[Perf] Linux/arm64: 55 Improvements on 5/12/2026 11:35:36 PM +00:00

[performanceautofiler]

Run Information

Name	Value
Architecture	arm64
OS	azurelinux 3.0
Queue	CobaltAzureLinux
Baseline	6820c801e8b84f3fc9f96ecb6e0838b829345951
Compare	71df7dbecba25333e2bc10872a09e6ab32624933
Diff	Diff
Configs	CompilationMode:tiered, R2RType:r2r_interpreter, RunKind:micro

Improvements in System.Net.Http.Tests.SocketsHttpHandlerPerfTest

Benchmark	Baseline	Test	Test/Base	Test Quality	Edge Detector	Baseline IR	Compare IR	IR Ratio
Get - Duration of single invocation 📝 - Benchmark Source ADX - Test Multi Config Graph	404.61 μs	216.61 μs	0.54	0.37	False
Get_EnumerateHeaders_Unvalidated - Duration of single invocation ADX - Test Multi Config Graph	381.73 μs	214.39 μs	0.56	0.43	False
Get - Duration of single invocation 📝 - Benchmark Source ADX - Test Multi Config Graph	390.39 μs	219.85 μs	0.56	0.40	False
Get_EnumerateHeaders_Unvalidated - Duration of single invocation ADX - Test Multi Config Graph	378.80 μs	213.41 μs	0.56	0.45	False
Get_EnumerateHeaders_Validated - Duration of single invocation 📝 - Benchmark Source ADX - Test Multi Config Graph	405.51 μs	229.44 μs	0.57	0.37	False
Get - Duration of single invocation 📝 - Benchmark Source ADX - Test Multi Config Graph	198.24 μs	118.78 μs	0.60	0.47	False
Get_EnumerateHeaders_Validated - Duration of single invocation 📝 - Benchmark Source ADX - Test Multi Config Graph	389.55 μs	241.49 μs	0.62	0.41	False
Get_EnumerateHeaders_Validated - Duration of single invocation 📝 - Benchmark Source ADX - Test Multi Config Graph	212.90 μs	132.05 μs	0.62	0.41	False
Get_EnumerateHeaders_Validated - Duration of single invocation 📝 - Benchmark Source ADX - Test Multi Config Graph	212.94 μs	135.79 μs	0.64	0.44	False
Get_EnumerateHeaders_Unvalidated - Duration of single invocation ADX - Test Multi Config Graph	203.64 μs	132.07 μs	0.65	0.40	False
Get_EnumerateHeaders_Unvalidated - Duration of single invocation ADX - Test Multi Config Graph	197.07 μs	131.14 μs	0.67	0.38	False
Get_EnumerateHeaders_Unvalidated - Duration of single invocation ADX - Test Multi Config Graph	311.92 μs	226.38 μs	0.73	0.38	False
Get - Duration of single invocation 📝 - Benchmark Source ADX - Test Multi Config Graph	173.49 μs	128.89 μs	0.74	0.49	False
Get - Duration of single invocation 📝 - Benchmark Source ADX - Test Multi Config Graph	348.59 μs	259.47 μs	0.74	0.35	False
Get_EnumerateHeaders_Validated - Duration of single invocation 📝 - Benchmark Source ADX - Test Multi Config Graph	307.43 μs	230.94 μs	0.75	0.28	False
Get - Duration of single invocation 📝 - Benchmark Source ADX - Test Multi Config Graph	293.63 μs	225.65 μs	0.77	0.37	False
Get_EnumerateHeaders_Validated - Duration of single invocation 📝 - Benchmark Source ADX - Test Multi Config Graph	352.47 μs	286.74 μs	0.81	0.38	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.Net.Http.Tests.SocketsHttpHandlerPerfTest*'

Details

System.Net.Http.Tests.SocketsHttpHandlerPerfTest.Get(ssl: True, chunkedResponse: False, responseLength: 1)

ETL Files

Histogram

JIT Disasms

System.Net.Http.Tests.SocketsHttpHandlerPerfTest.Get_EnumerateHeaders_Unvalidated(ssl: True, chunkedResponse: False, responseLength: 1)

ETL Files

Histogram

JIT Disasms

System.Net.Http.Tests.SocketsHttpHandlerPerfTest.Get(ssl: True, chunkedResponse: True, responseLength: 1)

ETL Files

Histogram

JIT Disasms

System.Net.Http.Tests.SocketsHttpHandlerPerfTest.Get_EnumerateHeaders_Unvalidated(ssl: True, chunkedResponse: True, responseLength: 1)

ETL Files

Histogram

JIT Disasms

System.Net.Http.Tests.SocketsHttpHandlerPerfTest.Get_EnumerateHeaders_Validated(ssl: True, chunkedResponse: True, responseLength: 1)

ETL Files

Histogram

JIT Disasms

System.Net.Http.Tests.SocketsHttpHandlerPerfTest.Get(ssl: False, chunkedResponse: False, responseLength: 1)

ETL Files

Histogram

JIT Disasms

System.Net.Http.Tests.SocketsHttpHandlerPerfTest.Get_EnumerateHeaders_Validated(ssl: True, chunkedResponse: False, responseLength: 1)

