1st March 2008
User documentation
This subroutine calculates scaling factors for a sparse symmetric or unsymmetric matrix \(\mathbf{A}\) = \({\{a _{ij}\}} _{m \times n}\) to make the \(p\)-norms of all the rows and columns be approximately equal to 1. In the symmetric case, the scaling is symmetric. The matrix may be stored by columns, in the cooordinate format, or as a packed dense matrix. If \(m \ne n\), it uses the \(\infty\)-norm. Otherwise, the user may choose any \(p\)-norm, with \(p \ge 1\).