I am trying to run for the first time the Gromacs 5.1.4 on cluster supported by GPU. However, I am not able to figure out how to proceed such as which commands to use first and so on. Is there anybody having experience in working with Gromacs 5.1.4 with GPU supported cluster or with any other versions.
Thanks a lot for your help in advance!
Best regards
Pankaj
You will have to approach your cluster admin. There will be some some script which will specify the walltime, queue name, and other details including mpirun calling protocol. Contact your admin
Hi Pankaj, Sarath is right. You will have to consult the IT guy , who is supposed to be admin of the cluster. He will instruct you. You may need short script to fire a job on the cluster. See the link below.
http://supercomputing.iitd.ac.in/?soft
Hello Pankaj,
For the processing of simulation system there is nothing different than other versions of Gromacs. but for firing mdrun on GPU cluster , you must check the que types for your cluster and install the G_5.x.x with -DGMX_GPU=ON flag.
There are no differences while calling mdrun on GPU or CPU clusters it is just the matter of how the batch submission system handles the assignment of GPUs for a GPU compatible code.
for example
#!/bin/bash
#SBATCH -p gpu
#SBATCH --qos=gpu
#SBATCH -U usergroup
#SBATCH --nodes=1
#SBATCH --ntasks=2
#SBATCH --gres=gpu:2
#SBATCH --cpus-per-task=16
#SBATCH --mem=4096
#SBATCH --job-name="simulation"
#SBATCH --time=00:00:00
#SBATCH --mail-type=end
#SBATCH [email protected]
#======START===============================
echo "On which nodes it executes :"
echo $SLURM_JOB_NODELIST
echo "Which MPI Implementation is used :"
echo "Print current shell limits :"
#ulimit -a
echo "Now run the MPI tasks..."
EXE=/home/softwares/GMX5-GPU/bin
mpirun -np 2 $EXE/mdrun_mpi -s name.tpr
I hope that helps.
Well, there is a difference between calculation on CPU's and GPU's. I was able to figure out how to run the calculations efficiently only after the help of our cluster admin. But of course it always depends on your cluster architecture.
I have attached the PBS script which I use.
- Badges
- Science topic