As the fabrication technology advances, the number of transistors in modern computer systems keeps growing exponentially. This growth creates tremendous potential while imposing big challenges to the electronic design automation community on how to construct such complicated systems. In order to effectively utilize the computing resources, a number of fundamental problems need to be addressed. As the scale and complexity of these problems grow, we must look towards new optimizations methods, rather than simply perform iterative improvements on existing techniques.

In this dissertation, I will report our research work on constructing new heuristic algorithms using the Ant Colony metaheuristic for effectively and efficiently solving a range of difficult architectural design problems. We investigate three NP-hard problems in this context, namely system partitioning, operation scheduling and design space exploration. Results show that Ant Colony metaheuristic is a very promising approach for solving these problems, and the algorithms we propose generally achieve better quality solutions with much improved stability compared to conventional methods. Moreover, by establishing the theoretical link between timing and resource constrained scheduling, we propose an effective design exploration framework that leverages duality between the scheduling problems. To our best knowledge, our work is the first extensive study on applying the Ant Colony metaheuristics to the architectural design field.