Hybrid Dynamic Data Race Detection in SystemC

Data races are one of the most common problems in concurrent programs. As SystemC standard allows non-deterministic scheduling of processes, this leads to data races. Hence, different executions of the same concurrent program may lead to unexpected results due to race conditions. We develop a hybrid dynamic data race detection algorithm for SystemC/TLM designs that adopts the well-studied dynamic race detection algorithms; lockset and happens-before. Experiments show that our solution has fewer false positives than lockset and fewer false negatives than happens-before algorithms. Our implementation uses dynamic binary instrumentation allowing us to work on designs for which source codes may not be available such as pre-compiled IPs.