Adaptive Acceleration of SSOR for Solving Large Linear Systems.
Abstract
Symmetric successive overrelaxation (SSOR) for solving large, sparse systems of linear equations involves the estimation of a parameter omega. An adaptive procedure is outlined for improving the estimates for omega and the spectral radius S(Sigma omega) of the iteration matrix Sigma omega. These estimates are then used in the SSOR method with Chebyshev acceleration. The objective is to achieve convergence in only a few more iterations than would be required if the best possible values of omega and S(Sigma omega) were used from the outset. The method is applied to obtain finite difference solutions of a number of generalized Dirichlet problems. In certain cases, the number of iterations required using the adaptive precedure increases like h to the minus 1/2 power, where h is the mesh size. (Author)
Document Details
- Document Type
- Technical Report
- Publication Date
- Mar 01, 1979
- Accession Number
- ADA069133
Entities
People
- Vitalius Benokraitis
Organizations
- Ballistic Research Laboratory