From: corbett@lupa.eng.sun.com (Robert Corbett)
Subject: Re: % on negative numbers
Date: 04 Dec 1998
Newsgroups: comp.std.c

In article <3669579e.3725172@news.dial.pipex.com>,
Mathew Hendry <scampi@dial.pipex.com> wrote:
>
>I'm curious. Are most machines consistent in their implementation of
>division/modulo operations involving negative numbers?

Except in cases involving overflow, which presumably are not covered
by the C standard, every architecture I have seen truncates toward
zero when the mathematical result of an integer division is not an
integer.  The reason is simple.  Unlike multiplication, there is no
reasonable way to do signed division directly in hardware.  (BTW,
the direct hardware implementations of signed multiplication are
very clever; they are well worth studying).  The usual implementation
of signed division does an unsigned division and then fixes up the
result, negating if necessary.  Since the result will already have
been chopped, getting any result other than truncate toward zero
would take extra time and circuitry.  Hardware designers tend not to
like to use either if they can avoid it.

Most machines implement modulo using software, so there is greater
variation.

					Sincerely,
					Bob Corbett