Factorized-sparse-approximate-inverse preconditionings of linear systems with unsymmetric matrices

The paper considers the application of factorized-sparse-approximate-inverse (FSAI) preconditionings to linear algebraic systems with nonsingular unsymmetric coefficient matrices. Special attention is paid to methods for optimizing the sparsity pattern of such preconditioners. It is numerically demonstrated that the efficiency of FSAI preconditionings can substantially be improved by applying the so-called post-filtering and prefiltering in order to sparsify preconditioning matrices.