[gmx-users] Building a Larger Liquid Phase - Truncate 'Vacuum'

Justin Lemkul jalemkul at vt.edu
Thu Jul 30 18:34:38 CEST 2015

On 7/29/15 4:25 PM, Elton Carvalho wrote:
> On Thu, Jun 25, 2015 at 7:39 PM, Justin Lemkul <jalemkul at vt.edu> wrote:
>> On 6/25/15 2:46 PM, John Degenstein wrote:
>>> I suppose in retrospect that I could simply use a single processor which I
>>> think would reduce the number of DD cells and thus prevent this error.
>> Or just use OpenMP parallelization.  Any time you drastically change the
>> size of the system, this will be the case.
> Is DD treated differently if you run MPI or OpenMP parallelization? If
> so, how? I don't recall reading that in the docs and I always assumed
> that DD worked distributing one domain per thread (be it OpenMP or
> MPI).

There are as many DD cells as there are (thread-)MPI processes.  These can be 
further subdivided into multiple OpenMP threads.  What I was talking about above 
is the use of only OpenMP, e.g.

mdrun -ntmpi 1 -nt X

There is no use of DD in this case.



Justin A. Lemkul, Ph.D.
Ruth L. Kirschstein NRSA Postdoctoral Fellow

Department of Pharmaceutical Sciences
School of Pharmacy
Health Sciences Facility II, Room 629
University of Maryland, Baltimore
20 Penn St.
Baltimore, MD 21201

jalemkul at outerbanks.umaryland.edu | (410) 706-7441


More information about the gromacs.org_gmx-users mailing list