[MLIR]`vectorizeAffineForOp` triggers assertion failure `it != strategy.reductionLoops.end()

[BealZephyr]

Test commit: c3440e2

Steps to reproduce:

mlir-opt test.mlir -affine-super-vectorizer-test -vectorize-affine-loop-nest

Test case:

func.func @main(%in: memref<256x512xf32>, %out: memref<256xf32>) {
 %cst = arith.constant 1.000000e+00 : f32
 %final_red = affine.for %j = 0 to 512 iter_args(%red_iter = %cst) -> (f32) {
   %add = arith.addf %red_iter, %red_iter : f32
   affine.yield %add : f32
 }
 return
}

Crash trace:

mlir-opt: /home/ty/llvm-project/mlir/lib/Dialect/Affine/Transforms/SuperVectorize.cpp:1307: mlir::Operation* vectorizeAffineForOp(mlir::AffineForOp, {anonymous}::VectorizationState&): Assertion `it != strategy.reductionLoops.end() && "Reduction descriptors not found when vectorizing a reduction loop"' failed.
PLEASE submit a bug report to https://github.com/llvm/llvm-project/issues/ and include the crash backtrace.
Stack dump:
0.	Program arguments: /home/ty/llvm-project/build/bin/mlir-opt test.mlir -affine-super-vectorizer-test -vectorize-affine-loop-nest
 #0 0x0000555e1883b0da llvm::sys::PrintStackTrace(llvm::raw_ostream&, int) (/home/ty/llvm-project/build/bin/mlir-opt+0x22b0da)
 #1 0x0000555e1883836c SignalHandler(int) Signals.cpp:0:0
 #2 0x00007feaf1af4980 __restore_rt (/lib/x86_64-linux-gnu/libpthread.so.0+0x12980)
 #3 0x00007feaf07a5e87 raise /build/glibc-CVJwZb/glibc-2.27/signal/../sysdeps/unix/sysv/linux/raise.c:51:0
 #4 0x00007feaf07a77f1 abort /build/glibc-CVJwZb/glibc-2.27/stdlib/abort.c:81:0
 #5 0x00007feaf07973fa __assert_fail_base /build/glibc-CVJwZb/glibc-2.27/assert/assert.c:89:0
 #6 0x00007feaf0797472 (/lib/x86_64-linux-gnu/libc.so.6+0x30472)
 #7 0x0000555e1891ba1a vectorizeAffineForOp(mlir::AffineForOp, (anonymous namespace)::VectorizationState&) SuperVectorize.cpp:0:0
 #8 0x0000555e1891c7bb mlir::WalkResult llvm::function_ref<mlir::WalkResult (mlir::Operation*)>::callback_fn<vectorizeLoopNest(std::vector<llvm::SmallVector<mlir::AffineForOp, 2u>, std::allocator<llvm::SmallVector<mlir::AffineForOp, 2u>>>&, mlir::VectorizationStrategy const&)::'lambda'(mlir::Operation*)>(long, mlir::Operation*) SuperVectorize.cpp:0:0
 #9 0x0000555e18914d7b mlir::WalkResult mlir::detail::walk<mlir::ForwardIterator>(mlir::Operation*, llvm::function_ref<mlir::WalkResult (mlir::Operation*)>, mlir::WalkOrder) (/home/ty/llvm-project/build/bin/mlir-opt+0x304d7b)
#10 0x0000555e189164f3 vectorizeLoopNest(std::vector<llvm::SmallVector<mlir::AffineForOp, 2u>, std::allocator<llvm::SmallVector<mlir::AffineForOp, 2u>>>&, mlir::VectorizationStrategy const&) SuperVectorize.cpp:0:0
#11 0x0000555e18916cd7 mlir::vectorizeAffineLoopNest(std::vector<llvm::SmallVector<mlir::AffineForOp, 2u>, std::allocator<llvm::SmallVector<mlir::AffineForOp, 2u>>>&, mlir::VectorizationStrategy const&) (/home/ty/llvm-project/build/bin/mlir-opt+0x306cd7)
#12 0x0000555e19d943ef (anonymous namespace)::VectorizerTestPass::runOnOperation() (.part.261) TestVectorizationUtils.cpp:0:0
#13 0x0000555e19d952fb (anonymous namespace)::VectorizerTestPass::runOnOperation() TestVectorizationUtils.cpp:0:0
#14 0x0000555e1a155bee mlir::detail::OpToOpPassAdaptor::run(mlir::Pass*, mlir::Operation*, mlir::AnalysisManager, bool, unsigned int) (/home/ty/llvm-project/build/bin/mlir-opt+0x1b45bee)
#15 0x0000555e1a155f2a mlir::detail::OpToOpPassAdaptor::runPipeline(mlir::OpPassManager&, mlir::Operation*, mlir::AnalysisManager, bool, unsigned int, mlir::PassInstrumentor*, mlir::PassInstrumentation::PipelineParentInfo const*) (/home/ty/llvm-project/build/bin/mlir-opt+0x1b45f2a)
