# Bench (Performance Benchmarks)

**Status:** Implemented
**Source:** `src/Bench/`
**Owner:** Platform Team

> **Note:** This folder documents **performance benchmarks**. For **competitive benchmarking** (accuracy comparison with other scanners), see [`../benchmark/`](../benchmark/).

## Purpose

Bench provides performance benchmark infrastructure for StellaOps modules. Measures throughput, latency, and resource usage to detect regressions and validate performance targets.

## Components

**Benchmark Projects:**
- `StellaOps.Bench.LinkNotMerge` - Link-Not-Merge correlation performance
- `StellaOps.Bench.LinkNotMerge.Vex` - LNM VEX statement performance
- `StellaOps.Bench.Notify` - Notification delivery throughput
- `StellaOps.Bench.PolicyEngine` - Policy evaluation performance
- `StellaOps.Bench.ScannerAnalyzers` - Language analyzer performance

## Scanner Vendor Parity Tracking

`StellaOps.Bench.ScannerAnalyzers` now supports vendor parity tracking for offline benchmark runs:
- Scenario-level vendor ingestion from JSON or SARIF artifacts (`vendorResults[]` in benchmark config).
- Optional Stella finding ingestion (`stellaFindingsPath`) for exact overlap comparisons.
- Deterministic parity outputs in benchmark JSON and Prometheus exports:
  - overlap counts and percentages
  - scanner-only / vendor-only counts
  - parity score (Jaccard-style overlap over union)

## Usage

```bash
# Run all benchmarks
dotnet run -c Release --project src/Bench/StellaOps.Bench/LinkNotMerge/StellaOps.Bench.LinkNotMerge

# Run with specific runtime
dotnet run -c Release --project src/Bench/StellaOps.Bench/Notify/StellaOps.Bench.Notify
```

## Related Documentation

- Competitive Benchmark: `../benchmark/architecture.md`
- Scanner: `../scanner/architecture.md`
- Policy: `../policy/architecture.md`
- Notify: `../notify/architecture.md`