An Investigation of Three Computer Programs for the Solution of AX = B Where A is Symmetric and Sparse.

This report documents an investigation of three recent programs for the solution of AX B where A is symmetric and sparse the Yale Sparse Matrix Package the Munksgaard subroutines and the Mesztenyi-Rheinboldt subroutines. The first two programs compute in-core solutions the third uses random access to obtain its solution. The performance of these three programs is compared with that of a previously developed out-of-core equation solver, CSKYDG2. The two in-core equation solvers especially the first are faster than CSKYDG2 the third is slower. All three provide the same degree of accuracy as CSKYDG2, but they require large amounts of core storage. It would appear that the two in-core equation solvers are not suited for use on the CDC 6000 series of computers in their present form due to the limited amount of core storage available. Although the third equation solver does not require as much core storage, it does not perform as well as existing out-of-core euqation solvers which use the same or lesser amounts of core storage. Author