A parallel processed scheme for the eigenproblem of positive definite matrices (Q1347172)

The largest eigenvalue of a symmetric matrix is computed by a Jacobi type algorithm that in each iteration zeroes out the first column with a Householder reflection from the left. Application from the right fills the zeros again, but the norm of the first column will tend to zero with a rate comparable to the power method. Similar methods have been proposed from time to time, see e.g. \textit{C. D. LaBudde} [Math. Comput. 18, 118-123 (1964; Zbl 0115.344)] or \textit{A. Ruhe} [BIT 20, 88-96 (1980; Zbl 0435.65029)].