OpenEuphoria

Re: 32-bit random numbers

• Posted by Tommy Carlier <tommy.carlier at pandora.be> Jul 03, 2004
• 590 views

Juergen Luethje wrote:
> Hi, this function can not take a big 32-bit integer atom as argument.
> If n > 2*#3FFFFFFF, then (n - #3FFFFFFF) > #3FFFFFFF, that means the
> argument to the first call of rand() will be too big, and Euphoria
> aborts with a syntax error.
> 
> Also, I want the rand32() function to behave like Euphoria's built-in
> rand() function. That means, the returned values must have a uniform
> distribution. Are you sure, that the sum
>       rand(n - #3FFFFFFF) + rand(#3FFFFFFF)
> is uniformly distributed?
> 
> Sorry, I don't have a statistical test for this at hand ATM.

You're totally right. I just typed in some function here, without really
thinking about it a lot. I just wanted to show how easy it would be to create a
function that generates 32-bit random numbers. The function I gave was just to
push others in the right direction.

As Robert already pointed out: the returned values are indeed not uniformly
distributed. Instead of adding 2 random integers, a different formula should be
used. I'm sure there are enough smart Euphorians to solve this problem.

--
tommy online: http://users.pandora.be/tommycarlier
Euphoria Message Board: http://uboard.proboards32.com