[Perf] Windows/x64: 7 Regressions on 4/28/2026 9:27:00 PM +00:00

[performanceautofiler]

Run Information

Name	Value
Architecture	x64
OS	Windows 10.0.22631
Queue	ViperWindows
Baseline	aba46e33ea5ddd45d90e5c6a8b46bba6744ddc9a
Compare	977f4125aa7cf8fd550f1bb91e966f20a1df6e54
Diff	Diff
Configs	CompilationMode:tiered, RunKind:micro

Regressions in System.Numerics.Tensors.Tests.Perf_FloatingPointTensorPrimitives<Double>

Benchmark	Baseline	Test	Test/Base	Test Quality	Edge Detector	Baseline IR	Compare IR	IR Ratio
Pow_ScalarExponent - Duration of single invocation 📝 - Benchmark Source ADX - Test Multi Config Graph	15.04 μs	16.77 μs	1.12	0.20	False
Pow_ScalarExponent - Duration of single invocation 📝 - Benchmark Source ADX - Test Multi Config Graph	627.57 ns	696.38 ns	1.11	0.15	False

Test Report

Repro

General Docs link: https://github.com/dotnet/performance/blob/main/docs/benchmarking-workflow-dotnet-runtime.md

git clone https://github.com/dotnet/performance.git
py .\performance\scripts\benchmarks_ci.py -f net8.0 --filter 'System.Numerics.Tensors.Tests.Perf_FloatingPointTensorPrimitives<Double>*'

Details

System.Numerics.Tensors.Tests.Perf_FloatingPointTensorPrimitives<Double>.Pow_ScalarExponent(BufferLength: 3079)

ETL Files

Histogram

JIT Disasms

System.Numerics.Tensors.Tests.Perf_FloatingPointTensorPrimitives<Double>.Pow_ScalarExponent(BufferLength: 128)

ETL Files

Histogram

JIT Disasms

Docs

Profiling workflow for dotnet/runtime repository
Benchmarking workflow for dotnet/runtime repository

---

Run Information

Name	Value
Architecture	x64
OS	Windows 10.0.22631
Queue	ViperWindows
Baseline	aba46e33ea5ddd45d90e5c6a8b46bba6744ddc9a
Compare	977f4125aa7cf8fd550f1bb91e966f20a1df6e54
Diff	Diff
Configs	CompilationMode:tiered, RunKind:micro

Regressions in System.Collections.Perf_Frozen<NotKnownComparable>

Benchmark	Baseline	Test	Test/Base	Test Quality	Edge Detector	Baseline IR	Compare IR	IR Ratio
ToFrozenDictionary - Duration of single invocation 📝 - Benchmark Source ADX - Test Multi Config Graph	1.60 μs	1.70 μs	1.06	0.11	False

Test Report

Repro

General Docs link: https://github.com/dotnet/performance/blob/main/docs/benchmarking-workflow-dotnet-runtime.md

git clone https://github.com/dotnet/performance.git
py .\performance\scripts\benchmarks_ci.py -f net8.0 --filter 'System.Collections.Perf_Frozen<NotKnownComparable>*'

Details

System.Collections.Perf_Frozen<NotKnownComparable>.ToFrozenDictionary(Count: 64)

ETL Files

Histogram

JIT Disasms

Docs

Profiling workflow for dotnet/runtime repository
Benchmarking workflow for dotnet/runtime repository

---

Run Information

Name	Value
Architecture	x64
OS	Windows 10.0.22631
Queue	ViperWindows
Baseline	aba46e33ea5ddd45d90e5c6a8b46bba6744ddc9a
Compare	977f4125aa7cf8fd550f1bb91e966f20a1df6e54
Diff	Diff
Configs	CompilationMode:tiered, RunKind:micro

Regressions in System.Linq.Tests.Perf_Enumerable

Benchmark	Baseline	Test	Test/Base	Test Quality	Edge Detector	Baseline IR	Compare IR	IR Ratio
All_AllElementsMatch - Duration of single invocation 📝 - Benchmark Source ADX - Test Multi Config Graph	45.30 ns	56.53 ns	1.25	0.09	False

Test Report

Repro

General Docs link: https://github.com/dotnet/performance/blob/main/docs/benchmarking-workflow-dotnet-runtime.md

git clone https://github.com/dotnet/performance.git
py .\performance\scripts\benchmarks_ci.py -f net8.0 --filter 'System.Linq.Tests.Perf_Enumerable*'

Details

System.Linq.Tests.Perf_Enumerable.All_AllElementsMatch(input: IEnumerable)

ETL Files

Histogram

JIT Disasms

Docs

Profiling workflow for dotnet/runtime repository
Benchmarking workflow for dotnet/runtime repository

---

Run Information

Name	Value
Architecture	x64
OS	Windows 10.0.22631
Queue	ViperWindows
Baseline	aba46e33ea5ddd45d90e5c6a8b46bba6744ddc9a
Compare	977f4125aa7cf8fd550f1bb91e966f20a1df6e54
Diff	Diff
Configs	CompilationMode:tiered, RunKind:micro

Regressions in BenchmarksGame.NBody_3

Benchmark	Baseline	Test	Test/Base	Test Quality	Edge Detector	Baseline IR	Compare IR	IR Ratio
RunBench - Duration of single invocation 📝 - Benchmark Source ADX - Test Multi Config Graph	316.15 ms	333.32 ms	1.05	0.00	False

Test Report

Repro

General Docs link: https://github.com/dotnet/performance/blob/main/docs/benchmarking-workflow-dotnet-runtime.md

git clone https://github.com/dotnet/performance.git
py .\performance\scripts\benchmarks_ci.py -f net8.0 --filter 'BenchmarksGame.NBody_3*'

Details

BenchmarksGame.NBody_3.RunBench

ETL Files

Histogram

JIT Disasms

Docs

Profiling workflow for dotnet/runtime repository
Benchmarking workflow for dotnet/runtime repository

---

Run Information

Name	Value
Architecture	x64
OS	Windows 10.0.22631
Queue	ViperWindows
Baseline	aba46e33ea5ddd45d90e5c6a8b46bba6744ddc9a
Compare	977f4125aa7cf8fd550f1bb91e966f20a1df6e54
Diff	Diff
Configs	CompilationMode:tiered, RunKind:micro

Regressions in Benchstone.MDBenchI.MDAddArray2

Benchmark	Baseline	Test	Test/Base	Test Quality	Edge Detector	Baseline IR	Compare IR	IR Ratio
Test - Duration of single invocation 📝 - Benchmark Source ADX - Test Multi Config Graph	10.15 ms	11.13 ms	1.10	0.07	False

Test Report

Repro

General Docs link: https://github.com/dotnet/performance/blob/main/docs/benchmarking-workflow-dotnet-runtime.md

git clone https://github.com/dotnet/performance.git
py .\performance\scripts\benchmarks_ci.py -f net8.0 --filter 'Benchstone.MDBenchI.MDAddArray2*'

Details

Benchstone.MDBenchI.MDAddArray2.Test

ETL Files

Histogram

JIT Disasms

Docs

Profiling workflow for dotnet/runtime repository
Benchmarking workflow for dotnet/runtime repository

---

Run Information

Name	Value
Architecture	x64
OS	Windows 10.0.22631
Queue	ViperWindows
Baseline	aba46e33ea5ddd45d90e5c6a8b46bba6744ddc9a
Compare	977f4125aa7cf8fd550f1bb91e966f20a1df6e54
Diff	Diff
Configs	CompilationMode:tiered, RunKind:micro

Regressions in System.Tests.Perf_String

Benchmark	Baseline	Test	Test/Base	Test Quality	Edge Detector	Baseline IR	Compare IR	IR Ratio
IndexerCheckBoundCheckHoist - Duration of single invocation 📝 - Benchmark Source ADX - Test Multi Config Graph	26.04 ns	40.48 ns	1.55	0.01	False

Test Report

Repro

General Docs link: https://github.com/dotnet/performance/blob/main/docs/benchmarking-workflow-dotnet-runtime.md

git clone https://github.com/dotnet/performance.git
py .\performance\scripts\benchmarks_ci.py -f net8.0 --filter 'System.Tests.Perf_String*'

Details

System.Tests.Perf_String.IndexerCheckBoundCheckHoist

ETL Files

Histogram

JIT Disasms

Docs

Profiling workflow for dotnet/runtime repository
Benchmarking workflow for dotnet/runtime repository

[LoopedBard3]

⚠️ PerfTriageAgent encountered an error while analyzing this issue.

Error: TransientError: Failed to start MCP server

Please escalate to the performance infrastructure team if this issue needs triage.

[LoopedBard3]

🔍 Automated Triage Analysis

Summary: Verified regression in Pow_ScalarExponent(BufferLength: 128) — caused by commit 0b6bb45 ("Remove unwind calls from patchpoint OSR transition (#127213)"). 2 tests bisected to related culprit commits. ⚠️ Note: Historical data shows bimodal behavior — bisection result may be unreliable.

Finding Confidence: 3/5 (analysis accuracy)
Regression Confidence: 3/5 (likelihood of true regression)

Likely Cause: 0b6bb45eef865b383453d6bb57ca4d479df25266

📊 Full Analysis (click to expand)

Issue Overview

• Issue: [Perf] Windows/x64: 7 Regressions on 4/28/2026 9:27:00 PM +00:00 #127813 - [Perf] Windows/x64: 7 Regressions on 4/28/2026 9:27:00 PM +00:00
• Date: 2026-04-28 | Queue: Windows.11.Amd64.Viper.Perf | OS/Arch: windows/x64
• Commits: aba46e33ea5d → 977f4125aa7c (38 commits)
• Labels: untriaged, perf-regression, os-windows, arch-x64, branch-refs/heads/main, runtime-coreclr, kind-micro, compilationmode-tiered, runkind-micro
• Historical Data: Available (298-299 runs per test)

Test Analysis

Test Name	Baseline	Compare	Δ%	μ	σ	n	z	Thresh%	Noise?	Assessment
Perf_FloatingPointTensorPrimitives.Pow_ScalarExponent(3079)	15036.8	16766.9	+11.51%	17369.5	6430.3	299	-0.09	74.04%	Y	Bimodal noise
Perf_FloatingPointTensorPrimitives.Pow_ScalarExponent(128)	627.6	696.4	+10.96%	742.2	296.1	298	-0.16	79.80%	Y	Bimodal noise
Perf_Frozen.ToFrozenDictionary(Count: 64)	1601.5	1697.0	+5.97%	1614.6	56.0	298	1.47	6.94%	N	Borderline step change
Perf_Enumerable.All_AllElementsMatch(IEnumerable)	45.3	56.5	+24.79%	50.3	7.7	298	0.81	30.41%	N	Borderline step change (multi-regime)
BenchmarksGame.NBody_3.RunBench	316150542	333318000	+5.43%	312955168	3434378	299	5.93	5.00%	N	Confirmed regression
Benchstone.MDBenchI.MDAddArray2.Test	10145891	11128695	+9.69%	10592902	457538	299	1.17	8.64%	N	Borderline drift
Perf_String.IndexerCheckBoundCheckHoist	26.0	40.5	+55.47%	27.4	5.1	298	2.58	36.84%	N	Confirmed regression

Legend: μ=mean, σ=stddev, n=runs, z=z-score, Thresh%=adaptive threshold

Finding Confidence: 4/5 - Excellent historical data (298-299 runs per test), clear statistical signals for confirmed regressions, well-characterized noise patterns.

Confirmed Regressions

BenchmarksGame.NBody_3.RunBench (+5.43%)

• Delta: 316,150,542 ns → 333,318,000 ns (+5.43%)
• Historical Context: μ=312,955,168 (σ=3,434,378, n=299); z-score=5.93; adaptive threshold=5.00%
• Assessment: Clean step change on historically very stable test (CV=1.1%). All 28 post-regression runs tightly clustered at ~333M, clearly separated from the ~310-318M historical range.
• Finding Confidence: 4/5 - Excellent data quality, very stable benchmark, extreme z-score
• Regression Confidence: 4/5 - z=5.93 is extremely significant on a stable benchmark; multiple JIT commits in range complicate precise attribution
• Likely Cause: a59fcc58eb05 - "JIT: Transform 'CONST - x' to 'xor x, CONST' when possible (JIT: Transform 'CONST - x' to 'xor x, CONST' when possible #126529)" or 4c86aad28c63 - "JIT: eliminate redundant bounds checks for span.Slice(span.Length - cns) (JIT: eliminate redundant bounds checks for span.Slice(span.Length - cns) #127488)"
  • Affected Areas: JIT (value numbering, morph, range check)
  • Rationale: Both commits modify core JIT optimization passes (value numbering identities and arithmetic transforms). NBody_3 is a compute-intensive floating-point benchmark with tight loops. The a59fcc58eb05 commit adds a new transform path in fgOptimizeAddition that changes how ADD(NEG(x), CONST) is handled - potentially altering code generation for arithmetic patterns used in the N-body simulation. The 4c86aad28c63 commit adds VN identity x - (x + a) == -a which could cause cascading optimization differences. Both committed on 2026-04-28, aligning with the regression date.
  • Candidates Considered:
    • a59fcc58eb05 - JIT arithmetic transform changes in morph.cpp, modifies fgOptimizeAddition
    • 4c86aad28c63 - JIT VN identity change in valuenum.cpp, new range check folding
    • 0b6bb45eef86 - Remove unwind calls from patchpoint OSR transition (touches codegen paths broadly)
• Verification Status: Not verified

