Research on Parallel Algorithm of PageRank based on GPU

被引：0

作者：

Jiang, Hao ^{[1
]}

Gao, Jian-Ming ^{[2
]}

机构：

[1] Southeast Univ, Sch Comp Sci & Engn, Nanjing 210096, Jiangsu, Peoples R China

[2] Southeast Univ, Sch Comp Sci & Engn, Nanjing 210096, Jiangsu, Peoples R China

来源：

2016 INTERNATIONAL CONFERENCE ON COMPUTER SCIENCE AND INFORMATION SECURITY (CSIS 2016) | 2016年

关键词：

PageRank; CUDA; Parallel Algorithm; GPU;

D O I：

暂无

中图分类号：

TP18 [人工智能理论];

学科分类号：

081104 ; 0812 ; 0835 ; 1405 ;

摘要：

As the key technology of Google search engine, PageRank algorithm is the most classical mining algorithm of Web structure. In order to improve its computational efficiency, this paper mainly analyzes the PageRank algorithm and proposes a matrix-block parallel algorithm based on the CUDA platform. Through the powerful parallel computing power of GPU, we using a lot of GPU threads in parallel processing to speed up the convergence of the algorithm, so as to enhance the computational efficiency of the PageRank algorithm. Finally, we implement the PageRank parallel algorithm based on CUDA, and the experimental results show that the maximum speed of the algorithm faster than the traditional algorithm on CPU is up to 37.5 times.

引用

页码：391 / 397

页数：7