ETL Files

Histogram

JIT Disasms

System.Net.Http.Tests.SocketsHttpHandlerPerfTest.Get_EnumerateHeaders_Validated(ssl: False, chunkedResponse: True, responseLength: 1)

ETL Files

Histogram

JIT Disasms

System.Net.Http.Tests.SocketsHttpHandlerPerfTest.Get_EnumerateHeaders_Validated(ssl: False, chunkedResponse: False, responseLength: 1)

ETL Files

Histogram

JIT Disasms

System.Net.Http.Tests.SocketsHttpHandlerPerfTest.Get_EnumerateHeaders_Unvalidated(ssl: False, chunkedResponse: False, responseLength: 1)

ETL Files

Histogram

JIT Disasms

System.Net.Http.Tests.SocketsHttpHandlerPerfTest.Get_EnumerateHeaders_Unvalidated(ssl: False, chunkedResponse: True, responseLength: 1)

ETL Files

Histogram

JIT Disasms

System.Net.Http.Tests.SocketsHttpHandlerPerfTest.Get_EnumerateHeaders_Unvalidated(ssl: False, chunkedResponse: False, responseLength: 100000)

ETL Files

Histogram

JIT Disasms

System.Net.Http.Tests.SocketsHttpHandlerPerfTest.Get(ssl: False, chunkedResponse: True, responseLength: 1)

ETL Files

Histogram

JIT Disasms

System.Net.Http.Tests.SocketsHttpHandlerPerfTest.Get(ssl: False, chunkedResponse: True, responseLength: 100000)

ETL Files

Histogram

JIT Disasms

System.Net.Http.Tests.SocketsHttpHandlerPerfTest.Get_EnumerateHeaders_Validated(ssl: False, chunkedResponse: False, responseLength: 100000)

ETL Files

Histogram

JIT Disasms

System.Net.Http.Tests.SocketsHttpHandlerPerfTest.Get(ssl: False, chunkedResponse: False, responseLength: 100000)

ETL Files

Histogram

JIT Disasms

System.Net.Http.Tests.SocketsHttpHandlerPerfTest.Get_EnumerateHeaders_Validated(ssl: False, chunkedResponse: True, responseLength: 100000)

ETL Files

Histogram

JIT Disasms

Docs

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

---

Run Information

Name	Value
Architecture	arm64
OS	azurelinux 3.0
Queue	CobaltAzureLinux
Baseline	6820c801e8b84f3fc9f96ecb6e0838b829345951
Compare	52090b9f07a14ec5db3a19b379548f469fefcc27
Diff	Diff
Configs	CompilationMode:tiered, R2RType:r2r_interpreter, RunKind:micro

Improvements in System.IO.Tests.StreamReaderReadLineTests

Benchmark	Baseline	Test	Test/Base	Test Quality	Edge Detector	Baseline IR	Compare IR	IR Ratio
ReadLineAsync - Duration of single invocation 📝 - Benchmark Source ADX - Test Multi Config Graph	3771.97 μs	239.01 μs	0.06	0.16	False
ReadLineAsync - Duration of single invocation 📝 - Benchmark Source ADX - Test Multi Config Graph	18.48 ms	1.50 ms	0.08	0.18	False
ReadLineAsync - Duration of single invocation 📝 - Benchmark Source ADX - Test Multi Config Graph	9231.82 μs	878.47 μs	0.10	0.15	False
ReadLineAsync - Duration of single invocation 📝 - Benchmark Source ADX - Test Multi Config Graph	928.26 μs	103.27 μs	0.11	0.16	False
ReadLineAsync - Duration of single invocation 📝 - Benchmark Source ADX - Test Multi Config Graph	288.20 μs	40.71 μs	0.14	0.19	False
ReadLineAsync - Duration of single invocation 📝 - Benchmark Source ADX - Test Multi Config Graph	88.46 μs	20.17 μs	0.23	0.19	False
ReadLineAsync - Duration of single invocation 📝 - Benchmark Source ADX - Test Multi Config Graph	79.31 μs	19.38 μs	0.24	0.08	False
ReadLineAsync - Duration of single invocation 📝 - Benchmark Source ADX - Test Multi Config Graph	60.47 μs	17.98 μs	0.30	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.IO.Tests.StreamReaderReadLineTests*'

Details

System.IO.Tests.StreamReaderReadLineTests.ReadLineAsync(LineLengthRange: [ 1, 8])

ETL Files

Histogram

JIT Disasms

System.IO.Tests.StreamReaderReadLineTests.ReadLineAsync(LineLengthRange: [ 0, 0])

ETL Files

Histogram

JIT Disasms

System.IO.Tests.StreamReaderReadLineTests.ReadLineAsync(LineLengthRange: [ 1, 1])

ETL Files

Histogram

JIT Disasms

System.IO.Tests.StreamReaderReadLineTests.ReadLineAsync(LineLengthRange: [ 9, 32])

ETL Files

Histogram

JIT Disasms

System.IO.Tests.StreamReaderReadLineTests.ReadLineAsync(LineLengthRange: [ 33, 128])

