X-architecture Steiner Tree Algorithm with Limited Routing Length inside Obstacle

Abstract

Steiner minimal tree construction is a key step in the physical design of Very Large Scale Integration (VLSI). Further considering X-architecture with better wirelength optimization and allowing wires to pass through obstacles to a certain extent before signal distortion, a novel X-architecture Steiner Minimal Tree with Limited Routing Length inside Obstacle (XSMT-LRLO) problem is formed. Therefore, the XSMT-LRLO based on Discrete Particle Swarm Optimization algorithm (XSMT-LRLO-DPSO) is proposed. Firstly, in order to significantly reduce the times of evaluations, a preprocessing strategy based on a lookup table is proposed. Secondly, XSMT-LRLO-DPSO is effectively en-coded by adopting the edge-point pairs encoding method adapted to an evolutionary iterative process. Then, aiming at the XSMT-LRLO problem, which is a discrete problem, a discrete update strategy based on mutation operation and crossover operation is proposed. Finally, adjustment and refinement strategies are introduced to respectively improve the obstacles bypassing ability and wirelength optimization ability of the proposed algorithm. The experimental results show that the proposed algorithm makes full use of the routing resources within the obstacles, and effectively saves routing resources. Compared with similar algorithms, the proposed algorithm has the strongest wirelength optimization ability.