On Tue, Nov 05, 2013, Soumendranath Bhakat wrote:
>
> We are running our MD jobs in Amber in GPU cluster. For submission of our
> jobs we are using ./sub_apv.sh & and vi sub_apv.sh is showing a script like
> this:
>
> #!/bin/sh
> source /etc/profile.d/modules.sh
> module add intel-XE/c8000 mvapich2/c8000 cuda/5.0 amber/k20
> exe=/export/home/embele/amber/amber12/bin/pmemd.cuda
>
> np=12
>
> time mpirun -np $np $exe -O -i Partial_Mini.in -o Partial_Mini.out -p
> com_solvated.top -c com_solvated.crd -ref com_solvated.crd -r
> Partial_Mini.rst
Several points here:
1. pmemd.cuda.MPI only runs on a small number of GPUs, maybe 2 or 4 at most.
If you have 12 GPUs, you best bet is to run 12 separate jobs.
2. The MPI version is pmemd.cuda.MPI. Your command line above is running 12
copies of the same single-GPU code (pmemd.cuda) all writing to the same files.
At the very best you will get extremely slow performance, and corrupt files
are certainly possible.
3. Because of the way email formats text, it is hard to tell where lines
begin and end. I don't immediately see any reason why the second job doesn't
run when the first is done.
...dac
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Tue Nov 05 2013 - 04:30:03 PST
