Reduction of a general matrix to tridiagonal form using a hypercube multiprocessor
Recently there has been a renewed interest in finding reliable methods of reducing general matrices to tridiagonal form. We have developed a serial reduction algorithm that appears to be very reliable in practice. In this paper we describe a parallel version of our algorithm, which has been implemented using a portable library developed at Oak Ridge National Laboratory (ORNL). The library allows the code to be portable across most commercial hypercubes. The algorithm was developed as one step in the process of finding eigenvalues of nonsymmetric matrices. Our original parallel eigenvalue routines reduced the matrix to Hessenberg form and then applied QR iteration, but their performance was disappointing. Our new parallel routines reduce the matrix to tridiagonal form and then apply LR iteration. Using an iPSC/2, we compare the performance of the new parallel routines with our previous parallel routines and show that the new routines are nearly an order of magnitude faster, allowing us to solve much larger problems than previously attempted. 14 refs., 6 figs., 2 tabs.
- Research Organization:
- Oak Ridge National Lab., TN (USA)
- DOE Contract Number:
- AC05-84OR21400
- OSTI ID:
- 6446177
- Report Number(s):
- CONF-890372-3; ON: DE89008721
- Resource Relation:
- Conference: 4. hypercube, concurrent computers and applications, Monterey, CA, USA, 6 Mar 1989; Other Information: Portions of this document are illegible in microfiche products
- Country of Publication:
- United States
- Language:
- English
Similar Records
Reduction of a general matrix to tridiagonal form
The design of a parallel, dense linear algebra software library: Reduction to Hessenberg, tridiagonal, and bidiagonal form