[Perf] Linux/arm64: 15 Regressions on 8/28/2026 11:31:31 PM +00:00

Open

#79,268 opened on Sep 10, 2026 1 comment 0 reactions 0 assignees View on GitHub

arch-arm64 branch-refs/heads/main compilationmode-tiered kind-micro os-linux perf-regression runkind-micro runtime-coreclr untriaged

Dominant language: No language data
Stars: 10
Forks: 8
PR merge metrics: PR metrics pending

Description

### Run Information

Name | Value
-- | --
Architecture | arm64
OS | azurelinux 3.0
Queue | CobaltAzureLinux
Baseline | [100da3f94464f99616ee4a401670bec491ebc787](https://github.com/dotnet/runtime/commit/100da3f94464f99616ee4a401670bec491ebc787)
Compare | [aa036afce592ad80e938a35bd376222fb232cba9](https://github.com/dotnet/runtime/commit/aa036afce592ad80e938a35bd376222fb232cba9)
Diff | [Diff](https://github.com/dotnet/runtime/compare/100da3f94464f99616ee4a401670bec491ebc787...aa036afce592ad80e938a35bd376222fb232cba9)
Configs | CompilationMode:tiered, RunKind:micro

### Regressions in System.Buffers.Tests.RentReturnArrayPoolTests<Object>

Benchmark | Baseline | Test | Test/Base | Test Quality | Edge Detector | Baseline IR | Compare IR | IR Ratio
-- | -- | -- | -- | -- | -- | -- | -- | --
|

[SingleSerial - Duration of single invocation]()
📝 - [Benchmark Source]()
[ADX - Test Multi Config Graph]()

| 191.15 ns | 468.59 ns | 2.45 | 0.33 | False | | |
|

[SingleSerial - Duration of single invocation]()
📝 - [Benchmark Source]()
[ADX - Test Multi Config Graph]()

| 171.65 ns | 379.27 ns | 2.21 | 0.35 | False | | |
|

[MultipleSerial - Duration of single invocation]()
📝 - [Benchmark Source]()
[ADX - Test Multi Config Graph]()

| 708.78 ns | 936.31 ns | 1.32 | 0.35 | False | | |

![graph]()
![graph]()
![graph]()
[Test Report]()

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

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

### System.Buffers.Tests.RentReturnArrayPoolTests<Object>.SingleSerial(RentalSize: 4096, ManipulateArray: False, Async: True, UseSharedPool: False)

#### ETL Files

#### Histogram

#### JIT Disasms

### System.Buffers.Tests.RentReturnArrayPoolTests<Object>.SingleSerial(RentalSize: 4096, ManipulateArray: False, Async: True, UseSharedPool: True)

#### ETL Files

#### Histogram

#### JIT Disasms

### System.Buffers.Tests.RentReturnArrayPoolTests<Object>.MultipleSerial(RentalSize: 4096, ManipulateArray: False, Async: True, UseSharedPool: False)

#### ETL Files

#### Histogram

#### JIT Disasms

### Docs

[Profiling workflow for dotnet/runtime repository](https://github.com/dotnet/performance/blob/master/docs/profiling-workflow-dotnet-runtime.md)
[Benchmarking workflow for dotnet/runtime repository](https://github.com/dotnet/performance/blob/master/docs/benchmarking-workflow-dotnet-runtime.md)

---

### Run Information

### Regressions in System.Buffers.Tests.RentReturnArrayPoolTests<Byte>

Benchmark | Baseline | Test | Test/Base | Test Quality | Edge Detector | Baseline IR | Compare IR | IR Ratio
-- | -- | -- | -- | -- | -- | -- | -- | --
|

[SingleSerial - Duration of single invocation]()
📝 - [Benchmark Source]()
[ADX - Test Multi Config Graph]()

| 175.11 ns | 515.42 ns | 2.94 | 0.22 | False | | |
|

[SingleSerial - Duration of single invocation]()
📝 - [Benchmark Source]()
[ADX - Test Multi Config Graph]()

| 125.58 ns | 331.06 ns | 2.64 | 0.39 | False | | |
|

[MultipleSerial - Duration of single invocation]()
📝 - [Benchmark Source]()
[ADX - Test Multi Config Graph]()

| 674.61 ns | 911.10 ns | 1.35 | 0.27 | False | | |

![graph]()
![graph]()
![graph]()
[Test Report]()

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

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

### System.Buffers.Tests.RentReturnArrayPoolTests<Byte>.SingleSerial(RentalSize: 4096, ManipulateArray: False, Async: True, UseSharedPool: False)

#### ETL Files

#### Histogram

#### JIT Disasms

### System.Buffers.Tests.RentReturnArrayPoolTests<Byte>.SingleSerial(RentalSize: 4096, ManipulateArray: False, Async: True, UseSharedPool: True)

