[mono][sgen] Fix incorrect condition when checking if we should do a major collection #120432
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
…major collection When we are in this method, major sweep hasn't yet finished so we check with some imprecise conservative values. min_heap_size is a conservative lower limit of the actual heap size. Allowance represents how much more we allow the heap to grow after the previous major collection. We were comparing the min_heap_size with the allowance by mistake, when we should have compared it with the next major trigger size (which is the heap_size after the last collection plus the allowance).
Contributor
There was a problem hiding this comment.
Pull Request Overview
This PR fixes an incorrect condition in the Mono SGEN garbage collector that was causing excessive major garbage collections. The bug was introduced in a previous change and resulted in performance regression from .NET 8 to .NET 9.
Key changes:
- Corrects the major collection trigger logic by comparing minimum heap size against the proper threshold
- Fixes a condition that was incorrectly comparing heap size with allowance instead of total trigger size
Contributor
|
Tagging subscribers to this area: @BrzVlad |
Member
Author
|
/ba-g wasm unrelated |
Member
Author
|
/backport to release/10.0 |
Contributor
|
Started backporting to release/10.0: https://github.com/dotnet/runtime/actions/runs/18348341019 |
Member
Author
|
/backport to release/9.0-staging |
Contributor
|
Started backporting to release/9.0-staging: https://github.com/dotnet/runtime/actions/runs/18348346199 |
steveisok
pushed a commit
that referenced
this pull request
Nov 4, 2025
…we should do a major collection (#120532) Backport of #120432 to release/10.0 /cc @BrzVlad ## Customer Impact - [x] Customer reported - [ ] Found internally This impacts all platforms using mono. A customer reported a 40x regression on an XML Serialization benchmark when moving from .NET8 to .NET9. This regression is caused by the runtime triggering excessive GCs due to a bug in the heuristic. This is more likely to happen if the GC concurrent sweep phase takes longer and the application allocates large objects frequently. This likely impacts many applications, but with less noticeable performance impact. ## Regression - [x] Yes - [ ] No This regressed in .NET9 ## Testing Tested the fix on sample benchmark that it works correctly. ## Risk Low risk. The fix is localized and the previous condition was blatantly wrong by mistake. Co-authored-by: Vlad Brezae <[email protected]>
Sign up for free
to subscribe to this conversation on GitHub.
Already have an account?
Sign in.
2 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.
When we are in this method, major sweep hasn't yet finished so we check with some imprecise conservative values. min_heap_size is a conservative lower limit of the actual heap size. Allowance represents how much more we allow the heap to grow after the previous major collection. We were comparing the min_heap_size with the allowance by mistake, when we should have compared it with the next major trigger size (which is the heap_size after the last collection plus the allowance).
Wrong condition added in #98154, causing excessive GCs in some scenarios and regression from .net8 to .net9.
Fixes #119580