[gmx-developers] Re: NMR refinement

[Bogdan Costescu]

On Tue, Jan 15, 2013 at 3:01 PM, santhosh <mesanthu at gmail.com> wrote:
> @Bogdan :
>>>>>Is this really one vector for the whole molecular system or for a
> group ? This looks suspiciously like the pull code...
>
> Its really one vector(3NX1) for one protein system(N atoms).

Ah, we are using the term vector with different meanings: I use it to
represent a physical vector, a set of 3 real values giving the
magnitude of the 3 components in space defining f.e. a vector force;
you use the term as a storage object, a 1D array.

So you want to calculate an additional force for each atom, which
should be added to the force already calculated by GROMACS. I seem to
remember similar requests, so please look through the archives...

If the force for each atom can be calculated from its position,
possibly relative to other nearby atoms, and possibly including some
atom property (like charge), then you don't need to care how the
system is partitioned. However, if the force for an atom depends on
all atoms, then you need to gather the positions of atoms from all
nodes, calculate forces on one node and distribute them again to all
nodes; in this case, take a look at the code gathering atom
coordinates before writing to a trajectory.

Cheers,
Bogdan