A New Approach for Task Level Computational Resource Bi-partitioning

An essential problem for hardware/software codesign is the partitioning of an application onto the computational resources. This paper presents a novel approach for the task level resource partitioning problem. Our approach is based on the Ant System algorithm, a meta-heuristic method inspired by the study of the behaviors of ants. In our algorithm, a collection of agents cooperate using distributed and local heuristic information to effectively explore the search space. An iterative stochastic decision making process is carried by the agents in order to optimally allocate tasks onto either the general purpose processor or the reconfigurable logic. Experiments show that the proposed algorithm provides robust results that are qualitatively close to the optimal with minor computational cost.