Pangolin v1.0, a conservative 2-D advection model towards large-scale parallel calculation
Résumé
To exploit the possibilities of parallel comput-ers, we designed a large-scale bidimensional atmospheric advection model named Pangolin. As the basis for a fu-ture chemistry-transport model, a finite-volume approach for advection was chosen to ensure mass preservation and to ease parallelization. To overcome the pole restriction on time steps for a regular latitude–longitude grid, Pangolin uses a quasi-area-preserving reduced latitude–longitude grid. The features of the regular grid are exploited to reduce the mem-ory footprint and enable effective parallel performances. In addition, a custom domain decomposition algorithm is pre-sented. To assess the validity of the advection scheme, its re-sults are compared with state-of-the-art models on algebraic test cases. Finally, parallel performances are shown in terms of strong scaling and confirm the efficient scalability up to a few hundred cores.
Origine | Publication financée par une institution |
---|