@javelin/perf

A runnerless, single-threaded benchmarking library for Node.

Getting Started

1npm i @javelin/perf

src/fibonacci.perf.ts

1import {fibonacci} from "./fibonacci"
2import {perf} from "@javelin/perf"
3
4perf("100 elements", () => fibonacci(100))

1> ts-node src/fibonacci.perf.ts
2
3src/fibonacci.perf.ts
4
5 100 elements   106,321 ops/s     =   4.13

Setup

1perf("iterator next()", () => {
2  let it = fibonacci.iterator({start: 1_000})
3  return () => it.next()
4})

Options

Run just a single benchmark in a file:

1perf.only("name", () => {})

Don't throw an error when a benchmark degrades in performance.

1perf("name", () => {}, {
2  throwOnFailure: false,
3})

Configuration

1# Throw an error when performance degrades, halting test execution.
2THROW_ON_FAILURE=false
3# Write failing test stats to perf results files.
4WRITE_FAILURES=false
5# Number of times to execute each individual test.
6PERF_RUNS=15_000
7# Number of samples to ignore from upper and lower extremes when computing perf results.
8PERF_SAMPLES_TO_DISCARD_PER_EXTREME=100
9# Perf results file extension.
10MODULE_RESULTS_EXTENSION=".perf-results"
11# Perf file extension.
12MODULE_EXTENSION=".perf.ts"