Ok.
Many thanks.
Saverio
________________________________
Da: Adrian Roitberg via AMBER <amber.ambermd.org>
Inviato: giovedì 11 aprile 2024 16:22
A: saverio lemme via AMBER <amber.ambermd.org>
Oggetto: Re: [AMBER] Help for the value of #SBATCH --cpus-per-task on a cluster when using only one GPU for pmemd.cuda_SPFP.MPI
If you are going to use only one gpu, you should not use
pmemd.cuda_SPFP.MPI but rather use pmemd.cuda_SPFP
Also, set
#SBATCH --cpus-per-task to 1, delete the export OMP lines ans do not use mpirun
your script shopuld end up looking like this:
#!/bin/bash
#SBATCH -N1 --ntasks-per-node=1
#SBATCH --cpus-per-task=1
#SBATCH --gres=gpu:1
module load profile/lifesc
module load amber/2022
. $AMBERHOME/amber.sh
$AMBERHOME/bin/pmemd.cuda_SPFP -O -i 10_prod_md.in -o 10_prod_md_step_000_050_ns.out -p target.prmtop -c 9_relaxing.rst7
-r 10_prod_md_step_000_050_ns.rst7 -inf 10_prod_md_step_000_050_ns.info -ref 9_relaxing.rst7 -x 10_prod_md_step_000_050_ns.nc
On 4/11/24 7:02 AM, saverio lemme via AMBER wrote:
> [External Email]
>
> Hi all,
> I should like to execute an MD simulation on the Leonardo
> cluster using only one GPU.
> Each node has 1 x CPU Intel Xeon 8358 32 core, 2,6 GHz and
> 4x Nvidia custom Ampere GPU 64GB HBM2.
> I have modified a working script ,prepared for using all the four GPUs ,for using only one GPU as follows:
>
> #!/bin/bash
>
> #SBATCH -N1 --ntasks-per-node=1
> #SBATCH --cpus-per-task=8
> #SBATCH --gres=gpu:1
> module load profile/lifesc
> module load amber/2022
> . $AMBERHOME/amber.sh
> export OMP_NUM_THREADS=8
> export OMP_PROC_BIND=true
>
> mpirun -np 1 $AMBERHOME/bin/pmemd.cuda_SPFP.MPI -O -i 10_prod_md.in -o 10_prod_md_step_000_050_ns.out -p target.prmtop -c 9_relaxing.rst7
> -r 10_prod_md_step_000_050_ns.rst7 -inf 10_prod_md_step_000_050_ns.info -ref 9_relaxing.rst7 -x 10_prod_md_step_000_050_ns.nc
>
> Which is the right value for
> #SBATCH --cpus-per-task
> and
> export OMP_NUM_THREADS,
> in order to avoid to waste my computing resources quota, when using only one GPU?
> Any suggestion will be appreciated.
>
> Thanks.
>
> Saverio
> _______________________________________________
> AMBER mailing list
> AMBER.ambermd.org
> https://eur03.safelinks.protection.outlook.com/?url=http%3A%2F%2Flists.ambermd.org%2Fmailman%2Flistinfo%2Famber&data=05%7C02%7Csaverio.lemme%40alumni.uniroma2.eu%7C9169aa375ebd440eb81608dc5a32ed05%7C24c5be2ad76440c5997582d08ae47d0e%7C0%7C0%7C638484421989336537%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C0%7C%7C%7C&sdata=8EXliqDjIauxPHO0N2WOA5aAcAXnSgJf6dPLcPg55Ec%3D&reserved=0<http://lists.ambermd.org/mailman/listinfo/amber>
--
Dr. Adrian E. Roitberg
V.T. and Louise Jackson Professor in Chemistry
Department of Chemistry
University of Florida
roitberg.ufl.edu
352-392-6972
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
https://eur03.safelinks.protection.outlook.com/?url=http%3A%2F%2Flists.ambermd.org%2Fmailman%2Flistinfo%2Famber&data=05%7C02%7Csaverio.lemme%40alumni.uniroma2.eu%7C9169aa375ebd440eb81608dc5a32ed05%7C24c5be2ad76440c5997582d08ae47d0e%7C0%7C0%7C638484421989344212%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C0%7C%7C%7C&sdata=aThZO33k5PyJoxQO6yjjWSAzmZQlf8QdMIE4%2B68lAHM%3D&reserved=0<http://lists.ambermd.org/mailman/listinfo/amber>
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Thu Apr 11 2024 - 08:00:03 PDT
