5883, %8423
  %8425 = icmp eq i64 %8424, 0
  br i1 %8425, label %ComputeEnd5880, label %ComputeLoop5879, !amdgpu.uniform !42

ComputeEnd5880:                                   ; preds = %ComputeLoop5879
  %.lcssa8900 = phi i32 [ %8420, %ComputeLoop5879 ]
  %.lcssa8899 = phi i32 [ %8421, %ComputeLoop5879 ]
  %div8.i.i.i.i.i3564 = lshr i32 %spec.store.select.i.i.i3563, 5
  %idxprom.i.i.i.i.i3565 = zext nneg i32 %div8.i.i.i.i.i3564 to i64
  %8426 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i.i.i3565
  %8427 = trunc i64 %4813 to i32
  %8428 = lshr i64 %4813, 32
  %8429 = trunc i64 %8428 to i32
  %8430 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %8427, i32 0)
  %8431 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %8429, i32 %8430)
  %8432 = icmp eq i32 %8431, 0
  %8433 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %8432)
  %8434 = extractvalue { i1, i64 } %8433, 0
  %8435 = extractvalue { i1, i64 } %8433, 1
  br i1 %8434, label %4817, label %4819

Flow7851:                                         ; preds = %4837, %if.then.i.i.i.i3574
  %8436 = phi i1 [ false, %4837 ], [ true, %if.then.i.i.i.i3574 ]
  br i1 %8436, label %if.end14.i.i.i3645, label %Flow7852, !amdgpu.uniform !42

ComputeLoop5884:                                  ; preds = %if.then12.i.i.i3579, %ComputeLoop5884
  %Accumulator5886 = phi i32 [ %8440, %ComputeLoop5884 ], [ -1, %if.then12.i.i.i3579 ]
  %ActiveBits5887 = phi i64 [ %8443, %ComputeLoop5884 ], [ %4834, %if.then12.i.i.i3579 ]
  %8437 = call i64 @llvm.cttz.i64(i64 %ActiveBits5887, i1 true)
  %8438 = trunc i64 %8437 to i32
  %8439 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i.i.i3585, i32 %8438)
  %8440 = and i32 %Accumulator5886, %8439
  %8441 = shl i64 1, %8437
  %8442 = xor i64 %8441, -1
  %8443 = and i64 %ActiveBits5887, %8442
  %8444 = icmp eq i64 %8443, 0
  br i1 %8444, label %ComputeEnd5885, label %ComputeLoop5884, !amdgpu.uniform !42

ComputeEnd5885:                                   ; preds = %ComputeLoop5884
  %.lcssa8901 = phi i32 [ %8440, %ComputeLoop5884 ]
  %8445 = trunc i64 %4834 to i32
  %8446 = lshr i64 %4834, 32
  %8447 = trunc i64 %8446 to i32
  %8448 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %8445, i32 0)
  %8449 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %8447, i32 %8448)
  %8450 = icmp eq i32 %8449, 0
  %8451 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %8450)
  %8452 = extractvalue { i1, i64 } %8451, 0
  %8453 = extractvalue { i1, i64 } %8451, 1
  br i1 %8452, label %4835, label %4837

ComputeLoop5888:                                  ; preds = %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit3653, %ComputeLoop5888
  %Accumulator5890 = phi i32 [ %8457, %ComputeLoop5888 ], [ -1, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit3653 ]
  %ActiveBits5891 = phi i64 [ %8460, %ComputeLoop5888 ], [ %4914, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit3653 ]
  %8454 = call i64 @llvm.cttz.i64(i64 %ActiveBits5891, i1 true)
  %8455 = trunc i64 %8454 to i32
  %8456 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i21.i.i3644, i32 %8455)
  %8457 = and i32 %Accumulator5890, %8456
  %8458 = shl i64 1, %8454
  %8459 = xor i64 %8458, -1
  %8460 = and i64 %ActiveBits5891, %8459
  %8461 = icmp eq i64 %8460, 0
  br i1 %8461, label %ComputeEnd5889, label %ComputeLoop5888, !amdgpu.uniform !42

ComputeEnd5889:                                   ; preds = %ComputeLoop5888
  %.lcssa8895 = phi i32 [ %8457, %ComputeLoop5888 ]
  %8462 = trunc i64 %4914 to i32
  %8463 = lshr i64 %4914, 32
  %8464 = trunc i64 %8463 to i32
  %8465 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %8462, i32 0)
  %8466 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %8464, i32 %8465)
  %8467 = icmp eq i32 %8466, 0
  %8468 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %8467)
  %8469 = extractvalue { i1, i64 } %8468, 0
  %8470 = extractvalue { i1, i64 } %8468, 1
  br i1 %8469, label %4915, label %4918

ComputeLoop5892:                                  ; preds = %for.cond.i.i1.i3662, %ComputeLoop5892
  %Accumulator5894 = phi i32 [ %8475, %ComputeLoop5892 ], [ 0, %for.cond.i.i1.i3662 ]
  %OldValuePhi5895 = phi i32 [ %8474, %ComputeLoop5892 ], [ poison, %for.cond.i.i1.i3662 ]
  %ActiveBits5896 = phi i64 [ %8478, %ComputeLoop5892 ], [ %4923, %for.cond.i.i1.i3662 ]
  %8471 = call i64 @llvm.cttz.i64(i64 %ActiveBits5896, i1 true)
  %8472 = trunc i64 %8471 to i32
  %8473 = call i32 @llvm.amdgcn.readlane.i32(i32 %shl.i.i.i.i.i3674, i32 %8472)
  %8474 = call i32 @llvm.amdgcn.writelane.i32(i32 %Accumulator5894, i32 %8472, i32 %OldValuePhi5895)
  %8475 = or i32 %Accumulator5894, %8473
  %8476 = shl i64 1, %8471
  %8477 = xor i64 %8476, -1
  %8478 = and i64 %ActiveBits5896, %8477
  %8479 = icmp eq i64 %8478, 0
  br i1 %8479, label %ComputeEnd5893, label %ComputeLoop5892, !amdgpu.uniform !42

ComputeEnd5893:                                   ; preds = %ComputeLoop5892
  %.lcssa8890 = phi i32 [ %8474, %ComputeLoop5892 ]
  %.lcssa8889 = phi i32 [ %8475, %ComputeLoop5892 ]
  %div8.i.i.i.i.i3670 = lshr i32 %spec.store.select.i.i.i3669, 5
  %idxprom.i.i.i.i.i3671 = zext nneg i32 %div8.i.i.i.i.i3670 to i64
  %8480 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i.i.i3671
  %8481 = trunc i64 %4923 to i32
  %8482 = lshr i64 %4923, 32
  %8483 = trunc i64 %8482 to i32
  %8484 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %8481, i32 0)
  %8485 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %8483, i32 %8484)
  %8486 = icmp eq i32 %8485, 0
  %8487 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %8486)
  %8488 = extractvalue { i1, i64 } %8487, 0
  %8489 = extractvalue { i1, i64 } %8487, 1
  br i1 %8488, label %4927, label %4929

Flow7832:                                         ; preds = %4947, %if.then.i.i.i.i3680
  %8490 = phi i1 [ false, %4947 ], [ true, %if.then.i.i.i.i3680 ]
  br i1 %8490, label %if.end14.i.i.i3751, label %Flow7833, !amdgpu.uniform !42

ComputeLoop5897:                                  ; preds = %if.then12.i.i.i3685, %ComputeLoop5897
  %Accumulator5899 = phi i32 [ %8494, %ComputeLoop5897 ], [ -1, %if.then12.i.i.i3685 ]
  %ActiveBits5900 = phi i64 [ %8497, %ComputeLoop5897 ], [ %4944, %if.then12.i.i.i3685 ]
  %8491 = call i64 @llvm.cttz.i64(i64 %ActiveBits5900, i1 true)
  %8492 = trunc i64 %8491 to i32
  %8493 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i.i.i3691, i32 %8492)
  %8494 = and i32 %Accumulator5899, %8493
  %8495 = shl i64 1, %8491
  %8496 = xor i64 %8495, -1
  %8497 = and i64 %ActiveBits5900, %8496
  %8498 = icmp eq i64 %8497, 0
  br i1 %8498, label %ComputeEnd5898, label %ComputeLoop5897, !amdgpu.uniform !42

ComputeEnd5898:                                   ; preds = %ComputeLoop5897
  %.lcssa8891 = phi i32 [ %8494, %ComputeLoop5897 ]
  %8499 = trunc i64 %4944 to i32
  %8500 = lshr i64 %4944, 32
  %8501 = trunc i64 %8500 to i32
  %8502 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %8499, i32 0)
  %8503 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %8501, i32 %8502)
  %8504 = icmp eq i32 %8503, 0
  %8505 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %8504)
  %8506 = extractvalue { i1, i64 } %8505, 0
  %8507 = extractvalue { i1, i64 } %8505, 1
  br i1 %8506, label %4945, label %4947

ComputeLoop5901:                                  ; preds = %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit3759, %ComputeLoop5901
  %Accumulator5903 = phi i32 [ %8511, %ComputeLoop5901 ], [ -1, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit3759 ]
  %ActiveBits5904 = phi i64 [ %8514, %ComputeLoop5901 ], [ %5024, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit3759 ]
  %8508 = call i64 @llvm.cttz.i64(i64 %ActiveBits5904, i1 true)
  %8509 = trunc i64 %8508 to i32
  %8510 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i21.i.i3750, i32 %8509)
  %8511 = and i32 %Accumulator5903, %8510
  %8512 = shl i64 1, %8508
  %8513 = xor i64 %8512, -1
  %8514 = and i64 %ActiveBits5904, %8513
  %8515 = icmp eq i64 %8514, 0
  br i1 %8515, label %ComputeEnd5902, label %ComputeLoop5901, !amdgpu.uniform !42

ComputeEnd5902:                                   ; preds = %ComputeLoop5901
  %.lcssa8885 = phi i32 [ %8511, %ComputeLoop5901 ]
  %8516 = trunc i64 %5024 to i32
  %8517 = lshr i64 %5024, 32
  %8518 = trunc i64 %8517 to i32
  %8519 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %8516, i32 0)
  %8520 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %8518, i32 %8519)
  %8521 = icmp eq i32 %8520, 0
  %8522 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %8521)
  %8523 = extractvalue { i1, i64 } %8522, 0
  %8524 = extractvalue { i1, i64 } %8522, 1
  br i1 %8523, label %5025, label %5028

ComputeLoop5905:                                  ; preds = %for.cond.i.i.i3828, %ComputeLoop5905
  %Accumulator5907 = phi i32 [ %8529, %ComputeLoop5905 ], [ 0, %for.cond.i.i.i3828 ]
  %OldValuePhi5908 = phi i32 [ %8528, %ComputeLoop5905 ], [ poison, %for.cond.i.i.i3828 ]
  %ActiveBits5909 = phi i64 [ %8532, %ComputeLoop5905 ], [ %5161, %for.cond.i.i.i3828 ]
  %8525 = call i64 @llvm.cttz.i64(i64 %ActiveBits5909, i1 true)
  %8526 = trunc i64 %8525 to i32
  %8527 = call i32 @llvm.amdgcn.readlane.i32(i32 %shl.i.i.i.i.i3840, i32 %8526)
  %8528 = call i32 @llvm.amdgcn.writelane.i32(i32 %Accumulator5907, i32 %8526, i32 %OldValuePhi5908)
  %8529 = or i32 %Accumulator5907, %8527
  %8530 = shl i64 1, %8525
  %8531 = xor i64 %8530, -1
  %8532 = and i64 %ActiveBits5909, %8531
  %8533 = icmp eq i64 %8532, 0
  br i1 %8533, label %ComputeEnd5906, label %ComputeLoop5905, !amdgpu.uniform !42

ComputeEnd5906:                                   ; preds = %ComputeLoop5905
  %.lcssa8876 = phi i32 [ %8528, %ComputeLoop5905 ]
  %.lcssa8875 = phi i32 [ %8529, %ComputeLoop5905 ]
  %div8.i.i.i.i.i3836 = lshr i32 %spec.store.select.i.i.i3835, 5
  %idxprom.i.i.i.i.i3837 = zext nneg i32 %div8.i.i.i.i.i3836 to i64
  %8534 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i.i.i3837
  %8535 = trunc i64 %5161 to i32
  %8536 = lshr i64 %5161, 32
  %8537 = trunc i64 %8536 to i32
  %8538 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %8535, i32 0)
  %8539 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %8537, i32 %8538)
  %8540 = icmp eq i32 %8539, 0
  %8541 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %8540)
  %8542 = extractvalue { i1, i64 } %8541, 0
  %8543 = extractvalue { i1, i64 } %8541, 1
  br i1 %8542, label %5165, label %5167

Flow7787:                                         ; preds = %5185, %if.then.i.i.i.i3846
  %8544 = phi i1 [ false, %5185 ], [ true, %if.then.i.i.i.i3846 ]
  br i1 %8544, label %if.end14.i.i.i3918, label %Flow7788, !amdgpu.uniform !42

ComputeLoop5910:                                  ; preds = %if.then12.i.i.i3851, %ComputeLoop5910
  %Accumulator5912 = phi i32 [ %8548, %ComputeLoop5910 ], [ -1, %if.then12.i.i.i3851 ]
  %ActiveBits5913 = phi i64 [ %8551, %ComputeLoop5910 ], [ %5182, %if.then12.i.i.i3851 ]
  %8545 = call i64 @llvm.cttz.i64(i64 %ActiveBits5913, i1 true)
  %8546 = trunc i64 %8545 to i32
  %8547 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i.i.i3857, i32 %8546)
  %8548 = and i32 %Accumulator5912, %8547
  %8549 = shl i64 1, %8545
  %8550 = xor i64 %8549, -1
  %8551 = and i64 %ActiveBits5913, %8550
  %8552 = icmp eq i64 %8551, 0
  br i1 %8552, label %ComputeEnd5911, label %ComputeLoop5910, !amdgpu.uniform !42

ComputeEnd5911:                                   ; preds = %ComputeLoop5910
  %.lcssa8877 = phi i32 [ %8548, %ComputeLoop5910 ]
  %8553 = trunc i64 %5182 to i32
  %8554 = lshr i64 %5182, 32
  %8555 = trunc i64 %8554 to i32
  %8556 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %8553, i32 0)
  %8557 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %8555, i32 %8556)
  %8558 = icmp eq i32 %8557, 0
  %8559 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %8558)
  %8560 = extractvalue { i1, i64 } %8559, 0
  %8561 = extractvalue { i1, i64 } %8559, 1
  br i1 %8560, label %5183, label %5185

ComputeLoop5914:                                  ; preds = %_ZN22__llvm_libc_22_0_0_git15write_to_stderrENS_3cpp11string_viewE.exit.i3907, %ComputeLoop5914
  %Accumulator5916 = phi i32 [ %8565, %ComputeLoop5914 ], [ -1, %_ZN22__llvm_libc_22_0_0_git15write_to_stderrENS_3cpp11string_viewE.exit.i3907 ]
  %ActiveBits5917 = phi i64 [ %8568, %ComputeLoop5914 ], [ %5280, %_ZN22__llvm_libc_22_0_0_git15write_to_stderrENS_3cpp11string_viewE.exit.i3907 ]
  %8562 = call i64 @llvm.cttz.i64(i64 %ActiveBits5917, i1 true)
  %8563 = trunc i64 %8562 to i32
  %8564 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i21.i.i3916, i32 %8563)
  %8565 = and i32 %Accumulator5916, %8564
  %8566 = shl i64 1, %8562
  %8567 = xor i64 %8566, -1
  %8568 = and i64 %ActiveBits5917, %8567
  %8569 = icmp eq i64 %8568, 0
  br i1 %8569, label %ComputeEnd5915, label %ComputeLoop5914, !amdgpu.uniform !42

ComputeEnd5915:                                   ; preds = %ComputeLoop5914
  %.lcssa8867 = phi i32 [ %8565, %ComputeLoop5914 ]
  %8570 = trunc i64 %5280 to i32
  %8571 = lshr i64 %5280, 32
  %8572 = trunc i64 %8571 to i32
  %8573 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %8570, i32 0)
  %8574 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %8572, i32 %8573)
  %8575 = icmp eq i32 %8574, 0
  %8576 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %8575)
  %8577 = extractvalue { i1, i64 } %8576, 0
  %8578 = extractvalue { i1, i64 } %8576, 1
  br i1 %8577, label %5281, label %5284

ComputeLoop5918:                                  ; preds = %for.cond.i.i.i3942, %ComputeLoop5918
  %Accumulator5920 = phi i32 [ %8583, %ComputeLoop5918 ], [ 0, %for.cond.i.i.i3942 ]
  %OldValuePhi5921 = phi i32 [ %8582, %ComputeLoop5918 ], [ poison, %for.cond.i.i.i3942 ]
  %ActiveBits5922 = phi i64 [ %8586, %ComputeLoop5918 ], [ %5289, %for.cond.i.i.i3942 ]
  %8579 = call i64 @llvm.cttz.i64(i64 %ActiveBits5922, i1 true)
  %8580 = trunc i64 %8579 to i32
  %8581 = call i32 @llvm.amdgcn.readlane.i32(i32 %shl.i.i.i.i.i3954, i32 %8580)
  %8582 = call i32 @llvm.amdgcn.writelane.i32(i32 %Accumulator5920, i32 %8580, i32 %OldValuePhi5921)
  %8583 = or i32 %Accumulator5920, %8581
  %8584 = shl i64 1, %8579
  %8585 = xor i64 %8584, -1
  %8586 = and i64 %ActiveBits5922, %8585
  %8587 = icmp eq i64 %8586, 0
  br i1 %8587, label %ComputeEnd5919, label %ComputeLoop5918, !amdgpu.uniform !42

ComputeEnd5919:                                   ; preds = %ComputeLoop5918
  %.lcssa8862 = phi i32 [ %8582, %ComputeLoop5918 ]
  %.lcssa8861 = phi i32 [ %8583, %ComputeLoop5918 ]
  %div8.i.i.i.i.i3950 = lshr i32 %spec.store.select.i.i.i3949, 5
  %idxprom.i.i.i.i.i3951 = zext nneg i32 %div8.i.i.i.i.i3950 to i64
  %8588 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i.i.i3951
  %8589 = trunc i64 %5289 to i32
  %8590 = lshr i64 %5289, 32
  %8591 = trunc i64 %8590 to i32
  %8592 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %8589, i32 0)
  %8593 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %8591, i32 %8592)
  %8594 = icmp eq i32 %8593, 0
  %8595 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %8594)
  %8596 = extractvalue { i1, i64 } %8595, 0
  %8597 = extractvalue { i1, i64 } %8595, 1
  br i1 %8596, label %5293, label %5295

Flow7768:                                         ; preds = %5313, %if.then.i.i.i.i3960
  %8598 = phi i1 [ false, %5313 ], [ true, %if.then.i.i.i.i3960 ]
  br i1 %8598, label %if.end14.i.i.i4029, label %Flow7769, !amdgpu.uniform !42

ComputeLoop5923:                                  ; preds = %if.then12.i.i.i3965, %ComputeLoop5923
  %Accumulator5925 = phi i32 [ %8602, %ComputeLoop5923 ], [ -1, %if.then12.i.i.i3965 ]
  %ActiveBits5926 = phi i64 [ %8605, %ComputeLoop5923 ], [ %5310, %if.then12.i.i.i3965 ]
  %8599 = call i64 @llvm.cttz.i64(i64 %ActiveBits5926, i1 true)
  %8600 = trunc i64 %8599 to i32
  %8601 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i.i.i3971, i32 %8600)
  %8602 = and i32 %Accumulator5925, %8601
  %8603 = shl i64 1, %8599
  %8604 = xor i64 %8603, -1
  %8605 = and i64 %ActiveBits5926, %8604
  %8606 = icmp eq i64 %8605, 0
  br i1 %8606, label %ComputeEnd5924, label %ComputeLoop5923, !amdgpu.uniform !42

ComputeEnd5924:                                   ; preds = %ComputeLoop5923
  %.lcssa8863 = phi i32 [ %8602, %ComputeLoop5923 ]
  %8607 = trunc i64 %5310 to i32
  %8608 = lshr i64 %5310, 32
  %8609 = trunc i64 %8608 to i32
  %8610 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %8607, i32 0)
  %8611 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %8609, i32 %8610)
  %8612 = icmp eq i32 %8611, 0
  %8613 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %8612)
  %8614 = extractvalue { i1, i64 } %8613, 0
  %8615 = extractvalue { i1, i64 } %8613, 1
  br i1 %8614, label %5311, label %5313

