(3.231.29.122) 您好!臺灣時間:2021/02/25 21:17
字體大小: 字級放大   字級縮小   預設字形  
回查詢結果

詳目顯示:::

我願授權國圖
: 
twitterline
研究生:郭育吟
研究生(外文):Yu-Yin Kuo
論文名稱:應用機器學習方法在可繞度導向之巨集電路擺置
論文名稱(外文):Routability-driven Macro Placement with Machine-Learning Technique
指導教授:鄭維凱鄭維凱引用關係
指導教授(外文):Wei-Kai Cheng
學位類別:碩士
校院名稱:中原大學
系所名稱:資訊工程研究所
學門:工程學門
學類:電資工程學類
論文種類:學術論文
論文出版年:2017
畢業學年度:105
語文別:中文
論文頁數:64
中文關鍵詞:機器學習巨集擺放半周線長繞線壅塞
外文關鍵詞:machine learningmacro placementHPWLrouting congestion
相關次數:
  • 被引用被引用:0
  • 點閱點閱:510
  • 評分評分:系統版面圖檔系統版面圖檔系統版面圖檔系統版面圖檔系統版面圖檔
  • 下載下載:0
  • 收藏至我的研究室書目清單書目收藏:0
現今的晶片製程進步,使晶片面積縮小且更省電,而各元件擺放位置及繞線路徑對晶片效能有直接的影響,其中巨集與標準元件的擺放佈局規劃扮演重要的角色。在超大型積體電路的實體設計中,巨集擺置為平面規劃的第一步驟,因此巨集擺置結果的優劣會影響整個平面規劃,在傳統上並不會刻意對巨集擺置的結果做評估,必須要到標準元件擺置及繞線完成後才做整體評估,這樣的方式造成若評估結果不好,會導致要重做平面規劃的整個流程,耗費巨大的時間成本。而在平面規劃中,通常會以半周線長及繞線壅塞程度進行評估。
因此本論文提出以機器學習的技術,事先評估巨集擺置的結果,透過蒐集先前的歷史資料,使用多種不同機器學習演算法進行分析比較,並建立出機器學習模組來預測半周線長及繞線壅塞程度,讓巨集擺置的結果能以有效率的方式進行判斷。我們的實驗結果證明利用機器學習的技術能夠有效地預估半周線長及繞線壅塞程度,提供巨集擺置結果的參考值,讓平面規劃的流程更順暢。
Macro placement is the first and important step in floorplan. Macros'' location directly effects next two steps, standard cell placement and routing. However, it is a time-consuming work to evaluate macro placement''s result.
In this thesis, Our key observation is that the evaluation have to wait until standard cell placement and routing are finished. That''s why we said that evaluate macro placement result is a tedious work. To address this problem, we propose a effective prediction method with machine learning technique. This method predict HPWL and routing congestion after macro placement are finished rather than standard cell placement and routing are finished. It take short time that we can get HPWL and routing congestion. Experiment results show that our prediction is accurate and effective.

目錄:
中文摘要 I
ABSTRACT II
目錄 III
圖目錄 IV
表目錄 V
第一章、 前言 1
第二章、 問題描述 3
2.1 研究動機 3
2.2 問題描述 5
第三章、 相關研究 6
3.1 機器學習相關研究 6
3.1.1 機器學習之資料前處理 6
3.1.2 機器學習之資料集 7
3.1.3 機器學習方法 8
3.2 機器學習之EDA領域應用實例 10
第四章、 程式流程 12
4.1 機器學習流程 12
4.2 巨集擺置器(Macro placer) 13
4.3 機器學習特徵定義 17
4.4 機器學習演算法及特徵選擇 28
第五章、 實驗結果 30
5.1 實驗方法 30
5.2 實驗結果 33
5.2.1 繞線壅塞程度預估(國際積體電路設計競賽Benchmark) 33
5.2.2 半周線長預估(國際積體電路設計競賽Benchmark) 37
5.2.3 繞線壅塞程度預估(國際積體電路設計競賽+ISPD Benchmark) 42
5.2.4 半周線長預估(國際積體電路設計競賽+ISPD Benchmark) 46
5.3 綜合比較 51
第六章、 結論 57
參考文獻 58

圖目錄:
圖1-1: Physical design flow. 1
圖2-1: Floorplan flow. 4
圖3-1: Data transformation flow. 6
圖3-2: Split dataset. 7
圖4-1: Azure machine learning flow. 13
圖4-2: Macro placer principle. 14
圖4-3: Movable inside boundary. 15
圖4-4: Macro placer type. 16
圖4-5: Macro placement area. 18
圖4-6: Total macro area vs. macro placement area. 19
圖4-7: Macro pin connection. 20
圖4-8: Distance form pin to block boundary. 21
圖4-9: Distribution of macro. 22
圖4-11: Distance from center point of macro placement to center point of unused rectangle. 24
圖4-12: Distance between pin and center point of unused rectangle. 25
圖4-13: Distance from center point of macro placement area and center point of macro. 26
圖4-14: 9% of macro placement area. 27

