-
Notifications
You must be signed in to change notification settings - Fork 5.7k
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
oneDNN kernels code cleanup #50743
oneDNN kernels code cleanup #50743
Conversation
你的PR提交成功,感谢你对开源项目的贡献! |
@zyfncg could you please review and approve |
@zyfncg thank you very much for review. Could |
@XieYunshen Could you please review this PR for |
I think that problems with low coverage may occur due to our naming unification. Along with PHI migration, we started renaming |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
LGTM
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
TODOs have to be removed. resolved
Problems with low coverage are resolved. @zyfncg could you approve once more and merge this PR? |
PR types
Others
PR changes
Others
Describe
We used to have onednn_reuse.h which included multiple methods and utils related to oneDNN kernels. Goal is to split this header file into smaller, dedicated imports. It will improve readability and linking process. Changes done in this PR:
matmul
,matmul_v2
andfused_matmul
kernels,matmul_v2
kernel was migrated to PHI and this file has onlymatmul_v1
kernels left,SetOutMemDescWithLogicalLayoutFusesSupport
to smaller methods, because onlytranspose2
can have all 3 attributes used in this method. Bothfc
andelementwise
kernels have support just for one attribute, so it can be skipped.