Journal of Computer Applications

• Typical applications • Previous Articles     Next Articles

Fast operation of large-scale high-precision matrix based on GPU

<a href="http://www.joca.cn/EN/article/advancedSearchResult.do?searchSQL=((([Author]) AND 1[Journal]) AND year[Order])" target="_blank"></a>C<a href="http://www.joca.cn/EN/article/advancedSearchResult.do?searchSQL=((([Author]) AND 1[Journal]) AND year[Order])" target="_blank"></a>h<a href="http://www.joca.cn/EN/article/advancedSearchResult.do?searchSQL=((([Author]) AND 1[Journal]) AND year[Order])" target="_blank"></a>a<a href="http://www.joca.cn/EN/article/advancedSearchResult.do?searchSQL=((([Author]) AND 1[Journal]) AND year[Order])" target="_blank"></a>n<a href="http://www.joca.cn/EN/article/advancedSearchResult.do?searchSQL=((([Author]) AND 1[Journal]) AND year[Order])" target="_blank"></a>g<a href="http://www.joca.cn/EN/article/advancedSearchResult.do?searchSQL=((([Author]) AND 1[Journal]) AND year[Order])" target="_blank"></a> <a href="http://www.joca.cn/EN/article/advancedSearchResult.do?searchSQL=((([Author]) AND 1[Journal]) AND year[Order])" target="_blank"></a>S<a href="http://www.joca.cn/EN/article/advancedSearchResult.do?searchSQL=((([Author]) AND 1[Journal]) AND year[Order])" target="_blank"></a>U<a href="http://www.joca.cn/EN/article/advancedSearchResult.do?searchSQL=((([Author]) AND 1[Journal]) AND year[Order])" target="_blank"></a> <a href="http://www.joca.cn/EN/article/advancedSearchResult.do?searchSQL=((([Author]) AND 1[Journal]) AND year[Order])" target="_blank"></a>Z<a href="http://www.joca.cn/EN/article/advancedSearchResult.do?searchSQL=((([Author]) AND 1[Journal]) AND year[Order])" target="_blank"></a>h<a href="http://www.joca.cn/EN/article/advancedSearchResult.do?searchSQL=((([Author]) AND 1[Journal]) AND year[Order])" target="_blank"></a>o<a href="http://www.joca.cn/EN/article/advancedSearchResult.do?searchSQL=((([Author]) AND 1[Journal]) AND year[Order])" target="_blank"></a>n<a href="http://www.joca.cn/EN/article/advancedSearchResult.do?searchSQL=((([Author]) AND 1[Journal]) AND year[Order])" target="_blank"></a>g<a href="http://www.joca.cn/EN/article/advancedSearchResult.do?searchSQL=((([Author]) AND 1[Journal]) AND year[Order])" target="_blank"></a>-<a href="http://www.joca.cn/EN/article/advancedSearchResult.do?searchSQL=((([Author]) AND 1[Journal]) AND year[Order])" target="_blank"></a>l<a href="http://www.joca.cn/EN/article/advancedSearchResult.do?searchSQL=((([Author]) AND 1[Journal]) AND year[Order])" target="_blank"></a>i<a href="http://www.joca.cn/EN/article/advancedSearchResult.do?searchSQL=((([Author]) AND 1[Journal]) AND year[Order])" target="_blank"></a>a<a href="http://www.joca.cn/EN/article/advancedSearchResult.do?searchSQL=((([Author]) AND 1[Journal]) AND year[Order])" target="_blank"></a>n<a href="http://www.joca.cn/EN/article/advancedSearchResult.do?searchSQL=((([Author]) AND 1[Journal]) AND year[Order])" target="_blank"></a>g<a href="http://www.joca.cn/EN/article/advancedSearchResult.do?searchSQL=((([Author]) AND 1[Journal]) AND year[Order])" target="_blank"></a> <a href="http://www.joca.cn/EN/article/advancedSearchResult.do?searchSQL=((([Author]) AND 1[Journal]) AND year[Order])" target="_blank"></a>F<a href="http://www.joca.cn/EN/article/advancedSearchResult.do?searchSQL=((([Author]) AND 1[Journal]) AND year[Order])" target="_blank"></a>U<a href="http://www.joca.cn/EN/article/advancedSearchResult.do?searchSQL=((([Author]) AND 1[Journal]) AND year[Order])" target="_blank"></a> <a href="http://www.joca.cn/EN/article/advancedSearchResult.do?searchSQL=((([Author]) AND 1[Journal]) AND year[Order])" target="_blank"></a>Y<a href="http://www.joca.cn/EN/article/advancedSearchResult.do?searchSQL=((([Author]) AND 1[Journal]) AND year[Order])" target="_blank"></a>u<a href="http://www.joca.cn/EN/article/advancedSearchResult.do?searchSQL=((([Author]) AND 1[Journal]) AND year[Order])" target="_blank"></a>-<a href="http://www.joca.cn/EN/article/advancedSearchResult.do?searchSQL=((([Author]) AND 1[Journal]) AND year[Order])" target="_blank"></a>c<a href="http://www.joca.cn/EN/article/advancedSearchResult.do?searchSQL=((([Author]) AND 1[Journal]) AND year[Order])" target="_blank"></a>h<a href="http://www.joca.cn/EN/article/advancedSearchResult.do?searchSQL=((([Author]) AND 1[Journal]) AND year[Order])" target="_blank"></a>e<a href="http://www.joca.cn/EN/article/advancedSearchResult.do?searchSQL=((([Author]) AND 1[Journal]) AND year[Order])" target="_blank"></a>n<a href="http://www.joca.cn/EN/article/advancedSearchResult.do?searchSQL=((([Author]) AND 1[Journal]) AND year[Order])" target="_blank"></a> <a href="http://www.joca.cn/EN/article/advancedSearchResult.do?searchSQL=((([Author]) AND 1[Journal]) AND year[Order])" target="_blank"></a>T<a href="http://www.joca.cn/EN/article/advancedSearchResult.do?searchSQL=((([Author]) AND 1[Journal]) AND year[Order])" target="_blank"></a>A<a href="http://www.joca.cn/EN/article/advancedSearchResult.do?searchSQL=((([Author]) AND 1[Journal]) AND year[Order])" target="_blank"></a>N<a href="http://www.joca.cn/EN/article/advancedSearchResult.do?searchSQL=((([Author]) AND 1[Journal]) AND year[Order])" target="_blank"></a>   