ComputeLoop5927:                                  ; preds = %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIcEERS1_T_.exit4037, %ComputeLoop5927
  %Accumulator5929 = phi i32 [ %8619, %ComputeLoop5927 ], [ -1, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIcEERS1_T_.exit4037 ]
  %ActiveBits5930 = phi i64 [ %8622, %ComputeLoop5927 ], [ %5365, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIcEERS1_T_.exit4037 ]
  %8616 = call i64 @llvm.cttz.i64(i64 %ActiveBits5930, i1 true)
  %8617 = trunc i64 %8616 to i32
  %8618 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i21.i.i4028, i32 %8617)
  %8619 = and i32 %Accumulator5929, %8618
  %8620 = shl i64 1, %8616
  %8621 = xor i64 %8620, -1
  %8622 = and i64 %ActiveBits5930, %8621
  %8623 = icmp eq i64 %8622, 0
  br i1 %8623, label %ComputeEnd5928, label %ComputeLoop5927, !amdgpu.uniform !42

ComputeEnd5928:                                   ; preds = %ComputeLoop5927
  %.lcssa8857 = phi i32 [ %8619, %ComputeLoop5927 ]
  %8624 = trunc i64 %5365 to i32
  %8625 = lshr i64 %5365, 32
  %8626 = trunc i64 %8625 to i32
  %8627 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %8624, i32 0)
  %8628 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %8626, i32 %8627)
  %8629 = icmp eq i32 %8628, 0
  %8630 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %8629)
  %8631 = extractvalue { i1, i64 } %8630, 0
  %8632 = extractvalue { i1, i64 } %8630, 1
  br i1 %8631, label %5366, label %5369

ComputeLoop5931:                                  ; preds = %for.cond.i.i1.i4046, %ComputeLoop5931
  %Accumulator5933 = phi i32 [ %8637, %ComputeLoop5931 ], [ 0, %for.cond.i.i1.i4046 ]
  %OldValuePhi5934 = phi i32 [ %8636, %ComputeLoop5931 ], [ poison, %for.cond.i.i1.i4046 ]
  %ActiveBits5935 = phi i64 [ %8640, %ComputeLoop5931 ], [ %5374, %for.cond.i.i1.i4046 ]
  %8633 = call i64 @llvm.cttz.i64(i64 %ActiveBits5935, i1 true)
  %8634 = trunc i64 %8633 to i32
  %8635 = call i32 @llvm.amdgcn.readlane.i32(i32 %shl.i.i.i.i.i4058, i32 %8634)
  %8636 = call i32 @llvm.amdgcn.writelane.i32(i32 %Accumulator5933, i32 %8634, i32 %OldValuePhi5934)
  %8637 = or i32 %Accumulator5933, %8635
  %8638 = shl i64 1, %8633
  %8639 = xor i64 %8638, -1
  %8640 = and i64 %ActiveBits5935, %8639
  %8641 = icmp eq i64 %8640, 0
  br i1 %8641, label %ComputeEnd5932, label %ComputeLoop5931, !amdgpu.uniform !42

ComputeEnd5932:                                   ; preds = %ComputeLoop5931
  %.lcssa8852 = phi i32 [ %8636, %ComputeLoop5931 ]
  %.lcssa8851 = phi i32 [ %8637, %ComputeLoop5931 ]
  %div8.i.i.i.i.i4054 = lshr i32 %spec.store.select.i.i.i4053, 5
  %idxprom.i.i.i.i.i4055 = zext nneg i32 %div8.i.i.i.i.i4054 to i64
  %8642 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i.i.i4055
  %8643 = trunc i64 %5374 to i32
  %8644 = lshr i64 %5374, 32
  %8645 = trunc i64 %8644 to i32
  %8646 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %8643, i32 0)
  %8647 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %8645, i32 %8646)
  %8648 = icmp eq i32 %8647, 0
  %8649 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %8648)
  %8650 = extractvalue { i1, i64 } %8649, 0
  %8651 = extractvalue { i1, i64 } %8649, 1
  br i1 %8650, label %5378, label %5380

Flow7759:                                         ; preds = %5398, %if.then.i.i.i.i4064
  %8652 = phi i1 [ false, %5398 ], [ true, %if.then.i.i.i.i4064 ]
  br i1 %8652, label %if.end14.i.i.i4135, label %Flow7760, !amdgpu.uniform !42

ComputeLoop5936:                                  ; preds = %if.then12.i.i.i4069, %ComputeLoop5936
  %Accumulator5938 = phi i32 [ %8656, %ComputeLoop5936 ], [ -1, %if.then12.i.i.i4069 ]
  %ActiveBits5939 = phi i64 [ %8659, %ComputeLoop5936 ], [ %5395, %if.then12.i.i.i4069 ]
  %8653 = call i64 @llvm.cttz.i64(i64 %ActiveBits5939, i1 true)
  %8654 = trunc i64 %8653 to i32
  %8655 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i.i.i4075, i32 %8654)
  %8656 = and i32 %Accumulator5938, %8655
  %8657 = shl i64 1, %8653
  %8658 = xor i64 %8657, -1
  %8659 = and i64 %ActiveBits5939, %8658
  %8660 = icmp eq i64 %8659, 0
  br i1 %8660, label %ComputeEnd5937, label %ComputeLoop5936, !amdgpu.uniform !42

ComputeEnd5937:                                   ; preds = %ComputeLoop5936
  %.lcssa8853 = phi i32 [ %8656, %ComputeLoop5936 ]
  %8661 = trunc i64 %5395 to i32
  %8662 = lshr i64 %5395, 32
  %8663 = trunc i64 %8662 to i32
  %8664 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %8661, i32 0)
  %8665 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %8663, i32 %8664)
  %8666 = icmp eq i32 %8665, 0
  %8667 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %8666)
  %8668 = extractvalue { i1, i64 } %8667, 0
  %8669 = extractvalue { i1, i64 } %8667, 1
  br i1 %8668, label %5396, label %5398

ComputeLoop5940:                                  ; preds = %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit4143, %ComputeLoop5940
  %Accumulator5942 = phi i32 [ %8673, %ComputeLoop5940 ], [ -1, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit4143 ]
  %ActiveBits5943 = phi i64 [ %8676, %ComputeLoop5940 ], [ %5475, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit4143 ]
  %8670 = call i64 @llvm.cttz.i64(i64 %ActiveBits5943, i1 true)
  %8671 = trunc i64 %8670 to i32
  %8672 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i21.i.i4134, i32 %8671)
  %8673 = and i32 %Accumulator5942, %8672
  %8674 = shl i64 1, %8670
  %8675 = xor i64 %8674, -1
  %8676 = and i64 %ActiveBits5943, %8675
  %8677 = icmp eq i64 %8676, 0
  br i1 %8677, label %ComputeEnd5941, label %ComputeLoop5940, !amdgpu.uniform !42

ComputeEnd5941:                                   ; preds = %ComputeLoop5940
  %.lcssa8847 = phi i32 [ %8673, %ComputeLoop5940 ]
  %8678 = trunc i64 %5475 to i32
  %8679 = lshr i64 %5475, 32
  %8680 = trunc i64 %8679 to i32
  %8681 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %8678, i32 0)
  %8682 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %8680, i32 %8681)
  %8683 = icmp eq i32 %8682, 0
  %8684 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %8683)
  %8685 = extractvalue { i1, i64 } %8684, 0
  %8686 = extractvalue { i1, i64 } %8684, 1
  br i1 %8685, label %5476, label %5479

ComputeLoop5944:                                  ; preds = %for.cond.i.i.i4212, %ComputeLoop5944
  %Accumulator5946 = phi i32 [ %8691, %ComputeLoop5944 ], [ 0, %for.cond.i.i.i4212 ]
  %OldValuePhi5947 = phi i32 [ %8690, %ComputeLoop5944 ], [ poison, %for.cond.i.i.i4212 ]
  %ActiveBits5948 = phi i64 [ %8694, %ComputeLoop5944 ], [ %5612, %for.cond.i.i.i4212 ]
  %8687 = call i64 @llvm.cttz.i64(i64 %ActiveBits5948, i1 true)
  %8688 = trunc i64 %8687 to i32
  %8689 = call i32 @llvm.amdgcn.readlane.i32(i32 %shl.i.i.i.i.i4224, i32 %8688)
  %8690 = call i32 @llvm.amdgcn.writelane.i32(i32 %Accumulator5946, i32 %8688, i32 %OldValuePhi5947)
  %8691 = or i32 %Accumulator5946, %8689
  %8692 = shl i64 1, %8687
  %8693 = xor i64 %8692, -1
  %8694 = and i64 %ActiveBits5948, %8693
  %8695 = icmp eq i64 %8694, 0
  br i1 %8695, label %ComputeEnd5945, label %ComputeLoop5944, !amdgpu.uniform !42

ComputeEnd5945:                                   ; preds = %ComputeLoop5944
  %.lcssa8838 = phi i32 [ %8690, %ComputeLoop5944 ]
  %.lcssa8837 = phi i32 [ %8691, %ComputeLoop5944 ]
  %div8.i.i.i.i.i4220 = lshr i32 %spec.store.select.i.i.i4219, 5
  %idxprom.i.i.i.i.i4221 = zext nneg i32 %div8.i.i.i.i.i4220 to i64
  %8696 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i.i.i4221
  %8697 = trunc i64 %5612 to i32
  %8698 = lshr i64 %5612, 32
  %8699 = trunc i64 %8698 to i32
  %8700 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %8697, i32 0)
  %8701 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %8699, i32 %8700)
  %8702 = icmp eq i32 %8701, 0
  %8703 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %8702)
  %8704 = extractvalue { i1, i64 } %8703, 0
  %8705 = extractvalue { i1, i64 } %8703, 1
  br i1 %8704, label %5616, label %5618

Flow7714:                                         ; preds = %5636, %if.then.i.i.i.i4230
  %8706 = phi i1 [ false, %5636 ], [ true, %if.then.i.i.i.i4230 ]
  br i1 %8706, label %if.end14.i.i.i4302, label %Flow7715, !amdgpu.uniform !42

ComputeLoop5949:                                  ; preds = %if.then12.i.i.i4235, %ComputeLoop5949
  %Accumulator5951 = phi i32 [ %8710, %ComputeLoop5949 ], [ -1, %if.then12.i.i.i4235 ]
  %ActiveBits5952 = phi i64 [ %8713, %ComputeLoop5949 ], [ %5633, %if.then12.i.i.i4235 ]
  %8707 = call i64 @llvm.cttz.i64(i64 %ActiveBits5952, i1 true)
  %8708 = trunc i64 %8707 to i32
  %8709 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i.i.i4241, i32 %8708)
  %8710 = and i32 %Accumulator5951, %8709
  %8711 = shl i64 1, %8707
  %8712 = xor i64 %8711, -1
  %8713 = and i64 %ActiveBits5952, %8712
  %8714 = icmp eq i64 %8713, 0
  br i1 %8714, label %ComputeEnd5950, label %ComputeLoop5949, !amdgpu.uniform !42

ComputeEnd5950:                                   ; preds = %ComputeLoop5949
  %.lcssa8839 = phi i32 [ %8710, %ComputeLoop5949 ]
  %8715 = trunc i64 %5633 to i32
  %8716 = lshr i64 %5633, 32
  %8717 = trunc i64 %8716 to i32
  %8718 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %8715, i32 0)
  %8719 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %8717, i32 %8718)
  %8720 = icmp eq i32 %8719, 0
  %8721 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %8720)
  %8722 = extractvalue { i1, i64 } %8721, 0
  %8723 = extractvalue { i1, i64 } %8721, 1
  br i1 %8722, label %5634, label %5636

ComputeLoop5953:                                  ; preds = %_ZN22__llvm_libc_22_0_0_git15write_to_stderrENS_3cpp11string_viewE.exit.i4291, %ComputeLoop5953
  %Accumulator5955 = phi i32 [ %8727, %ComputeLoop5953 ], [ -1, %_ZN22__llvm_libc_22_0_0_git15write_to_stderrENS_3cpp11string_viewE.exit.i4291 ]
  %ActiveBits5956 = phi i64 [ %8730, %ComputeLoop5953 ], [ %5731, %_ZN22__llvm_libc_22_0_0_git15write_to_stderrENS_3cpp11string_viewE.exit.i4291 ]
  %8724 = call i64 @llvm.cttz.i64(i64 %ActiveBits5956, i1 true)
  %8725 = trunc i64 %8724 to i32
  %8726 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i21.i.i4300, i32 %8725)
  %8727 = and i32 %Accumulator5955, %8726
  %8728 = shl i64 1, %8724
  %8729 = xor i64 %8728, -1
  %8730 = and i64 %ActiveBits5956, %8729
  %8731 = icmp eq i64 %8730, 0
  br i1 %8731, label %ComputeEnd5954, label %ComputeLoop5953, !amdgpu.uniform !42

ComputeEnd5954:                                   ; preds = %ComputeLoop5953
  %.lcssa8829 = phi i32 [ %8727, %ComputeLoop5953 ]
  %8732 = trunc i64 %5731 to i32
  %8733 = lshr i64 %5731, 32
  %8734 = trunc i64 %8733 to i32
  %8735 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %8732, i32 0)
  %8736 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %8734, i32 %8735)
  %8737 = icmp eq i32 %8736, 0
  %8738 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %8737)
  %8739 = extractvalue { i1, i64 } %8738, 0
  %8740 = extractvalue { i1, i64 } %8738, 1
  br i1 %8739, label %5732, label %5735

ComputeLoop5957:                                  ; preds = %for.cond.i.i.i4326, %ComputeLoop5957
  %Accumulator5959 = phi i32 [ %8745, %ComputeLoop5957 ], [ 0, %for.cond.i.i.i4326 ]
  %OldValuePhi5960 = phi i32 [ %8744, %ComputeLoop5957 ], [ poison, %for.cond.i.i.i4326 ]
  %ActiveBits5961 = phi i64 [ %8748, %ComputeLoop5957 ], [ %5740, %for.cond.i.i.i4326 ]
  %8741 = call i64 @llvm.cttz.i64(i64 %ActiveBits5961, i1 true)
  %8742 = trunc i64 %8741 to i32
  %8743 = call i32 @llvm.amdgcn.readlane.i32(i32 %shl.i.i.i.i.i4338, i32 %8742)
  %8744 = call i32 @llvm.amdgcn.writelane.i32(i32 %Accumulator5959, i32 %8742, i32 %OldValuePhi5960)
  %8745 = or i32 %Accumulator5959, %8743
  %8746 = shl i64 1, %8741
  %8747 = xor i64 %8746, -1
  %8748 = and i64 %ActiveBits5961, %8747
  %8749 = icmp eq i64 %8748, 0
  br i1 %8749, label %ComputeEnd5958, label %ComputeLoop5957, !amdgpu.uniform !42

ComputeEnd5958:                                   ; preds = %ComputeLoop5957
  %.lcssa8824 = phi i32 [ %8744, %ComputeLoop5957 ]
  %.lcssa8823 = phi i32 [ %8745, %ComputeLoop5957 ]
  %div8.i.i.i.i.i4334 = lshr i32 %spec.store.select.i.i.i4333, 5
  %idxprom.i.i.i.i.i4335 = zext nneg i32 %div8.i.i.i.i.i4334 to i64
  %8750 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i.i.i4335
  %8751 = trunc i64 %5740 to i32
  %8752 = lshr i64 %5740, 32
  %8753 = trunc i64 %8752 to i32
  %8754 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %8751, i32 0)
  %8755 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %8753, i32 %8754)
  %8756 = icmp eq i32 %8755, 0
  %8757 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %8756)
  %8758 = extractvalue { i1, i64 } %8757, 0
  %8759 = extractvalue { i1, i64 } %8757, 1
  br i1 %8758, label %5744, label %5746

Flow7695:                                         ; preds = %5764, %if.then.i.i.i.i4344
  %8760 = phi i1 [ false, %5764 ], [ true, %if.then.i.i.i.i4344 ]
  br i1 %8760, label %if.end14.i.i.i4413, label %Flow7696, !amdgpu.uniform !42

ComputeLoop5962:                                  ; preds = %if.then12.i.i.i4349, %ComputeLoop5962
  %Accumulator5964 = phi i32 [ %8764, %ComputeLoop5962 ], [ -1, %if.then12.i.i.i4349 ]
  %ActiveBits5965 = phi i64 [ %8767, %ComputeLoop5962 ], [ %5761, %if.then12.i.i.i4349 ]
  %8761 = call i64 @llvm.cttz.i64(i64 %ActiveBits5965, i1 true)
  %8762 = trunc i64 %8761 to i32
  %8763 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i.i.i4355, i32 %8762)
  %8764 = and i32 %Accumulator5964, %8763
  %8765 = shl i64 1, %8761
  %8766 = xor i64 %8765, -1
  %8767 = and i64 %ActiveBits5965, %8766
  %8768 = icmp eq i64 %8767, 0
  br i1 %8768, label %ComputeEnd5963, label %ComputeLoop5962, !amdgpu.uniform !42

ComputeEnd5963:                                   ; preds = %ComputeLoop5962
  %.lcssa8825 = phi i32 [ %8764, %ComputeLoop5962 ]
  %8769 = trunc i64 %5761 to i32
  %8770 = lshr i64 %5761, 32
  %8771 = trunc i64 %8770 to i32
  %8772 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %8769, i32 0)
  %8773 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %8771, i32 %8772)
  %8774 = icmp eq i32 %8773, 0
  %8775 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %8774)
  %8776 = extractvalue { i1, i64 } %8775, 0
  %8777 = extractvalue { i1, i64 } %8775, 1
  br i1 %8776, label %5762, label %5764

ComputeLoop5966:                                  ; preds = %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIcEERS1_T_.exit4421, %ComputeLoop5966
  %Accumulator5968 = phi i32 [ %8781, %ComputeLoop5966 ], [ -1, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIcEERS1_T_.exit4421 ]
  %ActiveBits5969 = phi i64 [ %8784, %ComputeLoop5966 ], [ %5816, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIcEERS1_T_.exit4421 ]
  %8778 = call i64 @llvm.cttz.i64(i64 %ActiveBits5969, i1 true)
  %8779 = trunc i64 %8778 to i32
  %8780 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i21.i.i4412, i32 %8779)
  %8781 = and i32 %Accumulator5968, %8780
  %8782 = shl i64 1, %8778
  %8783 = xor i64 %8782, -1
  %8784 = and i64 %ActiveBits5969, %8783
  %8785 = icmp eq i64 %8784, 0
  br i1 %8785, label %ComputeEnd5967, label %ComputeLoop5966, !amdgpu.uniform !42

ComputeEnd5967:                                   ; preds = %ComputeLoop5966
  %.lcssa = phi i32 [ %8781, %ComputeLoop5966 ]
  %8786 = trunc i64 %5816 to i32
  %8787 = lshr i64 %5816, 32
  %8788 = trunc i64 %8787 to i32
  %8789 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %8786, i32 0)
  %8790 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %8788, i32 %8789)
  %8791 = icmp eq i32 %8790, 0
  %8792 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %8791)
  %8793 = extractvalue { i1, i64 } %8792, 0
  %8794 = extractvalue { i1, i64 } %8792, 1
  br i1 %8793, label %5817, label %5820

ComputeLoop5970:                                  ; preds = %for.cond.i.i1.i4430, %ComputeLoop5970
  %Accumulator5972 = phi i32 [ %8799, %ComputeLoop5970 ], [ 0, %for.cond.i.i1.i4430 ]
  %OldValuePhi5973 = phi i32 [ %8798, %ComputeLoop5970 ], [ poison, %for.cond.i.i1.i4430 ]
  %ActiveBits5974 = phi i64 [ %8802, %ComputeLoop5970 ], [ %5828, %for.cond.i.i1.i4430 ]
  %8795 = call i64 @llvm.cttz.i64(i64 %ActiveBits5974, i1 true)
  %8796 = trunc i64 %8795 to i32
  %8797 = call i32 @llvm.amdgcn.readlane.i32(i32 %shl.i.i.i.i.i4442, i32 %8796)
  %8798 = call i32 @llvm.amdgcn.writelane.i32(i32 %Accumulator5972, i32 %8796, i32 %OldValuePhi5973)
  %8799 = or i32 %Accumulator5972, %8797
  %8800 = shl i64 1, %8795
  %8801 = xor i64 %8800, -1
  %8802 = and i64 %ActiveBits5974, %8801
  %8803 = icmp eq i64 %8802, 0
  br i1 %8803, label %ComputeEnd5971, label %ComputeLoop5970, !amdgpu.uniform !42

