Calling LLVM intrinsics with packed result types

farnoy · June 22, 2025, 6:23pm

How can I call this intrinsic?

declare {i32, i1} @llvm.nvvm.vote.sync(i32 %membermask, i32 %mode, i1 %predicate)

I’ve tried the following, but it crashes the compiler

var result = llvm_intrinsic[
    "llvm.nvvm.vote.sync",
    _RegisterPackType[Int32, Bool]
](Int32(_FULL_MASK), Int32(3), value)
return result[0].cast[dtype]()

INFO: Analyzed target //max/kernels/test/gpu/basics:test_prefix_sum.mojo.test (0 packages loaded, 0 targets configured).
ERROR: /modular/max/kernels/test/gpu/basics/BUILD.bazel:45:14: //max/kernels/test/gpu/basics:test_prefix_sum.mojo.test compiling mojo object failed: (Aborted): mojo failed: error executing MojoCompile command (from target //max/kernels/test/gpu/basics:test_prefix_sum.mojo.test) external/rules_mojo++mojo+mojo_toolchain_linux_x86_64/bin/mojo build '-strip-file-prefix=.' --emit object -o bazel-out/k8-fastbuild/bin/max/kernels/test/gpu/basics/test_prefix_sum.mojo.test.lo ... (remaining 24 arguments skipped)

