臺灣博碩士論文加值系統

在先進製程的實體設計流程中，繞線是非常複雜且花時間的階段。當繞線完成後，若電路未通過設計規則檢查，則需要進行調整後再重新繞線以產生新的電路，如此反覆繞線會耗費大量的時間在只為了觀察設計規則檢查之結果上，因此使用一個預測方法用以取代實際繞線就能觀察設計規則檢查後的結果是非常重要的。且根據實驗，發現在繞線時巨集元件周圍容易出現違反設計規則，因此如何將巨集元件資訊用以輔助預測也是極為重要的。此外，由於繞線器演算法中的隨機因子，即使是使用同樣的擺置結果做為繞線輸入，每次的繞線結果也可能會有些許不同。因此，該預測方法不只要能快速地得到設計規則檢查結果，還要能大致符合每次的繞線結果。本論文提出一個基於YOLOv3能檢查違反設計規則的預測器，用以根據擺置結果並搭配巨集元件資訊預測違反設計規則的位置及大小，同時考慮繞線器的隨機因子，並適合用以產生全域繞線器之繞線指引。本預測器使用 2015 ISPD contest 之測資並以 Synopsys ICC-II 進行擺置，再使用擺置特徵資料與對應違反設計規則的結果進行訓練及測試。實驗結果顯示，我們的方法所預測出在巨集元件周圍的結果與實際的結果之間的分布狀況很吻合。

In an advanced physical design flow, the routing stage is complex and time-consuming. After the routing stage, the Design Rule Check (DRC) will be performed to check the manufacturability, and the found DRC hotspots (i.e. DRC violations) must be eliminated. However, the DRC hotspot elimination process takes lots of time since any modification on the design must iteratively re-perform routing. Therefore, it's crucial to develop a predictor which can predict DRC results without performing routing. Empirically, DRC hotspots usually happen around macros and it is also important to take the macro information into consideration. Besides, the routing results with the same placement may be different due to the non-deterministic behavior a router could have. Therefore, the DRC hotspot predictor needs to predict all the possible DRC hotspots generated by different routing results. We propose a DRC hotspot predictor from a given placement based on YOLOv3. We use the 2015 ISPD contest benchmarks and Synopsys ICC-II to collect training and testing data. Our experimental results show that the distributions around macros of the predicted DRC hotspots and the ground truth DRC hotspots are well-matched.

1 Introduction 1
1.1 Motivation 1
1.2 Previous Work 2
1.3 Our Contribution 3
2 Preliminaries 4
2.1 Object Detection 4
2.2 YOLOv3 4
2.2.1 Anchor 6
2.2.2 Loss Function 6
2.3 GR Routing Guide 8
2.4 Problem Description 8
3 Methodology 9
3.1 Overall Flow 9
3.2 Feature Extraction 9
3.2.1 Placer­generated Features 10
3.2.2 Horizontal/vertical macro­aware Rudy Features 11
3.2.3 Randomly Generated Macro­aware GR Routing Guides 13
3.2.4 Image Generation 15
3.3 Labeled Bounding Box Generation 15
3.4 Data Augmentation 16
3.5 Model Training 16
4 Experimental Results 17
4.1 Experiment Setup 17
4.2 Results 19
5 Conclusion 30
References 31

[1] S. Batterywala, N. Shenoy, W. Nicholls, and H. Zhou, “Track assignment: A desirable intermediate step between global routing and detailed routing,” in Proceedings of International Conference on Computer­Aided Design, pp. 59–66, 2002.
[2] M. B. Alawieh, W. Li, Y. Lin, L. Singhal, M. A. Iyer, and D. Z. Pan, “High­definition routing congestion prediction for large­scale fpgas,” in Proceedings of Asia and South Pacific Design Automation Conference, pp. 26–31, 2020.
[3] J. Chen, J. Kuang, G. Zhao, D. J.­H. Huang, and E. F. Young, “Pros: A plug­in for routability optimization applied in the state­of­the­art commercial eda tool using deep learning,” in Proceedings of International Conference On Computer Aided Design, 2020.
[4] W.­T. Hung, J.­Y. Huang, Y.­C. Chou, C.­H. Tsai, and M. Chao, “Transforming global routing report into drc violation map with convolutional neural network,” in Proceedings of International Symposium on Physical Design, pp. 57–64, 2020.
[5] W. Zeng, A. Davoodi, and R. O. Topaloglu, “Explainable drc hotspot prediction with random forest and shap tree explainer,” in Proceedings of Design, Automation and Test in Europe, pp. 1151–1156, 2020.
[6] R. Liang, H. Xiang, D. Pandey, L. Reddy, S. Ramji, G.­J. Nam, and J. Hu, “Drc hotspot prediction at sub­10nm process nodes using customized convolutional network,” in Proceedings of International Symposium on Physical Design, pp. 135–142, 2020.
[7] Z. Xie, Y.­H. Huang, G.­Q. Fang, H. Ren, S.­Y. Fang, Y. Chen, and J. Hu, “Routenet: Routability prediction for mixed­size designs using convolutional neural network,” in Proceedings of International Conference on Computer­Aided Design, 2018.
[8] W.­H. Liu, C.­K. Koh, and Y.­L. Li, “Case study for placement solutions in ispd11 and dac12 routability­driven placement contests,” in Proceedings of International Symposium on Physical Design, pp. 114–119, 2013.
[9] J. Long, E. Shelhamer, and T. Darrell, “Fully convolutional networks for semantic segmentation,” in Proceedings of Conference on Computer Vision and Pattern Recognition, pp. 3431–3440, 2015.
[10] J. Redmon and A. Farhadi, “Yolov3: An incremental improvement,” CoRR, vol. abs/1804.02767, 2018.
[11] T.­Y. Lin, P. Dollár, R. Girshick, K. He, B. Hariharan, and S. Belongie, “Feature pyramid networks for object detection,” in Proceedings of Conference on Computer Vision and Pattern Recognition, pp. 936–944, 2017.
[12] S. H. Rezatofighi, N. Tsoi, J. Gwak, A. Sadeghian, I. Reid, and S. Savarese, “Generalized intersection over union: A metric and a loss for bounding box regression,” Proceedings of Conference on Computer Vision and Pattern Recognition, pp. 658–666, 2019.
[13] T.­Y. Lin, P. Goyal, R. Girshick, K. He, and P. Dollár, “Focal loss for dense object detection,” in Proceedings of International Conference on Computer Vision, pp. 2999–3007, 2017.
[14] P. Spindler and F. M. Johannes, “Fast and accurate routing demand estimation for efficient routability­driven placement,” in Proceedings of Design, Automation and Test in Europe, pp. 1–6, 2007.
[15] A. Caldwell, A. Kahng, S. Mantik, I. Markov, and A. Zelikovsky, “On wirelength estimations for row­based placement,” IEEE Transactions on Computer­Aided Design of Integrated Circuits and Systems, vol. 18, no. 9, pp. 1265–1278, 1999.