Skip to content

Cp2k

Installed versions

Resource Version
Dardel/cpe23.12 2024.3, 2024.1
Dardel-GPU/cpe23.12.gpu 2024.3, 2024.1

General information

CP2K is a program to perform atomistic and molecular simulations of solid state, liquid, molecular, and biological systems. For more information please visit https://www.cp2k.org.

How to use

The CP2K installation contained in this module was built with support for the programs and libraries ELPA, Libint-CP2K, libvori, libxc, libxsmm, PLUMED, and spglib. To display info on which environment variables are set when loading the module, use 

ml PDC/<version>
ml show cp2k/2024.3-cpeGNU-23.12
To load the CP2K module 
ml PDC/<version>
ml cp2k/2024.3-cpeGNU-23.12
Below follows an example job script for CP2K, for running on a single Dardel node using 16 MPI ranks and 8 threads. You need to replace pdc.staff with an active project that you belong to.

#!/bin/bash

#SBATCH -J cp2k-test
#SBATCH -A pdc.staff
#SBATCH -p main
#SBATCH -t 1-00:00:00

#SBATCH --nodes=1
#SBATCH --ntasks-per-node=16
#SBATCH --cpus-per-task=16

ml PDC/<version>
ml cp2k/2024.3-cpeGNU-23.12

export OMP_NUM_THREADS=8
export OMP_PLACES=cores

srun cp2k.psmp -i inputfile.inp -o logfile.log

Assuming the script is named jobscriptCP2K.sh, it can be submitted using 

sbatch jobscriptCP2K.sh

It can happen that the threads exceed their stack space which results in segmentation fault. If this happens, try increasing the stack space by adding the lines 

export OMP_STACKSIZE=256M
ulimit -Ss unlimited
to the job script. For the CP2K benchmark case H2O-dft-ls.NREP4.inp the optimal setting for running on 8 nodes on Dardel is

#!/bin/bash

#SBATCH -J cp2k-test
#SBATCH -A pdc.staff
#SBATCH -p main
#SBATCH -t 01:00:00

#SBATCH --nodes=8
#SBATCH --ntasks-per-node=16
#SBATCH --cpus-per-task=16

ml PDC/23.12
ml cp2k/2024.3-cpeGNU-23.12

export SRUN_CPUS_PER_TASK=$SLURM_CPUS_PER_TASK
export OMP_NUM_THREADS=8
export OMP_PLACES=cores

srun cp2k.psmp -i inputfile.inp -o logfile.log

For using CP2K together with PLUMED, we suggest to use 128 MPI ranks and 1 thread per node

#!/bin/bash

#SBATCH -J cp2k-test
#SBATCH -A pdc.staff
#SBATCH -p main
#SBATCH -t 1-00:00:00

#SBATCH --nodes=1
#SBATCH --ntasks-per-node=128
#SBATCH --cpus-per-task=1

ml PDC/23.12
ml cp2k/2024.3-cpeGNU-23.12

export SRUN_CPUS_PER_TASK=$SLURM_CPUS_PER_TASK
export OMP_NUM_THREADS=1
export OMP_PLACES=cores

srun cp2k.psmp -i inputfile.inp -o logfile.log

The executables for Dardel GPU nodes have been built with backend for AMD GPUs, with linking to GPU builds of the libraries COSMA, SpFFT, and SPLA.

To load the CP2K module

ml PDC/23.12
ml cp2k/2024.3-cpeGNU-23.12-gpu

Below follows an example job script for CP2K, for running on a single Dardel GPU node using 8 MPI tasks (corresponding to one MPI task per GPU) and 8 OpenMP threads. You need to replace pdc.staff with an active project that you belong to. Note: This script is a simple template. For efficient calculation the script needs to be augmented with settings to pin appropriately the computation threads to the CCDs and GCD.

Assuming the script is named jobscriptCP2K.sh, it can be submitted using 

sbatch jobscriptCP2K.sh

Please consult the official CP2K documentation for more details https://www.cp2k.org/

Known issue

If geometry optimization got stuck in first optimization step, consider adding PREFERRED_DIAG_LIBRARY SL in the &GLOBAL section. See https://github.com/cp2k/cp2k/issues/1696 for details.