Parallel dynamic grid refinement for industrial applications

This paper covers a description of the algorithms and data-structures used in the parallel DLR TAU-code Adaptation tool. Attention is given to the basic algorithms of the parallel anisotropic refinement and de-refinement, as well as the data-structures for distributed grids and grid hierarchy. The parallelization issues for distributed data are discussed along with the dynamic repartitioning and the load-balancing algorithm. To demonstrate the dynamic character of the TAU-Code adaptation, we choose a simulation for a complex aircraft configuration and perform a parallel CFD calculation using varying numbers of CPUs. The obtained results show good parallel scalability with respect to memory and CPU-time consumption.