Research On Test Scheduling Of 3D NoC Collaborative Optimization

System on chip(SoC) based on bus appear low communication efficiency, poor scalability, long transmission delay and high power communication problems when used to large scale parallel computing. Network on chip(NoC) which use routing and packet switching emerged to solve the bottleneck problems of SoC and become a mainstream technology. But with the rapid development of semiconductor integrated technology, NoC technology is limited by planar structure, so some scholars have proposed merge 3D IC into NoC become 3D NoC chip. However the increase of species and quantity of IP cores in 3D NoC make the test become more and more important, at the same time the increase of complexity of chip also make the test become more difficult. Therefore, how to achieve efficient test schedule is one of the problems for 3D NoC technology.Based on the study of 3D NoC system basic structure and key technologies related on IP core testing, using the bandwidth division multiplexing implement dynamic bandwidth allocation scheme to research on test scheduling of 3D NoC, combining reuse NoC as test access mechanism(TAM). Under the dual condition of power consumption constrain, Design the multiple population cloud evolution algorithm to optimize test sequence produced by bandwidth allocation, implement hybridization technique in elite group, and achieve the shortest test time by collaborative optimization bandwidth and schedule. The experiment consider the ITC’02 test circuit as simulation object, compare with the non-implement dynamic bandwidth allocation, comparative results show that the method and algorithm proposed of the paper can reduce the test time.Simultaneously, under the condition of TAM bandwidth, combine with Pareto set, cloud model and hybrid technology variation collaborative optimization time and power during test 3D NoC. Implementation non-dominated sorting for individual of population, given the appropriate level for all individuals, removal of extreme individuals with similar characteristics, using cloud mode to update individuals with elitism retain strategy, using hybridization and mutation policy find optimal scheduling scheme set in evolution later period. The experiments show that the algorithm can collaborative optimization test time and power, and choose different scheduling scheme which is needed in the practical.