Solving sparse linear least-squares problems on some supercomputers by using large dense blocks (Q1371662)

``On many high-speed computers a dense matrix technique is preferable to sparse matrix technique when the matrices are not large, because the high computational speed compensates fully the disadvantages of using more arithmetic operations and more storage.'' Statements like this one in the text are very dangerous at least when matrices from the discretization of partial differential equations are involved. Fortunately, the authors consider the case in which the matrix consists of blocks of dense submatrices and the fast routines for dense matrix computations are applied to the blocks. In particular linear least squares problems are considered. \{Reviewer's remark: Efficient algorithms for sparse matrix problems often use quite different methods for the solution and cannot be considered as a reduction of the corresponding method for dense matrices\}.