计算机科学 ›› 2012, Vol. 39 ›› Issue (10): 157-159.

• 数据库与数据挖掘 • 上一篇    下一篇

异构平台上多维线性哈希的研究

刘 勇,赵秦德,赖正文,黄东平,王憬星   

  1. (广西民族大学信息科学与工程学院 南宁530006); (广西交通科学研究院南宁530002);(华南理工大学软件学院 广州510006)
  • 出版日期:2018-11-16 发布日期:2018-11-16

Research for Multidimensional Linear Hashing on Heterogeneous Platforms

  • Online:2018-11-16 Published:2018-11-16

摘要: 目前多维数据广泛应用于多个领域,但其复杂性影响了多维数据的操作效率。为提高对多维数据的处理能力,提出一种在CPU/GPU异构平台上的多维线性哈希并行计算方案。该方案通过对传统线性哈希表数据结构的扩展,可实现对哈希表的快速创建和查询。同时,在多个处理器平台上进行的实验对提出的方案的有效性进行了验证。实验结果表明,当处理的数据规模较大时,提出的方案由于充分利用了GPU强大的并行处理能力,在创建哈希表和查询数据上,比传统的CPU方案性能分别提高了约25倍和38倍,充分显示出提出的方案在处理多维数据时的优势。

关键词: 异构平台,图形处理器,多维线性哈希,计算统一设备架构

Abstract: Nowadays,multidimensional data is used in plenty of areas, but for the complex of multidimensional data, the efficiency of computing is not perfect, In order to speed up the query and manipulation of multidimensional data,a multi-dimensional linear hashing parallel computing solution on CPU/GPU heterogeneous platforms was proposed. Based on the extension of traditional hashing table data structure, it is able to achieve quick creating and query. Experiments on different platforms were done. The experimental results show that the proposed solution is about 25 times and 38 times faster than traditional solution on hashing table creating and data qucry,and it reflects the advantage of the proposed solution.

Key words: Heterogeneous platforms, Graphical processing unit, Multidimensional linear hashing, Compute unified device architcturc

No related articles found!
Viewed
Full text


Abstract

Cited

  Shared   
  Discussed   
No Suggested Reading articles found!