[Perf] Linux/arm64: System.Text.RegularExpressions.Tests.Perf_Regex_Industry_RustLang_Sherlock Regressions on 12/6/2024 6:16:01 PM

[performanceautofiler]

Run Information

Name	Value
Architecture	arm64
OS	ubuntu 22.04
Queue	AmpereUbuntu
Baseline	2f8151352a759e682de49dab04cb5081185f2454
Compare	2d81cbb28a8feb1af5b790664b511b264d032877
Diff	Diff
Configs	CompilationMode:tiered, RunKind:micro

Regressions in System.Text.RegularExpressions.Tests.Perf_Regex_Industry_RustLang_Sherlock

Benchmark	Baseline	Test	Test/Base	Test Quality	Edge Detector	Baseline IR	Compare IR	IR Ratio
Count - Duration of single invocation 📝 - Benchmark Source ADX - Test Multi Config Graph	207.97 μs	219.74 μs	1.06	0.01	False
Count - Duration of single invocation 📝 - Benchmark Source ADX - Test Multi Config Graph	2.32 ms	2.48 ms	1.07	0.02	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
python3 .\performance\scripts\benchmarks_ci.py -f net8.0 --filter 'System.Text.RegularExpressions.Tests.Perf_Regex_Industry_RustLang_Sherlock*'

System.Text.RegularExpressions.Tests.Perf_Regex_Industry_RustLang_Sherlock.Count(Pattern: "Sherlock|Holmes|Watson", Options: Compiled)

ETL Files

Histogram

JIT Disasms

System.Text.RegularExpressions.Tests.Perf_Regex_Industry_RustLang_Sherlock.Count(Pattern: "\p{Lu}", Options: None)

ETL Files

Histogram

JIT Disasms

Docs

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

[dotnet-policy-service]

Tagging subscribers to this area: @dotnet/area-system-text-regularexpressions
See info in area-owners.md if you want to be subscribed.

[LoopedBard3]

Most likely #109283 from range 8fca0a1...c7f4149. FYI @stephentoub.

[stephentoub]

Most likely #109283 from range 8fca0a1...c7f4149. FYI @stephentoub.

It'd be odd for this to have any meaningful impact on throughput. It's effectively adding a null check as part of compilation, but it should have zero impact on the generated IL for the dynamic method.

[steveharter]

@LoopedBard3 can you provide a new historical run of this benchmark to see if it stabilized or not?

[LoopedBard3]

The reports should be updated, one way to access them would be to use the Benchmark links starting with "Count - " and clicking the graph in the difference column on the linked page.
Otherwise, here are the updated graphs up through the 30th:

It seems both are still displaying the bimodal behavior.