This paper considers a few variants of Krylov subspace techniques for solving linear systems on parallel computers. The goal of these variants is to avoid global dot-products which hamper parallelism in this class of methods. They are based on replacing the standard Euclidean inner product with a discrete inner product over polynomials. The set of knots for the discrete inner product is obtained by estimating eigenvalues of the coefficient matrix.
|Original language||English (US)|
|Number of pages||22|
|Journal||Applied Numerical Mathematics|
|State||Published - Jun 1999|
|Event||Proceedings of the 1997 15th IMACS World Congress on Scientific Computation, Modelling and Applied Mathematics - Berlin, Ger|
Duration: Aug 24 1997 → Aug 29 1997
Bibliographical noteFunding Information:
I Expanded version of a presentation given at the 1997 IMACS World Congress (Berlin, August 1997). This work was partially supported by NSF grant CCR-9618827, DARPA grant number NIST 60NANB2D1272, and by the Minnesota Supercomputer Institute. ∗ Corresponding author. Present address: LIP6 – Université Paris 6, 4, place Jussieu, 75252 Paris Cedex 05, France.