Hi,
Cpptraj.MPI will make use of parallelism when processing single
trajectories for almost all actions - the input trajectory frames will be
divided among all MPI processes. So the RDF calculation can benefit. If you
have a chance to compile the OpenMP+MPI code however I really recommend
doing that; then RDF will benefit from both levels of parallelization.
Typically I use 1 MPI thread per socket and leave the rest for OpenMP (so
e.g. given two nodes each with 2 six core processors I’d use 4 MPI threads
and OMP_NUM_THREADS set to 6). See the following for more info on
parallelism in cpptraj:
https://onlinelibrary.wiley.com/doi/full/10.1002/jcc.25382
Hope this helps,
-Dan
On Wed, May 8, 2019 at 5:51 AM Lorenzo Gontrani <lorenzo.gontrani.gmail.com>
wrote:
> Dear Amber developers,
>
> I am facing the problem of rdf calculations on HUGE trajectories (500
> frames with about 3,5 million atoms). On the linux machine I have access
> to, the cpptraj version was compiled only with MPI and not OpenMP
> support. Is MPI parallelization restricted to data analysis on multiple
> trajectories, or does it work for rdf too?
>
> Thanks a lot for any help
>
> Lorenzo G.
>
> --
> ==========================================
> Lorenzo Gontrani
> Research associate - EDXD group
> University of Rome "La Sapienza"
> ResearcherID: L-6061-2014
> Scopus Author ID: 6506613970
> OrcID:0000-0001-8212-7029
>
> GSM +39 338 7615798
> Email lorenzo DOT gontrani AT uniroma1 DOT com
> Webpage: http://webcaminiti/gontrani.html
> =========================================
> Rispetta l'ambiente: se non è necessario, non stampare
> questa e-mail
> Protect the environment: do not print this e-mail, unless
> necessary
>
>
> _______________________________________________
> AMBER mailing list
> AMBER.ambermd.org
> http://lists.ambermd.org/mailman/listinfo/amber
>
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Wed May 08 2019 - 04:30:03 PDT
