Address crash in iree-reduce with complex tensors (#24674) See issue https://github.com/iree-org/iree/issues/24673. Details are repeated here. `iree-reduce` crashes in `reduceLinalgOnTensorsDelta` when it tries to replace a `linalg` op that produces a **complex**-typed tensor. The strategy builds a zero constant via `Builder::getZeroAttr(elementType)`, which returns a **null** attribute for `complex` (it only supports int/index/float). Creating an `arith.constant` from the null attribute aborts the process. The proposed fix simply does an early-out if the type is not int, float, or index instead of crashing. Repro: `interesting.sh` ```bash #!/bin/bash # Interestingness oracle for iree-reduce. # # A reduced candidate is "interesting" (exit 0) only while it still contains the # complex-typed linalg.generic (identified by the complex.create in its body). # This keeps the reducer from deleting the op via an earlier strategy, so it # reaches reduceLinalgOnTensorsDelta with the complex op still present. grep -q "complex.create" "$1" ``` `complex_repro.mlir` ``` // Minimal reproducer for the iree-reduce crash in reduceLinalgOnTensorsDelta. // // The linalg.generic writes a complex<f32> tensor. When the reducer decides to // replace this op with a linalg.fill of a zero constant, it calls // Builder::getZeroAttr(complex<f32>), which returns a null attribute. Building // an arith.constant from that null attribute crashes ("Failed to infer result // type(s)"). // // The op survives the earlier reduction strategies because the interestingness // check (interesting.sh) requires "complex.create" to remain present, so the // reducer reaches reduceLinalgOnTensorsDelta with the complex op still in place. func.func @main(%arg0: tensor<4xf32>) -> tensor<4xcomplex<f32>> { %empty = tensor.empty() : tensor<4xcomplex<f32>> %0 = linalg.generic { indexing_maps = [affine_map<(d0) -> (d0)>, affine_map<(d0) -> (d0)>], iterator_types = ["parallel"] } ins(%arg0 : tensor<4xf32>) outs(%empty : tensor<4xcomplex<f32>>) { ^bb0(%in: f32, %out: complex<f32>): %c = complex.create %in, %in : complex<f32> linalg.yield %c : complex<f32> } -> tensor<4xcomplex<f32>> return %0 : tensor<4xcomplex<f32>> } ``` Run using: ```bash chmod +x interesting.sh iree-reduce ./interesting.sh ./complex_repro.mlir -o reduced.mlir ``` Crashes with: ``` LLVM ERROR: Failed to infer result type(s): "arith.constant"(...) {} : () -> ( ??? ) Please report issues to https://github.com/iree-org/iree/issues and include the crash backtrace. Stack dump: 0. Program arguments: ../../fix-undef-global-hoist/iree-compiler/tools/iree-reduce ./interesting.sh ./complex_repro.mlir -o reduced.mlir #0 0x00007aa82ee7cf5b llvm::sys::PrintStackTrace(llvm::raw_ostream&, int) fix-undef-global-hoist/iree/third_party/llvm-project/llvm/lib/Support/Unix/Signals.inc:881:13 #1 0x00007aa82ee7ab83 llvm::sys::RunSignalHandlers() fix-undef-global-hoist/iree/third_party/llvm-project/llvm/lib/Support/Signals.cpp:109:18 #2 0x00007aa82ee7de71 SignalHandler(int, siginfo_t*, void*) fix-undef-global-hoist/iree/third_party/llvm-project/llvm/lib/Support/Unix/Signals.inc:448:38 #3 0x00007aa828e45330 (/lib/x86_64-linux-gnu/libc.so.6+0x45330) #4 0x00007aa828e9eb2c __pthread_kill_implementation ./nptl/pthread_kill.c:44:76 #5 0x00007aa828e9eb2c __pthread_kill_internal ./nptl/pthread_kill.c:78:10 #6 0x00007aa828e9eb2c pthread_kill ./nptl/pthread_kill.c:89:10 #7 0x00007aa828e4527e raise ./signal/../sysdeps/posix/raise.c:27:6 #8 0x00007aa828e288ff abort ./stdlib/abort.c:81:7 #9 0x00007aa82ee0850b llvm::report_fatal_error(llvm::Twine const&, bool) fix-undef-global-hoist/iree/third_party/llvm-project/llvm/lib/Support/ErrorHandling.cpp:137:5 #10 0x00007aa82ee0858e (fix-undef-global-hoist/iree-compiler/lib/libIREECompiler.so+0x580858e) #11 0x00007aa837dabc21 (fix-undef-global-hoist/iree-compiler/lib/libIREECompiler.so+0xe7abc21) #12 0x00007aa834c50cc1 llvm::SmallVectorTemplateCommon<mlir::Type, void>::begin() fix-undef-global-hoist/iree/third_party/llvm-project/llvm/include/llvm/ADT/SmallVector.h:276:45 #13 0x00007aa834c50cc1 llvm::SmallVectorTemplateCommon<mlir::Type, void>::end() fix-undef-global-hoist/iree/third_party/llvm-project/llvm/include/llvm/ADT/SmallVector.h:278:27 #14 0x00007aa834c50cc1 llvm::SmallVector<mlir::Type, 2u>::~SmallVector() fix-undef-global-hoist/iree/third_party/llvm-project/llvm/include/llvm/ADT/SmallVector.h:1242:46 #15 0x00007aa834c50cc1 mlir::arith::ConstantOp::build(mlir::OpBuilder&, mlir::OperationState&, mlir::TypedAttr) fix-undef-global-hoist/iree-compiler/llvm-project/tools/mlir/include/mlir/Dialect/Arith/IR/ArithOps.cpp.inc:4379:1 #16 0x00007aa834c131be mlir::arith::ConstantOp::create(mlir::OpBuilder&, mlir::Location, mlir::TypedAttr) fix-undef-global-hoist/iree-compiler/llvm-project/tools/mlir/include/mlir/Dialect/Arith/IR/ArithOps.cpp.inc:4384:55 #17 0x00007aa8337db6af mlir::Operation::getInlineOpResult(unsigned int) fix-undef-global-hoist/iree/third_party/llvm-project/mlir/include/mlir/IR/Operation.h:1035:61 #18 0x00007aa8337db6af mlir::Operation::getOpResultImpl(unsigned int) fix-undef-global-hoist/iree/third_party/llvm-project/mlir/include/mlir/IR/Operation.h:1045:14 #19 0x00007aa8337db6af mlir::Operation::getResult(unsigned int) fix-undef-global-hoist/iree/third_party/llvm-project/mlir/include/mlir/IR/Operation.h:432:54 #20 0x00007aa8337db6af mlir::OpTrait::OneTypedResult<mlir::Type>::Impl<mlir::arith::ConstantOp>::getResult() fix-undef-global-hoist/iree/third_party/llvm-project/mlir/include/mlir/IR/OpDefinition.h:708:33 #21 0x00007aa8337db6af mlir::OpTrait::OneTypedResult<mlir::Type>::Impl<mlir::arith::ConstantOp>::operator mlir::Value() fix-undef-global-hoist/iree/third_party/llvm-project/mlir/include/mlir/IR/OpDefinition.h:713:54 #22 0x00007aa8337db6af mlir::iree_compiler::Reducer::reduceLinalgOnTensorsDelta(mlir::iree_compiler::Reducer::ChunkManager&, mlir::iree_compiler::Reducer::WorkItem&) fix-undef-global-hoist/iree/compiler/src/iree/compiler/Reducer/Strategies/ReduceLinalgOnTensorsDelta.cpp:114:18 #23 0x00007aa834d464dd mlir::iree_compiler::Reducer::Delta::checkChunk(mlir::iree_compiler::Reducer::Chunk, llvm::function_ref<void (mlir::iree_compiler::Reducer::ChunkManager&, mlir::iree_compiler::Reducer::WorkItem&)>, llvm::ArrayRef<mlir::iree_compiler::Reducer::Chunk>, llvm::DenseSet<mlir::iree_compiler::Reducer::Chunk, llvm::DenseMapInfo<mlir::iree_compiler::Reducer::Chunk, void>>&) fix-undef-global-hoist/iree/compiler/src/iree/compiler/Reducer/Framework/Delta.cpp:39:8 #24 0x00007aa834d46aa7 llvm::LogicalResult::failed() const fix-undef-global-hoist/iree/third_party/llvm-project/llvm/include/llvm/Support/LogicalResult.h:43:43 #25 