PARALLEL PERFORMANCE OF APPLICATIONS ON SUPERCOMPUTERS

Parallel processing performance on a suite of application programs has been measured and analyzed. Many previous tests of parallel processing used synthetic or artificial programs to demonstrate parallel processing capabilities. For this test we have used the Perfect Benchmarks suite to test parallel processing on applications. Parallel processing is not the exclusive domain of so-called 'massively parallel processors'. Many other computer systems, ranging from high-end workstations, to minicomputers, to mainframes, to supercomputers, theoretically have parallel processing capabilities. In this article we list the performance for one and multiple CPUs of various supercomputers. The CRAY Y-MP system was able to exploit parallel processing on twelve of the 13 benchmark programs. We will report these results and will discuss parameters which characterize the efficiency of parallel processing.