Skip to content

.NET 10 vs main (Mid-July data) CoreCLR Performance Regression Tracking issue #5281

Description

@LoopedBard3

[.NET 10 vs main] CoreCLR performance regression tracking

This issue tracks the actionable CoreCLR results from the reviewed release/10.0 vs main windowed comparison. Draft generated 2026-08-17. GitHub issue/PR states refreshed from live metadata on 2026-08-17. User decisions and AI-attributed notes applied on 2026-08-17.

Use the checkboxes in place: leave an item unchecked while investigation, filing, ownership, or verification remains; check it when that work is finished. Feel free to edit or append status directly on the same checklist line as work progresses. There is intentionally no separate Completed section.

Within each section, items are ordered by worst reviewed percentage (largest first). Regression count is used only to break ties.

Inclusion rules

Included: needs-investigation, confirmed runtime regressions, unmarked runtime/autofiling/missed items, and missed regressions marked truly-missed.

Excluded: fixed, acceptable, not-applicable, noise, improved, baseline-outlier, covered, and linked.

Actionable summary

Area Items
Needs investigation (runtime + ungrouped missed) 37
Active tracked runtime regressions 13
Autofiling issues needing transfer/resolution 28
Missed regression date groups needing filing 9
Total checkboxes 87

Needs investigation

Show 37 items

Items requiring a second look before deciding whether to file, transfer, accept, or close them.

  • dotnet/runtime#125085needs-investigation; closed-completed; 162 CoreCLR reviewed regressions; worst +1122.4%; regression date 2026-02-20. Review note: Some regressions still real, many seem to have been noise. Need to double check System.Numerics.Tests.Perf_BigInteger and System.Collections.Concurrent.IsEmpty at least Representative chart.

  • dotnet/runtime#128998needs-investigation; closed-not-planned; 44 CoreCLR reviewed regressions; worst +1122.4%; regression date 2026-06-02. Review note: Many System.Buffers.Tests.RentReturnArrayPoolTests tests were (and are) regressed while this was closed as noise. AI Note: Verify whether sibling fix PR #126622 resolved the ArrayPool regressions before closing this tracker item. (evidence) Representative chart.

  • dotnet/runtime#129180acceptable; closed-completed; 36 CoreCLR reviewed regressions; worst +584.4%; regression date 2026-06-06. Review note: Still regressed, initial look into done, but last update was June 10th and not part of a milestone. AI Note (user-approved): Closed as an accepted Zen4 REX-prefix microarchitectural trade-off with no viable mitigation. (evidence) Representative chart.

  • dotnet/runtime#120068needs-investigation; closed-completed; 49 CoreCLR reviewed regressions; worst +132.3%; regression date 2025-09-22. Review note: Seems System.Collections.Concurrent.IsEmpty may still be hitting issues. Representative chart.

  • dotnet/runtime#125501not-applicable; closed-completed; 1 CoreCLR reviewed regressions; worst +110.2%; regression date 2026-03-10. Review note: Cause wasn't fully tracked down and it seems to now be bimodal. AI Note (user-approved): Closed after investigation concluded this was not an actual performance regression. (evidence) Representative chart.

  • dotnet/runtime#130697needs-investigation; open; 30 CoreCLR reviewed regressions; worst +95.0%; regression date 2026-07-08. Review note: Added to milestone but nothing new in a month or so Representative chart.

  • dotnet/runtime#125781needs-investigation; open; 7 CoreCLR reviewed regressions; worst +45.0%; regression date 2026-03-17. Review note: No real follow up since March Representative chart.

  • dotnet/runtime#129860needs-investigation; open; 90 CoreCLR reviewed regressions; worst +39.4%; regression date 2026-06-24. Review note: Still very real and no non-wasm work has started or been surfaced recently. Representative chart.

  • System.Threading.Tests.Perf_CancellationToken.RegisterAndUnregister_Serialneeds-investigation; Ubuntu-arm64-Ampere; +39.3%; 40.7 → 56.7. Review note: Nearly recovered, maybe still a 1-4 % regression when bimodal noise is taken into account.

  • dotnet/runtime#126154fixed; closed-completed; 5 CoreCLR reviewed regressions; worst +35.2%; regression date 2026-03-23. Review note: Last updated Jun 9th. User update: Fixed and nearly unregressed. Representative chart.

