Interconnect tree optimization algorithm in nanometer very large scale integration designs

Abstract

This thesis proposes a graph-based maze routing and buffer insertion algorithm for nanometer Very Large Scale Integration (VLSI) layout designs. The algorithm is called Hybrid Routing Tree and Buffer insertion with Look-Ahead (HRTB-LA). In recent VLSI designs, interconnect delay becomes a dominant factor compared to gate delay. The well-known technique to minimize the interconnect delay is by inserting buffers along the interconnect wires. In conventional buffer insertion algorithms, the buffers are inserted on the fixed routing paths. However, in a modern design, there are macro blocks that prohibit any buffer insertion in their respective area. Most of the conventional buffer insertion algorithms do not consider these obstacles. In the presence of buffer obstacles, post routing algorithm may produce poor solution. On the other hand, simultaneous routing and buffer insertion algorithm offers a better solution, but it was proven to be NP-complete. Besides timing performance, power dissipation of the inserted buffers is another metric that needs to be optimized. Research has shown that power dissipation overhead due to buffer insertions is significantly high. In other words, interconnect delay and power dissipation move in opposite directions. Although many methodologies to optimize timing performance with power constraint have been proposed, no algorithm is based on grid graph technique. Hence, the main contribution of this thesis is an efficient algorithm using a hybrid approach for multi-constraint optimization in multi-terminal nets. The algorithm uses dynamic programming to compute the interconnect delay and power dissipation of the inserted buffers incrementally, while an effective runtime is achieved with the aid of novel look-ahead and graph pruning schemes. Experimental results prove that HRTB-LA is able to handle multi-constraint optimizations and produces up to 47% better solution compared to a post routing buffer insertion algorithm in comparable runtime.