A trade-off Oriented placement tool

High quality placement results are always produced at the cost of significant runtimes. In this paper, we study the trade-off between the overall quality and the runtime for standard-cell placement problems. We implemented and studied a class of schemes to achieve the runtime vs. quality trade-off. We developed a new trade-off oriented placement tool (TOOP) which is controlled by decision trees. TOOP can adjust itself based on user's requests and netlist properties. Compared to Cadence QPlace, even the fastest mode of TOOP (lowest quality) can produce placements with similar or better layout. TOOP also shows much stronger ability to produce routable placement when compared to Capo.