Repository navigation
[executorch][cuda] Autotune SDPA on a visible mask and tune pack GQA with the tile - #24004
Merged
Merged
Conversation
…with the tile Pull Request resolved: #23727 D123681196 declared representative KV lengths for SDPA kernels that read KV_LEN_ptr. Masked calls still autotuned unrepresentative work: fabricated boolean masks were all false, so candidate kernels skipped visible KV blocks. Pack-GQA selection also relied on a fixed tile-utilization heuristic rather than the candidate tile. Declare Mask_ptr beside KV_LEN_ptr on the forward, decode-split-K and small-query-split-K kernels, using an all-visible mask for each representative KV-length scenario. Each candidate therefore sweeps the declared visible extent. Calls without masks pass zero and remain untouched. Make PACK_GQA a forward-kernel config dimension. Tune packed and unpacked candidates for GQA, prune packing without GQA, and derive the launch-grid head count from the chosen config. The non-power-of-two kernel keeps its fixed-BLOCK_M rule. Existing masked numerical/reference and autotune-input tests cover the contract. ghstack-source-id: 451457287 @exported-using-ghexport Differential Revision: [D124908769](https://our.internmc.facebook.com/intern/diff/D124908769/)
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/24004
Note: Links to docs will display an error until the docs builds have been completed. ⏳ No Failures, 192 PendingAs of commit 0f9a035 with merge base 897d28f ( This comment was automatically generated by Dr. CI and updates every 15 minutes. |
meta-cla
Bot
added
the
CLA Signed
This PR needs a
|
shoumikhin
approved these changes
Oct 20, 2026
This branch was successfully deployed
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
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
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 was created by the merge bot to help merge the original PR into the main branch.
ghstack PR number: #23727 by @Gasoonjia
^ Please use this as the source of truth for the PR details, comments, and reviews
ghstack PR base: https://github.com/pytorch/executorch/tree/gh/Gasoonjia/43/base
ghstack PR head: https://github.com/pytorch/executorch/tree/gh/Gasoonjia/43/head
Merge bot PR base: https://github.com/pytorch/executorch/tree/main
Merge bot PR head: https://github.com/pytorch/executorch/tree/gh/Gasoonjia/43/orig
Differential Revision: D124908769
@diff-train-skip-merge