Study On Standard Cell Placement Algorithms Based On Quadratic Objective Function

The development of nanometer technology highly increases the VLSI integration level. It is possible to integrate billions of transistors on a single chip. So we need faster and more effective placement algorithm. A lot of algorithms, like stochastic, partition, and analytical algorithms, have been presented since 1980s. Recent years, algorithms based on quadratic objective function have been attracting interest from researchers for its advantage of handling large-scale circuits. Its disadvantage is the difference between quadratic wire length and half perimeter wire length. Also the initial placement of quadratic algorithm will involve a lot of overlaps. We have to deal with these problems.In this paper, we present a partition algorithm to remove overlaps. This algorithm makes full use of module position information. Compare to iterative method based on module density gradient, our algorithm is faster and more effective. After removing overlaps, we use some optimization method, like GlobalMove, to further reduce wire length. The theoretical analysis and the experimental results on the ISPD04 dataset show that this algorithm can solve the standard cells placement problem effectively.