The low-frequency wave propagation code LEMan has been parallelized. Due to large memory requirement but fast computation with the cold model, the parallelization is limited to a low number of processors. The specific block-tridiagonal structure of the matrix to be solved has been taken into account for the MPI implementation. It has then been compared with the performance of OpenMP in order to determine the optimal method depending on the case studied.
David Atienza Alonso, Tomas Teijeiro Campo, Una Pale
Arjen Lenstra, Robert Granger, Thorsten Kleinjung, Benjamin Pierre Charles Wesolowski
Wenzel Alban Jakob, Delio Aleardo Vicini, Sébastien Nicolas Speierer