ETL Files

Histogram

JIT Disasms

System.IO.Tests.StreamReaderReadLineTests.ReadLineAsync(LineLengthRange: [ 0, 1024])

ETL Files

Histogram

JIT Disasms

System.IO.Tests.StreamReaderReadLineTests.ReadLineAsync(LineLengthRange: [ 129, 1024])

ETL Files

Histogram

JIT Disasms

System.IO.Tests.StreamReaderReadLineTests.ReadLineAsync(LineLengthRange: [1025, 2048])

ETL Files

Histogram

JIT Disasms

Docs

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

---

Run Information

Name	Value
Architecture	arm64
OS	azurelinux 3.0
Queue	CobaltAzureLinux
Baseline	6820c801e8b84f3fc9f96ecb6e0838b829345951
Compare	52090b9f07a14ec5db3a19b379548f469fefcc27
Diff	Diff
Configs	CompilationMode:tiered, R2RType:r2r_interpreter, RunKind:micro

Improvements in System.Net.Security.Tests.SslStreamTests

Benchmark	Baseline	Test	Test/Base	Test Quality	Edge Detector	Baseline IR	Compare IR	IR Ratio
ReadWriteAsync - Duration of single invocation 📝 - Benchmark Source ADX - Test Multi Config Graph	102.89 μs	56.46 μs	0.55	0.39	False
ConcurrentReadWrite - Duration of single invocation 📝 - Benchmark Source ADX - Test Multi Config Graph	81.25 μs	48.86 μs	0.60	0.51	False
ConcurrentReadWriteLargeBuffer - Duration of single invocation 📝 - Benchmark Source ADX - Test Multi Config Graph	87.88 μs	56.97 μs	0.65	0.38	False
WriteReadAsync - Duration of single invocation 📝 - Benchmark Source ADX - Test Multi Config Graph	63.97 μs	42.11 μs	0.66	0.31	False
LargeReadWriteAsync - Duration of single invocation 📝 - Benchmark Source ADX - Test Multi Config Graph	168.22 μs	113.64 μs	0.68	0.42	False
LargeWriteReadAsync - Duration of single invocation 📝 - Benchmark Source ADX - Test Multi Config Graph	70.89 μs	50.77 μs	0.72	0.33	False
DefaultHandshakeContextIPv6Async - Duration of single invocation 📝 - Benchmark Source ADX - Test Multi Config Graph	3.38 ms	2.87 ms	0.85	0.18	False
DefaultHandshakeIPv4Async - Duration of single invocation 📝 - Benchmark Source ADX - Test Multi Config Graph	3.94 ms	3.68 ms	0.94	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.Net.Security.Tests.SslStreamTests*'

Details

System.Net.Security.Tests.SslStreamTests.ReadWriteAsync

ETL Files

Histogram

JIT Disasms

System.Net.Security.Tests.SslStreamTests.ConcurrentReadWrite

ETL Files

Histogram

JIT Disasms

System.Net.Security.Tests.SslStreamTests.ConcurrentReadWriteLargeBuffer

ETL Files

Histogram

JIT Disasms

System.Net.Security.Tests.SslStreamTests.WriteReadAsync

ETL Files

Histogram

JIT Disasms

System.Net.Security.Tests.SslStreamTests.LargeReadWriteAsync

ETL Files

Histogram

JIT Disasms

System.Net.Security.Tests.SslStreamTests.LargeWriteReadAsync

ETL Files

Histogram

JIT Disasms

System.Net.Security.Tests.SslStreamTests.DefaultHandshakeContextIPv6Async

ETL Files

Histogram

JIT Disasms

System.Net.Security.Tests.SslStreamTests.DefaultHandshakeIPv4Async

ETL Files

Histogram

JIT Disasms

Docs

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

[performanceautofiler]

Run Information

Name	Value
Architecture	arm64
OS	azurelinux 3.0
Queue	CobaltAzureLinux
Baseline	6820c801e8b84f3fc9f96ecb6e0838b829345951
Compare	52090b9f07a14ec5db3a19b379548f469fefcc27
Diff	Diff
Configs	CompilationMode:tiered, R2RType:r2r_interpreter, RunKind:micro

Improvements in System.IO.Tests.StreamReaderReadToEndTests

Benchmark	Baseline	Test	Test/Base	Test Quality	Edge Detector	Baseline IR	Compare IR	IR Ratio
ReadToEndAsync - Duration of single invocation 📝 - Benchmark Source ADX - Test Multi Config Graph	92.81 ms	60.97 ms	0.66	0.13	False
ReadToEndAsync - Duration of single invocation 📝 - Benchmark Source ADX - Test Multi Config Graph	95.16 ms	62.62 ms	0.66	0.11	False
ReadToEndAsync - Duration of single invocation 📝 - Benchmark Source ADX - Test Multi Config Graph	93.82 ms	62.23 ms	0.66	0.14	False
ReadToEndAsync - Duration of single invocation 📝 - Benchmark Source ADX - Test Multi Config Graph	92.98 ms	61.75 ms	0.66	0.10	False
ReadToEndAsync - Duration of single invocation 📝 - Benchmark Source ADX - Test Multi Config Graph	93.21 ms	61.91 ms	0.66	0.06	False
ReadToEndAsync - Duration of single invocation 📝 - Benchmark Source ADX - Test Multi Config Graph	92.18 ms	62.46 ms	0.68	0.09	False
ReadToEndAsync - Duration of single invocation 📝 - Benchmark Source ADX - Test Multi Config Graph	92.58 ms	63.15 ms	0.68	0.11	False
ReadToEndAsync - Duration of single invocation 📝 - Benchmark Source ADX - Test Multi Config Graph	90.07 ms	62.47 ms	0.69	0.10	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.IO.Tests.StreamReaderReadToEndTests*'

