Cluster-based Placement for multilevel hierarchical FPGA

Abstract

In this paper, we present a multilevel hierarchical FPGA (MFPGA) architecture model and propose a cluster-based placement algorithm for this model. The algorithm has a multi-scale optimized V-shape flow including constructive bottom-up clustering process and top-down placement process. Experimental results indicate that our algorithm improves total wire-length and logic utilization by more than 15% and 10% on average for MCNC benchmark designs compared with the start-of-art vpr tool.