5.9: [ClosureSpecializer] Don't release trivial noescape apply argument twice. #66285
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
When a specialization is created, in the original function, releases are added in two different places: (1) `ClosureSpecCloner::populateCloned` (2) `rewriteApplyInst` In the former, releases are added for closures which are guaranteed or trivial noescape (but with owned convention). In the latter, releases are added for closures that are owned. Previously, when emitting releases at (2), whether the closure was trivial noescape wasn't considered. The result was inserting the releases twice, an overrelease. Here, fix (2) to recognize trivial noescape as not +1. rdar://110115795
|
@swift-ci please test |
tbkka
approved these changes
Jun 1, 2023
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Outdated suggestions cannot be applied.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.
Description: Analogous to #66274 but for
applyinstead oftry_apply.ClosureSpecializer creates releases for the closure passed as an parameter to a
applythat is specialized. These releases are created in (1)ClosureSpecCloner::populateClonedand (2)rewriteApplyInst, depending on the effective ownership of the closure: in (1) they are inserted for closures passed without an increment of the retain count (done e.g. for closures passed as an@guaranteedparameter); in (2) they are inserted for closures passed with an increment of the retain count (done e.g. for closures passed as an@ownedparameter).Among the closures that are passed without an increment of the retain count are those which are noescape (
SILFunctionType::isTrivialNoEscape). That is true even when the closure is passed as an@ownedparameter. The condition of (1) under which releases are added for the closure correctly included consideration of theisTrivialNoEscapepredicate of the closure's type; consequently that code inserts releases in both of theapply's destination blocks. The condition of (2) under which releases were added for the closure incorrectly failed to consider theisTrivialNoEscapepredicate; consequently, that code also inserted releases in both of the destination blocks. The result was adding two releases in each destination block, one of each of which was an overrelease.The fix is just to consider the
isTrivialNoEscapepredicate in (2).Risk: Low. The patch adds a predicate to the bailout condition of (2) that complements an existing use of the same predicate in the condition of (1).
Scope: Narrow. This only affects releases for noescape closures passed as owned arguments to
applys which are specialized.Original PR: #66275
Reviewed By: Arnold Schwaigofer ( @aschwaighofer )
Testing: Added test that previously inserted double release is no longer present.
Resolves: rdar://110115795