| name | perf |
| description | Performance optimization with benchmarking first. Use when asked to optimize, improve speed, reduce allocations, or profile code.
|
| disable-model-invocation | true |
| argument-hint | <class.method to optimize> |
You are tasked with optimizing the performance of: $ARGUMENTS
Follow this workflow in order:
-
Create a BenchmarkDotNet benchmark file:
- Path:
PhotoManager/Benchmarks/PhotoManager.Benchmarks/{Category}/{ClassName}{MethodName}Benchmarks.cs
- Example:
PhotoManager/Benchmarks/PhotoManager.Benchmarks/Common/HashingHelperCalculateHashBenchmarks.cs
- Use attributes:
[MemoryDiagnoser], [Orderer(SummaryOrderPolicy.FastestToSlowest)],
[RankColumn]
- Mark the CURRENT implementation as
[Benchmark(Baseline = true)]
-
Keep the original method unchanged in its source file — do NOT modify it yet.
-
Add optimized implementations IN THE BENCHMARK FILE ONLY:
- Each optimization gets its own
[Benchmark] method
- Name them clearly:
Optimized_ApproachName, Optimized_Stackalloc, etc.
-
Run the benchmark:
dotnet run --project PhotoManager/Benchmarks/PhotoManager.Benchmarks/PhotoManager.Benchmarks.csproj -c Release -- --filter "*BenchmarkClassName"
-
Evaluate results (speed is priority, but allocation shouldn't increase significantly):
- Speed: Lower Mean/Median is better
- Memory: Lower Allocated is better
- Check the Ratio column compared to baseline
-
Implement the winner in the original source file:
- Replace ONLY if there's clear improvement
- Add comment:
// ClassName.MethodName: X% faster
- Check
GlobalUsings.cs before adding any using directives
-
Verify: Run dotnet test --filter "FullyQualifiedName~ClassName" PhotoManager/PhotoManager.slnx
to ensure no regressions. Build must produce zero warnings.
-
Keep all variants in the benchmark file (original + all optimizations) for future reference.
Return benchmark results summary with percentage improvement.
