High Performance Preconditioning

TLDR: It is shown how a rather high performance can be achieved for the more effective preconditioners, such as successive over-relaxation and incomplete decompositions, on most vector computers if used in a straightforward manner.

Abstract: The discretization of second-order elliptic partial differential equations over three-dimensional rectangular regions, in general, leads to very large sparse linear systems. Because of their huge order and their sparseness, these systems can only be solved by iterative methods using powerful computers, e.g., vector supercomputers. Most of those methods are only attractive when used in combination with a so-called preconditioning matrix. Unfortunately, the more effective preconditioners, such as successive over-relaxation and incomplete decompositions, do not perform very well on most vector computers if used in a straightforward manner. In this paper it is shown how a rather high performance can be achieved for these preconditioners.