Path: utzoo!attcan!utgpu!jarvis.csri.toronto.edu!mailrus!wuarchive!udel!burdvax!dave
From: dave@PRC.Unisys.COM (David Lee Matuszek)
Newsgroups: comp.lang.misc
Subject: Re: Fast conversions, another urban myth?
Message-ID: <11550@burdvax.PRC.Unisys.COM>
Date: 25 Sep 89 18:53:55 GMT
References: <832@dms.UUCP> <688@UALTAVM.BITNET> <136@bbxsda.UUCP> <4125@cbnewsh.ATT.COM> <150@bbxsda.UUCP> <826@maxim.erbe.se>
Sender: news@PRC.Unisys.COM
Organization: Unisys Corporation, Paoli Research Center; Paoli, PA
Lines: 26

In article <826@maxim.erbe.se> prc@erbe.se (Robert Claeson) writes:

>Gimme a library of accurate, infinite-precision arithmetic routines, anytime.

What, no smiley face?  Well, maybe you're serious, then.  Just in case
you are...

Robert Gregory and another numerical analyst (whose name currently
escapes me) worked out a reasonably efficient way to perform
mathematical operations with zero loss of precision.  They wrote a
book on the subject, published, I think, by Prentice-Hall.  At one
time I sorta understood it, so conceptually it can't be all that
difficult.

The method uses a very nonstandard representation of numbers, so it
really ought to be implemented at the hardware level; but it seemed
efficient enough to be feasible to implement in software, if you were
really serious about accuracy.  (The expensive part is converting
to/from their representation.)  I'm a little surprised that no one has
ever done anything with it--but then I'm not a numerical analyst
myself (anything but!), so maybe I wouldn't have heard anyway.

-- Dave Matuszek (dave@prc.unisys.com)
-- Unisys Corp. / Paoli Research Center / PO Box 517 / Paoli PA  19301
-- Any resemblance between my opinions and those of my employer is improbable.
* 20th anniversary?  Yeah, but it's 17 years since the LAST man on the moon! *