隨著先進製程的演進,光罩費用急遽升高,個別公司單獨負擔產品設計驗證的成本也隨之大幅增加,「多專案晶圓」(Multi-Project Wafer;MPW) 是一個嶄新的服務方案,可以將不同的設計放在同一組光罩裡,進而降低整體製造之成本;然而,在多專案晶圓設計中,佈局規劃與晶圓切割是兩個重要的特性與限制;因此,本研究提出一MPW平面規劃佈局模型,並透過模擬退火演算法搜尋最佳解;本研究考量光罩使用率、專案排列密度優化、最高的晶圓應用與產出最多晶片數量,經由實驗結果顯示優於傳統MPW相關研究,達到改善效能與節省成本之目的。
As the manufacturing technology advances into the deep sub-micron era, the cost of mask has raised. Cost of the product design and verification reflected on the company. Multi-Project Wafer (MPW) is a new service for the IC design house. It puts different dies onto the same set of masks to reduce the production costs. Mask floorplanning and wafer dicing are two important constraints. In this thesis, we proposed a MPW model and use Simulated Annealing algorithm for simultaneous reticle floorplanning and wafer dicing. We will consider area minimization, density optimization, wafer utilization and die maximization, and their combinations. Experimental results have demonstrated that our model and algorithms can improve the mask floorplanning problem with multiple optimization objectives and constraints.