ComputeEnd5971:                                   ; preds = %ComputeLoop5970
  %.lcssa8963 = phi i32 [ %8798, %ComputeLoop5970 ]
  %.lcssa8962 = phi i32 [ %8799, %ComputeLoop5970 ]
  %div8.i.i.i.i.i4438 = lshr i32 %spec.store.select.i.i.i4437, 5
  %idxprom.i.i.i.i.i4439 = zext nneg i32 %div8.i.i.i.i.i4438 to i64
  %8804 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i.i.i4439
  %8805 = trunc i64 %5828 to i32
  %8806 = lshr i64 %5828, 32
  %8807 = trunc i64 %8806 to i32
  %8808 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %8805, i32 0)
  %8809 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %8807, i32 %8808)
  %8810 = icmp eq i32 %8809, 0
  %8811 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %8810)
  %8812 = extractvalue { i1, i64 } %8811, 0
  %8813 = extractvalue { i1, i64 } %8811, 1
  br i1 %8812, label %5832, label %5834

Flow7686:                                         ; preds = %5852, %if.then.i.i.i.i4448
  %8814 = phi i1 [ false, %5852 ], [ true, %if.then.i.i.i.i4448 ]
  br i1 %8814, label %if.end14.i.i.i4519, label %Flow7687, !amdgpu.uniform !42

ComputeLoop5975:                                  ; preds = %if.then12.i.i.i4453, %ComputeLoop5975
  %Accumulator5977 = phi i32 [ %8818, %ComputeLoop5975 ], [ -1, %if.then12.i.i.i4453 ]
  %ActiveBits5978 = phi i64 [ %8821, %ComputeLoop5975 ], [ %5849, %if.then12.i.i.i4453 ]
  %8815 = call i64 @llvm.cttz.i64(i64 %ActiveBits5978, i1 true)
  %8816 = trunc i64 %8815 to i32
  %8817 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i.i.i4459, i32 %8816)
  %8818 = and i32 %Accumulator5977, %8817
  %8819 = shl i64 1, %8815
  %8820 = xor i64 %8819, -1
  %8821 = and i64 %ActiveBits5978, %8820
  %8822 = icmp eq i64 %8821, 0
  br i1 %8822, label %ComputeEnd5976, label %ComputeLoop5975, !amdgpu.uniform !42

ComputeEnd5976:                                   ; preds = %ComputeLoop5975
  %.lcssa8964 = phi i32 [ %8818, %ComputeLoop5975 ]
  %8823 = trunc i64 %5849 to i32
  %8824 = lshr i64 %5849, 32
  %8825 = trunc i64 %8824 to i32
  %8826 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %8823, i32 0)
  %8827 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %8825, i32 %8826)
  %8828 = icmp eq i32 %8827, 0
  %8829 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %8828)
  %8830 = extractvalue { i1, i64 } %8829, 0
  %8831 = extractvalue { i1, i64 } %8829, 1
  br i1 %8830, label %5850, label %5852

ComputeLoop5979:                                  ; preds = %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit4527, %ComputeLoop5979
  %Accumulator5981 = phi i32 [ %8835, %ComputeLoop5979 ], [ -1, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit4527 ]
  %ActiveBits5982 = phi i64 [ %8838, %ComputeLoop5979 ], [ %5931, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit4527 ]
  %8832 = call i64 @llvm.cttz.i64(i64 %ActiveBits5982, i1 true)
  %8833 = trunc i64 %8832 to i32
  %8834 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i21.i.i4518, i32 %8833)
  %8835 = and i32 %Accumulator5981, %8834
  %8836 = shl i64 1, %8832
  %8837 = xor i64 %8836, -1
  %8838 = and i64 %ActiveBits5982, %8837
  %8839 = icmp eq i64 %8838, 0
  br i1 %8839, label %ComputeEnd5980, label %ComputeLoop5979, !amdgpu.uniform !42

ComputeEnd5980:                                   ; preds = %ComputeLoop5979
  %.lcssa8958 = phi i32 [ %8835, %ComputeLoop5979 ]
  %8840 = trunc i64 %5931 to i32
  %8841 = lshr i64 %5931, 32
  %8842 = trunc i64 %8841 to i32
  %8843 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %8840, i32 0)
  %8844 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %8842, i32 %8843)
  %8845 = icmp eq i32 %8844, 0
  %8846 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %8845)
  %8847 = extractvalue { i1, i64 } %8846, 0
  %8848 = extractvalue { i1, i64 } %8846, 1
  br i1 %8847, label %5932, label %5935

ComputeLoop5983:                                  ; preds = %for.cond.i.i1.i4536, %ComputeLoop5983
  %Accumulator5985 = phi i32 [ %8853, %ComputeLoop5983 ], [ 0, %for.cond.i.i1.i4536 ]
  %OldValuePhi5986 = phi i32 [ %8852, %ComputeLoop5983 ], [ poison, %for.cond.i.i1.i4536 ]
  %ActiveBits5987 = phi i64 [ %8856, %ComputeLoop5983 ], [ %5944, %for.cond.i.i1.i4536 ]
  %8849 = call i64 @llvm.cttz.i64(i64 %ActiveBits5987, i1 true)
  %8850 = trunc i64 %8849 to i32
  %8851 = call i32 @llvm.amdgcn.readlane.i32(i32 %shl.i.i.i.i.i4548, i32 %8850)
  %8852 = call i32 @llvm.amdgcn.writelane.i32(i32 %Accumulator5985, i32 %8850, i32 %OldValuePhi5986)
  %8853 = or i32 %Accumulator5985, %8851
  %8854 = shl i64 1, %8849
  %8855 = xor i64 %8854, -1
  %8856 = and i64 %ActiveBits5987, %8855
  %8857 = icmp eq i64 %8856, 0
  br i1 %8857, label %ComputeEnd5984, label %ComputeLoop5983, !amdgpu.uniform !42

ComputeEnd5984:                                   ; preds = %ComputeLoop5983
  %.lcssa8953 = phi i32 [ %8852, %ComputeLoop5983 ]
  %.lcssa8952 = phi i32 [ %8853, %ComputeLoop5983 ]
  %div8.i.i.i.i.i4544 = lshr i32 %spec.store.select.i.i.i4543, 5
  %idxprom.i.i.i.i.i4545 = zext nneg i32 %div8.i.i.i.i.i4544 to i64
  %8858 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i.i.i4545
  %8859 = trunc i64 %5944 to i32
  %8860 = lshr i64 %5944, 32
  %8861 = trunc i64 %8860 to i32
  %8862 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %8859, i32 0)
  %8863 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %8861, i32 %8862)
  %8864 = icmp eq i32 %8863, 0
  %8865 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %8864)
  %8866 = extractvalue { i1, i64 } %8865, 0
  %8867 = extractvalue { i1, i64 } %8865, 1
  br i1 %8866, label %5948, label %5950

Flow7666:                                         ; preds = %5968, %if.then.i.i.i.i4554
  %8868 = phi i1 [ false, %5968 ], [ true, %if.then.i.i.i.i4554 ]
  br i1 %8868, label %if.end14.i.i.i4625, label %Flow7667, !amdgpu.uniform !42

ComputeLoop5988:                                  ; preds = %if.then12.i.i.i4559, %ComputeLoop5988
  %Accumulator5990 = phi i32 [ %8872, %ComputeLoop5988 ], [ -1, %if.then12.i.i.i4559 ]
  %ActiveBits5991 = phi i64 [ %8875, %ComputeLoop5988 ], [ %5965, %if.then12.i.i.i4559 ]
  %8869 = call i64 @llvm.cttz.i64(i64 %ActiveBits5991, i1 true)
  %8870 = trunc i64 %8869 to i32
  %8871 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i.i.i4565, i32 %8870)
  %8872 = and i32 %Accumulator5990, %8871
  %8873 = shl i64 1, %8869
  %8874 = xor i64 %8873, -1
  %8875 = and i64 %ActiveBits5991, %8874
  %8876 = icmp eq i64 %8875, 0
  br i1 %8876, label %ComputeEnd5989, label %ComputeLoop5988, !amdgpu.uniform !42

ComputeEnd5989:                                   ; preds = %ComputeLoop5988
  %.lcssa8954 = phi i32 [ %8872, %ComputeLoop5988 ]
  %8877 = trunc i64 %5965 to i32
  %8878 = lshr i64 %5965, 32
  %8879 = trunc i64 %8878 to i32
  %8880 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %8877, i32 0)
  %8881 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %8879, i32 %8880)
  %8882 = icmp eq i32 %8881, 0
  %8883 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %8882)
  %8884 = extractvalue { i1, i64 } %8883, 0
  %8885 = extractvalue { i1, i64 } %8883, 1
  br i1 %8884, label %5966, label %5968

ComputeLoop5992:                                  ; preds = %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit4633, %ComputeLoop5992
  %Accumulator5994 = phi i32 [ %8889, %ComputeLoop5992 ], [ -1, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit4633 ]
  %ActiveBits5995 = phi i64 [ %8892, %ComputeLoop5992 ], [ %6045, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit4633 ]
  %8886 = call i64 @llvm.cttz.i64(i64 %ActiveBits5995, i1 true)
  %8887 = trunc i64 %8886 to i32
  %8888 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i21.i.i4624, i32 %8887)
  %8889 = and i32 %Accumulator5994, %8888
  %8890 = shl i64 1, %8886
  %8891 = xor i64 %8890, -1
  %8892 = and i64 %ActiveBits5995, %8891
  %8893 = icmp eq i64 %8892, 0
  br i1 %8893, label %ComputeEnd5993, label %ComputeLoop5992, !amdgpu.uniform !42

ComputeEnd5993:                                   ; preds = %ComputeLoop5992
  %.lcssa8948 = phi i32 [ %8889, %ComputeLoop5992 ]
  %8894 = trunc i64 %6045 to i32
  %8895 = lshr i64 %6045, 32
  %8896 = trunc i64 %8895 to i32
  %8897 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %8894, i32 0)
  %8898 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %8896, i32 %8897)
  %8899 = icmp eq i32 %8898, 0
  %8900 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %8899)
  %8901 = extractvalue { i1, i64 } %8900, 0
  %8902 = extractvalue { i1, i64 } %8900, 1
  br i1 %8901, label %6046, label %6049

ComputeLoop5996:                                  ; preds = %for.cond.i.i1.i4643, %ComputeLoop5996
  %Accumulator5998 = phi i32 [ %8907, %ComputeLoop5996 ], [ 0, %for.cond.i.i1.i4643 ]
  %OldValuePhi5999 = phi i32 [ %8906, %ComputeLoop5996 ], [ poison, %for.cond.i.i1.i4643 ]
  %ActiveBits6000 = phi i64 [ %8910, %ComputeLoop5996 ], [ %6058, %for.cond.i.i1.i4643 ]
  %8903 = call i64 @llvm.cttz.i64(i64 %ActiveBits6000, i1 true)
  %8904 = trunc i64 %8903 to i32
  %8905 = call i32 @llvm.amdgcn.readlane.i32(i32 %shl.i.i.i.i.i4655, i32 %8904)
  %8906 = call i32 @llvm.amdgcn.writelane.i32(i32 %Accumulator5998, i32 %8904, i32 %OldValuePhi5999)
  %8907 = or i32 %Accumulator5998, %8905
  %8908 = shl i64 1, %8903
  %8909 = xor i64 %8908, -1
  %8910 = and i64 %ActiveBits6000, %8909
  %8911 = icmp eq i64 %8910, 0
  br i1 %8911, label %ComputeEnd5997, label %ComputeLoop5996, !amdgpu.uniform !42

ComputeEnd5997:                                   ; preds = %ComputeLoop5996
  %.lcssa8942 = phi i32 [ %8906, %ComputeLoop5996 ]
  %.lcssa8941 = phi i32 [ %8907, %ComputeLoop5996 ]
  %div8.i.i.i.i.i4651 = lshr i32 %spec.store.select.i.i.i4650, 5
  %idxprom.i.i.i.i.i4652 = zext nneg i32 %div8.i.i.i.i.i4651 to i64
  %8912 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i.i.i4652
  %8913 = trunc i64 %6058 to i32
  %8914 = lshr i64 %6058, 32
  %8915 = trunc i64 %8914 to i32
  %8916 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %8913, i32 0)
  %8917 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %8915, i32 %8916)
  %8918 = icmp eq i32 %8917, 0
  %8919 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %8918)
  %8920 = extractvalue { i1, i64 } %8919, 0
  %8921 = extractvalue { i1, i64 } %8919, 1
  br i1 %8920, label %6062, label %6064

Flow7647:                                         ; preds = %6082, %if.then.i.i.i.i4661
  %8922 = phi i1 [ false, %6082 ], [ true, %if.then.i.i.i.i4661 ]
  br i1 %8922, label %if.end14.i.i.i4732, label %Flow7648, !amdgpu.uniform !42

ComputeLoop6001:                                  ; preds = %if.then12.i.i.i4666, %ComputeLoop6001
  %Accumulator6003 = phi i32 [ %8926, %ComputeLoop6001 ], [ -1, %if.then12.i.i.i4666 ]
  %ActiveBits6004 = phi i64 [ %8929, %ComputeLoop6001 ], [ %6079, %if.then12.i.i.i4666 ]
  %8923 = call i64 @llvm.cttz.i64(i64 %ActiveBits6004, i1 true)
  %8924 = trunc i64 %8923 to i32
  %8925 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i.i.i4672, i32 %8924)
  %8926 = and i32 %Accumulator6003, %8925
  %8927 = shl i64 1, %8923
  %8928 = xor i64 %8927, -1
  %8929 = and i64 %ActiveBits6004, %8928
  %8930 = icmp eq i64 %8929, 0
  br i1 %8930, label %ComputeEnd6002, label %ComputeLoop6001, !amdgpu.uniform !42

ComputeEnd6002:                                   ; preds = %ComputeLoop6001
  %.lcssa8943 = phi i32 [ %8926, %ComputeLoop6001 ]
  %8931 = trunc i64 %6079 to i32
  %8932 = lshr i64 %6079, 32
  %8933 = trunc i64 %8932 to i32
  %8934 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %8931, i32 0)
  %8935 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %8933, i32 %8934)
  %8936 = icmp eq i32 %8935, 0
  %8937 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %8936)
  %8938 = extractvalue { i1, i64 } %8937, 0
  %8939 = extractvalue { i1, i64 } %8937, 1
  br i1 %8938, label %6080, label %6082

ComputeLoop6005:                                  ; preds = %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit4740, %ComputeLoop6005
  %Accumulator6007 = phi i32 [ %8943, %ComputeLoop6005 ], [ -1, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit4740 ]
  %ActiveBits6008 = phi i64 [ %8946, %ComputeLoop6005 ], [ %6177, %_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIPKcEERS1_T_.exit4740 ]
  %8940 = call i64 @llvm.cttz.i64(i64 %ActiveBits6008, i1 true)
  %8941 = trunc i64 %8940 to i32
  %8942 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i21.i.i4731, i32 %8941)
  %8943 = and i32 %Accumulator6007, %8942
  %8944 = shl i64 1, %8940
  %8945 = xor i64 %8944, -1
  %8946 = and i64 %ActiveBits6008, %8945
  %8947 = icmp eq i64 %8946, 0
  br i1 %8947, label %ComputeEnd6006, label %ComputeLoop6005, !amdgpu.uniform !42

ComputeEnd6006:                                   ; preds = %ComputeLoop6005
  %.lcssa8933 = phi i32 [ %8943, %ComputeLoop6005 ]
  %8948 = trunc i64 %6177 to i32
  %8949 = lshr i64 %6177, 32
  %8950 = trunc i64 %8949 to i32
  %8951 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %8948, i32 0)
  %8952 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %8950, i32 %8951)
  %8953 = icmp eq i32 %8952, 0
  %8954 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %8953)
  %8955 = extractvalue { i1, i64 } %8954, 0
  %8956 = extractvalue { i1, i64 } %8954, 1
  br i1 %8955, label %6178, label %6181
}

; Function Attrs: convergent mustprogress nofree norecurse nounwind
define internal fastcc void @_ZN22__llvm_libc_22_0_0_git7testing10TestLoggerlsIcEERS1_T_(i8 noundef signext range(i8 10, 33) %ch) unnamed_addr #8 align 2 !type !195 {
entry:
  %0 = tail call i32 @llvm.amdgcn.mbcnt.lo(i32 -1, i32 0)
  %1 = tail call noundef i32 @llvm.amdgcn.mbcnt.hi(i32 -1, i32 %0)
  %sh_prom.i.i.i = zext i32 %1 to i64
  br label %for.cond.i.i, !amdgpu.uniform !42

for.cond.i.i:                                     ; preds = %Flow25, %entry
  %2 = phi i32 [ poison, %entry ], [ %40, %Flow25 ]
  %3 = phi i64 [ poison, %entry ], [ %41, %Flow25 ]
  %4 = phi i32 [ poison, %entry ], [ %42, %Flow25 ]
  %retval.sroa.6.0.i.i = phi i32 [ undef, %entry ], [ %45, %Flow25 ]
  %retval.sroa.2.0.i.i = phi i64 [ undef, %entry ], [ %44, %Flow25 ]
  %retval.sroa.8.0.i.i = phi i32 [ undef, %entry ], [ %43, %Flow25 ]
  %index.0.i.i = phi i32 [ 0, %entry ], [ %47, %Flow25 ]
  %5 = load i32, ptr addrspace(1) @__llvm_rpc_client, align 8, !tbaa !68
  %cmp.not.i.i = icmp ult i32 %index.0.i.i, %5
  %spec.store.select.i.i = select i1 %cmp.not.i.i, i32 %index.0.i.i, i32 0
  %6 = tail call noundef i64 @llvm.amdgcn.ballot.i64(i1 true)
  %7 = and i32 %spec.store.select.i.i, 31
  %8 = lshr i64 %6, %sh_prom.i.i.i
  %9 = trunc i64 %8 to i32
  %conv4.i.i.i.i = and i32 %9, 1
  %shl.i.i.i.i = shl nuw i32 %conv4.i.i.i.i, %7
  %10 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop, !amdgpu.uniform !42

11:                                               ; preds = %ComputeEnd
  %sunkaddr = getelementptr inbounds i8, ptr addrspace(1) %97, i64 40
  %12 = atomicrmw or ptr addrspace(1) %sunkaddr, i32 %.lcssa30 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %13, !amdgpu.uniform !42

13:                                               ; preds = %11, %ComputeEnd
  %14 = phi i32 [ %12, %11 ], [ poison, %ComputeEnd ]
  call void @llvm.amdgcn.end.cf.i64(i64 %106)
  %15 = call i32 @llvm.amdgcn.readfirstlane.i32(i32 %14)
  %16 = or i32 %15, %.lcssa31
  %shl5.i.i.i.i = shl nuw i32 1, %7
  %and.i.i.i.i = and i32 %shl5.i.i.i.i, %16
  %tobool3.i.i.i = icmp ne i32 %and.i.i.i.i, 0
  %17 = tail call i64 @llvm.amdgcn.ballot.i64(i1 %tobool3.i.i.i)
  %and.i.i.i.i.i = and i64 %17, %6
  %cmp.i.not.i.i = icmp ne i64 %6, %and.i.i.i.i.i
  br i1 %cmp.i.not.i.i, label %if.then.i.i.i, label %Flow24, !amdgpu.uniform !42

if.then.i.i.i:                                    ; preds = %13
  fence syncscope("agent") acquire
  %18 = load <2 x i64>, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !75
  %19 = extractelement <2 x i64> %18, i32 0
  %20 = inttoptr i64 %19 to ptr
  %21 = extractelement <2 x i64> %18, i32 1
  %22 = inttoptr i64 %21 to ptr
  %idxprom.i.i.i = zext i32 %spec.store.select.i.i to i64
  %arrayidx.i.i.i = getelementptr inbounds nuw i32, ptr %20, i64 %idxprom.i.i.i, !amdgpu.uniform !42
  %23 = load atomic i32, ptr %arrayidx.i.i.i monotonic, align 4
  %24 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %23)
  %arrayidx.i45.i.i = getelementptr inbounds nuw i32, ptr %22, i64 %idxprom.i.i.i, !amdgpu.uniform !42
  %25 = load atomic i32, ptr %arrayidx.i45.i.i monotonic, align 4
  %26 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %25)
  %cmp.i46.not.i.i = icmp ne i32 %24, %26
  br i1 %cmp.i46.not.i.i, label %if.then12.i.i, label %Flow22, !amdgpu.uniform !42

if.then12.i.i:                                    ; preds = %if.then.i.i.i
  fence syncscope("agent") release
  %27 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %6, i1 true)
  %iszero.i.i.i.i.i.i = icmp ne i64 %6, 0
  %cmp2.i.i.i.i.i = icmp eq i64 %27, %sh_prom.i.i.i
  %cmp.i.i.i.i.i = select i1 %iszero.i.i.i.i.i.i, i1 %cmp2.i.i.i.i.i, i1 false
  %conv4.i.i51.i.i = zext i1 %cmp.i.i.i.i.i to i32
  %shl.i.i52.i.i = shl nuw i32 %conv4.i.i51.i.i, %7
  %xor.i.i.i.i = xor i32 %shl.i.i52.i.i, -1
  %28 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop2, !amdgpu.uniform !42

