NetNews Offline 2/NetNews Offline Volume 2.iso/news/comp/lang/c-part1/902

Text File | 1996-08-05 | 1.7 KB | 55 lines

Path: mail2news.demon.co.uk!genesis.demon.co.uk
From: Lawrence Kirby <[e-mail]>
Newsgroups: comp.lang.c
Subject: Re: pseudo-random numbers
Date: Wed, 10 Jan 96 00:14:37 GMT
Organization: none
Message-ID: <[e-mail]>
References: <[e-mail]>
Reply-To: [e-mail]
X-NNTP-Posting-Host: genesis.demon.co.uk
X-Newsreader: Demon Internet Simple News v1.27
X-Mail2News-Path: genesis.demon.co.uk

In article <[e-mail]>
           [e-mail] "Jason Stephenson" writes:

>I know what the c.l.c. FAQ says about pseudo-random numbers, so I wonder why
>I seem to get "better" results with "rand() % n + 1" than I get when I use
>"n * rand() / RAND_MAX + 1"  Am I missing some parenthesis or something?

Your code might overflow when calculating n * rand(). Use the versions in
section 13.16 of the FAQ, they don't have this problem. You could also try
the following which should give a even distribution:


#include <stdlib.h>

int randrange(int range)

{
    const int threshold = RAND_MAX - RAND_MAX % range;
    const int divisor = RAND_MAX / range;
    int     randval;

    while ((randval = rand()) >= threshold)
        ;

    return randval / divisor;
}


>Oh yeah, and I usually call "srand(clock())" once in the initialization of
>my program.

Bad idea. On some systems clock() returns values that start at zero at
program startup resulting in a high chance of the same seed on repeated
runs of the program. time() is better (although you don't strictly know
what type time_t is).

--
-----------------------------------------
Lawrence Kirby | [e-mail]
Wilts, England | [e-mail]
-----------------------------------------