A model-based extensible framework for efficient application design using FPGA

Abstract

For an FPGA designer, several choices are available in terms of target FPGA devices, IP-cores, algorithms, synthesis options, runtime reconfiguration, degrees of parallelism, among others, while implementing a design. Evaluation of design alternatives in the early stages of the design cycle is important because the choices made can have a critical impact on the performance of the final design. However, a large number of alternatives not only results in a large number of designs, but also makes it a hard problem to efficiently manage, simulate, and evaluate them. In this article, we present a framework for FPGA-based application design that addresses the aforementioned issues. This framework supports a hierarchical modeling approach that integrates application and device modeling techniques and allows development of a library of models for design reuse. The framework integrates a high-level performance estimator for rapid estimation of the latency, area, and energy of the designs. In addition, a design space exploration tool allows efficient evaluation of candidate designs against the given performance requirements. The framework also supports extension through integration of widely used tools for FPGA-based design while presenting a unified environment for different target FPGAs. We demonstrate our framework through the modeling and performance estimation of a signal processing kernel and the design of end-to-end applications.