基于云进化算法的3D NoC测试规划

Test Scheduling for 3-D Network on Chip Based on Cloud Evolutionary Algorithm

摘要:

针对三维片上网络(Three Dimensional Network-on-Chip, 3D NoC) IP核的测试问题,采用云进化算法优化测试规划,完成3D NoC测试；该方法首先通过平分搜索范围的方式形成第一代种群,依据3D NoC结构特点建立功耗模型,在满足功耗约束的情况下进行测试,采用种群精英个体保留策略选择优秀个体,并利用云模型的随机性和稳定性特点进行迭代寻优,旨在降低总的测试时间,获得最佳测试规划;以ITC′02测试标准电路作为实验对象,实验结果表明,在获得相同测试时间下,云进化算法比遗传算法具有更好的寻优能力,收敛代数提高了约50%,有效提高了测试效率。

Abstract:

For the test scheduling of 3D NoC (Three Dimensional Network-on-Chip, 3D NoC) resources cores, this paper presents a test scheduling method based on cloud evolutionary algorithm to complete test. Firstly, formed the first generation of population by means of share the whole scope, secondly, established power model based on the structural characteristics of 3D NoC, then complete test under the condition of dual power consumption constrain, thirdly, select excellent individuals by use the population elite reserve strategy, finally, use the randomness and stable tendency property of cloud model complete the iterative optimization and then reduce the total test time. Experiments with ITC′02 test circuit as the simulation object, the experimental results show that, For the same test time, cloud evolution algorithm has better search ability than the genetic algorithm, convergence is improved by about 50%, and improve the efficiency of the test.