A Dynamic Hardware-software Partitioning Algorithm Based On Reconfigurable System

The development of programmable logic devices provides the foundation of hardware for the reconfigurable system to tend towards to practicality in deed. Currently, the application of reconfigurable system mainly centralizes in many areas such as the science and technology of national defence, the avigation and spaceflight. In order to improve the popularization of the application about the reconfigurable system, a dynamic hardware-software partitioning method is needed. The dynamic hardware-software partitioning method based on configurable system is not very perfect recently, and even more the hardware-software partitioning method based on function doesn't have a ready system. The paper has put up a research on a model which bases on the hardware-software partitioning method based on function,mainly completed work including:1) bring forward to make use of the integer encapsulation of the hardware partitioning and software partitioning method based on function to construct a dynamic hardware-software partitioning model based on function. The paper analyses the whole realized process of dynamic hardware-software partitioning under the UCOS II operating system concretely. This partitioning process has characteristics of possessing very good hardware-software transparency and computer automatic accomplished completely.2) integrate the partitioning model, confirm the object and target of hardware-software partitioning algorithm, bring forward the strategy of reducing partitioning object to improve the efficiency of partitioning algorithm, and analyse and prove the validity of this method.3) analyse the algorithm in existence , the paper brings forward a sort of integrating Greedy algorithm with other algorithm to resolve the target problems. The experiment indicates that the Greedy algorithm has superduper perform efficiency. However, when the requirement of function proportion is comparatively big, it always daps into the parts and then makes the resource waste very obviously. It is proved that it is very necessary to integrate Greedy algorithm with other algorithms.4) this paper brings forward a concept of FFSD (Fix-Fcope-Scatter-Degree) model, and integrates the thought of the model with Tabu search algorithm, and then it makes use of transforming double f to decrease the production of the unfeasible result. The experiment has proved that Tabu search algorithm possesses very good function proportion requirement adaptability, and its performance time steadies around 0.1ms all the time. Under the circumstance of the similar result quality, compared with the simulation anneal algorithm it has better performance efficiency. Therefore it makes use of integrating this algorithm with Greedy algorithm to resolve the target problems is feasible and effective.