[oneDNN] Extend shape of elementwise_add/sub for swin_transformer #59421
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
|
你的PR提交成功，感谢你对开源项目的贡献! |
|
Hi @yuanlehome, could you please help check CI coverage? Since I saw the result is kind of unexpected, where header file got 0% coverage... However the upper level code (in elementwise_kernel.cc) got called thousands of times, and there is no condition causing return or something before entering header file. Thanks! |
|
@xinyu-intel, @vivienfanghuagood, @yuanlehome, hi, would you mind helping review this PR? Thanks~ |
yuanlehome
reviewed
Nov 30, 2023
yuanlehome
approved these changes
Dec 1, 2023
xinyu-intel
approved these changes
Dec 1, 2023
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
PR types
Performance optimization
PR changes
Others
Description
This PR together with #59394 are aiming for optimizing swin_transformer in #59252.
The reported regression is caused by perf regression of elementwise_add/sub and fused matmul. Due to rare input shapes, these two kernels can not utilize corresponding broadcast on oneDNN, hence the regression. Here we workaround to extend shape of input/residual_data and keep utilizing optimization of oneDNN