Performance guided high level algorithm partitioning

In recent times the trend in system engineering goes towards system modeling and analysis at higher abstraction levels. This provides the major advantage of simplifying the procedure of design reuse and simultaneously allows an abstracted possibility of a joint system evaluation in respect of certain design parameters. Traditionally MATLAB/SIMULINK has been chosen for an abstracted examination of the implemented algorithm. This approach lacks a correlation to the actual available hardware components which finally should implement the functionality. To reduce this modeling gap the presented approach chose SystemC as modeling language allowing a closer consideration of the underlying platform during the system/algorithm modeling process. This simulation model is following expanded by performance parameters provided from an Instruction Set Simulator and a High Level Synthesis tool, respectively. By utilizing this additional information, metrics for an optimum algorithm partitioning are deduced. In the second part of the paper, a practical example is presented and exemplarily implemented to show its suitability in such a high level design process.