一种以GPU异构计算为基础在簇结构中加速R树索引的方法
首发时间:2017-07-12
摘要:随着社会的信息化,外部存储器的容量得到了飞速的提升,然而其读写性能却发展缓慢。为此人们提出了多种多样的树形索引结构,在有限的读写速度下提高数据的响应速度,但数据的离散性仍阻碍其性能的进一步提升。与此同时GPU作为一种从设备其性能的发展为人们展示了计算机发展的另一个方向:异构计算,利用异构计算我们提出了一种可以快速对统一的数据结构进行处理的并行R树结构,并通过实验证明其在处理较大规模数据时性能优于传统R树。
关键词: 计算机软件与理论 SIMD R-tree 异构计算 簇
For information in English, please click here
An R-tree Index Acceleration Algorithms in Cluster Consturcture Based on GPU
Abstract:With the informatization development of the society, the capacity of the external memory is growing rapidly but performance is growing slower. For this reason, a variety of tree index structures have been proposed to improve the response speed of data at a limited reading and writing speed. However the discreteness of data is still hindering the further improvement of performance. Meanwhile GPU performance developing shows another direction of computer development: heterogeneous computing. With heterogeneous computing we propose a kind of parallel R-tree to handle uniform data structure and prove that its performance is better than triditional R-tree by experiment.
Keywords: computer software and theory SIMD R-tree heterogeneous computing cluster
基金:
论文图表:
引用
No.4738295119356414****
同行评议
共计0人参与
勘误表
一种以GPU异构计算为基础在簇结构中加速R树索引的方法
评论
全部评论0/1000