Use --sandbox_debug to see verbose messages from the sandbox and retain the sandbox build root for debugging
double free or corruption (fasttop)
Please submit a bug report to https://github.com/modular/modular/issues and include the crash backtrace along with all the relevant source codes.
 #0 0x0000557716bdda5b llvm::sys::PrintStackTrace(llvm::raw_ostream&, int) Signals.cpp:0:0
 #1 0x0000557716bdb82b llvm::sys::RunSignalHandlers() Signals.cpp:0:0
 #2 0x0000557716bde10a SignalHandler(int, siginfo_t*, void*) Signals.cpp:0:0
 #3 0x00007fcd96e45330 (/lib/x86_64-linux-gnu/libc.so.6+0x45330)
 #4 0x00007fcd96e9eb2c pthread_kill (/lib/x86_64-linux-gnu/libc.so.6+0x9eb2c)
 #5 0x00007fcd96e4527e raise (/lib/x86_64-linux-gnu/libc.so.6+0x4527e)
 #6 0x00007fcd96e288ff abort (/lib/x86_64-linux-gnu/libc.so.6+0x288ff)
 #7 0x00007fcd96e297b6 (/lib/x86_64-linux-gnu/libc.so.6+0x297b6)
 #8 0x00007fcd96ea8ff5 (/lib/x86_64-linux-gnu/libc.so.6+0xa8ff5)
 #9 0x00007fcd96eab32a (/lib/x86_64-linux-gnu/libc.so.6+0xab32a)
#10 0x00007fcd96eaddae cfree (/lib/x86_64-linux-gnu/libc.so.6+0xaddae)
#11 0x00005577136e3b4c compileOffloads(mlir::ModuleOp, llvm::MapVector<M::TargetInfoAttr, M::KGEN::OffloadInfo, llvm::DenseMap<M::TargetInfoAttr, unsigned int, llvm::DenseMapInfo<M::TargetInfoAttr, void>, llvm::detail::DenseMapPair<M::TargetInfoAttr, unsigned int>>, llvm::SmallVector<std::pair<M::TargetInfoAttr, M::KGEN::OffloadInfo>, 0u>>&, mlir::SymbolTable const&, M::KGEN::CompilationOptions, M::KGEN::ElaborateGeneratorsOptions) KGENCompiler.cpp:0:0
#12 0x000055771376969a M::KGEN::Elaborator::run(mlir::ModuleOp, llvm::ArrayRef<std::pair<M::KGEN::GeneratorOp, M::KGEN::ParameterExprArrayAttr>>) Elaborator.cpp:0:0
#13 0x000055771377752b (anonymous namespace)::ElaborateGeneratorsPass::runOnOperation() Elaborator.cpp:0:0
#14 0x00005577143de24b mlir::detail::OpToOpPassAdaptor::run(mlir::Pass*, mlir::Operation*, mlir::AnalysisManager, bool, unsigned int) Pass.cpp:0:0
#15 0x00005577143e0e80 mlir::PassManager::run(mlir::Operation*) Pass.cpp:0:0
#16 0x0000557713c1041d void llvm::detail::UniqueFunctionBase<void, M::AsyncRT::AnyAsyncValueRef&&>::CallImpl<M::Cache::cachedTransform(mlir::Operation*, M::RCRef<M::Cache::BlobCache<M::Cache::TransformCacheKey>>, M::AsyncRT::AnyAsyncValueRef, mlir::PassManager&, std::function<void (mlir::Operation*)> const&, std::function<void (mlir::Operation*)> const&)::$_0::operator()(mlir::Operation*, M::RCRef<M::WriteableBuffer>, M::AsyncRT::AnyAsyncValueRef) const::'lambda'(M::AsyncRT::AnyAsyncValueRef&&)>(void*, M::AsyncRT::AnyAsyncValueRef&) CachedTransform.cpp:0:0
#17 0x00005577136a3b97 void M::AsyncRT::AnyAsyncValueRef::andThen<false>(llvm::unique_function<void (M::AsyncRT::AnyAsyncValueRef&&)>&&) && ObjectCompiler.cpp:0:0
#18 0x0000557713c10017 M::AsyncRT::AnyAsyncValueRef llvm::detail::UniqueFunctionBase<M::AsyncRT::AnyAsyncValueRef, M::RCRef<M::WriteableBuffer>, M::AsyncRT::AnyAsyncValueRef>::CallImpl<M::AsyncRT::AnyAsyncValueRef M::Cache::cachedTransform<M::Cache::cachedTransform(mlir::Operation*, M::RCRef<M::Cache::BlobCache<M::Cache::TransformCacheKey>>, M::AsyncRT::AnyAsyncValueRef, mlir::PassManager&, std::function<void (mlir::Operation*)> const&, std::function<void (mlir::Operation*)> const&)::$_0, M::Cache::cachedTransform(mlir::Operation*, M::RCRef<M::Cache::BlobCache<M::Cache::TransformCacheKey>>, M::AsyncRT::AnyAsyncValueRef, mlir::PassManager&, std::function<void (mlir::Operation*)> const&, std::function<void (mlir::Operation*)> const&)::$_1>(mlir::Operation*, M::RCRef<M::Cache::BlobCache<M::Cache::TransformCacheKey>>, M::AsyncRT::AnyAsyncValueRef, M::RCRef<M::WriteableBuffer>, M::Cache::cachedTransform(mlir::Operation*, M::RCRef<M::Cache::BlobCache<M::Cache::TransformCacheKey>>, M::AsyncRT::AnyAsyncValueRef, mlir::PassManager&, std::function<void (mlir::Operation*)> const&, std::function<void (mlir::Operation*)> const&)::$_0&&, M::Cache::cachedTransform(mlir::Operation*, M::RCRef<M::Cache::BlobCache<M::Cache::TransformCacheKey>>, M::AsyncRT::AnyAsyncValueRef, mlir::PassManager&, std::function<void (mlir::Operation*)> const&, std::function<void (mlir::Operation*)> const&)::$_1&&, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char>>*)::'lambda'(M::RCRef<M::WriteableBuffer>, M::AsyncRT::AnyAsyncValueRef)>(void*, M::RCRef<M::WriteableBuffer>&, M::AsyncRT::AnyAsyncValueRef&) CachedTransform.cpp:0:0
