Krylov subspace acceleration of waveform relaxation
- Univ. of Notre Dame, IN (United States)
Standard solution methods for numerically solving time-dependent problems typically begin by discretizing the problem on a uniform time grid and then sequentially solving for successive time points. The initial time discretization imposes a serialization to the solution process and limits parallel speedup to the speedup available from parallelizing the problem at any given time point. This bottleneck can be circumvented by the use of waveform methods in which multiple time-points of the different components of the solution are computed independently. With the waveform approach, a problem is first spatially decomposed and distributed among the processors of a parallel machine. Each processor then solves its own time-dependent subsystem over the entire interval of interest using previous iterates from other processors as inputs. Synchronization and communication between processors take place infrequently, and communication consists of large packets of information - discretized functions of time (i.e., waveforms).
- Research Organization:
- Front Range Scientific Computations, Inc., Lakewood, CO (United States)
- OSTI ID:
- 440702
- Report Number(s):
- CONF-9604167--Vol.2; ON: DE96015307
- Country of Publication:
- United States
- Language:
- English
Similar Records
The effects of problem partitioning, allocation, and granularity on the performance of multiple-processor systems
Time-parallel iterative methods for parabolic PDES: Multigrid waveform relaxation and time-parallel multigrid