[ExecuTorch][WebGPU] Migrate multi-dispatch ops to make_compute_pipeline#20869
Open
JCNTH wants to merge 1 commit into
Open
[ExecuTorch][WebGPU] Migrate multi-dispatch ops to make_compute_pipeline#20869JCNTH wants to merge 1 commit into
JCNTH wants to merge 1 commit into
Conversation
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/20869
Note: Links to docs will display an error until the docs builds have been completed. ❌ 1 Cancelled JobAs of commit aa943fc with merge base aceeb40 ( CANCELLED JOB - The following job was cancelled. Please retry:
This comment was automatically generated by Dr. CI and updates every 15 minutes. |
This PR needs a
|
This was referenced Jul 10, 2026
[ExecuTorch][WebGPU] Add relu op (shared unary handler; sigmoid adopts make_compute_pipeline)
#20863
Open
This was referenced Jul 10, 2026
meta-cla
Bot
added
the
CLA Signed
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Stack from ghstack (oldest at bottom):
Route the multi-dispatch ops through
utils::make_compute_pipeline(layout:"auto"), extending the dedup to the remaining pipeline-building ops:rope: its two dispatches (xq, xk) now each build their own pipeline via the helper, so the shared bind-group-layout + its double-free guard are removed.cat: the per-input dispatch loop builds a bundle per input via the helper.sdpa/sdpa_fd_decode: the internalbuild_dispatchhelper builds each pass via the helper; the hand-ordered read_write-vs-read-only binding count is now derived from the shader by the driver.This trades one amortized shader module for a per-call shader compile in these ops — build-time only, zero runtime cost. Byte-behavior preserved: same passes, binding indices/types/buffers/sizes, dispatch counts, uniforms,
own_uniform_buffer, resize hooks, and validations. sdpa's softmax pass (wg_size==0) passesconstant_count=0, matching the prior null-constants path.Differential Revision: D110836676