Newsgroups: comp.parallel.pvm,comp.parallel.mpi,comp.parallel
From: gale@wind.hpc.pko.dec.com (Israel Gale)
Subject: HPF for Cray and IBM
Organization: Digital Equipment Corporation
Date: 30 Apr 1996 14:33:49 -0400
Message-ID: <4m7k8h$kbb@usenet.srv.cis.pitt.edu>

[Sorry, I lost the beginning of this thread]

> Does HPF run on a Cray T-3D ? Or an IBM cluster of RS-6000 stations?
> The reason I ask is that I spend a lot of time right now maintaining or
> developing parallel code on these two machines with PVM etc. My applications
> mainly involve domain-decomposition of data over the number of parallel
> processors. I would like to cut my development time, as we are constantly
> adding features to our (serial) code, which are then put into the parallel
> version (by hand by me).
> Suggestions?

Although I would have to see your code to be sure, domain
decomposition is the sort of thing that works well in HPF.

IBM just released their "XL HPF" HPF compiler this month.  It is a
subset-HPF compiler that should give you good performance.  There is
every reason to believe that this (already solid) compiler should
improve with each release.  (http://204.146.47.71:80/ad/fortran/xlhpf/)

As for Cray, the PGI (Portland Group) HPF compiler now works with
Cray's f90 compiler.  As far as I know, PGI's is the best third-party
(i.e. not written by a hardware vendor) HPF compiler.
(http://www.pgroup.com/) 

Obviously, I don't have any inside knowledge about IBM, Cray, or PGI,
so I would recommend contacting them directly for more information.

-Israel Gale                **********************************************
                            * For information on Digital Fortran/HPF,    *
                            * write to me (gale@hpc.pko.dec.com), or see *
                            * http://www.digital.com/info/hpc/fortran/   *
                            **********************************************

