Newsgroups: comp.parallel.pvm
From: wjiang@thud.cs.utk.edu (weicheng jiang   )
Subject: Re: Patch #8 for PVM version 3.3 is available on Netlib
Organization: Computer Science Dept, University of Tennessee, Knoxville
Date: 1 Aug 1995 13:42:17 GMT
Message-ID: <3vlavpINNq6g@CS.UTK.EDU>

The SP2 port in PVM3.3.8 is based on IBM's MPI-F, not MPICH. IBM's MPI-F offers
similar communication performance as MPL.

You must have IBM's MPI-F installed on your SP2 to take advantage of this 
special port.

-Weicheng Jiang
PVM Group

In article <3vkp20$raf@upsn2.u-psud.fr>, giga@cri.u-psud.fr (Gilles.GHATTAS) writes:
|> In article <DCLxLA.852@cunews.carleton.ca>, montoya@mrco.carleton.ca (Frederic Montoya) writes:
|> > Erik Demaine (edemaine@ug.cs.dal.ca) wrote:
|> > > Bob Manchek (manchek@rudolph.cs.utk.edu) wrote:
|> > > : + added port to IBM SP-2, using MPI.
|> > 
|> > > Is this implemented on MPI or MPL?  If the former, does that mean one must
|> > > first install e.g. MPICH?
|> 
|> 
|> I think PVM3.3.8 for SP/2 is implemented on MPI because I downloaded three month ago
|> a beta version of PVM3.3. This beta version needs 'mpicc' to compile. So you need to 
|> install MPICH before installing PVM3.3.8 (if you want to use the High Performance 
|> Switch of your IBM SP).
|>  
|> > > You may be interested to know (you might already) that IBM made their
|> > > own implementation of PVM (not sure what version) on the SP-2 native on
|> > > MPL.  If you're interested, you can contact me (our SP2 has PVMe).
|> 
|> There is only a problem with PVMe: PVMe is *not* up to date.
|> But we have obtained better performance with PVMe (*8-10) for bandwith and latency than 
|> PVM.
|> So we intensively use PVMe.
|> 
|> May be we 'll use PVM3.3.8 (rather than PVMe) in the future because the person who has
|> given me the beta version of PVM for SP said me: The performances are better with PVM for
|> SP than with PVMe. Is anyone has observed this?
|> 
|> At this time, we need to install and test it. But the problem is that I can't install 
|> MPICH because of a xlc bug. (I'm waiting for a PTF from IBM to fix the bug). 
|> 
|> SeeYA
|> 
|> Gilles (giga@cri.u-psud.fr).
|> 

