[CSC 435] MPI

[Andrew J. Pounds]

 From zeus and all of the systems in Lab 204 (except for those that are 
down) I have removed ALL versions of the MPI runtime environment and any 
development files.   Due to system updates there were multiple versions 
of MPI out there in addition to the "official release" from the OpenMPI 
team.

I then reinstalled a version of MPI that is appropriate for the version 
of CentOS that is on Zeus and the systems in lab 204.   To avoid 
confusion, during the install process I made sure that there were no 
references to the CUDA libraries (so you should no longer see those 
warnings).  It took about four and a half hours, but we should have a 
working system.

Here is what I need you to try immediately:

1. In you .bash_profile make sure that /usr/lib64/openmpi/bin is in your 
executable PATH

2. Also in your .bash_profile make sure that your LD_LIBRARY_PATH 
include /usr/lib64/openmpi/lib

3. log out of zeus and log back in to make sure that all your paths are 
set correctly

4. on zeus REMAKE your MPI files so that the libraries are all synced 
properly with the execution environment

5. on zeus try to run a single node mpi  executable

mpirun -n 2 -host zeus mpimmm

6. Try to run the PBS/TORQUE script in the job manager that was failing 
the other day and see if it now works


Please keep me posted and let me know if it is or is not working.

Systems 4, 8, and 21 are down.  System 4 is hopeless, I will try to fix 
8 and 21 later.   It would be great if you could knock out a bunch of 
your timings today for the C version of MPIMMM.

-- 
Andrew J. Pounds, Ph.D.  (pounds_aj at mercer.edu)
Professor of Chemistry and Computer Science
Mercer University,  Macon, GA 31207   (478) 301-5627
http://faculty.mercer.edu/pounds_aj

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://theochem.mercer.edu/pipermail/csc435/attachments/20160424/91fbf3be/attachment.html>