@ostera/rebench

:timer_clock: ReBench

A benchmarking package for ReasonML

Status: Usable ALPHA. The default announcer and cycle printers are not the most beautiful but they get the job done for now. Ideally in the future a default on-complete handler will inspect the data and show the clear winner. But this was the minimum I needed to start benchmarking some code.

Usage

Begin by adding the package to your dev dependencies:

ostera/reactor λ yarn add rebench --dev

And including it in your bsconfig.json file too:

{
  // ...
  "suffix": ".bs.js",
  "bs-dependencies": [
    "rebench",
  ],
  "warnings": {
    "error" : "+101"
  },
  // ... 
}

Now you're good! Make sure to run bsb -make-world :)

Writting Benchmarks

You begin by creating a ReBench suite with ReBench.make(). To that suite you can add as many test cases as you feel like with add(name, test_case, suite). This makes it possible to chain your calls and provides a very fluent API for writting the benchmarks.

let run = size =>
  ReBench.(
    make()
    |> add("ReStruct.BatchedQueue.tail", tail_batched_queue(size))
    |> add("ReStruct.BankersQueue.tail", tail_bankers_queue(size))
    |> add("ReStruct.Lazy.RealTimeQueue.tail", tail_realtime_queue(size))
    |> on(Start, Utils.default_announcer(~size, ~name="Queue.Tail"))
    |> on(Cycle, Utils.default_printer)
    |> on(Complete, _e => Js.log("Done!"))
    |> run(run_opts(~async=false))
  );

Running this benchmark has the following output:

ostera/restruct λ make bench
Benchmark: Queue.Tail (size: 10)
 => ReStruct.BatchedQueue.tail - 365964 ops
 => ReStruct.BankersQueue.tail - 5512905 ops
 => ReStruct.Lazy.RealTimeQueue.tail - 3797100 ops

Benchmark: Queue.Tail (size: 1000)
 => ReStruct.BatchedQueue.tail - 3167 ops
 => ReStruct.BankersQueue.tail - 5572714 ops
 => ReStruct.Lazy.RealTimeQueue.tail - 3344608 ops

Benchmark: Queue.Tail (size: 100000)
 => ReStruct.BatchedQueue.tail - 12 ops
 => ReStruct.BankersQueue.tail - 5365455 ops
 => ReStruct.Lazy.RealTimeQueue.tail - 3057600 ops

Done!