Effective Monitoring Memory Operations for Dynamic Race Detection through Hierarchical Filtering Method

Abstract

Data races are the hardest defect to handle in multithreaded programs due to the non-deterministic interleaving of concurrent threads. It incurs the expensive costs of dynamic data race detection to monitor all of memory operations to shared memory locations. This paper presents a hierarchical filtering method that removes unnecessary monitoring memory operations from three levels of binary image, and evaluates empirically the effectiveness of the filtering method for dynamic data race detection. The empirical results using a set of benchmarks show that our filtering method reduces the average runtime overhead to over 50% of dynamic data race detection.