An Evolutionary Approach for Data Locality Optimization

摘要：

Memory hierarchy is the most important parameter in the efficiency of the process. In current computers the efficiency of the process is limited by time which is affected by memory delays. Algorithms which affect the programs with nested loops containing a high amount of data make it impossible to store data in cache. Our approach, EADLL (Evolutionary Approach for Data Locality in Loops) algorithm is type of nested loop tiling method using an evolutionary algorithm introduced in this paper. EADLL algorithm increases data locality of the nested loops in code. The result shows that the amount of cache miss is reduced after implying EADLL algorithm.

关键词： Cache misses reuse vectors Ming

作者: Sahar Heron Shahriar Lotfi

作者单位: Computer Science Department Islamic Azad University, Shabestar Branch Shabestar, Iran Computer Science Department University of Tabriz Tabriz, Iran

会议类型: 国际会议

会议名称: 2010 Second Asia-Pacific Conference on Information Processing(2010年第二届亚太地区信息处理国际会议 APCIP 2010)

会议地点: 南昌

会议语种:英文

页码: 510-513

在线出版日期: 2010-09-17（万方平台首次上网日期，不代表论文的发表时间）