Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Posting-Version: version B 2.10.2 9/5/84; site ism780c.UUCP
Path: utzoo!watmath!clyde!cbosgd!ukma!psuvm.bitnet!psuvax1!burdvax!sdcrdcf!ism780c!tim
From: tim@ism780c.UUCP (Tim Smith)
Newsgroups: net.arch
Subject: Re: Integer division
Message-ID: <548@ism780c.UUCP>
Date: Mon, 10-Feb-86 15:57:07 EST
Article-I.D.: ism780c.548
Posted: Mon Feb 10 15:57:07 1986
Date-Received: Wed, 12-Feb-86 20:36:55 EST
References: <11610@ucbvax.BERKELEY.EDU> <5100003@ccvaxa>
Reply-To: tim@ism780c.UUCP (Tim Smith)
Organization: Interactive Systems Corp., Santa Monica, CA
Lines: 20

In article <5100003@ccvaxa> aglew@ccvaxa.UUCP writes:
>
>The only architectural question here is - what should the
>actual hardware support, if it has a MOD instruction?

The architectural issue is how should division round.  The behaviour
of the MOD function is just a side effect of how division rounds.
Thus, the architectural issue is which of the following is best:

        1.      round toward zero
        2.      round toward -infinity
        3.      round toward +infinity
        4.      round to closest integer

( have I left any out? ).  1 and 2 are the ones most people consider
reasonable.  With 2 we have arithmetic shift right and division by two
are the same ( on a "normal" machine ), and a - (a/b)*b is the
mod function.
--
Tim Smith       sdcrdcf!ism780c!tim || ima!ism780!tim || ihnp4!cithep!tim