Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Path: utzoo!watmath!clyde!caip!think!nike!ucbcad!ucbvax!NOSC.ARPA!broman%bugs
From: broman%bugs@NOSC.ARPA (Vincent Broman)
Newsgroups: net.lang.mod2
Subject: Re:  Open Array Parameters and matrices
Message-ID: <8610141544.AA01110@bugs.ARPA>
Date: Tue, 14-Oct-86 11:44:52 EDT
Article-I.D.: bugs.8610141544.AA01110
Posted: Tue Oct 14 11:44:52 1986
Date-Received: Fri, 17-Oct-86 20:12:02 EDT
Sender: daemon@ucbvax.BERKELEY.EDU
Organization: The ARPA Internet
Lines: 16

I don't see what is so abominably inefficient about a nice clean
indexing function on an array descriptor.  That would be the Natural
Solution to this problem as well as more general ones, such as how to
store symmetric arrays in half the space, how to switch from dense
to sparse storage of a matrix, etc, etc.

The function call overhead is all the additional burden required,
since the arithmetic of index computation must be done in any case.
The only other inefficiency might be when a smart optimizing compiler
can simplify some sequential accesses to the array, but isn't smart enough
(or given opportunity) to expand an indexing function call inline
in order to optimize it.

Vincent Broman, code 632,  Naval Ocean Systems Center, San Diego, CA 92152, USA
Phone: +1 619 225 2365        {seismo,caip,ihnp4,ucbvax}!\\\\\\\
Arpa: broman@bugs.nosc.mil Uucp: {floyd,moss,bang,hp-sdd,sdcsvax}!noscvax!broman