[Perf] Linux/x64: 9 Regressions on 7/9/2024 1:59:56 PM

[performanceautofiler]

Run Information

Name	Value
Architecture	x64
OS	ubuntu 22.04
Queue	ViperUbuntu
Baseline	13cf2649ad245ede1ed6e790079955db4c2fea54
Compare	5795e8c056c14741b80dfa304a974f5c0d72432a
Diff	Diff
Configs	CompilationMode:tiered, RunKind:micro

Regressions in System.Numerics.Tensors.Tests.Perf_NumberTensorPrimitives<Single>

Benchmark	Baseline	Test	Test/Base	Test Quality	Edge Detector	Baseline IR	Compare IR	IR Ratio
Max_Vector - Duration of single invocation 📝 - Benchmark Source ADX - Test Multi Config Graph	381.39 ns	574.28 ns	1.51	0.44	False
Abs - Duration of single invocation 📝 - Benchmark Source ADX - Test Multi Config Graph	7.41 ns	10.37 ns	1.40	0.37	False
Max_Scalar - Duration of single invocation 📝 - Benchmark Source ADX - Test Multi Config Graph	12.42 ns	16.79 ns	1.35	0.19	False
Max - Duration of single invocation 📝 - Benchmark Source ADX - Test Multi Config Graph	507.13 ns	656.26 ns	1.29	0.05	True
MaxMagnitude_Scalar - Duration of single invocation 📝 - Benchmark Source ADX - Test Multi Config Graph	15.59 ns	18.97 ns	1.22	0.12	True
MaxMagnitude_Vector - Duration of single invocation 📝 - Benchmark Source ADX - Test Multi Config Graph	19.04 ns	20.28 ns	1.07	0.12	False
BitwiseAnd_Scalar - Duration of single invocation 📝 - Benchmark Source ADX - Test Multi Config Graph	8.34 ns	10.24 ns	1.23	0.32	False
Max - Duration of single invocation 📝 - Benchmark Source ADX - Test Multi Config Graph	23.75 ns	32.94 ns	1.39	0.04	False
Max_Vector - Duration of single invocation 📝 - Benchmark Source ADX - Test Multi Config Graph	15.46 ns	18.24 ns	1.18	0.22	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.Numerics.Tensors.Tests.Perf_NumberTensorPrimitives<Single>*'

System.Numerics.Tensors.Tests.Perf_NumberTensorPrimitives<Single>.Max_Vector(BufferLength: 3079)

ETL Files

Histogram

JIT Disasms

System.Numerics.Tensors.Tests.Perf_NumberTensorPrimitives<Single>.Abs(BufferLength: 128)

ETL Files

Histogram

JIT Disasms

System.Numerics.Tensors.Tests.Perf_NumberTensorPrimitives<Single>.Max_Scalar(BufferLength: 128)

ETL Files

Histogram

JIT Disasms

System.Numerics.Tensors.Tests.Perf_NumberTensorPrimitives<Single>.Max(BufferLength: 3079)

ETL Files

Histogram

JIT Disasms

System.Numerics.Tensors.Tests.Perf_NumberTensorPrimitives<Single>.MaxMagnitude_Scalar(BufferLength: 128)

ETL Files

Histogram

JIT Disasms

System.Numerics.Tensors.Tests.Perf_NumberTensorPrimitives<Single>.MaxMagnitude_Vector(BufferLength: 128)

ETL Files

Histogram

JIT Disasms

System.Numerics.Tensors.Tests.Perf_NumberTensorPrimitives<Single>.BitwiseAnd_Scalar(BufferLength: 128)

ETL Files

Histogram

JIT Disasms

System.Numerics.Tensors.Tests.Perf_NumberTensorPrimitives<Single>.Max(BufferLength: 128)

ETL Files

Histogram

JIT Disasms

System.Numerics.Tensors.Tests.Perf_NumberTensorPrimitives<Single>.Max_Vector(BufferLength: 128)

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-numerics
See info in area-owners.md if you want to be subscribed.

[DrewScoggins]

Looks like this pr is related #103837

[DrewScoggins]

Additional Regressions:

• [Perf] Linux/x64: 2 Regressions on 7/10/2024 12:47:13 AM perf-autofiling-issues#38324
• [Perf] Linux/x64: 9 Regressions on 7/9/2024 12:32:58 AM perf-autofiling-issues#38322
• [Perf] Windows/x64: 20 Regressions on 7/10/2024 12:47:13 AM perf-autofiling-issues#38301
• [Perf] Windows/x64: 3 Regressions on 7/10/2024 12:47:13 AM perf-autofiling-issues#38295
• [Perf] Linux/x64: 6 Regressions on 7/10/2024 12:47:13 AM perf-autofiling-issues#38272

[tannergooding]

This is expected, the behavior of Max was normalized across all hardware. MaxNative is the API to use if you don't care about xplat behavioral differences and want to keep the perf