© 1991 by Institute of Mathematics and its Applications
The Factorization of Sparse Symmetric Indefinite Matrices
Atlas Centre, Rutherford Appleton Laboratory Oxon OX11 0QX
Department of Mathematics, Utah State University Logan, Utah 84322-3900, USA
The Harwell multifrontal code MA27 is able to solve symmetric indefinite systems of linear equations such as those that arise from least-squares and constrained optimization algorithms, but may sometimes lead to many more arithmetic operations being needed to factorize the matrix than is required by other strategies. In this paper, we report on the results of our investigation of this problem. We have concentrated on seeking new strategies that preserve the multifrontal principle but follow the sparsity structure more closely in the case when some of the diagonal entries are zero.