Hi everyone!
I have some problems with running gmx_mpi on more then one nodes. I have 4 nodes to run with 64 threads, but when i use slurm script with “mpirun gmx_mpi …”, gromacs starts the process on 4 nodes with… 32 cores in total, when 4 nodes will be with 32x4 cores and 64x4 threads.
Thanks everyone for replying!
OpenMPI version 4.0.2
In .log file i see this. Gromacs using 4 nodes, but only with 16 threads, whats looks weird… And using 64 mpi threads by default. Can i use all cores and threads of 4 or at least 2 nodes? It will be 128 threads.
Slurm script that i used to run job:
#!/bin/bash -l
#SBATCH --job-name=GROMACS
#SBATCH --time=2400:00:00
#SBATCH --partition=general
##SBATCH --partition=fatMemNode2Tb
##SBATCH --partition=gpunode
#SBATCH -e output_err_%j
#SBATCH -o output_%j
##SBATCH --ntasks=8
#SBATCH --nodes=4
#SBATCH -n 64
export PATH=/home/vddayneko/soft/gromacs2023.3_mpi/bin:$PATH
##export GMX_OPENMP_MAX_THREADS=256
##export OMP_NUM_THREADS=32
gmx_mpi grompp -f em.mdp -c complex_solv_ions.gro -p topol.top -o em.tpr
sleep 5
mpirun gmx_mpi mdrun -deffnm em -v
sleep 5