[all-commits] [llvm/llvm-project] f88150: [MLIR][XeGPU][VectorToXeGPU] Fixed transposed tran...

[Andrey Pavlenko via All-commits]

  Branch: refs/heads/main
  Home:   https://github.com/llvm/llvm-project
  Commit: f88150e1d5fc8591b0d71619d02e5b553161c1f8
      https://github.com/llvm/llvm-project/commit/f88150e1d5fc8591b0d71619d02e5b553161c1f8
  Author: Andrey Pavlenko <andrey.a.pavlenko at gmail.com>
  Date:   2026-07-13 (Mon, 13 Jul 2026)

  Changed paths:
    M mlir/lib/Conversion/VectorToXeGPU/VectorToXeGPU.cpp
    M mlir/test/Conversion/VectorToXeGPU/transfer-read-to-xegpu.mlir

  Log Message:
  -----------
  [MLIR][XeGPU][VectorToXeGPU] Fixed transposed transfer_read/write (#197281)

The problem was that adjustStridesForPermutation() permutes strides, but
they are required to compute base offset from transfer read indices.
Added optional argument to computeMemrefMeta() to preserve the original
strides.



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