On Wed, Mar 08, 2023, Vilhelm Ekberg via AMBER wrote:
>
> I am interested in running QM/MM in parallel on a cluster with 64 available
> cores per node. I will be using multi-sander (Amber22) with 2 groups, Some
> benchmarks QM/MM runs on my system with the parameters:
>
>
> #SBATCH --nodes 1
> #SBATCH --cpus-per-task 1
> #SBATCH --ntasks 2
>
> took roughly 4.5 days to complete. I am wondering how I should set the
> SBATCH parameters in order to optimize the parallelization? I guess
> one relevant question also is how much the job can be sped up with
> parallelization?
You don't say what sort of QM you are doing, or by which program. That is
crucial for parallelization: the only speed-up will come from parallelizing
the QM part. Also, I'm not sure why you have 2 groups in sander. But I
think the key thing is to know more about how you are calling whatever QM
program is doing the electronic structure stuff.
....regards...dac
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Wed Mar 15 2023 - 07:30:04 PDT
