This folder contains BenchmarkDotNet measurements for ModularityKit.Mutator.

• commit execution without policy pressure
• strict engine execution with policy checks
• simulate and validate only paths
• batch execution overhead
• throughput for single mutation execution across multiple state sizes
• throughput for batch mutation execution across multiple state sizes and batch sizes
• policy evaluation overhead for no policy, synchronous policy, asynchronous policy, and mixed multi policy runs
• interception, audit/history, and logging diagnostics overhead in the core runtime
• parallel execution, state gate contention, and concurrent batch scheduling pressure in the core runtime

The throughput benchmarks use cloned array backed state so state size effects remain visible in the actual mutation path rather than being hidden behind an artificial inner loop.

Build first:

dotnet build Benchmarks/ModularityKit.Mutator.Benchmarks.csproj -c Release

Run a specific benchmark:

dotnet Benchmarks/bin/Release/net10.0/ModularityKit.Mutator.Benchmarks.dll --filter '*MutationEngineCommitBenchmarks*'

Run the throughput-focused suite:

dotnet Benchmarks/bin/Release/net10.0/ModularityKit.Mutator.Benchmarks.dll --filter '*MutationEngineThroughputBenchmarks*'

Run the policy evaluation suite:

dotnet Benchmarks/bin/Release/net10.0/ModularityKit.Mutator.Benchmarks.dll --filter '*PolicyEvaluationSingleBenchmarks*'
dotnet Benchmarks/bin/Release/net10.0/ModularityKit.Mutator.Benchmarks.dll --filter '*PolicyEvaluationMultiBenchmarks*'

Run the diagnostics suite:

dotnet Benchmarks/bin/Release/net10.0/ModularityKit.Mutator.Benchmarks.dll --anyCategories Diagnostics

Run the concurrency suite:

dotnet Benchmarks/bin/Release/net10.0/ModularityKit.Mutator.Benchmarks.dll --anyCategories Concurrency

Key parameters reported by BenchmarkDotNet:

• StateSize controls the size of the cloned mutation state
• BatchSize controls the number of mutations executed in the batch benchmark

• The benchmark harness is configured for the current environment.
• Results are emitted by BenchmarkDotNet when the runner can write artifacts.