Skip to content

Conversation

@jondea
Copy link
Contributor

@jondea jondea commented Jan 9, 2026

This flag appears to have been copied verbatim from x64, and it doesn't seem to apply to AArch64. Specifically, on x64, turning the flag on enables the use of fewer instructions. On AArch64, it uses more instructions. The comment about when it is turned on also doesn't make sense, vector length is orthogonal to features.

Removing it to make it easier to add new functionality to jit_brgemm_kernel_t.

Checklist

General

  • Do all unit and benchdnn tests (make test and make test_benchdnn_*) pass locally for each commit?
  • Have you formatted the code using clang-format?

This flag appears to have been copied verbatim from x64, and it doesn't seem to
apply to AArch64. Specifically, on x64, turning the flag on enables the use of
fewer instructions. On AArch64, it uses _more_ instructions. The comment about
when it is turned on also doesn't make sense, vector length is orthogonal to
features.
@jondea jondea requested a review from a team as a code owner January 9, 2026 10:39
@github-actions github-actions bot added the platform:cpu-aarch64 Codeowner: @oneapi-src/onednn-cpu-aarch64 label Jan 9, 2026
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

platform:cpu-aarch64 Codeowner: @oneapi-src/onednn-cpu-aarch64

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant