dSFMT.c File Reference

double precision SIMD-oriented Fast Mersenne Twister (dSFMT) based on IEEE 754 format. More...

#include <string.h>
#include <assert.h>
#include "dSFMT.h"
#include "dSFMT-params.h"

Data Structures
union	W128_T
	128-bit data structure More...
Typedefs
typedef W128_T	w128_t
	128-bit data type
Functions
static void	lshift128 (w128_t *out, const w128_t *in, int shift)
	This function simulates SIMD 128-bit left shift by the standard C.
static void	gen_rand_all (void)
	This function fills the internal state array with double precision floating point pseudorandom numbers of the IEEE 754 format.
static void	gen_rand_array (w128_t array[], int size)
	This function fills the user-specified array with double precision floating point pseudorandom numbers of the IEEE 754 format.
static uint32_t	ini_func1 (uint32_t x)
	This function represents a function used in the initialization by init_by_array.
static uint32_t	ini_func2 (uint32_t x)
	This function represents a function used in the initialization by init_by_array.
static void	convert_co (w128_t array[], int size)
	This function converts the double precision floating point numbers which distribute uniformly in the range [1, 2) to those which distribute uniformly in the range [0, 1).
static void	convert_oc (w128_t array[], int size)
	This function converts the double precision floating point numbers which distribute uniformly in the range [1, 2) to those which distribute uniformly in the range (0, 1].
static void	convert_oo (w128_t array[], int size)
	This function converts the double precision floating point numbers which distribute uniformly in the range [1, 2) to those which distribute uniformly in the range (0, 1).
static int	sfmt_idxof (int i)
	This function simulate a 32-bit array index overlapped to 64-bit array of LITTLE ENDIAN in BIG ENDIAN machine.
static void	initial_mask (void)
	This function initializes the internal state array to fit the IEEE 754 format.
static void	period_certification ()
	This function certificate the period of 2^{SFMT_MEXP}-1.
static void	do_recursion (w128_t *r, w128_t *a, w128_t *b, w128_t *c, w128_t *lung)
	This function represents the recursion formula.
const char *	get_idstring (void)
	This function returns the identification string.
int	get_min_array_size (void)
	This function returns the minimum size of array used for fill_array functions.
double	genrand_close1_open2 (void)
	This function generates and returns double precision pseudorandom number which distributes uniformly in the range [1, 2).
void	fill_array_close1_open2 (double array[], int size)
	This function generates double precision floating point pseudorandom numbers which distribute in the range [1, 2) to the specified array[] by one call.
void	fill_array_open_close (double array[], int size)
	This function generates double precision floating point pseudorandom numbers which distribute in the range (0, 1] to the specified array[] by one call.
void	fill_array_close_open (double array[], int size)
	This function generates double precision floating point pseudorandom numbers which distribute in the range [0, 1) to the specified array[] by one call.
void	fill_array_open_open (double array[], int size)
	This function generates double precision floating point pseudorandom numbers which distribute in the range (0, 1) to the specified array[] by one call.
void	init_gen_rand (uint32_t seed)
	This function initializes the internal state array with a 32-bit integer seed.
void	init_by_array (uint32_t init_key[], int key_length)
	This function initializes the internal state array, with an array of 32-bit integers used as the seeds.
Variables
static w128_t	sfmt [SFMT_N+1]
	the 128-bit internal state array
static double *	psfmt64 = &sfmt[0].d[0]
	the double pointer to the 128-bit internal state array
static int	sfmt_idx
	index counter to the internal state array as double
static int	is_sfmt_initialized = 0
	a flag: it is 0 if and only if the internal state is not yet initialized.

---

Detailed Description

double precision SIMD-oriented Fast Mersenne Twister (dSFMT) based on IEEE 754 format.

Author:

Mutsuo Saito (Hiroshima University)

Makoto Matsumoto (Hiroshima University)

Copyright (C) 2007 Mutsuo Saito, Makoto Matsumoto and Hiroshima University. All rights reserved.

The new BSD License is applied to this software, see LICENSE.txt

---

Typedef Documentation

typedef union W128_T w128_t

128-bit data type

---

Function Documentation

static void convert_co	(	w128_t	array[],
		int	size
	)			[inline, static]

This function converts the double precision floating point numbers which distribute uniformly in the range [1, 2) to those which distribute uniformly in the range [0, 1).

Parameters:

	array	array of double precision floating point numbers
	size	size of the array

static void convert_oc	(	w128_t	array[],
		int	size
	)			[inline, static]

This function converts the double precision floating point numbers which distribute uniformly in the range [1, 2) to those which distribute uniformly in the range (0, 1].

Parameters:

	array	array of double precision floating point numbers
	size	size of the array

static void convert_oo	(	w128_t	array[],
		int	size
	)			[inline, static]

This function converts the double precision floating point numbers which distribute uniformly in the range [1, 2) to those which distribute uniformly in the range (0, 1).

Parameters:

	array	array of double precision floating point numbers
	size	size of the array

static void do_recursion	(	w128_t *	r,
		w128_t *	a,
		w128_t *	b,
		w128_t *	c,
		w128_t *	lung
	)			[inline, static]

This function represents the recursion formula.

Parameters:

	r	output
	a	a 128-bit part of the internal state array
	b	a 128-bit part of the internal state array
	c	a 128-bit part of the internal state array
	lung	a 128-bit part of the internal state array

void fill_array_close1_open2	(	double	array[],
		int	size
	)

This function generates double precision floating point pseudorandom numbers which distribute in the range [1, 2) to the specified array[] by one call.