#19 0x0000557713c0fb18 void llvm::detail::UniqueFunctionBase<void, M::AsyncRT::AsyncValueRef<std::optional<M::RCRef<M::Buffer>>>&&>::CallImpl<M::AsyncRT::AnyAsyncValueRef M::Cache::cachedTransform<M::AsyncRT::AnyAsyncValueRef M::Cache::cachedTransform<M::Cache::cachedTransform(mlir::Operation*, M::RCRef<M::Cache::BlobCache<M::Cache::TransformCacheKey>>, M::AsyncRT::AnyAsyncValueRef, mlir::PassManager&, std::function<void (mlir::Operation*)> const&, std::function<void (mlir::Operation*)> const&)::$_0, M::Cache::cachedTransform(mlir::Operation*, M::RCRef<M::Cache::BlobCache<M::Cache::TransformCacheKey>>, M::AsyncRT::AnyAsyncValueRef, mlir::PassManager&, std::function<void (mlir::Operation*)> const&, std::function<void (mlir::Operation*)> const&)::$_1>(mlir::Operation*, M::RCRef<M::Cache::BlobCache<M::Cache::TransformCacheKey>>, M::AsyncRT::AnyAsyncValueRef, M::RCRef<M::WriteableBuffer>, M::Cache::cachedTransform(mlir::Operation*, M::RCRef<M::Cache::BlobCache<M::Cache::TransformCacheKey>>, M::AsyncRT::AnyAsyncValueRef, mlir::PassManager&, std::function<void (mlir::Operation*)> const&, std::function<void (mlir::Operation*)> const&)::$_0&&, M::Cache::cachedTransform(mlir::Operation*, M::RCRef<M::Cache::BlobCache<M::Cache::TransformCacheKey>>, M::AsyncRT::AnyAsyncValueRef, mlir::PassManager&, std::function<void (mlir::Operation*)> const&, std::function<void (mlir::Operation*)> const&)::$_1&&, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char>>*)::'lambda'(M::RCRef<M::WriteableBuffer>, M::AsyncRT::AnyAsyncValueRef)>(M::AsyncRT::EncodedLocation, M::RCRef<M::Cache::BlobCache<M::Cache::TransformCacheKey>> const&, M::AsyncRT::AnyAsyncValueRef, M::RCRef<M::WriteableBuffer>, M::Cache::cachedTransform(mlir::Operation*, M::RCRef<M::Cache::BlobCache<M::Cache::TransformCacheKey>>, M::AsyncRT::AnyAsyncValueRef, mlir::PassManager&, std::function<void (mlir::Operation*)> const&, std::function<void (mlir::Operation*)> const&)::$_0, llvm::unique_function<M::AsyncRT::AnyAsyncValueRef (M::RCRef<M::Buffer>)>, bool, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char>>*)::'lambda'(M::AsyncRT::AsyncValueRef<std::optional<M::RCRef<M::Buffer>>>&&)>(void*, M::AsyncRT::AsyncValueRef<std::optional<M::RCRef<M::Buffer>>>&) CachedTransform.cpp:0:0
#20 0x0000557713c1889c void (anonymous namespace)::WorkQueueThread::runItemsImpl<(anonymous namespace)::WorkQueueThread::runOnThread()::$_0, (anonymous namespace)::WorkQueueThread::runOnThread()::$_1>((anonymous namespace)::WorkQueueThread::runOnThread()::$_0, (anonymous namespace)::WorkQueueThread::runOnThread()::$_1, bool, llvm::StringLiteral, llvm::StringLiteral) ThreadPoolWorkQueue.cpp:0:0
#21 0x0000557713c18707 (anonymous namespace)::WorkQueueThread::runOnThread() ThreadPoolWorkQueue.cpp:0:0
#22 0x00007fcd9a0ecdb4 (/lib/x86_64-linux-gnu/libstdc++.so.6+0xecdb4)
#23 0x00007fcd96e9caa4 (/lib/x86_64-linux-gnu/libc.so.6+0x9caa4)
#24 0x00007fcd96f29c3c (/lib/x86_64-linux-gnu/libc.so.6+0x129c3c)
INFO: Found 1 test target...
Target //max/kernels/test/gpu/basics:test_prefix_sum.mojo.test failed to build
Use --verbose_failures to see the command lines of failed build steps.
INFO: Elapsed time: 31.331s, Critical Path: 31.28s
INFO: 12 processes: 2 internal, 10 processwrapper-sandbox.
ERROR: Build did NOT complete successfully
//max/kernels/test/gpu/basics:test_prefix_sum.mojo.test         FAILED TO BUILD

