Structured multi-block grid partitioning using balanced cut trees

Kurzfassung

An algorithm to partition structured multi-block hexahedral grids for a load balanced assignment of the partitions to a given number of bins is presented. It uses a balanced hierarchical cut tree data structure to partition the structured blocks into structured partitions. The refinement of the cut tree attempts to generate equally shaped partitions with a low amount of additional surface. A multi-block load balancing approach is presented that guarantees to satisfy an upper bound of load imbalance. The partition quality of the algorithm is compared to established recursive edge bisection approaches and an unstructured partitioning using METIS. Two generic and two turbomachinery test cases demonstrate the superior quality and fast runtime of the present algorithm at generating load balanced structured partitions.