基于粒子群算法的多约束3D NoC协同测试规划

Cooperative test scheduling of 3D NoC under multiple constraints based on the particle swarm optimization algorithm

为了提高三维片上网络(3D NoC)资源内核的测试效率,对多约束下的3D NoC进行测试规划。在硅通孔(TSV)数量、功耗以及带宽约束下,分别将TSV位置、IP核测试数据分配作为两个寻优变量,利用离散粒子群算法协同进化,以减少测试时间并提高TSV利用率。在算法中引入全局次优极值对粒子进行指导,提高全局搜索能力;并通过自适应参数调整策略增加种群多样性,从而改善粒子搜索的停滞现象。以国际标准测试集ITC'02中的电路作为仿真对象,仿真结果表明,算法能够有效地完成在多约束下对TSV位置的寻优并合理分配通信资源,缩短了测试时间,提高了TSV利用率。

To improve the testing efficiency of the IP core in three-dimensional network-on-chip（ 3D NoC）,research on test scheduling of the 3D NoC under the multiple constraints is conducted. A new method taking the TSV allocation and the IP core test data assignment as two optimization variables respectively,and co-evolved them by the discrete particle swarm optimization algorithm is proposed to minimize the test time and fully utilize the limited TSVs under the multiple restrictions such as the limited number of TSVs,power and the constant bandwidth. To enhance the global searching capacity and increase the diversity of population to refine the stagnation phenomenon,strategies of introducing the global sub-optimal extreme＇s influence on the update of particle and adjusting the parameters by self-adaption are designed. Taking ITC ＇02 test benchmark as the experiment object,simulation results demonstrate that the proposed method can effectively accomplish the placement optimization of TSVs and the allocation of the communication resource under the multiple constraints,and therefore the test time is shortened and the TSVs＇ utilization is improved.