From: Tony Kocurko Newsgroups: comp.parallel.pvm Subject: Re: Practical Applications Date: Tue, 06 Oct 1998 15:33:42 +0000 Organization: Memorial University of Newfoundland Message-Id: <361A3855.4DB6E67@waves.esd.mun.ca> References: <6r1vqb$639$1@nnrp1.dejanews.com> <6v9egq$stc$1@supernews.com> Mime-Version: 1.0 Content-Type: multipart/alternative; boundary="------------EAB7F5EFA7F1C12C5624D177" --------------EAB7F5EFA7F1C12C5624D177 Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit Chris Stoddard wrote: > So the questions at this point are: > Is anybody working on Practical Applications for PVM/MPI? > if so what are they working on? > If not, why? is it too impractical or impossible? Chris, Take a look at http://weland.esd.mun.ca, specifically the slide show that starts at the bottom of the page. It shows how I wrote a code that helps us look for oil. Is that practical enough an application for you? In a nutshell, the original, serial version of the program that I wrote ran in 5.25 days on a single 50 MHz SuperSPARC. The first PVM version ran in a little less than a day on a mixed bag of old, beater SPARCs and an i486. Based on that speed up, we were able to pry loose some moolah and built a modest farm of 8 PCs (16 cpus and 1 Gb of total RAM on a 100 Mb/s full-duplex switched LAN), which ran the code in about an hour and a half. All of the parallel code uses PVM. Tony Kocurko --------------------------------------------------------------------------- Seismological Systems Manager "Someone who had begun to read geometry Department of Earth Sciences with Euclid, when he had learned the Alexander Murray Building ER-4063 first proposition, asked Euclid, `But Memorial University of Newfoundland what shall I get by learning these St. John's, Newfoundland things?' whereupon Euclid called his Canada A1B 3X5 slave and said `Give him three-pence since he must make gain out of what he phone : 1-709-737-8898 learns.'" FAX : 1-709-737-2589 - Stobaeus e-mail: tony@waves.esd.mun.ca web : http://www.esd.mun.ca/~music/bio8.html --------------------------------------------------------------------------- --------------EAB7F5EFA7F1C12C5624D177 Content-Type: text/html; charset=us-ascii Content-Transfer-Encoding: 7bit Chris Stoddard wrote:

  So the questions at this point are:
     Is anybody working on Practical Applications for PVM/MPI?
     if so what are they working on?
     If not, why? is it too impractical or impossible?

Chris,
  Take a look at http://weland.esd.mun.ca, specifically the slide show
that starts at the bottom of the page. It shows how I wrote a code that
helps us look for oil. Is that practical enough an application for you?
In a nutshell, the original, serial version of the program that I wrote
ran in 5.25 days on a single 50 MHz SuperSPARC. The first PVM version
ran in a little less than a day on a mixed bag of old, beater SPARCs and
an i486. Based on that speed up, we were able to pry loose some moolah
and built a modest farm of 8 PCs (16 cpus and 1 Gb of total RAM on a
100 Mb/s full-duplex switched LAN), which ran the code in about an hour
and a half. All of the parallel code uses PVM.
Tony Kocurko
---------------------------------------------------------------------------
Seismological Systems Manager       "Someone who had begun to read geometry
Department of Earth Sciences         with Euclid,  when he had  learned the
Alexander Murray Building ER-4063    first proposition, asked Euclid,  `But
Memorial University of Newfoundland  what  shall I  get by  learning  these
St. John's, Newfoundland             things?'  whereupon Euclid  called his
Canada      A1B 3X5                  slave and said  `Give him  three-pence
                                     since he must make gain out of what he
phone : 1-709-737-8898               learns.'"
FAX   : 1-709-737-2589                                           - Stobaeus
e-mail: tony@waves.esd.mun.ca
web   : http://www.esd.mun.ca/~music/bio8.html
---------------------------------------------------------------------------
  --------------EAB7F5EFA7F1C12C5624D177--