Re: [ESPResSo-devel] harmonic potential

[Axel Arnold]

Am Dienstag 21 Oktober 2008 14:39 schrieben Sie:
> Hello,
>
> can't you introduce Verlet lists for bonded potentials
> ONCE AND FOR ALL, which are simply not updated?

Hmm, the Verlet lists are not the problem; bonded interactions are not in them 
at all.

The problem is with the cell lists. If we allow harmonic bonds from any 
particle to any particle, we need to give up the current scheme of neighbor 
communication only, because harmonically bonded particles can be potentially 
everywhere. That can be done by e.g. broadcasting the positions of all bonded 
particles, but will kill the parallel performance. Moreover, I doubt that 
under normal circumstances, one needs harmonic bonds that span half the 
simulation box or more.

Therefore I vote for introducing a sort of breaking distance; one can still 
choose it relatively large, if necessary (half the local box size of a 
processor).

Regards,
Axel

-- 
Dr. Axel Arnold 
Fraunhofer SCAI
Schloss Birlinghoven, 53754 Sankt Augustin, Germany
Tel: +49 2241 14 2575