Speaker
Description
With the advent of super-computing, first-principle simulations became essential for the generation of contemporary knowledge, notably in condensed matter (1), but also in more applied fields, e.g., where the development of applications is based on the usage of intense laser radiation on solid materials (2–5).
Although numerous open source scientific codes are nowadays available, mastering these can require a long-term investment, making them not directly available to a wide range of users. In addition, long-term access to super-computing centers for research purposes remains challenging in the context of the current energy crisis, making benchmarking actions time-demanding in comparison to nowadays scientific production requirements. As a result, a benchmark of versatile and experimentally validated simulation techniques acquired along years of research experience can be of high value.
In this work, the data generated using time-dependent density functional theory (TDDFT) using the Octopus code (1) on a variety of European supercomputers was benchmarked by reusing production runs that were validated both on established theories (2) and on dedicated experiments (6). From the ~2,000 available simulations, the run-average of the time needed per iteration was acquired on a variety of 10 European super-computing facilities, counting among them some that were listed in the Top 500 world chart (top500.org).
By correlating the time measurements with specifications of the employed libraries, processors, compiler chains and their parameters, the results demonstrate the modularity of the employed code and suggest guidelines for researchers, developers and machine providers in view of offering a wider usage of the TDDFT in near future.
This work was supported by the European Regional Development Fund and the state budget of the Czech Republic (project BIATRI: CZ.02.1.01/0.0/0.0/15 003/0000445, project HiLASE CoE: No. CZ.02.1.01/0.0/0.0/15 006/0000674). Computational support was provided by the Ministry of Education, Youth and Sports of the Czech Republic through the e-INFRA CZ (ID:90140). The numerical results of this research have been partially achieved using the DECI resource Navigator based in Portugal at University of Cambria with support from the PRACE aisbl, and ELI Beamlines for the extra computational support (Eclipse HPC cluster).
References
1. N. Tancogne-Dejean et al., J Chem Phys. 152, 124119 (2020).
2. T. J.-Y. Derrien et al., Phys. Rev. B. 104, L241201 (2021).
3. I. Gnilitskyi et al., Sci. Rep. 7, 8485 (2017).
4. I. Gnilitskyi, L. Orazi, T. J.-Y. Derrien, N. M. Bulgakova, T. Mocek, Method of ultrafast laser writing of highly-regular periodic structures on metallic materials (2016), (available at https://patentscope.wipo.int/search/en/detail.jsf?docId=WO18010707).
5. T. J.-Y. Derrien, Y. Levy, N. M. Bulgakova, in Ultrafast Laser Nanostructuring: The Pursuit of Extreme Scales (Springer, 2022; https://link.springer.com/book/9783031147517), vol. 239 of Springer Series in Optical Sciences.
6. P. Suthar, F. Trojánek, P. Malý, T. J.-Y. Derrien, M. Kozák, Commun. Phys. 5, 288 (2022).