Details

System.IO.Tests.StreamReaderReadToEndTests.ReadToEndAsync(LineLengthRange: [ 1, 1])

ETL Files

Histogram

JIT Disasms

System.IO.Tests.StreamReaderReadToEndTests.ReadToEndAsync(LineLengthRange: [ 0, 0])

ETL Files

Histogram

JIT Disasms

System.IO.Tests.StreamReaderReadToEndTests.ReadToEndAsync(LineLengthRange: [ 1, 8])

ETL Files

Histogram

JIT Disasms

System.IO.Tests.StreamReaderReadToEndTests.ReadToEndAsync(LineLengthRange: [ 0, 1024])

ETL Files

Histogram

JIT Disasms

System.IO.Tests.StreamReaderReadToEndTests.ReadToEndAsync(LineLengthRange: [1025, 2048])

ETL Files

Histogram

JIT Disasms

System.IO.Tests.StreamReaderReadToEndTests.ReadToEndAsync(LineLengthRange: [ 9, 32])

ETL Files

Histogram

JIT Disasms

System.IO.Tests.StreamReaderReadToEndTests.ReadToEndAsync(LineLengthRange: [ 129, 1024])

ETL Files

Histogram

JIT Disasms

System.IO.Tests.StreamReaderReadToEndTests.ReadToEndAsync(LineLengthRange: [ 33, 128])

ETL Files

Histogram

JIT Disasms

Docs

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

---

Run Information

Name	Value
Architecture	arm64
OS	azurelinux 3.0
Queue	CobaltAzureLinux
Baseline	6820c801e8b84f3fc9f96ecb6e0838b829345951
Compare	52090b9f07a14ec5db3a19b379548f469fefcc27
Diff	Diff
Configs	CompilationMode:tiered, R2RType:r2r_interpreter, RunKind:micro

Improvements in System.Formats.Tar.Tests.Perf_TarWriter

Benchmark	Baseline	Test	Test/Base	Test Quality	Edge Detector	Baseline IR	Compare IR	IR Ratio
PaxTarEntry_WriteEntry_Async - Duration of single invocation 📝 - Benchmark Source ADX - Test Multi Config Graph	12.75 μs	5.89 μs	0.46	0.17	False
GnuTarEntry_WriteEntry_Async - Duration of single invocation 📝 - Benchmark Source ADX - Test Multi Config Graph	3.42 μs	1.58 μs	0.46	0.13	False
V7TarEntry_WriteEntry_Async - Duration of single invocation 📝 - Benchmark Source ADX - Test Multi Config Graph	2.87 μs	1.47 μs	0.51	0.20	False
UstarTarEntry_WriteEntry_Async - Duration of single invocation 📝 - Benchmark Source ADX - Test Multi Config Graph	2.93 μs	1.52 μs	0.52	0.24	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.Formats.Tar.Tests.Perf_TarWriter*'

Details

System.Formats.Tar.Tests.Perf_TarWriter.PaxTarEntry_WriteEntry_Async

ETL Files

Histogram

JIT Disasms

System.Formats.Tar.Tests.Perf_TarWriter.GnuTarEntry_WriteEntry_Async

ETL Files

Histogram

JIT Disasms

System.Formats.Tar.Tests.Perf_TarWriter.V7TarEntry_WriteEntry_Async

ETL Files

Histogram

JIT Disasms

System.Formats.Tar.Tests.Perf_TarWriter.UstarTarEntry_WriteEntry_Async

ETL Files

Histogram

JIT Disasms

Docs

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

---

Run Information

Name	Value
Architecture	arm64
OS	azurelinux 3.0
Queue	CobaltAzureLinux
Baseline	6820c801e8b84f3fc9f96ecb6e0838b829345951
Compare	52090b9f07a14ec5db3a19b379548f469fefcc27
Diff	Diff
Configs	CompilationMode:tiered, R2RType:r2r_interpreter, RunKind:micro

Improvements in System.Threading.Tests.Perf_SemaphoreSlim

Benchmark	Baseline	Test	Test/Base	Test Quality	Edge Detector	Baseline IR	Compare IR	IR Ratio
ReleaseWaitAsync_WithCancellationToken - Duration of single invocation 📝 - Benchmark Source ADX - Test Multi Config Graph	7.37 μs	3.98 μs	0.54	0.55	False
ReleaseWaitAsync_WithCancellationTokenAndTimeout - Duration of single invocation 📝 - Benchmark Source ADX - Test Multi Config Graph	7.55 μs	4.63 μs	0.61	0.55	False
ReleaseWaitAsync_WithTimeout - Duration of single invocation 📝 - Benchmark Source ADX - Test Multi Config Graph	6.71 μs	4.53 μs	0.68	0.65	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.Threading.Tests.Perf_SemaphoreSlim*'

