Improving Interpolation in BoomerAMG
With new more aggressive coarsening algorithms that while reducing memory also degrade convergence often dramatically, it was imperative that new interpolation routines be implemented to recover this degradation. The implementation details and results for three new interpolation routines, standard, extended, and F-F, are presented in this paper. The project was focused on parallel implementation, so there is little theoretical analysis. The references contain much of the algorithmic design issues and analysis, if further understanding or exploration is needed, see [1][2][3]. It will be shown throughout this paper that long-range interpolation is needed for a number of problems, for some cases, there is almost a reduction in iterations by 2 orders of magnitude.
- Research Organization:
- Lawrence Livermore National Lab. (LLNL), Livermore, CA (United States)
- Sponsoring Organization:
- USDOE
- DOE Contract Number:
- W-7405-ENG-48
- OSTI ID:
- 894324
- Report Number(s):
- UCRL-TR-224345; TRN: US200701%%392
- Country of Publication:
- United States
- Language:
- English
Similar Records
Distance-Two Interpolation for Parallel Algebraic Multigrid
PipeSight: A High-Performance Computing Platform for Pipeline Integrity Management