| 基于GPU的异构并行环境下程序优化策略研究 |
| |
| 引用本文： | 刘星,孙文君,曹雅楠,顾进广.基于GPU的异构并行环境下程序优化策略研究[J].培训与研究,2010(8):88-91. |
| |
| 作者姓名： | 刘星 孙文君 曹雅楠 顾进广 |
| |
| 作者单位： | [1]武汉科技大学计算机科学与技术学院,武汉430081 [2]清华大学工程物理系,北京100084 |
| |
| 基金项目： | 国家自然科学基金（60803160）; 湖北省自然科学基金重点项目（2009CDA136 2009CDA034）; 湖北省教育厅科学研究项目（Q20101110 D2009110） |
| |
| 摘 要： | 由CPU＋GPU组成的异构并行环境可以有效地提高计算系统的运算速度和处理能力。本文设计了几种有效的优化策略。首先依据计算情况和数据规模对并行计算单元进行合理的资源动态分配;然后通过共享存储器优化、指令优化及内存合并技术,对算法进一步优化,将其改造成为适用于异构环境处理的并行模式,以提高算法的运行速度。论文通过基于异构分布并行环境的蛋白质折叠结构预测算法进行实验,结果证明此优化方法是可行的、有效的,算法的运行速度有很大的提高,获得了较好的加速比。
|
| 关 键 词： | 异构并行环境 遗传退火算法 优化策略 |
Optimization Strategy of Programs Based on Heterogeneous Parallel Environment |
| |
| LIU-Xing,SUN Wen-jun,CAO Ya-nan,GU Jin-guang.Optimization Strategy of Programs Based on Heterogeneous Parallel Environment[J].Training and Research-Journal of Hubei College of Education,2010(8):88-91. |
| |
| Authors: | LIU-Xing SUN Wen-jun CAO Ya-nan GU Jin-guang |
| |
| Institution: | 1.College of Computer Science and Technology,Wuhan University of Science and Technology,Wuhan 430081,China;)2.Department of Engineering Physics,Tsinghua University,Beijing 100084,China) |
| |
| Abstract: | The heterogeneous parallel environment which formed by the CPU and GPU,can improve the execution speed and processing power of computing system.Several optimization approaches are designed to transform the program in the heterogeneous environment.Firstly,it dynamically allocates the resource of the system according to the size of the data and the status of the calculation.And then it realizes the algorithm on the heterogeneous environment,based on share memory optimization,instruction optimization and memory coalescing technology.Thus,the algorithm is transformed into a parallel processing mode which can enhance the execution efficiency.An experiment based on optimization of protein folding structure prediction algorithm shows the validity of the strategy.The high execution speed and better rate can be got by using this method,which has great meaning and value on the research of the folding structure of protein. |
| |
| Keywords: | heterogeneous parallel environment genetic annealing algorithm optimization strategy |
| 本文献已被 维普 万方数据 等数据库收录！ |
|
