ci: aarch64: Initial nightly performance test #2837
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
Adding a comprehensive performance test to the nightly pipeline. For now, matmul, conv, eltwise and reorders are tested but more tests are to be added in the future. Other minor changes: - Fix benchdnn comparison script - Update precommit performance tests list - Add fast math tests
Ryo-not-rio
force-pushed
the
ryo-not-rio/performance-test-ci
branch
from
64fa5ad to
6f1a913
Compare
jondea
reviewed
Mar 11, 2025
| @@ -0,0 +1,46 @@ | |||
| #! /bin/bash | |||
There was a problem hiding this comment.
Why do we need a separate script for fast math? The BF16 paths should all be exercised by setting the --attr-fast-math benchdnn flag?
There was a problem hiding this comment.
ahhhh was not aware, good to know!
Sqvid
requested changes
Mar 12, 2025
| "$2 --reorder --mode=P --perf-template=%prb%,%-time% --batch=${SCRIPT_DIR}/inputs/reorder_nightly >> $4" | ||
| ) | ||
|
|
||
| for i in {1..5} |
There was a problem hiding this comment.
I think we should make the number of runs a variable so it always stays in sync between the loop, and the print statements below.
Radu2k
approved these changes
Mar 12, 2025
Sqvid
requested changes
Mar 13, 2025
| for i in {1..5} | ||
| N = 5 | ||
|
|
||
| for i in {1..$N} |
There was a problem hiding this comment.
I don't think this works when N is a variable. I think you want for i in $(seq ${N})
There was a problem hiding this comment.
I see you implemented the seq change but keep in mind by doing seq 0 $N rather than seq $N you have changed the number of loops from N to N+1. Was this intended?
| for i in {1..5} | ||
| N = 5 | ||
|
|
||
| for i in {1..$N} |
Ryo-not-rio
force-pushed
the
ryo-not-rio/performance-test-ci
branch
from
7e3fc41 to
ce967d5
Compare
Ryo-not-rio
force-pushed
the
ryo-not-rio/performance-test-ci
branch
from
ce967d5 to
54e3b69
Compare
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.
Adding a comprehensive performance test to the nightly pipeline. For now, matmul, conv, eltwise and reorders are tested but more tests are to be added in the future.
Other minor changes:
A run of the new pipeline can be seen in https://github.com/oneapi-src/oneDNN/actions/runs/13707641078. It is currently failing as it is catching a regression that needs to be solved but the pipeline itself is working as it should.