From: Juergen Hoffmann <juergen.hoffmann@physik.uni-karlsruhe.de>
Newsgroups: comp.parallel.pvm
Subject: Re: pvm vs. mpi
Date: Tue, 11 May 1999 16:01:46 +0200
Organization: Institut fuer Angewandte Physik
Message-Id: <3738384A.41C6@physik.uni-karlsruhe.de>
References: <3736D96A.41C6@physik.uni-karlsruhe.de>
    <373736AE.41C6@iwr.uni-heidelberg.de>
Mime-Version: 1.0
Content-Type: text/plain; charset=iso-8859-1
Content-Transfer-Encoding: 8bit
Xref: ukc comp.parallel.pvm:8369


Hello Hoai Tran Van,

>   I'm not an expert in MPI & PVM programming. But both are the same
> (message passing interface). Few days ago, I got a paper discussing
> about the differences between them. If you need it, i will try to take
> and send the internet address of its.

Yesterday night I found and read this paper:
<a href=http://www.epm.ornl.gov/pvm/PVMvsMPI.ps> 
PVM and MPI: A Comparison of Features</a>
Paper published in Calculateurs Paralleles Vol. 8 No. 2 (1996).
Describes differences between PVM and MPI and under what circumstances
one package is favored over the other.

>   Anyway, you can learn one of them and switch easily to another. Some
> guys said : MPI for parallel computer and PVM for networked
> workstations.

Yes, this seems to be one point. I will use a RS/6000 single processor,
a RS/6000 double processor both with AIX, two Intel 486 Linux "machines"
and two Intel 386 Linux "machines" all in the same network segment.
So with all the remaks made here and the information I got from 
the paper and the LAM and PVM Homepages I decided today to use PVM.

Thanks to everybody for their remarks about this!

Juergen

-- 
Dipl.-Phys. Jürgen Hoffmann
Institut für Angewandte Physik
Universität Karlsruhe
D-76128 Karlsruhe
Tel.: +49-721-608-7478   Fax: +49-721-607593 
email: juergen.hoffmann@physik.uni-karlsruhe.de