Re: Shock: rand function is not uniform!

In message <[email protected]>
          James Taylor <[email protected]> wrote:

> On Mon 18 Feb, Richard Proctor wrote:
> >
> > The rand function in the current riscos perl is based on the C
> > function "rand", which is well known to be less the ideal.
>
> Less than ideal? That's rather an understatement. Did you run
> the program I attached which plots a graph of its distribution?
> >From the graph it is clear that the probability of getting a
> value in the middle of rand's range is around about a *quarter*
> of the probability of getting values at the two extremes of
> its range. Now I don't think you could describe that as
> anything less than an unmitigated disaster!

Did you run it over an infinite period of time ? The only thing
that is clear from empirical evidence is that on that run, things
were that way. Whilst it might be less than ideal, a simple
incrementing function would satisfy your test, and still be less
than ideal.

> > if/when we move riscos perl to the more recent version which
> > uses the C fucntion rand48 which is more random.
>
> I really don't think the problem with Perl's rand has anything
> to do with any minor randomness failings in C's rand. I have

You keep saying 'C's rand'; I don't know what you mean. C's rand
is implementation specific, surely. If you mean UnixLib rand, please say
that.

> written an equivalent test program in C and it is clear from
> the graph that whatever failings C's rand might have, it is

Again, unclear... I don't know what compiler or library you've linked
against and it's impossible to give any comment.

> at least approximately uniform. The same goes for C's
> _ANSI_rand() function which is supposed to be even less

C doesn't have any such function. UnixLib might; be clear.

> good. I have also tested Perl's rand function on various
> Unix boxes and found that it behaves properly. In comparison,
> RISC OS Perl's rand function is *so* broken that I can only
> think that there's a bug in the way it was implemented.

Sorry if that seems critical, but your comments are unclear.

-- 
Gerph {djf0-.3w6e2w2.226,6q6w2q2,2.3,2m4}
URL: http://www.movspclr.co.uk/
.... Eyes to the heavens, screaming at the sky;
    Trying to send you messages, but choking on goodbye.