Newsgroups: comp.parallel.pvm
From: jeff@adeno.wistar.upenn.edu (Jeffrey Taylor)
Subject: PVM_ARCH = LINUXMP
Organization: University of Pennsylvania
Date: 29 Jan 1997 19:54:35 GMT
Mime-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Message-ID: <5co9tr$dmf@netnews.upenn.edu>

I run PVM programs on my SMP Linux system everyday and I can't wait any longer
for a shared memory PVM.  If you have already done this port, please let me
know!  If you are willing to help me do it (beta-testing and debugging) let me
know!  If you have any ideas or suggestions on the subject please send the to
me!  

This is a copy of an E-mail I sent to the PVM developers today:

Dear PVM developers,
	I, and a growing number of others, am using PVM on a multi-processor 
shared memory system running Linux.  At this time the only way to do this is to
use PVM_ARCH=LINUX which uses sockets and the loopback device.  This works but 
is probably not very efficient.  I have begun porting the shared memory code 
that is used with the other multi processors in the PVM distribution to work on
Linux by making the new PVM_ARCH type "LINUXMP".
        With very little effort, I am at the point where the daemon and 
libraries compile ok but there are still some runtime errors.  I admit to being
a novice in shared memory programming but I believe that the existing code will
port easily and not require any substantial changes.  Before going any further,
I wanted to inquire as to the status of this project on your end.  
	The PVM project has shown great support for LINUX systems from the early
going and I therefore assume that there has been some thought on your end 
concerning LINUX SMP systems.  I don't want to duplicate a lot of work if 
someone (perhaps better qualified) is already working on a LINUX shared memory 
PVM.  Perhaps, if someone is working on such a project, I could be of service 
in testing the code and squashing bugs, please let me know!  In any event, 
unless I hear from you to the contrary, I will continue the work I have started
and when/if I get it working perhaps it can be included in the next PVM release.



Jeff Taylor

-- 
=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
Jeffrey S. Taylor			        Molecular Modeling	
The Wistar Institute and			Software development
The University of Pennsylvania			Molecular Biology
Philadelphia, PA 19104          E-Mail    Jeff@aries.wistar.upenn.edu	
(215) 898-2202			WWW     http://adeno.wistar.upenn.edu