Parallel sparse Cholesky factorization

In this paper we describe algorithms for the ordering and numerical factorization step in parallel sparse Cholesky factorization. Direct methods for solving sparse positive definite systems play an important role in many scientific applications such as linear programming and structural engineering. The importance of direct methods is mainly due to their generality and robustness. The paper describes minimum degree and nested dissection based ordering methods and presents a scalable parallel algorithm for the factorization of sparse matrices. The interested reader will find many references to the relevant literature.