ConcSys: Reliable and Efficient Complex, Concurrent Software Systems

今天的许多和明天的大多数复杂的软件系统都是并发的。现代智能手机、笔记本电脑和台式电脑都有多核处理器，只能通过并发软件来利用。汽车软件系统建立在各种计算设备上，每个计算设备都具有多个核心，并且彼此同时交互。科学计算利用互连的计算机集群来实现大规模并行。所有这些系统的共同点是对并发软件的需求，其中执行的多个线程可以并行进行，并且偶尔彼此同步。并发软件的编程一直都很困难，传统上都是由少数专家掌握的。然而，并行的日益流行将并发编程提上了普通程序员的议事日程。ConcSys项目开发程序分析和软件系统，帮助普通程序员使复杂的并发系统比今天更可靠和高效。为了实现这一目标，该项目结合了可伸缩的静态分析、精确的动态分析和自动测试生成。这种组合是有益的，因为动态分析解决了可伸缩静态分析固有的不精确性，而自动测试生成提供了动态分析程序的驱动程序。为了支持我们的工作，我们开发了一个框架，用于以严格和可比较的方式评估我们的技术。ConcSys项目专注于适用于具有数百万行代码的大型真实世界系统的方法，因此，将有助于使未来的软件系统可靠和高效。

Many of today's and most of tomorrow's complex software systems are concurrent. Modern smart phones, laptops, and desktop computers have multi-core processors, which can be exploited only by concurrent software. Automotive software systems build upon various computing devices, which each have multiple cores and which interact with each other concurrently. Scientific computing leverages clusters of interconnected computers to achieve large-scale parallelism. Common to all these systems is the need for concurrent software, where multiple threads of execution can proceed in parallel and occasionally synchronize with each other. Programming concurrent software has always been difficult and traditionally has been mastered by a small set of experts. However, the increasing prevalence of parallelism is bringing concurrent programming on the agenda of ordinary programmers.The ConcSys project develops program analyses and software systems that help ordinary programmers to make complex, concurrent systems significantly more reliable and efficient than they are today. To achieve this goal, the project combines scalable static analysis, precise dynamic analysis, and automatic test generation. This combination is beneficial because dynamic analysis addresses the inherent imprecision of scalable static analysis, while automatic test generation provides a driver for dynamically analyzing a program. To underpin our work, we develop a framework for evaluating our techniques in a rigorous and comparable way. The ConcSys project focuses on approaches that are applicable to large, real-world systems with millions of lines of code and therefore, will contribute towards making tomorrow's software systems reliable and efficient.