Applying PSO To Test Case Optimization And GPGPU Acceleration

Regression testing is a very efficient way to assure the quality of software during the development. Test case prioritization, a technique to search the best sequence of test cases execution, is essentially vital to regression testing. At the same time in regressing testing for real projects, single-objective test cases prioritization has been gradually replaced by multi-objective prioritization. As a result, multi-objective optimization technique, especially applying Evolution Algorithms to address optimization problems is a hot spot in current research.However, MOTCP itself is time-consuming technique, even based on the evolutionary algorithm, which is a research hot spot and successfully applied into different optimization problems in many fields. Especially, evolution algorithms simulate the natural feature and evolve basing on the population. So, with the increasing scale of the population as well as the optimization problem, the performance of EAs declines sharply.To address this problem, this paper presents a test suite prioritization technique based on Particle Swarm Optimization, where the corresponding particle representation, position and speed updated methods are proposed. Meanwhile, we analysis the performance of our method deeply and use the GPGPU acceleration technique to improve the performance.Empirical studies are conducted to study the effect caused by different types of particle updating methods and size of particle swarm. Compared to the NSGA-Ⅱ based test case prioritization, the proposed technique is more efficient in test suite prioritization in real programs. In addition, we use GPGPU to accelerate the fitness function computing section and greatly improve the performance of our method.