• 中国期刊全文数据库
  • 中国学术期刊综合评价数据库
  • 中国科技论文与引文数据库
  • 中国核心期刊(遴选)数据库
林孔成, 孙希延, 纪元法, 等. 融合早期时钟流程与CCOpt的时钟树优化[J]. 桂林电子科技大学学报, 2024, 44(2): 118-126. DOI: 10.16725/j.1673-808X.2023199
引用本文: 林孔成, 孙希延, 纪元法, 等. 融合早期时钟流程与CCOpt的时钟树优化[J]. 桂林电子科技大学学报, 2024, 44(2): 118-126. DOI: 10.16725/j.1673-808X.2023199
LIN Kongcheng, SUN Xiyan, JI Yuanfa, et al. CCOpt clock tree optimization and early clock flow[J]. Journal of Guilin University of Electronic Technology, 2024, 44(2): 118-126. DOI: 10.16725/j.1673-808X.2023199
Citation: LIN Kongcheng, SUN Xiyan, JI Yuanfa, et al. CCOpt clock tree optimization and early clock flow[J]. Journal of Guilin University of Electronic Technology, 2024, 44(2): 118-126. DOI: 10.16725/j.1673-808X.2023199

融合早期时钟流程与CCOpt的时钟树优化

CCOpt clock tree optimization and early clock flow

  • 摘要: 随着集成电路工艺节点的演进,芯片集成规模的不断扩大以及工作频率的提高,传统时钟树综合策略无法满足目前芯片设计的时序要求。时钟协同优化(CCOpt)技术不能有效解决时钟树综合后的绕线拥塞和时钟门控路径的时序优化不佳问题。为此,提出一种融合早期时钟流程和CCOpt技术的时钟树优化方法,通过在标准单元布局阶段提前构建时钟树,对时钟树绕线、时钟单元驱动和间距进行约束,并运用和调控CCOpt的有用偏差技术,对数据和时钟路径同时进行优化来完成时钟树综合。将该方法应用到6 nm工艺下的PCIe模块进行验证,实验结果表明,该方法能有效缓解时钟树综合后的绕线拥塞,最差时序违例值降低了63.6%,时钟门控路径时序优化了20.3%,时钟网络功耗降低了1.54%,整体面积减小1.8%,有效提高了芯片的性能。

     

    Abstract: With the evolution of IC process nodes, the continuous expansion of chip integration scale and the increase of operating frequency, the traditional clock tree synthesis strategy can't meet the timing requirements of current chip design. The clock concurrent optimization technology can't effectively solve the problem of routing congestion after clock tree synthesis and poor timing optimization of clock gated paths. To solve this problem, a clock tree optimization method combining early clock flow and CCOpt technology is proposed. The clock tree is constructed in advance in the placement stage, the clock tree routing, clock cell drivers and spacing are constrained, and the useful skew technology of CCOpt is used to optimize data and clock path at the same time to complete the clock tree synthesis. The proposed method is applied to the PCIe block in 6 nm process. The results show that the proposed method can effectively alleviate the routing congestion problem after clock tree synthesis. The worst negative slack is reduced by 63.6%, the clock gate path timing is optimized by 20.3%, the clock network power consumption is reduced by 1.54%, and the overall area is reduced by 1.8%. The performance of the chip is effectively improved.

     

/

返回文章
返回