CPA-DA: Efficient Sequential Synthesis and Optimization for High-Performance Circuits

项目编号：0811270题目：CPA-DA：高性能电路的高效顺序合成与优化机构：西北大学摘要处理速度和功耗已成为现代高性能VLSI电路中最重要的设计标准。它们不仅依赖于组合部分，而且还依赖于存储元件（锁存器和触发器）和时钟机制。在项目中，我们开发了一个基于增量重定时和其他网络流技术的高效顺序综合和优化框架。该框架基于结构操作，如重新计时、重新合成和扫描，这些操作几乎涵盖了所有可能的顺序转换。基于增量重定时、受限时钟倾斜调度和顺序楼层规划的有效算法的顺序优化套件已开发并将向公众提供。结合研究和教育活动的结果提高了所有类型的纳米VLSI电路的性能和设计生产力，从高速微处理器到无处不在的pda，手机等中的片上系统（soc）。性能和设计生产力的提高转化为客户更低的成本和更好的功能，从而使所有消费者，特别是社会经济上处于不利地位的消费者，都能获得负担得起的信息技术。此外，高性能计算机，通过优化性能和提高设计效率，将有助于其他重要科学领域的研究，如遗传学，生物信息学和医学。

PI: Zhou, HaiProposal No: 0811270Title: CPA-DA: Efficient Sequential Synthesis and Optimization for High-Performance CircuitsInstitution: Northwestern UniversityABSTRACTProcessing speed and power consumption have become the most important design criteria in modern high-performance VLSI circuits. They are not only dependent on the combinational part but also on the memory elements (latches and flip-flops) and the clocking mechanism. In the project, we develop an efficient sequential synthesis and optimization framework based on incremental retiming and other network flow techniques. The framework is based on structural operations such as retiming, re-synthesis, and sweep, which are shown to cover almost all possible sequential transformations. A sequential optimization suite based on efficient algorithms for incremental retiming, constrained clock skew scheduling, and sequential floor-planning is developed and will be made available to the public.The outcomes of combined research and education activities improve the performance and design productivity of nanometer VLSI circuits of all types, from high-speed microprocessors to the omnipresent systems-on-a-chips (SOCs) found in PDAs, cell phones, and more. The improvement on performance and design productivity translates to lower cost and better functionality to customers, thus giving all consumers, particularly socio-economically disadvantaged, access to affordable information technology. Furthermore, the high performance computers, enhanced by optimized performance and improved design productivity, will aid research in other important scientific areas such as genetics, bioinformatics, and medicine.