摘要
面向基于平台的系统芯片设计,提出具有初始信息素的蚂蚁寻优软硬件划分算法AOwIP·基本思想是:①利用基于平台的设计方法中已有参考设计的软硬件划分结果作为初始划分解,进行适当变换后生成初始信息素分布·②在所生成初始信息素分布的基础上,利用蚂蚁算法正反馈、高效收敛的优势寻求最优划分解·该算法利用基于平台的设计方法强调系统重用的优势,克服蚂蚁算法在求解软硬件划分问题时缺乏初始信息素的不足·实验表明,AOwIP算法有效提高了蚂蚁算法的最优解搜索效率·
Ant optimization algorithm gets to optimal results efficiently, but it lacks initial pheromone at the beginning, which limits its further improvement. A hardware/software partitioning algorithm is presented for platform-based system-on-a-chip design. This algorithm is based on ant optimization with initial pheromone called AOwIP, and the basic ideas are： a） Use the partitioning result of reference design provided by platform-based design method as initial partition of current design, and convert the initial partition into initial pheromone for ant algorithm, and b） Based on the initial pheromone generated, the AOwIP makes use of such advantages as positive feedback and efficient convergence of the ant algorithm to search for the optimal partitioning scheme. The AOwIP adopts the system level reusing feature of platform- based design to avoid the ant algorithm＇s shortcoming of lacking initial pheromone. Experiments show that the AOwIP improves the performance of ant algorithm for hardware/software partitioning problems.
出处
《计算机研究与发展》
EI
CSCD
北大核心
2005年第12期2176-2183,共8页
Journal of Computer Research and Development
基金
国家自然科学基金项目(90207019)
国家"八六三"高技术研究发展计划基金项目(2002AA1Z1480)~~
关键词
蚂蚁算法
基于平台的设计
软硬件划分
系统芯片
ant algorithm
platform-based design
hardware/software partitioning
system-on-a-chip（SoC）
