Skip to content

Latest commit

 

History

History

quantum-espresso

README.md

Quantum-Espresso Benchmark

Based on the PRACE Unified European Applications Benchmark Suite medium testcase, modified to be relativistic.

  1. Download source code from https://gitlab.com/QEF/q-e/-/archive/qe-7.3/q-e-qe-7.3.tar.gz (md5sum: ac279db8330c257ed24449088148c247).

  2. Extract source code: tar -xzf q-e-qe-7.3.tar.gz

  3. Modify the environment appropriately to use the correct compiler and optimization flags for the system in question. Make sure the mpirun command is available. In the reference benchmark StdEnv/2023 with openmpi/4.1.5 and gcc/8.3.0 are used. There is a list of commands that were run to build Quantum-Espresso on Graham in the build_qe_GP.sh file.

  4. In the benchmark directory of quantum-espresso run:

    export QE_HOME="/path/to/quantum-espresso-install"
export PATH="$QE_HOME/bin:$PATH"

export OMP_NUM_THREADS=1
srun pw.x -inp Ta2O5-2x2xz-552.in > run_GP.out

    or:

    mpirun -np NUM_PROCS pw.x -inp Ta2O5-2x2xz-552.in > run.out

    where NUM_PROCS corresponds to the number of MPI processes being launched depending on which architecture is being used. Alternatively, if running Quantum Espresso in hybrid mode, --map-by node:pe=$NUM_CPUS_PER_NODE will need to be added as an mpirun argument, and possibly OMP_NUM_THREADS=$NUM_CPUS_PER_NODE. An example SLURM script is provided, submit_qe.sh that was used to produce the reference benchmark on Graham.

  5. Note the benchmark problem size fits on a minimum of 8 nodes of GP(Graham) with 125GB memory per node.

  6. Note that the code will automatically calculate the running time of the code, and output these numbers. The final output line which reads "total cpu time spent up to now is" (the line just before "End of self-consistent calculation"), in seconds, should be reported in the "QE" tab of the benchmark spreadsheet.

  7. All modified source code, output logs, and solution files are to be provided in with the response.