Executed 0 out of 1 test: 1 fails to build.

Repro code:

For some reason it’s unable to find this intrinsic on Godbolt, but it crashes the compiler with the above trace when compiled locally.

clattner · June 28, 2025, 6:46pm

Wow, that’s a terrible error message. The signature looks right to me, assuming value is a Bool.

@weiwei.chen can you take a look at this when you get a chance? Both to see if there is a simple fix, but also to see if we can make the elaborator not crash on bad input?

clattner · June 28, 2025, 6:48pm

Actually, compiler explorer is generating a reasonable error message:

open-source/max/mojo/stdlib/stdlib/sys/intrinsics.mojo:101:14: error: could not find LLVM intrinsic: "llvm.nvvm.vote.sync" : !kgen.string
<source>:14:10: note: called from
        ](Int32(_FULL_MASK), Int32(3), t)
         ^
open-source/max/mojo/stdlib/stdlib/sys/intrinsics.mojo:101:14: error: LLVM Translation failed for operation: llvm.call_intrinsic
<source>:14:10: note: called from
        ](Int32(_FULL_MASK), Int32(3), t)
         ^
open-source/max/mojo/stdlib/stdlib/sys/intrinsics.mojo:101:14: note: see current operation: %3 = "llvm.call_intrinsic"(%2, %1, %0) <{fastmathFlags = #llvm.fastmath<none>, intrin = "llvm.nvvm.vote.sync" : !kgen.string, op_bundle_sizes = array<i32>, operandSegmentSizes = array<i32: 3, 0>}> : (i32, i32, i1) -> !llvm.struct<(i32, i1)>

farnoy · June 28, 2025, 8:32pm

It complains about the missing intrinsic in Godbolt, but it does find it (I guess) locally when I’m compiling with an nvidia GPU. But I get that stacktrace double free error with no more details.

sora · June 28, 2025, 9:39pm

I was able to see the error locally. Could you open an issue on GH for this?

