Path: utzoo!utgpu!news-server.csri.toronto.edu!cs.utexas.edu!sun-barr!olivea!genie!udel!haven.umd.edu!uvaarpa!murdoch!astsun.astro.Virginia.EDU!gl8f
From: gl8f@astsun.astro.Virginia.EDU (Greg Lindahl)
Newsgroups: comp.sys.super
Subject: Re: Massively Parallel LINPACK on the Intel Touchstone Delta machine
Message-ID: <1991Jun13.171812.15043@murdoch.acc.Virginia.EDU>
Date: 13 Jun 91 17:18:12 GMT
References: <1991Jun06.205144.22611@ariel.unm.edu> <1991Jun10.144354.695@chpc.utexas.edu> <11771@mirsa.inria.fr>
Sender: usenet@murdoch.acc.Virginia.EDU
Organization: Department of Astronomy, University of Virginia
Lines: 23

In article <11771@mirsa.inria.fr> furnish@zig.inria.fr (Geoffrey Furnish) writes:

>In a language written for a serial machine like a CRAY, etc, I'd have had
>to change every single set of double loops into tripple loops, and add a
>subscript to every single array reference in the program.

Er, there's nothing about FORTRAN 90 that makes it impossible to write
a Cray compiler for it. On the other hand, there are array expressions
in F90 which cannot be optimized as well as the equivalent serial
expression.  Ask on comp.lang.fortran for examples.

Array notation can be nice in some situations. It isn't for everyone.

For example, I have a program which has a 3d data array, and a
one-dimensional operator which must be applied in the x, y, and z
directions. This "operator" is about 3,000 lines of code. Tell me how
I can write all those expressions into 3 sets of 3d-array code without
writing my own preprocessor? Hm?

Nice, yes. Solves everything, no. BTW, my code is typical for
high-accuracy astrophysical fluid dynamics.

-- greg