Message-Id: <3.0.6.32.19990220021733.008a69d0@pop.globalserve.net>
X-Sender: derbyshire AT pop DOT globalserve DOT net
X-Mailer: QUALCOMM Windows Eudora Light Version 3.0.6 (32)
Date: Sat, 20 Feb 1999 02:17:33 -0500
To: djgpp AT delorie DOT com
From: Paul Derbyshire <pderbysh AT usa DOT net>
Subject: Re: An inline assembler RNG for C
In-Reply-To: <20678zseao.fsf@Sky.inp.nsk.su>
References: <Paul Derbyshire's message of "Thu, 18 Feb 1999 10:32:18 -0500">
 <George Marsaglia's message of "Wed, 17 Feb 1999 13:35:29 -0500">
 <36CB0BF1 DOT 23FF9E2 AT stat DOT fsu DOT edu>
 <3 DOT 0 DOT 6 DOT 32 DOT 19990218103218 DOT 008d7180 AT pop DOT globalserve DOT net>
Mime-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
Reply-To: djgpp AT delorie DOT com

At 01:27 AM 2/19/99 +0600, you wrote:
>This extreme example does not give much choice for coefficient `a',
>maybe in the original generator good choice of `a' prevents
>such thing, no matter what initial x and c are used, except for x=0,
>c=0.

The original poster stated as much...indeed, he stated the repeat time
should be a very large prime number in the same ballpark as 2^32, using a
from the list of suggestions he included.

-- 
   .*.  "Clouds are not spheres, mountains are not cones, coastlines are not
-()  <  circles, and bark is not smooth, nor does lightning travel in a
   `*'  straight line."    -------------------------------------------------
        -- B. Mandelbrot  |http://surf.to/pgd.net
_____________________ ____|________     Paul Derbyshire     pderbysh AT usa DOT net
Programmer & Humanist|ICQ: 10423848|