yarandom.c

/* yarandom.c -- Yet Another Random Number Generator.
 * Copyright (c) 1997-2010 by Jamie Zawinski <jwz@jwz.org>
 *
 * Permission to use, copy, modify, distribute, and sell this software and its
 * documentation for any purpose is hereby granted without fee, provided that
 * the above copyright notice appear in all copies and that both that
 * copyright notice and this permission notice appear in supporting
 * documentation.  No representations are made about the suitability of this
 * software for any purpose.  It is provided "as is" without express or 
 * implied warranty.
 */

/* The unportable mess that is rand(), random(), drand48() and friends led me
   to ask Phil Karlton <karlton@netscape.com> what the Right Thing to Do was.
   He responded with this.  It is non-cryptographically secure, reasonably
   random (more so than anything that is in any C library), and very fast.

   I don't understand how it works at all, but he says "look at Knuth,
   Vol. 2 (original edition), page 26, Algorithm A.  In this case n=55,
   k=24 and m=2^32."

   So there you have it.

   ---------------------------
   Note: xlockmore 4.03a10 uses this very simple RNG:

	if ((seed = seed % 44488 * 48271 - seed / 44488 * 3399) < 0)
	  seed += 2147483647;
	return seed-1;

   of which it says

	``Dr. Park's algorithm published in the Oct. '88 ACM  "Random Number
	  Generators: Good Ones Are Hard To Find" His version available at
	  ftp://cs.wm.edu/pub/rngs.tar Present form by many authors.''

   Karlton says: ``the usual problem with that kind of RNG turns out to
   be unexepected short cycles for some word lengths.''

   Karlton's RNG is faster, since it does three adds and two stores, while the
   xlockmore RNG does two multiplies, two divides, three adds, and one store.

   Compiler optimizations make a big difference here:
       gcc -O:     difference is 1.2x.
       gcc -O2:    difference is 1.4x.
       gcc -O3:    difference is 1.5x.
       SGI cc -O:  difference is 2.4x.
       SGI cc -O2: difference is 2.4x.
       SGI cc -O3: difference is 5.1x.
   Irix 6.2; Indy r5k; SGI cc version 6; gcc version 2.7.2.1.
 */


#if 0
	#ifdef HAVE_CONFIG_H
	# include "config.h"
	#endif

	#ifdef HAVE_UNISTD_H
	# include <unistd.h>  /* for getpid() */
	#endif
	#include <sys/time.h> /* for gettimeofday() */
#endif

#include "xlockmore.h"

#include "yarandom.h"
# undef ya_rand_init


/* The following 'random' numbers are taken from CRC, 18th Edition, page 622.
   Each array element was taken from the corresponding line in the table,
   except that a[0] was from line 100. 8s and 9s in the table were simply
   skipped. The high order digit was taken mod 4.
 */
#define VectorSize 55
static unsigned int a[VectorSize] = {
 035340171546, 010401501101, 022364657325, 024130436022, 002167303062, /*  5 */
 037570375137, 037210607110, 016272055420, 023011770546, 017143426366, /* 10 */
 014753657433, 021657231332, 023553406142, 004236526362, 010365611275, /* 14 */
 007117336710, 011051276551, 002362132524, 001011540233, 012162531646, /* 20 */
 007056762337, 006631245521, 014164542224, 032633236305, 023342700176, /* 25 */
 002433062234, 015257225043, 026762051606, 000742573230, 005366042132, /* 30 */
 012126416411, 000520471171, 000725646277, 020116577576, 025765742604, /* 35 */
 007633473735, 015674255275, 017555634041, 006503154145, 021576344247, /* 40 */
 014577627653, 002707523333, 034146376720, 030060227734, 013765414060, /* 45 */
 036072251540, 007255221037, 024364674123, 006200353166, 010126373326, /* 50 */
 015664104320, 016401041535, 016215305520, 033115351014, 017411670323  /* 55 */
};

static unsigned int i1 = 0, i2 = 0;

unsigned int
ya_random (void)
{
  int ret = a[i1] + a[i2];
  a[i1] = ret;
  if (++i1 >= VectorSize) i1 = 0;
  if (++i2 >= VectorSize) i2 = 0;
  return ret;
}

void
ya_rand_init(unsigned int seed)
{
  int i;
  if (seed == 0)
    {
    	/*
      struct timeval tp;
#ifdef GETTIMEOFDAY_TWO_ARGS
      struct timezone tzp;
      gettimeofday(&tp, &tzp);
#else
      gettimeofday(&tp);
#endif*/

      /* Since the multiplications will have a larger effect on the
         upper bits than the lower bits, after every addition in the
         seed, perform a bitwise rotate by an odd number, resulting
         in a better distribution of randomness throughout the bits.
         -- Brian Carlson, 2010.
       */
#define ROT(X,N) (((X)<<(N)) | ((X)>>((sizeof(unsigned int)*8)-(N))))
      //seed = (999 * tp.tv_sec);
      //seed = ROT (seed, 11);
      //seed += (1001 * tp.tv_usec);
      //seed = ROT (seed, 7);
      //seed += (1003 * getpid());
      //seed = ROT (seed, 13);

      seed = GetCurrentProcessId();
      seed <<= 16;
      seed ^= (DWORD)(DWORD_PTR)GetForegroundWindow();
      seed <<= 8;
      seed ^= GetTickCount();
    }

  a[0] += seed;
  for (i = 1; i < VectorSize; i++)
    {
      seed = seed*999;
      seed = ROT (seed, 9);
      seed += a[i-1]*1001;
      seed = ROT (seed, 15);
      a[i] += seed;
    }

  i1 = a[0] % VectorSize;
  i2 = (i1 + 24) % VectorSize;
}