Gromacs MD run doesn’t recognize GPU

Scientific support
1

hello,i am trying to run gromacs using the GPU partitation:

#!/bin/bash
# em.sh - Energy Minimization using GPU
#SBATCH --job-name=em_gpu
#SBATCH --output=em_gpu_%j.out
#SBATCH --error=em_gpu_%j.err
#SBATCH --ntasks=1
#SBATCH --cpus-per-task=16
#SBATCH --mem=64G
#SBATCH --time=2:00:00
#SBATCH --partition=gpu
#SBATCH --gres=gpu:l40s:1

module load cuda-toolkit/12.9.1
module load cudnn/8.8.0.121
source /shared/software/miniconda/bin/activate gromacs-2020.2

export OMP_NUM_THREADS=16
nvidia-smi

gmx grompp -f em.mdp -c solv_ions.gro -p topol.top -o em.tpr -maxwarn 2
gmx mdrun -v -deffnm em -ntomp 16 -nb gpu -pme gpu -pin on

however, i got the following error:

Function:    bool gmx::decideWhetherToUseGpusForPme(bool, gmx::TaskTarget, const std::vector<int>&, const gmx_hw_info_t&, const t_inputrec&, const gmx_mtop_t&, int, int, bool)
MPI rank:    0 (out of 5)

Feature not implemented:
Cannot compute PME interactions on a GPU, because PME GPU does not support not
a dynamical integrator.

how is it possiblemto solve it?
thank you for your help.

2

Hello,

I think it’s related to GROMACS usage (not a cluster/node/GPU issue).
But I have no clue. Maybe you could try posting on the GROMACS forum?
For example: GPU Inconsistency User Input/Feature Not Implemented - User discussions - GROMACS forums

Best regards

1 « J'aime »
3

thank you for your help.

4

i think the GROMACS environment or modules were not compiled with GPU support that is why it keep giving this error .