Date of Award

1989

Document Type

Dissertation

Degree Name

Doctor of Philosophy (PhD)

Department

Computer Science

First Advisor

S. Sitharama Iyengar

Abstract

One of the most important forms of routing strategies is called "channel routing". This approach allows us to reduce the extremely difficult VLSI layout problem to a collection of simpler subproblems. For channel routing problems, most frequently mentioned heuristic algorithms use parameters derived from experiments to approach the routing solution without carefully considering the effect of each selected wire segment to the final routing solution. In this dissertation, we propose a new channel routing algorithm in the two-layer restricted-Manhattan routing model (2-RM) in detail. There are three phases involved in developing the new routing algorithm. In the first phase, we distinguish one type of wire from the others using some optimality criteria, which makes the selection of a set of best horizontal wire segments for a track more effective so that good performance of the generated routing solutions can be achieved. In the second phase, we develop a theoretical framework related to two major data structures, column density and vertical constraint graph, which effectively improves search efficiency and routing performance. Finally in the third phase, we develop an efficient powerful heuristic channel routing algorithm based on the concepts shown in phase one and the theoretical framework proposed in phase two. We highlight the application of our algorithm to the channel routing problems in the three-layer restricted-Manhattan overlap (3-RM-O) and three-layer Manhattan overlay (3-M-O) routing models. On many tests we have conducted on the examples known in the literature, our algorithm has performed as well or better than the existing algorithms in both 2-RM and 3-M-O routing models. Our experiments show that our approach has the potential to outperform other algorithms in other routing models.

Pages

161

DOI

10.31390/gradschool_disstheses.4850

Share

COinS