“…In fact, if one can show that the entries in the QR factorization converge rapidly to machine precision, one can avoid the computation of the entire factorization and use the limits directly instead in the factors, as soon as convergence is achieved, leading to signiÿcant computational savings in solving the linear system Ax = c, or the least squares problem min Ax − d 2 (where A = is still a Toeplitz matrix) by the QR factorization. If A is a symmetric, diagonally dominant, tridiagonal Toeplitz matrix, it is shown in Reference [14] that the diagonals of the LU factors of A converge and computational savings are possible. Similar properties also hold for cyclic reduction, see Reference [15]; but for the QR factorization we are not aware of any results in the literature.…”