This talk gives an overview of ongoing work at UIC on so-called `middle-down' placement techniques. In such strategies the target layout area is dissected by a grid resulting in a placement-based partitioning problem. This partitioning problem is solved and the solution refined to ultimately yield a detailed placement. This can be viewed as a generalization of `top-down' strategies in which we begin with a 1x2 or 2x2 disection of the layout area; in our work we typically begin with a much finer grid. Algorithmic techniques for solving the various problems arising from this sort of flow.
Back to Multilevel Optimization in VLSICAD