Message-Id:	<2.2.32.19970530194714.006c3788@gate>
Mime-Version:	1.0
Date:	Fri, 30 May 1997 12:47:14 -0700
To:	"Matthew Bennett" <Bennett AT btinternet DOT com>
From:	"Alan Wilson" <awilson AT wilshire DOT com>
Subject:	Re: Random numbers
Cc:	djgpp AT delorie DOT com

At 11:09 AM 5/30/97 GMT, you wrote:
>Hi all,
>
>When using the 'random' command in the form:
>
>printf("%d\n", random() & 10);
>
>The numbers produced are always either 0, 2, 8 or 10. If I use say 2 as my
>range, the numbers produced are only 0 or 2.  I don't seem to be able to
>get a well varied set of random numbers (i.e. all the numbers in the range
>eventually appear). Using the command form:
>
>prinf("%d\n", random(10));
>
>produces the same sequence of very large numbers - whatever range I use.
>
>Does anyone know how I can get a sequence of in-range, varied numbers?
>For what I'm writing, it would also need to be seeded (the 'srandom'
>command is what I'm currently using).
>
>Thanks,


The following is some comments that some people sent me after I asked a
similar qustion about random numbers...Hope it helps.
------------------------

"if you were more
interested in algorithms to generate random numbers I'd suggest you get
(or rather, borrow) Knuth's Seminumerical Algorithms - The Art of Computer
Programming, Volume 2. Half the book is about random numbers, how to
generate them, how random they are, etc. Only get this if you are heavily
into mathematics, though!

You could, of course, also look at the source code for the rand[om]()
functions in libc.

-- 
George Foot <mert0407 AT sable DOT ox DOT ac DOT uk>"
-------------------

"On getting random numbers :

If you do a histogram of rand() % 100 or random() % 100 you will 
find that its not uniform. If you image that rand return a uniform 
number from 0-31 and you % 10 you map 

[ rand val ] -> (val % 10)
[0,10,20,30] -> 0
[1,11,21,31] -> 1
[2,12,22]    -> 2
[3,13,23]    -> 3
[4,14,24]    -> 4
  . . .         .
[9,19,29]    -> 9


Depending on what your doing it may be close enough not to matter. 

Somthing like this gets rid of this problem. 

int rnd(int lo,int hi)
{
   int rng = hi - lo + 1;
   return ( lo + rng * rand() / MAX_INT );
}


I only spent a couple days trying to find why a modeler I was 
writing always favored low number 8^)....eheft AT dnaco DOT net (Eric Heft)"
----------------------------

"The easyest way is to use the random function from the stdlib library. like so

x = random() % 100;

You can make more realistic random numbers by using the current time as a
seed. I can't remember the seed command but I think it's randomize(x).

Hope this is helpfull, you should also consult the documentation that comes
with DJGPP.

Cheers,
 Timothy Gunn <n2209977 AT student DOT fit DOT qut DOT edu DOT au>"

Alan Wilson

- Raw text -