[gmx-users] MPI simulation with CHARMM27 force field and CMAP dihedrals problem

[Koivuniemi, Artturi]

Hi,

I have tried to simulate a protein in water with the CHARMM27 force field and the GROMACS simulation package. Without the CMAP correction the simulation runs just fine, but when adding the CMAP correction to the dihedrals, the protein quickly starts to unfold and the simulation stops (with no error message in log or supercomputer's output file). I'm running my simulations on a supercomputer with 64 processors and I'm using mdrun_mpi. When running on my desktop with four cores I have no problems. Could it be that the current mpi version of mdrun in the supercomputer is the reason for this, or is the CMAP correction unavailable for parallel simulations using the mpi as in the case of GPUs when using an implicit solvent model?

Best regards,
Artturi Koivuniemi