Data-Flow Analysis of Large Software Systems

基于数据流分析的软件工具可以使许多软件工程任务自动化。然而，这些工具在实践中很少使用，因为现有的数据流分析器对于用包含复杂语言结构（如指针、数组和函数值变量）的语言编写的大型软件系统来说过于昂贵或过于不精确。本研究的主要目标是开发在实践中足够高效和精确的数据流分析，以分析处于开发、测试和维护中的程序。本研究主要涉及三个方面：为C程序开发高效的数据流分析；进行大规模实验，对算法进行评估，指导后续工作方向；并研究了这些技术在面向对象语言中的应用。拟议的研究将产生若干切实的结果：它将提供有关如何进行实际数据流分析的信息；它将构建原型并组装一组实验对象，用于评估研究结果和演示算法的实用性；它将使用有关数据流分析的发现作为开发面向对象程序的此类技术的起点。

Software tools based on data-flow analyses can automate many software engineering tasks. These tools, however, are rarely used in practice because existing data-flow analyzers are prohibitively expensive or too imprecise for use with large software systems written in languages that contain complex language constructs such as pointers, arrays, and function-valued variables. The primary goal of this research is to develop data-flow analyses that are sufficiently efficient and precise in practice to analyze programs when those programs are under development, testing, and maintenance. The research involves three primary efforts: development of efficient and effective data-flow analyses for C Programs; performance of large-scale experimentation to evaluate the algorithms and to guide the direction of the later work; and investigation of the application of the techniques to object-oriented languages. The proposed research will have several tangible results: it will provide information about how practical data-flow analyses can be created; it will construct prototypes and assemble a set of experimental subjects for use in evaluation of the results of the research and demonstration of the practicality of the algorithms; and it will use discoveries about data-flow analyses as a starting point for developing such techniques for object-oriented programs.