Dependence-aware, unbounded sound predictive race detection

What is it about?

Data races are a class of bugs where two or more accesses to shared memory occur without proper synchronization. Data races are an important source of issues for shared memory concurrent programs. Data race detection methods can detect and report data races, making it easy to find and solve data races. Predictive data race detection analyzes a single execution of a program, and predicts data races that may occur in many other executions. We designed two new predictive data race detection methods, intended to be used at testing time, that can detect more races than prior works. Our methods provide soundness or completeness, finding no false positives or no false negatives.

Featured Image