xrandom 0.0.2

Fast random number generators: xorshift32, xorshift64, xorshift128 and xorshift128+

xrandom

Fast random number generators in Dart.

Xorshift algorithms are known among the fastest random number generators, requiring very small code and state.

Speed

Generating 100 million of random numbers with AOT-compiled binary.

Time (lower is better)	nextInt	nextDouble	nextBool
Random (dart:math)	2323	3107	2264
Xorshift32	1269	1930	1467

Simplicity

It's compatible with the standard Random

import 'package:xrandom/xrandom.dart';

Random xrandom = Xorshift32();

var a = xrandom.nextBool(); 
var b = xrandom.nextDouble();
var c = xrandom.nextInt(n);

Determinism

Xrandom's classes have a deterministic method. By creating the object like that, you'll get same sequence of numbers every time.

test('my test', () {
    final xrandom = Xorshift32.deterministic();
    // run this test twice ;)
    expect(xrandom.nextInt(1000), 119);
    expect(xrandom.nextInt(1000), 240);
    expect(xrandom.nextInt(1000), 369);    
});    

You can achieve the same determinism by creating the Random with a seed argument. However, this does not protect you from dart:math implementation updates.

In contrast to dart:math, xrandom uses xorshift32, a very specific algorithm. Therefore, the predictability of the Xrandom's deterministic sequences can be relied upon. (but not until the library reaches stable release status)

Compatibility

Class	64-bit platforms	JavaScript
Xorshift32	yes	yes
Xorshift128	yes	yes
Xoshiro128pp	yes	yes
Xorshift64	yes	no
Xorshift128p	yes	no

The library has been thoroughly tested to match reference numbers generated by C algorithms. The sources in C are taken directly from scientific publications by George Marsaglia and Sebastiano Vigna, the inventors of the algorithms. The Xorshift128+ results are also matched to reference values from JavaScript xorshift library, that tested the 128+ similarly.

Testing is done in the GitHub Actions cloud on Windows, Ubuntu and macOS in VM and Node.js modes.

Classes

Class	Algorithm	Algorithm author	Published
Xorshift32	xorshift32	G. Marsaglia	2003
Xorshift64	xorshift64	G. Marsaglia	2003
Xorshift128	xorshift128	G. Marsaglia	2003
Xorshift128p	xorshift128+	S. Vigna	2015
Xoshiro128pp	xoshiro128++ 1.0	D. Blackman and S. Vigna	2019

Speed optimizations

nextFloat, unlike nextDouble, does not seek to create the highest quality 64-bit floating-point number. This method prefers speed to accuracy. It transforms a single 32-bit integer into a double. Therefore, the result is limited to a maximum of 2^32-1 values.

Time (lower is better)	nextDouble	nextFloat
Random (dart:math)	3107	-
Xorshift32	1930	696
Xorshift64	2164	1340
Xorshift128	3164	1293
Xorshift128p	3023	1373
Xoshiro128pp	4535	2086

The nextInt32() and nextInt64() do not accept any arguments. They return the raw output of the RNGs.

var xrandom = Xorshift128p(); 
xrandom.nextInt32();  // 32-bit unsigned 
xrandom.nextInt64();  // 64-bit signed

Time (lower is better)	nextInt	nextInt32	nextInt64
Random (dart:math)	2323	-	-
Xorshift32	1269	767	-
Xorshift64	1974	1292	1381
Xorshift128	1967	1301	-
Xorshift128p	2015	1364	1521
Xoshiro128pp	2546	2048	-

More benchmarks

Time (lower is better)	nextInt	nextDouble	nextBool
Random (dart:math)	2323	3107	2264
Xorshift32	1269	1930	1467
Xorshift64	1974	2164	1393
Xorshift128	1967	3164	1479
Xorshift128p	2015	3023	1413
Xoshiro128pp	2546	4535	1500

---

All the benchmarks on this page are from AOT-compiled binaries running on AMD A9-9420e with Ubuntu 20.04. Time is measured in milliseconds.