#16 0x0000555e1a156312 mlir::detail::OpToOpPassAdaptor::runOnOperationAsyncImpl(bool)::'lambda'(mlir::detail::OpToOpPassAdaptor::runOnOperationAsyncImpl(bool)::OpPMInfo&)::operator()(mlir::detail::OpToOpPassAdaptor::runOnOperationAsyncImpl(bool)::OpPMInfo&) const Pass.cpp:0:0
#17 0x0000555e1a155335 mlir::detail::OpToOpPassAdaptor::runOnOperationAsyncImpl(bool) (/home/ty/llvm-project/build/bin/mlir-opt+0x1b45335)
#18 0x0000555e1a155b32 mlir::detail::OpToOpPassAdaptor::run(mlir::Pass*, mlir::Operation*, mlir::AnalysisManager, bool, unsigned int) (/home/ty/llvm-project/build/bin/mlir-opt+0x1b45b32)
#19 0x0000555e1a156a43 mlir::detail::OpToOpPassAdaptor::runPipeline(mlir::OpPassManager&, mlir::Operation*, mlir::AnalysisManager, bool, unsigned int, mlir::PassInstrumentor*, mlir::PassInstrumentation::PipelineParentInfo const*) (.constprop.510) Pass.cpp:0:0
#20 0x0000555e1a1570cd mlir::PassManager::run(mlir::Operation*) (/home/ty/llvm-project/build/bin/mlir-opt+0x1b470cd)
#21 0x0000555e1a14b043 performActions(llvm::raw_ostream&, std::shared_ptr<llvm::SourceMgr> const&, mlir::MLIRContext*, mlir::MlirOptMainConfig const&) MlirOptMain.cpp:0:0
#22 0x0000555e1a14b711 processBuffer(llvm::raw_ostream&, std::unique_ptr<llvm::MemoryBuffer, std::default_delete<llvm::MemoryBuffer>>, mlir::MlirOptMainConfig const&, mlir::DialectRegistry&, llvm::ThreadPool*) MlirOptMain.cpp:0:0
#23 0x0000555e1a14b901 mlir::LogicalResult llvm::function_ref<mlir::LogicalResult (std::unique_ptr<llvm::MemoryBuffer, std::default_delete<llvm::MemoryBuffer>>, llvm::raw_ostream&)>::callback_fn<mlir::MlirOptMain(llvm::raw_ostream&, std::unique_ptr<llvm::MemoryBuffer, std::default_delete<llvm::MemoryBuffer>>, mlir::DialectRegistry&, mlir::MlirOptMainConfig const&)::'lambda'(std::unique_ptr<llvm::MemoryBuffer, std::default_delete<llvm::MemoryBuffer>>, llvm::raw_ostream&)>(long, std::unique_ptr<llvm::MemoryBuffer, std::default_delete<llvm::MemoryBuffer>>, llvm::raw_ostream&) MlirOptMain.cpp:0:0
#24 0x0000555e1a21d810 mlir::splitAndProcessBuffer(std::unique_ptr<llvm::MemoryBuffer, std::default_delete<llvm::MemoryBuffer>>, llvm::function_ref<mlir::LogicalResult (std::unique_ptr<llvm::MemoryBuffer, std::default_delete<llvm::MemoryBuffer>>, llvm::raw_ostream&)>, llvm::raw_ostream&, bool, bool) (/home/ty/llvm-project/build/bin/mlir-opt+0x1c0d810)
#25 0x0000555e1a146a23 mlir::MlirOptMain(llvm::raw_ostream&, std::unique_ptr<llvm::MemoryBuffer, std::default_delete<llvm::MemoryBuffer>>, mlir::DialectRegistry&, mlir::MlirOptMainConfig const&) (/home/ty/llvm-project/build/bin/mlir-opt+0x1b36a23)
#26 0x0000555e1a14924e mlir::MlirOptMain(int, char**, llvm::StringRef, mlir::DialectRegistry&, bool) (/home/ty/llvm-project/build/bin/mlir-opt+0x1b3924e)
#27 0x0000555e188070c0 main (/home/ty/llvm-project/build/bin/mlir-opt+0x1f70c0)
#28 0x00007feaf0788c87 __libc_start_main /build/glibc-CVJwZb/glibc-2.27/csu/../csu/libc-start.c:344:0
#29 0x0000555e18819f2a _start (/home/ty/llvm-project/build/bin/mlir-opt+0x209f2a)

[llvmbot]

@llvm/issue-subscribers-mlir-affine

[caojoshua]

assertion source

This issue is triggered by the existence of iter_args in the affine.for. The code path through -affine-super-vectorizer-test initializes a VectorizationStrategy without anything in the LoopReductions, raising the assertion. If we instead use affine-super-vectorizer, the reductions are initialized through runOnOperation.

[Lewuathe]

I have submitted a patch to emit the message instead of crashing.
https://reviews.llvm.org/D147765