北京航空航天大学学报 ›› 2004, Vol. 30 ›› Issue (11): 1111-1115.

• 论文 • 上一篇    下一篇

软件流水中的循环展开优化

李文龙, 刘利, 汤志忠   

  1. 清华大学 计算机科学与技术系, 北京 100084
  • 收稿日期:2004-06-25 出版日期:2004-11-30 发布日期:2010-09-24
  • 作者简介:李文龙 (1977-),男,辽宁鞍山人,博士生, liwenlong@tsinghua.org.cn.
  • 基金资助:

    国家自然科学基金资助项目(60173010)

Loop unrolling optimization for software pipelining

Li Wenlong, Liu Li, Tang Zhizhong   

  1. Dept. of Computer Science and Technology, Tsinghua Universi ty, Beijing 100084, China
  • Received:2004-06-25 Online:2004-11-30 Published:2010-09-24

摘要: 在软件流水中应用循环展开可以实现分数值的启动间距,提高资源的利用率,同时基于展开的优化技术可以降低程序的资源需求和关键路径的长度.提出了基于程序特性的展开因子算法UTBPC(Unrolling Times Based Program Characteristics),解决了循环展开的核心问题——展开因子的确定,同时提出了基于展开的软件数据预取优化技术,提高了软件数据预取的效率.所有这些都在ORC(Open Research Compiler)中实现,并对SPEC?2000中的程序进行了测试,平均性能提高了2.6%.实验结果表明新提出的算法和基于展开的数据预取优化技术提高了编译器的整体性能.

Abstract: Loop unrolling can m ake software pipelining achieve fractional initiation interval and improve resou rce utilization. Optimizations based on unrolling can reduce resource requiremen ts and the heights of critical paths. An algorithm named UTBPC(unrolling times based program characteristics) for determining unrolling factors and unrol ling b ased optimization for software data prefetching were proposed. These optimizatio ns were implemented in ORC(open research compiler), and SPEC CPU2000 benchmarks were tested in Itanium processor. The average perf ormance was improv ed by 2.6%. The results show that UTBPC algorithm and unrolling base d optimization for software data prefetching can improve the overall performance of compilers.

中图分类号: 


版权所有 © 《北京航空航天大学学报》编辑部
通讯地址:北京市海淀区学院路37号 北京航空航天大学学报编辑部 邮编:100191 E-mail:jbuaa@buaa.edu.cn
本系统由北京玛格泰克科技发展有限公司设计开发