【作者】 闫颖；

【导师】 程利新；

【作者基本信息】 哈尔滨工程大学 ， 计算机系统结构， 2010， 硕士

【摘要】 高级综合是数字系统设计自动化的关键技术之一,是近年来国内外研究、开发和应用的热门课题。高级综合工具的出现简化了复杂集成电路的设计过程,缩短了设计周期。高级综合理论的研究和算法的改进,对于提高综合质量非常重要。本文主要关注高级综合中调度和分配两个关键步骤的优化问题。首先,本文在分析以往高级综合系统所采用的技术基础上,提出了一种新的高级综合调度优化策略。采用一种基于模拟退火算法的方案,优化高级综合的调度过程,目的是以在较短的时间内逼近全局最优解。为了在调度过程中进一步考虑调度和分配的相互作用,提出了同时考虑时间、造价和功能单元利用率的能量函数,将分配结果作为计算调度方案能量函数的要素之一。资源分配采用了一种快速的团划分算法实现,尽可能的满足能量函数易于计算的需求。其次,为了在实际应用中进一步减少算法执行时间和改善结果质量,本文引入了对基本模拟退火算法的几种改进方案。采用改进算法流程和增加新功能等手段,来减轻模拟退火算法中一些固有缺陷的影响。通过加温退火、升温过程、记忆功能和返回搜索等改进策略,使算法具备了自适应初始温度选择、避免过早陷入局部最优和防止错过全局最优等功能,从而进一步提升了时序调度的优化效果。本文实现了所提出的方案,通过实验验证了使用该方案优化调度和分配问题的效果。在本文的最后,给出了算法的部分关键参数和主要的实验数据。更多还原

【Abstract】 High-level synthesis is one of the key technologies of digital system design automation, and has become a hot topic of investigation, development, and application at home and abroad in recent years. The advent of the high level synthesis tools simplifies the design process of complex integrated circuits, and shortens design cycles, and the study of high level synthesis theory and the improvement of related algorithm is of great importance. In this paper, attention is concentrated on the scheduling and allocation of high level synthesis.First, a new optimization strategy of high level synthesis is put forward, based on researching the technology adopted in other high level synthesis system. The scheduling of high level synthesis is optimized with a solution based on simulation annealing algorithm. This method further considers the interaction of scheduling with allocation in the process of scheduling, and presents a new form of energy function involving three factors: time, cost and Utility Ratio. The result of allocation is treated as a factor of energy function. The allocation is completed by a quickly clique partitioning algorithm, to meet the needs of energy function which must be computed easily.Second, several improvements of simulation annealing algorithm is imported, for the decrease of operation time and promotion of effect in the practice. The process and increasing new capabilities is perfected, to avoid the defects of simulation annealing algorithm. Heating annealing, tempering annealing, memorial annealing and annealing with back searching are introduced to solve the problems such as the selection of initial temperature, to avoid falling into local optimum too early and prevention of miss global optimum. In doing so, the effect of scheduling optimization is further improved. The scheme which is put forward in this paper is implemented, and the effects of the application of optimizing the scheduling and allocation process are tested. Finally, the key parameters and experiment data are given to verify the effects.更多还原