Loading [MathJax]/extensions/TeX/cellcolor_ieee.js
Optimization of Macro Placement Using Genetic Algorithm | IEEE Conference Publication | IEEE Xplore

Optimization of Macro Placement Using Genetic Algorithm

Publisher: IEEE

Abstract:

This thesis presents a novel approach to macro-layout in physical design, combining genetic algorithms with traditional design techniques. Genetic algorithms simulate the...View more

Abstract:

This thesis presents a novel approach to macro-layout in physical design, combining genetic algorithms with traditional design techniques. Genetic algorithms simulate the process of genetic evolution by natural selection, whereby the most environmentally adapted individuals are obtained through the evolution of reproduction from generation to generation. In this paper, the macro layout is represented by individuals, and the characteristics of the macro layout are used to constrain it. In the algorithm presented in this paper, each individual's adaptation action to their environmental context is used to represent the line length. The smaller the line length, the greater the macro layout's conformity to its constraints. Finally, the physical design of different macro layouts is compared with the timing obtained by STA tool. The final conversion time violation of macro layout applying the algorithm of this paper achieved an average reduction of 6.8% in violation and 17% in the number of violation paths compared to manual macro layout. The macro layout has been optimized.
Date of Conference: 10-12 May 2024
Date Added to IEEE Xplore: 09 September 2024
ISBN Information:

ISSN Information:

Publisher: IEEE
Conference Location: Guangzhou, China

I. Introduction

In modern industry, layout methods are typically divided into two stages: macro placement and standard cell placement. In the former, macros are placed in legal locations around the chip boundaries, while in the latter, standard cells are placed in the remaining space after all macros have been fixed[1]. As illustrated in Fig. 1, this process is repeated until the desired layout is achieved. However, the evaluation of layout results is typically conducted manually only after the wiring step, which can lead to multiple iterations in the physical design flow and a reduction in overall productivity. As machine learning AI algorithms continue to gain popularity and are increasingly applied in various fields, there has been a proliferation of machine learning-based algorithms that utilize machine learning to accelerate the physical design process.

Basic macro placement flow

References

References is not available for this document.