作者投稿和查稿 主编审稿 专家审稿 编委审稿 远程编辑

计算机工程 ›› 2012, Vol. 38 ›› Issue (24): 279-282. doi: 10.3969/j.issn.1000-3428.2012.24.066

• 开发研究与设计技术 • 上一篇    下一篇

CUDA程序到Cell平台的源代码移植

岳 峰,庞建民,张一弛,余 勇   

  1. (解放军信息工程大学信息工程学院,郑州 450002)
  • 收稿日期:2012-02-02 修回日期:2012-04-27 出版日期:2012-12-20 发布日期:2012-12-18
  • 作者简介:岳 峰(1985-),男,博士研究生,主研方向:动态编译,系统虚拟化;庞建民,教授、博士、博士生导师;张一驰,博士研究生;余 勇,硕士研究生
  • 基金资助:
    国家“863”计划基金资助项目(2009AA012201);“核高基”重大专项(2009ZX01036-001-001);河南省重大科技攻关计划基金资助项目(092101210501)

Source Code Migration of CUDA Program to Cell Platform

YUE Feng, PANG Jian-min, ZHANG Yi-chi, YU Yong   

  1. (Institute of Information Engineering, PLA Information Engineering University, Zhengzhou 450002, China)
  • Received:2012-02-02 Revised:2012-04-27 Online:2012-12-20 Published:2012-12-18

摘要: 相对于传统的串行程序移植,并行系统间的代码移植因体系结构间的巨大差异而变得极为复杂。为此,针对统一计算设备架构(CUDA)程序向其他异构多核平台的移植,提出CUDA架构到Cell的映射方案。通过模型映射、并行粒度提升、共享变量清除和运行时优化,使CUDA程序的大规模并行线程可以在Cell平台上正确执行。实验结果证明,翻译后的程序在Cell的执行效率可达到Cell平台上手动编写程序的72%。

关键词: 源代码移植, 异构多核, 模型映射, 共享变量清除, 运行时优化

Abstract: Compared with traditional serial program migration, parallel program migration becomes sophisticated for the huge diversity of different architectures. To migrate Compute Unified Device Architecture(CUDA) programs to other heterogeneous multi-cores, a method of mapping CUDA architecture to Cell is proposed. Through executing model mapping, enhancing parallel granularity, memory mapping and optimization, the mass threads in CUDA can execute correctly in Cell architecture by source code migration. Experimental result shows the executing speed of translated programs can achieve 72% of native compiled programs.

Key words: source code migration, heterogeneous multi-core, model mapping, shared variable removing, runtime optimization

中图分类号: