A priori worst case error bounds for floating-point computations | IEEE Journals & Magazine | IEEE Xplore