Double Ordering and Fill-In for the LU Factorization

We present a new method, called reversed double ordering, for reordering arbitrary matrices prior to LU factorization. This reordering creates a variable-band matrix. We compare the fill-in of the LU factorization for sparse matrices with respect to reversed double ordering, column minimum degree ordering, and the reversed Cuthill--McKee algorithm. Moreover, we combine the first two reorderings with good success.