Reasoning about Data Structures, Concurrency, and Resources

摘要[0541021]本文研究了计算机程序的规范和验证，以及确保验证的合理性所需的语义。特别感兴趣的是：分离逻辑，它处理使用共享可变数据结构或共享变量并发的程序。目标是将逻辑扩展到使用安全类型系统和自动存储回收的高级语言，以及允许在数据结构中嵌入代码指针的机器级语言。无粒度语义，通过将竞争条件（即，通过并发进程同时访问相同的存储）视为灾难性事件来处理共享变量并发性，而不强加任何默认级别的原子操作。其目标是通过避免在具有不可接受行为的程序之间进行无用的区分来简化对程序的理解。本研究的智力价值在于，它将大大增加分离逻辑的话语领域，并为共享变量同步的这一逻辑和其他逻辑提供可靠的论证。更广泛的影响是，在一类重要的有用但困难的计算机程序中，避免错误将变得更加容易。最终，应该有可能在逻辑中自动进行证明检查，这样该类中的程序就可以伴随机器检查其正确性的证明。

Abstract0541021John C. ReynoldsCarnegie Mellon UniversityReasoning about Shared Structure and ConcurrencyThe specification and verification of computer programs is investigated, along with the semantics needed to insure the soundness of verification. Of specific interest are:Separation Logic, which treats programs employing shared mutable data structures or shared-variable concurrency. The goal is to extend the logic to high-level languages using safe type systems and automaticstorage reclamation, and also to machine-level languages permitting pointers to code to be embedded within data structures.Grainless Semantics, which treats shared-variable concurrency without imposing any default level of atomic operations, by regarding race conditions (i.e., simultaneous access to the same storage by concurrentprocesses) as catastrophic events. The goal is to simplify the understanding of programs by avoiding useless distinctions between programs with unacceptable behavior.The intellectual merit of this research is that it will substantially increase the domain of discourse of separation logic, and facilitate soundness arguments for this and other logics for shared-variableconcurrency.The broader impact is that it will become easier to avoid errors in an important class of useful but difficult computer programs. Eventually, it should be possible to automate proof-checking in the logic so thatprograms in this class can be accompanied by machine-checkable proofs of their correctness.