générateur de nombre aléatoire sous VC++ meilleur que rand ?

C Library Functions                                   drand48(3C)
 
NAME
     drand48,  erand48,  lrand48,  nrand48,   mrand48,   jrand48,
     srand48,  seed48,  lcong48  - generate uniformly distributed
     pseudo-random numbers
 
SYNOPSIS
     #include <stdlib.h>
 
     double drand48(void)
 
     double erand48(unsigned short xi[3] );
 
     long lrand48(void)
 
     long nrand48(unsigned short xi[3] );
 
     long mrand48(void)
 
     long jrand48(unsigned short xi[3] );
 
     void srand48(long seedval);
 
     unsigned short *seed48(unsigned short seed16v[3] );
 
     void lcong48(unsigned short param[7] );
 
DESCRIPTION
     This family of  functions  generates  pseudo-random  numbers
     using  the  well-known linear congruential algorithm and 48-
     bit integer arithmetic.
 
     Functions  drand48()  and  erand48()   return   non-negative
     double-precision floating-point values uniformly distributed
     over the interval [0.0, 1.0].
 
     Functions lrand48() and nrand48() return  non-negative  long
     integers uniformly distributed over the interval [0, 2**31].
 
     Functions  mrand48()  and  jrand48()  return   signed   long
     integers uniformly distributed over the interval [-2**31 , 2
    **31 ].
 
     Functions srand48(), seed48(), and lcong48() are initializa-
     tion  entry  points,  one  of which should be invoked before
     either  drand48(),  lrand48(),  or  mrand48()   is   called.
     (Although  it  is not recommended practice, constant default
     initializer  values  will  be  supplied   automatically   if
     drand48(), lrand48(), or mrand48() is called without a prior
     call to an initialization entry point.) Functions erand48(),
     nrand48(),  and  jrand48()  do not require an initialization
     entry point to be called first.
 
 
 
 
SunOS 5.7           Last change: 22 Jan 1993                    1
 
 
 
 
 
 
C Library Functions                                   drand48(3C)
 
 
 
     All the routines work by generating  a  sequence  of  48-bit
     integer  values,  Xi  , according to the linear congruential
     formula
 
      X n+1= (aX n+c)  mod m n>=0.
 
     The parameter m = 2**48; hence 48-bit integer arithmetic  is
     performed. Unless lcong48() has been invoked, the multiplier
     value a and the addend value c are given by
 
               a = 5DEECE66D16 = 2736731631558
               c = B16 = 138 .
 
     The value  returned  by  any  of  the  functions  drand48(),
     erand48(),  lrand48(), nrand48(), mrand48(), or jrand48() is
     computed by first generating  the  next  48-bit  Xi  in  the
     sequence.  Then the appropriate number of bits, according to
     the type of data item to be returned, are  copied  from  the
     high-order  (leftmost)  bits  of Xi and transformed into the
     returned value.
 
     The functions drand48(), lrand48(), and mrand48() store  the
     last  48-bit  Xi generated in an internal buffer. Xi must be
     initialized prior to being invoked. The functions erand48(),
     nrand48(), and jrand48() require the calling program to pro-
     vide storage for the  successive  Xi  values  in  the  array
     specified  as  an  argument  when the functions are invoked.
     These routines do not have to be  initialized;  the  calling
     program  must place the desired initial value of Xi into the
     array and pass it as an argument. By using  different  argu-
     ments,  functions  erand48(), nrand48(), and jrand48() allow
     separate modules of a  large  program  to  generate  several
     independent  streams  of pseudo-random numbers, that is, the
     sequence of numbers in each stream will not depend upon how
     many times the routines have been called to generate numbers
     for the other streams.
 
     The initializer function srand48() sets  the  high-order  32
     bits  of  Xi  to  the 32 bits contained in its argument. The
     low-order 16 bits of Xi  are  set  to  the  arbitrary  value
     330E16 .
 
     The initializer function seed48() sets the value  of  Xi  to
     the  48-bit  value specified in the argument array. In addi-
     tion, the previous value of  Xi  is  copied  into  a  48-bit
     internal  buffer,  used  only  by seed48(), and a pointer to
     this buffer is the value returned by seed48(). This returned
     pointer,  which can just be ignored if not needed, is useful
     if a program is to be restarted from a given point  at  some
     future  time  - use the pointer to get at and store the last
     Xi value, and then use  this  value  to  reinitialize  using
 
 
 
SunOS 5.7           Last change: 22 Jan 1993                    2
 
 
 
 
 
 
C Library Functions                                   drand48(3C)
 
 
 
     seed48() when the program is restarted.
 
     The initialization function lcong48()  allows  the  user  to
     specify  the  initial  Xi  the  multiplier  value a, and the
     addend value c. Argument array elements  param[0-2]  specify
     Xi, param[3-5] specify the multiplier a, and param[6] speci-
     fies the 16-bit addend c. After lcong48() has been called, a
     subsequent call to either srand48() or seed48() will restore
     the ``standard'' multiplier and  addend  values,  a  and  c,
     specified above.
 
ATTRIBUTES
     See attributes(5) for descriptions of the  following  attri-
     butes:
     ___________________________________________________________
    |       ATTRIBUTE TYPE        |       ATTRIBUTE VALUE       |
    |_____________________________|_____________________________|
    | MT-Level                    | Safe                        |
    |_____________________________|_____________________________|
 
SEE ALSO
     rand(3C), attributes(5)