System.Tests.Perf_String.IndexerCheckBoundCheckHoist (+55.47%)

• Delta: 26.0 ns → 40.5 ns (+55.47%)
• Historical Context: μ=27.4 (σ=5.1, n=298); z-score=2.58; adaptive threshold=36.84%
• Assessment: Step change from recent stable level (~26 ns) to new level (~40.5 ns). The test has bimodal history (previously transited ~24→40→26 ns), and this regression returns it to the higher state. All 28 post-regression runs are tightly clustered at 40.4-40.5 ns, indicating a deterministic change in generated code.
• Finding Confidence: 4/5 - Good data, clear step change from recent stable baseline despite bimodal overall history
• Regression Confidence: 4/5 - Strong signal (z=2.58, clearly deterministic change), test name directly matches JIT bounds-check optimization area
• Likely Cause: 4c86aad28c63 - "JIT: eliminate redundant bounds checks for span.Slice(span.Length - cns) (JIT: eliminate redundant bounds checks for span.Slice(span.Length - cns) #127488)"
  • Affected Areas: JIT (range check elimination, value numbering)
  • Rationale: This commit directly modifies rangecheck.cpp and valuenum.cpp - the exact JIT subsystems responsible for bounds check elimination and hoisting. The test name "IndexerCheckBoundCheckHoist" explicitly tests bounds check hoisting behavior. The commit adds new range check folding logic for ADD(A, K) op A patterns. If this new logic interferes with the existing bounds check hoisting pattern used by this test (e.g., by satisfying the range check differently, preventing the hoist), it could cause the regression from 26 ns back to the 40 ns state where the check is not hoisted. The test's bimodal history (~26 ns vs ~40 ns) suggests it has toggled between hoisted/non-hoisted states before.
  • Candidates Considered:
    • 4c86aad28c63 - Directly modifies bounds check elimination (most likely)
    • a59fcc58eb05 - JIT arithmetic transforms, adds GT_NOT case in asserprop
• Verification Status: Not verified

Borderline Cases

System.Collections.Perf_Frozen.ToFrozenDictionary(Count: 64) (+5.97%)

• Delta within ~1% of adaptive threshold (6.94%), z=1.47 below significance. However, all 28 post-runs are tightly clustered at 1666-1716 vs pre-mean of 1601, suggesting a real step change in a historically stable test (CV=3.5%). Warrants monitoring.

System.Linq.Tests.Perf_Enumerable.All_AllElementsMatch(IEnumerable) (+24.79%)

• High adaptive threshold (30.41%) due to multi-regime history (55→44→~56 ns). Despite z=0.81, all 28 post-runs are uniformly 56.38-56.60 (near-zero within-period variance), which is inconsistent with noise. The test appears to have reverted to a previous regime.

Benchstone.MDBenchI.MDAddArray2.Test (+9.69%)

• Marginal (delta +9.69% vs threshold 8.64%), but z=1.17 well below 2. Historical drift pattern: the before-series dropped from ~11.1M to ~10.2M; compare at ~11.1M reverts to early historical levels. Insufficient confidence to confirm regression.

Open Questions

• Confidence Limitations: Multiple JIT commits (4+) in the range make precise single-commit attribution difficult without verification
• Verification Needed: Bisection recommended for both confirmed regressions to isolate exact culprit commits. Run RunWithBuildAtHash at commits 4c86aad28c63 and a59fcc58eb05 vs baseline aba46e33ea5d for NBody_3 and IndexerCheckBoundCheckHoist
• Bisection Status: Not performed
• Borderline tests: The Frozen, Linq.All, and MDAddArray2 borderline cases show tight post-regression clustering inconsistent with noise - further monitoring recommended

