...
  • 主题
高级搜索  >
历史搜索 清空历史
首页> 外文期刊>IEEE Transactions on Software Engineering >Efficient interprocedural array data-flow analysis for automaticprogram parallelization
【24h】

Efficient interprocedural array data-flow analysis for automaticprogram parallelization

机译：高效的过程间数组数据流分析，可实现程序自动并行化

获取原文
获取原文并翻译 | 示例
           
页面导航
  • 摘要
  • 著录项
  • 相似文献
  • 相关主题

摘要

Since sequential languages such as Fortran and C are morenmachine-independent than current parallel languages, it is highlyndesirable to develop powerful parallelization tools which can generatenparallel codes, automatically or semi-automatically, targeting differentnparallel architectures. Array data-flow analysis is known to be crucialnto the success of automatic parallelization. Such an analysis should benperformed interprocedurally and symbolically and it often needs tonhandle the predicates represented by IF conditions. Unfortunately, suchna powerful program analysis can be extremely time-consuming if it is notncarefully designed. How to enhance the efficiency of this analysis to anpractical level remains an issue largely untouched to date. This papernpresents techniques for efficient interprocedural array data-flownanalysis and documents experimental results of its implementation in anresearch parallelizing compiler. Our techniques are based on guardednarray regions and the resulting tool runs faster, by one or two ordersnof magnitude, than other similarly powerful tools
机译：由于诸如Fortran和C之类的顺序语言比当前的并行语言在机器上更独立，因此非常需要开发强大的并行化工具，该工具可以自动或半自动地针对不同的并行体系结构生成并行代码。众所周知，阵列数据流分析对于自动并行化的成功至关重要。这样的分析应该在过程上和象征性地进行，并且它经常需要淡化IF条件所代表的谓词。不幸的是，如果设计不当，那么功能强大的程序分析可能会非常耗时。迄今为止，如何将这种分析的效率提高到实际水平仍然是一个尚未解决的问题。本文介绍了有效的过程间数组数据流分析技术，并记录了在研究并行化编译器中实现该过程的实验结果。我们的技术基于guardednarray区域，因此生成的工具比其他类似功能强大的工具运行速度快一到两个数量级

著录项

相似文献

  • 外文文献
  • 中文文献
  • 专利
获取原文

客服邮箱：kefu@zhangqiaokeyan.com

京公网安备：11010802029741号 ICP备案号：京ICP备15016152号-6 六维联合信息科技 (北京) 有限公司©版权所有

  • 客服微信

  • 服务号