Dory: 一個簡單但是有效的廣域佈線器

Abstract

傳統超大型積體電路實體設計流程包含了規劃、置放、廣域佈線以及細部佈線。一個快速的廣域佈線器可以幫助置放器更加準確的預估線長以及可繞度。高品質的廣域佈線器也可以幫助細部佈線的可繞度。在這篇論文中，我們發展了一個高效率擁擠導向的全域佈線器，叫做 Dory。可以很快的產生比整數線性規劃基底的廣域佈線器更好的的結果。Dory 是根據 FastRoute 2.0 為基底，加上我們所發展出的歷史基底繞線、改良的繞線流程以及簡化的多重源頭多重目標迷宮繞線。實驗結果也展現出 Dory 可以在很快的時間之內減少許多的擁擠度。

在 IBM98 的測試檔方面，和現有的學術廣域佈線器 FastRoute 2.0 和 BoxRouter 相比，我們可以完成所有的電路而沒有任何的溢出，並且繞線的長度也較短。在 ISPD2007 的測試檔方面，Dory 在線長和擁擠度比 FastRoute 更好，不過花費更多的時間。Dory 比 FGR, MaizeRouter, BoxRouter 快，但是在線長和擁擠度方面比較差

Traditional VLSI physical design flow is composed of floorplanning, placement, global routing and detailed routing. A fast global router can help placer in accurately estimating wire length and routability. A high-quality global router increases routability for detailed routers. In this paper, we develop a high-performance congestion-driven global router, called Dory, to fast produce better routing results as compared to a ILP-based global router. Based on the routing flow of FastRoute 2.0, we develop a history-based routing, an enhanced routing flow and simplified multi-source multi-sink maze routings. Experimental results reveal that Dory decreases many overflows at little cost of runtime.

On IBM98 benchmarks, compared to state-of-art academic global router FastRoute 2.0, and BoxRouter, we can complete all circuits without any overflow with less wirelength. On ISPD2007 benchmarks, Dory is superior to FastRoute in overflow and wire length at the cost of worse runtime, and also superior to FGR, MaizeRouter and BoxRouter in runtime at the cost of worse overflow and wire length.