29:                                               ; preds = %ComputeEnd3
  %sunkaddr16 = getelementptr inbounds i8, ptr addrspace(1) %97, i64 40
  %30 = atomicrmw and ptr addrspace(1) %sunkaddr16, i32 %.lcssa32 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %31, !amdgpu.uniform !42

31:                                               ; preds = %29, %ComputeEnd3
  call void @llvm.amdgcn.end.cf.i64(i64 %124)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %Flow22, !amdgpu.uniform !42

if.end14.i.i:                                     ; preds = %Flow22
  %32 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %6, i1 true)
  %iszero.i.i.i.i.i = icmp ne i64 %6, 0
  %cmp2.i.i.i.i = icmp eq i64 %32, %sh_prom.i.i.i
  %cmp.i.i.i.i = select i1 %iszero.i.i.i.i.i, i1 %cmp2.i.i.i.i, i1 false
  %33 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %cmp.i.i.i.i)
  %34 = extractvalue { i1, i64 } %33, 0
  %35 = extractvalue { i1, i64 } %33, 1
  br i1 %34, label %if.then16.i.i, label %if.end22.i.i

if.then16.i.i:                                    ; preds = %if.end14.i.i
  %36 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !76
  %opcode.i.i = getelementptr inbounds nuw %"struct.rpc::Header", ptr %36, i64 %idxprom.i.i.i, i32 1
  store i32 1660944387, ptr %opcode.i.i, align 8, !tbaa !77
  %arrayidx21.i.i = getelementptr inbounds nuw %"struct.rpc::Header", ptr %36, i64 %idxprom.i.i.i
  store i64 %6, ptr %arrayidx21.i.i, align 8, !tbaa !80
  br label %if.end22.i.i, !amdgpu.uniform !42

Flow23:                                           ; preds = %if.end22.i.i, %Flow22
  %37 = phi i32 [ %24, %if.end22.i.i ], [ %retval.sroa.8.0.i.i, %Flow22 ]
  %38 = phi i64 [ %6, %if.end22.i.i ], [ %retval.sroa.2.0.i.i, %Flow22 ]
  %39 = phi i32 [ %spec.store.select.i.i, %if.end22.i.i ], [ %retval.sroa.6.0.i.i, %Flow22 ]
  br label %cleanup26.i.i, !amdgpu.uniform !42

if.end22.i.i:                                     ; preds = %if.then16.i.i, %if.end14.i.i
  call void @llvm.amdgcn.end.cf.i64(i64 %35)
  tail call void @llvm.amdgcn.wave.barrier()
  br label %Flow23, !amdgpu.uniform !42

Flow24:                                           ; preds = %cleanup26.i.i, %13
  %40 = phi i32 [ %39, %cleanup26.i.i ], [ %2, %13 ]
  %41 = phi i64 [ %38, %cleanup26.i.i ], [ %3, %13 ]
  %42 = phi i32 [ %37, %cleanup26.i.i ], [ %4, %13 ]
  %43 = phi i32 [ %37, %cleanup26.i.i ], [ %retval.sroa.8.0.i.i, %13 ]
  %44 = phi i64 [ %38, %cleanup26.i.i ], [ %retval.sroa.2.0.i.i, %13 ]
  %45 = phi i32 [ %39, %cleanup26.i.i ], [ %retval.sroa.6.0.i.i, %13 ]
  %46 = phi i1 [ %cmp.i46.not.i.i, %cleanup26.i.i ], [ true, %13 ]
  br i1 %46, label %for.inc.i.i, label %Flow25, !amdgpu.uniform !42

cleanup26.i.i:                                    ; preds = %Flow23
  br label %Flow24, !amdgpu.uniform !42

for.inc.i.i:                                      ; preds = %Flow24
  %inc.i.i = add i32 %spec.store.select.i.i, 1
  br label %Flow25, !amdgpu.uniform !42

Flow25:                                           ; preds = %for.inc.i.i, %Flow24
  %47 = phi i32 [ %inc.i.i, %for.inc.i.i ], [ poison, %Flow24 ]
  %48 = phi i1 [ false, %for.inc.i.i ], [ true, %Flow24 ]
  br i1 %48, label %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i, label %for.cond.i.i, !amdgpu.uniform !42

_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i: ; preds = %Flow25
  %.lcssa35 = phi i32 [ %40, %Flow25 ]
  %.lcssa34 = phi i64 [ %41, %Flow25 ]
  %.lcssa33 = phi i32 [ %42, %Flow25 ]
  fence acquire
  %49 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %mul.i.i.i.i.i = shl i32 %.lcssa35, 6
  %idxprom.i19.i.i.i.i = zext i32 %mul.i.i.i.i.i to i64
  %arrayidx.i20.i.i.i.i = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %49, i64 %idxprom.i19.i.i.i.i
  %arrayidx.i22.i.i.i.i = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %arrayidx.i20.i.i.i.i, i64 %sh_prom.i.i.i
  store i64 1, ptr %arrayidx.i22.i.i.i.i, align 8, !tbaa !82
  %arrayidx6.i.i.i.i.i.i = getelementptr inbounds nuw i8, ptr %arrayidx.i22.i.i.i.i, i64 8
  store i8 %ch, ptr %arrayidx6.i.i.i.i.i.i, align 8
  %tobool.not.i.i.i.i.i = icmp eq i32 %.lcssa33, 0
  %conv.i.i.i.i.i = zext i1 %tobool.not.i.i.i.i.i to i32
  fence release
  %50 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %idxprom.i23.i.i.i.i = zext i32 %.lcssa35 to i64
  %arrayidx.i24.i.i.i.i = getelementptr inbounds nuw i32, ptr %50, i64 %idxprom.i23.i.i.i.i
  store atomic i32 %conv.i.i.i.i.i, ptr %arrayidx.i24.i.i.i.i monotonic, align 4
  %51 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !84
  %arrayidx.i.i5.i = getelementptr inbounds nuw %"struct.rpc::Header", ptr %51, i64 %idxprom.i23.i.i.i.i, !amdgpu.uniform !42
  %52 = load i64, ptr %arrayidx.i.i5.i, align 8, !tbaa !80
  br label %while.cond.i.i.i, !amdgpu.uniform !42

while.cond.i.i.i:                                 ; preds = %Flow21, %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i
  %phi.broken = phi i64 [ %67, %Flow21 ], [ 0, %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i ]
  %port.sroa.37.0.i = phi i32 [ %conv.i.i.i.i.i, %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i ], [ %65, %Flow21 ]
  %idx.0.i.i.i = phi i64 [ 56, %_ZN3rpc6Client4openILj1660944387EEENS_4PortILb0EEEv.exit.i ], [ %64, %Flow21 ]
  %cmp.i.i.i = icmp eq i64 %idx.0.i.i.i, 0
  %53 = tail call i64 @llvm.amdgcn.ballot.i64(i1 %cmp.i.i.i)
  %and.i.i.i.i6.i = and i64 %53, %52
  %tobool.not.i.i.i = icmp ne i64 %and.i.i.i.i6.i, 0
  %54 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i22.i.i = getelementptr inbounds nuw i32, ptr %54, i64 %idxprom.i23.i.i.i.i, !amdgpu.uniform !42
  %55 = load atomic i32, ptr %arrayidx.i22.i.i monotonic, align 4
  %56 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %tobool.not.i.i.i)
  %57 = extractvalue { i1, i64 } %56, 0
  %58 = extractvalue { i1, i64 } %56, 1
  br i1 %57, label %cond.false.i8.i.i.i, label %Flow21

cond.false.i8.i.i.i:                              ; preds = %while.cond.i.i.i
  %59 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %55)
  %60 = icmp ne i32 %59, %port.sroa.37.0.i
  br i1 %60, label %while.body.i.i21.i.i.i.preheader, label %Flow20, !amdgpu.uniform !42

while.body.i.i21.i.i.i.preheader:                 ; preds = %cond.false.i8.i.i.i
  br label %while.body.i.i21.i.i.i, !amdgpu.uniform !42

Flow20:                                           ; preds = %Flow19, %cond.false.i8.i.i.i
  br label %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i, !amdgpu.uniform !42

while.body.i.i21.i.i.i:                           ; preds = %while.body.i.i21.i.i.i.preheader, %while.body.i.i21.i.i.i
  tail call void @llvm.amdgcn.s.sleep(i32 2)
  %61 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i22.i.i.i = getelementptr inbounds nuw i32, ptr %61, i64 %idxprom.i23.i.i.i.i, !amdgpu.uniform !42
  %62 = load atomic i32, ptr %arrayidx.i.i.i22.i.i.i monotonic, align 4
  %63 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %62)
  %cmp.i.not.i.i23.i.i.i = icmp eq i32 %63, %port.sroa.37.0.i
  br i1 %cmp.i.not.i.i23.i.i.i, label %Flow19, label %while.body.i.i21.i.i.i, !amdgpu.uniform !42

Flow19:                                           ; preds = %while.body.i.i21.i.i.i
  br label %Flow20, !amdgpu.uniform !42

Flow21:                                           ; preds = %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i, %while.cond.i.i.i
  %64 = phi i64 [ %add.i.i.i, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i ], [ poison, %while.cond.i.i.i ]
  %65 = phi i32 [ %conv.i.i26.i.i.i, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i ], [ poison, %while.cond.i.i.i ]
  %66 = phi i1 [ false, %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i ], [ true, %while.cond.i.i.i ]
  call void @llvm.amdgcn.end.cf.i64(i64 %58)
  %67 = call i64 @llvm.amdgcn.if.break.i64(i1 %66, i64 %phi.broken)
  %68 = call i1 @llvm.amdgcn.loop.i64(i64 %67)
  br i1 %68, label %cond.false.i.i, label %while.cond.i.i.i

_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i: ; preds = %Flow20
  fence acquire
  br i1 %cmp.i.i.i, label %if.then.i.i.i.i.i.i, label %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i, !amdgpu.uniform !42

if.then.i.i.i.i.i.i:                              ; preds = %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i
  %69 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 32), align 8, !tbaa !81
  %arrayidx.i20.i35.i.i.i = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %69, i64 %idxprom.i19.i.i.i.i
  %arrayidx.i22.i38.i.i.i = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %arrayidx.i20.i35.i.i.i, i64 %sh_prom.i.i.i
  store i8 %ch, ptr %arrayidx.i22.i38.i.i.i, align 1
  br label %_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i, !amdgpu.uniform !42

_ZN3rpc4PortILb0EE4sendIZNS1_6send_nEPKPKvPyEUlPNS_6BufferEjE0_EEvT_.exit.i.i.i: ; preds = %if.then.i.i.i.i.i.i, %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit.i.i.i.i
  %conv.i.i26.i.i.i = xor i32 %port.sroa.37.0.i, 1
  fence release
  %70 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 16), align 8, !tbaa !83
  %arrayidx.i24.i29.i.i.i = getelementptr inbounds nuw i32, ptr %70, i64 %idxprom.i23.i.i.i.i
  store atomic i32 %conv.i.i26.i.i.i, ptr %arrayidx.i24.i29.i.i.i monotonic, align 4
  %add.i.i.i = add i64 %idx.0.i.i.i, 64
  br label %Flow21, !amdgpu.uniform !42

cond.false.i.i:                                   ; preds = %Flow21
  %.lcssa29 = phi i64 [ %67, %Flow21 ]
  %port.sroa.37.0.i.lcssa = phi i32 [ %port.sroa.37.0.i, %Flow21 ]
  %.lcssa28 = phi i32 [ %55, %Flow21 ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa29)
  %71 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %.lcssa28)
  %72 = icmp ne i32 %71, %port.sroa.37.0.i.lcssa
  %73 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %72)
  %74 = extractvalue { i1, i64 } %73, 0
  %75 = extractvalue { i1, i64 } %73, 1
  br i1 %74, label %while.body.i.i.i.preheader, label %Flow18

while.body.i.i.i.preheader:                       ; preds = %cond.false.i.i
  br label %while.body.i.i.i, !amdgpu.uniform !42

Flow18:                                           ; preds = %Flow, %cond.false.i.i
  call void @llvm.amdgcn.end.cf.i64(i64 %75)
  br label %_ZN22__llvm_libc_22_0_0_git15write_to_stderrENS_3cpp11string_viewE.exit, !amdgpu.uniform !42

while.body.i.i.i:                                 ; preds = %while.body.i.i.i.preheader, %while.body.i.i.i
  %phi.broken26 = phi i64 [ 0, %while.body.i.i.i.preheader ], [ %79, %while.body.i.i.i ]
  tail call void @llvm.amdgcn.s.sleep(i32 2)
  %76 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !85
  %arrayidx.i.i.i8.i = getelementptr inbounds nuw i32, ptr %76, i64 %idxprom.i23.i.i.i.i, !amdgpu.uniform !42
  %77 = load atomic i32, ptr %arrayidx.i.i.i8.i monotonic, align 4
  %78 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %77)
  %cmp.i.not.i.i.i = icmp eq i32 %78, %port.sroa.37.0.i.lcssa
  %79 = call i64 @llvm.amdgcn.if.break.i64(i1 %cmp.i.not.i.i.i, i64 %phi.broken26)
  %80 = call i1 @llvm.amdgcn.loop.i64(i64 %79)
  br i1 %80, label %Flow, label %while.body.i.i.i

Flow:                                             ; preds = %while.body.i.i.i
  %.lcssa27 = phi i64 [ %79, %while.body.i.i.i ]
  call void @llvm.amdgcn.end.cf.i64(i64 %.lcssa27)
  br label %Flow18, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git15write_to_stderrENS_3cpp11string_viewE.exit: ; preds = %Flow18
  fence acquire
  tail call void @llvm.amdgcn.wave.barrier()
  fence syncscope("agent") release
  %81 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %.lcssa34, i1 true)
  %iszero.i.i.i.i.i13.i = icmp ne i64 %.lcssa34, 0
  %cmp2.i.i.i.i14.i = icmp eq i64 %81, %sh_prom.i.i.i
  %cmp.i.i.i.i15.i = select i1 %iszero.i.i.i.i.i13.i, i1 %cmp2.i.i.i.i14.i, i1 false
  %82 = and i32 %.lcssa35, 31
  %conv4.i.i.i19.i = zext i1 %cmp.i.i.i.i15.i to i32
  %shl.i.i.i20.i = shl nuw i32 %conv4.i.i.i19.i, %82
  %xor.i.i.i21.i = xor i32 %shl.i.i.i20.i, -1
  %83 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  br label %ComputeLoop6, !amdgpu.uniform !42

84:                                               ; preds = %ComputeEnd7
  %div8.i.i.i16.i = lshr i32 %.lcssa35, 5
  %idxprom.i.i.i17.i = zext nneg i32 %div8.i.i.i16.i to i64
  %85 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i17.i
  %sunkaddr17 = getelementptr inbounds i8, ptr addrspace(1) %85, i64 40
  %86 = atomicrmw and ptr addrspace(1) %sunkaddr17, i32 %.lcssa syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %87, !amdgpu.uniform !42

87:                                               ; preds = %84, %ComputeEnd7
  call void @llvm.amdgcn.end.cf.i64(i64 %141)
  tail call void @llvm.amdgcn.wave.barrier()
  ret void

ComputeLoop:                                      ; preds = %for.cond.i.i, %ComputeLoop
  %Accumulator = phi i32 [ %92, %ComputeLoop ], [ 0, %for.cond.i.i ]
  %OldValuePhi = phi i32 [ %91, %ComputeLoop ], [ poison, %for.cond.i.i ]
  %ActiveBits = phi i64 [ %95, %ComputeLoop ], [ %10, %for.cond.i.i ]
  %88 = call i64 @llvm.cttz.i64(i64 %ActiveBits, i1 true)
  %89 = trunc i64 %88 to i32
  %90 = call i32 @llvm.amdgcn.readlane.i32(i32 %shl.i.i.i.i, i32 %89)
  %91 = call i32 @llvm.amdgcn.writelane.i32(i32 %Accumulator, i32 %89, i32 %OldValuePhi)
  %92 = or i32 %Accumulator, %90
  %93 = shl i64 1, %88
  %94 = xor i64 %93, -1
  %95 = and i64 %ActiveBits, %94
  %96 = icmp eq i64 %95, 0
  br i1 %96, label %ComputeEnd, label %ComputeLoop, !amdgpu.uniform !42

ComputeEnd:                                       ; preds = %ComputeLoop
  %.lcssa31 = phi i32 [ %91, %ComputeLoop ]
  %.lcssa30 = phi i32 [ %92, %ComputeLoop ]
  %div8.i.i.i.i = lshr i32 %spec.store.select.i.i, 5
  %idxprom.i.i.i.i = zext nneg i32 %div8.i.i.i.i to i64
  %97 = getelementptr inbounds nuw i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i.i.i
  %98 = trunc i64 %10 to i32
  %99 = lshr i64 %10, 32
  %100 = trunc i64 %99 to i32
  %101 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %98, i32 0)
  %102 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %100, i32 %101)
  %103 = icmp eq i32 %102, 0
  %104 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %103)
  %105 = extractvalue { i1, i64 } %104, 0
  %106 = extractvalue { i1, i64 } %104, 1
  br i1 %105, label %11, label %13

Flow22:                                           ; preds = %31, %if.then.i.i.i
  %107 = phi i1 [ false, %31 ], [ true, %if.then.i.i.i ]
  br i1 %107, label %if.end14.i.i, label %Flow23, !amdgpu.uniform !42

ComputeLoop2:                                     ; preds = %if.then12.i.i, %ComputeLoop2
  %Accumulator4 = phi i32 [ %111, %ComputeLoop2 ], [ -1, %if.then12.i.i ]
  %ActiveBits5 = phi i64 [ %114, %ComputeLoop2 ], [ %28, %if.then12.i.i ]
  %108 = call i64 @llvm.cttz.i64(i64 %ActiveBits5, i1 true)
  %109 = trunc i64 %108 to i32
  %110 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i.i, i32 %109)
  %111 = and i32 %Accumulator4, %110
  %112 = shl i64 1, %108
  %113 = xor i64 %112, -1
  %114 = and i64 %ActiveBits5, %113
  %115 = icmp eq i64 %114, 0
  br i1 %115, label %ComputeEnd3, label %ComputeLoop2, !amdgpu.uniform !42

ComputeEnd3:                                      ; preds = %ComputeLoop2
  %.lcssa32 = phi i32 [ %111, %ComputeLoop2 ]
  %116 = trunc i64 %28 to i32
  %117 = lshr i64 %28, 32
  %118 = trunc i64 %117 to i32
  %119 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %116, i32 0)
  %120 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %118, i32 %119)
  %121 = icmp eq i32 %120, 0
  %122 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %121)
  %123 = extractvalue { i1, i64 } %122, 0
  %124 = extractvalue { i1, i64 } %122, 1
  br i1 %123, label %29, label %31

ComputeLoop6:                                     ; preds = %ComputeLoop6, %_ZN22__llvm_libc_22_0_0_git15write_to_stderrENS_3cpp11string_viewE.exit
  %Accumulator8 = phi i32 [ -1, %_ZN22__llvm_libc_22_0_0_git15write_to_stderrENS_3cpp11string_viewE.exit ], [ %128, %ComputeLoop6 ]
  %ActiveBits9 = phi i64 [ %83, %_ZN22__llvm_libc_22_0_0_git15write_to_stderrENS_3cpp11string_viewE.exit ], [ %131, %ComputeLoop6 ]
  %125 = call i64 @llvm.cttz.i64(i64 %ActiveBits9, i1 true)
  %126 = trunc i64 %125 to i32
  %127 = call i32 @llvm.amdgcn.readlane.i32(i32 %xor.i.i.i21.i, i32 %126)
  %128 = and i32 %Accumulator8, %127
  %129 = shl i64 1, %125
  %130 = xor i64 %129, -1
  %131 = and i64 %ActiveBits9, %130
  %132 = icmp eq i64 %131, 0
  br i1 %132, label %ComputeEnd7, label %ComputeLoop6, !amdgpu.uniform !42

ComputeEnd7:                                      ; preds = %ComputeLoop6
  %.lcssa = phi i32 [ %128, %ComputeLoop6 ]
  %133 = trunc i64 %83 to i32
  %134 = lshr i64 %83, 32
  %135 = trunc i64 %134 to i32
  %136 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %133, i32 0)
  %137 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %135, i32 %136)
  %138 = icmp eq i32 %137, 0
  %139 = call { i1, i64 } @llvm.amdgcn.if.i64(i1 %138)
  %140 = extractvalue { i1, i64 } %139, 0
  %141 = extractvalue { i1, i64 } %139, 1
  br i1 %140, label %84, label %87
}