Details

System.Threading.Tests.Perf_SemaphoreSlim.ReleaseWaitAsync_WithCancellationToken

ETL Files

Histogram

JIT Disasms

System.Threading.Tests.Perf_SemaphoreSlim.ReleaseWaitAsync_WithCancellationTokenAndTimeout

ETL Files

Histogram

JIT Disasms

System.Threading.Tests.Perf_SemaphoreSlim.ReleaseWaitAsync_WithTimeout

ETL Files

Histogram

JIT Disasms

Docs

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

---

Run Information

Name	Value
Architecture	arm64
OS	azurelinux 3.0
Queue	CobaltAzureLinux
Baseline	6820c801e8b84f3fc9f96ecb6e0838b829345951
Compare	52090b9f07a14ec5db3a19b379548f469fefcc27
Diff	Diff
Configs	CompilationMode:tiered, R2RType:r2r_interpreter, RunKind:micro

Improvements in System.IO.Tests.Perf_File

Benchmark	Baseline	Test	Test/Base	Test Quality	Edge Detector	Baseline IR	Compare IR	IR Ratio
ReadAllLinesAsync - Duration of single invocation 📝 - Benchmark Source ADX - Test Multi Config Graph	311.66 μs	127.56 μs	0.41	0.34	False
AppendAllLinesAsync - Duration of single invocation 📝 - Benchmark Source ADX - Test Multi Config Graph	66.38 μs	49.70 μs	0.75	0.63	False
AppendAllTextAsync - Duration of single invocation 📝 - Benchmark Source ADX - Test Multi Config Graph	61.16 μs	49.21 μs	0.80	0.52	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.IO.Tests.Perf_File*'

Details

System.IO.Tests.Perf_File.ReadAllLinesAsync

ETL Files

Histogram

JIT Disasms

System.IO.Tests.Perf_File.AppendAllLinesAsync

ETL Files

Histogram

JIT Disasms

System.IO.Tests.Perf_File.AppendAllTextAsync(size: 10000)

ETL Files

Histogram

JIT Disasms

Docs

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

---

Run Information

Name	Value
Architecture	arm64
OS	azurelinux 3.0
Queue	CobaltAzureLinux
Baseline	6820c801e8b84f3fc9f96ecb6e0838b829345951
Compare	52090b9f07a14ec5db3a19b379548f469fefcc27
Diff	Diff
Configs	CompilationMode:tiered, R2RType:r2r_interpreter, RunKind:micro

Improvements in System.Numerics.Tensors.Tests.Perf_NumberTensorPrimitives<Int32>

Benchmark	Baseline	Test	Test/Base	Test Quality	Edge Detector	Baseline IR	Compare IR	IR Ratio
IndexOfMax - Duration of single invocation 📝 - Benchmark Source ADX - Test Multi Config Graph	71.34 ns	56.73 ns	0.80	0.08	False
IndexOfMax - Duration of single invocation 📝 - Benchmark Source ADX - Test Multi Config Graph	1.73 μs	1.47 μs	0.85	0.13	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_NumberTensorPrimitives<Int32>*'

Details

System.Numerics.Tensors.Tests.Perf_NumberTensorPrimitives<Int32>.IndexOfMax(BufferLength: 128)

ETL Files

Histogram

JIT Disasms

System.Numerics.Tensors.Tests.Perf_NumberTensorPrimitives<Int32>.IndexOfMax(BufferLength: 3079)

ETL Files

Histogram

JIT Disasms

Docs

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

---

Run Information

Name	Value
Architecture	arm64
OS	azurelinux 3.0
Queue	CobaltAzureLinux
Baseline	6820c801e8b84f3fc9f96ecb6e0838b829345951
Compare	52090b9f07a14ec5db3a19b379548f469fefcc27
Diff	Diff
Configs	CompilationMode:tiered, R2RType:r2r_interpreter, RunKind:micro

Improvements in System.Net.WebSockets.Tests.SocketSendReceivePerfTest

Benchmark	Baseline	Test	Test/Base	Test Quality	Edge Detector	Baseline IR	Compare IR	IR Ratio
ReceiveSend - Duration of single invocation 📝 - Benchmark Source ADX - Test Multi Config Graph	106.05 μs	79.94 μs	0.75	0.41	False
SendReceive - Duration of single invocation 📝 - Benchmark Source ADX - Test Multi Config Graph	32.22 μs	25.77 μs	0.80	0.52	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.Net.WebSockets.Tests.SocketSendReceivePerfTest*'

Details

System.Net.WebSockets.Tests.SocketSendReceivePerfTest.ReceiveSend

ETL Files

Histogram

JIT Disasms

System.Net.WebSockets.Tests.SocketSendReceivePerfTest.SendReceive

ETL Files

Histogram

JIT Disasms

Docs

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

[LoopedBard3]

🔍 Automated Triage Analysis

Summary: 55 tests reported as improvements, spanning 7 distinct areas. All tests are async operations running under the R2R interpreter configuration (runtime-async). The historical analysis classified 49 as Improvement, 4 as Regression (metric direction ambiguity — see note), 2 as Borderline, and 1 as an unknown metadata note. Bisection skipped — autofile run configuration is not supported by the bisection harness.

