[MLIR]`OneShotBufferizePass` crashes: Assertion `memref::CastOp::areCastCompatible(resultBuffer->getType(), *resultMemRefType) && "CallOp::bufferize: cast incompatible"' failed.

[BealZephyr]

Test commit: da0730b

Steps to reproduce:

mlir-opt test.mlir -one-shot-bufferize=allow-unknown-ops

Test case:

func.func @main(%arg0: tensor<*xi32>) {
  %0 = tensor.cast %arg0 : tensor<*xi32> to tensor<*xi32>
  return 
}

Crash trace:

mlir-opt: /home/ty/llvm-project/mlir/lib/Dialect/Tensor/Transforms/BufferizableOpInterfaceImpl.cpp:101: mlir::LogicalResult mlir::tensor::{anonymous}::CastOpInterface::bufferize(mlir::Operation*, mlir::RewriterBase&, const mlir::bufferization::BufferizationOptions&) const: Assertion `memref::CastOp::areCastCompatible(resultBuffer->getType(), *resultMemRefType) && "CallOp::bufferize: cast incompatible"' 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 -one-shot-bufferize=allow-unknown-ops
 #0 0x000055fd675f9f2a llvm::sys::PrintStackTrace(llvm::raw_ostream&, int) (/home/ty/llvm-project/build/bin/mlir-opt+0xd63f2a)
 #1 0x000055fd675f709c SignalHandler(int) Signals.cpp:0:0
 #2 0x00007f2d809a7980 __restore_rt (/lib/x86_64-linux-gnu/libpthread.so.0+0x12980)
 #3 0x00007f2d7f658e87 raise /build/glibc-CVJwZb/glibc-2.27/signal/../sysdeps/unix/sysv/linux/raise.c:51:0
 #4 0x00007f2d7f65a7f1 abort /build/glibc-CVJwZb/glibc-2.27/stdlib/abort.c:81:0
 #5 0x00007f2d7f64a3fa __assert_fail_base /build/glibc-CVJwZb/glibc-2.27/assert/assert.c:89:0
 #6 0x00007f2d7f64a472 (/lib/x86_64-linux-gnu/libc.so.6+0x30472)
 #7 0x000055fd68972646 mlir::bufferization::detail::BufferizableOpInterfaceInterfaceTraits::FallbackModel<mlir::tensor::(anonymous namespace)::CastOpInterface>::bufferize(mlir::bufferization::detail::BufferizableOpInterfaceInterfaceTraits::Concept const*, mlir::Operation*, mlir::RewriterBase&, mlir::bufferization::BufferizationOptions const&) BufferizableOpInterfaceImpl.cpp:0:0
 #8 0x000055fd6796a862 mlir::bufferization::bufferizeOp(mlir::Operation*, mlir::bufferization::BufferizationOptions const&, bool, mlir::bufferization::OpFilter const*, mlir::bufferization::BufferizationStatistics*) (/home/ty/llvm-project/build/bin/mlir-opt+0x10d4862)
 #9 0x000055fd679693a4 (anonymous namespace)::OneShotBufferizePass::runOnOperation() Bufferize.cpp:0:0
#10 0x000055fd694a43ce mlir::detail::OpToOpPassAdaptor::run(mlir::Pass*, mlir::Operation*, mlir::AnalysisManager, bool, unsigned int) (/home/ty/llvm-project/build/bin/mlir-opt+0x2c0e3ce)
#11 0x000055fd694a5223 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
#12 0x000055fd694a58ad mlir::PassManager::run(mlir::Operation*) (/home/ty/llvm-project/build/bin/mlir-opt+0x2c0f8ad)
#13 0x000055fd694977b0 performActions(llvm::raw_ostream&, std::shared_ptr<llvm::SourceMgr> const&, mlir::MLIRContext*, mlir::MlirOptMainConfig const&) MlirOptMain.cpp:0:0
#14 0x000055fd69497f55 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
#15 0x000055fd69498111 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
#16 0x000055fd69578290 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+0x2ce2290)
#17 0x000055fd694936f9 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+0x2bfd6f9)
#18 0x000055fd6949843a mlir::MlirOptMain(int, char**, llvm::StringRef, mlir::DialectRegistry&) (/home/ty/llvm-project/build/bin/mlir-opt+0x2c0243a)
#19 0x000055fd675bc9cd main (/home/ty/llvm-project/build/bin/mlir-opt+0xd269cd)
#20 0x00007f2d7f63bc87 __libc_start_main /build/glibc-CVJwZb/glibc-2.27/csu/../csu/libc-start.c:344:0
#21 0x000055fd675cf66a _start (/home/ty/llvm-project/build/bin/mlir-opt+0xd3966a)

[Lewuathe]

Tried to fix the issue here. https://reviews.llvm.org/D149239

[Lewuathe]

Lowering memref.cast op with unranked memref for input and output to another dialect (e.g. LLVM) is not allowed. We can support a bufferization from unranked to unranked, but it may cause runtime errors later after lowering more.

So what we can do is to make the pass return the failure instead of catching in the assertion?