表目錄:
表5-1: RC parameter. 30
表5-2 Predict RC with boosted decision tree regression algorithm(Cad). 33
表5-3 Predict RC with decision forest regression algorithm(Cad). 34
表5-4 Predict RC with linear regression algorithm(Cad). 35
表5-5 Predict RC with Poisson regression algorithm(Cad). 35
表5-6 Predict RC with neural network regression algorithm(Cad). 36
表5-7 Predict HPWL with boosted decision tree regression algorithm(Cad). 37
表5-8 Predict HPWL with decision forest regression algorithm(Cad). 38
表5-9 Predict HPWL with linear regression algorithm(Cad). 39
表5-10 Predict HPWL with Poisson regression algorithm(Cad). 40
表5-11 Predict HPWL with neural network regression algorithm(Cad). 41
表5-12 Predict RC with boosted decision tree regression algorithm(Cad+ISPD). 42
表5-13 Predict RC with decision forest regression algorithm(Cad+ISPD). 43
表5-14 Predict RC with linear regression algorithm(Cad+ISPD). 44
表5-15 Predict RC with Poisson regression algorithm(Cad+ISPD). 44
表5-16 Predict RC with neural network regression algorithm(Cad+ISPD). 45
表5-17 Predict HPWL with boosted decision tree regression algorithm(Cad+ISPD). 46
表5-18 Predict HPWL with decision forest regression algorithm(Cad+ISPD). 47
表5-19 Predict HPWL with linear regression algorithm(Cad+ISPD). 48
表5-20 Predict HPWL with Poisson regression algorithm(Cad+ISPD). 49
表5-21 Predict HPWL with neural network regression algorithm(Cad+ISPD). 50
表5-21 The best result of predictive RC(Cad). 51
表5-22 The best result of predictive HPWL(Cad). 52
表5-23 The best result of predictive RC(Cad+ISPD). 53
表5-24 The best result of predictive HPWL(Cad+ISPD). 54
表5-25 Using decision forest regression algorithm predict RC with different features. 55
表5-26 Using decision forest regression algorithm predict HPWL with different features. 56
[1]Amir H. Salek, Jinan Lou, and Massoud Pedram, “An integrated logical and physical design flow for deep submicron circuits”, IEEE Transactions on Computer-Aided Design of Integrated Circuits and Systems, Volume 18, Issue 9, Sep 1999.
[2]Microsoft corp., “Microsoft Azure Machine Learning studio-Regression”.
[3]Yaser S. Abu-Mostafa, M. Magdon-Ismail, H-T. Lin, “Learning From Data”, 2012.
[4]E. Alpaydin, “Introduction to Machine Learning”, 2004.
[5]Richard S. Sutton and Andrew G. Barto , “Reinforcement Learning:
An Introduction”, 2012
[6]Li-C Wang, Magdy S. Abadir “Data Mining In EDA -Basic Principles, Promises, and Constraints”, 51st ACM/EDAC/IEEE Design Automation Conference (DAC), pp.1-6, 2014.
[7]W-H. Chang, C-H. Lin, S-P. Mu, L-D. Chen, C-H. Tsai, Y-C. Chiu, and M. C-T. Chao, “Generating Routing-Driven Power Distribution Networks with Machine-Learning Technique”, IEEE Transactions on Computer-Aided Design of Integrated Circuits and Systems, vol.36, no.8, pp.1237-1250, 2017.
[8]J. Z. Yan, N. Viswanathan and C. Chu, “Handling Complexities in Modern Large-Scale Mixed-Size Placement”, 46th ACM/IEEE Design Automation Conference, 2009.
[9] T-C. Chen, P-H. Yuh and Y-W. Chang “MP-Trees: A Packing-Based Macro Placement Algorithm for Modern Mixed-Size Designs”, IEEE Transactions on Computer-Aided Design of Integrated Circuits and Systems, vol.27, no.9, pp.1621-1634, 2008.
[10] S. N. Adya and I. L. Markov, “Consistent Placement of Macro-Blocks Using Floorplanning and Standard-Cell Placement”, ACM Transactions on Design Automation of Electronic System, vol.10, 2005.
[11]B. Yang, M. Wu, S. Liu, MediaTek Inc., “Routability-driven macro placement”, 2015 Cad contest - Problem D.
[12]ISPD 2006 ACM International Symposium on Physical Design, 2006.
[13]T-C. Chen, Z-W. Jiang, T-C., H-C. Chen and Y-W. Chang, “NTUplace3: An Analytical Placer for Large-Scale Mixed-Size Designs With Preplaced Blocks and Density Constraints”, In Transaction on Computer-Aided Design, vol 27, no. 7, pp. 1228-1240, July 2008.
[14] Yi-F. Chen, C-C. Huang, C-H. Chiou, Y-W. Chang and C-J. Wang, “Routability-Driven Blockage-Aware Macro Placement”, 51st ACM/EDAC/IEEE Design Automation Conference (DAC), pp. 1-6, 2014.
[15] Microsoft corp., “Microsoft Azure Machine Learning studio-Evaluate-model-performance”.
[16] Hamid Shojaei, Azadeh Davoodi, and Jeffrey T. Linderoth, “Congestion analysis for global routing via integer programming”, Computer-Aided Design (ICCAD), IEEE/ACM International Conference,pp. 256- 262,2011.
QRCODE
 
 
 
 
 
                                                                                                                                                                                                                                                                                                                                                                                                               
第一頁 上一頁 下一頁 最後一頁 top
無相關論文
 
無相關期刊
 
無相關點閱論文
 
系統版面圖檔 系統版面圖檔