Execution method and system for parallel dynamic symbol based on Actor model

Abstract

The invention relates to an execution method and system for a parallel dynamic symbol based on an Actor model. The method comprises the steps of combining a parallel framework of the configured Actor model into an amended dynamic symbol execution tool; taking out a task from a preconstructed path stack to be solved by multiple work nodes of the Actor model through communication nodes, and obtaining a task path constraint value of the task through a task survey meter; using a constraint solver to solve the task path constraint value and obtain a solution value Valuation; substituting the solution value Valuation into a path in the path stack to be solved by adopting a recursive method to generate a path to be explored, and storing the path to be explored in a preconstructed path stack to be explored. According to the execution method for the parallel dynamic symbol, parallelization on two levels is achieved, parallel analysis of program paths can be conducted on multiple nodes, constraint solution and path exploring can be simultaneously conducted, program time consumption is reduced, and the capacity of testing a large-scale program can be strengthened.

Description

technical field [0001] The invention mainly relates to the technical field of software testing, in particular to a parallel dynamic symbolic execution method and system based on Actor model. Background technique [0002] The symbolic execution technology proposed in 1976 is a static software analysis method, which uses symbolic values ​​instead of specific values ​​as the input of the program, and is mainly used for software path exploration, test case generation and software defect detection. Dynamic symbolic execution is a method proposed during the development of symbolic execution. It combines the advantages of concrete execution and symbolic execution. In some cases, path constraints may not be able to solve for specific values ​​during path exploration by symbolic execution and cannot continue. Path exploration, at this time, the use of dynamic symbolic execution can solve this problem through its specific execution. [0003] The efficiency of the traditional symbolic...

AI Technical Summary

Problems solved by technology

However, Cloud9 needs to maintain the entire symbolic execution tree during the execution process, and the path search consumes a lot of extra time. In addition, Cloud9 is aimed at classic symbolic execution tools rather than dynamic symbolic execution. It is different from this paper in the task generation and distribution mechanism. of

[0006] (2) The second is to consider the implementation of parallelization in constraint solving: the Concurrent Bounded Model Checking (CBMC) method proposed by Quoc-Sang Phan et al. is implemented by multiple threads. When the main thread performs path exploration, it collects The path constraint does not solve the constraint, but waits until the path exploration is completed, then analyzes its structure and converts it into a series of disjoint constraint subsets and solves it concurrently with multiple threads. This method can improve the solution efficiency, but the disadvantage is that it does not It is not truly parallel, and the constraint solving analysis starts after all path explorations are completed, which wastes the waiting time in the middle and reduces the overall efficiency

[0007] One of the main problems in the above method (1) is the maintenance tree problem. When multi-node execution tasks, this method needs to maintain a whole symbolic execution tree, which consumes a lot of extra time when exploring. In addition, This method is based on classic symbolic execution tools, not dynamic symbolic execution tools, such task generation and allocation need to waste a lot of system resources and time

In method (2), parallel technology is used to solve the problem, and the constraints of each path are assigned to multiple nodes for solution. However, this method is only aimed at the application scenario of model detection, and only parallelizes the constraint solving process.

Images