[all-commits] [llvm/llvm-project] 10e126: [mlir][bufferization] Support matching permutation...

[Ziliang Zhang via All-commits]

  Branch: refs/heads/main
  Home:   https://github.com/llvm/llvm-project
  Commit: 10e12674cb9c52611d1f3c3c4c92ce43b586b7fe
      https://github.com/llvm/llvm-project/commit/10e12674cb9c52611d1f3c3c4c92ce43b586b7fe
  Author: Ziliang Zhang <zzl.coding at gmail.com>
  Date:   2026-09-25 (Fri, 25 Sep 2026)

  Changed paths:
    M mlir/lib/Dialect/Linalg/Transforms/BufferizableOpInterfaceImpl.cpp
    M mlir/test/Dialect/Linalg/one-shot-bufferize-analysis.mlir
    M mlir/test/Dialect/Linalg/one-shot-bufferize.mlir

  Log Message:
  -----------
  [mlir][bufferization] Support matching permutation maps in Linalg analysis (#226079)

Generalize `LinalgOpInterface::bufferizesToElementwiseAccess` to
recognize participating tensor operands with identical permutation
indexing maps, instead of requiring identity maps.

This enables in-place bufferization after loop interchange. Projected
and mismatching maps, reduction iterators, and sparse operands remain
conservatively rejected.

The tests cover the analysis decision as well as end-to-end
bufferization without an allocation or copy.



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