Date: Wed, 02 Nov 1994 20:58:01 -0500 (CDT)
From: Aaron Ucko <UCKO AT VAX1 DOT ROCKHURST DOT EDU>
Subject: Re: Random Number Generators
To: elf AT netcom DOT com
Cc: djgpp AT sun DOT soe DOT clarkson DOT edu
Organization: Rockhurst College; Kansas City, MO

>>    int main(int argc, char* argv[])
>>    {
>>       int i;
>>       srand(atoi(argv[1]));
>>       for(i=0; i<30; ++i)
>> 	 printf(" %d", rand()&3);
>>       putchar('\n');
>>    }
>> 
>>         printf(" %d", (rand()>>16)&3);
>> 
>
>If the authors of these pseudo-random number code fragments are
>interested, my research has shown that neither of these techniques is
>very good for selecting uniform deviates (as they are sometimes
>called).  In fact, it is generally not a good idea to select bits from
>a sequence of random numbers because you violate the foundation of the
>number generator.  They generate random sequences of integers, but not
>necessarily random sequences of bits.
>
>You would do much better to partitioning the random number space:
>
>  deviate = rand ();
>  if (deviate < RAND_MAX/3)
>     result = 0;
>  else if (deviate > (RAND_MAX*2)/3)
>     result = 2;
>  else
>     result = 1;
>
>Granted, this is an imperfect partition, but it will more closely
>generate a sequence with a random distribution of 0's, 1's and 2's.

Erm...partititioning into 2^n equal partitions DOES have the effect of
examining only the first n bits, you know.  But why use rand() anyway when you
can use Berkeley's random() or somebody-or-other's rand48(), both of which are
included and much more `random.'  Just my semi-nybble's worth.

--- Aaron Ucko (ucko AT vax1 DOT rockhurst DOT edu; finger for PGP public key) -=- httyp!
-=*=-Just because you're paranoid doesn't mean they aren't out to get you.-=*=-
Geek code 2.1 [finger hayden AT vax1 DOT mankato DOT msus DOT edu for explanation]: 
 GCS/M/S d(-) H s g+ p? !au a-- w+ v+ C++(+++)>++++ U-(S+)>++++ P+ L>++ 3(-) 
 E-(----) !N>++ K- W(--) M-(--) V(--) po-(--) Y+(++) t(+) !5 j R G tv--(-) 
 b+++ !D(--) B--(---) e>++++(*) u++(@) h!() f(+) r-(--)>+++ n+(-) y?