Performance evaluation of scientific programs on advanced architecture computers

Recently a number of advanced architecture machines have become commercially available. These new machines promise better cost-performance then traditional computers, and some of them have the potential of competing with current supercomputers, such as the Cray X/MP, in terms of maximum performance. This paper describes an on-going project to evaluate a broad range of advanced architecture computers using a number of complete scientific application programs. The computers to be evaluated include (1) distributed-memory machines such as the NCUBE, INTEL and Caltech/JPL hypercubes, and the MEIKO computing surface, (2) shared-memory, bus architecture machines such as the Sequent Balance and the Alliant, (3) very long instruction word machines such as the Multiflow Trace 7/200 computer, (4) “traditional” supercomputers such as the Cray X/MP and Cray-2, and (5) SIMD machines such as the Connection Machine. Currently 11 application codes from a number of scientific disciplines have been selected, although it is not intended to run all codes on all machines. Results are presented for two of the codes (QCD and missile tracking), and future work is proposed