Hardware/software co-design has been an area of research for a few decades. Currently co-design is utilized to create hardware coprocessors for compute intensive tasks of a system (which otherwise, performed in software, will not meet the performance goals). Design of correct hardware coprocessors with area, timing and power constraints is a time consuming task. In this paper, we present a methodology to alleviate this problem up to a certain extent. First, we show how to adopt a high-level synthesis tool in design space exploration to converge towards efficient hardware coprocessors. Second, we show, through a series of case studies that, a system-level approach, keeping platform specific optimizations in mind, can help in doing such an exploration efficiently.