Xorswift

Xorshift pseudorandom number generator library.
Specialized to generate tons of random numbers.

Xorshift is

• one of the fastest RNGs
• non-cryptographically-secure

See original paper for more details of Xorshift itself.

• Xorshift RNGs

XorshiftGenerator

struct XorshiftGenerator: RandomNumberGenerator {
    /// Create `XorshiftGenerator`.
    /// - precondition: At least one of seeds must be non-zero.
    public init(x: UInt32, y: UInt32, z: UInt32, w: UInt32)

    /// Create `XorshiftGenerator` seeded with another `generator`.
    public init<G: RandomNumberGenerator>(using generator: inout G)

    /// Create `XorshiftGenerator` seeded with `SystemRandomNumberGenerator`.
    public init()
}

Random API available

XorshiftGenerator conforms to RandomNumberGenerator of standard library.
You can use Random APIs with XorshiftGenerator.

var gen = XorshiftGenerator()
Int.random(in: 0..<10, using: &gen)
UInt.random(in: 0..<10, using: &gen)
Float.random(in: 0..<10, using: &gen)
Double.random(in: 0..<10, using: &gen)

let array = [0, 1, 2, 3]
array.randomElement(using: &gen)

Filling UInt32 area

XorshiftGenerator has method fill which fills given area with random UInt32s.
It is faster than arc4random_buf.

func testPerformance_arc4random() {
    let count = 1_000_000
    var a = [UInt32](repeating: 0, count: count)
    measure {
        for _ in 0..<100 {
            arc4random_buf(&a, MemoryLayout<UInt32>.size * a.count)
        }
    } // 0.136sec
}

func testPerformance_xorshift() {
    let count = 1_000_000
    var a = [UInt32](repeating: 0, count: count)
    measure {
        var gen = XorshiftGenerator()
        for _ in 0..<100 {
            gen.fill(start: &a, count: a.count)
        }
    } // 0.067sec
}

Float/Double utility to sample from uniform/normal distributions

XorshiftGenerator also has several methods to sample multiple floating point random numbers from uniform/normal distributions.
See Uniform.swift and Normal.swift for detail.