[all-commits] [llvm/llvm-project] f53b62: [MLIR][Linalg] Fix empty tensor assumptions for li...

[maxbartel via All-commits]

  Branch: refs/heads/main
  Home:   https://github.com/llvm/llvm-project
  Commit: f53b6249c24005d1a6208cd9e355595eb6519dc0
      https://github.com/llvm/llvm-project/commit/f53b6249c24005d1a6208cd9e355595eb6519dc0
  Author: maxbartel <bartel at roofline.ai>
  Date:   2025-10-08 (Wed, 08 Oct 2025)

  Changed paths:
    M mlir/include/mlir/Dialect/Linalg/IR/LinalgRelayoutOps.td
    M mlir/lib/Dialect/Linalg/Transforms/Transforms.cpp
    M mlir/test/Dialect/Linalg/decompose-pack.mlir

  Log Message:
  -----------
  [MLIR][Linalg] Fix empty tensor assumptions for linalg.pack decomposition (#160246)

The original code seemed to assume that the tiling dimensions for the
tensor.empty op before applying the transposing were always the last
dimensions. However, pack allows you to choose any dimension to tile.

The easiest way I found to solve this is to prefill the SmallVector with
1s of size (srcRank - numberOfTiles) and then appending the tile sizes.

This way I could also get rid of the first loop in the code.



To unsubscribe from these emails, change your notification settings at https://github.com/llvm/llvm-project/settings/notifications