Lazy sequentialization has emerged as one of the most promising approaches for concurrent program analysis but the only efficient implementation given so far works just for bounded programs. This restricts the approach to bug-finding purposes. In this paper, we describe and evaluate a new lazy sequentialization translation that does not unwind loops and thus allows to analyze unbounded computations, even with an unbounded number of context switches. In connection with an appropriate sequential backend verification tool it can thus also be used for the safety verification of concurrent programs, rather than just for bug-finding. The main technical novelty of our translation is the simulation of the thread resumption in a way that does not use gotos and thus does not require that each statement is executed at most once. We have implemented this translation in the UL-CSeq tool for C99 programs that use the pthreads API.We evaluate UL-CSeq on several benchmarks, using different sequential verification backends on the sequentialized program, and show that it is more effective than previous approaches in proving the correctness of the safe benchmarks, and still remains competitive with state-of-the-art approaches for finding bugs in the unsafe benchmarks.
|Titolo:||Lazy sequentialization for the safety verification of unbounded concurrent programs|
|Data di pubblicazione:||2016|
|Appare nelle tipologie:||4.1.1 Proceedings con DOI|