[Perf] Linux/x64: 1 Regression on 9/6/2024 10:52:37 PM

[performanceautofiler]

Run Information

Name	Value
Architecture	x64
OS	ubuntu 22.04
Queue	TigerUbuntu
Baseline	d4601d390b6ca9177d62a5ac29b74d30cc16d63d
Compare	5c4686f831d34c2c127e943d0f0d144793eeb0ad
Diff	Diff
Configs	CompilationMode:tiered, RunKind:micro

Regressions in System.Text.Tests.Perf_StringBuilder

Benchmark	Baseline	Test	Test/Base	Test Quality	Edge Detector	Baseline IR	Compare IR	IR Ratio
ctor_capacity - Duration of single invocation 📝 - Benchmark Source ADX - Test Multi Config Graph	19.53 ns	21.61 ns	1.11	0.04	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.Tests.Perf_StringBuilder*'

System.Text.Tests.Perf_StringBuilder.ctor_capacity(length: 100)

ETL Files

Histogram

JIT Disasms

Docs

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

[DrewScoggins]

Could be #107218

[AaronRobinsonMSFT]

Yep. A 2 nanosecond regression here is acceptable.

[stephentoub]

What is the cause / connection with #107218 ?

[AaronRobinsonMSFT]

We converted the GC.AllocateUninitializedArray<T>() to use a QCall in some instances. It is possible there will be a small regression in some cases. Two ns seems aboput right, but I could see it being "slightly" more in some cases.