Convergence for Increasing Numbers of Blocks

Kurzfassung

In order to exploit CPU power of parallel machines, it is often necessary to use all available processors. Running parallel codes on parallel machines with growing numbers of processors requires a subdivision of the grid into at least the same number of blocks. From an algorithmic point of view the algorithm per block remains the same, but the overall computation is influenced by the subdivision. In this section we present some investigations, how a subdivision into more and more blocks influences the convergence. Here, we consider the parallel codes FLOWer and LSS, which use different discretizations and different solvers.