Research On Test Case Generation Method Of Parallel Program Based On Path Coverage

Software testing is an essential link in the process of developing software, it starts from the requirements analysis phase until the interactive software products. Through the test of product can make software products more in line with the needs of users, and can reduce development cycle and development costs. At the heart of the test is valid test data,the existing method of test data generation is mainly to the serial program, with the progress of era of parallel computing to high-speed development, the parallel technology is widely used in image processing, such as weather forecast problem.Compared to the serial program, the structure of the parallel program is more complex,it contains multiple processes between the need to transfer data through special statement,realizing the information sharing between process, make full use of parallelism. Academics have resource competition of parallel programs, such as deadlock problem test scheme is proposed. In between processes can right under the premise of sharing information, need to test the internal structure of the program, thus enhance the reliability of the parallel program. So it is necessary to explore the structure of the parallel program under cover, the existing test program structure covering method is based on a single coverage criterion of test data. Due to the nature of the parallel program, the existing method of generating test data serial program is no longer suitable for parallel programs. This article research meets the path coverage criterion of parallel program test data generation.Message passing interface(Message passing interface, MPI) has strong portability, the advantages of simple operation, is widely used to write parallel programs. MPI parallel program is divided into communication determine the MPI parallel program and not sure of the MPI parallel program due to communication mechanism, in the article, first of all,analysis these two kinds of parallel program, understand the composition of a parallel program, the basic concepts such as path. Second, the study of the existing group of intelligent algorithm, combined with the characteristics of MPI parallel program proposed is applicable to solving the MPI parallel program test data cross collaborative evolutionary artificial colony algorithm(Cross co- evolutionary artificial bee colony algorithm,CCABC), the algorithm of crossover mechanism was introduced and cooperative co-evolution. Third, selection of communication to determine the MPI parallel program,establish mathematical model and the design of fitness function, application CCABC algorithm generated path coverage criterion of test data, through specific examples to verify the effectiveness of the algorithm of CCABC. Finally, using CCABC algorithm to generate test data of the communication of uncertain MPI parallel program.The programon the number of test data in produce proposes two solutions, one is transformed into communication to determine the MPI parallel program, and then in solving test data.Another is the definition of similarity path set is given, after looking for similar path set in the path of to the target then application CCABC algorithm generation test data. Through experiment and comparison experiment proves the feasibility and efficiency of the scheme.