CAREER: Hybrid Atomicity Checking

职业：混合原子性检查

• 批准号: 0644130
• 负责人: Stephen Freund
• 金额: $ 40万
• 依托单位: Williams College
• 依托单位国家: 美国
• 项目类别: Continuing Grant
• 财政年份: 2007
• 资助国家: 美国
• 起止时间: 2007-04-15 至 2013-03-31
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=0644130&HistoricalAwards=false
• 关键词: CAREER; Hybrid; Atomicity; Checking

The construction and validation of reliable multi-threaded programs is extremely difficult. Threads can improve performance on multi-processor machines and multi-core processors by enabling a program to execute more than one routine simultaneously, but unintended interactions between threads are hard to recognize during testing and are a common source of errors in deployed systems.This research develops hybrid checkers that prevent unintended thread interactions by ensuring that a program's routines are atomic. A routine is atomic if its execution is not affected by and does not interfere with concurrently-executing threads. Previous work on static atomicity checkers (that inspect source code) and dynamic atomicity checkers (that monitor running programs) demonstrate the potential advantages of enforcing atomicity requirements. However, these approaches have precision or coverage limitations that reduce their ability to check large systems effectively. Hybrid checkers synthesize the best aspects of both techniques without suffering from these limitations.The impacts of hybrid atomicity checkers, and their integration into a broad educational program, include improved software quality and better software engineering practices. Specifically, hybrid checkers provide a cost-effective mechanism for finding errors resistant to testing, are more usable and scalable than existing tools, and support a design methodology that encourages precisely specifying interactions between threads.

构建和验证可靠的多线程程序是极其困难的。线程可以通过使程序同时执行多个例程来提高多处理器机器和多核处理器上的性能，但是在测试期间很难识别线程之间的意外交互，并且是已部署系统中常见的错误来源。本研究开发了混合检查器，通过确保程序的例程是原子的来防止意外的线程交互。如果例程的执行不受并发执行线程的影响，也不与并发执行线程发生干扰，那么例程就是原子性的。以前关于静态原子性检查器（检查源代码）和动态原子性检查器（监视正在运行的程序）的工作证明了强制原子性需求的潜在优势。然而，这些方法具有精度或覆盖范围的限制，这降低了它们有效检查大型系统的能力。混合检查器综合了这两种技术的最佳方面，而没有这些限制。混合原子性检查器的影响，以及它们与广泛的教育计划的集成，包括改进的软件质量和更好的软件工程实践。具体来说，混合检查器提供了一种经济有效的机制来查找不受测试影响的错误，比现有工具更可用和可扩展，并且支持一种鼓励精确指定线程之间交互的设计方法。