Floord Planning Based On Module Dimension Pretreatment

With the development of very large scale integrate circulate, more and more modules has been arranged on the chip, and the dimension of semiconductor become smaller and smaller, the Integration of chip is bigger than before, the circulate's Frequency is higher than before, and the floor planning become more and more Complex, then the floor planning of very large scale integrate circulate become more important. In this paper, a new layout method has been proposed based on the Organization of the evolutionary algorithm, witch been Generalized as followed:After we further studied the very large scale integrate floor planning, we proposed the idea named as " clustering according to line for soft module", and we list some define related to floor planning. So we proposed the method for floor planning based on hard module's length's clustering. Before floor planning, we cluster the modules which their length is equal, when arrange the module which is in the same cluster as one unite, we take the area utilization as the goal. Though the arithmetic operators named as training and splitting, we rectify the chip's width and the result of clustering, until we get the satisfying result.After we further studied the soft module's character which area is fixed and shape is not fixed, we proposed the method named as arranging in line for soft module. This method arranges the modules'which area is closed in the same line, and clusters these modules as one class. If some module in one class can not satisfy the restriction of length-width ratio, then we adjust the number of modules in one class, and make all the modules satisfy the restriction of length-width ratio, then we improve the area utilization ratio. According to the method named as clustering in line, this method can calculate it's wide-length ratio precision, and get the best area utilization of one line, then the last area utilization can get the best result.According to the character of loss corner of hard rectangle modules, we proposed the idea of filling in corner. This method look for the lack of each hard rectangle modules, then find each soft modules to fill in these lack, and transform the hard rectangle modules to hard modules. After this step, we cluster each hard module, and arrange these modules as the method mentioned upwards. When arrange, we fill in the gap with the remaining soft modules, so it can improve area utilization.The three method as we mentioned, we tested them with the standard module data aggregate, and we compare this method with moving modal sequences, the result prove that our method is effective.