Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Posting-Version: version B 2.10.2 9/18/84; site harvard.ARPA
Path: utzoo!watmath!clyde!burl!ulysses!allegra!mit-eddie!think!harvard!stew
From: stew@harvard.ARPA (Stew Rubenstein)
Newsgroups: net.lang.c
Subject: Re: pointer alignment on a Vax 780
Message-ID: <214@harvard.ARPA>
Date: Wed, 26-Jun-85 02:04:42 EDT
Article-I.D.: harvard.214
Posted: Wed Jun 26 02:04:42 1985
Date-Received: Fri, 28-Jun-85 00:17:12 EDT
References: <478@crystal.UUCP>
Distribution: net
Organization: Aiken Computation Laboratory, Harvard
Lines: 12

> Are there any alignment constraints on the Vax 780?
> Could one, for example, set an integer pointer to
> the middle of a machine word? 
> 
> If there aren't alignment constraints, is it faster
> to compare two word-aligned integers?

There are no constraints, but there is definitely a speed penalty for
fetching non-aligned data.  longwords should be aligned on longword
boundaries, quadwords on quadwords, etc., for best performance.  I have
even heard of differences in benchmark timing based on the alignment
of the instruction stream, though this is harder to predict and control.