Setting parameters in the Yambo new options for large systems
- Name: Setting parameters in the Yambo new options for large systems
- EuroHPC machine used: LUMI
- Topic: Physical sciences; Electronic engineering
Overview of the project
The Scaling memory use in Yambo code for spectroscopic applications project concerns the first technical part of the larger project focused on the fiber lasers and the spectroscopy of the rare earths in glass. The whole problem requires accurate calculations of the absorption and emission spectra with the non-linear optical effects and then using them in the classical equations for the light propagation. In the quantum-mechanical part, the elementary cells contain 100-200 atoms treated with the ab initio second-quantization methods, thus being very demanding for the memory use.
The first benchmark project is intended to obtain the parallelization parameters and memory use of the YAMBO code used for solving the Bethe-Salpeter equation without phonons, which means only the absorption spectrum at zero temperature, and its scalability versus the number of atoms and the level of accuracy in the numerical integration. It is planned to continue with the next benchmark projects for the calculations with phonons, which means applications to finite temperature absorption and emission spectra, as well as the nonlinear effects in the future.
How did EPICURE support the project and what were the benefits of the support?
“The challenge was to pursue very large calculations that used to break (not go to the end) due to memory consumption and/or very complex parallelization architecture of the code (different in each part of the code). It was necessary to prepare tables with the input and script setups that show the code performance. These tables serve me now as “technical manual” for planning future production grants and setting the calculations in the recent grant. EPICURE support prepared the diagnostic tools and helped to solved problems with the cutting-edge calculations for large systems. The tables could be officially published for all users of the YAMBO code.
The most important benefit from the EPICURE support was the ability to perform the calculations that were not feasible before. Of course, we saved computing and human time having an overview for the performance of the code for large systems. Knowledge of the specific parallelization setup for each part in these sequenced-runs of this block-code makes a step towards further automation.”