Parallel Performance of two different applications of a Domain Decomposition technique to the Jacobi-Davidson method
- Speaker:
- Menno Genseberger, Delft Hydraulics
- Date/time:
- Friday 6 July 2007, 13:00
- Location:
- M.I. room K8b
Abstract
The Jacobi-Davidson method is an iterative method suitable for computing solutions of large eigenvalue problems. Most computational work of Jacobi-Davidson is due to a so-called correction equation at the intermediate level. In previous research a strategy for the computation of (approximate) solutions of the correction equation was proposed. The strategy is based on a domain decomposition technique in order to reduce wall clock time and local memory requirements.
This talk discusses the aspect that the original strategy can be improved by taking into account the relation of the intermediate level with the top level of the Jacobi-Davidson method. This results in a different application of the domain decomposition technique to the Jacobi-Davidson method. Although the two approaches look similar, there are subtle differences in implementation and the consequences in terms of computational time for large scale eigenvalue problems are nontrivial. Therefore, the parallel performance of the two approaches has been investigated with scaling experiments.