| RSoC is based on the Dynamic Reconfigurable technology which can change part of its function while running. So the system based on RSoC can switch the tasks' implementation, that is to say, the task running in the hardware may be switched to the software. So the dynamic hardware software partitioning is required, and it requires a balance of performance and quality. In order to address the problem of the dynamic HW/SW partitioning, this thesis does the following work.At first, an adaptive HW/SW partitioning algorithm is proposed. The design idea is that at first we should quantify the complexity of the specific partitioning problem, and then the parameters of the algorithm are adjusted according to the quantitative values. The experiments show that the adaptive adjustment technology can enhance the performance of the partitioning algorithm.Then, a new evaluation function based on the priority is proposed according to the features of the HW/SW partitioning. In the filed of HW/SW partitioning, the aim of the evaluation function is to evaluate the result of the partitioning. Experiments show that this evaluation function can get a relative fair evaluation result.At last, a new HW/SW partitioning algorithm (SSS-DPSO) is proposed. The Discrete Particle Swarm (DPSO) can be implemented easily and has the advantage of high local searching ability. However it will be trapped in the local optimum points because of its poor global searching ability. The situation can be improved by combining the progressive Search Space Smoothing technique (SSS). SSS can create a good searching environment for the DPSO by smoothing the search space of the partitioning problem. One of the difficulties of applying the SSS is how to smooth the problem's search space. In this thesis, we realize a high-quality smoothing method based on reducing the differences between every two tasks. Experiments show that the proposed algorithm can improve the quality and performance of the HW/SW partitioning. |
PDF Full Text Request