摘要
针对异构多核片上网络(NoC)的任务映射问题,根据IP核的选择以及IP核向NoC平台中位置映射的两个阶段的不同特点,分别提出能耗和延时的粗略估算模型和精确计算模型。为避免离散空间搜索解落入局部最优,设计了混沌扰动机制。提出了带混沌扰动机制的改进型离散粒子群优化算法,以搜索能耗和延时优化的多目标NoC映射方案,该算法比传统优化算法在能耗和延时上有显著的性能提高。
Heterogeneous multi-core cooperative networks-on-chip(NoC) mapping was split into two stages:assigning the tasks to the suitable IP cores,and then mapping the IP cores to the appropriate NoC tiles.To deal with the different characteristics of these two successive stages,a coarse model and an accurate model of energy consumption or delay estimation were proposed respectively.A discrete particle swarm optimization with chaotic disturbance was proposed to solve the multi-objective NoC mapping problems,where a chaotic disturbance mechanism was designed to avoid obtaining local optimal solutions.The simulation results are better than that obtained by traditional schemes significantly.
出处
《计算机科学》
CSCD
北大核心
2011年第9期298-303,共6页
Computer Science
基金
国家中长期重大专项项目(2011ZX03003-003-04)
国家自然科学基金项目(60873076)资助
关键词
片上网络
映射
多目标优化
离散粒子群优化
NoC
Mapping
Multi-objective optimization
Discrete particle swarm optimization