Simultaneous Multi-die Floorplanning and Technology Assignment

TL;DR

This paper introduces a novel approach to multi-die floorplanning that simultaneously optimizes technology assignment, area, wirelength, performance, power, and cost, using machine learning for rapid PPA estimation.

Contribution

It is the first systematic study to treat technology choice as a variable in multi-die floorplanning, integrating machine learning for rapid estimation and outperforming greedy methods.

Findings

Systematic optimization significantly outperforms greedy approaches.

Incorporates machine learning for rapid PPA estimation.

Validates methods on 2.5D and 3D ICs with commercial tools.

Abstract

In heterogeneous integration, different dies may employ distinct technologies, making floorplanning across multiple dies inherently coupled with technology assignment. By assuming a fixed technology, almost all prior floorplanning studies were developed without addressing the challenge of technology assignment. This work presents the first systematic study of multi-die floorplanning that treats technology choice as a variable. To address the challenge of variable block areas, we incorporate a recent machine learning technique for rapid PPA estimation. Our methods jointly optimize area, wirelength, performance, power, and cost, thereby highlighting the importance of technology assignment. Experimental evaluations, validated with a commercial tool for both 2.5D and 3D ICs, demonstrate that our systematic optimizations significantly outperform a greedy approach.