Re: [AMBER] Job submission script for SGE

From: Jason Swails <>
Date: Fri, 14 Aug 2015 07:46:27 -0400

On Thu, Aug 13, 2015 at 7:50 PM, Himansu <> wrote:

> Dear Amber users,
> Recently we installed AMBER14 and AMBER TOOLS 15 in our SGE HPC. We are
> using Tutorials to verify the installation.

​Don't use the tutorials simply to verify the installation -- that's what
the test suite is for. The manual has instructions on running the test

Currently we are using
> Tutorial: A room-temperature ionic liquid. The serial one is working but
> not the parallel one. I am attaching the job submission script (
> that I used for the parallel run for other software such as TURBOMOLE.
> Seeking your suggestions in this regard.

​Your submission script is *way* too complicated. What's more is that you
are running all steps -- including the setup steps -- in the compute node.
The "typical" workflow is to do the first steps by hand on your workstation
(or interactively on the login node) -- preparing the PDB file, running
leap to get the prmtop and inpcrd file -- and then use a submission script
to run either sander or pmemd (perhaps several sander or pmemd jobs

​Also, clusters like the one you are trying to run on are all different. I
suggest consulting your cluster documentation for how to run
MPI-parallelized programs (in general, you will need some kind of "mpirun"
or "mpiexec" command to run sander.MPI or pmemd.MPI with more than one
processor). Also try simplifying your submission script significantly so
that it only runs a single sander.MPI simulation and see if you can get it
to work. That will be easier both for you to debug and for us to help
with​ if you still have problems.

Good luck,

Jason M. Swails
Rutgers University
Postdoctoral Researcher
AMBER mailing list
Received on Fri Aug 14 2015 - 05:00:04 PDT
Custom Search