Exploiting Model-Knowledge in High-Level Synthesis

High-level synthesis tools are gaining more and more acceptance in industrial design flows. While they increase productivity in implementing a single complex hardware module, synthesizing and optimizing many hardware components simultaneously is still an open problem. Here, domain-specific models and specifications are seen as a key ingredient to raise the level of abstraction in future design flows. In this paper, we present a novel model-based synthesis framework which provides for efficient high-level-synthesis of streaming applications modeled as a set of communicating processes. The underlying formal dataflow model of computation enables model-based optimizations like efficient data caching, and naturally exposes parallelism contained in the application which can also be exploited by the proposed synthesis framework. Using a Motion-JPEG decoder as case-study, we will show how this model-based synthesis approach improves the overall quality of generated implementations in terms of performance and resource utilization.