Skip to main content

Bench

Bench is a benchmark harness that allows you to organize a number of benchmarks / performance tests into a Benchmark suite that can be executed with a single command.

Each test must be registered with a unique id which allows Bench to do compare results across runs and perform limited regression testing.

Precise, repeatable performance measuring on a modern multi-tasking OS is hard, as subsequent runs might be scheduled different by the OS. The consistency between repeated benchmark tests tend to be within +- 10-15% and is generally sufficient for verifying incremental improvements.

Usage​

import {Bench} from '@probe.gl/bench';

const bench = new Bench()
  .group('Utility tests')
  .add('Math.sqrt', () => Math.sqrt(100))
  ;

bench.run();

Methods​

constructor​

new Bench({})

group(id)​

Adds a group header.

bench.group(id)

add​

Adds a test case. Supports multiple signatures:

bench.add(id: string, testCaseProps: TestCaseProps, testFunc: () => unknown) bench.add(id: string, testFunc: () => unknown)

Parameters

  • id (String) - The unique string for this test. Used as the description of the test in the results.
  • testFunc (Function, options) - Function run for each test iteration.

Options

  • priority=0 (Number, optional) - allows controlling which bench cases execute. Can also be specified through the options object.
  • initialize=: () => any initialization function called once before testFunc iterations start.
  • multiplier=1 : Number Multiplier applied to the number of actual iterations. Use this if each test case already performs a number of iterations. Affects reporting only.
  • unit='iterations': Can be used to customize the output message for bench results.
  • _throughput=false : Number Use with Bench.addAsync to specify that _throughput iterations should be run in parallel. Note that automatic iteration selection is not available in this case.

Returns: itself for chaining.

addAsync​

Adds an async test case. Use when testFunc returns a promise. Supports same signatures as add.

bench.add(id: string, testCaseProps: TestCaseProps, testFunc: () => Promise<unknown>) bench.add(id: string, testFunc: () => Promise<unknown>)

When using addAsync, testFunc is expected to return a promise.

run()​

bench.run()

calibrate​

bench.calibrate(id, func1, func2, opts)