[ET-VK][ez] Fix using a temporary variable when creating ComputePipeline
#9405
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
…line` ## Context Fix changes introduced in #8634 (D70021032). This change decoupled local work group size from additional specialization constants. As part of this change, when creating a `VkComputePipeline` a temporary `SpecVarList` is created to merge the `WorkgroupSize` with additional specialization constants. However, this can be an issue with some Vulkan drivers because the `SpecVarList` is a temporary, and thus will be destroyed at the end of the function call. The pointer stored in the `VkSpecializationInfo` will be invalidated, leading to undefined behaviour. This diff fixes this by restoring the behaviour of `ComputePipeline::Descriptor` storing the `specialization_constants` directly. Also fix the fact that the `VkSpecializationMapEntry` vector was also a temporary when creating a `VkComputePipeline` by storing it in `ComputePipeline`. Differential Revision: [D71488015](https://our.internmc.facebook.com/intern/diff/D71488015/) [ghstack-poisoned]
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/9405
Note: Links to docs will display an error until the docs builds have been completed. ❌ 1 New FailureAs of commit ed7e765 with merge base 6daff83 ( NEW FAILURE - The following job has failed:
This comment was automatically generated by Dr. CI and updates every 15 minutes. |
SS-JIA
added a commit
that referenced
this pull request
Mar 19, 2025
…line` ## Context Fix changes introduced in #8634 (D70021032). This change decoupled local work group size from additional specialization constants. As part of this change, when creating a `VkComputePipeline` a temporary `SpecVarList` is created to merge the `WorkgroupSize` with additional specialization constants. However, this can be an issue with some Vulkan drivers because the `SpecVarList` is a temporary, and thus will be destroyed at the end of the function call. The pointer stored in the `VkSpecializationInfo` will be invalidated, leading to undefined behaviour. This diff fixes this by restoring the behaviour of `ComputePipeline::Descriptor` storing the `specialization_constants` directly. Also fix the fact that the `VkSpecializationMapEntry` vector was also a temporary when creating a `VkComputePipeline` by storing it in `ComputePipeline`. Differential Revision: [D71488015](https://our.internmc.facebook.com/intern/diff/D71488015/) ghstack-source-id: 272783852 Pull Request resolved: #9405
facebook-github-bot
added
the
CLA Signed
|
This pull request was exported from Phabricator. Differential Revision: D71488015 |
This PR needs a
|
trivedivivek
approved these changes
Mar 20, 2025
facebook-github-bot
merged commit e9899fd
into
gh/SS-JIA/197/base
80 of 84 checks passed
facebook-github-bot
temporarily deployed
to
cherry-pick-bot
GitHub Actions
Inactive
SS-JIA
added a commit
that referenced
this pull request
Mar 20, 2025
…line` (#9464) This PR was created by the merge bot to help merge the original PR into the main branch. ghstack PR number: #9405 by @SS-JIA ^ 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/SS-JIA/197/base ghstack PR head: https://github.com/pytorch/executorch/tree/gh/SS-JIA/197/head Merge bot PR base: https://github.com/pytorch/executorch/tree/main Merge bot PR head: https://github.com/pytorch/executorch/tree/gh/SS-JIA/197/orig @diff-train-skip-merge Co-authored-by: Stephen Jia <[email protected]>
oscarandersson8218
pushed a commit
to oscarandersson8218/executorch
that referenced
this pull request
Mar 21, 2025
…line` (pytorch#9464) This PR was created by the merge bot to help merge the original PR into the main branch. ghstack PR number: pytorch#9405 by @SS-JIA ^ 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/SS-JIA/197/base ghstack PR head: https://github.com/pytorch/executorch/tree/gh/SS-JIA/197/head Merge bot PR base: https://github.com/pytorch/executorch/tree/main Merge bot PR head: https://github.com/pytorch/executorch/tree/gh/SS-JIA/197/orig @diff-train-skip-merge Co-authored-by: Stephen Jia <[email protected]>
SS-JIA
added a commit
that referenced
this pull request
Mar 26, 2025
## Context As title; similar to #9016 since the interface for `ComputePipeline` descriptor was reverted in #9405. Differential Revision: [D71706868](https://our.internmc.facebook.com/intern/diff/D71706868/) [ghstack-poisoned]
SS-JIA
added a commit
that referenced
this pull request
Mar 26, 2025
## Context As title; similar to #9016 since the interface for `ComputePipeline` descriptor was reverted in #9405. Differential Revision: [D71706868](https://our.internmc.facebook.com/intern/diff/D71706868/) ghstack-source-id: 274200673 Pull Request resolved: #9648
SS-JIA
added a commit
that referenced
this pull request
Mar 26, 2025
## Context As title; similar to #9016 since the interface for `ComputePipeline` descriptor was reverted in #9405. Differential Revision: [D71706868](https://our.internmc.facebook.com/intern/diff/D71706868/) [ghstack-poisoned]
SS-JIA
added a commit
that referenced
this pull request
Mar 26, 2025
Pull Request resolved: #9648 ## Context As title; similar to #9016 since the interface for `ComputePipeline` descriptor was reverted in #9405. ghstack-source-id: 274260087 @exported-using-ghexport Differential Revision: [D71706868](https://our.internmc.facebook.com/intern/diff/D71706868/)
kirklandsign
pushed a commit
that referenced
this pull request
Mar 27, 2025
## Context As title; similar to #9016 since the interface for `ComputePipeline` descriptor was reverted in #9405. Differential Revision: [D71706868](https://our.internmc.facebook.com/intern/diff/D71706868/) [ghstack-poisoned]
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.
Stack from ghstack (oldest at bottom):
ComputePipeline#9405Context
Fix changes introduced in #8634 (D70021032).
This change decoupled local work group size from additional specialization constants. As part of this change, when creating a
VkComputePipelinea temporarySpecVarListis created to merge theWorkgroupSizewith additional specialization constants.However, this can be an issue with some Vulkan drivers because the
SpecVarListis a temporary, and thus will be destroyed at the end of the function call. The pointer stored in theVkSpecializationInfowill be invalidated, leading to undefined behaviour.This diff fixes this by restoring the behaviour of
ComputePipeline::Descriptorstoring thespecialization_constantsdirectly.Also fix the fact that the
VkSpecializationMapEntryvector was also a temporary when creating aVkComputePipelineby storing it inComputePipeline.Differential Revision: D71488015