基于分组的并行程序多路径覆盖测试数据进化生成

尽管并行软件测试已经得到软件工程界的广泛关注,但是,如何高效生成覆盖并行软件多条路径的测试数据,相关的研究还比较少。本文研究消息传递并行程序多路径覆盖测试数据生成问题,并提出基于分组的测试数据进化生成方法。首先根据并行程序包含的进程数、可用的计算资源以及路径相似度,将目标路径分成若干组,并基于每组目标路径,建立多路径覆盖测试数据生成问题的数学模型;然后采用多种群并行遗传算法求解上述模型,使得一次运行遗传算法,生成覆盖所有目标路径的测试数据。性能分析表明,所提出的目标路径分组方法不但能够保证不同组包含的目标路径相差很少,而且同一组的目标路径之间具有很大的相似度。将所提方法应用于4个基准程序的测试中,并与已有方法比较,结果表明,所提方法在保证路径覆盖率的前提下,可大大缩减个体评价次数和耗时。

Evolutionary generation of test data for multi-paths coverage of parallel programs by grouping

Parallel software testing has attracted wide-ranging attention in the community of software engineering.However,less research is concerned with effectively generating test data for multi-paths coverage of parallel programs.We investigated the problem of generating test data for multi-paths coverage of message-passing parallel programs and proposed a grouping-based method of evolutionarily generating test data.First,target paths were divided into several groups according to the number of processes in a parallel program,available computation resources,and the similarities of target paths,and the mathematical model of generating test data for multi-paths coverage was built based on the target paths belonging to each group.Secondly,a multi-population parallel genetic algorithm was employed to solve the above model,so that the test data covering all target paths could be generated in one run of the genetic algorithm.Performance analysis indicates that the proposed method guarantees not only a small difference in the number of target paths belonging to different groups,but also a great similarity among target paths in the same group.The proposed method was applied to four benchmark programs,and compared with existing methods.The experi-mental results show that the proposed method greatly reduces the number of evaluated individuals and the consumption time on the promise of meeting the path coverage rate.