Analysis of Dissection Algorithms for Vector Computers.

Recently two dissection algorithms one-way and incomplete nested dissection have been developed for sloving the sparse positive definite linear systems arising from n by n grid problems. Concurrently, vector computers such as the CDC STAR-100 and TI ASC have been developed for large scientific applications. An analysis of the use of dissection algorithms on vector computers dictates that vectors of maximum length be utilized thereby implying little or no dissection on the other hand, minimizing multiplication counts suggests that considerable dissection be performed. In this paper we discuss the resolution of this conflict by minimizing the total time required by vectorized versions of the two algorithms. Author