Performance Benchmarks

Performance benchmarks run on-demand via GitHub Actions to track ContentLoader, Registry, and ECS system performance across releases.

Latest Run

Performance benchmark results will appear here after the next benchmark run.

Benchmark Coverage

We measure:

• ContentLoader.LoadSimplePack — Time to load a basic pack (10 units, 5 buildings)
• ContentLoader.LoadComplexPack — Time to load a full pack (100+ definitions, cross-references)
• Registry.QueryUnits — Time to query 1000 units by faction + role
• Registry.RegisterBulkUnits — Time to register 100 units with conflict detection

Running Locally

dotnet run --project src/Tools/Benchmarks/DINOForge.Tools.Benchmarks.csproj -- --configuration Release

Results appear in BenchmarkDotNet.Artifacts/results/.

Baseline Management

• First run: Creates docs/benchmarks/baseline.json as the baseline
• Subsequent runs: Compare against baseline; >10% regression fails the workflow
• On success: Baseline is updated with new results (moving target)

To reset baseline:

rm docs/benchmarks/baseline.json

The next benchmark run will establish a new baseline.

Historical Data

Benchmark history is available via GitHub Actions artifacts:

• Workflow: .github/workflows/benchmarks.yml
• Artifacts: Retained for 90 days
• Latest run: Available in docs/benchmarks/latest-raw.json

CI Integration

Benchmarks run on Ubuntu latest (ubuntu-latest) with:

• .NET 11.0.x
• Release configuration
• Isolated from game instance (no ECS)

Benchmarks do NOT run on Windows (where game deployment occurs) to avoid interference from game processes.