A composite step conjugate gradients squared algorithm for solving nonsymmetric linear systems (Q1334205): Difference between revisions

A new variant of the conjugate gradients squared (CGS) algorithm for solving non-symmetric linear systems, which is based on squaring the composite step bi-conjugate gradient (CSBCG) method [see \textit{R. E. Bank} and \textit{T. F. Chan}, ibid. 7, No. 1, 1-16 (1994; Zbl 0809.65025)], is presented. The created composite step CGS (CSCGS) combines the advantages of CGS (no multiplications by the transposed matrix and a faster convergence rate than BCG) with the handling of pivot (also called true, Galerkin or hard) break-downs. The CSCGS algorithm is given and details of implementation are widely discussed. Three numerical examples prove the better performance over CGS.