Performance evaluation of the LBM solver Musubi on various HPC architectures

This contribution presents the performance of the Lattice Boltzmann implementation Musubi on four different High Performance Computing architectures. Musubi is maintained within the APES simulation framework that makes use of a distributed octree mesh representation and includes a mesh generation and a postprocessing tool to enable end-to-end parallel simulation work flows. An unstructured representation of the mesh is used, so only fluid elements are stored and computed for any arbitrary complex geometry with minimum user interference. Elements are serialized according to a space filling curve to ensure good spatial locality. The validity of our approach is demonstrated by the good performance and scaling behavior on the four HPC systems with minimal porting efforts.