基于锁集合算法的增强型数据竞争检测方法

Enhanced data race detection approach based on lockset algorithm

针对锁集合算法误报率高的缺点,采用增强型数据竞争检测方法对锁集合算法及算法实现两方面进行改进.增强型锁集合算法提出了更强必要条件的检测算法.该算法通过分析线程结束情况和理解变量的线程信息,精化数据竞争检测条件.算法实现采用系统体系结构扩展的方法,通过扩展锁表部件,将集合操作精确映射到位图操作,进一步提高检测精度和减少误报.实现了无干扰免插桩检测技术,避免了二进制代码插桩带来的干扰性和不确定性问题.与同类系统的对比实验结果表明,该方法能够有效地将误报率减小到0.67,并且具有很高的检测效率.

Because too many false-positive faults were produced in the traditional lockset algorithm, an enhanced data race detection approach based on lockset algorithm was brought forward to overcome the drawback from two ways, e.g. lockset-based algorithm definition and implementation. A more enhanced prerequisite algorithm was presented. The enhanced lockset algorithm refines the data race condition by catching the end of the thread and analyzing the right thread information of various variables. On the other hand, based on the architecture extension technology, a new hardware component, called locks table, is extended to accurately map set operation to bit operation. In this way, the detection precision can be improved. Meanwhile, a non-instrumentation detection method was implemented to avoid the probe-effect and indetermination. Compared with similar approaches, the enhanced lockset algorithm can effectively reduce the false-positive rate to be 0.67 and achieve good detection efficiency.