[XPU][OptRed] Revamp -tritonintelgpu-optimize-reduction-locality
#2800
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
Original implementation had two critical issues: - *Functional*: It did not preserve register order, so it was computing a different reduction. - *Performance*: When converting back to the original tensor type, it did: `reshape(convert_layout(res))`. That means the `reshape` operation served as an anchor and the suboptimal slice layout was propagated. This was fixed as follows: - Keep register order. - Do `convert_layout(reshape(res))` when converting back to the original type, thus propagating the more optimal layout. See implementation for further details. Signed-off-by: victor-eds <[email protected]>
etiotto
approved these changes
Nov 26, 2024
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Outdated suggestions cannot be applied.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.
Original implementation had two critical issues:
reshape(convert_layout(res)). That means thereshapeoperation served as an anchor and the suboptimal slice layout was propagated.This was fixed as follows:
convert_layout(reshape(res))when converting back to the original type, thus propagating the more optimal layout.See implementation for further details.
Closes #2752