GROMACS version: 2020.4
GROMACS modification: No
I am having 8 nodes with 40 cores each (without hyperthreading).
I am trying to run gmx_mpi on these 320 core (8x40) but it gives mpi or mpi thread error.
I tried many combinations but all failed.
combinatons tried
$mpiexec.hydra -f $PBS_NODEFILE -np 320 gmx_mpi mdrun -s md_0_1.tpr -deffnm md_0_10
$mpiexec.hydra -f $PBS_NODEFILE -np 320 gmx_mpi mdrun -ntomp 4 -s md_0_1.tpr -deffnm md_0_10
$mpiexec.hydra -f $PBS_NODEFILE -np 320 gmx_mpi mdrun -ntomp 6 -nt 6 -s md_0_1.tpr -deffnm md_0_10
$mpiexec.hydra -f $PBS_NODEFILE -np 320 gmx_mpi mdrun -ntomp 40 -s md_0_1.tpr -deffnm md_0_10
Can you suggest me, how will I use all the core of each node with maximized performance?
See the detailed configuration here: https://ccf.iiita.ac.in/hpc.html