The project's benchmark harness

The benchmark harness runs the project's benchmarks and keeps their results, so a change can be measured against what came before. Which harness that is depends on the language, and is found in the project, not assumed.

Finding it

Look for benchmark code (a benches/ directory in Rust, Benchmark functions in Go test files, a JMH source set), for a bench target in the task file, and in CI for a job that runs it and stores results. Saved baselines live with that job, in a separate branch or in a store.

Using it

• Run it as found: cargo bench, go test -run='^$' -bench=., pytest --benchmark-only.
• One machine: a dedicated machine, or the old and the new code run in turns on the same one, so its noise falls on both.
• Several runs against a saved baseline: go test -count=10 compared with benchstat; --benchmark-save and --benchmark-compare in pytest-benchmark; --save-baseline and --baseline in criterion.

Examples: pytest-benchmark (Python), criterion (Rust), go test -bench (Go), JMH (Java), BenchmarkDotNet (.NET).