Parser benchmarks

VcsToolkit.Benchmarks is a manual BenchmarkDotNet tool for before/after parser comparisons. It measures the unified-diff parser, Git porcelain-v2 and conflict parsers, and jj template and conflict parsers against deterministic fixtures committed under benchmarks/VcsToolkit.Benchmarks/fixtures/.

Benchmarks are not a CI gate. Hosted runners are noisy, so timing changes there are not reliable enough to block merges. Run them locally when changing parser hot paths and compare results on the same machine.

From the repository root:

cd benchmarks/VcsToolkit.Benchmarks
dotnet run -c Release

Run one benchmark group with BenchmarkDotNet's filter option:

dotnet run -c Release -- --filter "*DiffBenchmarks*"
dotnet run -c Release -- --filter "*GitPorcelainBenchmarks*"
dotnet run -c Release -- --filter "*JjBenchmarks*"

The benchmark project is included in the solution for deterministic build ordering, but it is non-packable and .github/workflows/ci.yml does not run it.