The Principal Investigators' will perform a set of benchmarks of high-performance advanced-architecture computers on a total of nine scientific programs spanning many fields and algorithms. The Principal Investigators' use the CRAY-XMP as a yardstick and compare this to the leading commercially available parallel machines including the hypercube, and other distributed memory, VLIW and Shared Memory architectures. The study will involve leading computational scientists in each field and experts in specific machine architectures. These will be interfaced by groups at Caltech and at the San Diego Supercomputer Center, San Diego, experienced on using and optimizing implementations on vector and parallel computers. This will be a one-year pilot project that will develop methodology for larger scale performance evaluation for advanced architecture computers and begin to gather the data needed to compare the price-performance of emerging supercomputer architectures with that of current supercomputers. The Program Director for Computational Science and Engineering recommends support in the sum of $195,700 for a period of one year.
