计算机科学 ›› 2015, Vol. 42 ›› Issue (1): 50-53.doi: 10.11896/j.issn.1002-137X.2015.01.011

• 2013年全国理论计算机科学学术年会 • 上一篇    下一篇

渐进式智能回溯向量化代码调优方法

赵博,赵荣彩,徐金龙,高伟   

  1. 信息工程大学 郑州450002 数学工程与先进计算国家重点实验室 郑州450002,信息工程大学 郑州450002 数学工程与先进计算国家重点实验室 郑州450002,信息工程大学 郑州450002 数学工程与先进计算国家重点实验室 郑州450002,信息工程大学 郑州450002 数学工程与先进计算国家重点实验室 郑州450002
  • 出版日期:2018-11-14 发布日期:2018-11-14
  • 基金资助:
    本文受核高基国家科技重大专项(2009ZX01036)资助

Method of Progressive Intelligent Backtracking Vector Code Adjustment and Optimization

ZHAO Bo, ZHAO Rong-cai, XU Jin-long and GAO Wei   

  • Online:2018-11-14 Published:2018-11-14

摘要: 为了充分发挥高性能计算机的计算能力,缓解程序员设计和编写并行程序的压力,扩充可用软件集合,设计并实现了利用交互界面深入挖掘程序中的可向量化语句,优化生成代码中的向量化语句,提高生成代码的执行效率。该方法对充分发挥高性能计算机的计算能力,增强系统可用性和扩展应用范围具有重要的意义,同时能够提供有效的辅助手段和工具支持。渐进式智能回溯向量化代码调优架构通过对用户提交的串行程序进行程序分析和变换,采用串行程序分析、数据依赖分析、向量化分析等技术手段,根据分析结果对程序进行变换和优化,自动生成最终的向量化代码。该方法通过分析串行程序中潜在的并行性,将其自动变换为等价的向量化代码形式,大大简化了程序员的工作。

关键词: 渐进式,静态调优,动态调优,动静结合

Abstract: In order to fully develop the computing ability of high-performance computer and relieve the pressure of designing and writing parallel programs for programmers,with the expansion of available software sets we designed and realized the vector program through the interactive interface.Using this method,we could optimize the generated vector code to improve the efficiency for the implementation of generated code.Our process is of great significance to express the ability of high-performance computing,enhance the availability and extend range of application.Furthermore,it can provide available supplementary means and tool support.The method of progressive intelligent backtracking vector code adjustment and optimization can automatically generate the parallel code after analysis and transformation of the serial code given by the user using the following methods such as serial code analysis,data dependence analysis and parallelization analysis,etc.The work of this article can greatly reduce the work of programmers by transforming the serial code into parallel code automatically with the analysis of parallelization in the serial code.

Key words: Progressive,Static adjustment,Dynamic adjustment,Static combined dynamic

[1] Stewart J.An investigation of SIMD instruction sets.University of Ballarat School of Information Technology and Mathematical Sciences,2005..http://noisymime.org/blogimages/SI-MD.pdf
[2] Nuzman D,Rosen I,Zaks A.Auto-Vectorization of interleaved data for SIMD[C]∥Proc.of the ACM SIGPLAN Conf.on Programming Language Design and Implementation.Ottawa:ACM Press,2006:132-143
[3] 魏帅,赵荣彩,姚远.面向SLP的多重循环向量化[J].软件学报,2012(7):1717-1728
[4] 李玉祥,施慧,陈莉.面向非多媒体程序的SIMD向量化算法的研究及改进[J].小型微型计算机系统,2009(10):1927-1935
[5] 白书敬,李中升,漆锋滨.反馈式编译优化技术浅析[J].高性能计算技术,2005,10(5):1-5
[6] 郝云龙,赵荣彩,侯永生,等.反馈式编译在循环级性能分析中的应用[J].计算机工程,2011,5(5):32-34
[7] 姚远,赵荣彩.基于Profile信息的连续性分析算法及其优化[J].计算机工程,2012(9):28-31

No related articles found!
Viewed
Full text


Abstract

Cited

  Shared   
  Discussed   
No Suggested Reading articles found!