| 摘要: |
| 提出了一种具有数据流特征的Java并行程序设计模型,并针对该模型提出了一种基于运行时信息反馈的自适应优化算法,使得运行时系统可以利用数据流程序所暴露出的数据并行性,加速程序的运行.此外,在该模型中加入了数据流多态的概念,扩展了该模型的面向对象特性.在一个实际的开放源码Java虚拟机中实现了上述程序设计模型及优化方法.在实际多核多线程机器上的实验结果表明,所提出的程序设计模型及优化能够充分利用硬件的并行处理能力,显著地提高了程序的性能. |
| 关键词: 数据流 并行程序设计模型 可管理运行时环境 运行时优化 |
| DOI: |
| 分类号: |
| 基金项目:Supported by the National Basic Research Program of China under Grant No.2005CB321602 (国家重点基础研究发展计划(973)) |
|
| Dataflow-Style Java Parallel Programming Model and Runtime Optimization |
|
LIU Tao,FAN Bin,WU Cheng-Yong,ZHANG Zhao-Qing
|
| Abstract: |
| This paper presents a dataflow-style Java parallel programming model with a runtime profile based thread duplication algorithm to exploit data level parallelism. Furthermore, a new dataflow polymorphism feature is introduced. This model has been implemented in an open source Java virtual machine. Evaluations on real machine show good speedup for benchmark applications. |
| Key words: dataflow parallel programming model managed runtime environment runtime optimization |