使用在工程變更命令上以機器學習為基的動態電路壓降預測

Machine-learning-based Dynamic IR Drop Prediction for ECO

方彥鈞(Yen-Chun Fang)

指導教授 ： 李建模

國立臺灣大學/電機資訊學院/電子工程學研究所/碩士(2018年)

https://doi.org/10.6342/NTU201803184

摘要

在積體電路設計中，為了確保每一個元件的壓降都不超過一定的限制，往往會需要不斷反覆的實施工程變更命令。這麼做十分浪費時間因為我們需要不斷的在相似的電路設計上做動態電路壓降模擬。在本研究中，我們利用工程變更命令前的電路訓練機器學習模型，並用以預測工程變更命令後電路中的壓降。為了增加預測的準確度，我們提出了27種考慮電路時序、電路功耗和電路實體的特徵，我們的方法是可擴充的因為我們的特徵維度是固定的，不會受到電路大小及元件庫影響。同時，我們提出了使用在電路壓降嚴重區域的區域模型，以此同時改善模型的準確度以及運行時間。我們的實驗結果顯示對於一個500萬元件的電路，預測結果與實際值得相關係數可達0.97，平均絕對誤差可達2.9mV。我們對10萬個元件做預測的動作可以在2分鐘內完成。我們提出的方法提供了一個快速的電路壓降預測，並可加速工程變更命令所需時間。

關鍵字

機器學習 ； 電路壓降 ； 電源供應雜訊

並列摘要

During design signoff, many iterations of Engineer Change Order (ECO) are needed to ensure IR drop of each cell instance meets the specified limit. It is a waste of resources because repeated dynamic IR drop simulations take a very long time on very similar designs. In this work, we train a machine learning model, based on data before ECO, and predict IR drop after ECO. To increase our prediction accuracy, we propose 27 timing-aware, power-aware, and physical-aware features. Our method is scalable because the feature dimension is fixed, independent of design size and cell library. Also, we propose to build regional models for cell instances near IR drop violations to improves both prediction accuracy and training time. Our experiments show that our prediction correlation coefficient is 0.97 and average error is 2.9mV on a 5-million-cell industry design. Our IR drop prediction for 100K cell instances can be completed within 2 minutes. Our proposed method provides a fast IR drop prediction to speedup ECO.

並列關鍵字

machine learning ； IR drop ； power supply noise

參考文獻

[Ahmed 07] N. Ahmed, M. Tehranipoor, and V. Jayaram. "Transition delay fault test pattern generation considering supply voltage noise in a SOC design." Proc. 44th annual Design Automation Conf. ACM, 2007, pp. 533-538.

[Apache 15] Apache RedHawk User Manual, 2015.

[Bastani 08] P. Bastani, K. Killpack, L. -C. Wang, and E. Chiprout "Speedpath prediction based on learning from a small set of examples." Proc. 45th annual Design Automation Conf.. ACM, 2008, pp. 217-222.

[Burd 00] T. D. Burd and R. W. Brodersen, “Design issues for dynamic voltage scaling”, Proc. 2000 int’l. Symp. on Low Power Electronics and Design. ACM, 2000, pp, 9 - 14

[Chen 01] T.-H. Chen, and C. C.-P. Chen, “Efficient large-scale power grid analysis based on preconditioned krylov-subspace iterative methods,” Proc. 38th annual Design Automation Conf.. 2001, pp. 559-562.