| dc.description.abstract | The HW/SW co-design philosophy introduced almost two decades ago provides solutions to common challenges in the electronic system design industry. It provides techniques for improving the efficiency of conventional design processes for electronic systems, by introducing methods for simultaneous design of hardware and software, as well as techniques for analyzing and finding correct implementation alternatives for selected system functions. This thesis describes the selection of a HW/SW co-design methodology for a software-based Pedestrian Detection System, as well as results from applying the methodology on the system itself.
In this design methodology, the system functions are evaluated from pre-defined design constraints for performance, energy consumption, excessive hardware and software parts, and system reliability. The functions not meeting the constraints are further selected for partitioning into either hardware or software architectures that result in potentially new implementation suggestions for the functions selected. The evaluation has been performed with evaluation criteria in the form of a cost function based on the constrained factors, and the results from applying the methodology on the system have been described. The data used for the evaluation criteria were counts of instructions executed, data reads/writes, as well as cache hit/miss counts of these. These counts are run time attributes of the Pedestrian Detection System's system functions, and are meant to indicate run time and performance demand on the system. The result of the evaluation indicated that no new implementation alternatives were needed. | |