farnoy · June 29, 2025, 1:56pm

github.com/modular/modular

[BUG] Compiler crash using LLVM intrinsic returning heterogenous pack type

opened 01:55PM - 29 Jun 25 UTC

farnoy

bug mojo modular-repo

### Bug description Continued from https://forum.modular.com/t/calling-llvm-int…rinsics-with-packed-result-types/1708/5 Trying to use this intrinsic ```llvm declare {i32, i1} @llvm.nvvm.vote.sync(i32 %membermask, i32 %mode, i1 %predicate) ``` with ```mojo var result = llvm_intrinsic[ "llvm.nvvm.vote.sync", _RegisterPackType[Int32, Bool] ](Int32(_FULL_MASK), Int32(3), value) return result[0].cast[dtype]() ``` ### Actual behavior Crashes with: ``` INFO: Analyzed target //max/kernels/test/gpu/basics:test_prefix_sum.mojo.test (0 packages loaded, 0 targets configured). ERROR: /modular/max/kernels/test/gpu/basics/BUILD.bazel:45:14: //max/kernels/test/gpu/basics:test_prefix_sum.mojo.test compiling mojo object failed: (Aborted): mojo failed: error executing MojoCompile command (from target //max/kernels/test/gpu/basics:test_prefix_sum.mojo.test) external/rules_mojo++mojo+mojo_toolchain_linux_x86_64/bin/mojo build '-strip-file-prefix=.' --emit object -o bazel-out/k8-fastbuild/bin/max/kernels/test/gpu/basics/test_prefix_sum.mojo.test.lo ... (remaining 24 arguments skipped) Use --sandbox_debug to see verbose messages from the sandbox and retain the sandbox build root for debugging double free or corruption (fasttop) Please submit a bug report to https://github.com/modular/modular/issues and include the crash backtrace along with all the relevant source codes. #0 0x0000557716bdda5b llvm::sys::PrintStackTrace(llvm::raw_ostream&, int) Signals.cpp:0:0 #1 0x0000557716bdb82b llvm::sys::RunSignalHandlers() Signals.cpp:0:0 #2 0x0000557716bde10a SignalHandler(int, siginfo_t*, void*) Signals.cpp:0:0 #3 0x00007fcd96e45330 (/lib/x86_64-linux-gnu/libc.so.6+0x45330) #4 0x00007fcd96e9eb2c pthread_kill (/lib/x86_64-linux-gnu/libc.so.6+0x9eb2c) #5 0x00007fcd96e4527e raise (/lib/x86_64-linux-gnu/libc.so.6+0x4527e) #6 0x00007fcd96e288ff abort (/lib/x86_64-linux-gnu/libc.so.6+0x288ff) #7 0x00007fcd96e297b6 (/lib/x86_64-linux-gnu/libc.so.6+0x297b6) #8 0x00007fcd96ea8ff5 (/lib/x86_64-linux-gnu/libc.so.6+0xa8ff5) #9 0x00007fcd96eab32a (/lib/x86_64-linux-gnu/libc.so.6+0xab32a) #10 0x00007fcd96eaddae cfree (/lib/x86_64-linux-gnu/libc.so.6+0xaddae) #11 0x00005577136e3b4c compileOffloads(mlir::ModuleOp, llvm::MapVector<M::TargetInfoAttr, M::KGEN::OffloadInfo, llvm::DenseMap<M::TargetInfoAttr, unsigned int, llvm::DenseMapInfo<M::TargetInfoAttr, void>, llvm::detail::DenseMapPair<M::TargetInfoAttr, unsigned int>>, llvm::SmallVector<std::pair<M::TargetInfoAttr, M::KGEN::OffloadInfo>, 0u>>&, mlir::SymbolTable const&, M::KGEN::CompilationOptions, M::KGEN::ElaborateGeneratorsOptions) KGENCompiler.cpp:0:0 #12 0x000055771376969a M::KGEN::Elaborator::run(mlir::ModuleOp, llvm::ArrayRef<std::pair<M::KGEN::GeneratorOp, M::KGEN::ParameterExprArrayAttr>>) Elaborator.cpp:0:0 #13 0x000055771377752b (anonymous namespace)::ElaborateGeneratorsPass::runOnOperation() Elaborator.cpp:0:0 #14 0x00005577143de24b mlir::detail::OpToOpPassAdaptor::run(mlir::Pass*, mlir::Operation*, mlir::AnalysisManager, bool, unsigned int) Pass.cpp:0:0 #15 0x00005577143e0e80 mlir::PassManager::run(mlir::Operation*) Pass.cpp:0:0 #16 0x0000557713c1041d