template<typename T>

tf::Xorshift class

class to create a fast xorshift-based pseudo-random number generator

Template parameters
T	unsigned integral type used as the internal state (supported uint32_t and uint64_t)

This class implements a lightweight xorshift pseudo-random number generator suitable for performance-critical paths such as schedulers, work-stealing victim selection, and randomized backoff. The implementation is branchless on the hot path and has a very small state footprint (one machine word). All operations are integer-only.

Constructors, destructors, conversion operators

Xorshift() defaulted

constructs an uninitialized Xorshift generator

Xorshift(T value)

constructs a Xorshift generator with the given seed

Public functions

void seed(T value)

seeds the generator with a new value

auto operator()() -> T

generates the next pseudo-random value

Function documentation

template<typename T>

tf::Xorshift<T>::Xorshift() defaulted

constructs an uninitialized Xorshift generator

The internal state is not initialized. The user must call seed() with a non-zero value before generating numbers.

template<typename T>

tf::Xorshift<T>::Xorshift(T value)

constructs a Xorshift generator with the given seed

Parameters
value	the new seed value to use

The seed value must be non-zero.

template<typename T>

void tf::Xorshift<T>::seed(T value)

seeds the generator with a new value

Parameters
value	the new seed value

The seed value must be non-zero. A zero seed results in a degenerated generator that always returns zero.

template<typename T>

T tf::Xorshift<T>::operator()()

generates the next pseudo-random value

Returns	a pseudo-random value of type T

For 32-bit state, this function implements the Xorshift32 algorithm. For 64-bit state, this function implements the Xorshift64 algorithm with a multiplicative output transformation to improve distribution.