The number of pseudorandom numbers is specified by the argument size, which must be at least (SFMT_MEXP / 128) * 2 and a multiple of two. The function get_min_array_size() returns this minimum size. The generation by this function is much faster than the following fill_array_xxx functions.

For initialization, init_gen_rand() or init_by_array() must be called before the first call of this function. This function can not be used after calling genrand_xxx functions, without initialization.

Parameters:

	array	an array where pseudorandom numbers are filled by this function. The pointer to the array must be "aligned" (namely, must be a multiple of 16) in the SIMD version, since it refers to the address of a 128-bit integer. In the standard C version, the pointer is arbitrary.
	size	the number of 64-bit pseudorandom integers to be generated. size must be a multiple of 2, and greater than or equal to (SFMT_MEXP / 128) * 2.

Note:

memalign or posix_memalign is available to get aligned memory. Mac OSX doesn't have these functions, but malloc of OSX returns the pointer to the aligned memory block.

void fill_array_close_open	(	double	array[],
		int	size
	)

This function generates double precision floating point pseudorandom numbers which distribute in the range [0, 1) to the specified array[] by one call.

This function is the same as fill_array_close1_open2() except the distribution range.

Parameters:

	array	an array where pseudorandom numbers are filled by this function.
	size	the number of pseudorandom numbers to be generated. see also

See also:

fill_array_close1_open2()

void fill_array_open_close	(	double	array[],
		int	size
	)

This function generates double precision floating point pseudorandom numbers which distribute in the range (0, 1] to the specified array[] by one call.

This function is the same as fill_array_close1_open2() except the distribution range.

Parameters:

	array	an array where pseudorandom numbers are filled by this function.
	size	the number of pseudorandom numbers to be generated. see also

See also:

fill_array_close1_open2()

void fill_array_open_open	(	double	array[],
		int	size
	)

This function generates double precision floating point pseudorandom numbers which distribute in the range (0, 1) to the specified array[] by one call.

This function is the same as fill_array_close1_open2() except the distribution range.

Parameters:

	array	an array where pseudorandom numbers are filled by this function.
	size	the number of pseudorandom numbers to be generated. see also

See also:

fill_array_close1_open2()

static void gen_rand_all

(

void

)

[inline, static]

This function fills the internal state array with double precision floating point pseudorandom numbers of the IEEE 754 format.

static void gen_rand_array	(	w128_t	array[],
		int	size
	)			[inline, static]

This function fills the user-specified array with double precision floating point pseudorandom numbers of the IEEE 754 format.

Parameters:

	array	an 128-bit array to be filled by pseudorandom numbers.
	size	number of 128-bit pseudorandom numbers to be generated.

double genrand_close1_open2

(

void

)

[inline]

This function generates and returns double precision pseudorandom number which distributes uniformly in the range [1, 2).

This is the primitive and faster than generating numbers in other ranges. init_gen_rand() or init_by_array() must be called before this function.

Returns:

double precision floating point pseudorandom number

const char* get_idstring

(

void

)

This function returns the identification string.

The string shows the Mersenne exponent, and all parameters of this generator.

Returns:

id string.

int get_min_array_size

(

void

)

This function returns the minimum size of array used for fill_array functions.

Returns:

minimum size of array used for fill_array functions.

static uint32_t ini_func1

(

uint32_t

x

)

[inline, static]

This function represents a function used in the initialization by init_by_array.

Parameters:

x

32-bit integer

Returns:

32-bit integer

static uint32_t ini_func2

(

uint32_t

x

)

[inline, static]

This function represents a function used in the initialization by init_by_array.

Parameters:

x

32-bit integer

Returns:

32-bit integer

void init_by_array	(	uint32_t	init_key[],
		int	key_length
	)

This function initializes the internal state array, with an array of 32-bit integers used as the seeds.

Parameters:

	init_key	the array of 32-bit integers, used as a seed.
	key_length	the length of init_key.

void init_gen_rand

(

uint32_t

seed

)

This function initializes the internal state array with a 32-bit integer seed.

Parameters:

seed

a 32-bit integer used as the seed.

void initial_mask

(

void

)

[static]

This function initializes the internal state array to fit the IEEE 754 format.

static void lshift128	(	w128_t *	out,
		const w128_t *	in,
		int	shift
	)			[inline, static]

This function simulates SIMD 128-bit left shift by the standard C.

The 128-bit integer given in in is shifted by (shift * 8) bits. This function simulates the LITTLE ENDIAN SIMD.

Parameters:

	out	the output of this function
	in	the 128-bit data to be shifted
	shift	the shift value

static void period_certification

(

void

)

[static]

This function certificate the period of 2^{SFMT_MEXP}-1.

static int sfmt_idxof

(

int

i

)

[inline, static]

This function simulate a 32-bit array index overlapped to 64-bit array of LITTLE ENDIAN in BIG ENDIAN machine.

---

Variable Documentation

int is_sfmt_initialized = 0 [static]

a flag: it is 0 if and only if the internal state is not yet initialized.

double* psfmt64 = &sfmt[0].d[0] [static]

the double pointer to the 128-bit internal state array

w128_t sfmt[SFMT_N+1] [static]

the 128-bit internal state array

int sfmt_idx [static]

index counter to the internal state array as double

---