BAVC: Classifying Benign Atomicity Violations via Machine Learning

Abstract

The reality of multi-core hardware has made concurrent programs pervasive. Unfortunately, writing correct concurrent programs is difficult. Atomicity violation, which is caused by concurrent executions unexpectedly violating the atomicity of a certain code region, is one of the most common concurrency errors. However, atomicity violation bugs are hard to find using traditional testing and debugging techniques. In this paper, we investigate an approach based on machine learning techniques (specifically decision tree and support vector machine (SVM)) for classifying the benign atomicity violations from the harmful ones. A benign atomicity violation is known not to affect the program's correctness even it happens. We formulate our problem as a supervised-learning problem and apply these two machine learning techniques to classify the atomicity violation report. Our experimental evaluation shows that the proposed method is effective in identifying the benign atomicity violation warnings.