基于LSTM的逻辑综合阶段延时预测方法

doi:10.6040/j.issn.1672-3961.0.2025.057

摘要/Abstract

摘要： 为了提高数字集成电路逻辑综合阶段的延时评估精度与效率,提出一种基于长短期记忆网络(long short-term memory,LSTM)的逻辑综合阶段延时预测方法。将时序路径表示为由标准单元构成的有序序列,提取并构建包含单元类型、扇出、负载电容和输入转换时间等关键特征参数的结构化序列数据;通过LSTM时序建模中的上下文记忆能力,捕捉路径中各级单元之间复杂的时序依赖关系,实现对路径延时的高精度预测。试验结果表明,对比现有对单元延时和线延时进行累加的机器学习估算方法,在预测精度上,基于LSTM的预测方法在保证高准确率的前提下,对不同类型的案例具有更好的适应性;在运行速度上,在多数测试案例上实现2.8~3.2倍加速。在无工艺信息的通用门级网表上验证本研究方法的预测能力,其表现优于传统静态时序分析方法,验证了该方法在早期设计阶段的有效性和应用前景。

关键词: 长短期记忆网络, 逻辑综合, 静态时序分析, 路径延时预测, 机器学习

Abstract: To improve the prediction accuracy and efficiency for logic synthesis in digital integrated circuit design process, a logic synthesis delay predicting method based on long short-term memory(LSTM)was proposed. The timing path was treated as an ordered sequence of standard cells, and key feature parameters such as cell type, fanout, load capacitance, and input transition time were extracted and organized into structured sequence data. With the context memory capability of LSTM-based timing modeling, the complex timing dependencies between cells at different levels in the path were captured, achieving high-precision prediction of path delay. Experimental results showed that, compared to existing machine learning-based estimation methods that accumulate cell delays and wire delays, the LSTM-based prediction method demonstrated better adaptability to different types of cases while maintaining accuracy. In terms of running speed, a speedup of 2.8 to 3.2 times was achieved in most test cases. The prediction method was also validated on generic netlists without technology information and the performance was superior to traditional static timing analysis methods, demonstrating its effectiveness and potential for early-stage design applications.

Key words: long short-term memory, logic synthesis, static timing analysis, path delay predicting, machine learning

中图分类号:

TP319

王庆康,周冉冉,王永. 基于LSTM的逻辑综合阶段延时预测方法[J]. 山东大学学报 (工学版), 2026, 56(2): 43-51.

WANG Qingkang, ZHOU Ranran, WANG Yong. A logic synthesis delay predicting method based on LSTM[J]. Journal of Shandong University(Engineering Science), 2026, 56(2): 43-51.

