考慮保持佈局拓撲原樣的超大型積體電路製程遷移自動化技術

Abstract

製程遷移自動化技術在細胞函式庫與全客制化積木重新設計的議題上伴演著很重要的腳色，人工的佈局重新設計需要耗費大量的人力資源及時間。傳統的製程遷移自動化有著會導致怖局物體扭曲與拓樸關係消滅的缺點，這可能會影響到原本佈局的特徵並使得遷移後的佈局難以被使用者所辨認與驗證。拓樸佈局只記錄物體間的拓樸關係、而忽略了實際的圖形資訊，這個特性增加了在後佈局設計階段做動作的彈性。 這份論文包括兩個部份，首先我們討論一個未成功的工作，為了解決細胞層級製程遷移問題而對三角化編碼圖形所做的改進。接下來我們提出一個新的矩形拓樸佈局模型，並基於這個模型提出了一套細胞層級製程遷移自動化的流程，包括：裝置與繞線擷取、拓墣佈局建立、線性方程式建構、單向壓縮包含掃描線的演算法與線性方程式的解決方法、繞線回復這些步驟。實驗數據顯示我們的演算法很好的保留了細胞佈局的拓樸關係，並且比最近使用線性規劃的製程遷移研究還要快幾倍。

Layout migration plays an important role in the re-design of a cell library and full-custom block. Handcrafted re-design requires a lot of man power and design time. Traditional layout migration suffers the disadvantage of shape distortion of objects and destruction to topology between objects. It may alter the features produced by original layout topology and also makes the migrated layout hard to be recognized and verified by designers. Topological layout only records the topological relation of layout objects and ignores its geometrical information. This property improves the flexibility to perform ECO operations in post-layout stage. This thesis contains two parts. Firstly, we discuss our unsuccessful work on enhancing triangulation encoding graph (TEG) for solving cell-level migration. Secondly, we propose a rectangular topological layout model and a cell-level migration flow based on the proposed topological layout model. The migration flow contains device and wire extraction, topological layout construction, linear equation construction, unidirectional compression combining scan-line algorithm and linear equation solver, and wire restoration. Experimental results show that cell topology is well preserved and a several times runtime speedup is achieved as compared with recent migration research based on ILP formulation.