Citation Context

... either dynamic or static. Dynamic race detectors are based on either the happens-before relation [1, 13-15, 35, 43, 46],the lockset algorithm [2, 10, 11, 38, 45, 49, 50], or a combination of the two =-=[16, 28, 39, 40, 54]-=-. Static race detectors are eitherprimarily flow insensitive type-based systems [7, 8, 19, 20, 41, 44], flow sensitive static versions of the lockset algorithm [12, 17, 47],or path sensitive model che...

Citation Context

...c race detection, happens-before [7, 26, 28] and lockset algorithms [6, 34, 38] have been thoroughly studied. Many extensions and combinations have been proposed to reduce false positives or overhead =-=[27, 29, 43]-=-. Static data race detection techniques include race type-safe systems [3, 10], static-versions of the lockset algorithm [8] and model checking techniques [17]. As discussed in Section 1, all data rac...

Citation Context

...78-1-60558-392-1/09/06. . . $5.00 making them extremely difficult to detect, reproduce, and eliminate. Consequently, much prior work has focused on static [1, 5, 3, 19, 4, 12, 15, 26, 40] and dynamic =-=[33, 38, 27, 42, 30, 11, 34, 42, 30]-=- analysis tools for detecting race conditions. In general, dynamic race detectors fall into two categories, depending on whether they report false alarms. Precise race detectors never produce false al...

Citation Context

... described in Section 3.4, is sufficient to order all accesses to shared memory. Unfortunately, our experience suggests that most concurrent programs contain data-races, many of which are intentional =-=[46]-=-. In this case, the happens-before graph should include additional edges that determine the order of racy accesses. Neglecting these edges may result in inability to replay a given execution. One poss...

Citation Context

...happensbefore race detection has been successfully extended to classify harmful races from benign races [34], but they suffer from the same limitations as happens-before techniques. Hybrid techniques =-=[14, 37, 38, 54]-=- combine lockset with happens-before to make dynamic race detection both precise and predictive. Despite the combination, hybrid techniques could report many false warnings. One characteristics that d...

Citation Context

...ing input generation techniques to improve the effectiveness of concurrent program testing. Concurrency bug detection Much research has been conducted on detecting different types of concurrency bugs =-=[19, 20, 29, 5, 13, 27, 7]-=-. In general, exposing software bugs is complementary to bug detection, as most dynamic bug detectors require bugs to manifest during the bug detection runs. Concurrency testing tools like CHESS, Race...

Citation Context

...es on-line, either with instrumentation support [33] and with hardware support [1, 29]. There have also been attempts to ameliorate the performance cost of dynamic analysis using static optimizations =-=[11, 38, 23, 25]-=-. Alternatively, if we can efficiently record sufficient information about a program’s execution to allow us to deterministically replay the execution, then we can do off-line analysis. The advantage ...

Citation Context

... the program’s documentation. 7. Related Work Autolocker is complementary to dynamic race detection techniques, such as those based on locksets [26, 32], happens-before relationships [4, 33], or both =-=[34]-=-. These tools detect the absence of locks and thus could be used to infer missing lock annotations. However, their success depends on the specific execution and may have false negatives. Flanagan and ...

Citation Context

... variables such as barrier synchronization. The other approach to dynamic data-race detection is based on computing the happens-before relation [7, 19, 21] using vector clocks [16]. Hybrid techniques =-=[18, 25]-=- combine lockset and happens-before analysis. For example, RaceTrack [25] uses a basic 3 This overhead is not representative as Multiset executions consist entirely of shared variable accesses each in...

Citation Context

...add(x); } } class Vector { synchronized void add(Object o) { ... } synchronized boolean contains(Object o) { ... } } For race conditions, a variety of dynamic race detection tools have been developed =-=[36, 33, 47]-=-, including complete race detectors that never produce false alarms [34, 37, 7]. For atomicity, a variety of dynamic analyses have also been developed (e.g. [11, 44, 1, 43, 46]), but these tools are a...