Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Posting-Version: version B 2.10.1 6/24/83; site kobold.UUCP
Path: utzoo!linus!security!genrad!grkermit!masscomp!kobold!tjt
From: tjt@kobold.UUCP
Newsgroups: net.arch
Subject: Re: Complement Arithmetic
Message-ID: <262@kobold.UUCP>
Date: Fri, 27-Jan-84 21:22:22 EST
Article-I.D.: kobold.262
Posted: Fri Jan 27 21:22:22 1984
Date-Received: Wed, 1-Feb-84 01:32:31 EST
References: <1425@burdvax.UUCP>
Organization: Masscomp, Westford, MA
Lines: 17

As I recall, the biggest problem with "positive and negative zero" in
one's complement arithmetic is the natural tendency to regard the
"positive" zero as the true zero, whereas any additive or subtractive
operation would naturally produce the "negative" zero.  Some machines
attempted to automagically convert this to "positive" zero, but usually
failed for some case or other.  I don't remember the details but I
think it was something to do with a direct implementation of
subtraction of a positive number from a negative number when the
subtraction is implemented differently than complementing the second
operand and adding.

While we're at it, what about sign-magnitude representations?  These
have the same "positive and negative zero" problem, although we seem to
be able to tolerate it for floating point.
-- 
	Tom Teixeira,  Massachusetts Computer Corporation.  Westford MA
	...!{ihnp4,harpo,decvax}!masscomp!tjt   (617) 692-6200 x275