Computational Intelligence For Vlsi Layout

As the kernel and foundation of the Information Technology (IT), Integrate Circuit (1C) has been developing rapidly featuring larger and larger integration scale, more miniaturization, higher performance and reliability. The higher the complexity of the 1C design is, the higher the performance requirement of the EDA tools is. Physical design is a complex key link of the VLSI design flow. And the placement is the most important step in the link. Most placement problems are NP complete, which can only be solved by some heuristic algorithms. As the rapid increasing of the VLSI integration scale, it becomes urgent to find effective algorithms for solving placement problems, aiming at improving the placement quality and shortening the computation time.The main work of this dissertation is lies on applying intelligent algorithms to solve the VLSI placement problems. After a brief review of current placement algorithms, we introduce two algorithms, which are Tabu Search (TS) algorithm and Fuzzy Tabu Search (FTS) algorithm, to solve the VLSI placement problems.TS is a flexible framework of a variety of strategies originating from artificial intelligence and is therefore open to further improvement. It has been shown that TS is superior to many other algorithms both in the time required to obtain a solution and in the quality of the solution of many optimization problems. It becomes more and more popular in optimization algorithms. We develop a TS approach for solving the VLSI placement problems in this dissertation. The computer simulation shows that TS is better than genetic algorithm in the both speed and quality of placement.However, the performance of TS strongly depends on the key parameters, which guide the search in the TS procedure. They are usually determined by experience and always keep constant during the entire search procedure. This approach would not be efficient in many cases. We develop an FTS which use a fuzzy system to determine some parameters needed in the TS. Benefits of the methodology are illustrated by the numerical results.