[llvm] [AMDGPU] Enable i8 GEP promotion for vector allocas (PR #166132)

Sun Dec 7 20:24:14 PST 2025

llvm-ci wrote:

LLVM Buildbot has detected a new failure on builder `openmp-offload-amdgpu-runtime-2` running on `rocm-worker-hw-02` while building `llvm` at step 10 "Add check check-libc-amdgcn-amd-amdhsa".

Full details are available at: https://lab.llvm.org/buildbot/#/builders/10/builds/18635

<details>
<summary>Here is the relevant piece of the build log for the reference</summary>

```
Step 10 (Add check check-libc-amdgcn-amd-amdhsa) failure: test (failure)
...
[2611/3298] Running hermetic test libc.test.src.math.smoke.lroundf_test.__hermetic__
[==========] Running 3 tests from 1 test suite.
[ RUN      ] LlvmLibcRoundToIntegerTest.InfinityAndNaN
[       OK ] LlvmLibcRoundToIntegerTest.InfinityAndNaN (10 us)
[ RUN      ] LlvmLibcRoundToIntegerTest.RoundNumbers
[       OK ] LlvmLibcRoundToIntegerTest.RoundNumbers (6 us)
[ RUN      ] LlvmLibcRoundToIntegerTest.SubnormalRange
[       OK ] LlvmLibcRoundToIntegerTest.SubnormalRange (789 us)
Ran 3 tests.  PASS: 3  FAIL: 0
[2612/3298] Linking CXX executable libc/test/src/stdlib/libc.test.src.stdlib.heap_sort_test.__hermetic__.__build__
FAILED: libc/test/src/stdlib/libc.test.src.stdlib.heap_sort_test.__hermetic__.__build__ 
: && /home/botworker/builds/openmp-offload-amdgpu-runtime-2/llvm.build/./bin/clang++ --target=amdgcn-amd-amdhsa -fPIC -fno-semantic-interposition -fvisibility-inlines-hidden -Werror=date-time -Werror=unguarded-availability-new -Wall -Wextra -Wno-unused-parameter -Wwrite-strings -Wcast-qual -Wmissing-field-initializers -Wimplicit-fallthrough -Wcovered-switch-default -Wno-noexcept-type -Wnon-virtual-dtor -Wdelete-non-virtual-dtor -Wsuggest-override -Wstring-conversion -Wno-pass-failed -Wmisleading-indentation -Wctad-maybe-unsupported -fdiagnostics-color -ffunction-sections -fdata-sections -O3 -DNDEBUG -Wl,--color-diagnostics    --target=amdgcn-amd-amdhsa -Wno-multi-gpu -mcpu=native -flto -Wl,-mllvm,-amdgpu-lower-global-ctor-dtor=0 -nostdlib -static -Wl,-mllvm,-amdhsa-code-object-version=6 libc/startup/gpu/amdgpu/CMakeFiles/libc.startup.gpu.amdgpu.crt1.dir/start.cpp.o libc/test/src/stdlib/CMakeFiles/libc.test.src.stdlib.heap_sort_test.__hermetic__.__build__.dir/heap_sort_test.cpp.o -o libc/test/src/stdlib/libc.test.src.stdlib.heap_sort_test.__hermetic__.__build__  libc/test/UnitTest/libLibcTest.hermetic.a  libc/test/UnitTest/libLibcHermeticTestSupport.hermetic.a  libc/test/src/stdlib/liblibc.test.src.stdlib.heap_sort_test.__hermetic__.libc.a && :
PLEASE submit a bug report to https://github.com/llvm/llvm-project/issues/ and include the crash backtrace and instructions to reproduce the bug.
Stack dump:
