Parallelization of the SurfWB-UC code for the numerical simulation of nonlinear depth-averaged shallow water waves
Résumé
The SurfWB-UC model is an implementation of a set of algorithms aimed at solving the Non-Linear Shallow Water equations, with applications ranging from river to estuarine and tsunami flows. As a predictive tool, the computed results may be required in real time. For research purposes, reducing computation time from days to hours could enhance results and foster a deeper understanding of the simulated phenomena. This work presents an MPI parallel version of the code and conducts a performance analysis on a particularly simple yet challenging test case, which was run on the cluster machine at the Escuela de Ingeniería, Pontificia Universidad Católica de Chile. Experiments suggest that the code is scalable and can achieve a speedup of order 10 for the range of cores currently available in the cluster. Further tests should be conducted for debugging, handling exceptions, and complementing the performance analysis developed herein. This work is the result of a collaboration at the hydraulics department, presented as a final project in the high-performance computing workshop organized by the computational mathematics institute (Instituto de Matemática computacional, IMC) of the Pontificia Universidad Católica de Chile in 2014.
Origine | Fichiers produits par l'(les) auteur(s) |
---|