Numerical experience with a superfast real Toeplitz solver (Q1825588): Difference between revisions

The numerical stability of superfast real Toeplitz solvers for linear system of equations \(Mx=b\) are considered, where M is a Hermitian positive definite Toeplitz matrix. Knowledge of the numerical reliability for superfast methods is important because they allow for the solutions of problems which would require a prohibitive amount of time for a fast Toeplitz solver. A generalized Schur algorithm and its use in the superfast solution of positive definite Toeplitz system are described. Experimental results generated using a FORTRAN code of the generalized Schur algorithm are presented for real positive definite Toeplitz systems. The algorithm displays favorable behavior in the growth rates of the resulting residuals and errors are comparable with those of the Szegö recursions.