Finding Confidence: 4/5 (analysis accuracy)
Improvement Confidence: 5/5 (likelihood of true improvement)

📊 Full Analysis (click to expand)

Issue Overview

• Issue: [Perf] Linux/arm64: 55 Improvements on 5/12/2026 11:35:36 PM +00:00 #72977 - [Perf] Linux/arm64: 55 Improvements on 5/12/2026 11:35:36 PM +00:00
• Type: Performance Improvement
• Date: 2026-05-12 | Queue: AzureLinux.3.Cobalt.Arm64.Perf | OS/Arch: linux/arm64
• Commits: 6820c801e8b8 → 71df7dbecba2 (22 commits)
• Labels: untriaged, perf-improvement, os-linux, branch-refs/heads/main, arch-arm64, runtime-coreclr, kind-micro, compilationmode-tiered, r2rtype-r2r_interpreter, runkind-micro
• Historical Data: Available (95–97 runs per test, excellent statistical confidence)
• Run Configuration: Tiered compilation, R2R interpreter mode (runtime-async enabled)

Test Analysis

55 tests reported as improvements, spanning 7 distinct areas. All tests are async operations running under the R2R interpreter configuration (runtime-async). The historical analysis classified 49 as Improvement, 4 as Regression (metric direction ambiguity — see note), 2 as Borderline, and 1 as an unknown metadata note.

Note on "Regression" classifications: The historical analysis tool classified 4 SocketsHttpHandler tests as "Regression" due to unknown metric direction. Since these are latency benchmarks (lower = better) and the auto-filer reported them as improvements, they are genuine improvements with -23% to -26% deltas. They are treated as improvements in this analysis.

System.Net.Http — SocketsHttpHandler (17 tests)

Test Name	Baseline	Compare	Δ%	μ	σ	n	z	Thresh%	Noise?	Assessment
Get(ssl:T, chunked:F, len:1)	404.61	216.61	-46.5%	380.29	20534.6	97	-7.97	10.8%	N	Step change, very stable
Get_Headers_Unvalidated(ssl:T, chunked:F, len:1)	381.73	214.39	-43.8%	—	—	97	-9.17	—	N	Step change
Get(ssl:T, chunked:T, len:1)	390.39	219.85	-43.7%	—	—	97	-8.00	—	N	Step change
Get_Headers_Unvalidated(ssl:T, chunked:T, len:1)	378.80	213.41	-43.7%	—	—	97	-7.20	—	N	Step change
Get_Headers_Validated(ssl:T, chunked:T, len:1)	405.51	229.44	-43.4%	—	—	97	-8.28	—	N	Step change
Get(ssl:F, chunked:F, len:1)	198.24	118.78	-40.1%	—	—	97	-5.05	—	N	Step change
Get_Headers_Validated(ssl:T, chunked:F, len:1)	389.55	241.49	-38.0%	—	—	97	-7.48	—	N	Step change
Get_Headers_Validated(ssl:F, chunked:T, len:1)	212.90	132.05	-38.0%	—	—	97	-5.33	—	N	Step change
Get_Headers_Validated(ssl:F, chunked:F, len:1)	212.94	135.79	-36.2%	—	—	97	-5.27	—	N	Step change
Get_Headers_Unvalidated(ssl:F, chunked:F, len:1)	203.64	132.07	-35.1%	—	—	97	-4.82	—	N	Step change
Get_Headers_Unvalidated(ssl:F, chunked:T, len:1)	197.07	131.14	-33.5%	—	—	97	-5.41	—	N	Step change
Get_Headers_Unvalidated(ssl:F, chunked:F, len:100k)	311.92	226.38	-27.4%	—	—	97	-4.58	—	N	Step change
Get(ssl:F, chunked:T, len:1)	173.49	128.89	-25.7%	—	—	96	-4.72	—	N	Step change
Get(ssl:F, chunked:T, len:100k)	348.59	259.47	-25.6%	—	—	96	-5.85	—	N	Step change
Get_Headers_Validated(ssl:F, chunked:F, len:100k)	307.43	230.94	-24.9%	—	—	96	-4.67	—	N	Step change
Get(ssl:F, chunked:F, len:100k)	293.63	225.65	-23.1%	—	—	96	-3.73	—	N	Step change
Get_Headers_Validated(ssl:F, chunked:T, len:100k)	352.47	286.74	-18.6%	—	—	97	-3.35	—	N	Step change

System.IO — StreamReader ReadLineAsync (8 tests)