0.	Program arguments: /home/botworker/builds/openmp-offload-amdgpu-runtime-2/llvm.build/bin/ld.lld --no-undefined -shared -plugin-opt=mcpu=gfx90a -plugin-opt=O3 --lto-CGO3 -plugin-opt=-function-sections=1 -plugin-opt=-data-sections=1 -L/home/botworker/builds/openmp-offload-amdgpu-runtime-2/llvm.build/bin/../lib/amdgcn-amd-amdhsa --color-diagnostics -mllvm -amdgpu-lower-global-ctor-dtor=0 -mllvm -amdhsa-code-object-version=6 libc/startup/gpu/amdgpu/CMakeFiles/libc.startup.gpu.amdgpu.crt1.dir/start.cpp.o libc/test/src/stdlib/CMakeFiles/libc.test.src.stdlib.heap_sort_test.__hermetic__.__build__.dir/heap_sort_test.cpp.o libc/test/UnitTest/libLibcTest.hermetic.a libc/test/UnitTest/libLibcHermeticTestSupport.hermetic.a libc/test/src/stdlib/liblibc.test.src.stdlib.heap_sort_test.__hermetic__.libc.a -o libc/test/src/stdlib/libc.test.src.stdlib.heap_sort_test.__hermetic__.__build__
1.	Running pass 'Function Pass Manager' on module 'ld-temp.o'.
2.	Running pass 'AMDGPU Promote Alloca' on function '@_ZN49LlvmLibcHeapSortTest_SameElementThreeElementArray3RunEv'
 #0 0x00007bd51120e100 llvm::sys::PrintStackTrace(llvm::raw_ostream&, int) (/home/botworker/builds/openmp-offload-amdgpu-runtime-2/llvm.build/bin/../lib/libLLVMSupport.so.22.0git+0x20e100)
 #1 0x00007bd51120adbf llvm::sys::RunSignalHandlers() (/home/botworker/builds/openmp-offload-amdgpu-runtime-2/llvm.build/bin/../lib/libLLVMSupport.so.22.0git+0x20adbf)
 #2 0x00007bd51120af12 SignalHandler(int, siginfo_t*, void*) Signals.cpp:0:0
 #3 0x00007bd510842520 (/lib/x86_64-linux-gnu/libc.so.6+0x42520)
 #4 0x00007bd50c854df1 llvm::Instruction::eraseFromParent() (/home/botworker/builds/openmp-offload-amdgpu-runtime-2/llvm.build/bin/../lib/../lib/libLLVMCore.so.22.0git+0x254df1)
 #5 0x00007bd50ef80f7d (anonymous namespace)::AMDGPUPromoteAllocaImpl::tryPromoteAllocaToVector(llvm::AllocaInst&)::'lambda'(llvm::Instruction*, llvm::Twine)::operator()(llvm::Instruction*, llvm::Twine) const AMDGPUPromoteAlloca.cpp:0:0
 #6 0x00007bd50ef89288 (anonymous namespace)::AMDGPUPromoteAllocaImpl::tryPromoteAllocaToVector(llvm::AllocaInst&) AMDGPUPromoteAlloca.cpp:0:0
 #7 0x00007bd50ef8bb21 (anonymous namespace)::AMDGPUPromoteAllocaImpl::run(llvm::Function&, bool) (.part.0) AMDGPUPromoteAlloca.cpp:0:0
 #8 0x00007bd50ef8dc33 (anonymous namespace)::AMDGPUPromoteAlloca::runOnFunction(llvm::Function&) AMDGPUPromoteAlloca.cpp:0:0
 #9 0x00007bd50c8a4e93 llvm::FPPassManager::runOnFunction(llvm::Function&) (/home/botworker/builds/openmp-offload-amdgpu-runtime-2/llvm.build/bin/../lib/../lib/libLLVMCore.so.22.0git+0x2a4e93)
#10 0x00007bd50c8a52e9 llvm::FPPassManager::runOnModule(llvm::Module&) (/home/botworker/builds/openmp-offload-amdgpu-runtime-2/llvm.build/bin/../lib/../lib/libLLVMCore.so.22.0git+0x2a52e9)
#11 0x00007bd50c8a5c37 llvm::legacy::PassManagerImpl::run(llvm::Module&) (/home/botworker/builds/openmp-offload-amdgpu-runtime-2/llvm.build/bin/../lib/../lib/libLLVMCore.so.22.0git+0x2a5c37)
#12 0x00007bd510ebeb29 codegen(llvm::lto::Config const&, llvm::TargetMachine*, std::function<llvm::Expected<std::unique_ptr<llvm::CachedFileStream, std::default_delete<llvm::CachedFileStream>>> (unsigned int, llvm::Twine const&)>, unsigned int, llvm::Module&, llvm::ModuleSummaryIndex const&) LTOBackend.cpp:0:0
#13 0x00007bd510ec09ad llvm::lto::backend(llvm::lto::Config const&, std::function<llvm::Expected<std::unique_ptr<llvm::CachedFileStream, std::default_delete<llvm::CachedFileStream>>> (unsigned int, llvm::Twine const&)>, unsigned int, llvm::Module&, llvm::ModuleSummaryIndex&) (/home/botworker/builds/openmp-offload-amdgpu-runtime-2/llvm.build/bin/../lib/../lib/libLLVMLTO.so.22.0git+0x569ad)
