Spaz's hrBench

High-resolution micro-benchmarking in Node.js.

Install and Usage

npm install spazmodius/hrbench -D

const Benchmark = require('@spazmodius/hrbench')
const crypto = require('crypto')

Benchmark('Getting crypto bytes')
	.test('small', () => crypto.randomBytes(15))
	.test('medium', () => crypto.randomBytes(150))

Concepts

Concept	Definition
Benchmark	A suite of tests whose performance will be compared to each other.
Test	A single operation in a benchmark, to be timed and compared to other tests in the same Benchmark.
Cycle	A single execution of a test. There are many cycles in a round.
Round	Many repeated cycles, allowing enough time to pass for stable timing.

A benchmark times each test by running it in a tight loop (a round). Each round will consist of enough cycles, typically, to take about 1/10th of a second. Rounds themselves will be repeated as necessary to reach statistical confidence.

Once all of the tests have been timed to satisfactory statistical confidence, their timings will be summarized, ordered from fastest to slowest, and printed to the console.

Ansync tests

No, sorry.

API

[new] Benchmark( [name], [options] )

Argument	Type	Optional	Description
name	string	✓	Name of this benchmark.
options	object	✓	Options

Starts a benchmark, to compare the performance of a set of tests.

Options

name

Naming the benchmark is useful if there are multiple benchmarks in the file.

maxCycles

Set the maximum number of cycles per round.

Typically, the number of cycles is automatically estimated so as to produce rounds approximately 1/10th of a second each. However, limiting the number of cycles per round is useful for memory-intensive tests, since there will be no garbage-collection between cycles during a round.

benchmark.test( [name], op, [noop] )

Argument	Type	Optional	Description
name	string	✓	Name of this test.
op	function		The operation to time.
noop	function or null	✓	Operation that represents doing nothing.

Adds a test to the benchmark.

If a name is not specified, then one will be generated from the name or contents of the operation.

The noop function will be timed just like the operation, and subtracted from the results. This will calibrate for the overhead of looping and invoking the operation. By default, noop is an empty function call. See Demo3 for an example of a non-trivial noop.

However, passing null will time an empty loop without any function call at all. This has the effect of including the cost of invocation in the reported results. See Demo1 for an example of where this might be useful.