On-the-fly detection of access anomalies

摘要: Access anomalies are a common class of bugs in shared-memory parallel programs. An access anomaly occurs when two concurrent execution threads both write (or one thread reads and the other writes) same shared memory location without coordination. Approaches to detection include static analysis, post-mortem trace on-the-fly monitoring.A general algorithm for is presented, which can be applied programs with nested fork-join synchronization operations. The advantage over analysis that amount storage used greatly reduced by data compression techniques discarding information as soon it becomes obsolete. In required at any time depends only on number V variables being monitored N threads, not synchronizations. Data achieved use called merging subtraction. Upper bounds shown × N2