An Approach for Dynamic Hardware /Software Partitioning Based on DPBIL

HW/SW partitioning is one of the key steps in high-level modeling of embedded system co-design. In some kinds of data-dependence applications, the partitioning environment is dynamic. An approach based on a kind of evolutionary algorithm called DPBIL is proposed for dynamic HW/SW partitioning problem. DPBIL is inspired by Dualism or complementarity in the nature and operates on a pair of dual probability vectors to adapt itself to the dynamic environment. We systematically constructed the system model, defined the dynamic environment and conducted experimental analysis via comparison with the existing method. To enhance the exploration performance of DPBIL, a threshold constraint mechanism is introduced. Experimental results indicated that our approach is feasible and efficient for dynamic HW/SW partitioning, and has much better performance than traditional ones.