math.rng

Class XorShift1024Star

java.lang.Object

math.rng.AbstractRng64

math.rng.XorShift1024Star

All Implemented Interfaces:

PseudoRandom, PseudoRandomStream, SplittablePseudoRandom

public class XorShift1024Star
extends AbstractRng64
implements SplittablePseudoRandom

1024-bit Xorshift pseudo random generator suggested by George Marsaglia (2003) as studied by Sebastiano Vigna (2016). It has better statistical properties than XorShift64Star and other 64-bit xorshift generators and is about 25% faster despite its larger state size.

This generator has a period of 2¹⁰²⁴ − 1.

At least the two lowest bits of a xorshift* generator satisfy a linear recurrence and will therefore fail linearity tests (MatrixRank and LinearComp in TestU01) which were designed to catch these linear relations. However, low linear complexity of the lowest bits has rarely an effect in practice, and certainly has no impact at all when the generator is only used to generate floating-point numbers.

This generator is 16-dimensionally equidistributed.

Field Summary

Fields inherited from class math.rng.AbstractRng64

DOUBLE_NORM, FLOAT_NORM, nextGaussian, unused

Constructor Summary

Constructors

Modifier	Constructor and Description
	XorShift1024Star()
	XorShift1024Star(long seed)
	XorShift1024Star(long[] seed)
protected	XorShift1024Star(long[] seed, boolean unused)

Method Summary

Modifier and Type	Method and Description
static SplittablePseudoRandom	getDefault()
long	nextLong()
XorShift1024Star	split() Constructs and returns a new SplittablePseudoRandom instance that shares no mutable state with the current instance.

Methods inherited from class math.rng.AbstractRng64

beta, beta, cauchy, cauchy, chiSquare, chiSquare, doubles, doubles, doubles, doubles, exponential, exponential, gamma, gamma, getAlgorithm, getSeed, ints, ints, ints, ints, logNormal, logNormal, longs, longs, longs, longs, next, nextBoolean, nextBytes, nextDouble, nextDouble, nextDoubles, nextFloat, nextFloat, nextGaussian, nextGaussian, nextInt, nextInt, nextInt, nextLong, nextLong, nextLongs, normal, normal, saveSeed, saveSeed, studentT, studentT, weibull, weibull

Methods inherited from class java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Methods inherited from interface math.rng.PseudoRandom

getAlgorithm, getSeed, next, nextBoolean, nextBytes, nextDouble, nextDouble, nextDoubles, nextFloat, nextFloat, nextGaussian, nextGaussian, nextInt, nextInt, nextInt, nextLong, nextLong, nextLongs

Methods inherited from interface math.rng.PseudoRandomStream

beta, beta, cauchy, cauchy, chiSquare, chiSquare, doubles, doubles, doubles, doubles, exponential, exponential, gamma, gamma, ints, ints, ints, ints, logNormal, logNormal, longs, longs, longs, longs, normal, normal, studentT, studentT, weibull, weibull

Constructor Detail

XorShift1024Star

public XorShift1024Star()

XorShift1024Star

public XorShift1024Star(long seed)

XorShift1024Star

public XorShift1024Star(long[] seed)

XorShift1024Star

protected XorShift1024Star(long[] seed,
                           boolean unused)

Method Detail

nextLong

public final long nextLong()

Specified by:

nextLong in interface PseudoRandom

Specified by:

nextLong in class AbstractRng64

split

public XorShift1024Star split()

Description copied from interface: SplittablePseudoRandom

Constructs and returns a new SplittablePseudoRandom instance that shares no mutable state with the current instance. However, with very high probability, the values collectively generated by the two instances have the same statistical properties as if the same quantity of values were generated by a single thread using a single SplittablePseudoRandom instance. Instances of SplittablePseudoRandom are not thread-safe. They are designed to be split, not shared, across threads.

Specified by:

split in interface SplittablePseudoRandom

Returns:

a new SplittablePseudoRandom instance that shares no mutable state with the current instance

getDefault

public static SplittablePseudoRandom getDefault()