[gmx-developers] mpi and thread command line options

[Roland Schulz]

On Tue, Jul 10, 2012 at 8:26 PM, Shirts, Michael (mrs5pt)
<mrs5pt at eservices.virginia.edu> wrote:
>> Additionally, do we want to keep "-nt" on the long run? To me it
>> sounds too general especially that the thread-MPI parallelization is
>> not even "proper" multithreading. So I'd rather make it more verbose
>> e.g. "-ntmpi" and have with the same functionality as the deprecated
>> -nt.
>
> The one thing about current -nt / thread-MPI is that it just works with
> pretty much linear scaling, no matter what the deal with the cluster is or
> the MPI installed. We end up running most of our batch calculations this
> way.  Having something provide full speed out of the box regardless of how
> good the cluster admin that is hard to break is very important to most
> users.  So I wouldn't hide it or obscure it.

Maybe we should preserve the meaning of "-nt" in that sense described
by Michael. Thus it would give the total number of cores for mdrun to
use and mdrun decides how many OpenMP threads / tMPI processes to use.
The simplest automatic would: usually use as many tMPI processes as
-nt (and thus not changing the behavior) but if a n GPUS are detected
then use n tMPI processes and nt/n OpenMPI threads.

Roland

>
> Best,
> ~~~~~~~~~~~~
> Michael Shirts
> Assistant Professor
> Department of Chemical Engineering
> University of Virginia
> michael.shirts at virginia.edu
> (434)-243-1821
>
>
>
> --
> gmx-developers mailing list
> gmx-developers at gromacs.org
> http://lists.gromacs.org/mailman/listinfo/gmx-developers
> Please don't post (un)subscribe requests to the list. Use the
> www interface or send it to gmx-developers-request at gromacs.org.
>
>
>
>



-- 
ORNL/UT Center for Molecular Biophysics cmb.ornl.gov
865-241-1537, ORNL PO BOX 2008 MS6309