Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Posting-Version: version B 2.10.1 6/24/83 SMI; site sun.uucp
Path: utzoo!linus!decvax!decwrl!sun!shannon
From: shannon@sun.uucp (Bill Shannon)
Newsgroups: net.micro.68k,net.arch
Subject: Re: bit addressing and floating point formats
Message-ID: <1661@sun.uucp>
Date: Mon, 3-Sep-84 17:59:36 EDT
Article-I.D.: sun.1661
Posted: Mon Sep  3 17:59:36 1984
Date-Received: Thu, 13-Sep-84 05:16:43 EDT
References: loral.444 <491@turtlevax.UUCP> <1185@rti-sel.UUCP>, <1136@hcrvax.UUCP> <3049@watcgl.UUCP>
Organization: Sun Microsystems, Inc.
Lines: 19

dmmartindale@watcgl.UUCP says:
	Thus you can use the "feature" of passing the address of a double where
	the address of its float portion is needed without having to add an offset.

It should be made clear that this is EXTREMELY non-portable.  It depends
on something which is not true on e.g. the Sun using IEEE floating point
format:

	The bit representation of a float is the same as the bit
	representation of a double except that it has fewer bits
	in the mantissa, and that the address of a double will
	point to a float with the "same" value.

On many machines, such as the Sun, this is NOT TRUE.  We have been
bitten quite a few times by code that passes the address of a float
to a routine expecting a pointer to a double.

					Bill Shannon
					Sun Microsystems, Inc.