[gmx-developers] MC integrator
=?ISO-8859-1?Q?Andr=E9_Assuncao_da_?=Silva Teixeira Ribeiro
aastr at iq.ufrj.br
Sun Aug 9 14:37:00 CEST 2009
Following Berk's suggestion I have checked the TPI code and used the same ideas to
optimize performance for the MC integrator. In the current form it will only work if the
MC trial move involved only one charge group.
I was also having a problem with calling do_force to evaluate energy without force
calculation. The array fr->f_novirsum is not assigned when we do not want forces,
however it was called inside ewald_LRcorrection when fr->bEwald is TRUE. A simple change
in the above routine fixes this.
Anyway, with the optimized ns and without calculation of forces the performance
increased by more than 100% when compared to my original code. However, due to some
changes pd decomposition is not working right now.
I will be now trying to work with flexible molecules.
The code can be found at
More information about the gromacs.org_gmx-developers