| This thesis presents fundamental algorithms for the Regularity Advisor, a regularity oriented knowledge base and watchdog that empowers integrated circuit designers and design tools with information about the impact of their decisions on regularity of the final product.; One of the primary reasons for the tremendous growth rate of semiconductor industry was the ability to accurately predict performance and manufacturability of the final product directly from the drawn layout. However, as dimensions of the allowed layout features kept shrinking, all the intricate physical phenomena involved in the manufacturing process made both performance and manufacturability difficult to predict and expensive to verify. Standard design and optimization techniques, including design rule checks and reticle enhancement techniques, became very inefficient. Restricting allowed layout configurations and increasing regularity of layout masks has been considered as the best approach to alleviate the above problems. Such an regularity oriented approach requires a methodology to evaluate regularity and its impact on final designs and design flows.; This thesis presents such a methodology. We introduce a set of data structures and algorithms for extracting and evaluating regularity of layouts, discovering correlations between regularity and various design decisions and layout attributes, comparing regularity-related effectiveness of different design flows. We also propose several novel regularity-guided design algorithms that exploit existence of regularity to optimize various design tasks.; We demonstrate example implementations of pattern representation and extraction methods. Based on those implementations we analyze regularity and applicability of the above methods to existing state-of-the-art IC designs. We also show potential benefits of regularity oriented approaches based upon an example from recently introduced FabBrix design flow. Finally, we compare various types of designs (memories, ASIC, SOC, etc.) and design flows based on which we conclude that current methodologies yield designs that are extremely irregular and expensive to verify and optimize and a new design methodologies are required to circumvent complexities in manufacturing process. |
