Exhaustive Data Path Optimization in High-Level Synthesis through Area Improvement

The stringent performance and area constraints and short time to market of modern digital systems drive us towards automated methods for producing high speed and low area architecture with optimum features. In this paper, we present a new algorithm, which automatically maximizes resource optimization of data path while meeting performance constraints. The main input of this algorithm is the control data flow graph (CDFG) of the application extracted from C code, while its output is defined as a set of resource instances along with their connections. An initial data path is extracted from the application source code and is improved until an optimized data path is obtained. The experimental results show that not only the refined data path outperforms data paths generated by some other tools, but also presents lower cost, less overhead and can be generated in less time.