Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Path: utzoo!mnetor!seismo!rutgers!ames!oliveb!pyramid!batcomputer!braner
From: braner@batcomputer.tn.cornell.edu (braner)
Newsgroups: comp.sys.atari.st
Subject: Re: problems with rnd# from assembler
Message-ID: <1084@batcomputer.tn.cornell.edu>
Date: Thu, 21-May-87 00:48:21 EDT
Article-I.D.: batcompu.1084
Posted: Thu May 21 00:48:21 1987
Date-Received: Sat, 23-May-87 13:19:59 EDT
References: <2088@pbhya.UUCP> <1021@viper.Lynx.MN.ORG> <2591@zen.berkeley.edu>
Reply-To: braner@tcgould.tn.cornell.edu.UUCP (braner)
Organization: Cornell Theory Center, Cornell University, Ithaca NY
Lines: 13
Keywords: midi/rnd#
Summary: I have my own random number generator

[]

I don't know whether xbios(17) works or not: I don't trust system-supplied
random number generators for statistical reasons (and purpose).

I have my own assembly-langauge random generator that yields a 32-bit
number in about 20 microseconds, and it passed all the statistical tests
I tried (uniformity, runs in halftiles, 2d lagged-correlation plots...).
(If you only need 16 bits you can make it faster still!)
I also have an AL extension to convert it to a FP number in (0,1).  It takes
about 60 microsecs to call from C and get a double var stuffed.

- Moshe Braner