Going with the flow: bridging the gap between theory and practice in physical design
Going with the flow: bridging the gap between theory and practice in physical design
复制标题
顺其自然：弥合物理设计理论与实践之间的差距
DOI：
10.1145/1735023.1735026
复制
发表时间：
2010
期刊：
影响因子：
--
通讯作者：
P. Groeneveld
中科院分区：
文献类型：
--
作者：
P. Groeneveld
实际的物理设计流程是许多算法的复杂链。各种各样的布局和路由器穿插着增量逻辑合成算法，修补了时间。总体目标是同时解决过多的设计目标，如LVS/DRC正确性，速度，面积，功耗和产量。每一个新的技术节点都会给流程增加额外的复杂性。串扰灵敏度的提高，以及更严格的功率和成品率要求，目前正在对28 nm造成严重破坏。
EDA研究主要集中在改进单个算法上。在过去的十年里，ISPD举办了比赛，以寻找最佳的放置和全球路由工具。虽然重要，但每个算法都是最优的真的重要吗？考虑到许多相互冲突的目标，“最佳”意味着什么？本演示侧重于基于最先进的商业物理设计工具集的经验的整体流程问题。
在多目标设计流程中预先确定优先级是很重要的。哪些影响占主导地位，哪些可以逐步解决？令人惊讶的是，从一个在其成本函数中使用冲突目标的算法中很难得到好的结果。相反，通常更可取的做法是让每个算法只解决一个问题，并在接下来的步骤中逐步修补不太重要的问题。本演示文稿还解决了多算法流的“颠簸”问题。许多局部最优值和有限的实验证据使得很难适当地调整算法链。
A practical Physical Design flow is an intricate chain of many algorithms. Assorted placers and routers are interspersed with incremental logical synthesis algorithms that patch up timing. The overall goal is to address a plethora of design objectives simultaneously such as LVS/DRC correctness, speed, area, power and yield. Each new technology node adds additional complications to flow. Increased crosstalk sensitivity, and tighter power and yield requirements are currently causing havoc in 28nm.
EDA research has been focused primarily on improving individual algorithms. Over the past decade ISPD has hosted contests to find the best placement and global routing tool. Though important, does it really matter for each algorithm to be optimal? And what does "optimal" mean given the many conflicting objectives? This presentation focuses on the overall flow issues based on the experiences with a state-of-the-art commercial physical design toolset.
It is important to be up-front about priorities in a multi-objective design flow. Which effects dominate and which can be addressed incrementally? It is surprisingly hard to get good results from an algorithm that uses conflicting objectives in its cost function. Instead, it is often preferable to let each algorithm address only a single issue and patch up the less important ones incrementally in the next steps. This presentation also addresses the "bumpiness" of multi-algorithm flows. The many local optima and limited experimental evidence make it hard to properly tune the chain of algorithms.