[XLA:GPU] Do not multi-output fuse sibling transposes with reductions. #28786
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
![copybara-service[bot]](https://avatars.githubusercontent.com/in/44061?s=60&v=4){kind=small}
copybara-service
bot
force-pushed
the
test_781870005
branch
2 times, most recently
from
0692bc2 to
7048b9e
Compare
A elemental (aka nested) transpose has a different read pattern than an unnested transpose or reduction, because the emitters for tiled transposes and parallel reductions ensure uniform read patterns. A multi-output fusion with roots that have different read patterns is generally not profitable. Input data will be read multiple times by different threads, which defeats the purpose of multi-output fusion. What is worse, increased register pressure can impact performance. PiperOrigin-RevId: 783214366
copybara-service
bot
force-pushed
the
test_781870005
branch
from
7048b9e to
c7f65c9
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Outdated suggestions cannot be applied.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.
[XLA:GPU] Do not multi-output fuse sibling transposes with reductions.
A elemental (aka nested) transpose has a different read pattern than an unnested transpose or reduction, because the emitters for tiled transposes and parallel reductions ensure uniform read patterns.
A multi-output fusion with roots that have different read patterns is generally not profitable. Input data will be read multiple times by different threads, which defeats the purpose of multi-output fusion. What is worse, increased register pressure can impact performance.