Vector processing on the Alliant FX/8 multiprocessor
Walid Abu-Sufah, Allen D. Malony · Proceedings of the International Conference on Parallel Processing · 1986
The Alliant FX/8 multiprocessor implements several high-speed computation ideas in software and hardware. Each of the 8 computational elements (CEs) has vector capabilities and multiprocessor support. Generally, the FX/8 delivers its highest processing rates when executing vector loops concurrently. In this paper, extensive empirical performance results are presented for vector processing on the FX/8. The vector kernels of the LANL BMK8al benchmark are used in the experiments. Each kernel is executed on 1 and 8 CEs and show the measured execution rate (in MFLOPS) as a function of vector length. The performance of 1 CE as a vector processor is analyzed by finding the vector lengths where vector processing exceeds that of scalar processing and calculating Hockney's n/sub 1/2/. For 8 CEs, upper/lower bounds are given on the achieved speedups and on the multiprocessing overhead. The speedup variation is shown as the number of CEs increases from 2 to 8. 25 refs., 8 figs., 7 tabs.