; Function Attrs: mustprogress nobuiltin nounwind allocsize(0)
define noundef nonnull ptr @_Znwm(i64 noundef %size) local_unnamed_addr #14 {
entry:
  %sub.i = add i64 %size, 7
  %div2.i = and i64 %sub.i, -8
  %0 = load ptr, ptr addrspace(1) @_ZN12_GLOBAL__N_13ptrE, align 8, !tbaa !55
  %add.ptr.i = getelementptr inbounds nuw i8, ptr %0, i64 %div2.i
  store ptr %add.ptr.i, ptr addrspace(1) @_ZN12_GLOBAL__N_13ptrE, align 8, !tbaa !55
  ret ptr %0
}

; Function Attrs: mustprogress nobuiltin nounwind allocsize(0)
define noundef nonnull ptr @_Znam(i64 noundef %size) local_unnamed_addr #14 {
entry:
  %sub.i = add i64 %size, 7
  %div2.i = and i64 %sub.i, -8
  %0 = load ptr, ptr addrspace(1) @_ZN12_GLOBAL__N_13ptrE, align 8, !tbaa !55
  %add.ptr.i = getelementptr inbounds nuw i8, ptr %0, i64 %div2.i
  store ptr %add.ptr.i, ptr addrspace(1) @_ZN12_GLOBAL__N_13ptrE, align 8, !tbaa !55
  ret ptr %0
}

; Function Attrs: mustprogress nobuiltin nounwind
define void @_ZdlPv(ptr noundef %0) local_unnamed_addr #15 {
entry:
  tail call void @llvm.trap()
  unreachable
}

; Function Attrs: cold noreturn nounwind memory(inaccessiblemem: write)
declare void @llvm.trap() #16

; Function Attrs: mustprogress nobuiltin nounwind
define void @_ZdlPvm(ptr noundef %ptr, i64 noundef %size) local_unnamed_addr #15 {
entry:
  tail call void @llvm.trap()
  unreachable
}

; Function Attrs: mustprogress nobuiltin nounwind
define void @_ZdlPvSt11align_val_t(ptr noundef %mem, i64 noundef %0) local_unnamed_addr #15 {
entry:
  tail call void @llvm.trap()
  unreachable
}

; Function Attrs: mustprogress nocallback nofree nosync nounwind willreturn memory(none)
declare i32 @llvm.amdgcn.mbcnt.lo(i32, i32) #17

; Function Attrs: mustprogress nocallback nofree nosync nounwind willreturn memory(none)
declare i32 @llvm.amdgcn.mbcnt.hi(i32, i32) #17

; Function Attrs: convergent mustprogress nocallback nofree nounwind willreturn memory(none)
declare i64 @llvm.amdgcn.ballot.i64(i1) #18

; Function Attrs: convergent mustprogress nocallback nofree nounwind willreturn memory(none)
declare i32 @llvm.amdgcn.readfirstlane.i32(i32) #18

; Function Attrs: mustprogress nocallback nofree nosync nounwind speculatable willreturn memory(none)
declare i64 @llvm.cttz.i64(i64, i1 immarg) #2

; Function Attrs: convergent mustprogress nocallback nofree nounwind willreturn
declare void @llvm.amdgcn.wave.barrier() #19

; Function Attrs: mustprogress nocallback nofree nosync nounwind speculatable willreturn memory(none)
declare i64 @llvm.umin.i64(i64, i64) #2

; Function Attrs: mustprogress nocallback nofree nosync nounwind willreturn
declare void @llvm.amdgcn.s.sleep(i32 immarg) #20

; Function Attrs: convergent mustprogress nounwind
define internal fastcc void @__cxa_finalize() unnamed_addr #21 {
entry:
  %0 = load i64, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @_ZN22__llvm_libc_22_0_0_git16atexit_callbacksE, i64 1024), align 8, !tbaa !43
  %cmp.i13.i = icmp ne i64 %0, 0
  br i1 %cmp.i13.i, label %_ZN22__llvm_libc_22_0_0_git11FixedVectorINS_10AtExitUnitELm64EE4backEv.exit.i.preheader, label %Flow3, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git11FixedVectorINS_10AtExitUnitELm64EE4backEv.exit.i.preheader: ; preds = %entry
  br label %_ZN22__llvm_libc_22_0_0_git11FixedVectorINS_10AtExitUnitELm64EE4backEv.exit.i, !amdgpu.uniform !42

Flow3:                                            ; preds = %Flow, %entry
  br label %_ZN22__llvm_libc_22_0_0_git19call_exit_callbacksERNS_11FixedVectorINS_10AtExitUnitELm64EEE.exit, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git11FixedVectorINS_10AtExitUnitELm64EE4backEv.exit.i: ; preds = %_ZN22__llvm_libc_22_0_0_git11FixedVectorINS_10AtExitUnitELm64EE4backEv.exit.i.preheader, %_ZN22__llvm_libc_22_0_0_git11FixedVectorINS_10AtExitUnitELm64EE4backEv.exit.i
  %1 = phi i64 [ %7, %_ZN22__llvm_libc_22_0_0_git11FixedVectorINS_10AtExitUnitELm64EE4backEv.exit.i ], [ %0, %_ZN22__llvm_libc_22_0_0_git11FixedVectorINS_10AtExitUnitELm64EE4backEv.exit.i.preheader ]
  %2 = getelementptr %"struct.__llvm_libc_22_0_0_git::AtExitUnit", ptr addrspace(1) @_ZN22__llvm_libc_22_0_0_git16atexit_callbacksE, i64 %1
  %arrayidx.i.i.i = getelementptr i8, ptr addrspace(1) %2, i64 -16, !amdgpu.uniform !42
  %3 = load <2 x i64>, ptr addrspace(1) %arrayidx.i.i.i, align 8, !tbaa !49
  %unit.sroa.0.0.copyload.i1 = extractelement <2 x i64> %3, i32 0
  %4 = inttoptr i64 %unit.sroa.0.0.copyload.i1 to ptr
  %unit.sroa.5.0.copyload.i2 = extractelement <2 x i64> %3, i32 1
  %5 = inttoptr i64 %unit.sroa.5.0.copyload.i2 to ptr
  tail call void @llvm.memset.p1.i64(ptr addrspace(1) noundef align 8 dereferenceable(16) %arrayidx.i.i.i, i8 0, i64 16, i1 false)
  %6 = load i64, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @_ZN22__llvm_libc_22_0_0_git16atexit_callbacksE, i64 1024), align 8, !tbaa !43
  %dec.i.i = add i64 %6, -1
  store i64 %dec.i.i, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @_ZN22__llvm_libc_22_0_0_git16atexit_callbacksE, i64 1024), align 8, !tbaa !43
  tail call void %4(ptr noundef %5) #36
  %7 = load i64, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @_ZN22__llvm_libc_22_0_0_git16atexit_callbacksE, i64 1024), align 8, !tbaa !43
  %cmp.i.i = icmp eq i64 %7, 0
  br i1 %cmp.i.i, label %Flow, label %_ZN22__llvm_libc_22_0_0_git11FixedVectorINS_10AtExitUnitELm64EE4backEv.exit.i, !amdgpu.uniform !42

Flow:                                             ; preds = %_ZN22__llvm_libc_22_0_0_git11FixedVectorINS_10AtExitUnitELm64EE4backEv.exit.i
  br label %Flow3, !amdgpu.uniform !42

_ZN22__llvm_libc_22_0_0_git19call_exit_callbacksERNS_11FixedVectorINS_10AtExitUnitELm64EEE.exit: ; preds = %Flow3
  store i64 0, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @_ZN22__llvm_libc_22_0_0_git16atexit_callbacksE, i64 1024), align 8, !tbaa !43
  ret void
}

; Function Attrs: mustprogress nocallback nofree nounwind willreturn memory(argmem: write)
declare void @llvm.memset.p1.i64(ptr addrspace(1) writeonly captures(none), i8, i64, i1 immarg) #11

; Function Attrs: convergent inlinehint mustprogress nounwind
define internal void @_ZN22__llvm_libc_22_0_0_git17stdc_at_exit_funcEPv(ptr noundef readonly captures(none) %payload) #22 comdat {
entry:
  tail call void %payload() #36
  ret void
}

; Function Attrs: cold convergent mustprogress noreturn nounwind
define internal fastcc void @_ZN22__llvm_libc_22_0_0_git4exitEi(i32 noundef %status) unnamed_addr #23 !section_prefix !58 {
entry:
  tail call fastcc void @__cxa_finalize() #36
  tail call fastcc void @_ZN22__llvm_libc_22_0_0_git8internal4exitEi(i32 noundef %status) #37
  unreachable
}

; Function Attrs: cold convergent mustprogress nofree norecurse noreturn nounwind
define internal fastcc void @_ZN22__llvm_libc_22_0_0_git8internal4exitEi(i32 noundef %status) unnamed_addr #24 !section_prefix !58 {
entry:
  %port = alloca %"struct.rpc::Port", align 8, addrspace(5)
  %0 = call ptr @llvm.amdgcn.addrspacecast.nonnull.p0.p5(ptr addrspace(5) %port)
  call void @llvm.lifetime.start.p5(ptr addrspace(5) %port) #34
  %1 = getelementptr inbounds nuw i8, ptr addrspace(5) %port, i32 8
  %2 = getelementptr inbounds nuw i8, ptr addrspace(5) %port, i32 16
  %3 = getelementptr inbounds nuw i8, ptr addrspace(5) %port, i32 20
  %4 = getelementptr inbounds nuw i8, ptr addrspace(5) %port, i32 24
  %5 = getelementptr inbounds nuw i8, ptr addrspace(5) %port, i32 28
  %6 = getelementptr inbounds nuw i8, ptr addrspace(5) %port, i32 30
  store <2 x i8> splat (i8 -86), ptr addrspace(5) %6, align 2
  %call = tail call fastcc <{ ptr, i64, i32, i32, i32, i8, i8 }> @_ZN3rpc6Client4openILj1660944385EEENS_4PortILb0EEEv() #38
  %newret = extractvalue <{ ptr, i64, i32, i32, i32, i8, i8 }> %call, 0
  %newret1 = extractvalue <{ ptr, i64, i32, i32, i32, i8, i8 }> %call, 1
  %newret3 = extractvalue <{ ptr, i64, i32, i32, i32, i8, i8 }> %call, 2
  %newret5 = extractvalue <{ ptr, i64, i32, i32, i32, i8, i8 }> %call, 3
  %newret7 = extractvalue <{ ptr, i64, i32, i32, i32, i8, i8 }> %call, 4
  %newret9 = extractvalue <{ ptr, i64, i32, i32, i32, i8, i8 }> %call, 5
  %newret11 = extractvalue <{ ptr, i64, i32, i32, i32, i8, i8 }> %call, 6
  store ptr %newret, ptr addrspace(5) %port, align 8
  store i64 %newret1, ptr addrspace(5) %1, align 8
  store i32 %newret3, ptr addrspace(5) %2, align 8
  store i32 %newret5, ptr addrspace(5) %3, align 4
  store i32 %newret7, ptr addrspace(5) %4, align 8
  %7 = insertelement <2 x i8> poison, i8 %newret9, i32 0
  %8 = insertelement <2 x i8> %7, i8 %newret11, i32 1
  store <2 x i8> %8, ptr addrspace(5) %5, align 4
  call fastcc void @"_ZN3rpc4PortILb0EE13send_and_recvIZN22__llvm_libc_22_0_0_git8internal4exitEiE3$_0ZNS4_4exitEiE3$_1EEvT_T0_"(ptr noundef nonnull align 8 dereferenceable(30) %0) #38
  call fastcc void @"_ZN3rpc4PortILb0EE4sendIZN22__llvm_libc_22_0_0_git8internal4exitEiE3$_2EEvT_"(ptr noundef nonnull align 8 dereferenceable(30) %0, i32 %status) #38
  call fastcc void @_ZN3rpc4PortILb0EE5closeEv(ptr noundef nonnull align 8 dereferenceable(30) %0) #38
  tail call void @llvm.amdgcn.endpgm()
  unreachable
}

; Function Attrs: convergent inlinehint mustprogress nofree norecurse nounwind
define internal fastcc <{ ptr, i64, i32, i32, i32, i8, i8 }> @_ZN3rpc6Client4openILj1660944385EEENS_4PortILb0EEEv() unnamed_addr #25 comdat align 2 !type !196 {
entry:
  %0 = tail call i32 @llvm.amdgcn.mbcnt.lo(i32 -1, i32 0)
  %1 = tail call noundef i32 @llvm.amdgcn.mbcnt.hi(i32 -1, i32 %0)
  %sh_prom.i = zext i32 %1 to i64
  br label %for.cond, !amdgpu.uniform !42

for.cond:                                         ; preds = %Flow18, %entry
  %2 = phi i32 [ poison, %entry ], [ %52, %Flow18 ]
  %3 = phi i64 [ poison, %entry ], [ %53, %Flow18 ]
  %4 = phi i32 [ poison, %entry ], [ %54, %Flow18 ]
  %retval.sroa.6.0 = phi i32 [ undef, %entry ], [ %57, %Flow18 ]
  %retval.sroa.2.0 = phi i64 [ undef, %entry ], [ %56, %Flow18 ]
  %retval.sroa.8.0 = phi i32 [ undef, %entry ], [ %55, %Flow18 ]
  %index.0 = phi i32 [ 0, %entry ], [ %59, %Flow18 ]
  %5 = load i32, ptr addrspace(1) @__llvm_rpc_client, align 8, !tbaa !68
  %cmp.not = icmp ult i32 %index.0, %5
  %spec.store.select = select i1 %cmp.not, i32 %index.0, i32 0
  %6 = tail call noundef i64 @llvm.amdgcn.ballot.i64(i1 true)
  %div8.i.i = lshr i32 %spec.store.select, 5
  %7 = and i32 %spec.store.select, 31
  %idxprom.i.i = zext nneg i32 %div8.i.i to i64
  %8 = getelementptr inbounds i32, ptr addrspace(1) @__llvm_rpc_client, i64 %idxprom.i.i
  %9 = lshr i64 %6, %sh_prom.i
  %10 = trunc i64 %9 to i32
  %conv4.i.i = and i32 %10, 1
  %shl.i.i = shl nuw i32 %conv4.i.i, %7
  %11 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  %12 = trunc i64 %11 to i32
  %13 = lshr i64 %11, 32
  %14 = trunc i64 %13 to i32
  %15 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %12, i32 0)
  %16 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %14, i32 %15)
  %17 = icmp eq i32 %16, 0
  br i1 %17, label %18, label %20, !amdgpu.uniform !42

18:                                               ; preds = %for.cond
  %sunkaddr = getelementptr inbounds i8, ptr addrspace(1) %8, i64 40
  %19 = atomicrmw or ptr addrspace(1) %sunkaddr, i32 %shl.i.i syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %20, !amdgpu.uniform !42

20:                                               ; preds = %18, %for.cond
  %21 = phi i32 [ %19, %18 ], [ poison, %for.cond ]
  %22 = call i32 @llvm.amdgcn.readfirstlane.i32(i32 %21)
  %23 = select i1 %17, i32 0, i32 %shl.i.i
  %24 = or i32 %22, %23
  %shl5.i.i = shl nuw i32 1, %7
  %and.i.i = and i32 %24, %shl5.i.i
  %tobool3.i = icmp ne i32 %and.i.i, 0
  %25 = tail call i64 @llvm.amdgcn.ballot.i64(i1 %tobool3.i)
  %and.i.i.i = and i64 %25, %6
  %cmp.i.not = icmp ne i64 %6, %and.i.i.i
  br i1 %cmp.i.not, label %if.then.i, label %Flow17, !amdgpu.uniform !42

if.then.i:                                        ; preds = %20
  fence syncscope("agent") acquire
  %26 = load <2 x i64>, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 8), align 8, !tbaa !75
  %27 = extractelement <2 x i64> %26, i32 0
  %28 = inttoptr i64 %27 to ptr
  %29 = extractelement <2 x i64> %26, i32 1
  %30 = inttoptr i64 %29 to ptr
  %idxprom.i = zext i32 %spec.store.select to i64
  %arrayidx.i = getelementptr inbounds nuw i32, ptr %28, i64 %idxprom.i, !amdgpu.uniform !42
  %31 = load atomic i32, ptr %arrayidx.i monotonic, align 4
  %32 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %31)
  %arrayidx.i45 = getelementptr inbounds nuw i32, ptr %30, i64 %idxprom.i, !amdgpu.uniform !42
  %33 = load atomic i32, ptr %arrayidx.i45 monotonic, align 4
  %34 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %33)
  %cmp.i46.not = icmp ne i32 %32, %34
  br i1 %cmp.i46.not, label %if.then12, label %Flow, !amdgpu.uniform !42

if.then12:                                        ; preds = %if.then.i
  fence syncscope("agent") release
  %35 = call i64 @llvm.amdgcn.ballot.i64(i1 true)
  %36 = trunc i64 %35 to i32
  %37 = lshr i64 %35, 32
  %38 = trunc i64 %37 to i32
  %39 = call i32 @llvm.amdgcn.mbcnt.lo(i32 %36, i32 0)
  %40 = call i32 @llvm.amdgcn.mbcnt.hi(i32 %38, i32 %39)
  %41 = icmp eq i32 %40, 0
  br i1 %41, label %42, label %46, !amdgpu.uniform !42

42:                                               ; preds = %if.then12
  %43 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %6, i1 true)
  %iszero.i.i.i.i = icmp ne i64 %6, 0
  %cmp2.i.i.i = icmp eq i64 %43, %sh_prom.i
  %cmp.i.i.i = select i1 %iszero.i.i.i.i, i1 %cmp2.i.i.i, i1 false
  %conv4.i.i51 = zext i1 %cmp.i.i.i to i32
  %shl.i.i52 = shl nuw i32 %conv4.i.i51, %7
  %xor.i.i = xor i32 %shl.i.i52, -1
  %sunkaddr15 = getelementptr inbounds i8, ptr addrspace(1) %8, i64 40
  %44 = atomicrmw and ptr addrspace(1) %sunkaddr15, i32 %xor.i.i syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  br label %46, !amdgpu.uniform !42

Flow:                                             ; preds = %46, %if.then.i
  %45 = phi i1 [ false, %46 ], [ true, %if.then.i ]
  br i1 %45, label %if.end14, label %Flow16, !amdgpu.uniform !42

46:                                               ; preds = %42, %if.then12
  tail call void @llvm.amdgcn.wave.barrier()
  br label %Flow, !amdgpu.uniform !42

if.end14:                                         ; preds = %Flow
  %47 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %6, i1 true)
  %iszero.i.i.i = icmp ne i64 %6, 0
  %cmp2.i.i = icmp eq i64 %47, %sh_prom.i
  %cmp.i.i = select i1 %iszero.i.i.i, i1 %cmp2.i.i, i1 false
  br i1 %cmp.i.i, label %if.then16, label %if.end22, !amdgpu.uniform !42

if.then16:                                        ; preds = %if.end14
  %48 = load ptr, ptr addrspace(1) getelementptr inbounds nuw (i8, ptr addrspace(1) @__llvm_rpc_client, i64 24), align 8, !tbaa !76
  %opcode = getelementptr inbounds nuw %"struct.rpc::Header", ptr %48, i64 %idxprom.i, i32 1
  store i32 1660944385, ptr %opcode, align 8, !tbaa !77
  %arrayidx21 = getelementptr inbounds nuw %"struct.rpc::Header", ptr %48, i64 %idxprom.i
  store i64 %6, ptr %arrayidx21, align 8, !tbaa !80
  br label %if.end22, !amdgpu.uniform !42

Flow16:                                           ; preds = %if.end22, %Flow
  %49 = phi i32 [ %32, %if.end22 ], [ %retval.sroa.8.0, %Flow ]
  %50 = phi i64 [ %6, %if.end22 ], [ %retval.sroa.2.0, %Flow ]
  %51 = phi i32 [ %spec.store.select, %if.end22 ], [ %retval.sroa.6.0, %Flow ]
  br label %cleanup26, !amdgpu.uniform !42

if.end22:                                         ; preds = %if.then16, %if.end14
  tail call void @llvm.amdgcn.wave.barrier()
  br label %Flow16, !amdgpu.uniform !42

