ICS '88 Proceedings of the 2nd international conference on Supercomputing
A fast GPU implementation for solving sparse ill-posed linear equation systems
PPAM'09 Proceedings of the 8th international conference on Parallel processing and applied mathematics: Part I
Computational Optimization and Applications
| Hi-index | 0.00 |
Some domain decomposition methods consist in solving an interface problem by the mean of the conjugate gradient method. The condensed operator is obtained by the computation of local independent problems in each substructure. This leads to computational errors that entail the conjugate gradient algorithm to converge slowly. The number of iterations can be greatly reduced by complete reorthogonalization. The cost of this process is low compared to the resolution of all the local subproblems to be performed at each iteration.