| 基于GPU的K-means并行算法研究与实现 |
| |
| 引用本文： | 原建伟,王坤,李爱国.基于GPU的K-means并行算法研究与实现[J].陕西理工学院学报(自然科学版),2012,28(5):44-48. |
| |
| 作者姓名： | 原建伟 王坤 李爱国 |
| |
| 作者单位： | 陕西工业职业技术学院 信息工程学院,陕西咸阳,712000 |
| |
| 摘 要： | 分析了K-means算法在GPU上实现并行计算的可能性,并在GTX8800 GT显卡上实现,研究了GPU的存储访问机制,在对数据进行合理组织基础上对算法进行改进,避免了存储体冲突的产生,提高了算法的健壮性.研究结果证明该方法在GPU上的并行运算速度明显快于CPU,加速比高.
|
| 关 键 词： | K均值算法 图形处理器 存储体冲突 CUDA |
Study and implementation of K-means based on GPU |
| |
| YUAN Jian-wei
,
WANG Kun
,
LI Ai-guo.Study and implementation of K-means based on GPU[J].Journal of Shananxi University of Technology:Natural Science Edition,2012,28(5):44-48. |
| |
| Authors: | YUAN Jian-wei WANG Kun LI Ai-guo |
| |
| Institution: | (College of Information Engineering,Shaanxi Polytechnic Institute, Xianyang 712000,China) |
| |
| Abstract: | The possibility of implementing parallel computing K-means algorithms was analyzed on GPU,and it was carried out on GTX 8800 GT graphic adapter.The algorithms was optimized after studying the principle of accessing the storage from GPU,then a method of avoiding bank conflict was proposed through the reasonable data designing and improving on algorithms.The result shows that this implementation can more effectively increase the K-means performance and have a good speedup. |
| |
| Keywords: | K-means GPU bank conflict CUDA |
| 本文献已被 CNKI 万方数据 等数据库收录！ |
|
