Benchmarking Bonsai Components

Bonsai_bench is a library which provides facilities for benchmarking Bonsai components.

Creating a benchmark

Bonsai_bench.Test.create creates a benchmark. It takes a component to benchmark, a way to extract that component's inject_action function, and a list of interactions to perform during the benchmark. For example:

let state =
  Bonsai_bench.Test.create
    ~name:"Bonsai.state"
    ~component:(Bonsai.state [%here] (module Int) ~default_model:0)
    ~get_inject:(fun (_, inject) -> inject)
    Bonsai_bench.Interaction.(many_with_stabilizations [ inject 1; reset_model ])

There are more examples in the example/ directory.

Running a benchmark

To run the benchmark, pass it to Bonsai_bench.bench, a thin wrapper around Core_bench_js.bench that handles necessary cleanup between tests:

let () =
  let quota = Core_bench_js.Quota.Span (Time.Span.of_sec 1.0) in
  Bonsai_bench.bench ~run_config:(Core_bench_js.Run_config.create () ~quota) [ state ]

Build the javascript-executables target for the directory with BUILD_PROFILE=fast-exe. Then, from a terminal, run: node path/to/executable/name.bc.js to see the results.

Code organization

• src/bonsai_bench.mli: Contains the API for running and creating tests. The list of possible interactions that can be performed during benchmarks can also be found there.