fen-bench

A small, sane JS micro-benchmark library.

---

This library implements the only way I've found of getting consistent performance data about JS functions. It attempts to account for how modern JS engines optimize, by carefully managing references, measuring continuously, and discarding old measurements.

Browser demo here

YMMV, but for me measuring performance this way has solved the problem of getting wildly varying results over time or across page reloads, as tends to happen with other libraries.

Usage:

npm install --save-dev fen-bench

var Bench = require('fen-bench')
var bench = new Bench()

// settings
bench.testDuration = 100            // ms to run each case
bench.pauseDuration = 15            // ms between cases
bench.maxMeasurements = 10          // how many data points to keep

// define test cases
bench.testCases.push(function testOne() {
    for (var sum = 0, i = 0; i < 1e6; i++) sum += i % 16
    // test cases must return a consistent value
    return sum
})

// runs continuously until stopped
bench.start()
setTimeout(() => bench.stop(), 5000)

// callback is called after each round of tests
bench.callback = () => {
    bench.results.forEach(res => {
        console.log(res.ops, 'calls/sec')
        console.log(res.dev, 'std. deviation')
    })
}

// clears currently stored results
bench.reset()

Standard disclaimers apply for test cases - they should return a consistent value that is derived from the work your function is doing. This prevents clever JS engines from marking the heavy part of your test case as dead code, and optimizing it into an empty function.

---

Author: https://github.com/fenomas

License: MIT