#### ETL Files

#### Histogram

#### JIT Disasms

### System.Buffers.Tests.RentReturnArrayPoolTests<Byte>.MultipleSerial(RentalSize: 4096, ManipulateArray: False, Async: True, UseSharedPool: False)

#### ETL Files

#### Histogram

#### JIT Disasms

### Docs

---

### Run Information

### Regressions in System.Threading.Tasks.Tests.Perf_AsyncMethods

Benchmark | Baseline | Test | Test/Base | Test Quality | Edge Detector | Baseline IR | Compare IR | IR Ratio
-- | -- | -- | -- | -- | -- | -- | -- | --
|

[Yield - Duration of single invocation]()
📝 - [Benchmark Source]()
[ADX - Test Multi Config Graph]()

| 92.46 ns | 371.88 ns | 4.02 | 0.42 | False | | |
|

[SingleYieldMethodInvocation - Duration of single invocation]()
📝 - [Benchmark Source]()
[ADX - Test Multi Config Graph]()

| 356.21 ns | 731.25 ns | 2.05 | 0.27 | False | | |

![graph]()
![graph]()
[Test Report]()

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

```cmd
git clone https://github.com/dotnet/performance.git
py .\performance\scripts\benchmarks_ci.py -f net8.0 --filter 'System.Threading.Tasks.Tests.Perf_AsyncMethods*'
```

### System.Threading.Tasks.Tests.Perf_AsyncMethods.Yield

#### ETL Files

#### Histogram

#### JIT Disasms

### System.Threading.Tasks.Tests.Perf_AsyncMethods.SingleYieldMethodInvocation

#### ETL Files

#### Histogram

#### JIT Disasms

### Docs

---

### Run Information

### Regressions in System.Net.Security.Tests.SslStreamTests

Benchmark | Baseline | Test | Test/Base | Test Quality | Edge Detector | Baseline IR | Compare IR | IR Ratio
-- | -- | -- | -- | -- | -- | -- | -- | --
|

[ConcurrentReadWriteLargeBuffer - Duration of single invocation]()
📝 - [Benchmark Source]()
[ADX - Test Multi Config Graph]()

| 47.47 μs | 55.22 μs | 1.16 | 0.38 | False | | |

![graph]()
[Test Report]()

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

```cmd
git clone https://github.com/dotnet/performance.git
py .\performance\scripts\benchmarks_ci.py -f net8.0 --filter 'System.Net.Security.Tests.SslStreamTests*'
```

### System.Net.Security.Tests.SslStreamTests.ConcurrentReadWriteLargeBuffer

#### ETL Files

#### Histogram

#### JIT Disasms

### Docs

---

### Run Information

### Regressions in System.Formats.Tar.Tests.Perf_TarFile

Benchmark | Baseline | Test | Test/Base | Test Quality | Edge Detector | Baseline IR | Compare IR | IR Ratio
-- | -- | -- | -- | -- | -- | -- | -- | --
|

[CreateFromDirectory_Path_Async - Duration of single invocation]()
📝 - [Benchmark Source]()
[ADX - Test Multi Config Graph]()

| 101.60 μs | 132.65 μs | 1.31 | 0.67 | False | | |

![graph]()
[Test Report]()

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

```cmd
git clone https://github.com/dotnet/performance.git
py .\performance\scripts\benchmarks_ci.py -f net8.0 --filter 'System.Formats.Tar.Tests.Perf_TarFile*'
```

### System.Formats.Tar.Tests.Perf_TarFile.CreateFromDirectory_Path_Async

#### ETL Files

#### Histogram

#### JIT Disasms

### Docs

---

### Run Information

### Regressions in Microsoft.AspNetCore.Server.Kestrel.Performance.PipeThroughputBenchmark

Benchmark | Baseline | Test | Test/Base | Test Quality | Edge Detector | Baseline IR | Compare IR | IR Ratio
-- | -- | -- | -- | -- | -- | -- | -- | --
|

[Parse_ParallelAsync - Duration of single invocation]()
📝 - [Benchmark Source]()
[ADX - Test Multi Config Graph]()

| 558.98 ns | 958.52 ns | 1.71 | 0.50 | False | | |

![graph]()
[Test Report]()

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

```cmd
git clone https://github.com/dotnet/performance.git
py .\performance\scripts\benchmarks_ci.py -f net8.0 --filter 'Microsoft.AspNetCore.Server.Kestrel.Performance.PipeThroughputBenchmark*'
```

