Parallel Computation of Flow Problems Using Explicit Schemes

Summary

Algorithms are presented for efficient computation of steady flows on parallel computers using explicit schemes. In parallel computing, one is concerned with computation and communication costs. Data parallelization is assumed where the flow domain is divided into sub-regions (blocks) which are connected to each other by smaller sub-regions (interfaces). In order to reduce the computation cost, two approaches are proposed: an explicit block solver with filtering in space, and an explicit block solver with filtering in time. For reducing the communication cost, an interface solver with a filter in time is developed.

For both filtering in space and time, the basic approach is to filter the high frequency components of the residual vector..For reducing the cost of communication, the information exchange between the neighboring blocks was conducted after several iteration steps. The resulting high frequency effects are then filtered in time.