OCaml Bench brings together the community's work on measuring the performance of the OCaml compiler, runtime and ecosystem. We want results that compiler developers can act on before a PR is merged, that release teams can use to compare versions (OCaml 5.x, trunk, OxCaml), and that anyone can reproduce on their own machine using the same tools that run in CI.
Start with the Sandmark wiki, which has an overview of the project, how the pieces fit together, and the design of the upcoming benchmarking service. The Sandmark 3.0 roadmap shows our current roadmap.
Questions, new benchmarks and bug reports are welcome on any of the repositories below. We suggest consulting the README.md and contribution guide for each project.
- sandmark: runs benchmarks on tuned machines providing visualisations of the results.
- macro-benches: a suite of realistic OCaml applications in the style of DaCapo.
- micro-benches: the classic suite of micro benchmarks, each exercising a particular part of the language or runtime, and runnable standalone with minimal setup.