GA-based algorithm for hardware/software partitioning with resource contentions

In embedded system design, hardware/software (HW/SW) partitioning is the NP-complete problem which searches reasonable composition of hardware and software components which not only satisfies the constraint of hardware but also optimizes the execution time. Blocks derive from the functional specification internally represented as control/data flow graph. In this paper, the model of the embedded system is extended so that resource contentions are taken into account. A GA-based algorithm is proposed on the basic of the model, in which “serial blocks” are calculated by NAP. The algorithm can lead to a more precise design space exploration. Experimental results show that, comparison with exhaustive search method, the algorithm we proposed can be conveniently implemented HW/SW partitioning with resource contentions.