Implementing techniques for elliptic problems on vector processors

To provide the arithmetic power required by large-scale numerical simulations, the fastest computers today incorporate vector processing. Two types of vector architecture are defined, and the variation in performance that can occur on a vector processor as a function of algorithm and implementation, the consequences of this variation, and the performance of some basic operators on the two classes of vector architecture are discussed. The performance of some higher-level operators that should be used with caution is also considered. Then the implementation of techniques for elliptic problems using the operators discussed previously is reviewed. Included are Fast Poisson solvers, dissection, and point, line, block, and conjugant gradient schemes. Finally, some areas of research are noted. 1 figure. (RWR