[gmx-users] problem of running multiple cores (Tom)

Justin Lemkul jalemkul at vt.edu
Sun Mar 16 00:33:43 CET 2014

Please don't reply to the entire digest.

On 3/15/14, 7:27 PM, Tom wrote:
> Justin
> Thanks for your quick response!
> Yes, I used  "periodic_molecules = yes"
> It is correct to use this, right?

Per your description of bonds across PBC, yes.  The periodic_molecules keyword 
is intended for this exact situation (in fact, that's its only purpose).

> I just can not figuire out why single core can work but can not work
> muli-cores.

The DD algorithm can't resolve the bonded geometry, while in serial perhaps 
something is coincidentally working.

> Is it because i did not assign the charge groups on my surface atoms?

No.  If you didn't use periodic_molecules, that's why it's failing.  It's not 
entirely clear what you mean by "did not assign charge groups."  Does that mean 
all the atoms are in a single group, or each atom is in its own group?  If the 
former, you will have problems with neighbor searching unless you are using the 
Verlet cutoff scheme.  If the latter, there is no problem as long as you are 
using PME.



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

Department of Pharmaceutical Sciences
School of Pharmacy
Health Sciences Facility II, Room 601
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