noiseSource Class Reference

A class that gives access to a RNG on the cpu and gpu. More...

#include <noiseSource.h>

Public Member Functions
	noiseSource ()
	base constructor
int	getInt (int minimum, int maximum)
	Get a reproducible integer.
Dscalar	getRealUniform (Dscalar minimum=0., Dscalar maximum=1.)
	Get a real from uniform distribution.
Dscalar	getRealNormal (Dscalar mean=0., Dscalar std=1.)
	Get a real from normal distribution.
void	initialize (int _N)
	Set the array size of the cuda rngs.
void	setReproducible (bool _rep)
	set reproducibility
void	setReproducibleSeed (int _seed)
	set the seed on a reproducible RNG run
void	initializeGPURNGs (int globalSeed=1337, int tempSeed=0)
	allow for whatever GPU RNG initialization is needed More...

Public Attributes
bool	Reproducible
	should the dynamics be reproducible?
int	N
	number of entries for the cuda RNG
int	RNGSeed
	The seed used by the random number generator, when non-reproducible dynamics have been set.
random_device	rd
	an initializer for non-reproducible random number generation on the cpu
mt19937	gen
	A reproducible Mersenne Twister.
mt19937	genrd
	A non-reproducible Mersenne Twister.
bool	initializeGPURNG
	A flag to determine whether the CUDA RNGs should be initialized or not (so that the program will run on systems with no GPU by setting this to false.
GPUArray< curandState >	RNGs
	An array random-number-generators for use on the GPU branch of the code.

Detailed Description

A class that gives access to a RNG on the cpu and gpu.

Provides features to some psuedo-rng functions. On the CPU side, one can call for a random integer (in a specified range), a random real with a uniform distribution, or a random real from a normal distribution. On the GPU side, provides access to a GPUArray of curandState objects, and functionality to initialize them.

Member Function Documentation

initializeGPURNGs()

void noiseSource::initializeGPURNGs	(	int	globalSeed = 1337,
		int	tempSeed = 0
	)

allow for whatever GPU RNG initialization is needed

Parameters

globalSeed	the global seed to use
offset	the value of the offset that should be sent to the cuda RNG... This is one part of what would be required to support reproducibly being able to load a state from a databse and continue the dynamics in the same way every time. This is not currently supported.

References ArrayHandle< T >::data, access_location::device, GPUArray< T >::getNumElements(), gpu_initialize_RNG_array(), N, access_mode::overwrite, Reproducible, GPUArray< T >::resize(), RNGs, and RNGSeed.

Referenced by updaterWithNoise::setReproducible().

---

The documentation for this class was generated from the following files:

• noiseSource.h
• noiseSource.cpp