基于粒子群算法的3D NoC测试优化方法

Optimization of 3D NoC Test Based on Particle Swarm Optimization

三维片上网络(3D NoC)中资源内核的测试问题日益突出,本文采用带分复用策略并结合改进离散粒子群算法对IP核测试数据分配及调度方案进行快速寻优.算法通过群体多样性的监控实现对粒子前行速度的动态调节,增强全局探索能力,并自适应调整惯性权重,以加强粒子的局部开发能力,进而改善粒子搜索的停滞现象.以ITC'02测试标准电路作为实验对象,仿真结果表明,本文方法能有效地完成3D NoC资源内核的最大化并行测试,减少了测试时间,提高了资源利用率.

The IP core test problem in three dimensional network-on-chip （3D NoC） is becoming increasingly obvious. In this paper, the improved discrete particle swarm optimization algorithm and a bandwidth division multiplexing method are used to optimal the test data assignment and scheduling for IP core. In order to enhance the global exploration ability and improve the stagnation phenomenon, the algorithm realizes the dynamic adjustment of the forward speed of particle by monitoring the population diversity, meanwhile, adjust the inertia weight adaptively to increase the local development ability. Taking ITC＇02 standard circuit as the test object, experiment was conducted; and the experiment results show that the proposed method can effectively maximize the parallel test of 3D NoC, thus shortened the test time as well as improved the resource utilization.