Re: [AMBER] cpptraj question

From: Jason Swails <>
Date: Sat, 16 Jan 2016 11:55:00 -0600

On Sat, Jan 16, 2016 at 10:38 AM, Robert Molt <> wrote:

> I am encountering an oddity in attempting to do multiple serial cpptraj
> jobs on a given node. The problem is either inherent to my script
> choices or inherent to cppraj. I feel relatively confident the script is
> right (he said before falling)...such that I am wondering if there is an
> element to summoning cpptraj that I do not understand. I know some
> softwares have locking mechanisms to prevent multiple serial spawns.

​Those mechanisms are implemented on the machine/operating system level,
not the software level in this case.

My script contents are as follows:
> cd Run_1
> cp ../analysis.ptraj ./
> aprun -n 1 cpptraj -i analysis.ptraj &

​It's unnecessary to copy analysis.ptraj here. Just do

cd Run_1
aprun -n 1 cpptraj -i ../analysis.ptraj

> cd ../
> cd Run_2
> cp ../analysis.ptraj ./
> aprun -n 1 cpptraj -i analysis.ptraj &
> Wait
> I have modified this slightly; it actually goes over 16 jobs to use a
> full node, but you get the picture. When I examine my files associated
> with the jobs, /nominally/ 16 jobs have started in that the job record
> indicates

​The fact that you're using aprun indicates you're using a Cray machine, I
think. In that case, there are ways to run MPMD (Multiple Program Multiple
Data parallel computing model) with aprun "the correct way" (rather than
sending stuff to the background and using a "wait" at the end.

You'll have to look at the aprun manpage, but I think it's something like
having a single aprun statement and separating distinct jobs with colons.
So something like:

aprun -n 1 cpptraj -i analysis.ptraj : -n 1 cpptraj -i analysis2.ptraj : ...

But it's been a long time since I've done this, so I'm not positive on all
the details.


Jason M. Swails
Rutgers University
Postdoctoral Researcher
AMBER mailing list
Received on Sat Jan 16 2016 - 10:00:03 PST
Custom Search