Post-Routing Analytical Models for Homogeneous FPGA Architectures

Abstract

The rapid growth in Field Programmable Gate Array (FPGA) architecture design space has led to an explosion in architectural choices that exceed well over 1,000,000 configurations. This makes searching for pareto-optimal solutions using a CAD-based incremental design process near impossible for hardware architects and application engineers. Designers need fast and accurate analytical models in order to evaluate the impact of their design choices on performance. Despite the proliferation of FPGA models, todays state-of the art modeling tools suffer from two drawbacks. First, they rely on circuit characteristics extracted from various stages of the FPGA CAD flow making them CAD dependent. Second, they lack ability to take routing architecture parameters into account. These two factors pose as a barrier for converging to the desired implementation rapidly. In this research, we address these two challenges and propose the first static power and post-routing wirelength models in academia. Our models are unique as they are CAD-independent, and they take both logic and routing architecture parameters into account. Using the static power model we are able to estimate the active and idle leakage power dissipation in homogeneous FPGAs with average correlation factor of 95% and mean percentage error of 17% over experimental results based on MCNC benchmarks. Using our wirelength model, we are able to obtain a low mean percentage error of 4.2% and an average correlation factor of 84% using MCNC and VTR benchmarks. We also show that utilizing wirelength model for architecture optimization process reduces the design space exploration time by 53% compared to the CAD-based process. We finally propose an algorithmic approach to estimate the logic density (i.e., number of LUTs) of multiplexer-based circuits, and address the problem of discrete effects in FPGA analytical models. We show that a model that generates logic density of a fundamental circuit element, such as a multiplexer, can be used to estimate performance metrics, such as critical path delay and power.