A Trade-off Oriented Placement Tool

High quality placement results are always produced at the cost ofsignificant runtimes. In this paper, we study the trade-offbetween the overall quality and the runtime for standard-cellplacement problems. We implemented and studied a class ofschemes to achieve the runtime vs. quality trade-off. Wedeveloped a new trade-off oriented placement tool (TOOP)which is controlled by decision trees. TOOP can adjust itselfbased on user's requests and netlist properties. Compared toCadence QPlace, even the fastest mode of TOOP (lowest quality)can produce placements with similar or better layout. TOOP alsoshows much stronger ability to produce routable placement whencompared to Capo.