Block HB (Happens-Before) based dynamic data race detection method and system

Abstract

The invention discloses a block HB (Happens-Before) based dynamic data race detection method and system. The method comprises the steps of carrying out compilation on a source representation by employing a front end of an LLVM (Low Level Virtual Machine) to obtain an IR (Intermediate Representation), and carrying out instrumentation on a memory access operation and a synchronization operation by employing a process PASS of the LLVM; in an operating process of the IR, obtaining physical time information of the memory access operation and the synchronization operation after the instrumentation, and then compressing and simplifying HB relationships; and carrying out data race detection based on the compressed HB relationships. By compressing and simplifying the HB relationships of an event, the overheads of a time and a memory caused by a large number of the HB relationships are reduced.

Description

technical field [0001] The invention belongs to the technical field of parallel computing data competition detection, and more specifically relates to a block partial order-based dynamic data competition detection method and system. Background technique [0002] With the rapid development of social networks, computer biology, scientific computing and security fields, the amount of data targeting these information is increasing day by day, which brings new challenges to some traditional algorithms. [0003] The increase in the amount of data makes people pay more and more attention to the time and memory overhead of the algorithm, which in turn has a certain impact on the field of data race detection. The current data race detection tools include static data race detection and dynamic data race detection, and the traditional static data race detection is mainly based on the lockset detection method, which mainly has the following disadvantages: (1) There are many false positi...

AI Technical Summary

Problems solved by technology

[0004] In view of the defects or improvement needs of the existing static data competition detection technology and dynamic data competition detection technology, the present invention provides a dynamic data competition detection method and system based on block partial order, the purpose of which is to ensure the accuracy of competition detection In this way, memory and time overheads are reduced, thereby solving the problems of large memory overhead, time overhead and low detection accuracy in the existing static data race detection technology and dynamic data race detection technology in terms of memory overhead, time overhead, and detection accuracy technical issues

Images