yjs/lib/prng/PRNG/Mt19937.js
2018-11-25 03:17:00 +01:00

68 lines
1.9 KiB
JavaScript

/**
* @module prng
*/
const N = 624
const M = 397
const twist = (u, v) => ((((u & 0x80000000) | (v & 0x7fffffff)) >>> 1) ^ ((v & 1) ? 0x9908b0df : 0))
const nextState = (state) => {
let p = 0
let j
for (j = N - M + 1; --j; p++) {
state[p] = state[p + M] ^ twist(state[p], state[p + 1])
}
for (j = M; --j; p++) {
state[p] = state[p + M - N] ^ twist(state[p], state[p + 1])
}
state[p] = state[p + M - N] ^ twist(state[p], state[0])
}
/**
* This is a port of Shawn Cokus's implementation of the original Mersenne Twister algorithm (http://www.math.sci.hiroshima-u.ac.jp/~m-mat/MT/MT2002/CODES/MTARCOK/mt19937ar-cok.c).
* MT has a very high period of 2^19937. Though the authors of xorshift describe that a high period is not
* very relevant (http://vigna.di.unimi.it/xorshift/). It is four times slower than xoroshiro128plus and
* needs to recompute its state after generating 624 numbers.
*
* @example
* const gen = new Mt19937(new Date().getTime())
* console.log(gen.next())
*
* @public
*/
export class Mt19937 {
/**
* @param {Number} seed The starting point for the random number generation. If you use the same seed, the generator will return the same sequence of random numbers.
*/
constructor (seed) {
this.seed = seed
const state = new Uint32Array(N)
state[0] = seed
for (let i = 1; i < N; i++) {
state[i] = (Math.imul(1812433253, (state[i - 1] ^ (state[i - 1] >>> 30))) + i) & 0xFFFFFFFF
}
this._state = state
this._i = 0
nextState(this._state)
}
/**
* Generate a random signed integer.
*
* @return {Number} A 32 bit signed integer.
*/
next () {
if (this._i === N) {
// need to compute a new state
nextState(this._state)
this._i = 0
}
let y = this._state[this._i++]
y ^= (y >>> 11)
y ^= (y << 7) & 0x9d2c5680
y ^= (y << 15) & 0xefc60000
y ^= (y >>> 18)
return y
}
}