Path: utzoo!utgpu!jarvis.csri.toronto.edu!mailrus!purdue!ames!oliveb!apple!desnoyer
From: desnoyer@Apple.COM (Peter Desnoyers)
Newsgroups: comp.lang.c
Subject: Re: Randomness of MSC 5.1 rand()
Message-ID: <30294@apple.Apple.COM>
Date: 8 May 89 17:31:50 GMT
References: <1598@oregon.uoregon.edu>
Organization: Apple Computer Inc, Cupertino, CA
Lines: 19

In article <1598@oregon.uoregon.edu> michelbi@oregon.uoregon.edu (Michel Biedermann) writes:
>How good (i.e. random) is the rand() function in MSC 5.1/QC 2.0?

Here's one easy test:
   int i; double sum=0;			
   for (i = 0; i < 20; i++){
     while( (double)rand()/MAX_RAND > 0.0001);	/* wait */
     sum += (double)rand()/MAX_RAND;	/* then do something random */
   }
   printf( "%f\n", sum/20.0);

On the VAX I use (BSD 4.3?) I get 0.44, on Mac MPW I get 0.43. (both
are reasonable.) On at least one version of Turbo Pascal for the PC
you will get 0.25 or less. (my boss had this persistant bug in a small
simulation of random packet arrivals - the average packet size was
max/4 or max/8, rather than max/2. Turns out his code looked like the
code above.)

				Peter Desnoyers