### Microsoft.AspNetCore.Server.Kestrel.Performance.PipeThroughputBenchmark.Parse_ParallelAsync(Length: 4096, Chunks: 1)

#### ETL Files

#### Histogram

#### JIT Disasms

### Docs

---

### Run Information

### Regressions in System.Threading.Tasks.ValueTaskPerfTest

Benchmark | Baseline | Test | Test/Base | Test Quality | Edge Detector | Baseline IR | Compare IR | IR Ratio
-- | -- | -- | -- | -- | -- | -- | -- | --
|

[CreateAndAwait_FromYieldingAsyncMethod - Duration of single invocation]()
📝 - [Benchmark Source]()
[ADX - Test Multi Config Graph]()

| 589.31 ns | 928.31 ns | 1.58 | 0.35 | False | | |

![graph]()
[Test Report]()

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

```cmd
git clone https://github.com/dotnet/performance.git
py .\performance\scripts\benchmarks_ci.py -f net8.0 --filter 'System.Threading.Tasks.ValueTaskPerfTest*'
```

### System.Threading.Tasks.ValueTaskPerfTest.CreateAndAwait_FromYieldingAsyncMethod

#### ETL Files

#### Histogram

#### JIT Disasms

### Docs

---

### Run Information

### Regressions in System.Threading.Channels.Tests.UnboundedChannelPerfTests

Benchmark | Baseline | Test | Test/Base | Test Quality | Edge Detector | Baseline IR | Compare IR | IR Ratio
-- | -- | -- | -- | -- | -- | -- | -- | --
|

[PingPong - Duration of single invocation]()
[ADX - Test Multi Config Graph]()

| 8.44 ms | 16.25 ms | 1.93 | 0.39 | False | | |

![graph]()
[Test Report]()

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

```cmd
git clone https://github.com/dotnet/performance.git
py .\performance\scripts\benchmarks_ci.py -f net8.0 --filter 'System.Threading.Channels.Tests.UnboundedChannelPerfTests*'
```

### System.Threading.Channels.Tests.UnboundedChannelPerfTests.PingPong

#### ETL Files

#### Histogram

#### JIT Disasms

### Docs

---

### Run Information

### Regressions in System.Threading.Channels.Tests.BoundedChannelPerfTests

Benchmark | Baseline | Test | Test/Base | Test Quality | Edge Detector | Baseline IR | Compare IR | IR Ratio
-- | -- | -- | -- | -- | -- | -- | -- | --
|

[PingPong - Duration of single invocation]()
[ADX - Test Multi Config Graph]()

| 8.48 ms | 16.09 ms | 1.90 | 0.37 | False | | |

![graph]()
[Test Report]()

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

```cmd
git clone https://github.com/dotnet/performance.git
py .\performance\scripts\benchmarks_ci.py -f net8.0 --filter 'System.Threading.Channels.Tests.BoundedChannelPerfTests*'
```

### System.Threading.Channels.Tests.BoundedChannelPerfTests.PingPong

#### ETL Files

#### Histogram

#### JIT Disasms

### Docs

---

### Run Information

### Regressions in System.Threading.Channels.Tests.SpscUnboundedChannelPerfTests

Benchmark | Baseline | Test | Test/Base | Test Quality | Edge Detector | Baseline IR | Compare IR | IR Ratio
-- | -- | -- | -- | -- | -- | -- | -- | --
|

[PingPong - Duration of single invocation]()
[ADX - Test Multi Config Graph]()

| 8.48 ms | 17.37 ms | 2.05 | 0.44 | False | | |

![graph]()
[Test Report]()

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

```cmd
git clone https://github.com/dotnet/performance.git
py .\performance\scripts\benchmarks_ci.py -f net8.0 --filter 'System.Threading.Channels.Tests.SpscUnboundedChannelPerfTests*'
```

### System.Threading.Channels.Tests.SpscUnboundedChannelPerfTests.PingPong

#### ETL Files

#### Histogram

#### JIT Disasms

### Docs

Contributor guide

No contributing guide indexed for this repository

Research direction

Start with the linked ArrayPoolTests.cs benchmark source and the reported baseline/compare commits. Follow the benchmarking and profiling workflow documents, then run the provided benchmarks_ci.py command for RentReturnArrayPoolTests. Done means determining the cause of the ARM64 regression and confirming the reported performance change is addressed.

Written by the indexing model from the issue text.

Assessment

Tech stack: azure, csharp, linux
Domain: performance
Issue type: Bug
Difficulty: 4/5
Estimated time: 3-5 days
Activity status: Active
Clarity: Needs clarification
Newbie friendliness: 38/100

[Perf] Linux/arm64: 15 Regressions on 8/28/2026 11:31:31 PM +00:00

Description

Contributor guide

Research direction

Assessment

Get new issues in your inbox