[Fix][Relax] Run destructors for non-trivially-destructible types in Arena - #20163
Open
OmarAzizi wants to merge 2 commits into
Open
[Fix][Relax] Run destructors for non-trivially-destructible types in Arena#20163OmarAzizi wants to merge 2 commits into
OmarAzizi wants to merge 2 commits into
Conversation
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
1 participant
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Summary
support::Arenanever calls destructors but frees raw memory pages.GraphPartitioner::Groupviolates this by holding a ref countedattrsmap, which leaks native memory on every pass invocation since~Group()is never called.Fixes #20056.
Fix
Implements option (b) from the issue:
Arena::make<T>()now registers invokes a destructor for non-trivially-destructible types, guarded byif constexprso trivially destructible types pay no extra cost. Fixes the leak at the source instead of per call site.Went with option (b) instead of (a) as the same pattern also exists in [
LiftedFunctionRewritePlan] in (rewrite_cuda_graph.cc) which would fixGroupbut leave that and any future case unfixed.Testing
Reproduced the leak using the script attached to the issue (updated for the
TensorStructInfo->Typeunification in #19853, otherwise unchanged): ~4.96 MiB/iter before this fix, ~0 MiB/iter after. Existing Relax fusion/partitioning tests pass.