MHP analysis method and device for parallel programs

Abstract

The embodiment of the invention provides an MHP analysis method and device for a parallel program. The MHP analysis method for the parallel program comprises the following steps: starting from a main thread of the parallel program to be analyzed, and selecting one thread at a time to simulate the execution of the parallel program to be analyzed, wherein, when the thread for simulating the execution cannot continue to execute, jumping to any one thread which can continue to execute to simulate execution until the simulation execution of all threads of the parallel program to be analyzed is finished, wherein the parallel program to be analyzed comprises at least two threads; according to the results of simulation execution of the parallel program to be analyzed, constructing a multithread control flow graph corresponding to the parallel program to be analyzed; processing the multithread control flow graph to obtain a conservative and correct multithread control flow graph, and no missing of possible parallel MHP result of each node of the conservative and correct multithread control flow graph; and performing the MHP analysis on the conservative and correct multithread control flow graph to obtain the MHP analysis result of the parallel program to be analyzed.

Description

technical field [0001] The embodiments of the present invention relate to the field of computer technology, and in particular to a parallel program MHP analysis method and device. Background technique [0002] With the development of computer technology, computers have developed from single-core to multi-core era, from stand-alone processing to large-scale cluster processing. The development of computers has also promoted the development of computer software. Up to now, hundreds of computer programming languages ​​have been created, such as popular C / C++, Java, Python, etc. Various large-scale software programs, such as operating systems, databases, and large-scale software games, often contain millions of lines of code. These large-scale software programs may take years to develop by a team. Then, how to debug software programs more effectively, and how to verify and optimize the correctness of programs has become a research hotspot in computer software development. [00...

AI Technical Summary

Problems solved by technology

However, the current method of constructing a multi-threaded GFG is too simple, and generally only examines the analysis of creation (creat) primitives and lock (lock) / unlock (unlock) primitives, while ignoring the importance of Happens-Before analysis. As a result, when performing MHP analysis based on CFG, the analysis results have a large number of false positives or false negatives

Images