A NEURAL NETWORK APPROACH FOR THE TIMING DRIVEN MACRO CELL PLACEMENT

In this paper, a mean field annealing neural network approach for the timing driven macro cell placement problem is proposed. In the algorithm, a three dimensional permute matrix of binary variables is used to map the problem to the neural network, the energy function including object item, overlap constrained item, and timing constrained item is presented, and then iteration procedure is put into practice with the mean annealing equation. Normalization of neurons proves that only one cell can be assigned to one position in the placement grid. Some practical constraints such as variable block aspect ratio, rotations and reflections, terminal locations, and block shape, can be supported. The algorithm is programmed with Visual C++ language, and experimental result shows that it is an effective method.