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!decvax!genrad!grkermit!masscomp!kobold!tjt
From: tjt@kobold.UUCP
Newsgroups: net.arch
Subject: Re: Re: Complement Arithmetic - (nf)
Message-ID: <271@kobold.UUCP>
Date: Mon, 6-Feb-84 11:19:10 EST
Article-I.D.: kobold.271
Posted: Mon Feb  6 11:19:10 1984
Date-Received: Thu, 9-Feb-84 07:46:31 EST
References: <5354@uiucdcs.UUCP>
Organization: Masscomp, Westford, MA
Lines: 25

parsec!wallach wants to why the IEEE floating point standard uses a
sign-magnitude representation while representing integers in
2's-complement notation.

I can't speak for the IEEE specifiers, but I always thought the basic
floating point representation of:

			----------------------
			| S | EXP | Mantissa |
			----------------------
							   EXP-offset
where S is a sign bit and the mantissa gets multiplied by B
used the offset notation of the exponent so that the machine integer
comparison instructions would also work for floating point.

I didn't study the IEEE specification closely enough to know if this
still works, especially with unnormalized numbers.

It is also true that using integer comparison instructions would not
trap "illegal" values (e.g. -0).


-- 
	Tom Teixeira,  Massachusetts Computer Corporation.  Westford MA
	...!{ihnp4,harpo,decvax}!masscomp!tjt   (617) 692-6200 x275