A PARALLEL ALGORITHM FOR THE SYMMETRICAL EIGENVALUE PROBLEM

We present a parallel divide-and-conquer algorithm for the symmetric t ridiagonal eigenvalue problem implemented on a transputer network. It is based on a serial algorithm already presented by Bini & Pan[1] whos e main features are: i) the insensitivity to how tightly eigenvalues a re clustered and ii) the computation of the eigenvalues independently of the eigenvectors. Our proposed algorithm is faster than the origina l one even running on a single transputer and it guarantees more accur acy in the computation of the eigenvalues. The parallel divide-and-con quer approach is known to be problem dependent. This important issue i s studied in order to foresee the behavior of a parallel computation f or a generic matrix. For this purpose we use the definition of efficac y of a parallel algorithm.