Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Posting-Version: version B 2.10 5/3/83; site pyuxss.UUCP
Path: utzoo!watmath!clyde!burl!ulysses!mhuxl!mhuxm!pyuxww!pyuxss!aaw
From: aaw@pyuxss.UUCP (Aaron Werman)
Newsgroups: net.arch
Subject: Re: Complement Arithmetic, -0 as a tag
Message-ID: <262@pyuxss.UUCP>
Date: Wed, 1-Feb-84 10:41:07 EST
Article-I.D.: pyuxss.262
Posted: Wed Feb  1 10:41:07 1984
Date-Received: Tue, 7-Feb-84 08:23:40 EST
References: <5260@uiucdcs.UUCP>
Organization: Central Services Org., Piscataway N.J.
Lines: 11

Why should checking for -0 be significantly faster then checking the twos
compliment tag -2**(word size) - 1 ? This would gain the arithmetic op tag,
solve the compliment problem, and be only slightly higher software
overhead to emulate larger wordsize fixed point arithmetic

The only rational way to do either tag test would be in hardware (as
opposed to interpreter horizontal microcode) in parallel, clogging further
a big bottleneck, register-ALU pathways, which is probably why few of
the ones compliment architectures have -0 propagation traps.
			{harpo,houxm,ihnp4}!pyuxss!aaw
			Aaron Werman