RND.h File Reference

Random Number Distributions, interface. More...

Classes
struct	_RND_lc0
	Structure giving the parameters of a simply linear congruent random number generator. More...
struct	_RND_normal_lc0
	Structure giving the parameters of a simple normal distribution random number generator. More...
struct	_RND_poisson_lc0
	Structure giving the parameters of a poisson distribution. More...
Typedefs
typedef struct _RND_lc0	RND_lc0
	Typedef for the struct _RND_lc0.
typedef struct _RND_normal_lc0	RND_normal_lc0
	Typedef for the struct _RND_normal0.
typedef struct _RND_poisson_lc0	RND_poisson_lc0
	Typedef for the struct _LSI_poisson_lco.
Functions
void	RND_lc0_init (RND_lc0 *ctx, unsigned int seed)
	Init.
unsigned int	RND_lc0_gen (RND_lc0 *ctx)
	Produces a random number in the range 0 - 2**32-1 using a linear congruent method.
int	RND_normal_lc0_gen (RND_normal_lc0 *ctx)
	Produces a normal distribution.

---

Detailed Description

Random Number Distributions, interface.

Author:

JJRussell - russell@slac.stanford.edu

    CVS $Id: RND.h,v 1.1 2004/07/19 20:04:00 ebacho Exp $

SYNOPSIS

Defines the interfaces for a number of common random number distributions.

The casual reader may well ask why provide these here when such objects are readily available on most platforms. The answer is that the algorithms generating these distributions are fully exposed so that they may be readily implemented on a more than one platforms. This means, that given the algorithm and its parameterization, the exact distribution may be generated on any machine. This is valuable when. for example, one generates test patterns using these algorithms and ships the results to another machine for comparison. If the algorithm and its parameterization are stored, the remote machine can exactly reproduce and check the distribution.

---

Typedef Documentation

RND_lc0

Typedef for the struct _RND_lc0.

This random number generator uses the formula

x_{n+1} = (a x_n + c)

Suggested choices are

• a = 69069
• c = 1

The seed value must be odd. It has been selected since it is relatively fast, involving only a multiply and an addition. Probably not a great choice for real math, but for generating random patterns, not bad

RND_normal_lc0

Typedef for the struct _RND_normal0.

This is a very simplistic generator, just adding 6 random numbers from the LC generator

RND_poisson_lc0

Typedef for the struct _LSI_poisson_lco.

This is just a scaled exponential distribution, using a linear congruent

---

Function Documentation

unsigned int RND_lc0_gen

(

RND_lc0 *

ctx

)

Produces a random number in the range 0 - 2**32-1 using a linear congruent method.

Returns:

A random number in the range 0 - 2**32 - 1

Parameters:

ctx

Context parameter

One should scale this number by a suitable factor to produce the desired distribution. No should not use a modulus function. For example, do not do

      r = RND__lc0_args (...) % 256;

To produce a number from 0 - 255. Rather do

     r = RND__lco_args (...) >> 24;

References _RND_lc0::a, _RND_lc0::c, RND__lc0_args(), and _RND_lc0::seed.

void RND_lc0_init	(	RND_lc0 *	ctx,
		unsigned int	seed
	)

Init.

Returns:

Init random number struct to suggested values

Parameters:

	ctx	Context parameter
	seed	random number seed

References _RND_lc0::a, _RND_lc0::c, and _RND_lc0::seed.

Referenced by SSRSIM_init().

int RND_normal_lc0_gen

(

RND_normal_lc0 *

ctx

)

Produces a normal distribution.

Returns:

A random number drawn from a normal distribution

Parameters:

ctx

Context parameter

References _RND_lc0::a, _RND_lc0::c, _RND_normal_lc0::lc0, _RND_normal_lc0::mean, RND__lc0_args(), _RND_lc0::seed, and _RND_normal_lc0::sigma.