Parallel Pipelining SOR Method for an Iterative Linear Equation Solver on a Distributed Memory Parallel Processor
スポンサーリンク
概要
- 論文の詳細を見る
In this paper a Parallel Pipelining SOR (ppSOR) method and an iterative linear equation solver using the ppSOR method are presented. The experimental results on a distributed memory parallel processor NEC cenju-3 with 16 PEs, are also pressented. It is shown that the convergence property of ppSOR method is as same as the multi-color method. However, this method have an ■ not to require to coloring nodes of the graph which is associated with the coefficient matrix to be solved. The experimental results show that 10 times speed up is achieved on a parallel processor with 16 PE compared to a single PE processor.
- 東海大学の論文
著者
関連論文
- Processor Architecture and Evaluation which Correspond to Deviation of Memory Latency
- Parallel Pipelining SOR Method for an Iterative Linear Equation Solver on a Distributed Memory Parallel Processor
- Design of A Memory Latency Tolerant Processor (SCALT)