  • Received:2008-10-20 Revised:2008-11-27 Online:2009-04-01 Published:2009-04-01
  • Contact: Chang SU

一种在GPU上高精度大型矩阵快速运算的实现

苏畅 付忠良 谭雨辰   

  1. 中国科学院成都计算机应用研究所 中科院成都计算机应用研究所 中科院成都计算机应用研究所
  • 通讯作者: 苏畅

Abstract: A fast calculation approach for large-scale matrix operation, which can be accomplished by Graphic Processing Unit (GPU), was designed. For taking full advantage of the parallel architecture of GPU to enhance the calculation speed, special matrix partitioning and memory allocation mechanism according to the features of GPU were designed to decrease the frequency of data access. Meanwhile Kahan's summation formula was introduced to ensure the precision of the calculation. The result shows that the approach can achieve better effect and greatly enhance the speed and the precision of the large matrix multiplication.

Key words: Graphic Processing Unit (GPU), matrix operation, high-precision, parallel architecture

摘要: 设计了一种在图形处理器(GPU)上完成大型矩阵快速运算的方法,主要通过使用Kahan求和公式来确保计算精度,根据GPU特点设计矩阵分块方式和内存分配机制来减少对数据访问频次,以发挥GPU的并行体系结构特性来提高计算速度。实验结果表明此方法能够取得较好的效果,可大大提升大型矩阵乘法的运算速度和精度。

关键词: 图形处理器, 矩阵运算, 高精度, 并行体系结构

CLC Number: