Practical Approach to Cell Replacement for Resolving Pin Inaccessibility

We propose a practical approach to the cell replacement problem for resolving the pin inaccessibility in the ECO (engineering-change-order) routing stage. The prior cell replacement method performs in two steps: (i) it prepares a subsidiary (i.e., alternative) cell library that includes for each cell type a set of diverse cell layouts. Then, (ii) it iteratively tries to replace the cells of routing failures with some cells in the subsidiary library during ECO routing in order to fix the routing failures. In this work, we downsize the subsidiary library produced in step (i) to speed up the sequential and time-consuming process of step (ii). Precisely, we propose a function based on the well-known formulation of Levenshtein distance to measure the degree of the pin topology difference between the layout of a cell type in the target library and a layout of the same type in the subsidiary library. Then, we update the subsidiary library to include, for each cell type, exactly one layout that has the biggest pin topology difference. Through experiments with benchmark circuits, it is shown that using the subsidiary library produced by our topology difference formulation enables to reduce the number of trials of cell replacements significantly over the conventional method while fixing almost the same amount of routing violations.