#14 0x00007bd510ead619 llvm::lto::LTO::runRegularLTO(std::function<llvm::Expected<std::unique_ptr<llvm::CachedFileStream, std::default_delete<llvm::CachedFileStream>>> (unsigned int, llvm::Twine const&)>) (/home/botworker/builds/openmp-offload-amdgpu-runtime-2/llvm.build/bin/../lib/../lib/libLLVMLTO.so.22.0git+0x43619)
#15 0x00007bd510eb2df2 llvm::lto::LTO::run(std::function<llvm::Expected<std::unique_ptr<llvm::CachedFileStream, std::default_delete<llvm::CachedFileStream>>> (unsigned int, llvm::Twine const&)>, llvm::FileCache) (/home/botworker/builds/openmp-offload-amdgpu-runtime-2/llvm.build/bin/../lib/../lib/libLLVMLTO.so.22.0git+0x48df2)
#16 0x00007bd5117ad8fc lld::elf::BitcodeCompiler::compile() (/home/botworker/builds/openmp-offload-amdgpu-runtime-2/llvm.build/bin/../lib/liblldELF.so.22.0git+0x1ad8fc)
#17 0x00007bd5116fca0d void lld::elf::LinkerDriver::compileBitcodeFiles<llvm::object::ELFType<(llvm::endianness)1, true>>(bool) (/home/botworker/builds/openmp-offload-amdgpu-runtime-2/llvm.build/bin/../lib/liblldELF.so.22.0git+0xfca0d)
#18 0x00007bd511720e24 void lld::elf::LinkerDriver::link<llvm::object::ELFType<(llvm::endianness)1, true>>(llvm::opt::InputArgList&) (/home/botworker/builds/openmp-offload-amdgpu-runtime-2/llvm.build/bin/../lib/liblldELF.so.22.0git+0x120e24)
#19 0x00007bd511728ca9 lld::elf::LinkerDriver::linkerMain(llvm::ArrayRef<char const*>) (/home/botworker/builds/openmp-offload-amdgpu-runtime-2/llvm.build/bin/../lib/liblldELF.so.22.0git+0x128ca9)
#20 0x00007bd5117292cc lld::elf::link(llvm::ArrayRef<char const*>, llvm::raw_ostream&, llvm::raw_ostream&, bool, bool) (/home/botworker/builds/openmp-offload-amdgpu-runtime-2/llvm.build/bin/../lib/liblldELF.so.22.0git+0x1292cc)
#21 0x00007bd5119e35e1 lld::unsafeLldMain(llvm::ArrayRef<char const*>, llvm::raw_ostream&, llvm::raw_ostream&, llvm::ArrayRef<lld::DriverDef>, bool) (/home/botworker/builds/openmp-offload-amdgpu-runtime-2/llvm.build/bin/../lib/liblldCommon.so.22.0git+0xe5e1)
#22 0x00005e1ec1509d16 lld_main(int, char**, llvm::ToolContext const&) (/home/botworker/builds/openmp-offload-amdgpu-runtime-2/llvm.build/bin/ld.lld+0x2d16)
#23 0x00005e1ec150955b main (/home/botworker/builds/openmp-offload-amdgpu-runtime-2/llvm.build/bin/ld.lld+0x255b)
#24 0x00007bd510829d90 __libc_start_call_main ./csu/../sysdeps/nptl/libc_start_call_main.h:58:16
#25 0x00007bd510829e40 call_init ./csu/../csu/libc-start.c:128:20
#26 0x00007bd510829e40 __libc_start_main ./csu/../csu/libc-start.c:379:5
#27 0x00005e1ec15095b5 _start (/home/botworker/builds/openmp-offload-amdgpu-runtime-2/llvm.build/bin/ld.lld+0x25b5)
clang++: error: unable to execute command: Segmentation fault (core dumped)
clang++: error: ld.lld command failed due to signal (use -v to see invocation)
[2613/3298] Linking CXX executable libc/test/src/math/smoke/libc.test.src.math.smoke.sin_test.__hermetic__.__build__
[2614/3298] Running hermetic test libc.test.src.math.smoke.fminimumf_test.__hermetic__

```

</details>

https://github.com/llvm/llvm-project/pull/166132