Congestion driven incremental placement algorithm based on network flow

Congestion is an important problem in VLSI physical design. The paper presents a new incremental placement to reduce local routing congestion with a global view. A minimum-cost flow problem is formulated to balance routing requirement globally, and reduce congestion with minimum deviation from input placement. Besides, a similar network flow method is adopted to remove row overflow and a heuristics, called crucial-part-first, is employed to resolve overlaps. Experimental results show that this algorithm can considerably reduce routing congestion with little wire-length sacrifice of initial placement.