摘要
大规模并行计算机(MPC)的出现和发展迫切要求有新的并行算法设计理论和技术来指导更实际的并行算法的设计。本文首先简单介绍了针对MPC提出的LogP和Barier-LogP并行计算模型,然后借助于Barrier-LogP模型从通信平衡、数据分配和重叠通信与计算这三个方面讨论了更实际的并行算法设计的一般方法和技巧。
The emergence and development of massively parallel computers require new parallel algorithm design theories and techniques. In this paper, we first introduce the LogP and Barrier LogP parallel computation models. Then we discuss three principles of more realistic parallel algorithm design with the aid of Barrier LogP model. The principles proposed are balancing interprocessor communication, optimizing data placement and overlapping communication with computation. Several algorithms designed in the Barrier LogP are also given.
出处
《计算机研究与发展》
EI
CSCD
北大核心
1996年第6期445-449,共5页
Journal of Computer Research and Development
基金
国家自然科学基金
863计划的资助
关键词
并行计算机
并行计算模型
并行算法
设计
Massively parallel computer, parallel computation model, parallel algorithm design, bitonic sorting.