Towards a cost-effective ILU preconditioner with high level fill (Q1198973): Difference between revisions

The utilization of preconditioned conjugate gradient methods provides a robust and competitive solution to large sparse matrix problems. The main factors which have a significant impact on the performance of the preconditioner are (i) the ordering of the unknowns in the original matrix \(A\); (ii) the sparsity pattern of the preconditioner \(M\); and (iii) the closeness of the respective spectra of \(M\) and \(A\). In a previous work [SIAM J. Matrix Analysis Appl. 13, No. 3, 944-961 (1992; Zbl 0760.65028)] the authors had proposed a method referred to as minimum discarded fill ordering strategy for general sparse matrices which more especially considers the effect of ordering on the preconditioner. In the present paper, several variants of this method are investigated. Detailed description of these ordering techniques are given, with test problems and numerical results.