Performance and Profiling Data of Plane-wave Calculations in Quantum ESPRESSO Simulation on Three Supercomputing Centres

This data was the work of the Thammasat University (Lampang Campus) team from benchmarking the Quantum Espresso simulation to complete a task of the 5th APAC HPC-AI 2022 student competition. The benchmarking and profiling results guided the team to adjust some parameters until achieving the best parameter setting guideline to obtain almost linear scalability of the plane-wave calculation (pw.x) on 48 - 1,536 CPUs. The findings led the team to win the best HPC performance award of the competition. The dataset comprises three parts i.e., the execution traces collected by the Extrae performance profiling tool, the results of a screening experiment running on three HPC centres, and the multi-node performance results when scaling out the workload on the Gadi supercomputer. The profiling data can be used to gain insights into the parallel and distributed execution behaviour of the pw.x function by using the Paraver visualization tool. The performance data of the screening and scaling out experiments could be used for testing different analysis methods or for finding a way to automate the parameter setting. The results also give guidance on computation resources required to calculate the total energy of the catalysts using the self-consistent field method implemented in the Quantum Espresso simulation.