Parallel FDTD Computations Optimized by Program Macro Data Flow Graph Redeployment

In this paper, we present and compare two methods for designing and profiling macro data flow graphs, which represent computation and communication patterns for the FDTD (finite difference time domain) problem in irregular computational areas. With the first method, optimized macro data flow graphs (MDFG) for FDTD computations are generated in three main phases: generation of initial MDFG based on wave propagation area partitioning, MDFG nodes merging with load balancing to obtain given number of macro nodes and communication optimization to minimize and balance internode data transmissions. With the second method, we use a modified CDC (connectivity-based distributed node clustering) algorithm to create an optimized macro data flow graph. The efficiency and execution time for both of these methods are compared. Computation efficiency for different several communication systems (MPI, RDMA RB, SHMEM) is discussed. All presented experimental results have been obtained by simulation