Enabling temporal blocking for a lattice Boltzmann flow solver through multicore-aware wavefront parallelization

This report presents first results of a pipelined wavefront parallelization approach applied to the lattice Boltzmann method in 3D. Threads executing successive time steps on a fixed spatial domain are scheduled to the same multi- core processor chip having a shared cache. The efficient reuse of data in cache by successive threads implements a multicore-aware temporal blocking parallelization in a rather simple way.