t8code - Extreme Scale Adaptive Mesh Refinement with Arbitrary Elements

Abstract

Meshes are a core part of almost any numerical simulation code. A good choice of mesh is crucial to obtain accurate and efficient results. Adaptive mesh refinement (AMR) is the process of refining a mesh in regions of interest and coarsening it elsewhere. AMR can significantly improve the accuracy of the solution and simultaneously reduce the number of DOFs and hence the memory footprint and runtime. Managing adaptive meshes in parallel, however, is hard -- especially when the meshes are refined and coarsened frequently. In the past decades the forest-of-trees approach using space-filling-curves has been established as a fast and reliable method that offers both geometric flexibility and excellent parallel scalability. However, it was mostly limited to hexahedral meshes. With the AMR library t8code, we extend this approach to meshes with arbitrary element shapes and refinement patterns. We present our algorithms and their implementation and demonstrate strong and weak scalability to several hundred thousand MPI ranks.