  • dotnet/runtime#128908needs-investigation; open; 1 CoreCLR reviewed regressions; worst +34.6%; regression date 2026-05-27. Review note: Added to Milestone 11 and given Pri3 tag (July 9th) AI Note (user-approved): The owner committed to investigate after snap and keep the work in .NET 11. (evidence) Representative chart.

  • dotnet/runtime#129585needs-investigation; open; 1 CoreCLR reviewed regressions; worst +33.3%; regression date 2026-06-16. Review note: Added to Milestone 11 on June 18th, still real but no follow-up since. AI Note: The fix is known but was deprioritized to .NET 12. (evidence) Representative chart.

  • dotnet/runtime#131159acceptable; closed-completed; 25 CoreCLR reviewed regressions; worst +32.4%; regression date 2026-07-14. Review note: Seems related to the same change, both CoreCLR and Wasm-Mono seem regressed. AI Note (user-approved): Closed as an accepted trade-off from scoping vectorization away from browser/WASM. (evidence) Representative chart.

  • dotnet/runtime#126884needs-investigation; open; 12 CoreCLR reviewed regressions; worst +32.4%; regression date 2026-04-12. Review note: In Milestone 11, last activity is a question from April 17th. AI Note: Normalization overhead is now the leading root-cause mechanism and an owner is engaged. (evidence) Representative chart.

  • dotnet/runtime#118641needs-investigation; PR-merged; 3 CoreCLR reviewed regressions; worst +32.1%; regression date 2025-08-18. Review note: Weird bounce back and still real on some, though the regression amount is pretty small. Representative chart.

  • dotnet/runtime#129998needs-investigation; open; 4 CoreCLR reviewed regressions; worst +29.7%; regression date 2026-06-23. Review note: Moved to Milestone 11 on June 29th, but regressions are still real. Representative chart.

  • dotnet/runtime#128208needs-investigation; open; 3 CoreCLR reviewed regressions; worst +28.4%; regression date 2026-05-07. Review note: Added to Milestone 11 on May 15, no more follow-up. AI Note (user-approved): PR #127845 provided a partial fix; remaining work was deferred to .NET 12. (evidence) Representative chart.

  • dotnet/runtime#125504needs-investigation; closed-not-planned; 13 CoreCLR reviewed regressions; worst +26.9%; regression date 2026-03-07. Review note: Assigned 07/15 but no confirmation. The CrC seem to be the main issues. AI Note (user-approved): Closed not planned and deferred to Future; the root causes are known but no fix landed. (evidence) Representative chart.

  • dotnet/runtime#127230acceptable; closed-completed; 11 CoreCLR reviewed regressions; worst +26.8%; regression date 2026-04-18. Review note: Has initial investigation work started, but no updates since Jun 03 AI Note (user-approved): Closed as an expected early-inline trade-off; improvements outweighed regressions overall. (evidence) Representative chart.

  • dotnet/runtime#128710needs-investigation; closed-not-planned; 2 CoreCLR reviewed regressions; worst +20.3%; regression date 2026-05-20. Review note: Still real and no action post transfer (May 28th) AI Note (user-approved): Closed not planned; likely caused by PR #128303 and closed without a fix. (evidence) Representative chart.

  • dotnet/runtime#114996needs-investigation; closed-completed; 5 CoreCLR reviewed regressions; worst +17.6%; regression date 2026-04-27. Review note: Need to ensure System.Buffers.Tests.ReadOnlySequenceTests are being tracked elsewhere. Representative chart.

  • dotnet/runtime#127012not-applicable; closed-completed; 3 CoreCLR reviewed regressions; worst +17.6%; regression date 2026-04-09. Review note: Added to 11 milestone and last updated Apr 27. Probably should just be considered as part of general runtime async work. AI Note (user-approved): Closed after the movement was attributed to bimodal test behavior rather than a true regression. (evidence) Representative chart.

  • BenchmarksGame.Fasta_1.RunBenchneeds-investigation; Ubuntu-arm64-Ampere; +17.4%; 581,025 → 682,015. Review note: No individual date seems to be the cause, just a slow decrease overtime.

  • System.Collections.ContainsKeyTrue<Int32, Int32>.ImmutableDictionary(Size: 512)needs-investigation; Win-x64-Viper; +13.3%; 3,773 → 4,276. Review note: Very noisy and mostly sideways, but the noisiness bounds shrink for an unclear reason.

