Generalized ADI preconditioning (Q1066614)

The two notes comment on the combination of different efficient tools for the solution of linear equations as the ADI-method, conjugate gradients and multigrid methods. For example, the author refers to the somewhat unexpected result of \textit{R. E. Bank} and \textit{C. C. Douglas} [SIAM J. Numer. Anal. 22, 617-633 (1985; reviewed above)] that conjugate gradients seem to work well when used as smoothers in multigrid algorithms. This is surprising since this approach does not attach any less importance to low-mode error reduction than to high-mode reduction. \{Reviewer's remark: There are indeed examples in which conjugate gradients are less efficient than classical smoothers and to which the author's comments apply.\}