From hillson@ait.nrl.navy.mil Tue Dec 28 12:18:09 1999
Date: Tue, 15 Jun 1999 13:56:08 -0500
From: Roger Hillson <hillson@ait.nrl.navy.mil>
To: Tom Pulikal <tapulika@npac.syr.edu>
Cc: tran@metsci.com, smith@ait.nrl.navy.mil,
     Jeff Steinman <steinman@ca.metsci.com>
Subject: Re: SPEEDES/MPI on Janeway

>Hi
>
>  I'm working with Dr.Furmanski on Speedes testing etc. here at NPAC. I
>was trying to compile the MPI version of Speedes on Janeway yesterday and
>got some errors.
Dear Tom,

What errors?  Did you get warnings, FATAL compilation errors, or runtime
errors?  Did all the executables link?

Also, per Tuan's instructions, be sure and use the correct version of
gnumake i.e.

Set your path to contains:
	~bjacob/bin/IRIX64
and make sure that when you execute :
	>which make
This version of make
	/afs/cmf/users/bjacob/bin/IRIX64/make
is found.

>I'm attaching the README file that was there with the
>distribution under src/comm/MPI/ along with this mail.

This is an obsolete version of the README file.  The source code is
current.  I will send you a copy of the current README.NEW file in a
separate message.  The compilation instructions are not correct (i.e. the
compilation instructins are for the SPEEDES 0.6 version).

>The distribution
>can be found at /scratch/tapulika/speedes061199mpi/ on Janeway. Also, I
>was trying to compile it using
>
>./Admin/spmake -j 16 COMM=MPI LIBEXT=a
>
>from the top level speedes directory. So what all environment variables
>should I set on Janeway inorder to get the MPI version of Speedes compiled
>on Janeway ?

(1)  You don't need to set any environmental variables to compile, assuming
that your paths are set correctly.  You do need to set some environmental
(usually) in order to run.

I will also send you two files.  One sets the paths in your environments
needed to link with mpi.  The other sets some Janeway environmental
variables, although at values which use up a lot of memory.  Tuan says he
will send you a modified setup file which sets the environmental variables
at lower values, but which are sufficient for running the ParNSS demo.

An important point:  you don't have to preset any environmental variables.
IRIX will issue a run-time warning if any MPI environmental variable needs
to be reset.  It is best to use the minimal values necessary in order to
convserve memory, etc.

BTW, you might read the paper I gave at the HPCMP conference last week on
the MPI interfaces.  I gave Dr. Furmanski a copy, but I could also send it
as an MS Word attachment.

Good luck, and call or send email if you need additional help.

Roger Hillson

>
>thanks,
>--Tom

========================================================================

Roger Hillson, PhD		Code 5583
Telephone: 202-404-7332	  	(Advanced Information Technology Branch)
Fax:	   202-767-1122		Naval Research Laboratory
Email:	   			4555 Overlook Ave, SW
hillson@ait.nrl.navy.mil	Washington, DC 20375-5337