# Benchmarks
See [BENCHMARK.md](../BENCHMARK.md) for the benchmark policy. In short:
authoritative Rustwright benchmark evidence should use capped, sharded Docker
runs instead of the developer machine's main Chrome or a long-lived host Chrome
process.
This directory contains two benchmark lanes:
- `run_benchmarks.py` and `automation_cases.py`: Rustwright API/equivalent
automation speed and parity cases. This is the primary speed lane for
Rustwright-vs-Playwright claims.
- Browser speed candidate folders: browser-engine, page-load, and external
benchmark scaffolds used as supporting evidence.
## Browser Speed Candidates
List the available browser-speed scaffolds:
```bash
python benchmarks/browser_speed/list.py
```
The browser-speed folders are intentionally lightweight. Setup commands place
downloaded tools, cloned repos, and outputs under ignored `.benchmark-data/`.
Useful starting points:
```bash
python benchmarks/crossbench/run.py --setup
python benchmarks/speedometer/run.py --repeat 20 --browser chrome-stable --dry-run
python benchmarks/tachometer/run.py --setup
python benchmarks/browsertime/run.py --setup
```
Use `bench-full` first for automation-library speed. Use these browser-speed
candidates to check browser-engine/config, page-load, and external synthetic
performance signals.