Skip to content

[Transform] Refinements on microkernel dialect lowering #324

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 7 commits into from
Sep 9, 2024

Conversation

huanghaixin008
Copy link
Contributor

@huanghaixin008 huanghaixin008 commented Sep 5, 2024

Tracking #325

This PR introduce a few minor refinements to the microkernel dialect lowering, including:

  • Fix segment fault from edge cases of microkernel.brgemmOp fusion
  • Support dynamic stride in batch dimension & leading dimension if their dimension value is 1
  • For VNNI cases, change from linalgx.batch_reduce_matmul_vnni to linalg.generic frontend

@huanghaixin008
Copy link
Contributor Author

@zhczhong Could you please help to check the DeepTileContractionOp changes regarding to not using linalgx part? I tried to change per comments but ended up with segment fault when running DeepTileContractionOp mlir tests.

@huanghaixin008 huanghaixin008 merged commit 866d2ee into main Sep 9, 2024
6 checks passed
zhczhong pushed a commit that referenced this pull request Sep 13, 2024
* [WIP] minor fix & improvements

* fix allow dynamic stride

* change linalgx frontend to linalg.generic

* fix clang-tify

* fix deepTileContractionOp

* fix clang-tidy
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

4 participants