The Slalom benchmark – it stands for Scalable Language-independent Ames Laboratory One-minute Measurement – is an increasingly popular benchmark capable of measuring architectures from single processors to massively parallel processing machines, according to Omri Selin’s Parallel Processing Report. The benchmark, designed by John Gustafson of Iowa State University’s Ames Laboratory, is described as inherently scalable, in contrast to other benchmarks such as Linpack, which often executed so quickly on supercomputers as to make the results meaningless. Some 100 systems have been measured, and Serlin lists the top performers. Top of the list comes the Intel Corp Delta with 400 40MHz 80860 processors, rated at 5,700 patches (the amount of work achieved in one minute) and 3200 MFLOPS. The next nine are as follows: Siemens-Nixdorf Informationssysteme AG S600 (one processor), Cray Research Inc Y-MP/8D (eight processors), Intel Delta (256), Cray Computer Corp 2S/4 (four), Cray Y-MP/8D (four), NCube Inc N-2 (1,024), Cray 2S/4 (two), Cray Y-MP/8D (two) and Intel Delta (64). Fujitsu Ltd’s VP 400 uniprocessor makes it in at number 13, MasPar Computer Corp’s 16,384-processor MP-1 at 17, IBM Corp’s 3090-200J with Vector Facility at 21, Alliant Computer Systems Corp’s 14-processor FX/2800 at 23, and the IBM RS/6000, not far behind the 3090, at 25. The RS/6000 Unix workstation’s rating is 1,610 patches and 63.5 MFLOPS.
