[Perf] Linux/arm64: 3 Regressions on 8/19/2026 11:48:00 PM +00:00

### Run Information

Name | Value
-- | --
Architecture | arm64
OS | azurelinux 3.0
Queue | CobaltAzureLinux
Baseline | [22645b47c400e6cc0aec97ddcb3f1998677b6d94](https://github.com/dotnet/runtime/commit/22645b47c400e6cc0aec97ddcb3f1998677b6d94)
Compare | [cb8ddc8cac0182e932903d921bd09300be7ae2a6](https://github.com/dotnet/runtime/commit/cb8ddc8cac0182e932903d921bd09300be7ae2a6)
Diff | [Diff](https://github.com/dotnet/runtime/compare/22645b47c400e6cc0aec97ddcb3f1998677b6d94...cb8ddc8cac0182e932903d921bd09300be7ae2a6)
Configs | CompilationMode:tiered, R2RType:r2r_interpreter, RunKind:micro

### Regressions in System.Tests.Perf_Single

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

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

| 267.23 ns | 337.22 ns | 1.26 | 0.10 | False | | |
|

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

| 474.81 ns | 548.03 ns | 1.15 | 0.10 | 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.Tests.Perf_Single*'
```

### System.Tests.Perf_Single.ToStringWithFormat(value: 12345, format: "G17")

#### ETL Files

#### Histogram

#### JIT Disasms

### System.Tests.Perf_Single.ToStringWithFormat(value: 12345, format: "F50")

#### 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

Name | Value
-- | --
Architecture | arm64
OS | azurelinux 3.0
Queue | CobaltAzureLinux
Baseline | [22645b47c400e6cc0aec97ddcb3f1998677b6d94](https://github.com/dotnet/runtime/commit/22645b47c400e6cc0aec97ddcb3f1998677b6d94)
Compare | [cb8ddc8cac0182e932903d921bd09300be7ae2a6](https://github.com/dotnet/runtime/commit/cb8ddc8cac0182e932903d921bd09300be7ae2a6)
Diff | [Diff](https://github.com/dotnet/runtime/compare/22645b47c400e6cc0aec97ddcb3f1998677b6d94...cb8ddc8cac0182e932903d921bd09300be7ae2a6)
Configs | CompilationMode:tiered, R2RType:r2r_interpreter, RunKind:micro

### Regressions in System.Tests.Perf_Double

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

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

| 288.18 ns | 345.78 ns | 1.20 | 0.13 | 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.Tests.Perf_Double*'
```

### System.Tests.Perf_Double.ToStringWithFormat(value: 12345, format: "G17")

#### 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)

Start with the benchmark sources in src/benchmarks/micro/libraries/System.Runtime/Perf.Single.cs and Perf.Double.cs, then compare the baseline and test commits linked in the issue. Run the provided benchmarks_ci.py commands for System.Tests.Perf_Single and System.Tests.Perf_Double on the specified configuration and inspect the linked reports. Done means the ARM64 regressions are reproduced, their cause is identified, and the performance result is addressed or explained.

Written by the indexing model from the issue text.