[all-commits] [llvm/llvm-project] 79ff70: [mlir][sparse] Better error handling when bufferiz...

[Matthias Springer via All-commits]

  Branch: refs/heads/main
  Home:   https://github.com/llvm/llvm-project
  Commit: 79ff70fda223f22cf409301b0f0c2bf451dbb6b3
      https://github.com/llvm/llvm-project/commit/79ff70fda223f22cf409301b0f0c2bf451dbb6b3
  Author: Matthias Springer <me at m-sp.org>
  Date:   2023-08-25 (Fri, 25 Aug 2023)

  Changed paths:
    M mlir/lib/Dialect/SparseTensor/Transforms/BufferizableOpInterfaceImpl.cpp
    A mlir/test/Dialect/SparseTensor/one_shot_bufferize_invalid.mlir

  Log Message:
  -----------
  [mlir][sparse] Better error handling when bufferizing sparse_tensor ops

sparse_tensor ops cannot be bufferized with One-Shot Bufferize. (They can only be analyzed.) The sparse compiler does the actual lowering to memref. Produce a proper error message instead of crashing.

This fixes #61311.

Differential Revision: https://reviews.llvm.org/D158728