Add support for more intrinsics in attention #248
Merged
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
This PR adds support for the 32x32x8 intrinsic in attention. This should allow for a wider search space when tuning attention. In order to do so, the following changes were required: - Add reshape as an anchor op in thread shape analysis, primarily to block the backward propagation of thread shapes from the MMA operands - Propagate thread shapes from MMA operands only if they are propagatable - Modify assignment of anchor ops to take indexing dims into account - Modify the computation of cluster size and stride - Simplification of code that adds reshapes Signed-off-by: Harsh Menon <[email protected]>
harsh-nod
requested review from
raikonenfnu,
Hardcode84,
martin-luecke and
erman-gurses
raikonenfnu
reviewed
Nov 1, 2024
raikonenfnu
reviewed
Nov 1, 2024
raikonenfnu
reviewed
Nov 1, 2024
raikonenfnu
reviewed
Nov 1, 2024
Signed-off-by: Harsh Menon <[email protected]>
raikonenfnu
approved these changes
Nov 1, 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Outdated suggestions cannot be applied.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.
This PR adds support for the 32x32x8 intrinsic in
attention. This should allow for a wider search space when tuning attention.
In order to do so, the following changes were required:
Add reshape as an anchor op in thread shape analysis, primarily to block the backward propagation of thread shapes from the MMA operands
Modify assignment of anchor ops to take indexing dims into account