  • dotnet/runtime#130044needs-investigation; open; 5 CoreCLR reviewed regressions; worst +10.3%; regression date 2026-06-25. Review note: Last updated nearly a month ago (investigation started though). AI Note: Assigned, marked as a bug, and milestoned to 11.0; no fix yet. (evidence) Representative chart.

  • IfStatements.IfStatements.OrArrayneeds-investigation; Ubuntu-x64-Viper; +9.3%; 30,495 → 33,324. Review note: Nothing clear from the .NET 11 timeframe

  • System.Reflection.Metadata.Perf_TypeName.ParseAndGetAssemblyQualifiedName(input: typeof(System.Collections.Generic.Dictionary<String, Boolean>))needs-investigation; Win-x64-Viper; +8.9%; 2,040 → 2,222. Review note: Got worse over time, nothing clear.

  • System.Reflection.Metadata.Perf_TypeName.ParseAndGetAssemblyQualifiedName(input: typeof(int))needs-investigation; Win-x64-Viper; +8.8%; 647.6 → 704.8. Review note: Got worse over time, nothing clear.

  • System.IO.Tests.Perf_Path.GetFullPathNoRedundantSegmentsneeds-investigation; Win-x64-Viper; +8.3%; 174.2 → 188.6. Review note: Very noisy, without clear regression dates.

  • System.Reflection.Metadata.Perf_TypeName.ParseAndGetAssemblyQualifiedName(input: typeof(System.Int32[,]))needs-investigation; Win-x64-Viper; +8.2%; 700.1 → 757.2. Review note: Got worse over time, nothing clear.

  • dotnet/runtime#121413needs-investigation; closed-completed; 2 CoreCLR reviewed regressions; worst +8.1%; regression date 2025-10-28. Review note: Added to Milestone 11, no activity since then (Nov 24, 2025) AI Note (user-approved): State is closed-completed, but the closure rationale was not captured; verify before replacing needs-investigation with fixed. (evidence) Representative chart.

  • System.Reflection.Metadata.Perf_TypeName.ParseAndGetAssemblyQualifiedName(input: typeof(System.Int32*))needs-investigation; Win-x64-Viper; +6.8%; 682.5 → 729.1. Review note: Got worse over time, nothing clear.

  • [System.Reflection.Metadata.Perf_TypeName.ParseAndGetAssemblyQualifiedName(input: typeof(System.Collections.Generic.Dictionary2[]))](https://pvscmdupload.z22.web.core.windows.net/reports/allTestHistory/refs/heads/main_x64_Windows%2010.0.22631/ViperWindows/System.Reflection.Metadata.Perf_TypeName.ParseAndGetAssemblyQualifiedName%28input%3A%20typeof%28System.Collections.Generic.Dictionary%602%5B%5D%29%29.html) — **needs-investigation**; Win-x64-Viper`; +6.7%; 2,120 → 2,263. Review note: Got worse over time, nothing clear.

  • System.Reflection.Metadata.Perf_TypeName.ParseAndGetAssemblyQualifiedName(input: typeof(System.Int32&))needs-investigation; Win-x64-Viper; +6.6%; 686.8 → 732.0. Review note: Got worse over time, nothing clear.

  • [System.Reflection.Metadata.Perf_TypeName.ParseAndGetFullName(input: typeof(System.Collections.Generic.Dictionary2[]))](https://pvscmdupload.z22.web.core.windows.net/reports/allTestHistory/refs/heads/main_x64_Windows%2010.0.22631/ViperWindows/System.Reflection.Metadata.Perf_TypeName.ParseAndGetFullName%28input%3A%20typeof%28System.Collections.Generic.Dictionary%602%5B%5D%29%29.html) — **needs-investigation**; Win-x64-Viper`; +6.6%; 1,532 → 1,632. Review note: Got worse over time, nothing clear.

  • System.Reflection.Metadata.Perf_TypeName.ParseAndGetAssemblyQualifiedName(input: typeof(System.Int32[*]))needs-investigation; Win-x64-Viper; +5.6%; 697.4 → 736.6. Review note: Got worse over time, nothing clear.

  • dotnet/runtime#123394needs-investigation; closed-completed; 1 CoreCLR reviewed regressions; worst +5.2%; regression date 2026-01-12. Review note: In Milestone 11 but no activity since Jan. AI Note (user-approved): GitHub is closed-completed, but no closure rationale was captured; verify before changing the mark. (evidence) Representative chart.