Flow17:                                           ; preds = %cleanup26, %20
  %52 = phi i32 [ %51, %cleanup26 ], [ %2, %20 ]
  %53 = phi i64 [ %50, %cleanup26 ], [ %3, %20 ]
  %54 = phi i32 [ %49, %cleanup26 ], [ %4, %20 ]
  %55 = phi i32 [ %49, %cleanup26 ], [ %retval.sroa.8.0, %20 ]
  %56 = phi i64 [ %50, %cleanup26 ], [ %retval.sroa.2.0, %20 ]
  %57 = phi i32 [ %51, %cleanup26 ], [ %retval.sroa.6.0, %20 ]
  %58 = phi i1 [ %cmp.i46.not, %cleanup26 ], [ true, %20 ]
  br i1 %58, label %for.inc, label %Flow18, !amdgpu.uniform !42

cleanup26:                                        ; preds = %Flow16
  br label %Flow17, !amdgpu.uniform !42

for.inc:                                          ; preds = %Flow17
  %inc = add i32 %spec.store.select, 1
  br label %Flow18, !amdgpu.uniform !42

Flow18:                                           ; preds = %for.inc, %Flow17
  %59 = phi i32 [ %inc, %for.inc ], [ poison, %Flow17 ]
  %60 = phi i1 [ false, %for.inc ], [ true, %Flow17 ]
  br i1 %60, label %cleanup27, label %for.cond, !amdgpu.uniform !42

cleanup27:                                        ; preds = %Flow18
  %.lcssa20 = phi i32 [ %52, %Flow18 ]
  %.lcssa19 = phi i64 [ %53, %Flow18 ]
  %.lcssa = phi i32 [ %54, %Flow18 ]
  %newret2 = insertvalue <{ ptr, i64, i32, i32, i32, i8, i8 }> <{ ptr addrspacecast (ptr addrspace(1) @__llvm_rpc_client to ptr), i64 poison, i32 poison, i32 poison, i32 poison, i8 poison, i8 poison }>, i64 %.lcssa19, 1
  %newret4 = insertvalue <{ ptr, i64, i32, i32, i32, i8, i8 }> %newret2, i32 64, 2
  %newret6 = insertvalue <{ ptr, i64, i32, i32, i32, i8, i8 }> %newret4, i32 %.lcssa20, 3
  %newret8 = insertvalue <{ ptr, i64, i32, i32, i32, i8, i8 }> %newret6, i32 %.lcssa, 4
  %newret10 = insertvalue <{ ptr, i64, i32, i32, i32, i8, i8 }> %newret8, i8 0, 5
  %newret12 = insertvalue <{ ptr, i64, i32, i32, i32, i8, i8 }> %newret10, i8 1, 6
  ret <{ ptr, i64, i32, i32, i32, i8, i8 }> %newret12
}

; Function Attrs: convergent inlinehint mustprogress nofree norecurse nounwind
define internal fastcc void @"_ZN3rpc4PortILb0EE13send_and_recvIZN22__llvm_libc_22_0_0_git8internal4exitEiE3$_0ZNS4_4exitEiE3$_1EEvT_T0_"(ptr noundef nonnull align 8 captures(none) dereferenceable(30) %this) unnamed_addr #25 align 2 !type !197 {
entry:
  %owns_buffer.i = getelementptr inbounds nuw i8, ptr %this, i64 29, !amdgpu.uniform !42
  %0 = load i8, ptr %owns_buffer.i, align 1, !tbaa !199, !range !152, !noundef !42
  %loadedv.i = trunc nuw i8 %0 to i1
  %loadedv.i.inv = xor i1 %loadedv.i, true
  br i1 %loadedv.i.inv, label %cond.false.i, label %Flow11, !amdgpu.uniform !42

Flow11:                                           ; preds = %cond.false.i, %entry
  %1 = phi i1 [ %11, %cond.false.i ], [ poison, %entry ]
  %2 = phi i32 [ %.pre.i7, %cond.false.i ], [ poison, %entry ]
  %3 = phi i1 [ false, %cond.false.i ], [ true, %entry ]
  br i1 %3, label %cond.true.i, label %cond.end.i, !amdgpu.uniform !42

cond.true.i:                                      ; preds = %Flow11
  %out.i = getelementptr inbounds nuw i8, ptr %this, i64 24, !amdgpu.uniform !42
  %4 = load i32, ptr %out.i, align 8, !tbaa !202
  br label %cond.end.i, !amdgpu.uniform !42

cond.false.i:                                     ; preds = %entry
  %5 = load ptr, ptr %this, align 8, !tbaa !203, !nonnull !42, !align !204
  %index.i = getelementptr inbounds nuw i8, ptr %this, i64 20, !amdgpu.uniform !42
  %6 = load <2 x i32>, ptr %index.i, align 4, !tbaa !205
  %7 = extractelement <2 x i32> %6, i32 0
  %.pre.i7 = extractelement <2 x i32> %6, i32 1
  %inbox.i.i = getelementptr inbounds nuw i8, ptr %5, i64 8, !amdgpu.uniform !42
  %8 = load ptr, ptr %inbox.i.i, align 8, !tbaa !85
  %idxprom.i.i = zext i32 %7 to i64
  %arrayidx.i.i = getelementptr inbounds nuw i32, ptr %8, i64 %idxprom.i.i, !amdgpu.uniform !42
  %9 = load atomic i32, ptr %arrayidx.i.i monotonic, align 4
  %10 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %9)
  %11 = icmp eq i32 %10, %.pre.i7
  br label %Flow11, !amdgpu.uniform !42

cond.end.i:                                       ; preds = %cond.true.i, %Flow11
  %12 = phi i32 [ %2, %Flow11 ], [ %4, %cond.true.i ]
  %cond.i = phi i1 [ %1, %Flow11 ], [ true, %cond.true.i ]
  %cond.i.inv = xor i1 %cond.i, true
  br i1 %cond.i.inv, label %while.body.lr.ph.i.i, label %Flow10, !amdgpu.uniform !42

while.body.lr.ph.i.i:                             ; preds = %cond.end.i
  %sunkaddr = getelementptr inbounds i8, ptr %this, i64 20, !amdgpu.uniform !42
  %13 = load i32, ptr %sunkaddr, align 4, !tbaa !206
  %14 = load ptr, ptr %this, align 8, !tbaa !203, !nonnull !42, !align !204
  %idxprom.i.i.i = zext i32 %13 to i64
  br label %while.body.i.i, !amdgpu.uniform !42

Flow10:                                           ; preds = %Flow9, %cond.end.i
  br label %"_ZN3rpc4PortILb0EE4sendIZN22__llvm_libc_22_0_0_git8internal4exitEiE3$_0EEvT_.exit", !amdgpu.uniform !42

while.body.i.i:                                   ; preds = %while.body.lr.ph.i.i, %while.body.i.i
  tail call void @llvm.amdgcn.s.sleep(i32 2)
  %sunkaddr1 = getelementptr inbounds i8, ptr %14, i64 8, !amdgpu.uniform !42
  %15 = load ptr, ptr %sunkaddr1, align 8, !tbaa !85
  %arrayidx.i.i.i = getelementptr inbounds nuw i32, ptr %15, i64 %idxprom.i.i.i, !amdgpu.uniform !42
  %16 = load atomic i32, ptr %arrayidx.i.i.i monotonic, align 4
  %17 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %16)
  %cmp.i.not.i.i = icmp eq i32 %17, %12
  br i1 %cmp.i.not.i.i, label %Flow9, label %while.body.i.i, !amdgpu.uniform !42

Flow9:                                            ; preds = %while.body.i.i
  br label %Flow10, !amdgpu.uniform !42

"_ZN3rpc4PortILb0EE4sendIZN22__llvm_libc_22_0_0_git8internal4exitEiE3$_0EEvT_.exit": ; preds = %Flow10
  %out5.i = getelementptr inbounds nuw i8, ptr %this, i64 24
  fence acquire
  %18 = load ptr, ptr %this, align 8, !tbaa !203, !nonnull !42, !align !204
  %sunkaddr2 = getelementptr inbounds i8, ptr %this, i64 20, !amdgpu.uniform !42
  %19 = load <2 x i32>, ptr %sunkaddr2, align 4, !tbaa !205
  %20 = extractelement <2 x i32> %19, i32 0
  %21 = extractelement <2 x i32> %19, i32 1
  %idxprom.i = zext i32 %20 to i64
  %tobool.not.i.i = icmp eq i32 %21, 0
  %conv.i.i = zext i1 %tobool.not.i.i to i32
  fence release
  %outbox.i.i = getelementptr inbounds nuw i8, ptr %18, i64 16, !amdgpu.uniform !42
  %22 = load ptr, ptr %outbox.i.i, align 8, !tbaa !83
  %arrayidx.i21.i = getelementptr inbounds nuw i32, ptr %22, i64 %idxprom.i
  store atomic i32 %conv.i.i, ptr %arrayidx.i21.i monotonic, align 4
  store i32 %conv.i.i, ptr %out5.i, align 8, !tbaa !202
  %receive.i = getelementptr inbounds nuw i8, ptr %this, i64 28
  store <2 x i8> zeroinitializer, ptr %receive.i, align 4, !tbaa !207
  %23 = load ptr, ptr %this, align 8, !tbaa !203, !nonnull !42, !align !204
  %24 = load i32, ptr %sunkaddr2, align 4, !tbaa !206
  %inbox.i.i8 = getelementptr inbounds nuw i8, ptr %23, i64 8, !amdgpu.uniform !42
  %25 = load ptr, ptr %inbox.i.i8, align 8, !tbaa !85
  %idxprom.i21.i = zext i32 %24 to i64
  %arrayidx.i22.i = getelementptr inbounds nuw i32, ptr %25, i64 %idxprom.i21.i, !amdgpu.uniform !42
  %26 = load atomic i32, ptr %arrayidx.i22.i monotonic, align 4
  %27 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %26)
  %28 = icmp ne i32 %27, %conv.i.i
  br i1 %28, label %while.body.i.i12.preheader, label %Flow8, !amdgpu.uniform !42

while.body.i.i12.preheader:                       ; preds = %"_ZN3rpc4PortILb0EE4sendIZN22__llvm_libc_22_0_0_git8internal4exitEiE3$_0EEvT_.exit"
  br label %while.body.i.i12, !amdgpu.uniform !42

Flow8:                                            ; preds = %Flow, %"_ZN3rpc4PortILb0EE4sendIZN22__llvm_libc_22_0_0_git8internal4exitEiE3$_0EEvT_.exit"
  br label %"_ZN3rpc4PortILb0EE4recvIZN22__llvm_libc_22_0_0_git8internal4exitEiE3$_1EEvT_.exit", !amdgpu.uniform !42

while.body.i.i12:                                 ; preds = %while.body.i.i12.preheader, %while.body.i.i12
  tail call void @llvm.amdgcn.s.sleep(i32 2)
  %sunkaddr4 = getelementptr inbounds i8, ptr %23, i64 8, !amdgpu.uniform !42
  %29 = load ptr, ptr %sunkaddr4, align 8, !tbaa !85
  %arrayidx.i.i.i13 = getelementptr inbounds nuw i32, ptr %29, i64 %idxprom.i21.i, !amdgpu.uniform !42
  %30 = load atomic i32, ptr %arrayidx.i.i.i13 monotonic, align 4
  %31 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %30)
  %cmp.i.not.i.i14 = icmp eq i32 %31, %conv.i.i
  br i1 %cmp.i.not.i.i14, label %Flow, label %while.body.i.i12, !amdgpu.uniform !42

Flow:                                             ; preds = %while.body.i.i12
  br label %Flow8, !amdgpu.uniform !42

"_ZN3rpc4PortILb0EE4recvIZN22__llvm_libc_22_0_0_git8internal4exitEiE3$_1EEvT_.exit": ; preds = %Flow8
  fence acquire
  %sunkaddr5 = getelementptr inbounds i8, ptr %this, i64 28
  store <2 x i8> splat (i8 1), ptr %sunkaddr5, align 4, !tbaa !207
  ret void
}

; Function Attrs: convergent inlinehint mustprogress nofree norecurse nounwind
define internal fastcc void @"_ZN3rpc4PortILb0EE4sendIZN22__llvm_libc_22_0_0_git8internal4exitEiE3$_2EEvT_"(ptr noundef nonnull align 8 captures(none) dereferenceable(30) %this, i32 %fill.coerce.0.val) unnamed_addr #25 align 2 !type !208 {
entry:
  %owns_buffer = getelementptr inbounds nuw i8, ptr %this, i64 29, !amdgpu.uniform !42
  %0 = load i8, ptr %owns_buffer, align 1, !tbaa !199, !range !152, !noundef !42
  %loadedv = trunc nuw i8 %0 to i1
  %loadedv.inv = xor i1 %loadedv, true
  br i1 %loadedv.inv, label %cond.false, label %Flow6, !amdgpu.uniform !42

Flow6:                                            ; preds = %cond.false, %entry
  %1 = phi i1 [ %11, %cond.false ], [ poison, %entry ]
  %2 = phi i32 [ %.pre4, %cond.false ], [ poison, %entry ]
  %3 = phi i1 [ false, %cond.false ], [ true, %entry ]
  br i1 %3, label %cond.true, label %cond.end, !amdgpu.uniform !42

cond.true:                                        ; preds = %Flow6
  %out = getelementptr inbounds nuw i8, ptr %this, i64 24, !amdgpu.uniform !42
  %4 = load i32, ptr %out, align 8, !tbaa !202
  br label %cond.end, !amdgpu.uniform !42

cond.false:                                       ; preds = %entry
  %5 = load ptr, ptr %this, align 8, !tbaa !203, !nonnull !42, !align !204
  %index = getelementptr inbounds nuw i8, ptr %this, i64 20, !amdgpu.uniform !42
  %6 = load <2 x i32>, ptr %index, align 4, !tbaa !205
  %7 = extractelement <2 x i32> %6, i32 0
  %.pre4 = extractelement <2 x i32> %6, i32 1
  %inbox.i = getelementptr inbounds nuw i8, ptr %5, i64 8, !amdgpu.uniform !42
  %8 = load ptr, ptr %inbox.i, align 8, !tbaa !85
  %idxprom.i = zext i32 %7 to i64
  %arrayidx.i = getelementptr inbounds nuw i32, ptr %8, i64 %idxprom.i, !amdgpu.uniform !42
  %9 = load atomic i32, ptr %arrayidx.i monotonic, align 4
  %10 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %9)
  %11 = icmp eq i32 %10, %.pre4
  br label %Flow6, !amdgpu.uniform !42

cond.end:                                         ; preds = %cond.true, %Flow6
  %12 = phi i32 [ %2, %Flow6 ], [ %4, %cond.true ]
  %cond = phi i1 [ %1, %Flow6 ], [ true, %cond.true ]
  %cond.inv = xor i1 %cond, true
  br i1 %cond.inv, label %while.body.lr.ph.i, label %Flow5, !amdgpu.uniform !42

while.body.lr.ph.i:                               ; preds = %cond.end
  %sunkaddr = getelementptr inbounds i8, ptr %this, i64 20, !amdgpu.uniform !42
  %13 = load i32, ptr %sunkaddr, align 4, !tbaa !206
  %14 = load ptr, ptr %this, align 8, !tbaa !203, !nonnull !42, !align !204
  %idxprom.i.i = zext i32 %13 to i64
  br label %while.body.i, !amdgpu.uniform !42

Flow5:                                            ; preds = %Flow, %cond.end
  br label %_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit, !amdgpu.uniform !42

while.body.i:                                     ; preds = %while.body.lr.ph.i, %while.body.i
  tail call void @llvm.amdgcn.s.sleep(i32 2)
  %sunkaddr1 = getelementptr inbounds i8, ptr %14, i64 8, !amdgpu.uniform !42
  %15 = load ptr, ptr %sunkaddr1, align 8, !tbaa !85
  %arrayidx.i.i = getelementptr inbounds nuw i32, ptr %15, i64 %idxprom.i.i, !amdgpu.uniform !42
  %16 = load atomic i32, ptr %arrayidx.i.i monotonic, align 4
  %17 = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 %16)
  %cmp.i.not.i = icmp eq i32 %17, %12
  br i1 %cmp.i.not.i, label %Flow, label %while.body.i, !amdgpu.uniform !42

Flow:                                             ; preds = %while.body.i
  br label %Flow5, !amdgpu.uniform !42

_ZN3rpc7ProcessILb0EE18wait_for_ownershipEyjjj.exit: ; preds = %Flow5
  %out6 = getelementptr inbounds nuw i8, ptr %this, i64 24
  fence acquire
  %lane_size = getelementptr inbounds nuw i8, ptr %this, i64 16, !amdgpu.uniform !42
  %18 = load ptr, ptr %this, align 8, !tbaa !203, !nonnull !42, !align !204
  %19 = load <2 x i32>, ptr %lane_size, align 8, !tbaa !205
  %packet.i = getelementptr inbounds nuw i8, ptr %18, i64 32, !amdgpu.uniform !42
  %20 = load ptr, ptr %packet.i, align 8, !tbaa !81
  %21 = extractelement <2 x i32> %19, i64 0
  %22 = extractelement <2 x i32> %19, i64 1
  %mul.i = mul i32 %22, %21
  %idxprom.i19 = zext i32 %mul.i to i64
  %arrayidx.i20 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %20, i64 %idxprom.i19
  %23 = tail call i32 @llvm.amdgcn.mbcnt.lo(i32 -1, i32 0)
  %24 = tail call noundef i32 @llvm.amdgcn.mbcnt.hi(i32 -1, i32 %23)
  %idxprom.i21 = zext i32 %24 to i64
  %arrayidx.i22 = getelementptr inbounds nuw %"struct.rpc::Buffer", ptr %arrayidx.i20, i64 %idxprom.i21
  store i32 %fill.coerce.0.val, ptr %arrayidx.i22, align 8, !tbaa !205
  %sunkaddr2 = getelementptr inbounds i8, ptr %this, i64 20, !amdgpu.uniform !42
  %25 = load <2 x i32>, ptr %sunkaddr2, align 4, !tbaa !205
  %26 = extractelement <2 x i32> %25, i32 0
  %27 = extractelement <2 x i32> %25, i32 1
  %tobool.not.i = icmp eq i32 %27, 0
  %conv.i = zext i1 %tobool.not.i to i32
  fence release
  %outbox.i = getelementptr inbounds nuw i8, ptr %18, i64 16, !amdgpu.uniform !42
  %28 = load ptr, ptr %outbox.i, align 8, !tbaa !83
  %idxprom.i23 = zext i32 %26 to i64
  %arrayidx.i24 = getelementptr inbounds nuw i32, ptr %28, i64 %idxprom.i23
  store atomic i32 %conv.i, ptr %arrayidx.i24 monotonic, align 4
  store i32 %conv.i, ptr %out6, align 8, !tbaa !202
  %receive = getelementptr inbounds nuw i8, ptr %this, i64 28
  store <2 x i8> zeroinitializer, ptr %receive, align 4, !tbaa !207
  ret void
}

; Function Attrs: convergent inlinehint mustprogress nofree norecurse nounwind willreturn
define internal fastcc void @_ZN3rpc4PortILb0EE5closeEv(ptr noundef nonnull readonly align 8 captures(none) dereferenceable(30) %this) unnamed_addr #26 comdat align 2 !type !210 {
entry:
  tail call void @llvm.amdgcn.wave.barrier()
  %0 = load <2 x i64>, ptr %this, align 8, !tbaa !57
  %1 = extractelement <2 x i64> %0, i32 0
  %2 = inttoptr i64 %1 to ptr
  %3 = extractelement <2 x i64> %0, i32 1
  %index5 = getelementptr inbounds nuw i8, ptr %this, i64 20, !amdgpu.uniform !42
  %4 = load i32, ptr %index5, align 4, !tbaa !206
  fence syncscope("agent") release
  %5 = tail call i32 @llvm.amdgcn.mbcnt.lo(i32 -1, i32 0)
  %6 = tail call noundef i32 @llvm.amdgcn.mbcnt.hi(i32 -1, i32 %5)
  %conv.i.i.i = zext i32 %6 to i64
  %7 = tail call range(i64 0, 65) i64 @llvm.cttz.i64(i64 %3, i1 true)
  %iszero.i.i.i.i = icmp ne i64 %3, 0
  %cmp2.i.i.i = icmp eq i64 %7, %conv.i.i.i
  %cmp.i.i.i = select i1 %iszero.i.i.i.i, i1 %cmp2.i.i.i, i1 false
  %div8.i.i = lshr i32 %4, 5
  %8 = and i32 %4, 31
  %idxprom.i.i = zext nneg i32 %div8.i.i to i64
  %9 = getelementptr inbounds nuw i32, ptr %2, i64 %idxprom.i.i
  %arrayidx.i.i1 = getelementptr inbounds nuw i8, ptr %9, i64 40
  %conv4.i.i = zext i1 %cmp.i.i.i to i32
  %shl.i.i = shl nuw i32 %conv4.i.i, %8
  %xor.i.i = xor i32 %shl.i.i, -1
  %10 = atomicrmw and ptr %arrayidx.i.i1, i32 %xor.i.i syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !42, !amdgpu.no.remote.memory !42
  tail call void @llvm.amdgcn.wave.barrier()
  ret void
}

