A note on conjugate gradient convergence. II (Q1576610)

Following their first paper on this subject [cf. \textit{A. E. Naiman}, \textit{I. M. Babuška} and \textit{H. C. Elman}, ibid. 76, No. 2, 209-230 (1997; Zbl 0905.65047)], the authors continue the investigation on conjugate gradient convergence for special system and very special right-hand sides (r.h.s.). Unfortunately, no motivation is given for this very narrow focus. For the particular r.h.s. the number of conjugate gradient steps to achieve tolerance is proportional to the number of equations. Actually, as the matrix is derived from the discretization of a boundary value problem, it would be more realistic if the given tolerance went also down in proportion to the discretization error.