From: mert0407 AT sable DOT ox DOT ac DOT uk (George Foot)
Newsgroups: comp.os.msdos.djgpp
Subject: Re: Random numbers
Date: 2 Jun 1997 06:34:15 GMT
Organization: Oxford University, England
Message-ID: <5mtph7$b3v@news.ox.ac.uk>
References: <01bc6ce9$72076bc0$363e63c3 AT 8652hvt73761> <5mr4ms$ipp AT freenet-news DOT carleton DOT ca>
NNTP-Posting-Host: sable.ox.ac.uk
Lines: 44
To: djgpp AT delorie DOT com
DJ-Gateway: from newsgroup comp.os.msdos.djgpp
Precedence: bulk

Paul Derbyshire (ao950 AT FreeNet DOT Carleton DOT CA) wrote:

: For values of x not a power of 2,

: random()%x

: is not as good as

: random()/(MAXINT/x)

As it stands, I disagree with this; the integer division MAXINT/x will
generally produce rounding errors which could be awkward (you will get
numbers which are too large). Performing the entire calucation cast to
double, though, can you justify this being a better generator? I remember
a thread about this a while ago, and I posted there what I thought was a
solution which would produce a fair selection of numbers. I was planning
to do some tests on it but never got around to it. 

My ideal function at the moment:

int gfrandom(int range)
{
 int a,b;
 a=1;
 while (a<range) a<<=1;
 a--;
 b=range;
 while (b>=range) b=random()&a;
 return b;
}

: Also, there's a name for a thousand million, "billion", which is somewhat
: less long and awkward :)

Fair enough, but note that in America a billion is 10^9 whereas (certainly
in the past) in Britain it is/was 10^12, which is more logical IMHO. I
tend to write `thousand million' to remove any ambiguity. I think now
`Britain' may sadly have changed to fit the American definition, which I
feel is less logical; but we shouldn't be dealing with units of 10^3
anyway...

-- 
George Foot <mert0407 AT sable DOT ox DOT ac DOT uk>
Merton College, Oxford