0x00007aa834d46aa7 llvm::failed(llvm::LogicalResult) fix-undef-global-hoist/iree/third_party/llvm-project/llvm/include/llvm/Support/LogicalResult.h:71:58 #26 0x00007aa834d46aa7 mlir::iree_compiler::Reducer::Delta::runDeltaPass(llvm::function_ref<void (mlir::iree_compiler::Reducer::ChunkManager&, mlir::iree_compiler::Reducer::WorkItem&)>, llvm::StringRef) fix-undef-global-hoist/iree/compiler/src/iree/compiler/Reducer/Framework/Delta.cpp:105:11 #27 0x00007aa82f5d68ea mlir::iree_compiler::Reducer::ireeRunReducingStrategies(mlir::OwningOpRef<mlir::Operation*>, mlir::iree_compiler::Reducer::ReducerConfig&) fix-undef-global-hoist/iree/compiler/src/iree/compiler/Reducer/iree_reduce_lib.cc:39:9 #28 0x00007aa82edba149 ireeReduceMainFromCL(int, char**, mlir::MLIRContext&) fix-undef-global-hoist/iree/compiler/src/iree/compiler/API/Internal/IREEReduceToolEntryPoint.cpp:102:26 #29 0x00007aa82edba149 ireeReduceRunMain fix-undef-global-hoist/iree/compiler/src/iree/compiler/API/Internal/IREEReduceToolEntryPoint.cpp:134:7 #30 0x00007aa828e2a1ca __libc_start_call_main ./csu/../sysdeps/nptl/libc_start_call_main.h:74:3 #31 0x00007aa828e2a28b call_init ./csu/../csu/libc-start.c:128:20 #32 0x00007aa828e2a28b __libc_start_main ./csu/../csu/libc-start.c:347:5 #33 0x00005a84cb2e26e5 _start (../../fix-undef-global-hoist/iree-compiler/tools/iree-reduce+0x16e5) ``` Signed-off-by: Paul Stark <paul.stark@cdprojektred.com> Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
IREE (Intermediate Representation Execution Environment, pronounced as “eerie”) is an MLIR-based end-to-end compiler and runtime that lowers Machine Learning (ML) models to a unified IR that scales up to meet the needs of the datacenter and down to satisfy the constraints and special considerations of mobile and edge deployments.
See our website for project details, user guides, and instructions on building from source.
Releases notes are published on GitHub releases.
| Package | Release status |
|---|---|
| GitHub release (stable) | |
| GitHub release (nightly) | |
iree-base-compiler | |
iree-base-runtime |
For more details on the release process, see https://iree.dev/developers/general/release-management/.
| Operating system | Build status |
|---|---|
| Linux | |
| macOS | |
| macOS |
For the full list of workflows see https://iree.dev/developers/general/github-actions/.
See our website for more information.
Community meeting recordings: IREE YouTube channel
| Date | Title | Recording | Slides |
|---|---|---|---|
| 2025-06-10 | Data-Tiling in IREE: Achieving High Performance Through Compiler Design (AsiaLLVM) | recording | slides |
| 2025-05-17 | Introduction to GPU architecture and IREE's GPU CodeGen Pipeline | recording | slides |
| 2025-02-12 | The Long Tail of AI: SPIR-V in IREE and MLIR (Vulkanised) | recording | slides |
| 2024-10-01 | Unveiling the Inner Workings of IREE: An MLIR-Based Compiler for Diverse Hardware | recording | |
| 2021-06-09 | IREE Runtime Design Tech Talk | recording | slides |
| 2020-08-20 | IREE CodeGen (MLIR Open Design Meeting) | recording | slides |
| 2020-03-18 | Interactive HAL IR Walkthrough | recording | |
| 2020-01-31 | End-to-end MLIR Workflow in IREE (MLIR Open Design Meeting) | recording | slides |
IREE is licensed under the terms of the Apache 2.0 License with LLVM Exceptions. See LICENSE for more information.