In the distributed environment, it is crucial that during the mesh adaptation process all mesh entities communicating over part boundaries perform correct updates such that the complete mesh remains valid after its modification. This work presents the infrastructure to support parallel mesh modification operations with boundary layers using both simple point-to-point data exchange and interaction with mesh migration. To increase the scalability of large-scale simulations, this study introduces a general-purpose communication package built on top of MPI which is aimed at improving inter-processor communications independently of the supercomputer architecture being considered. The package is developed to support parallel applications that rely on computation characterized by a large number of messages of various sizes, often small, that are focused within processor neighborhoods. The parallel anisotropic mesh adaptation procedures are tested on general curved 3D geometries and appropriately manage the geometrical boundary approximation while maintaining the semi-structured nature of the elements in the boundary layers.

Autorentext

Aleksandr Ovcharenko, PhD: Studied Computer Science at Rensselaer Polytechnic Institute. HPC Developer at Blue Brain Project, EPFL.