基于人工智能搜索和数据依赖分析的程序并行化

摘要

The front ⁃ end analysis technology in automatic parallelization are researched primarily. Aritificial intelligence search method was brought in and was modified. The extent and depth searching method supported by evaluation function was realized. The conventional data dependence analytical method was discussed on the basis of front⁃end analysis. For overcoming the insufficiency and deficiency of the traditional method,the array data flow analysis technology was created. Its flowchart algo⁃rithm is described and the ideas of terminally⁃written tree and write⁃write tree are exampled. These processes can improve effi⁃ciency of automatical parallelization well. The development of the parallelization in the future is also brought up.%　　主要研究了自动并行化中的前端分析技术，引入了人工智能搜索方法，并且对其进行改进，实现了在评估函数支持下进行广度和深度搜索的方法。在前端分析的基础上探讨了传统的数据依赖分析方法，针对其不足之处出现的数组数据流分析技术，对其流程图算法进行描述，并且对终写树和写写树的思想加以实例说明。这一流程经过测试后可以很好地提高程序并行化的效率。