Author(s): Jose Segovia-Burillo; Mario Morales-Hernandez; Sergio Martinez-Aranda; Pilar Garcia-Navarro
Linked Author(s): Pilar García-Navarro
Keywords: HPC; Non-hydrostatic models; Depth-averaged models; Shallow flows
Abstract: In the context of free-surface flow simulations, Shallow Water Equations (SWE) models solved using finite-volume numerical schemes are common. Certain configurations involving complex flows exhibit non-negligible vertical accelerations, and the implementation of a non-hydrostatic formulation can enhance the capture of flow dynamics and improve simulation accuracy. Here, we explore the implementation of 2D non-hydrostatic models in the open-source code SERGHEI from a High-Performance Computing (HPC) perspective. Two main approaches are followed based on the Divergence Integrated Incompressibility Condition (DIIC): a Hyperbolic-Elliptic (HE) model and a fully-Hyperbolic model (HR). Top-tier libraries such as Kokkos and Trilinos are used to efficiently calculate fluxes in the HR model and solve the sparse matrix system in the HE model, ensuring the performance and portability of the code. A performance-aware comparison of these two models is of great interest, due to the complexity of the problem and the bottlenecks regarding matrix inversions and
Year: 2026