Newsgroups: comp.parallel.mpi,comp.parallel.pvm
From: lindahl@cs.virginia.edu (Greg Lindahl)
Subject: Re: Why explicit message passing??
Organization: a guest of Shadow Island Games
Date: 2 Apr 1998 01:05:59 GMT
Message-ID: <6fuo9n$dm5@news1.newsguy.com>

Kamran Karimi <kamran@wallybox.cei.net> writes:

>  Thousands of people have access to networked workstations or PCs. Using
> their combined power is very attractive, as they are incredible sources of
> computing power. Now, should the programmers of such systems also use the
> same methods that the some big guys _have_ to use?
> 
> Unfortunately this is what is happening now.

The "big guys" have MPI, PVM, distributed shared memory, parallel
compilers, and various library systems that hide details. You can use
all of these on networked workstations or PCs.  Some of these work
well for some problems, and no one works for all. You can't cast
this as "shared memory or message passing, which is best?"

BTW, for the person who thought OO was going to resolve the
problem... which kind, OO shared memory or OO message passing? Most OO
systems hide the details, so you often don't have to care how it's
implemented, only how fast it goes on your problem.

-- g