An ILP Formulation for Minimizing the Number of Feedthrough Cells in a Standard Cell Placement

Jin-Tai Yan

doi:10.1155/1999/59138

Abstract

It is well known that standard cells have been widely used to implement VLSI circuits in the automation of physical design. Since one major aim of a cell-based design is to minimize total layout area in a standard cell placement, the number of feedthrough cells will be minimized to reduce total cell area in a standard cell placement. In this paper, first, we model a partitioning-based row assignment (PRA) problem to minimize the number of feedthrough cells in a standard cell placement. Furthermore, an integer linear programming (ILP) approach is proposed to solve the PRA problem in a standard cell placement. Finally, the ILP approach has been implemented and two standard-cell netlists, Primary 1 and Primary 2, have been tested by the proposed approach, Bose's approach [4] and an exhaustive search approach, respectively. The experimental results show that the ILP approach obtains fewer feedthrough cells than Bose's approach in a partitioning-based standard cell placement.

Highlights

It is well known that standard cells have been widely used to implement VLSI circuits in the automation of physical design
First, we model a partitioningbased row assignment (PRA) problem to minimize the number of feedthrough cells in a standard cell placement
Before we model a partitioning-based row assignment (PRA) problem in a standard cell placement, some important assumptions will be described as follows: First, assume that each feedthrough cell only contains one vertical net segment crossing one cell row in a standard cell placement

Summary

INTRODUCTION

It is well known that standard cells have been widely used to implement VLSI circuits in the automation of physical design. The minimization of the number of feedthrough cells in a standard cell placement is not considered in the row assignment step for a linear-assignmentbased approach. Bose and Saha [4] proposed a combined partitioning-based row assignment to minimize the number of feedthrough cells in a standard cell placement. According to the definition of a partitioningbased approach, we model a new partitioningbased approach for a standard cell placement This approach is divided into three phases: multi-way netlist partitioning, row assignment (inter-row placement) and multi-row linear placement (intra-row placement). K netlists are vertically linearized to minimize the number of feedthrough cells for a standard cell placement. First, we model a partitioningbased row assignment (PRA) problem to minimize the number of feedthrough cells in a standard cell placement. The experimental results show that the ILP approach obtains fewer feedthrough cells than Bose’s approach in a partitioning-based standard cell placement

PROBLEM DESCRIPTION AND DEFINITIONS

EXPERIMENTAL RESULTS

CONCLUSION