Test Name	Baseline	Compare	Δ%	n	z	Noise?	Assessment
ReadLineAsync([1, 1])	9231.82	878.47	-90.5%	97	-10.76	N	Massive step change
ReadLineAsync([1, 8])	3771.97	239.01	-93.7%	97	-11.70	N	Massive step change
ReadLineAsync([9, 32])	928.26	103.27	-88.9%	97	-10.09	N	Massive step change
ReadLineAsync([33, 128])	288.20	40.71	-85.9%	97	-9.50	N	Massive step change
ReadLineAsync([0, 0])	18.48	1.50	-91.9%	97	-12.67	N	Massive step change
ReadLineAsync([0, 1024])	88.46	20.17	-77.2%	97	-9.04	N	Massive step change
ReadLineAsync([129, 1024])	79.31	19.38	-75.6%	97	-11.33	N	Massive step change
ReadLineAsync([1025, 2048])	60.47	17.98	-70.3%	97	-10.63	N	Massive step change

System.IO — StreamReader ReadToEndAsync (8 tests)

Test Name	Baseline	Compare	Δ%	n	z	Noise?	Assessment
ReadToEndAsync([1, 1])	92.81	60.97	-34.3%	97	-5.93	N	Step change
ReadToEndAsync([0, 0])	95.16	62.62	-34.2%	97	-7.86	N	Step change
ReadToEndAsync([1, 8])	93.82	62.23	-33.7%	97	-6.91	N	Step change
ReadToEndAsync([0, 1024])	92.98	61.75	-33.6%	97	-7.48	N	Step change
ReadToEndAsync([1025, 2048])	93.21	61.91	-33.6%	97	-16.34	N	Step change
ReadToEndAsync([9, 32])	92.18	62.46	-32.2%	97	-7.23	N	Step change
ReadToEndAsync([129, 1024])	92.58	63.15	-31.8%	97	-9.15	N	Step change
ReadToEndAsync([33, 128])	90.07	62.47	-30.6%	97	-7.61	N	Step change

System.Net.Security — SslStream (8 tests)

Test Name	Baseline	Compare	Δ%	n	z	Noise?	Assessment
ReadWriteAsync	102.89	56.46	-45.1%	97	-5.39	N	Step change
ConcurrentReadWrite	81.25	48.86	-39.9%	97	-5.87	N	Step change
ConcurrentReadWriteLargeBuffer	87.88	56.97	-35.2%	97	-7.58	N	Step change
WriteReadAsync	63.97	42.11	-34.2%	97	-4.68	N	Step change
LargeReadWriteAsync	168.22	113.64	-32.5%	97	-5.93	N	Step change
LargeWriteReadAsync	70.89	50.77	-28.4%	97	-3.18	N	Step change
DefaultHandshakeContextIPv6Async	3.38	2.87	-15.1%	97	-2.80	N	Moderate step change
DefaultHandshakeIPv4Async	3.94	3.68	-6.6%	97	-2.45	N	Small improvement

System.Formats.Tar — TarWriter (4 tests)

Test Name	Baseline	Compare	Δ%	n	z	Noise?	Assessment
PaxTarEntry_WriteEntry_Async	12.75	5.89	-53.8%	97	-10.43	N	Step change
GnuTarEntry_WriteEntry_Async	3.42	1.58	-53.8%	97	-7.28	N	Step change
V7TarEntry_WriteEntry_Async	2.87	1.47	-48.8%	96	-4.54	N	Step change (high CV 10.5%)
UstarTarEntry_WriteEntry_Async	2.93	1.52	-48.1%	96	-4.28	N	Step change (high CV 10.7%)

System.Threading — SemaphoreSlim (3 tests)

Test Name	Baseline	Compare	Δ%	n	z	Noise?	Assessment
ReleaseWaitAsync_WithCancellationToken	7.37	3.98	-46.0%	97	-5.18	N	Step change
ReleaseWaitAsync_WithCancellationTokenAndTimeout	7.55	4.63	-38.7%	97	-3.29	N	Step change
ReleaseWaitAsync_WithTimeout	6.71	4.53	-32.5%	97	-3.12	N	Step change

System.IO — File operations (3 tests)

Test Name	Baseline	Compare	Δ%	n	z	Noise?	Assessment
ReadAllLinesAsync	311.66	127.56	-59.1%	97	-10.81	N	Step change
AppendAllLinesAsync	66.38	49.70	-25.1%	97	-3.07	N	Step change (high CV 11.1%)
AppendAllTextAsync(size: 10000)	61.16	49.21	-19.5%	97	-2.82	N	Step change

Other tests (4 tests)

Test Name	Baseline	Compare	Δ%	n	z	Noise?	Assessment
IndexOfMax(BufferLength: 128)	71.34	56.73	-20.5%	97	-3.48	N	Step change
IndexOfMax(BufferLength: 3079)	1.73	1.47	-15.0%	97	-1.98	Borderline	Near threshold
SocketSendReceivePerfTest.ReceiveSend	106.05	79.94	-24.6%	97	-6.41	N	Step change
SocketSendReceivePerfTest.SendReceive	32.22	25.77	-20.0%	97	-2.16	Borderline	Near threshold

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

Finding Confidence: 4/5 — Excellent historical data (95–97 runs per test), strong statistical power. All z-scores well beyond significance thresholds for non-borderline tests. Minor gap: metric unit reported as "unknown" for some test groups, but the direction (lower=better latency) is consistent with the improvement filing.