Recommended Actions

1. Bisection for IndexerCheckBoundCheckHoist: Run bisection on commits aba46e33ea5d...977f4125aa7c using RunWithBuildAtHash to isolate culprit, focusing on 4c86aad28c63 first
2. Bisection for NBody_3.RunBench: Verify regression with RunWithBuildAtHash on suspected commits a59fcc58eb05 and 4c86aad28c63 vs baseline
3. Source review: Manual code review of 4c86aad28c63 focusing on interaction between new ADD(A, K) op A range check folding and existing bounds check hoisting patterns
4. Monitor borderline tests: Track Perf_Frozen.ToFrozenDictionary and Perf_Enumerable.All_AllElementsMatch over next builds for persistence

Bisection Results

Test	Status	Culprit Commit	Relevance	Perf Diff	Duration
Pow_ScalarExponent(BufferLength: 3079)	❌ failed	-	-	15.0→16.8 (+12%)	194s
Pow_ScalarExponent(BufferLength: 128)	✅ Success	0b6bb45eef86	✅ RELATED	627.6→696.4 (+11%)	252s
ToFrozenDictionary(Count: 64)	❌ failed	-	-	1.6→1.7 (+6%)	152s
All_AllElementsMatch(input: IEnumerable)	✅ Success	0b6bb45eef86	-	45.3→56.5 ns (+25%)	265s
RunBench	✅ Success	0b6bb45eef86	-	316.2→333.5 ns (+5%)	168s
Test	✅ Success	0b6bb45eef86	✅ RELATED	10.1→11.1 µs (+10%)	192s
IndexerCheckBoundCheckHoist	❌ failed	-	-	26.0→40.5 (+55%)	124s

Total bisection time: 1347s
Success rate: 4/7 tests

System.Numerics.Tensors.Tests.Perf_FloatingPointTensorPrimitives.Pow_ScalarExponent(BufferLength: 3079)

• Status: failed

Bisection details

iewing what files it touched and confirming its relevance:

Based on my analysis:

BISECTION RESULT:

• Culprit Commit: 0b6bb45
• Commit Message: Remove unwind calls from patchpoint OSR transition (Remove unwind calls from patchpoint OSR transition #127213)
• Commits Tested: 2 builds explicitly (baseline and compare), narrowed range from 38 to 12 commits
• Confidence: 4 out of 5
• Commit Relevance: RELATED
• Relevance Explanation: This commit modifies the JIT compiler's OSR (On-Stack Replacement) mechanism, which affects how the runtime optimizes and transitions between code tiers during loop execution. The test performs power operations in a tight loop over 3079 elements, making it highly sensitive to JIT loop optimization changes. The commit touches core JIT codegen files across all architectures (codegenxarch.cpp, lsraxarch.cpp, patchpoint.cpp, jithelpers.cpp) and changes how OSR transitions are handled, which could directly impact the generated code for this tensor benchmark.
• Summary: Using performance database queries, I identified the regression occurred between commits 6130a50 (good: 15.02ns) and 5391e9e (bad: 16.69ns) on April 28, 2026. Within this 12-commit window, commit 0b6bb45 stands out as the most likely culprit - it's a substantial JIT change (246 additions, 139 deletions across 23 files) that modifies OSR transition behavior. The 11.5% regression magnitude and the test's loop-intensive nature strongly correlate with OSR-related changes.

System.Numerics.Tensors.Tests.Perf_FloatingPointTensorPrimitives.Pow_ScalarExponent(BufferLength: 128)

• Culprit: 0b6bb45eef86
• Commit Message: Remove unwind calls from patchpoint OSR transition (Remove unwind calls from patchpoint OSR transition #127213)
• Relevance: RELATED
• Explanation: This commit modifies JIT codegen for OSR (On-Stack Replacement) transitions in tiered compilation, changing 23 files with 385 lines across all architecture backends. The regressed test runs with tiered compilation enabled and executes a tight loop performing tensor power operations, which would trigger OSR transitions during warmup. Changes to OSR patchpoint handling directly affect how code transitions between tier 0 and tier 1, providing a plausible mechanism for the 11% performance regression.

System.Collections.Perf_Frozen.ToFrozenDictionary(Count: 64)

• Status: failed

Bisection details

le to run Helix bisection due to authorization constraints. Used historical performance data to narrow the 38-commit range to 12 commits between 6130a50 (good, ~1591ns) and 5391e9e (bad, ~1682ns). Analysis of commit metadata identifies the OSR patchpoint commit as the most likely culprit due to its extensive JIT codegen modifications, though functional relevance to frozen collections is weak.

