Dynamic mesh adaption on unstructured grids is a powerful tool for computing large-scale unsteady three-dimensional problems that require grid modifications to efficiently resolve solution features. By locally refining and coarsening the mesh to capture flowfield phenomena of interest, such procedures make standard computational methods more cost effective. We have developed a dynamic tetrahedral mesh adaption procedure that uses a data structure based on edges of the mesh. This makes the adaption method capable of performing anisotropic refinement and coarsening. Modifications have been made for a hexahedral adaption scheme to bypass the mesh quality problems associated with repeated anisotropic subdivision of a tetrahedral mesh.

An efficient parallel implementation of these adaptive methods is extremely difficult to achieve, primarily due to the load imbalance created by the dynamically-changing nonuniform grid. However, it is generally expected that unstructured adaptive-grid techniques will constitute a significant fraction of future high-performance computing. A novel method will be presented that dynamically balances the processor workloads with a global view. Mesh adaption, dynamic repartitioning, processor assignment, and data remapping are critical components of the framework that must be accomplished rapidly and efficiently so as not to cause a significant overhead to the numerical simulation. Potential bottlenecks are resolved to demonstrate that our scheme remains viable on a large number of processors.

Back to Workshop Schedule

1996-1997 Mathematics in High Performance Computing