Our proposed presentation would describe what we did to build continuous benchmarking websites that take a controlled experiment approach to running the operf-micro and sandmark benchmarking suites against tracked git branches of the OCaml compiler. Our goal was to produce tools to allow for the efficient upstreaming of multicore related functionality into the OCaml compiler.

The presentation will cover the available OCaml benchmarking suites; how other compiler communities handle performance benchmark; how we put together our experimental setup to collect controlled data; the interesting experience that should be shared from the project; interesting future work and extensions that could be pursued in the future.