; Function Attrs: cold convergent nocallback nofree noreturn nounwind
declare void @llvm.amdgcn.endpgm() #27

; Function Attrs: mustprogress nocallback nofree nosync nounwind willreturn
declare i64 @llvm.readsteadycounter() #20

; Function Attrs: nocallback nofree nounwind willreturn memory(argmem: readwrite)
declare void @llvm.memcpy.p0.p4.i64(ptr noalias writeonly captures(none), ptr addrspace(4) noalias readonly captures(none), i64, i1 immarg) #28

; Function Attrs: convergent nocallback nofree nounwind willreturn memory(none)
declare i32 @llvm.amdgcn.readlane.i32(i32, i32) #29

; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)
declare ptr @llvm.amdgcn.addrspacecast.nonnull.p0.p5(ptr addrspace(5)) #30

; Function Attrs: convergent nocallback nofree nounwind willreturn memory(none)
declare i32 @llvm.amdgcn.writelane.i32(i32, i32, i32) #29

; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)
declare i32 @llvm.amdgcn.mul.u24.i32(i32, i32) #30

; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)
declare noundef align 4 ptr addrspace(4) @llvm.amdgcn.kernarg.segment.ptr() #30

; Function Attrs: nocallback nofree nounwind willreturn
declare { i1, i64 } @llvm.amdgcn.if.i64(i1) #31

; Function Attrs: nocallback nofree nounwind willreturn
declare void @llvm.amdgcn.end.cf.i64(i64) #31

; Function Attrs: nocallback nofree nounwind willreturn memory(none)
declare i64 @llvm.amdgcn.if.break.i64(i1, i64) #32

; Function Attrs: nocallback nofree nounwind willreturn
declare i1 @llvm.amdgcn.loop.i64(i64) #31

; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)
declare { i64, i1 } @llvm.uadd.with.overflow.i64(i64, i64) #30

; Function Attrs: nocallback nofree nounwind willreturn
declare { i1, i64 } @llvm.amdgcn.else.i64.i64(i64) #31

attributes #0 = { convergent mustprogress nounwind "amdgpu-flat-work-group-size"="1,1" "amdgpu-max-num-workgroups"="1,1,1" "amdgpu-wave-limiter"="true" "frame-pointer"="all" "no-builtins" "no-trapping-math"="true" "stack-protector-buffer-size"="8" "uniform-work-group-size"="false" }
attributes #1 = { convergent mustprogress nounwind "frame-pointer"="all" "no-builtins" "no-trapping-math"="true" "stack-protector-buffer-size"="8" "uniform-work-group-size"="false" }
attributes #2 = { mustprogress nocallback nofree nosync nounwind speculatable willreturn memory(none) }
attributes #3 = { convergent mustprogress nounwind "amdgpu-wave-limiter"="true" "frame-pointer"="all" "no-builtins" "no-trapping-math"="true" "stack-protector-buffer-size"="8" "uniform-work-group-size"="false" }
attributes #4 = { cold convergent mustprogress noreturn nounwind "amdgpu-flat-work-group-size"="1,1" "amdgpu-max-num-workgroups"="1,1,1" "amdgpu-wave-limiter"="true" "frame-pointer"="all" "no-builtins" "no-trapping-math"="true" "stack-protector-buffer-size"="8" "uniform-work-group-size"="false" }
attributes #5 = { mustprogress nofree norecurse nosync nounwind willreturn memory(readwrite, argmem: write, inaccessiblemem: none) "amdgpu-agpr-alloc"="0" "amdgpu-memory-bound"="true" "amdgpu-no-completion-action" "amdgpu-no-default-queue" "amdgpu-no-dispatch-id" "amdgpu-no-dispatch-ptr" "amdgpu-no-flat-scratch-init" "amdgpu-no-heap-ptr" "amdgpu-no-hostcall-ptr" "amdgpu-no-implicitarg-ptr" "amdgpu-no-lds-kernel-id" "amdgpu-no-multigrid-sync-arg" "amdgpu-no-queue-ptr" "amdgpu-no-workgroup-id-x" "amdgpu-no-workgroup-id-y" "amdgpu-no-workgroup-id-z" "amdgpu-no-workitem-id-x" "amdgpu-no-workitem-id-y" "amdgpu-no-workitem-id-z" "no-builtins" "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-cpu"="gfx90a" "target-features"="+16-bit-insts,+atomic-buffer-global-pk-add-f16-insts,+atomic-fadd-rtn-insts,+atomic-fmin-fmax-global-f64,+ci-insts,+dl-insts,+dot1-insts,+dot10-insts,+dot2-insts,+dot3-insts,+dot4-insts,+dot5-insts,+dot6-insts,+dot7-insts,+dpp,+gfx8-insts,+gfx9-insts,+gfx90a-insts,+mai-insts,+s-memrealtime,+s-memtime-inst,+wavefrontsize64" "uniform-work-group-size"="false" }
attributes #6 = { mustprogress nofree norecurse nosync nounwind willreturn memory(none) "amdgpu-agpr-alloc"="0" "amdgpu-no-completion-action" "amdgpu-no-default-queue" "amdgpu-no-dispatch-id" "amdgpu-no-dispatch-ptr" "amdgpu-no-flat-scratch-init" "amdgpu-no-heap-ptr" "amdgpu-no-hostcall-ptr" "amdgpu-no-implicitarg-ptr" "amdgpu-no-lds-kernel-id" "amdgpu-no-multigrid-sync-arg" "amdgpu-no-queue-ptr" "amdgpu-no-workgroup-id-x" "amdgpu-no-workgroup-id-y" "amdgpu-no-workgroup-id-z" "amdgpu-no-workitem-id-x" "amdgpu-no-workitem-id-y" "amdgpu-no-workitem-id-z" "no-builtins" "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-cpu"="gfx90a" "target-features"="+16-bit-insts,+atomic-buffer-global-pk-add-f16-insts,+atomic-fadd-rtn-insts,+atomic-fmin-fmax-global-f64,+ci-insts,+dl-insts,+dot1-insts,+dot10-insts,+dot2-insts,+dot3-insts,+dot4-insts,+dot5-insts,+dot6-insts,+dot7-insts,+dpp,+gfx8-insts,+gfx9-insts,+gfx90a-insts,+mai-insts,+s-memrealtime,+s-memtime-inst,+wavefrontsize64" "uniform-work-group-size"="false" }
attributes #7 = { cold inlinehint mustprogress noreturn nounwind memory(inaccessiblemem: write) "amdgpu-agpr-alloc"="0" "amdgpu-no-completion-action" "amdgpu-no-default-queue" "amdgpu-no-dispatch-id" "amdgpu-no-dispatch-ptr" "amdgpu-no-flat-scratch-init" "amdgpu-no-heap-ptr" "amdgpu-no-hostcall-ptr" "amdgpu-no-implicitarg-ptr" "amdgpu-no-lds-kernel-id" "amdgpu-no-multigrid-sync-arg" "amdgpu-no-queue-ptr" "amdgpu-no-workgroup-id-x" "amdgpu-no-workgroup-id-y" "amdgpu-no-workgroup-id-z" "amdgpu-no-workitem-id-x" "amdgpu-no-workitem-id-y" "amdgpu-no-workitem-id-z" "no-builtins" "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-cpu"="gfx90a" "target-features"="+16-bit-insts,+atomic-buffer-global-pk-add-f16-insts,+atomic-fadd-rtn-insts,+atomic-fmin-fmax-global-f64,+ci-insts,+dl-insts,+dot1-insts,+dot10-insts,+dot2-insts,+dot3-insts,+dot4-insts,+dot5-insts,+dot6-insts,+dot7-insts,+dpp,+gfx8-insts,+gfx9-insts,+gfx90a-insts,+mai-insts,+s-memrealtime,+s-memtime-inst,+wavefrontsize64" "uniform-work-group-size"="false" }
attributes #8 = { convergent mustprogress nofree norecurse nounwind "amdgpu-agpr-alloc"="0" "amdgpu-no-completion-action" "amdgpu-no-default-queue" "amdgpu-no-dispatch-id" "amdgpu-no-dispatch-ptr" "amdgpu-no-flat-scratch-init" "amdgpu-no-heap-ptr" "amdgpu-no-hostcall-ptr" "amdgpu-no-implicitarg-ptr" "amdgpu-no-lds-kernel-id" "amdgpu-no-multigrid-sync-arg" "amdgpu-no-queue-ptr" "amdgpu-no-workgroup-id-x" "amdgpu-no-workgroup-id-y" "amdgpu-no-workgroup-id-z" "amdgpu-no-workitem-id-x" "amdgpu-no-workitem-id-y" "amdgpu-no-workitem-id-z" "no-builtins" "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-cpu"="gfx90a" "target-features"="+16-bit-insts,+atomic-buffer-global-pk-add-f16-insts,+atomic-fadd-rtn-insts,+atomic-fmin-fmax-global-f64,+ci-insts,+dl-insts,+dot1-insts,+dot10-insts,+dot2-insts,+dot3-insts,+dot4-insts,+dot5-insts,+dot6-insts,+dot7-insts,+dpp,+gfx8-insts,+gfx9-insts,+gfx90a-insts,+mai-insts,+s-memrealtime,+s-memtime-inst,+wavefrontsize64" "uniform-work-group-size"="false" }
attributes #9 = { mustprogress nocallback nofree nosync nounwind willreturn memory(argmem: readwrite) }
attributes #10 = { mustprogress nocallback nofree nosync nounwind willreturn memory(inaccessiblemem: write) }
attributes #11 = { mustprogress nocallback nofree nounwind willreturn memory(argmem: write) }
attributes #12 = { mustprogress nocallback nofree nounwind willreturn memory(argmem: readwrite) }
attributes #13 = { convergent mustprogress nounwind "no-builtins" "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-cpu"="gfx90a" "target-features"="+16-bit-insts,+atomic-buffer-global-pk-add-f16-insts,+atomic-fadd-rtn-insts,+atomic-fmin-fmax-global-f64,+ci-insts,+dl-insts,+dot1-insts,+dot10-insts,+dot2-insts,+dot3-insts,+dot4-insts,+dot5-insts,+dot6-insts,+dot7-insts,+dpp,+gfx8-insts,+gfx9-insts,+gfx90a-insts,+mai-insts,+s-memrealtime,+s-memtime-inst,+wavefrontsize64" "uniform-work-group-size"="false" }
attributes #14 = { mustprogress nobuiltin nounwind allocsize(0) "no-builtins" "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-cpu"="gfx90a" "target-features"="+16-bit-insts,+atomic-buffer-global-pk-add-f16-insts,+atomic-fadd-rtn-insts,+atomic-fmin-fmax-global-f64,+ci-insts,+dl-insts,+dot1-insts,+dot10-insts,+dot2-insts,+dot3-insts,+dot4-insts,+dot5-insts,+dot6-insts,+dot7-insts,+dpp,+gfx8-insts,+gfx9-insts,+gfx90a-insts,+mai-insts,+s-memrealtime,+s-memtime-inst,+wavefrontsize64" "uniform-work-group-size"="false" }
attributes #15 = { mustprogress nobuiltin nounwind "no-builtins" "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-cpu"="gfx90a" "target-features"="+16-bit-insts,+atomic-buffer-global-pk-add-f16-insts,+atomic-fadd-rtn-insts,+atomic-fmin-fmax-global-f64,+ci-insts,+dl-insts,+dot1-insts,+dot10-insts,+dot2-insts,+dot3-insts,+dot4-insts,+dot5-insts,+dot6-insts,+dot7-insts,+dpp,+gfx8-insts,+gfx9-insts,+gfx90a-insts,+mai-insts,+s-memrealtime,+s-memtime-inst,+wavefrontsize64" "uniform-work-group-size"="false" }
attributes #16 = { cold noreturn nounwind memory(inaccessiblemem: write) }
attributes #17 = { mustprogress nocallback nofree nosync nounwind willreturn memory(none) }
attributes #18 = { convergent mustprogress nocallback nofree nounwind willreturn memory(none) }
attributes #19 = { convergent mustprogress nocallback nofree nounwind willreturn }
attributes #20 = { mustprogress nocallback nofree nosync nounwind willreturn }
attributes #21 = { convergent mustprogress nounwind "amdgpu-flat-work-group-size"="1,1" "amdgpu-max-num-workgroups"="1,1,1" "amdgpu-memory-bound"="true" "frame-pointer"="all" "no-trapping-math"="true" "stack-protector-buffer-size"="8" "uniform-work-group-size"="false" }
attributes #22 = { convergent inlinehint mustprogress nounwind "frame-pointer"="all" "no-trapping-math"="true" "stack-protector-buffer-size"="8" "uniform-work-group-size"="false" }
attributes #23 = { cold convergent mustprogress noreturn nounwind "amdgpu-flat-work-group-size"="1,1" "amdgpu-max-num-workgroups"="1,1,1" "frame-pointer"="all" "no-trapping-math"="true" "stack-protector-buffer-size"="8" "uniform-work-group-size"="false" }
attributes #24 = { cold convergent mustprogress nofree norecurse noreturn nounwind "amdgpu-agpr-alloc"="0" "amdgpu-flat-work-group-size"="1,1" "amdgpu-max-num-workgroups"="1,1,1" "amdgpu-no-completion-action" "amdgpu-no-default-queue" "amdgpu-no-dispatch-id" "amdgpu-no-dispatch-ptr" "amdgpu-no-heap-ptr" "amdgpu-no-hostcall-ptr" "amdgpu-no-implicitarg-ptr" "amdgpu-no-lds-kernel-id" "amdgpu-no-multigrid-sync-arg" "amdgpu-no-queue-ptr" "amdgpu-no-workgroup-id-x" "amdgpu-no-workgroup-id-y" "amdgpu-no-workgroup-id-z" "amdgpu-no-workitem-id-x" "amdgpu-no-workitem-id-y" "amdgpu-no-workitem-id-z" "frame-pointer"="all" "no-trapping-math"="true" "stack-protector-buffer-size"="8" "uniform-work-group-size"="false" }
attributes #25 = { convergent inlinehint mustprogress nofree norecurse nounwind "amdgpu-agpr-alloc"="0" "amdgpu-flat-work-group-size"="1,1" "amdgpu-max-num-workgroups"="1,1,1" "amdgpu-no-completion-action" "amdgpu-no-default-queue" "amdgpu-no-dispatch-id" "amdgpu-no-dispatch-ptr" "amdgpu-no-flat-scratch-init" "amdgpu-no-heap-ptr" "amdgpu-no-hostcall-ptr" "amdgpu-no-implicitarg-ptr" "amdgpu-no-lds-kernel-id" "amdgpu-no-multigrid-sync-arg" "amdgpu-no-queue-ptr" "amdgpu-no-workgroup-id-x" "amdgpu-no-workgroup-id-y" "amdgpu-no-workgroup-id-z" "amdgpu-no-workitem-id-x" "amdgpu-no-workitem-id-y" "amdgpu-no-workitem-id-z" "frame-pointer"="all" "no-trapping-math"="true" "stack-protector-buffer-size"="8" "uniform-work-group-size"="false" }
attributes #26 = { convergent inlinehint mustprogress nofree norecurse nounwind willreturn "amdgpu-agpr-alloc"="0" "amdgpu-flat-work-group-size"="1,1" "amdgpu-max-num-workgroups"="1,1,1" "amdgpu-no-completion-action" "amdgpu-no-default-queue" "amdgpu-no-dispatch-id" "amdgpu-no-dispatch-ptr" "amdgpu-no-flat-scratch-init" "amdgpu-no-heap-ptr" "amdgpu-no-hostcall-ptr" "amdgpu-no-implicitarg-ptr" "amdgpu-no-lds-kernel-id" "amdgpu-no-multigrid-sync-arg" "amdgpu-no-queue-ptr" "amdgpu-no-workgroup-id-x" "amdgpu-no-workgroup-id-y" "amdgpu-no-workgroup-id-z" "amdgpu-no-workitem-id-x" "amdgpu-no-workitem-id-y" "amdgpu-no-workitem-id-z" "frame-pointer"="all" "no-trapping-math"="true" "stack-protector-buffer-size"="8" "uniform-work-group-size"="false" }
attributes #27 = { cold convergent nocallback nofree noreturn nounwind }
attributes #28 = { nocallback nofree nounwind willreturn memory(argmem: readwrite) }
attributes #29 = { convergent nocallback nofree nounwind willreturn memory(none) }
attributes #30 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }
attributes #31 = { nocallback nofree nounwind willreturn }
attributes #32 = { nocallback nofree nounwind willreturn memory(none) }
attributes #33 = { convergent nobuiltin nounwind "no-builtins" }
attributes #34 = { nounwind }
attributes #35 = { convergent nobuiltin noreturn nounwind "no-builtins" }
attributes #36 = { convergent nounwind }
attributes #37 = { convergent noreturn nounwind }
attributes #38 = { convergent }

