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

contrib/intel/jenkins: Disable ZE IPC path and use SAR instead #9903

Merged
merged 2 commits into from
Mar 19, 2024

Conversation

zachdworkin
Copy link
Contributor

The performance is really bad and takes very long. Until we debug it the ZE IPC path it will be disabled and will use SAR instead.

The performance is really bad and takes very long. Until we debug
it the ZE IPC path it will be disabled and will use SAR instead.

Signed-off-by: Zach Dworkin <zachary.dworkin@intel.com>
@j-xiong
Copy link
Contributor

j-xiong commented Mar 18, 2024

@zachdworkin The 'oneCCL-GPU-v3' stage is also slow with ZE IPC.

ZE IPC path is slow and needs to be debugged. Disabling it
forces oneccl-gpu tests to take the SAR path instead.

Signed-off-by: Zach Dworkin <zachary.dworkin@intel.com>
@zachdworkin zachdworkin force-pushed the jenkins_cb branch 2 times, most recently from 8f12a8e to d1bb80c Compare March 18, 2024 23:34
@j-xiong
Copy link
Contributor

j-xiong commented Mar 18, 2024

Wait until the 'v1.21.x' branch is created.

@j-xiong
Copy link
Contributor

j-xiong commented Mar 19, 2024

The v1.21.x branch has been created.

@j-xiong
Copy link
Contributor

j-xiong commented Mar 19, 2024

Doesn't seem to make any difference.

@aingerson
Copy link
Contributor

@j-xiong Based on the performance numbers in the log, it definitely did. It just doesn't look like it because it spent a long time waiting for nodes from other PRs that are still having the issue.

@zachdworkin zachdworkin merged commit ff2a621 into ofiwg:main Mar 19, 2024
12 checks passed
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.

3 participants