[Pw_forum] Trying to find reasonable parallelization parameters

[Axel Kohlmeyer]

On Thu, 22 Jan 2009, J. J. Ramsey wrote:

JJR> ----- Original Message ----
JJR> > From: Axel Kohlmeyer <akohlmey at cmm.chem.upenn.edu>
JJR> > To: J. J. Ramsey <jjr19 at uakron.edu>
JJR> > Cc: pw_forum at pwscf.org
JJR> > Sent: Thursday, January 22, 2009 5:49:36 PM
JJR> > Subject: Re: [Pw_forum] Trying to find reasonable parallelization parameters
JJR> > 
JJR> > oh... and one more thing.
JJR> > 
JJR> > _what_ MPI are you using?
JJR> 

JJR> OpenMPI 1.2.3. Unfortunately, I don't really have a whole lot of 
JJR> control over the MPI implementation. It's on a cluster where I have 

oh. you can compile your own. compiling MPI does not require
administrator privilege. i do it all the time when i am not
happy with how it is being set up (even in some of the 
supercomputing centers). in this case it would probably
be beneficial to upgrade to 1.2.8, but please try adding 
the following settings to your mpirun commandline as it is.

--mca btl_openib_use_srq 1

and

--mca mpi_leave_pinned 1

the former is the memory saver (and performance enhancer), the 
second may be needed if you get crashes (if not, don't use it).

...and as i said before, i get _much_ better performance with
-npernode 4 on nodes with 8 cores.

cheers,
   axel.

JJR> an account but not much if any say as to how it's administered. 
JJR> FWIW, Linux Networx is the provider of the cluster.


JJR> 
JJR> 
JJR> 
JJR>       
JJR> _______________________________________________
JJR> Pw_forum mailing list
JJR> Pw_forum at pwscf.org
JJR> http://www.democritos.it/mailman/listinfo/pw_forum
JJR> 

-- 
=======================================================================
Axel Kohlmeyer   akohlmey at cmm.chem.upenn.edu   http://www.cmm.upenn.edu
   Center for Molecular Modeling   --   University of Pennsylvania
Department of Chemistry, 231 S.34th Street, Philadelphia, PA 19104-6323
tel: 1-215-898-1582,  fax: 1-215-573-6233,  office-tel: 1-215-898-5425
=======================================================================
If you make something idiot-proof, the universe creates a better idiot.