摘要
SoC(System-on-a-Ch ip)系统芯片的嵌入式操作系统(Embedded Operating System)软/硬件自动划分(SoC-EOS划分)是一个NP完全问题,也是SoC软/硬件协同设计的一个关键步骤,它决定了SoC-EOS功能的软/硬件实现,其划分结果直接影响到SoC产品的开发效率和质量。引入了SoC-EOS划分问题的一个新模型,这有助于理解SoC-EOS划分问题的本质。提出了一种基于离散Hopfield神经网络的SoC-EOS划分方法,重新定义了神经网络的能量函数、运行方程和相关系数。最后,对该方法进行了仿真实验,并同遗传算法和蚂蚁算法进行了性能比较。实验结果表明,提出的神经网络方法是可行的和有效的。
The hardware-software automatic partitioning of embedded operating system in the SoC (SoC-EOS partitioning) is a NP-complete problem, and a crucial step in the hardware-software co-design of SoC. It determines the hardware-software implementation of SoC-EOS, and its result directly influences the development efficiency and quality of SoC prod- ucts. A new model for SoC-EOS partitioning was introduced, which can help on understanding the essence of the SoC- EOS partitioning. A discrete Hopfield neural network approach for implementing the SoC-EOS partitioning was proposed, where a novel energy function, operating equation and coefficients of the neural network are redefined. Then simulations were carried out with comparisons to the genetic algorithm and ant algorithm in the performance. Experimental results demonstrated the feasibility and effectiveness of the proposed method.
出处
《四川大学学报(工程科学版)》
EI
CAS
CSCD
北大核心
2006年第5期122-127,共6页
Journal of Sichuan University (Engineering Science Edition)
基金
国家自然科学基金资助项目(60572026)