Hierarchical Dynamic Load Balancing Strategy for a p-adaptive Discontinuous Galerkin Compressible LES Solver

The current work proposes a dynamic hierarchical load balancing strategy for a p-adaptive local discontinuous Galerkin compressible LES solver. Firstly, the load balancing problem is properly reformulated as a graph partitioning problem. Then, the graph partitioning step is solved exploiting the partitioning capabilities of the Zoltan library. Finally, this approach is validated on a classic benchmark for LES simulations, that is, the flow past a square cylinder. The hardware-informed load balancing capabilities and the overall performance of the proposed strategy are discussed. Lastly, the influence of load balancing on the results is investigated. Results demonstrate the goodness of the current framework, that allows for dynamically balanced adaptive simulations of complex and unsteady turbulent flows.