Random characterization of design automation algorithms

Randomly generated Directed Acyclic Graphs (DAGs) can be used to generate various kinds of EDA test data. For example, they can be used to characterize channel routing algorithms. This paper uses such data to characterize the relative performance of a number of different channel routing algorithms, with the aim of determining those factors that have the most effect on routing performance. Our studies show very little difference in the algorithms studied. Factors that have been considered to provide performance improvements are shown to be unimportant, and in some cases even detrimental to average routing performance. This study suggests that "well known" algorithms are not really well known at all, and that more extensive data is needed to characterize the algorithms that we use everyday.