!llvm.ident = !{!35}
!llvm.module.flags = !{!36, !37, !38, !39, !40, !41}

!0 = !{i64 16, !"_ZTS46LlvmLibcCharacterConverterUTF32To8Test_OneByte"}
!1 = !{i64 32, !"_ZTSM46LlvmLibcCharacterConverterUTF32To8Test_OneByteFvvE.virtual"}
!2 = !{i64 40, !"_ZTSM46LlvmLibcCharacterConverterUTF32To8Test_OneByteFvvE.virtual"}
!3 = !{i64 48, !"_ZTSM46LlvmLibcCharacterConverterUTF32To8Test_OneByteFvvE.virtual"}
!4 = !{i64 56, !"_ZTSM46LlvmLibcCharacterConverterUTF32To8Test_OneByteKFPKcvE.virtual"}
!5 = !{i64 16, !"_ZTSN22__llvm_libc_22_0_0_git7testing4TestE"}
!6 = !{i64 32, !"_ZTSMN22__llvm_libc_22_0_0_git7testing4TestEFvvE.virtual"}
!7 = !{i64 40, !"_ZTSMN22__llvm_libc_22_0_0_git7testing4TestEFvvE.virtual"}
!8 = !{i64 48, !"_ZTSMN22__llvm_libc_22_0_0_git7testing4TestEFvvE.virtual"}
!9 = !{i64 56, !"_ZTSMN22__llvm_libc_22_0_0_git7testing4TestEKFPKcvE.virtual"}
!10 = !{i64 16, !"_ZTS46LlvmLibcCharacterConverterUTF32To8Test_TwoByte"}
!11 = !{i64 32, !"_ZTSM46LlvmLibcCharacterConverterUTF32To8Test_TwoByteFvvE.virtual"}
!12 = !{i64 40, !"_ZTSM46LlvmLibcCharacterConverterUTF32To8Test_TwoByteFvvE.virtual"}
!13 = !{i64 48, !"_ZTSM46LlvmLibcCharacterConverterUTF32To8Test_TwoByteFvvE.virtual"}
!14 = !{i64 56, !"_ZTSM46LlvmLibcCharacterConverterUTF32To8Test_TwoByteKFPKcvE.virtual"}
!15 = !{i64 16, !"_ZTS48LlvmLibcCharacterConverterUTF32To8Test_ThreeByte"}
!16 = !{i64 32, !"_ZTSM48LlvmLibcCharacterConverterUTF32To8Test_ThreeByteFvvE.virtual"}
!17 = !{i64 40, !"_ZTSM48LlvmLibcCharacterConverterUTF32To8Test_ThreeByteFvvE.virtual"}
!18 = !{i64 48, !"_ZTSM48LlvmLibcCharacterConverterUTF32To8Test_ThreeByteFvvE.virtual"}
!19 = !{i64 56, !"_ZTSM48LlvmLibcCharacterConverterUTF32To8Test_ThreeByteKFPKcvE.virtual"}
!20 = !{i64 16, !"_ZTS47LlvmLibcCharacterConverterUTF32To8Test_FourByte"}
!21 = !{i64 32, !"_ZTSM47LlvmLibcCharacterConverterUTF32To8Test_FourByteFvvE.virtual"}
!22 = !{i64 40, !"_ZTSM47LlvmLibcCharacterConverterUTF32To8Test_FourByteFvvE.virtual"}
!23 = !{i64 48, !"_ZTSM47LlvmLibcCharacterConverterUTF32To8Test_FourByteFvvE.virtual"}
!24 = !{i64 56, !"_ZTSM47LlvmLibcCharacterConverterUTF32To8Test_FourByteKFPKcvE.virtual"}
!25 = !{i64 16, !"_ZTS60LlvmLibcCharacterConverterUTF32To8Test_CantPushMidConversion"}
!26 = !{i64 32, !"_ZTSM60LlvmLibcCharacterConverterUTF32To8Test_CantPushMidConversionFvvE.virtual"}
!27 = !{i64 40, !"_ZTSM60LlvmLibcCharacterConverterUTF32To8Test_CantPushMidConversionFvvE.virtual"}
!28 = !{i64 48, !"_ZTSM60LlvmLibcCharacterConverterUTF32To8Test_CantPushMidConversionFvvE.virtual"}
!29 = !{i64 56, !"_ZTSM60LlvmLibcCharacterConverterUTF32To8Test_CantPushMidConversionKFPKcvE.virtual"}
!30 = !{i64 16, !"_ZTS51LlvmLibcCharacterConverterUTF32To8Test_InvalidState"}
!31 = !{i64 32, !"_ZTSM51LlvmLibcCharacterConverterUTF32To8Test_InvalidStateFvvE.virtual"}
!32 = !{i64 40, !"_ZTSM51LlvmLibcCharacterConverterUTF32To8Test_InvalidStateFvvE.virtual"}
!33 = !{i64 48, !"_ZTSM51LlvmLibcCharacterConverterUTF32To8Test_InvalidStateFvvE.virtual"}
!34 = !{i64 56, !"_ZTSM51LlvmLibcCharacterConverterUTF32To8Test_InvalidStateKFPKcvE.virtual"}
!35 = !{!"clang version 22.0.0git (https://github.com/llvm/llvm-project.git 74b9484fd62d6be9bc49e154800ceef0d74ef24f)"}
!36 = !{i32 1, !"wchar_size", i32 4}
!37 = !{i32 8, !"PIC Level", i32 2}
!38 = !{i32 7, !"frame-pointer", i32 2}
!39 = !{i32 1, !"ThinLTO", i32 0}
!40 = !{i32 1, !"EnableSplitLTOUnit", i32 1}
!41 = !{i32 1, !"amdhsa_code_object_version", i32 600}
!42 = !{}
!43 = !{!44, !48, i64 1024}
!44 = !{!"_ZTSN22__llvm_libc_22_0_0_git11FixedVectorINS_10AtExitUnitELm64EEE", !45, i64 0, !48, i64 1024}
!45 = !{!"_ZTSN22__llvm_libc_22_0_0_git3cpp5arrayINS_10AtExitUnitELm64EEE", !46, i64 0}
!46 = !{!"omnipotent char", !47, i64 0}
!47 = !{!"Simple C++ TBAA"}
!48 = !{!"long", !46, i64 0}
!49 = !{!50, !50, i64 0}
!50 = !{!"any pointer", !46, i64 0}
!51 = !{!48, !48, i64 0}
!52 = !{!53, !53, i64 0}
!53 = !{!"p2 omnipotent char", !54, i64 0}
!54 = !{!"any p2 pointer", !50, i64 0}
!55 = !{!56, !56, i64 0}
!56 = !{!"p1 omnipotent char", !50, i64 0}
!57 = !{!46, !46, i64 0}
!58 = !{!"section_prefix", !"unlikely"}
!59 = !{!60, !62, i64 16}
!60 = !{!"_ZTSN22__llvm_libc_22_0_0_git7testing4TestE", !61, i64 8, !62, i64 16}
!61 = !{!"p1 _ZTSN22__llvm_libc_22_0_0_git7testing4TestE", !50, i64 0}
!62 = !{!"p1 _ZTSN22__llvm_libc_22_0_0_git7testing8internal10RunContextE", !50, i64 0}
!63 = !{!61, !61, i64 0}
!64 = !{!60, !61, i64 8}
!65 = !{!66, !67, i64 0}
!66 = !{!"_ZTSN22__llvm_libc_22_0_0_git7testing8internal10RunContextE", !67, i64 0}
!67 = !{!"_ZTSN22__llvm_libc_22_0_0_git7testing8internal10RunContext9RunResultE", !46, i64 0}
!68 = !{!69, !71, i64 0}
!69 = !{!"_ZTSN3rpc6ClientE", !70, i64 0}
!70 = !{!"_ZTSN3rpc7ProcessILb0EEE", !71, i64 0, !72, i64 8, !72, i64 16, !73, i64 24, !74, i64 32, !46, i64 40}
!71 = !{!"int", !46, i64 0}
!72 = !{!"p1 int", !50, i64 0}
!73 = !{!"p1 _ZTSN3rpc6HeaderE", !50, i64 0}
!74 = !{!"p1 _ZTSN3rpc6BufferE", !50, i64 0}
!75 = !{!72, !72, i64 0}
!76 = !{!69, !73, i64 24}
!77 = !{!78, !71, i64 8}
!78 = !{!"_ZTSN3rpc6HeaderE", !79, i64 0, !71, i64 8}
!79 = !{!"long long", !46, i64 0}
!80 = !{!78, !79, i64 0}
!81 = !{!70, !74, i64 32}
!82 = !{!79, !79, i64 0}
!83 = !{!70, !72, i64 16}
!84 = !{!70, !73, i64 24}
!85 = !{!70, !72, i64 8}
!86 = !{!87, !89}
!87 = distinct !{!87, !88, !"_ZN22__llvm_libc_22_0_0_git3cpp8internal13to_dec_stringIiEENS0_6stringET_: %agg.result"}
!88 = distinct !{!88, !"_ZN22__llvm_libc_22_0_0_git3cpp8internal13to_dec_stringIiEENS0_6stringET_"}
!89 = distinct !{!89, !90, !"_ZN22__llvm_libc_22_0_0_git3cpp9to_stringEi: %agg.result"}
!90 = distinct !{!90, !"_ZN22__llvm_libc_22_0_0_git3cpp9to_stringEi"}
!91 = !{!92, !48, i64 16}
!92 = !{!"_ZTSN22__llvm_libc_22_0_0_git15IntegerToStringIiNS_7details3FmtILh10ELb0ELb0ELb0ELm1EEEEE", !93, i64 0, !48, i64 16}
!93 = !{!"_ZTSN22__llvm_libc_22_0_0_git3cpp5arrayIcLm11EEE", !46, i64 0}
!94 = !{!95, !97, !99}
!95 = distinct !{!95, !96, !"_ZN22__llvm_libc_22_0_0_git3cpp8internal13to_dec_stringIiEENS0_6stringET_: %agg.result"}
!96 = distinct !{!96, !"_ZN22__llvm_libc_22_0_0_git3cpp8internal13to_dec_stringIiEENS0_6stringET_"}
!97 = distinct !{!97, !98, !"_ZN22__llvm_libc_22_0_0_git3cpp9to_stringEi: %agg.result"}
!98 = distinct !{!98, !"_ZN22__llvm_libc_22_0_0_git3cpp9to_stringEi"}
!99 = distinct !{!99, !100, !"_ZN22__llvm_libc_22_0_0_git7testing8internal13describeValueIcEENS_3cpp9enable_ifIXaasr3cppE13is_integral_vIT_ElestS5_Lm8EENS3_6stringEE4typeES5_: %agg.result"}
!100 = distinct !{!100, !"_ZN22__llvm_libc_22_0_0_git7testing8internal13describeValueIcEENS_3cpp9enable_ifIXaasr3cppE13is_integral_vIT_ElestS5_Lm8EENS3_6stringEE4typeES5_"}
!101 = !{i32 1, i32 4, i32 5, i32 10}
!102 = !{!103, !105, !107}
!103 = distinct !{!103, !104, !"_ZN22__llvm_libc_22_0_0_git3cpp8internal13to_dec_stringIiEENS0_6stringET_: %agg.result"}
!104 = distinct !{!104, !"_ZN22__llvm_libc_22_0_0_git3cpp8internal13to_dec_stringIiEENS0_6stringET_"}
!105 = distinct !{!105, !106, !"_ZN22__llvm_libc_22_0_0_git3cpp9to_stringEi: %agg.result"}
!106 = distinct !{!106, !"_ZN22__llvm_libc_22_0_0_git3cpp9to_stringEi"}
!107 = distinct !{!107, !108, !"_ZN22__llvm_libc_22_0_0_git7testing8internal13describeValueIcEENS_3cpp9enable_ifIXaasr3cppE13is_integral_vIT_ElestS5_Lm8EENS3_6stringEE4typeES5_: %agg.result"}
!108 = distinct !{!108, !"_ZN22__llvm_libc_22_0_0_git7testing8internal13describeValueIcEENS_3cpp9enable_ifIXaasr3cppE13is_integral_vIT_ElestS5_Lm8EENS3_6stringEE4typeES5_"}
!109 = !{!110, !112}
!110 = distinct !{!110, !111, !"_ZN22__llvm_libc_22_0_0_git3cpp8internal13to_dec_stringIiEENS0_6stringET_: %agg.result"}
!111 = distinct !{!111, !"_ZN22__llvm_libc_22_0_0_git3cpp8internal13to_dec_stringIiEENS0_6stringET_"}
!112 = distinct !{!112, !113, !"_ZN22__llvm_libc_22_0_0_git3cpp9to_stringEi: %agg.result"}
!113 = distinct !{!113, !"_ZN22__llvm_libc_22_0_0_git3cpp9to_stringEi"}
!114 = !{!115, !117, !119}
!115 = distinct !{!115, !116, !"_ZN22__llvm_libc_22_0_0_git3cpp8internal13to_dec_stringIiEENS0_6stringET_: %agg.result"}
!116 = distinct !{!116, !"_ZN22__llvm_libc_22_0_0_git3cpp8internal13to_dec_stringIiEENS0_6stringET_"}
!117 = distinct !{!117, !118, !"_ZN22__llvm_libc_22_0_0_git3cpp9to_stringEi: %agg.result"}
!118 = distinct !{!118, !"_ZN22__llvm_libc_22_0_0_git3cpp9to_stringEi"}
!119 = distinct !{!119, !120, !"_ZN22__llvm_libc_22_0_0_git7testing8internal13describeValueIiEENS_3cpp9enable_ifIXaasr3cppE13is_integral_vIT_ElestS5_Lm8EENS3_6stringEE4typeES5_: %agg.result"}
!120 = distinct !{!120, !"_ZN22__llvm_libc_22_0_0_git7testing8internal13describeValueIiEENS_3cpp9enable_ifIXaasr3cppE13is_integral_vIT_ElestS5_Lm8EENS3_6stringEE4typeES5_"}
!121 = !{!122, !124, !126}
!122 = distinct !{!122, !123, !"_ZN22__llvm_libc_22_0_0_git3cpp8internal13to_dec_stringIiEENS0_6stringET_: %agg.result"}
!123 = distinct !{!123, !"_ZN22__llvm_libc_22_0_0_git3cpp8internal13to_dec_stringIiEENS0_6stringET_"}
!124 = distinct !{!124, !125, !"_ZN22__llvm_libc_22_0_0_git3cpp9to_stringEi: %agg.result"}
!125 = distinct !{!125, !"_ZN22__llvm_libc_22_0_0_git3cpp9to_stringEi"}
!126 = distinct !{!126, !127, !"_ZN22__llvm_libc_22_0_0_git7testing8internal13describeValueIiEENS_3cpp9enable_ifIXaasr3cppE13is_integral_vIT_ElestS5_Lm8EENS3_6stringEE4typeES5_: %agg.result"}
!127 = distinct !{!127, !"_ZN22__llvm_libc_22_0_0_git7testing8internal13describeValueIiEENS_3cpp9enable_ifIXaasr3cppE13is_integral_vIT_ElestS5_Lm8EENS3_6stringEE4typeES5_"}
!128 = !{!129, !131}
!129 = distinct !{!129, !130, !"_ZN22__llvm_libc_22_0_0_git3cpp8internal13to_dec_stringIiEENS0_6stringET_: %agg.result"}
!130 = distinct !{!130, !"_ZN22__llvm_libc_22_0_0_git3cpp8internal13to_dec_stringIiEENS0_6stringET_"}
!131 = distinct !{!131, !132, !"_ZN22__llvm_libc_22_0_0_git3cpp9to_stringEi: %agg.result"}
!132 = distinct !{!132, !"_ZN22__llvm_libc_22_0_0_git3cpp9to_stringEi"}
!133 = !{!134, !136, !138}
!134 = distinct !{!134, !135, !"_ZN22__llvm_libc_22_0_0_git3cpp8internal13to_dec_stringIiEENS0_6stringET_: %agg.result"}
!135 = distinct !{!135, !"_ZN22__llvm_libc_22_0_0_git3cpp8internal13to_dec_stringIiEENS0_6stringET_"}
!136 = distinct !{!136, !137, !"_ZN22__llvm_libc_22_0_0_git3cpp9to_stringEi: %agg.result"}
!137 = distinct !{!137, !"_ZN22__llvm_libc_22_0_0_git3cpp9to_stringEi"}
!138 = distinct !{!138, !139, !"_ZN22__llvm_libc_22_0_0_git7testing8internal13describeValueIbEENS_3cpp9enable_ifIXaasr3cppE13is_integral_vIT_ElestS5_Lm8EENS3_6stringEE4typeES5_: %agg.result"}
!139 = distinct !{!139, !"_ZN22__llvm_libc_22_0_0_git7testing8internal13describeValueIbEENS_3cpp9enable_ifIXaasr3cppE13is_integral_vIT_ElestS5_Lm8EENS3_6stringEE4typeES5_"}
!140 = !{!138}
!141 = !{!142, !144, !146}
!142 = distinct !{!142, !143, !"_ZN22__llvm_libc_22_0_0_git3cpp8internal13to_dec_stringIiEENS0_6stringET_: %agg.result"}
!143 = distinct !{!143, !"_ZN22__llvm_libc_22_0_0_git3cpp8internal13to_dec_stringIiEENS0_6stringET_"}
!144 = distinct !{!144, !145, !"_ZN22__llvm_libc_22_0_0_git3cpp9to_stringEi: %agg.result"}
!145 = distinct !{!145, !"_ZN22__llvm_libc_22_0_0_git3cpp9to_stringEi"}
!146 = distinct !{!146, !147, !"_ZN22__llvm_libc_22_0_0_git7testing8internal13describeValueIbEENS_3cpp9enable_ifIXaasr3cppE13is_integral_vIT_ElestS5_Lm8EENS3_6stringEE4typeES5_: %agg.result"}
!147 = distinct !{!147, !"_ZN22__llvm_libc_22_0_0_git7testing8internal13describeValueIbEENS_3cpp9enable_ifIXaasr3cppE13is_integral_vIT_ElestS5_Lm8EENS3_6stringEE4typeES5_"}
!148 = !{!146}
!149 = !{!150, !151, i64 8}
!150 = !{!"_ZTSN22__llvm_libc_22_0_0_git7testing11TestOptionsE", !56, i64 0, !151, i64 8, !151, i64 9}
!151 = !{!"bool", !46, i64 0}
!152 = !{i8 0, i8 2}
!153 = !{!154, !156}
!154 = distinct !{!154, !155, !"_ZN22__llvm_libc_22_0_0_git3cpp8internal13to_dec_stringIiEENS0_6stringET_: %agg.result"}
!155 = distinct !{!155, !"_ZN22__llvm_libc_22_0_0_git3cpp8internal13to_dec_stringIiEENS0_6stringET_"}
!156 = distinct !{!156, !157, !"_ZN22__llvm_libc_22_0_0_git3cpp9to_stringEi: %agg.result"}
!157 = distinct !{!157, !"_ZN22__llvm_libc_22_0_0_git3cpp9to_stringEi"}
!158 = !{!159, !159, i64 0}
!159 = !{!"vtable pointer", !47, i64 0}
!160 = !{!150, !56, i64 0}
!161 = !{!150, !151, i64 9}
!162 = !{!163, !165}
!163 = distinct !{!163, !164, !"_ZN22__llvm_libc_22_0_0_git3cpp8internal13to_dec_stringIyEENS0_6stringET_: %agg.result"}
!164 = distinct !{!164, !"_ZN22__llvm_libc_22_0_0_git3cpp8internal13to_dec_stringIyEENS0_6stringET_"}
!165 = distinct !{!165, !166, !"_ZN22__llvm_libc_22_0_0_git3cpp9to_stringEy: %agg.result"}
!166 = distinct !{!166, !"_ZN22__llvm_libc_22_0_0_git3cpp9to_stringEy"}
!167 = !{!168, !48, i64 24}
!168 = !{!"_ZTSN22__llvm_libc_22_0_0_git15IntegerToStringIyNS_7details3FmtILh10ELb0ELb0ELb0ELm1EEEEE", !169, i64 0, !48, i64 24}
!169 = !{!"_ZTSN22__llvm_libc_22_0_0_git3cpp5arrayIcLm21EEE", !46, i64 0}
!170 = !{!171, !173}
!171 = distinct !{!171, !172, !"_ZN22__llvm_libc_22_0_0_git3cpp8internal13to_dec_stringIyEENS0_6stringET_: %agg.result"}
!172 = distinct !{!172, !"_ZN22__llvm_libc_22_0_0_git3cpp8internal13to_dec_stringIyEENS0_6stringET_"}
!173 = distinct !{!173, !174, !"_ZN22__llvm_libc_22_0_0_git3cpp9to_stringEy: %agg.result"}
!174 = distinct !{!174, !"_ZN22__llvm_libc_22_0_0_git3cpp9to_stringEy"}
!175 = !{!176, !178}
!176 = distinct !{!176, !177, !"_ZN22__llvm_libc_22_0_0_git3cpp8internal13to_dec_stringIyEENS0_6stringET_: %agg.result"}
!177 = distinct !{!177, !"_ZN22__llvm_libc_22_0_0_git3cpp8internal13to_dec_stringIyEENS0_6stringET_"}
!178 = distinct !{!178, !179, !"_ZN22__llvm_libc_22_0_0_git3cpp9to_stringEy: %agg.result"}
!179 = distinct !{!179, !"_ZN22__llvm_libc_22_0_0_git3cpp9to_stringEy"}
!180 = !{!181, !183}
!181 = distinct !{!181, !182, !"_ZN22__llvm_libc_22_0_0_git3cpp8internal13to_dec_stringIiEENS0_6stringET_: %agg.result"}
!182 = distinct !{!182, !"_ZN22__llvm_libc_22_0_0_git3cpp8internal13to_dec_stringIiEENS0_6stringET_"}
!183 = distinct !{!183, !184, !"_ZN22__llvm_libc_22_0_0_git3cpp9to_stringEi: %agg.result"}
!184 = distinct !{!184, !"_ZN22__llvm_libc_22_0_0_git3cpp9to_stringEi"}
!185 = !{!186, !188}
!186 = distinct !{!186, !187, !"_ZN22__llvm_libc_22_0_0_git3cpp8internal13to_dec_stringIiEENS0_6stringET_: %agg.result"}
!187 = distinct !{!187, !"_ZN22__llvm_libc_22_0_0_git3cpp8internal13to_dec_stringIiEENS0_6stringET_"}
!188 = distinct !{!188, !189, !"_ZN22__llvm_libc_22_0_0_git3cpp9to_stringEi: %agg.result"}
!189 = distinct !{!189, !"_ZN22__llvm_libc_22_0_0_git3cpp9to_stringEi"}
!190 = !{!191, !193}
!191 = distinct !{!191, !192, !"_ZN22__llvm_libc_22_0_0_git3cpp8internal13to_dec_stringIiEENS0_6stringET_: %agg.result"}
!192 = distinct !{!192, !"_ZN22__llvm_libc_22_0_0_git3cpp8internal13to_dec_stringIiEENS0_6stringET_"}
!193 = distinct !{!193, !194, !"_ZN22__llvm_libc_22_0_0_git3cpp9to_stringEi: %agg.result"}
!194 = distinct !{!194, !"_ZN22__llvm_libc_22_0_0_git3cpp9to_stringEi"}
!195 = !{i64 0, !"_ZTSMN22__llvm_libc_22_0_0_git7testing10TestLoggerEFRS1_cE"}
!196 = !{i64 0, !"_ZTSMN3rpc6ClientEFNS_4PortILb0EEEvE"}
!197 = !{i64 0, !198}
!198 = distinct !{}
!199 = !{!200, !151, i64 29}
!200 = !{!"_ZTSN3rpc4PortILb0EEE", !201, i64 0, !79, i64 8, !71, i64 16, !71, i64 20, !71, i64 24, !151, i64 28, !151, i64 29}
!201 = !{!"p1 _ZTSN3rpc7ProcessILb0EEE", !50, i64 0}
!202 = !{!200, !71, i64 24}
!203 = !{!200, !201, i64 0}
!204 = !{i64 8}
!205 = !{!71, !71, i64 0}
!206 = !{!200, !71, i64 20}
!207 = !{!151, !151, i64 0}
!208 = !{i64 0, !209}
!209 = distinct !{}
!210 = !{i64 0, !"_ZTSMN3rpc4PortILb0EEEFvvE"}