General Information

Quantum ESPRESSO (QE) is an integrated suite of Open-Source computer codes for electronic-structure calculations and materials modeling at the nanoscale. It is based on density-functional theory, plane waves, and pseudopotentials.

Discover available versions of Quantum ESPRESSO

ml av qe

Running Quantum ESPRESSO

QE is compiled using an Intel Classic compiler, linked against the libraries

• wannier90
• HDF5
• libXC

and can be parallelized using OpenMP, MPI or a combination thereof.

Tipps & Tricks

• Familiarize with QE's levels of parallelization.
• When the memory of a single node (while using all available CPU cores) is a concern, then try increasing the number of nodes. QE will automatically distribute calculations and datastructures across all tasks, so the memory per task decreases with increasing tasks.
• Pay attention to Notes and Warnings regarding parallelization at the beginning of QE output. If the number of tasks or nodes is not appropriate, QE will spill out warnings. Do not run such ill-parallelized calculations! Typical warnings are:
  • WARNING: too many processors for an effective parallelization!
  • suboptimal parallelization: some nodes have no k-points!

Sample Slurm Job

For QE the recommendation is to request relative ressources ( *-per-*= ) and first increase --tasks-per-node up to 128 before increasing --nodes to avoid inter-node communication, which is slower than intra-node communication. 

#!/usr/bin/env bash
#SBATCH --job-name=qe
#SBATCH --partition=epyc
#SBATCH --nodes=1
#SBATCH --tasks-per-node=16
#SBATCH --cpus-per-task=1 # recommended
#SBATCH --mem-per-cpu=4G
#SBATCH --mail-type=END,INVALID_DEPEND,TIME_LIMIT
# replace the email with your personal one in order to receive mail notifications:
#SBATCH --mail-user=noreply@physik.uni-augsburg.de
#SBATCH --time=1-0

ml purge
ml load qe/7.3

srun pw.x < job.in > job-${SLURM_JOB_ID}.out

Running Quantum ESPRESSO on GPU

A seperate module (look out for the red built for GPU(g)flag when using ml av qe ) provides a version of QE compiled with NVHPC-compiler which allows running most (but not all) functionalities of QE on GPU.

#!/usr/bin/env bash
#SBATCH --job-name=qe
#SBATCH --partition=epyc-gpu
#SBATCH --nodes=1
#SBATCH --tasks-per-node=2 # (between 1-3 for epyc-gpu nodes)
#SBATCH --cpus-per-task=1 # recommended
#SBATCH --gpus-per-task=1 # recommended
#SBATCH --mem-per-cpu=4G
#SBATCH --mail-type=END,INVALID_DEPEND,TIME_LIMIT
# replace the email with your personal one in order to receive mail notifications:
#SBATCH --mail-user=noreply@physik.uni-augsburg.de
#SBATCH --time=1-0

ml purge
ml load qe/7.3-ompi4.1-nvhpc24.1

srun pw.x < job.in > job-${SLURM_JOB_ID}.out

Pseudopotentials

All potentials of the have been made available via the environment variable PSEUDO_DIR as part of the modules. Serverl different versions of SSSP potentials are located in their respective $SUBDIR folders

Benchmarks

AUSURF112 (CPU, small size test case)

*normalized to a full socket (64 cores). Here the unoccupied cores were empty and CPU clock was higher, almost 200% for a serial calculation. Since this is hard to seperate, the reference (100%) is a full socket.

AUSURF112 (GPU, small size test case)

GRIR443 (CPU, medium size test case)

GRIR443 (GPU, medium size test case)

Support

If you have any problems with Quantum ESPRESSO please contact the team of IT-Physik (preferred) or the HPC-Servicedesk.