Move the transpose matmul pass to OSS and run it earlier in the flow #10433
+146
−1
Conversation
This file contains hidden or 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
mcremon-meta
requested review from
JacobSzwejbka,
larryliu0820 and
tarun292
as code owners
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/10433
Note: Links to docs will display an error until the docs builds have been completed. ✅ No FailuresAs of commit c2f2316 with merge base 6b877de ( This comment was automatically generated by Dr. CI and updates every 15 minutes. |
facebook-github-bot
added
the
CLA Signed
Contributor
|
This pull request was exported from Phabricator. Differential Revision: D73600069 |
mcremon-meta
added a commit
that referenced
this pull request
Apr 24, 2025
…10433) Summary: That pass is doing a lot more than it looks, and it's just easier to move it back to where it was. CPU backends will possibly see more cycles due to added permutes, but we don't care about that. All DSP backends should be more efficient on transposed matmuls. Should that not be the case in the future, we can re-evaluate. When we do the survey of passes and reorder them properly, we can think about this more. Differential Revision: D73600069
mcremon-meta
force-pushed
the
export-D73600069
branch
from
47c55ec to
8eb2b6e
Compare
Contributor
|
This pull request was exported from Phabricator. Differential Revision: D73600069 |
mcremon-meta
added a commit
that referenced
this pull request
Apr 24, 2025
…10433) Summary: Pull Request resolved: #10433 That pass is doing a lot more than it looks, and it's just easier to move it back to where it was. CPU backends will possibly see more cycles due to added permutes, but we don't care about that. All DSP backends should be more efficient on transposed matmuls. Should that not be the case in the future, we can re-evaluate. When we do the survey of passes and reorder them properly, we can think about this more. Differential Revision: D73600069
mcremon-meta
force-pushed
the
export-D73600069
branch
from
8eb2b6e to
689a082
Compare
skrtskrtfb
approved these changes
Apr 25, 2025
hsharma35
approved these changes
Apr 25, 2025
facebook-github-bot
pushed a commit
that referenced
this pull request
Apr 26, 2025
…10433) Summary: That pass is doing a lot more than it looks, and it's just easier to move it back to where it was. CPU backends will possibly see more cycles due to added permutes, but we don't care about that. All DSP backends should be more efficient on transposed matmuls. Should that not be the case in the future, we can re-evaluate. When we do the survey of passes and reorder them properly, we can think about this more. Reviewed By: hsharma35 Differential Revision: D73600069
facebook-github-bot
force-pushed
the
export-D73600069
branch
from
689a082 to
8c3b621
Compare
Contributor
|
This pull request was exported from Phabricator. Differential Revision: D73600069 |
mcremon-meta
added a commit
that referenced
this pull request
Apr 26, 2025
…10433) Summary: That pass is doing a lot more than it looks, and it's just easier to move it back to where it was. CPU backends will possibly see more cycles due to added permutes, but we don't care about that. All DSP backends should be more efficient on transposed matmuls. Should that not be the case in the future, we can re-evaluate. When we do the survey of passes and reorder them properly, we can think about this more. Reviewed By: hsharma35 Differential Revision: D73600069
mcremon-meta
force-pushed
the
export-D73600069
branch
from
8c3b621 to
13e7cc7
Compare
…10433) Summary: Pull Request resolved: #10433 That pass is doing a lot more than it looks, and it's just easier to move it back to where it was. CPU backends will possibly see more cycles due to added permutes, but we don't care about that. All DSP backends should be more efficient on transposed matmuls. Should that not be the case in the future, we can re-evaluate. When we do the survey of passes and reorder them properly, we can think about this more. Reviewed By: hsharma35 Differential Revision: D73600069
Contributor
|
This pull request was exported from Phabricator. Differential Revision: D73600069 |
mcremon-meta
force-pushed
the
export-D73600069
branch
from
13e7cc7 to
c2f2316
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.
Summary:
That pass is doing a lot more than it looks, and it's just easier to move it back to where it was. CPU backends will possibly see more cycles due to added permutes, but we don't care about that. All DSP backends should be more efficient on transposed matmuls. Should that not be the case in the future, we can re-evaluate.
When we do the survey of passes and reorder them properly, we can think about this more.
Differential Revision: D73600069