Path: utzoo!mnetor!uunet!lll-winken!lll-lcc!ames!eos!aurora!labrea!decwrl!16bits.dec.com!kruger
From: kruger@16bits.dec.com (Bear with me)
Newsgroups: comp.arch
Subject: overflow in optimized multiplication
Message-ID: <8803032326.AA28876@decwrl.dec.com>
Date: 4 Mar 88 02:21:00 GMT
Organization: Digital Equipment Corporation
Lines: 16

Lawrence Crowl writes that there is no easy way to tell when you've really
created an overflow condition in a partial computation. While this may be true
in current architectures, it does not have to be. There are two cases:

x*5 == x << 2 + x

in which you want to OR the overflow condition, and

x*7 == x << 3 - x

in which you want to XOR the overflows at each step.

Even with multiple levels of doing this, you can keep this methodology and
effectively keep track of overflow -- given appropriate (and minimal) hardware
support.

dov