EECS Publication
Finite-choice algorithm optimization in Conjugate Gradients
Jack Dongarra and Victor Eijkhout
We present computational aspects of mathematically equivalent implementations of the Conjugate Gradient method. Since the implementations have different performance characteristics, but compute the same quantities, an adaptive implementation can at run time pick the optimal choice.
Published 2003-01-01 05:00:00 as ut-cs-03-502 (ID:208)