Communication Coverage Testing Method Of Parallel Programs Based On Dominant Relation

Messaging-passing parallel program is a kind of widely used parallel program,which contains multiple parallel execution processes.Communication between processes is based on communication statements,and the communication is the main characteristic of the messagepassing parallel program.The core of communication testing is to efficiently generate test data covering communication.With the increase of program scale,communication between processes becomes more and more frequent,which makes communication testing more and more difficult.It is necessary to study the effective communication coverage method to improve the test efficiency.In this thesis,the communication coverage test of parallel programs for message-passing is studied from two aspects: communication coverage reduction and communication coverage test data reduction.Aiming at the problems of communication scale and communication uncertainty,and considering the characteristics of dominant relation,the theory and method of communication coverage reduction based on dominant relation were proposed.Firstly,based on the correspondence between communication and communication statements,the communication reduction problem was transformed into the reduction problem of communication statements,and the target scheduling sequences were obtained by reducing the scheduling sequence set.Then,the dominant relation among communication statements was analyzed,and the dominant relation diagram was generated to obtain the reduced set of communication statements.The communications were selected as the coverage targets according to the communication statements in the reduced set.Finally,based on the target scheduling sequence set,genetic algorithm was used to generate test data covering the target communications.Aiming at the problem of the scale of communication coverage test data generation,this thesis further proposed a test data reduction method based on the reduction set obtained by the statement dominance relation.Firstly,the communication statements in the set were divided and reduced by branching relation to obtain the communication statement groups.Then,the program under test was executed,the statement groups were sorted according to the execution order of the communication statement groups,and the communication statement queues were obtained by dividing the communication statement groups in turn.Finally,the fitness priority generation algorithm and the priority selection algorithm were given,which were applied to the genetic algorithm to generate the test data covering the communication statement queue,so as to realize the reduction of the test data.This thesis aimed at the communication coverage test of parallel programs.The communication coverage reduction method was proposed to reduce the target communications.The communication coverage test data reduction method was presented.By dividing the communication statement queues,the coverage targets were further reduced to reduce the size of the test data generation.On this basis,the communication coverage testing method of parallel programs based on dominant relation was completely introduced,and the realization process of communication coverage testing data generation was given.Experimental analysis shows that the proposed theory and method can reduce the generation time of test data and improves the efficiency of generation of communication coverage test data without reducing the communication coverage.