A Robust and Efficient ILU that Incorporates the Growth of the Inverse Triangular Factors

In this paper we present a new ILU decomposition which is based on an existing sparse direct solver. In contrast to many incomplete LU decompositions this ILU incorporates information about the inverse factors L-1 and U-1 which have direct influence on the dropping strategy. We demonstrate in several large scale examples that this implementation constructs a robust preconditioner.