I am trying to using quantum-espresso(QE), but I can not configure it right.
A template for QE looks like the following
The one I modified looks like the following, but it gives errors
Anyone can help?
Thank you
I am trying to using quantum-espresso(QE), but I can not configure it right.
A template for QE looks like the following
Anyone can help?
Thank you
It’s running now.
Thanks
How about:
#!/bin/bash
#SBATCH --job-name=qe
#SBATCH --ntasks=12
#SBATCH --cpus-per-task=1
#SBATCH --time=24:00:00
#SBATCH --mem-per-cpu=3GB
#SBATCH --mail-user=username@usc.edu
#SBATCH --mail-type=all
#SBATCH --nodes=1
#SBATCH --account=<account_id>
#SBATCH --partition=main
module purge
module load gcc/9.2.0
module load mpich/3.3.2
module load quantum-espresso/6.5-openblas-hdf5
module load fftw/3.3.8-mpich
module load openblas/0.3.7
module load netlib-scalapack/2.1.0-openblas-mpich
module load hdf5/1.10.6-mpich
cd ${SLURM_SUBMIT_DIR}
mpirun --bind-to core -np 12 pw.x -i x.md.in > x.md.out
Thank you, Molguin. Your slurm file works faster than the one above.
Depending on your system size, increasing the number of processors to run in parallel will speed up the time to solution.
The SBATCH slurm directives used to increase the number of parallel workers are:
#SBATCH --ntasks=value
#SBATCH --nodes=value
The quantum espresso code is a well MPI-parallelized code and exhibits excellent scaling, so I would recommend the following slurm directives:
#!/bin/bash
#SBATCH --job-name=qe
#SBATCH --ntasks=48
#SBATCH --cpus-per-task=1
#SBATCH --time=48:00:00
#SBATCH --mem=0 # use all of the available shared memory
#SBATCH --mail-user=username@usc.edu
#SBATCH --mail-type=all
#SBATCH --nodes=2
#SBATCH --account=<account_id>
#SBATCH --partition=main
#SBATCH --constraint=xeon-4116 # 24 cpus per node
#SBATCH --exclusive
module purge
module load gcc/9.2.0
module load mpich/3.3.2
module load quantum-espresso/6.5-openblas-hdf5
module load fftw/3.3.8-mpich
module load openblas/0.3.7
module load netlib-scalapack/2.1.0-openblas-mpich
module load hdf5/1.10.6-mpich
cd ${SLURM_SUBMIT_DIR}
mpirun --bind-to core -np 48 pw.x -i x.md.in > x.md.out
For 3 nodes, change the following:
#SBATCH --ntasks=72
#SBATCH --nodes=3
mpirun --bind-to core -np 72 pw.x -i x.md.in > x.md.out
For 4 nodes, change the following:
#SBATCH --ntasks=96
#SBATCH --nodes=4
mpirun --bind-to core -np 96 pw.x -i x.md.in > x.md.out
And so on.
Thank you, Molguin. That’s a really good explanation.
Hi Molguin,
There is something that seems missing in QE module. When I was trying to use pw_export.x, it gives the error.
For this missing file, I find a possible solution here: https://www.mail-archive.com/users@lists.quantum-espresso.org/msg38364.html
Is it possible to do it on our system?
Thank you