New library development method by FSM based cell pattern extraction

Abstract

The performance, power and area optimization with respect to an implementation cost is a fundamental problem in digital circuit design. With a custom design approach, finding patterns of frequent cell combinations can inspire the new cell development and optimization. A FSM (Frequent Subgraph Mining) method can help to develop new cell libraries, however, this requires huge engineering efforts for finding feasible operation conditions. This paper presents an optimized frequent subgraph mining platform by integrating various FSM methods. The experiment results with various designs demonstrate that the proposed method can reduce an overall runtime.