How to Vectorize the Algebraic Multi-level Iteration
Lutz Grosz
Abstract:
The paper discusses the optimal implementation of the
algebraic multi-level iteration (AMLI) on vector and parallel
computers. A suitable coarsening strategy for rectangular grids is
presented and its interaction with the processor architecture is
discussed. Examples show that the selected implementation of the
preconditioner can achieve nearly 90% of the possible computing
performance on a vector computer.
![[Back]](/images/prevpage.gif)
![[Index]](/images/index.gif)
![[Help]](/images/help.gif)
![[MSI]](/images/msi.gif)
![[ANU Online]](/images/online.gif)