Confirmed Improvements

Cross-cutting Async Infrastructure Improvement (55 tests, -6.6% to -93.7%)

The defining characteristic of this improvement is its breadth: all 55 affected tests are async operations spanning 7 completely unrelated namespaces (HTTP, SSL, StreamReader, TarWriter, SemaphoreSlim, File I/O, WebSockets). This pattern strongly points to a change in the core async runtime infrastructure rather than individual library optimizations.

The improvements are particularly dramatic for StreamReaderReadLineTests.ReadLineAsync (70–94% faster) and TarWriter.*_WriteEntry_Async (48–54% faster), with more moderate but still very significant improvements for HTTP handler tests (18–47%), SslStream tests (7–45%), and file I/O tests (20–59%).

Primary Attribution: Runtime Async Timestamp Instrumentation Fix

• Likely Cause: 408f6a04efa8 - "[runtime-async] Fix timestamp instrumentation (#127662)"

  • Author: Rachel Jarvi
  • Affected Areas: Core async infrastructure — AsyncHelpers.CoreCLR.cs, Task.cs
  • Rationale: This commit removes significant overhead from the async suspension hot path:
    1. Eliminated lock contention: Removed lock (continuationTimestamps) blocks and Dictionary.TryGetValue/TryAdd calls from TryAddRuntimeAsyncContinuationChainTimestamps — an entire overload was deleted
    2. Simplified instrumentation: Removed unnecessary parameter passing (curContinuation, newContinuation) through SuspendAsyncContext, HandleSuspended, and QueueSuspendedRuntimeAsyncContext
    3. Reduced allocations: By eliminating the chain-walk timestamp propagation, fewer dictionary entries are created per async suspension
    4. The commit message includes a benchmark showing 387.4ms → 404.2ms (new vs old), confirming measurable improvement
  • The run configuration uses R2R interpreter mode (runtime-async enabled), meaning all async operations go through this exact code path, explaining the universal improvement across all test categories
  • Finding Confidence: 4/5 — Strong historical data, clear code path analysis
  • Improvement Confidence: 5/5 — Surgical change to async hot path affecting Task infrastructure, extreme z-scores (|z| ≥ 2.4 for all tests, |z| ≥ 10 for many), perfect temporal alignment, universal async test impact

• Candidates Considered:

  • 408f6a04efa8 - PRIMARY: Runtime async timestamp instrumentation fix — removes lock contention and dictionary operations from async suspension hot path
  • 5e21845e3c25 - SECONDARY: Enable compilation of async thunks in composite mode — may contribute to R2R interpreter perf but primarily a compilation change, not runtime hot path
  • 5cd3e8432113 - UNLIKELY for most tests: TAR metadata block size limit — only relevant to TAR reading, not TAR writing benchmarks

• Verification Status: Not verified via RunWithBuildAtHash. Bisection recommended to confirm attribution, particularly to separate the contribution of 408f6a04efa8 from 5e21845e3c25.

Borderline Tests

Two tests were classified as borderline:

1. Perf_NumberTensorPrimitives.IndexOfMax(BufferLength: 3079): -15.0%, z=-1.98. Close to significance threshold. This test is not obviously async-related, so the improvement may be from a different mechanism or noise.
2. SocketSendReceivePerfTest.SendReceive: -20.0%, z=-2.16. Just above significance. Likely a genuine improvement given the broader async pattern, but with lower confidence due to CV=8.7%.

Open Questions

• Metric Direction: Historical analysis tool reported "metricUnit=unknown" for all tests. All tests are latency benchmarks (μs, lower=better), which is consistent with the improvement filing. 4 tests were misclassified as "Regression" by the tool due to this ambiguity.
• Dual Commit Contribution: Both 408f6a04efa8 (async instrumentation fix) and 5e21845e3c25 (async thunks in composite mode) could contribute to the improvements. Bisection would separate their individual contributions.
• Non-Async Tests: The IndexOfMax tests (System.Numerics.Tensors) are not obviously async. Their improvement may be from a different mechanism or incidental.
• Verification Needed: While confidence is high based on code analysis, bisection would provide definitive attribution.

Recommended Actions

1. Bisection — Run bisection on commits 6820c801e8b8...71df7dbecba2 using RunWithBuildAtHash to confirm 408f6a04efa8 as the primary cause. Test with a representative subset: one ReadLineAsync variant, one SocketsHttpHandler variant, and one SslStream variant.
2. Verify secondary commit — Test 5e21845e3c25 specifically to determine if async thunk compilation contributes additional improvement beyond the instrumentation fix.
3. Close as positive — If bisection confirms, this can be closed as a confirmed improvement attributed to the runtime-async instrumentation optimization.

---

Generated by PerfTriageAgent on 2026-05-14 09:56 UTC

[BrzVlad]

Looks like dotnet/runtime#126904 @jtschuster 👍