Synchronize no-GPU cache eviction with CPU streams#3566
Merged
Conversation
The no-GPU CPU allocator cache could release cached buffers while asynchronous CPU work was still queued on a non-default stream. mlx-lm generation can hit this by queuing the next token on a CPU stream and calling clear_cache between yielded tokens, which can leave queued work referencing freed cached memory. Synchronize CPU streams before evicting cached buffers from clear_cache or set_cache_limit, and protect get_cache_memory with the allocator mutex. Add a regression test that verifies clear_cache waits for queued CPU stream work before returning.
angeloskath
approved these changes
May 19, 2026
Member
angeloskath
left a comment
angeloskath
left a comment
There was a problem hiding this comment.
Looks great and makes total sense, thanks!
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Proposed changes
Follow up to #3554 - while testing variations on my AVX2 branch I found this bug on that change.
The no-GPU CPU allocator cache could release cached buffers while asynchronous CPU work was still queued on a non-default stream. mlx-lm generation can hit this by queuing the next token on a CPU stream and calling clear_cache between yielded tokens, which can leave queued work referencing freed cached memory.
Synchronize CPU streams before evicting cached buffers from clear_cache or set_cache_limit, and protect get_cache_memory with the allocator mutex. Add a regression test that verifies clear_cache waits for queued CPU stream work before returning.
Checklist
Put an
xin the boxes that apply.pre-commit run --all-filesto format my code / installed pre-commit prior to committing changes