void llvm::detail::UniqueFunctionBase<void, M::AsyncRT::AnyAsyncValueRef&&>::CallImpl<M::Cache::cachedTransform(mlir::Operation*, M::RCRef<M::Cache::BlobCache<M::Cache::TransformCacheKey>>, M::AsyncRT::AnyAsyncValueRef, mlir::PassManager&, std::function<void (mlir::Operation*)> const&, std::function<void (mlir::Operation*)> const&)::$_0::operator()(mlir::Operation*, M::RCRef<M::WriteableBuffer>, M::AsyncRT::AnyAsyncValueRef) const::'lambda'(M::AsyncRT::AnyAsyncValueRef&&)>(void*, M::AsyncRT::AnyAsyncValueRef&) CachedTransform.cpp:0:0 #17 0x00005577136a3b97 void M::AsyncRT::AnyAsyncValueRef::andThen<false>(llvm::unique_function<void (M::AsyncRT::AnyAsyncValueRef&&)>&&) && ObjectCompiler.cpp:0:0 #18 0x0000557713c10017 M::AsyncRT::AnyAsyncValueRef llvm::detail::UniqueFunctionBase<M::AsyncRT::AnyAsyncValueRef, M::RCRef<M::WriteableBuffer>, M::AsyncRT::AnyAsyncValueRef>::CallImpl<M::AsyncRT::AnyAsyncValueRef M::Cache::cachedTransform<M::Cache::cachedTransform(mlir::Operation*, M::RCRef<M::Cache::BlobCache<M::Cache::TransformCacheKey>>, M::AsyncRT::AnyAsyncValueRef, mlir::PassManager&, std::function<void (mlir::Operation*)> const&, std::function<void (mlir::Operation*)> const&)::$_0, M::Cache::cachedTransform(mlir::Operation*, M::RCRef<M::Cache::BlobCache<M::Cache::TransformCacheKey>>, M::AsyncRT::AnyAsyncValueRef, mlir::PassManager&, std::function<void (mlir::Operation*)> const&, std::function<void (mlir::Operation*)> const&)::$_1>(mlir::Operation*, M::RCRef<M::Cache::BlobCache<M::Cache::TransformCacheKey>>, M::AsyncRT::AnyAsyncValueRef, M::RCRef<M::WriteableBuffer>, M::Cache::cachedTransform(mlir::Operation*, M::RCRef<M::Cache::BlobCache<M::Cache::TransformCacheKey>>, M::AsyncRT::AnyAsyncValueRef, mlir::PassManager&, std::function<void (mlir::Operation*)> const&, std::function<void (mlir::Operation*)> const&)::$_0&&, M::Cache::cachedTransform(mlir::Operation*, M::RCRef<M::Cache::BlobCache<M::Cache::TransformCacheKey>>, M::AsyncRT::AnyAsyncValueRef, mlir::PassManager&, std::function<void (mlir::Operation*)> const&, std::function<void (mlir::Operation*)> const&)::$_1&&, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char>>*)::'lambda'(M::RCRef<M::WriteableBuffer>, M::AsyncRT::AnyAsyncValueRef)>(void*, M::RCRef<M::WriteableBuffer>&, M::AsyncRT::AnyAsyncValueRef&) CachedTransform.cpp:0:0 #19 0x0000557713c0fb18 void llvm::detail::UniqueFunctionBase<void, M::AsyncRT::AsyncValueRef<std::optional<M::RCRef<M::Buffer>>>&&>::CallImpl<M::AsyncRT::AnyAsyncValueRef M::Cache::cachedTransform<M::AsyncRT::AnyAsyncValueRef M::Cache::cachedTransform<M::Cache::cachedTransform(mlir::Operation*, M::RCRef<M::Cache::BlobCache<M::Cache::TransformCacheKey>>, M::AsyncRT::AnyAsyncValueRef, mlir::PassManager&, std::function<void (mlir::Operation*)> const&, std::function<void (mlir::Operation*)> const&)::$_0, M::Cache::cachedTransform(mlir::Operation*, M::RCRef<M::Cache::BlobCache<M::Cache::TransformCacheKey>>, M::AsyncRT::AnyAsyncValueRef, mlir::PassManager&, std::function<void (mlir::Operation*)> const&, std::function<void (mlir::Operation*)> const&)::$_1>(mlir::Operation*, M::RCRef<M::Cache::BlobCache<M::Cache::TransformCacheKey>>, M::AsyncRT::AnyAsyncValueRef, M::RCRef<M::WriteableBuffer>, M::Cache::cachedTransform(mlir::Operation*, M::RCRef<M::Cache::BlobCache<M::Cache::TransformCacheKey>>, M::AsyncRT::AnyAsyncValueRef, mlir::PassManager&, std::function<void (mlir::Operation*)> const&, std::function<void (mlir::Operation*)> const&)::$_0&&, M::Cache::cachedTransform(mlir::Operation*, M::RCRef<M::Cache::BlobCache<M::Cache::TransformCacheKey>>, M::AsyncRT::AnyAsyncValueRef, mlir::PassManager&, std::function<void (mlir::Operation*)> const&, std::function<void (mlir::Operation*)> const&)::$_1&&, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char>>*)::'lambda'(M::RCRef<M::WriteableBuffer>, M::AsyncRT::AnyAsyncValueRef)>(M::AsyncRT::EncodedLocation, M::RCRef<M::Cache::BlobCache<M::Cache::TransformCacheKey>> const&, M::AsyncRT::AnyAsyncValueRef, M::RCRef<M::WriteableBuffer>, M::Cache::cachedTransform(mlir::Operation*, M::RCRef<M::Cache::BlobCache<M::Cache::TransformCacheKey>>, M::AsyncRT::AnyAsyncValueRef, mlir::PassManager&, std::function<void (mlir::Operation*)> const&, std::function<void (mlir::Operation*)> const&)::$_0, llvm::unique_function<M::AsyncRT::AnyAsyncValueRef (M::RCRef<M::Buffer>)>, bool, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char>>*)::'lambda'(M::AsyncRT::AsyncValueRef<std::optional<M::RCRef<M::Buffer>>>&&)>(void*, M::AsyncRT::AsyncValueRef<std::optional<M::RCRef<M::Buffer>>>&) CachedTransform.cpp:0:0 #20 0x0000557713c1889c void (anonymous namespace)::WorkQueueThread::runItemsImpl<(anonymous namespace)::WorkQueueThread::runOnThread()::$_0, (anonymous namespace)::WorkQueueThread::runOnThread()::$_1>((anonymous namespace)::WorkQueueThread::runOnThread()::$_0, (anonymous namespace)::WorkQueueThread::runOnThread()::$_1, bool, llvm::StringLiteral, llvm::StringLiteral) ThreadPoolWorkQueue.cpp:0:0 #21 0x0000557713c18707 (anonymous namespace)::WorkQueueThread::runOnThread() ThreadPoolWorkQueue.cpp:0:0 #22 0x00007fcd9a0ecdb4 (/lib/x86_64-linux-gnu/libstdc++.so.6+0xecdb4) #23 0x00007fcd96e9caa4 (/lib/x86_64-linux-gnu/libc.so.6+0x9caa4) #24 0x00007fcd96f29c3c (/lib/x86_64-linux-gnu/libc.so.6+0x129c3c) INFO: Found 1 test target... Target //max/kernels/test/gpu/basics:test_prefix_sum.mojo.test failed to build Use --verbose_failures to see the command lines of failed build steps. INFO: Elapsed time: 31.331s, Critical Path: 31.28s INFO: 12 processes: 2 internal, 10 processwrapper-sandbox. ERROR: Build did NOT complete successfully //max/kernels/test/gpu/basics:test_prefix_sum.mojo.test FAILED TO BUILD Executed 0 out of 1 test: 1 fails to build. ``` ### Expected behavior To compile correctly ### Steps to reproduce https://mojo.godbo.lt/z/64Tvjfshe ```mojo var result = llvm_intrinsic[ "llvm.nvvm.vote.sync", _RegisterPackType[Int32, Bool] ](Int32(_FULL_MASK), Int32(3), value) return result[0].cast[dtype]() ``` ### System information - Mojo nightly

weiwei.chen · June 29, 2025, 7:40pm

sure, I’ll take a look at the GitHub issue. @farnoy thanks for file the issue!

Topic		Replies	Views
LLVM intrinsic problem Mojo debugging	7	190	April 4, 2025
Constrained Floating-Point LLVM Intrinsic Mojo discussion , llvm , 24_6	1	47	June 17, 2025
"No active MLIR context" with new `CustomOpLibrary` torch integration MAX	20	297	July 8, 2025
Mojo on Jetson Orin Mojo	3	102	June 16, 2025
`pack_bits` inside of vectorize fails constraint check Mojo 25_3	10	86	May 7, 2025

Calling LLVM intrinsics with packed result types

Related topics