Remarks on a displacement-rank inversion method for Toeplitz systems

Comments are made regarding the implementation of a Toeplitz-matrix inversion algorithm described by Bitmead and Anderson in 1]. We show that although the algorithm is asymptotically efficient with O(N(logN)²) operations, it requires a 10⁶×10⁶ matrix to break even with the class of algorithms whose operation count is of the order of O(N²) (as found in 4]).