[AOTI] Fix a small buffer mutation issue#169347
Closed
desertfire wants to merge 4 commits intogh/desertfire/614/basefrom
Closed
[AOTI] Fix a small buffer mutation issue#169347desertfire wants to merge 4 commits intogh/desertfire/614/basefrom
desertfire wants to merge 4 commits intogh/desertfire/614/basefrom
Conversation
Summary: Fix #169118. A small named buffer can be optimized as inlined constant, but if that named buffer is mutated later on, AOTI can generate wrong code if not respecting the mutation information. This PR fixes the issue by recording mutated_named_buffers when unlifting the input graph and pass to Inductor graph lowering. [ghstack-poisoned]
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/169347
Note: Links to docs will display an error until the docs builds have been completed. ✅ No FailuresAs of commit 6c7a272 with merge base c6ae757 ( This comment was automatically generated by Dr. CI and updates every 15 minutes. |
desertfire
added a commit
that referenced
this pull request
Dec 2, 2025
Summary: Fix #169118. A small named buffer can be optimized as inlined constant, but if that named buffer is mutated later on, AOTI can generate wrong code if not respecting the mutation information. This PR fixes the issue by recording mutated_named_buffers when unlifting the input graph and pass to Inductor graph lowering. ghstack-source-id: 9840de9 Pull Request resolved: #169347
Summary: Fix #169118. A small named buffer can be optimized as inlined constant, but if that named buffer is mutated later on, AOTI can generate wrong code if not respecting the mutation information. This PR fixes the issue by recording mutated_named_buffers when unlifting the input graph and pass to Inductor graph lowering. cc voznesenskym penguinwu EikanWang jgong5 Guobing-Chen XiaobingSuper zhuhaozhe blzheng wenzhe-nrv jiayisunx ipiszy kadeng muchulee8 amjames chauhang aakhundov coconutruben jataylo [ghstack-poisoned]
desertfire
added a commit
that referenced
this pull request
Dec 2, 2025
Summary: Fix #169118. A small named buffer can be optimized as inlined constant, but if that named buffer is mutated later on, AOTI can generate wrong code if not respecting the mutation information. This PR fixes the issue by recording mutated_named_buffers when unlifting the input graph and pass to Inductor graph lowering. ghstack-source-id: 842e21c Pull Request resolved: #169347
Summary: Fix #169118. A small named buffer can be optimized as inlined constant, but if that named buffer is mutated later on, AOTI can generate wrong code if not respecting the mutation information. This PR fixes the issue by recording mutated_named_buffers when unlifting the input graph and pass to Inductor graph lowering. cc voznesenskym penguinwu EikanWang jgong5 Guobing-Chen XiaobingSuper zhuhaozhe blzheng wenzhe-nrv jiayisunx ipiszy kadeng muchulee8 amjames chauhang aakhundov coconutruben jataylo [ghstack-poisoned]
desertfire
added a commit
that referenced
this pull request
Dec 2, 2025
Summary: Fix #169118. A small named buffer can be optimized as inlined constant, but if that named buffer is mutated later on, AOTI can generate wrong code if not respecting the mutation information. This PR fixes the issue by recording mutated_named_buffers when unlifting the input graph and pass to Inductor graph lowering. ghstack-source-id: bedd5e1 Pull Request resolved: #169347
yushangdi
approved these changes
Dec 3, 2025
yushangdi
reviewed
Dec 3, 2025
| self.named_buffers: dict[str, torch.Tensor] = ( | ||
| const_module.named_buffers if const_module else {} | ||
| ) | ||
| self.mutated_named_buffers: OrderedSet[torch.Tensor] = gm.meta.get( |
Contributor
There was a problem hiding this comment.
can or should we re-use self.mutated_buffers?
Contributor
Author
There was a problem hiding this comment.
self.mutated_buffers is only filled as we iterate through the graph, so by the point we check target in self.mutated_buffers, self.mutated_buffers may not have accurate enough mutation information.
Summary: Fix #169118. A small named buffer can be optimized as inlined constant, but if that named buffer is mutated later on, AOTI can generate wrong code if not respecting the mutation information. This PR fixes the issue by recording mutated_named_buffers when unlifting the input graph and pass to Inductor graph lowering. cc voznesenskym penguinwu EikanWang jgong5 Guobing-Chen XiaobingSuper zhuhaozhe blzheng wenzhe-nrv jiayisunx ipiszy kadeng muchulee8 amjames chauhang aakhundov coconutruben jataylo [ghstack-poisoned]
Contributor
Author
|
@pytorchbot merge |
Collaborator
Merge startedYour change will be merged once all checks pass (ETA 0-4 Hours). Learn more about merging in the wiki. Questions? Feedback? Please reach out to the PyTorch DevX Team |
umechand-amd
pushed a commit
to ROCm/pytorch
that referenced
this pull request
Dec 8, 2025
Summary: Fix pytorch#169118. A small named buffer can be optimized as inlined constant, but if that named buffer is mutated later on, AOTI can generate wrong code if not respecting the mutation information. This PR fixes the issue by recording mutated_named_buffers when unlifting the input graph and pass to Inductor graph lowering. Pull Request resolved: pytorch#169347 Approved by: https://github.com/yushangdi
JacobSzwejbka
pushed a commit
that referenced
this pull request
Dec 8, 2025
Summary: Fix #169118. A small named buffer can be optimized as inlined constant, but if that named buffer is mutated later on, AOTI can generate wrong code if not respecting the mutation information. This PR fixes the issue by recording mutated_named_buffers when unlifting the input graph and pass to Inductor graph lowering. Pull Request resolved: #169347 Approved by: https://github.com/yushangdi
tiendatngcs
pushed a commit
to tiendatngcs/pytorch-Dec25
that referenced
this pull request
Dec 10, 2025
Summary: Fix pytorch/pytorch#169118. A small named buffer can be optimized as inlined constant, but if that named buffer is mutated later on, AOTI can generate wrong code if not respecting the mutation information. This PR fixes the issue by recording mutated_named_buffers when unlifting the input graph and pass to Inductor graph lowering. ghstack-source-id: 6eec08b Pull Request resolved: pytorch/pytorch#169347
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
3 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):
Summary: Fix #169118. A small named buffer can be optimized as inlined constant, but if that named buffer is mutated later on, AOTI can generate wrong code if not respecting the mutation information. This PR fixes the issue by recording mutated_named_buffers when unlifting the input graph and pass to Inductor graph lowering.
cc @voznesenskym @penguinwu @EikanWang @jgong5 @Guobing-Chen @XiaobingSuper @zhuhaozhe @blzheng @wenzhe-nrv @jiayisunx @ipiszy @kadeng @muchulee8 @amjames @chauhang @aakhundov @coconutruben @jataylo