Newsgroups: comp.parallel.mpi
From: Nick Nevin <nevin@alex.osc.edu>
Subject: Re: Problem with LAM6.0
Organization: Ohio Supercomputer Center
Date: 08 Jun 1996 10:42:54 -0400
Message-ID: <vn2g286tl01.fsf@alex.osc.edu>


>>>>> " " == Franz-Josef Markus <markus@sylt.iti.mu-luebeck.de> writes:

     > in the meanwhile I found out that all works well if the spawn is
     > called by a process on node n0 (HOST) but not if it is called on
     > any other node.

     > This holds true for my simple test program and for the mentioned
     > scheduling software.

     > May this be a bug in LAM6.0???

Are you using the -s option to mpirun or in the spawn in order to ship
the binary from some node to the executing node? There is indeed a bug
with spawning in this case.  If you do not use -s, which is only really
necessary if you cannot access the binary you want to run on remote
nodes via some common network file system, then spawn works correctly
for all my tests. These tests include spawning from nodes other than n0.

I will look into the -s problem.

Regards, Nick.

-=-
Nick Nevin				nevin@osc.edu
Ohio Supercomputer Center		http://www.osc.edu/lam.html

