Open Access Open Access  Restricted Access Subscription Access
Open Access Open Access Open Access  Restricted Access Restricted Access Subscription Access

Sparse Symmetric Indefinite Systems and Polynomial Preconditioning


Affiliations
1 Department of Mathematics and Computer Science, Delaware State University Dover, Delaware 19901, United States
     

   Subscribe/Renew Journal


In this paper, a new algorithm is proposed to solve indefinite symmetric sparse linear systems using polynomial preconditioning. The preconditioning polynomials are derived using simple calculus techniques. In practice, this algorithm, unlike other algorithms proposed in the literature, does not require complete knowledge of the intervals containing the spectrum of the given matrix. This method is superior to CG on normal equations, SYMMLQ, and Saad's GCI algorithm. Also the algorithm is amenable for implementing on parallel machines. The performance of the algorithm is compared with Saad's GCI algorithm on the minisupercomputer Alliant FX/8.
User
Subscription Login to verify subscription
Notifications
Font Size

Abstract Views: 239

PDF Views: 0




  • Sparse Symmetric Indefinite Systems and Polynomial Preconditioning

Abstract Views: 239  |  PDF Views: 0

Authors

N. R. Nandakumar
Department of Mathematics and Computer Science, Delaware State University Dover, Delaware 19901, United States

Abstract


In this paper, a new algorithm is proposed to solve indefinite symmetric sparse linear systems using polynomial preconditioning. The preconditioning polynomials are derived using simple calculus techniques. In practice, this algorithm, unlike other algorithms proposed in the literature, does not require complete knowledge of the intervals containing the spectrum of the given matrix. This method is superior to CG on normal equations, SYMMLQ, and Saad's GCI algorithm. Also the algorithm is amenable for implementing on parallel machines. The performance of the algorithm is compared with Saad's GCI algorithm on the minisupercomputer Alliant FX/8.