Block reduction of matrices to condensed forms for eigenvalue computations (Q1822900)

The authors describe block algorithms for the reduction of a (real) symmetric matrix to a tridiagonal form and that of a general matrix to a bidiagonal form by using Householder transformations. The same approach can be used in the reduction to Hessenberg form. These reductions to condensed forms comprise a preliminary step in the computation of eigenvalues or singular values. The authors also show how these reductions may be pipelined with the divide and conquer technique for computing the eigensystem of a symmetric matrix or the singular value decomposition of a general matrix. These considerations have significant performance advantages on parallel-vector processors.