Skip to content
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

[GPU] do not need to broadcast for eltwise post ops constant argument in onednn gemm #27671

Merged

Conversation

songbell
Copy link
Contributor

@songbell songbell commented Nov 21, 2024

Details:

  • observe huge perf drop with onednn binary-ops per tensor mode (e.g. per_tensor POLICY input or integer 15 input)):

image

in this model, the argument of the binary is constant, but it has been broadcasted to match the gemm output tensor layout, yet it leads to huge perf drop

  • remove the unecessary broadcast for scalar argument which is later feed to onednn binary ops

Tickets:

  • 152186

@github-actions github-actions bot added the category: GPU OpenVINO GPU plugin label Nov 21, 2024
@songbell songbell marked this pull request as ready for review November 25, 2024 02:11
@songbell songbell requested review from a team as code owners November 25, 2024 02:11
@songbell songbell changed the title [draft-do not review]do not broadcast for scalar [GPU]do not need to broadcast for eltwise post ops constant argument in onednn gemm Nov 25, 2024
@songbell songbell requested a review from sshlyapn November 25, 2024 07:08
Signed-off-by: fishbell <[email protected]>
@sshlyapn sshlyapn added this to the 2025.0 milestone Nov 26, 2024
@vladimir-paramuzov vladimir-paramuzov added this pull request to the merge queue Nov 26, 2024
@vladimir-paramuzov vladimir-paramuzov changed the title [GPU]do not need to broadcast for eltwise post ops constant argument in onednn gemm [GPU] do not need to broadcast for eltwise post ops constant argument in onednn gemm Nov 26, 2024
Merged via the queue into openvinotoolkit:master with commit 4c5228b Nov 26, 2024
154 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
category: GPU OpenVINO GPU plugin
Projects
None yet
Development

Successfully merging this pull request may close these issues.

3 participants