Massive memory buys little speed for complete, in-core sparse Cholesky factorizations on some scalar computers (Q803714)

The author shows that the benefits of using large amounts of memory in direct factorizations of sparse, symmetric, positive-definite matrices more modest than pointed out by \textit{W. Murray} [Methods for linear programming, talk presented at the NATO Advanced Workshop on Algorithms and Model Formulations in Mathematical Programming, Bergen, Norway, 15-19 June 1987]. He gives the results of computational experiments performed on a variety of problems using IBM 3081 K, VAX 8550 and Sun-3/160. Relevant details about sparse Cholesky and data structures are described.