Research On Hybrid Algorithm Of Data Race Detection

With the popularity of multi-core architecture,the parallel ability and multi-threading technology of the system have been enhanced.The parallel programming development model of the system program has been continuously developed,and the multi-threading technology has gradually become an essential programming technology,which can improve the performance of the concurrent program of the operating system.Improvements in testing and the increase in affordable concurrency have driven the concurrency direction,although more comprehensive methods of analyzing increasingly complex programs have become the norm,many of these methods still have many flaws in detection.How to improve the false detection rate and missed detection rate of data race,as well as improve memory performance and reduce operating overhead are urgent problems to be solved.In this paper,the false detection rate and missed detection rate of existing test data race methods are improved,and the dynamic and static mixed data race detection DSrace algorithm(Dynamic and Static algorithm)is proposed.The static detection technology uses RELAY detection method,which has good accuracy and extensibility.The missed detection rate of data race detection is low,but the false detection rate is high.Combined with the dynamic data race detection technology,the false detection rate can be reduced.Dynamic data race detection uses an improved Lockset method to detect whether data race occured by thread access has performed read and write operations at the same memory,and at least one of operation is a write.After data race detection,the repeated filtering detection is used to analyze repeated data race to reduce memory redundancy and improve detection performance.In addition,the MVM algorithm(Message Verification Matching algorithm),a data race detection method in this paper,is proposed based on the conditions of data race.The MVM algorithm uses random numbers to indicate the type of access operation,then obtains access analysis information through alias analysis,calculates the value of the HASH function to detect data race,and last uses program slicing analysis to improve the accuracy rate of data race.Finally,through the realization of dynamic and static data race algorithm,the algorithm in this paper conducts experimental analysis on the detection accuracy,operating efficiency and scalability of data race.The experimental results show DSrace and MVM algorithm detect data races with higher accuracy than static and dynamic detectors,and the proposed algorithm has a better effect in reducing the false detection rate and the missed detection rate of data race,and does not have a great impact on the performance of the original program.