Author(s): Pablo Valles; Jose Segovia-Burillo; Mario Morales-Hernandez; Sergio Iserte; Antonio J. Pena; Daniel Caviedes-Voullieme
Linked Author(s):
Keywords: HPC; Malleability; Shallow water solver; Dynamic load balancing
Abstract: This work integrates the SERGHEI-SWE hydrodynamic model with the Dynamic Management of Resources library (DMRlib) to enable malleability, i.e., the runtime adjustment of MPI process counts thereby optimizing computational resource usage in shallow-flow simulations. That is, the approach allows to provide more computational resources when the demand is larger, i.e., when the flood extent is larger. The performance of the malleable SERGHEI-SWE is evaluated using several test cases. We find that the malleable mechanism preserves numerical accuracy, and that dynamic resource management can improve resource use efficiency, with runtimes faster than the minimal set of resources, and slower than those obtained with the largest set of resources. These findings highlight the potential of the integration for large-scale simulations while clearly pointing to the need for improved heuristics in the criteria leading to dynamically changing resource sets.
Year: 2026