BENCHMARK EVALUATION OF THE IBM SP2 FOR PARALLEL SIGNAL-PROCESSING

This paper evaluates the IBM SP2 architecture, the AIX parallel progra mming environment, and the IBM message-passing library (MPL) through S TAP (Space-Time Adaptive Processing) benchmark experiments. Only coars e-grain parallelism was exploited on the SP2 due to its high communica tion overhead. A new parallelization scheme is developed for programmi ng message passing multicomputers. Parallel STAP benchmark structures are illustrated with domain decomposition, efficient mapping of partit ioned programs, acid optimization of collective communication operatio ns, We measure the SP2 performance in terms of execution time, Gflop/s rate, speedup over a single SP2 node, and overall system utilization. With 256 nodes, the Maul SP2 demonstrated the best performance of 23 Gflop/s in executing the High-Order Post-Doppler program, correspondin g to a 34% system utilization. We have conducted a scalability analysi s to reveal the performance growth rate as a function of machine size and STAP problem size. Important lessons teamed from these parallel pr ocessing benchmark experiments are discussed in the context of real-ti me, adaptive, radar signal processing on massively parallel processors (MPP).