Philox4x32 Class

public sealed class Philox4x32 : CounterBasedRandomSource<CounterBasedGenerator32<Philox4x32Engine>, uint>, 
	IRandomSourceFactory<Philox4x32>, IJumpable, IAdvanceable

Public NotInheritable Class Philox4x32
	Inherits CounterBasedRandomSource(Of CounterBasedGenerator32(Of Philox4x32Engine), UInteger)
	Implements IRandomSourceFactory(Of Philox4x32), IJumpable, 
	IAdvanceable

public ref class Philox4x32 sealed : public CounterBasedRandomSource<CounterBasedGenerator32<Philox4x32Engine>, unsigned int>, 
	IRandomSourceFactory<Philox4x32^>, IJumpable, IAdvanceable

[<SealedAttribute>]
type Philox4x32 = 
    class
        inherit CounterBasedRandomSource<CounterBasedGenerator32<Philox4x32Engine>, uint32>
        interface IRandomSourceFactory<Philox4x32>
        interface IJumpable
        interface IAdvanceable
    end

Inheritance

Object  →  RandomSource<CounterBasedGenerator32<Philox4x32Engine>>  →  CounterBasedRandomSource<CounterBasedGenerator32<Philox4x32Engine>, UInt32>  →  Philox4x32

Implements

IAdvanceable, IJumpable, IRandomSourceFactory<Philox4x32>

Use the Philox4x32 class to represent a counter-based pseudo-random number generator that uses the Philox 4x32 algorithm by John K. Salmon et al.

Philox is a counter-based random number generator that produces a bijection from an n-bit input counter to an n-bit random output. This allows for direct computation of the Nth random number without computing all previous numbers.

Key features include excellent statistical quality (passes BigCrush), reproducibility, and the ability to generate arbitrary independent streams by setting different keys.

Counter-based generators are particularly useful for parallel computing where each thread can have its own counter range, ensuring no overlap in sequences.

Performance: Philox4x32 generates random numbers in blocks of 4, making it efficient for bulk generation.

Use cases:

• Parallel computing with independent, non-overlapping sequences
• Reproducible simulations requiring random access to sequences
• Applications requiring deterministic stream partitioning

Reference: John K. Salmon, Mark A. Moraes, Ron O. Dror, and David E. Shaw. "Parallel random numbers: as easy as 1, 2, 3." SC 2011.

Reference