Re: AMBER: Amber9's sander.MPI on x86_64

From: Thomas Steinbrecher <steinbrt.scripps.edu>
Date: Mon, 4 Jun 2007 10:26:55 -0700 (PDT)

Hi Karl,

the executable sander is for serial runs, if you start it by mpiexec, you
should only get 16 processes doing the same thing and overwriting their
output. All parallel runs have to use the sander.MPI executable. You can
also check your output file, it will contain information on how (and
if) the parallel computing is set up.

For the very bad speedup, have you tried if your network might be at fault
by running a 2 or 4 processor job on only one node? In my
(limited) experience it is not uncommon to see a drop in efficiency once
you run your job over more than one node on a big cluster.

Kind Regards,

Thomas

Dr. Thomas Steinbrecher
The Scripps Research Institute
10550 N. Torrey Pines Rd.
San Diego CA 92037, USA
-----------------------------------------------------------------------
The AMBER Mail Reflector
To post, send mail to amber.scripps.edu
To unsubscribe, send "unsubscribe amber" to majordomo.scripps.edu
Received on Wed Jun 06 2007 - 06:07:20 PDT
Custom Search