Active tracked regressions

Show 13 items

Existing runtime issues marked confirmed in the review. Treat every unchecked item here as a still-live regression, especially when its status has not been checked recently; check or edit the line when that changes.

  • dotnet/runtime#123159fixed; closed-completed; 43 CoreCLR reviewed regressions; worst +1122.4%; regression date 2026-01-06. Review note: Pretty sure some work is still being done for this. RentReturnArrayPoolTests seem to have unregressed from this change, Perf_File* slightly unregressed, but not all the way. User update: Fixed; checkbox should be checked in the public tracker. Representative chart.

  • dotnet/runtime#122186confirmed; closed-completed; 37 CoreCLR reviewed regressions; worst +1122.4%; regression date 2025-12-03. Review note: Still open and tracked (no recent updates). Many of the regressions seem unregressed, at least one System.Buffers.Tests.RentReturnArrayPoolTests<Object>.ProducerConsumer is still regressed above this issue. User update: RentReturnArrayPoolTests ProducerConsumer tests are still regressed. Representative chart.

  • dotnet/runtime#127326confirmed; closed-completed; 1 CoreCLR reviewed regressions; worst +164.7%; regression date 2026-04-13. Review note: Closed as not actionable, and wasn't a major regression contributor for the test. Representative chart.

  • dotnet/runtime#128594confirmed; open; 13 CoreCLR reviewed regressions; worst +129.9%; regression date 2026-05-21. Review note: Assigned 07/09, BigInt seems to be noise, Perf_Decimal is real. Representative chart.

  • dotnet/runtime#130309confirmed; open; 10 CoreCLR reviewed regressions; worst +99.6%; regression date 2026-07-03. Review note: In Milestone, last updated 3w ago. AI Note: Milestone moved from .NET 11 to .NET 12. (evidence) Representative chart.

  • dotnet/runtime#119499confirmed; closed-completed; 29 CoreCLR reviewed regressions; worst +71.1%; regression date 2025-09-04. Review note: Still some minor regression, but mostly unregressed by Improve performance of IsCollectible reflection properties runtime#119743 (still maybe a 10-15% regression for most it seems). Representative chart.

  • dotnet/runtime#130418confirmed; open; 13 CoreCLR reviewed regressions; worst +55.4%; regression date 2026-07-06. Review note: Seems to be real, untriaged currently. AI Note (user-approved): Upgraded to confirmed; assigned, milestoned to 11.0, and an active PR is in progress. (evidence) Representative chart.

  • dotnet/runtime#117775confirmed; open; 3 CoreCLR reviewed regressions; worst +48.0%; regression date 2025-10-14. Review note: General Investigation added to 12 milestone AI Note (user-approved): Milestone moved from .NET 11 to .NET 12; mark and state remain confirmed/open. (evidence) Representative chart.

  • dotnet/runtime#123677confirmed; open; 15 CoreCLR reviewed regressions; worst +47.4%; regression date 2026-01-23. Review note: Still real, set to P3 on milestone 11 July 9th AI Note: Priority was raised to P1; the owner committed to post-snap .NET 11 investigation. (evidence) Representative chart.

  • dotnet/runtime#130041fixed; closed-completed; 18 CoreCLR reviewed regressions; worst +40.0%; regression date 2026-06-29. Review note: In milestone, last updated 3w ago. User update: Fix is not 100%, but the associated project was closed. Representative chart.

  • dotnet/runtime#119770confirmed; open; 18 CoreCLR reviewed regressions; worst +37.6%; regression date 2025-09-10. Review note: Transferred and in work queue for 11, marked pri2 07/09. Tests are still regressed. AI Note: Priority was raised to P1; the owner committed to post-snap investigation. (evidence) Representative chart.

  • dotnet/runtime#122592confirmed; closed-not-planned; 1 CoreCLR reviewed regressions; worst +21.1%; regression date 2025-12-11. Review note: This is basically just noise around this regression issue. Representative chart.

  • dotnet/runtime#128910confirmed; closed-completed; 3 CoreCLR reviewed regressions; worst +16.8%; regression date 2026-05-27. Review note: All linked regressions seem to just be noise. Representative chart.

Autofiling issues needing transfer or resolution

Show 28 items

Uncorrelated autofiling issues that are unmarked or marked needs-investigation.

Missed regression groups needing filing

Show 9 items

Dated groups containing at least one needs-investigation or truly-missed regression. The checkbox represents the group as one candidate public tracking issue.

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions