Path: utzoo!utgpu!news-server.csri.toronto.edu!cs.utexas.edu!samsung!uakari.primate.wisc.edu!sdd.hp.com!news.cs.indiana.edu!ariel.unm.edu!ghostwheel.unm.edu!john
From: john@ghostwheel.unm.edu (John Prentice)
Newsgroups: comp.lang.fortran
Subject: Re: Array Notation
Message-ID: <1991Jan4.194728.29941@ariel.unm.edu>
Date: 4 Jan 91 19:47:28 GMT
Sender: news@ariel.unm.edu (News supported software)
Organization: University of New Mexico, Albuquerque, NM
Lines: 25

Let me bring this discussion full circle now.  The original question I
posted (in the parallel newsgroup?) had to do with the fact that the CM-2
and a few other machines invoke parallelism by the use of array syntax.
My argument was that this was a bit unfortunate since array syntax is
not part of any existing standard and Fortran Extended is yet to even be
ratified, much less a compiler written for it.  The response was that many
more compilers than I realized already support array syntax and that it is
therefore not as unportable as I argued (this point is moot for us by the way,
we are contractually obligated to use ANSI standard Fortran, so extensions
are not helpful since we can only use them when we have no other choice,
such as on the CM).  However, now people are pointing out that array syntax
is not a replacement for do-loops and that it may produce slower code, even
on a vector or parallel computer.  This would seem to drive us back toward
my original point, array syntax for invoking parallelism is unfortunate
until such time as Fortran Extended becomes available (whenver that might
be).  Even then what people are saying is that by writing my code to
run on a CM, which REQUIRES array syntax, I am going to hobble it on
most any other computer where the array syntax may defeat or confuse the
optimizer.  Is this a fair statement and if so, anybody have any suggestions
for how to address this problem short of having two versions of the code
(one for the CM and one for everyone else) ?

John Prentice
Amparo Corporation
Albuqueruqe, NM