| GPU编程模型中存储体冲突的研究 |
| |
| 引用本文： | 原建伟,李爱国,李文宇. GPU编程模型中存储体冲突的研究[J]. 河北工业科技, 2013, 30(1): 39-41,46 |
| |
| 作者姓名： | 原建伟 李爱国 李文宇 |
| |
| 作者单位： | 陕西工业职业技术学院信息工程学院 |
| |
| 摘 要： | GPU并行计算模型中使用共享内存是提高并行计算效率的重要途径,由于进程访问导致的存储体冲突,却会成倍地降低执行效率。经过分析存储体冲突产生的原因,提出了对算法的修改来解决存储体冲突的原则与方法,并通过对聚类算法的实施进行验证。
|
| 关 键 词： | 图形处理器 共享内存 并发存储访问 存储体冲突 矩阵运算 |
Study of bank conflict in GPU programming model |
| |
| YUAN Jianwei,LI Aiguo,LI Wenyu. Study of bank conflict in GPU programming model[J]. Hebei Journal of Industrial Science & Technology, 2013, 30(1): 39-41,46 |
| |
| Authors: | YUAN Jianwei LI Aiguo LI Wenyu |
| |
| Affiliation: | (College of Information Engineering,Shaanxi Polytechnic Institute,Xianyang Shaanxi 712000,China) |
| |
| Abstract: | It's an important way to improve the efficiency of parallel computing using shared memory in GPU parallel computing model.But it will exponentially reduce the computing efficiency because of the bank conflict caused by the processes accessing the memory.After analysis of the causes of bank conflict,a principle and solution through modifying algorithm was proposed.And it is verified through the implementation of K-means algorithm. |
| |
| Keywords: | GPU shared memory parallel storage access bank conflict matrix operation |
| 本文献已被 CNKI 等数据库收录！ |
| 点击此处可从《河北工业科技》浏览原始摘要信息 |
|
点击此处可从《河北工业科技》下载全文 |
|
