Hello everyone,
I have been trying to simulate a protein in water, but the following error appears:
Fatal error:
Too many LINCS warnings (1436)
If you know what you are doing you can adjust the warning threshold links in your MDP file or set the environment variable GMX_MAXCONSTRWARN to -1, but normally it is better to fix the problem
I already simulated the protein in vacuo and as a complex, the simulation went normally.
However, when I try to simulate only the protein in water the production doesn't start.
I am using an HPC to run the simulation. I use 2 GPUs and 8 Threads. I noticed that the problem doesn't appear when I don't use the GPU but the calculation takes a long time and becomes unfeasible.
I forward the .sh file
#!/bin/bash -v
#SBATCH --partition=GPUSP4
#SBATCH --ntasks=2 # number of tasks / mpi processes
#SBATCH --cpus-per-task=8 # Number OpenMP Threads per process
#SBATCH -J GMX_APOsv
#SBATCH --time=80:00:00
#SBATCH --gres=gpu:tesla:2
#OpenMP settings:
export OMP_NUM_THREADS=8
export MKL_NUM_THREADS=1
export OMP_PLACES=threads
export OMP_PROC_BIND=spread
echo $SLURM_JOB_ID #ID of job allocation
echo $SLURM_SUBMIT_DIR #Directory job where was submitted
echo $SLURM_JOB_NODELIST #File containing allocated hostnames
echo $SLURM_NTASKS #Total number of cores for job
module load Gromacs/2019.3-cuda
#run the application:
gmx mdrun -pin on -ntmpi 2 -ntomp 8 -deffnm md -v
Dear Leonardo Bruno Federico
Not sure but maybe you have to export the CUDA devices (=gpus) by adding something like
export CUDA_VISIBLE_DEVICES=0,1
and then specify the GPUs when running with something like
gmx mdrun -ntmpi 2 -ntomp 8 -pin on -gpu_id 01 -deffnm md -v
You say you have tried to run this job without GPUs and you don't get any error but it is very slow. Have you tried running it with only one GPU? And can you limit yourself to only one node? With something like
#SBATCH --nodes=1
Hello Nicola Piasentin
This I can run normally when I do not indicate the use of 2 GPUs but the calculation is very slow.
I use these settings for all the dynamics I have been running but for this study, I have not been successful.
I tried to make the changes you suggested, but the error came up again.
- Badges
- Science topic