[gmx-users] RE: RE: RE:RE: One question about Domain decomposition

xuji xuji at home.ipe.ac.cn
Tue Feb 17 14:02:32 CET 2009

Hi, Berk

Thank you for your explain!
I have a little parallel program about simulating long chain molecules. 
Because of the difficulties in parallel, the atoms of the chain molecules
are stored in the space grid. That is to say first I use domain decomposition
to split the whole simulation space into several small domains, 
and then in each domain I use the grids which are used in neighbor 
searching to store the atoms. It's convenient to do neighbor searching this way, 
but when I come to the bonds, angles and dihedrals forces it is hard to determine 
every atom's connected bonds, angles and dihedrals atoms. Especially it's hard to 
locate the memory position where the atom is stored according to its index.
So I come to Gromacs-4.0 for the high efficiency of it. So actually the element
of my program is atom but not charge group. So can you tell me a little more 
detail about the way you store the atoms of the chain molecule in Gromacs-4.0 
in domain decomposition? And how the bonds, angles and dihedrals are reconstructed
in each local domain?



More information about the gromacs.org_gmx-users mailing list