From: Jim choate <ravage@bga.com>
Date: Thu, 7 Jul 94 13:21:03 PDT
To: cypherpunks@toad.com
Subject: (fwd) Re: BSD random() - any good (source included)?
Message-ID: <199407072020.PAA29392@ivy.bga.com>
MIME-Version: 1.0
Content-Type: text/plain


Newsgroups: sci.math,sci.stat.math
Path: bga.com!news.sprintlink.net!hookup!ames!lll-winken.llnl.gov!overload.lbl.gov!dog.ee.lbl.gov!ihnp4.ucsd.edu!usc!howland.reston.ans.net!europa.eng.gtefsd.com!MathWorks.Com!news.kei.com!ssd.intel.com!carr
From: carr@ssd.intel.com (George Carr)
Subject: Re: BSD random() - any good (source included)?
Message-ID: <Cs4B6w.IqE@SSD.intel.com>
Sender: usenet@SSD.intel.com
Nntp-Posting-Host: shiva
Organization: Supercomputer Systems Divison, Intel Corp.
References: <deleydCs1L9A.2L6@netcom.com> <iFwLoc3w165w@hakatac.almanac.bc.ca>
Distribution: na
Date: Tue, 28 Jun 1994 17:28:08 GMT
Lines: 34
Xref: bga.com sci.math:14998 sci.stat.math:1241

In article <iFwLoc3w165w@hakatac.almanac.bc.ca>, rbursey@hakatac.almanac.bc.ca (robert bursey) writes:
|> deleyd@netcom.com writes:
|> 
|> > I did a research paper on Computer Generated Random Number Sequences in
|> > 1991.  Included are the results of testing numerous popular generators.
|> > The code used for testing the generators is also available if one
|> > is so inclined to do some testing of a particular generator.  (The only
|> > thing is a thorough test to determine the limits of the generator can
|> > take many hours of CPU time).
|> >
|> > Perhaps later this week I'll post the paper and see what the response
|> > is.  I'm always a bit apprehensive to post.  Never sure what the
|> > response will be.  Maybe someone will think it's interesting.
|> >
|> > David Deley
|> > deleyd@netcom.com
|> 
|> Does anybody know of a good test for randomness? I would definitely like to 
|> know how good computer RNG's are. Post away!

The classic reference is Volume 2 of Donald Knuth's The Art of Computer
Programming, Second Edition, Seminumerical Algorithms. I highly recommend
it to anyone wanting to know what "random" is all about.

If you really need to know whether your generator is random-enough for
your application you should expect to do your own testing and yes it will
require many hours of your time in addition to that of your computer.
-- 
George R. Carr, Jr.
Intel Supercomputer Systems Division    Parallel Systems Engineer
NOAA Forecast Systems Laboratory        carr@ssd.intel.com
Research Lab 3, Rm A227                 carr@neko.fsl.noaa.gov
3100 Marine St, R/E/FS5                 voice: 303-497-6130
Boulder, CO 80303                       fax:   303-497-6821