The implementation and use of sparse matrix techniques in general simulation programs

Abstract

Classical matrix algebra requires excessive computer assets (core storage and execution time) when applied to the solution of simultaneous linear equations. Most physical systems, whose simulation eventually requires the solution of these equations, produce ‘sparse’ matrices. The approach described herein exploits this sparseness to minimise core storage, execution time, and round off error while providing a flexible programming base for a variety of related simulation operations.