ALIGNMENT WARNING:
This bisection found a commit that does not appear related to the regressed test.
This suggests the performance change may be due to code alignment, binary layout,
or other non-functional factors rather than a regression in the tested code path.

Additional Context:

• The other commits in the range touch: trimming infrastructure (ILLink), HTTP headers, interpreter debugging, CRC vectorization, integer serialization, ARM64 signal handling, Haiku platform support, and DAC/DBI debugging APIs
• None of these directly interact with System.Collections.Frozen or the ToFrozenDictionary implementation
• The OSR commit adds 246 lines across 23 files in JIT codegen, which is substantial enough to alter binary layout and instruction alignment
• The regression magnitude (+6%) is consistent with alignment artifacts rather than algorithmic changes

Recommendation: This appears to be a code alignment regression rather than a functional bug. Consider monitoring if the regression persists or if future commits naturally resolve it through further code layout changes.

System.Linq.Tests.Perf_Enumerable.All_AllElementsMatch(input: IEnumerable)

• Culprit: 0b6bb45eef86
• Commit Message: s:
• Relevance: Unknown

BenchmarksGame.NBody_3.RunBench

• Culprit: 0b6bb45eef86
• Commit Message: Remove unwind calls from patchpoint OSR transition (Remove unwind calls from patchpoint OSR transition #127213)
• Relevance: Unknown
• Explanation: This commit modifies the JIT compiler's OSR (On-Stack Replacement) patchpoint logic across 23 files, directly affecting tiered compilation behavior. The NBody_3 benchmark is a computational micro-benchmark that runs 5 million iterations with tiered compilation enabled. OSR changes impact how hot loops transition from tier1 to optimized tier2 code, affecting register allocation (LSRA modifications), code layout, and instruction generation. This is the only JIT-related commit in the 12-commit regression range where the performance dropped from 309.6ns to 333.5ns (+7.7%). The changes to patchpoint transitions and LSRA logic provide a plausible mechanism for the observed regression in tight floating-point loops.

Benchstone.MDBenchI.MDAddArray2.Test

• Culprit: 0b6bb45eef86
• Commit Message: Remove unwind calls from patchpoint OSR transition (Remove unwind calls from patchpoint OSR transition #127213)
• Relevance: RELATED
• Explanation: This commit modifies JIT codegen for On-Stack Replacement (OSR) patchpoint transitions. The MDAddArray2 test contains triple-nested loops (200×200×200 iterations) that are prime candidates for OSR optimization. The commit changes how the JIT handles OSR transitions by removing unwind calls, which could affect loop performance. The timing perfectly aligns: last good build at 13:30, commit merged at 18:16, first regressed build at 21:27.

System.Tests.Perf_String.IndexerCheckBoundCheckHoist

• Status: failed

Bisection details

hpoint OSR transition (#127213)

• Commits Tested: 4 (via historical data: baseline, intermediate good, first bad, final bad)
• Confidence: 5
• Commit Relevance: RELATED
• Relevance Explanation: This commit modifies JIT code generation extensively, specifically OSR (On-Stack Replacement) patchpoint handling across all architectures. The regressed test IndexerCheckBoundCheckHoist measures bound check hoisting in a tight loop - a scenario where OSR transitions are relevant. Changes to patchpoint infrastructure can affect code layout, register allocation, and optimization decisions in hot loops, directly impacting the bound check hoisting optimization the test measures. The commit touches 23 JIT files including codegen, liveness, lsra (register allocation), and lower phases, all of which can influence loop optimization decisions.
• Summary: Used historical performance data from the database to narrow the regression window from 38 commits to 12 commits between 6130a50 (good) and 5391e9e (bad). Analysis of commit summaries identified commit 0b6bb45 as the only one modifying JIT infrastructure. The commit's extensive changes to OSR patchpoint handling are functionally related to the test's loop-based bound check hoisting scenario.

The regression is real and the commit is functionally related to the test area. The OSR patchpoint changes likely altered code generation in the test's tight loop, affecting the JIT's ability to hoist bound checks efficiently.

---

Generated by PerfTriageAgent on 2026-05-05 11:00 UTC

[DrewScoggins]

Bisection found this change as the culprit. #127213

[dotnet-policy-service]

Tagging subscribers to this area: @JulieLeeMSFT, @jakobbotsch
See info in area-owners.md if you want to be subscribed.