Computing the singular value decomposition on a distributed system of vector processors (Q1122941)

The two-sided Jacobi algorithm of \textit{G. E. Forsythe} and \textit{P. Henrici} [Trans. Am. Math. Soc. 94, 1-23 (1960; Zbl 0092.325)] is implemented on a LCAP system consisting of a ring of vector processors. Two block versions are compared on a series of random test matrices, and it is demonstrated that the computation is nearly perfectly sped up turning the speed of the subproblem solver into the decisive factor.