Multiway VLSI circuit partitioning based on dual net representation

J. Cong, W. Juan Labio, N. Shivakumar
1996 IEEE Transactions on Computer-Aided Design of Integrated Circuits and Systems  
In this paper, we study the area-balanced multi-way partitioning problem of VLSI circuits based on a new dual netlist representation named the hybrid dual netlist (HDN), and propose a general paradigm for multi-way circuit partitioning based on dual net transformation. Given a netlist we first compute a K-way partitioning of nets based on the HDN representation, and then transform the K-way net partition into a K-way module partitioning solution. The main contribution of our work is in the
more » ... lation and solution of the K-way module contention (K-MC) problem, which determines the best assignment of the modules in contention to partitions while maintaining user-specified area requirements, when we transform the net partition into a module partition. Under a natural definition of binding function between nets and modules, and preference function between partitions and modules, we show that the K-MC problem can be reduced to a min-cost max-flow problem. We present an efficient solution to the K-MC problem based on network flow computation. We apply our dual transformation paradigm to the well-known K-way FM partitioning algorithm (K-FM) and show that the new algorithm, named K-DualFM, reduces the net cutsize by 20% to 31% compared with the K-FM algorithm. We also apply the same paradigm to the K-MFFC-FM algorithm, a K-FM algorithm based on maximum fanout-free cone (MFFC) clustering reported in [10] , and show that the resulting algorithm, K-DualMFFC-FM reduces the net cutsize by 15% to 26% compared with K-MFFC-FM. Furthermore, we compare the K-DualFM algorithm with EIG1[18] and Paraboli [26], two recently proposed spectral-based bipartitioning algorithms. We showed that K-DualFM reduces the net cutsize by 56% on average when compared with EIG1 and produces comparable results with Paraboli.
doi:10.1109/43.494703 fatcat:mvlckfqmorgnfnn6xzpvqa4xqq