摘要
在并行程序开发过程中,对并行程序的有效优化能够充分发挥软硬件的执行效率。在进一步探讨Amdahl定律的基础上,针对数据密集型问题的单程序多数据(SPMD)并行计算,分析并行程序被有效优化后其计算时间及并行效率的变化规律,并给出了公式证明,有利于充分利用Amdahl定律指导并行程序的优化。实验结果表明了论证的有效性。
In developing parallel programs, effective parallel program optimization can give full play to the hardware and software execution efficiency. In this paper the Amdahl＇s law was re-observed. Based on this, the optimized changing trends of SPMD parallel codes＇ computing time and efficiency for data-intensive problems were analyzed, which were demonstrated by equations and could make the most use of Amdahl＇s law to guide parallel program optimization. The experimental results indicate the effectiveness of the discussions.
出处
《计算机应用》
CSCD
北大核心
2014年第A01期103-106,共4页
journal of Computer Applications
基金
上海高校青年教师培养计划项目
上海科学技术委员会创新计划项目(11511500200)
关键词
数据密集型
单程序多数据
优化
并行计算时间
并行效率
data-intensive
Single Program Multiple Data （SPMD）
optimization
parallel computing time
parallele/tqciency
