A modified relay-race algorithm for floorplanning in PCB and IC design

Mert Vatansever,Faik Başkaya

doi:10.3906/elk-1903-122

Abstract

Floorplanning is a fundamental design step in the physical design of printed circuit boards (PCBs) and integrated circuits (ICs), as it handles the complexity of layout design. From a computational point of view, the floorplanning problem is an NP hard problem, and the size of the search space grows exponentially with increasing numbers of modules. Thus, the algorithm used is an essential factor for speed and quality of the floorplanning process. Although polynomial-time floorplanning algorithms can be implemented when solution space is limited to slicing floorplans, optimal solutions often exist only in the nonslicing floorplan search space. Various stochastic algorithms such as simulated annealing (SA), the genetic algorithm (GA), and the relay race algorithm (RRA) can be used with nonslicing floorplans. In this paper, a modified relay race algorithm (MRRA) is proposed. Based on the experimental results utilizing MCNC benchmarks, MRRA improved both solution quality and run time for area optimization when compared with SA, GA, and RRA.

Highlights

The number of components in a circuit and the interconnections between these components increase rapidly as technology improves over time [1]
Various floorplanning algorithms have been proposed by researchers, including simulated annealing (SA), genetic algorithms (GAs), and the relay race algorithm (RRA)
Parameters Nr and Nf are determined according to the detailed analysis, which considers the number of modules in the circuit to improve the efficiency of the algorithm

Summary

Introduction

The number of components in a circuit and the interconnections between these components increase rapidly as technology improves over time [1]. (Γ+, Γ−) = (bacde, cabde) can be the sequence pair representation for one of the solutions of a floorplan that includes the module set a,b,c,d,e. Relay race algorithm Sheng et al proposed the RRA for floorplanning problems to approach a global optimal solution by exploring similar local optimal solutions more efficiently within shorter computation times [7]. Exchange move is the exchange of the order of two modules in both sequences Γ+ and Γ− Both rough search and focusing search follow similar procedures after the move method is selected. After completion of the rough search and focusing search, the relay operator generates a new solution from the current solution.

Move methods

Findings

Conclusion