基于大规模数据处理的程序分析技术研究

Static program analysis is widely used in various application areas to solve many practical problems, such as software defects and vulnerability detection, program optimization, software testing and debugging, software maintenance. Due to the importance of static analysis, researchers have always been studying it for a long time. Although people have made significant achievements on static analysis, it is still too challenging to perform sophisticated static analysis on large-scale modern software. The underlying reason is that sophisticated static analysis for large-scale modern software results in highly intensive computation and huge memory consumption, leading to poor scalability. This project aims to tackle the above research problem. From the perspective of big data processing, we propose a novel data-parallel solution for scalable and precise static analysis. Specifically, the project consists of four parts, namely the design of data-parallel algorithm for highly precise static analysis, the design of highly scalable computation model, the implementation and optimization of distributed system, and the development of practical analysis applications. By taking advantage of advanced methodologies and techniques in the community of programming languages, software engineering and big data systems, we study, devise and develop a user-friendly, scalable distributed system for sophisticated static analysis. Consequently, we can succeed in performing various precise static analysis on large-scale modern software.

静态程序分析作为程序设计语言及软件工程领域的关键分析技术，可以被应用到多个领域解决实际中的众多问题。作为软件工程领域重点研究课题，静态分析虽然近年来在科学研究上取得了不小的进步，但是将其广泛应用到实践中，尤其是对现代大规模软件进行高精度静态分析还有很大的难度。本项目拟针对以上研究难点，从大数据处理的全新角度，探究高可扩展性、高精度程序分析的技术方法。具体来讲，本项目将围绕着研究高精度静态分析的大数据并行化算法，设计高可扩展性分布式计算模型，优化实现分布式分析系统，以及开发一系列程序分析应用实例来展开工作，结合程序设计语言、软件工程、大数据处理、以及分布式系统领域的相关前沿理论和技术手段，研究、设计、开发用户友好的、高可扩展性、高精度程序分析分布式系统，从而一定程度上满足对大规模软件系统进行高精度静态分析的实际需求。

静态程序分析作为程序设计语言及软件工程领域的关键分析技术，可以被应用到多个领域解决实际中的众多问题。作为软件工程领域重点研究课题，静态分析虽然近年来在科学研究上取得了不小的进步，但是将其广泛应用到实践中，尤其是对现代大规模软件进行高精度静态分析还有很大的难度。本项目针对高精度静态分析的低可扩展性问题开展研究，具体内容包括：研究高精度静态分析的大数据并行化算法，设计高可扩展性分布式计算模型，优化实现分布式分析系统，以及开发一系列程序分析应用实例。项目取得了众多成果，提出了静态程序分析问题的数据并行化算法，设计了高可扩展性的程序分析计算模型，完成了多个高可扩展性静态程序分析系统的原型系统开发。相关实验表明，我们的高可扩展性分析系统可以在分钟级时间内完成对千万行规模软件的高精度分析，填补了相关技术空白，一定程度上满足对大规模软件系统进行高精度静态分析的实际需求。