arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2403.18569cs.LGcs.AI
收藏

PDNNet:用于动态 IR Drop 预测的 PDN 感知 GNN-CNN 异构网络

PDNNet: PDN-Aware GNN-CNN Heterogeneous Network for Dynamic IR Drop Prediction

School of Integrated Circuits, Peking University(北京大学集成电路学院);School of Microelectronics, Wuhan University(武汉大学微电子学院);Hubei Luojia Laboratory(湖北珞珈实验室);Institute of Electronic Design Automation, Peking University(北京大学电子设计自动化研究所);Beijing Advanced Innovation Center for Integrated Circuits(北京集成电路高精尖创新中心)

Yuxiang Zhao, Zhuomin Chai, Xun Jiang, Yibo Lin, Runsheng Wang, Ru Huang

AI总结:

针对大规模 IC 设计中动态 IR drop 仿真成本高且 CNN 方法忽视 PDN 配置的问题,提出 PDNGraph 与双分支 GNN-CNN 异构网络 PDNNet,实现更准确、可解释的预测,并较商业工具加速 545 倍。

摘要:

供电网络(PDN)上的 IR drop 与 PDN 的配置以及单元电流消耗密切相关。随着集成电路(IC)设计规模不断增大,动态 IR drop 仿真在计算上变得难以承受,基于机器学习的 IR drop 预测已被探索为一种颇具前景的解决方案。尽管基于 CNN 的方法已在若干工作中被适配用于 IR drop 预测任务,但其忽视 PDN 配置的缺点不可忽略。本文中,我们不仅考虑如何恰当地表示单元-PDN 关系,还考虑如何在特征聚合过程中遵循 IR drop 的物理本质对其建模。因此,我们提出一种新颖的图结构 PDNGraph,以统一表示 PDN 结构和细粒度的单元-PDN 关系。我们进一步提出一种双分支异构网络 PDNNet,融合两个并行的 GNN-CNN 分支,以便在学习过程中有效捕获上述特征。文中给出了若干关键设计,使动态 IR drop 预测兼具高效性与可解释性。我们是首个将图结构应用于基于深度学习的动态 IR drop 预测方法的工作。实验表明,PDNNet 优于最先进的基于 CNN 的方法,并且与商业工具相比实现了 545 倍加速,证明了本方法的优越性。

英文摘要:

IR drop on the power delivery network (PDN) is closely related to PDN's configuration and cell current consumption. As the integrated circuit (IC) design is growing larger, dynamic IR drop simulation becomes computationally unaffordable and machine learning based IR drop prediction has been explored as a promising solution. Although CNN-based methods have been adapted to IR drop prediction task in several works, the shortcomings of overlooking PDN configuration is non-negligible. In this paper, we consider not only how to properly represent cell-PDN relation, but also how to model IR drop following its physical nature in the feature aggregation procedure. Thus, we propose a novel graph structure, PDNGraph, to unify the representations of the PDN structure and the fine-grained cell-PDN relation. We further propose a dual-branch heterogeneous network, PDNNet, incorporating two parallel GNN-CNN branches to favorably capture the above features during the learning process. Several key designs are presented to make the dynamic IR drop prediction highly effective and interpretable. We are the first work to apply graph structure to deep-learning based dynamic IR drop prediction method. Experiments show that PDNNet outperforms the state-of-the-art CNN-based methods and achieves 545x speedup compared to the commercial tool, which demonstrates the superiority of our method.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