* Statistical analysis across multiple runs.
* Support for arbitrary shell commands.
* Constant feedback about the benchmark progress and current estimates.
* Warmup runs can be executed before the actual benchmark.
* Cache-clearing commands can be set up before each timing run.
* Statistical outlier detection to detect interference from other programs and caching effects.
* Export results to various formats: CSV, JSON, Markdown, AsciiDoc.
* Parameterized benchmarks (e.g. vary the number of threads).