CaDS Seminar 2022 - May 10

Daniel Caviedes-Voullième (SDL Terrestrial Systems)

Re-assessing (and hopefully improving) performance of the heterogeneous TSMP in the DEEP-SEA project

Abstract:
The Terrestrial Systems Modelling Platform (TSMP) is a scale-consistent, highly modular, massively parallel, fully integrated soil-vegetation-atmosphere modelling system. Currently, TSMP is based on the COSMO atmospheric model, the CLM land surface model, and the ParFlow hydrological model, linked together by means of the OASIS3-MCT library. Recently, ParFlow was ported to GPU, therefore enabling the possibility of running TSMP under a heterogeneous configuration (COSMO and CLM running on CPUs, and ParFlow running on GPUs). The different computational demands of each submodel inherently result in non-trivial load balancing across the submodels. This prompts the need to re-assess what we knew about optimal configurations and scaling behaviour. Consequently, performance, tracing and scaling studies are on their way, and first results are available. In particular, we are interested in understanding the differences between homogeneous (CPU-only) and heterogeneous (CPU-GPU) configurations, as we move into fully modular setups. This work is strongly related to the participation of TSMP in the DEEP-SEA project, for which such performance assessment is a first step. In this talk, we will also discuss conceptually some of the solutions that TSMP will explore in the DEEP-SEA project.