参考文献

[1] 贺旭, 王耀, 傅智勇, 等. 敏捷设计中基于机器学习的静态时序分析方法综述[J]. 计算机辅助设计与图形学学报, 2023, 35(4): 640-652. HE Xu, WANG Yao, FU Zhiyong, et al. A survey on machine learning-based technology for static timing analysis in agile design[J]. Journal of Computer-Aided Design & Computer Graphics, 2023, 35(4): 640-652.
[2] ACHARYAL C, SHARMA A, MISHRA N, et al. Switching activity factor-based ECSM characterization(SAFE): a novel technique for aging-aware static timing analysis[J]. IEEE Transactions on Computer-Aided Design of Integrated Circuits and Systems, 2024, 43(12): 4715-4725.
[3] 郭静静, 宗璟宜, 查佩文, 等. 基于机器学习的多输入切换效应的统计静态时序分析方法[J]. 微电子学, 2024, 54(3): 458-467. GUO Jingjing, ZONG Jingyi, ZHA Peiwen, et al. Machine learning-based method for statistical static timing analysis of multiple input switching effects[J]. Microelec-tronics, 2024, 54(3): 458-467.
[4] SHONIKER M, OLEYNIKOV O, COCKBURN B F, et al. Automatic selection of process corner simulations for faster design verification[J]. IEEE Transactions on Computer-Aided Design of Integrated Circuits and Systems, 2018, 37(6): 1312-1316.
[5] KAHNG A B, LIN B, NATH S. Enhanced metamodeling techniques for high-dimensional IC design estimation problems[C] //2013 Design, Automation & Test in Europe Conference & Exhibition(DATE). Grenoble, France: IEEE, 2013: 1861-1866.
[6] CHENG H H, JIANG I H, OU O. Fast and accurate wire timing estimation on tree and non-tree net structures[C] //2020 57th ACM/IEEE Design Automation Conference(DAC). San Francisco, USA: IEEE, 2020: 9218712.
[7] KAHNG A B, LUO M L, NATH S. SI for free: machine learning of interconnect coupling delay and transition effects[C] //2015 ACM/IEEE International Workshop on System Level Interconnect Prediction(SLIP). San Francisco, USA: IEEE, 2015: 7171706.
[8] KAHNG A B, MALLAPPA U, SAUL L. Using machine learning to predict path-based slack from graph-based timing analysis[C] //2018 IEEE 36th International Con-ference on Computer Design(ICCD). Orlando, USA: IEEE, 2018: 603-612.
[9] ZHAO X T, WANG T J, JIAO R, et al. Standard cells do matter: uncovering hidden connections for high-quality macro placement[C] //2024 Design, Automation & Test in Europe Conference & Exhibition(DATE). Valencia, Spain: IEEE, 2024: 25-27.
[10] BARBOZA E C, SHUKLA N, CHEN Y R, et al. Machine learning-based pre-routing timing prediction with reduced pessimism[C] //Proceedings of the 56th Annual Design Automation Conference 2019. Las Vegas, USA: ACM, 2019: 106.
[11] CAO P, HE G Q, YANG T. TF-predictor: transformer-based prerouting path delay prediction framework[J]. IEEE Transactions on Computer-Aided Design of Integrated Circuits and Systems, 2023, 42(7): 2227-2237.
[12] HE G Q, DING W J, YE Y Y, et al. An optimization-aware pre-routing timing prediction framework based on heterogeneous graph learning[C] //2024 29th Asia and South Pacific Design Automation Conference(ASP-DAC). Incheon: IEEE, 2024: 177-182.
[13] GUO Z Z, LIU M J, GU J Q, et al. A timing engine inspired graph neural network model for pre-routing slack prediction[C] //Proceedings of the 59th ACM/IEEE Design Automation Conference. San Francisco, USA: ACM, 2022: 1207-1212.
[14] KAHNG A B, MANTIK S. A system for automatic recording and prediction of design quality metrics[C] //Proceedings of the IEEE 2001. 2nd International Symposium on Quality Electronic Design. San Jose, USA: IEEE, 2001: 81-86.
[15] KAHNG A B, LIN B, NATH S. High-dimensional metamodeling for prediction of clock tree synthesis outcomes[C] //2013 ACM/IEEE International Work-shop on System Level Interconnect Prediction(SLIP). Austin, USA: IEEE, 2013: 6681685.
[16] LU Y C, LEE J, AGNESINA A, et al. GAN-CTS: a generative adversarial framework for clock tree prediction and optimization[C] //2019 IEEE/ACM International Conference on Computer-Aided Design(ICCAD). Westminster, USA: IEEE, 2019: 8942063.
[17] 张雷, 刘伟景, 赵启林. 一种高效实现时序优化的时钟树综合方案[J]. 上海电力大学学报, 2025, 41(1): 85-89. ZHANG Lei, LIU Weijing, ZHAO Qilin. A clock tree synthesis method for efficient timing optimization[J]. Journal of Shanghai University of Electric Power, 2025, 41(1): 85-89.
[18] KAHNG A B, KANG S, LEE H, et al. Learning-based approximation of interconnect delay and slew in signoff timing tools[C] //2013 ACM/IEEE International Work-shop on System Level Interconnect Prediction(SLIP). Austin, USA: IEEE, 2013: 6681682.
[19] HAN S S, KAHNG A B, NATH S, et al. A deep learning methodology to proliferate golden signoff timing[C] //2014 Design, Automation & Test in Europe Conference & Exhibition(DATE). Dresden, Germany: IEEE, 2014: 273.
[20] SONG Q Q, CHENG X, CAO P. Critical paths prediction under multiple corners based on BiLSTM network[C] //2023 60th ACM/IEEE Proceedings of Design Automation Conference(DAC). San Francisco, USA: IEEE, 2023: 10247984.
[21] NETO W L, MOREIRA M T, AMARU L, et al. Read your circuit: leveraging word embedding to guide logic optimization[C] //Proceedings of the 26th Asia and South Pacific Design Automation Conference. Tokyo, Japan: ACM, 2021: 530-535.
[22] BODAPATI S, NAJM F N. Prelayout estimation of individual wire lengths[J]. IEEE Transactions on Very Large Scale Integration(VLSI)Systems, 2001, 9(6): 943-958.
[23] FATHI B, BEHJAT L, RAKAI L M. Apre-placement net length estimation technique for mixed-size circuits[C] //Proceedings of International the 11th Workshop on System Level Interconnect Prediction. San Francisco, USA: ACM, 2009: 45-52.
[24] LIU Q, MA J G, ZHANG Q J. Neural network based pre-placement wirelength estimation[C] //2012 Inter-national Conference on Field-Programmable Technology. Seoul: IEEE, 2012: 16-22.
[25] HU B, MAREK-SADOWSKA M. Wire length prediction based clustering and its application in placement[C] //Proceedings of Design Automation Conference. Anaheim, USA: IEEE, 2003: 800-805.
[26] KAHNG A B, REDA S. Intrinsic shortest path length: a new, accurate a priori wirelength estimator[C] //ICCAD-2005. IEEE/ACM International Conference on Computer-Aided Design. San Jose, USA: IEEE, 2005: 173-180.
[27] LIU Q H, MAREK-SADOWSKA M. Pre-layout wire length and congestion estimation[C] //Proceedings of the 41st annual Design Automation Conference. San Diego, USA: ACM, 2004: 582-587.
[28] HYUN D, FAN Y P, SHIN Y. Accurate wirelength prediction for placement-aware synthesis through machine learning[C] //2019 Design, Automation & Test in Europe Conference & Exhibition(DATE). Florence, Italy: IEEE, 2019: 324-327.
[29] 蒋政涛, 贺旭, 李琼, 等. 基于随机森林的网表级时序预测模型[J]. 微电子学与计算机, 2022, 39(12): 107-114. JIANG Zhengtao, HE Xu, LI Qiong, et al. Random forest-based netlist-level timing prediction[J]. Micro-electronics & Computer, 2022, 39(12): 107-114.
[30] XIE Z Y, LIANG R J, XU X Q, et al. Preplacement net length and timing estimation by customized graph neural network[J]. IEEE Transactions on Computer-Aided Design of Integrated Circuits and Systems, 2022, 41(11): 4667-4680.
[31] ZHU L Y, GUO X F. Delay-driven physically-aware logic synthesis with informed search[C] //2023 IEEE 41st International Conference on Computer Design(ICCD). Washington, DC, USA: IEEE, 2023: 327-335.
[32] HOCHREITER S, SCHMIDHUBER J. Long short-term memory[J]. Neural Computation, 1997, 9(8): 1735-1780.

多维度评价

Viewed

Full text

Abstract

Cited

Shared

Discussed