Load Balancing for Adaptively Refined Grids

The solution of partial differential equations on a parallel computer is usually done by a data parallel approach. The grid is partitioned and mapped onto the processors. However, partitioning of unstructured meshes and adaptively refined meshes in general is an NP‐hard problem and heuristics are needed. In this paper a parallelisable and cheap method based on space‐filling curves is analysed. Quasi‐optimal estimates are derived for partitions of adaptively refined grids.