kylejeske / mitata

cross-runtime benchmarking lib and cli

mitata

cross-runtime benchmarking lib

Install

bun add mitata

npm install mitata

Examples

import { run, bench, group, baseline } from 'mitata';

// deno
// import { ... } from 'https://esm.sh/mitata';

// d8/jsc
// import { ... } from '<path to mitata>/src/cli.mjs';

bench('noop', () => {});
bench('noop2', () => {});

group('group', () => {
  baseline('baseline', () => {});
  bench('Date.now()', () => Date.now());
  bench('performance.now()', () => performance.now());
});

group({ name: 'group2', summary: false }, () => {
  bench('new Array(0)', () => new Array(0));
  bench('new Array(1024)', () => new Array(1024));
});

await run({
  avg: true, // enable/disable avg column (default: true)
  json: false, // enable/disable json output (default: false)
  colors: true, // enable/disable colors (default: true)
  min_max: true, // enable/disable min/max column (default: true)
  collect: false, // enable/disable collecting returned values into an array during the benchmark (default: false)
  percentiles: false, // enable/disable percentiles column (default: true)
});

output:

terminal screenshot (click to view)

cpu: Apple M1
runtime: bun 0.0.83 (arm64-darwin)

benchmark              time (avg)             (min … max)
---------------------------------------------------------
noop               323.34 ps/iter    (304.1 ps … 8.64 ns)
noop2              387.15 ps/iter   (304.1 ps … 16.79 ns)

• group
---------------------------------------------------------
baseline           782.71 ps/iter    (304.1 ps … 6.34 ns)
Date.now()          31.04 ns/iter    (30.9 ns … 34.59 ns)
performance.now()   92.71 ns/iter  (91.72 ns … 101.45 ns)

summary for group
  baseline
   39.66x faster than Date.now()
   118.45x faster than performance.now()

• group2
---------------------------------------------------------
new Array(0)         7.47 ns/iter   (6.46 ns … 269.07 ns)
new Array(1024)    286.39 ns/iter  (241.37 ns … 560.7 ns)

JIT bias

If you run benchmarks like this, you might notice that they get slower (only few nanoseconds) after the first few runs.

bench('noop', () => {});
bench('noop', () => {});
bench('noop', () => {});
bench('noop', () => {});
bench('noop', () => {});
bench('noop', () => {});
bench('noop', () => {});
bench('noop', () => {});
bench('noop', () => {});
bench('noop', () => {});
bench('noop', () => {});
bench('noop', () => {});
bench('noop', () => {});

I call this behavior "JIT bias". In simple words, v8 and JSC JIT expect us to pass the same function, so they optimize for it, but we break that promise and get deoptimized.

License

MIT © Evan