[Swift-devel] [Fwd: Re: MPI via globus (fwd)]

[Mike Wilde]

Not sure if this is relevant to us?

- Mike

-------- Original Message --------
Subject: Re: MPI via globus (fwd)
Date: Tue, 27 Feb 2007 15:01:24 -0600
From: Alain Roy <roy at cs.wisc.edu>
To: Stuart Martin <smartin at mcs.anl.gov>, Steve Gallo <smgallo at CCR.BUFFALO.EDU>
CC: OSG ITB <OSG-INT at OPENSCIENCEGRID.ORG>, Abhishek Singh Rana 
<rana at fnal.gov>,   Steve Gallo <smgallo at CCR.BUFFALO.EDU>, JP Navarro 
<navarro at mcs.anl.gov>
References: <Pine.SOL.4.64L.0702261219380.9975 at fsui02.fnal.gov> 
<B1A2BF3C-1C42-4F82-A0BD-0045C2BD3880 at mcs.anl.gov>

>There is a patch that TG gave to Alain that was added in VDT 1.6.1
>(correct me if I am wrong Alain).  It creates a new flag at the top
>of the pbs.pm script to turn off this default behavior for using the
>mpi paths found at post-install time.

If I understand it correctly, it's in VDT 1.6.1. I see a flag at the
top of pbs.pm that is:

     $mpisoftenv = 0; # 0=false, 1=true

Is that it?

-alain



-- 
Mike Wilde
Computation Institute, University of Chicago
Math & Computer Science Division
Argonne National Laboratory
Argonne, IL   60439    USA
tel 630-252-7497 fax 630-252-1997