可重构资源管理及硬件任务布局的算法研究

可重构系统具有微处理器的灵活性和接近于ASIC的计算速度,可重构硬件的动态部分重构能力能够实现计算和重构操作的重叠,使系统能够动态地改变运行任务,可重构资源管理和硬件任务布局方法是提高可重构系统性能的关键.提出了基于任务上边界计算最大空闲矩形的算法(TT-KAMER),能够有效地管理系统的空闲可重构资源;在此基础上使用FF和启发式BF算法进行硬件任务的布局.实验表明,算法能够有效地实现在线资源分配与任务布局,获得较高的资源利用率.

Algorithms of Reconfigurable Resource Management and Hardware Task Placement

Reconfigurable computing system has the flexibility of traditional CPU and the speed of ASIC approximately.Based on the ability of partially dynamic reconfiguration,the tasks can be dynamically reconfigured on the reconfigurable hardware at runtime.Some hardware tasks can run at the same time with the execution of the reconfiguration process of other tasks.To some extent,the runtime reconfiguration overhead can be hidden and the system performance can be improved.Reconfigurable computing has become one of the most important computing methods.With the improvement of the size and integration,such as FPGA,more and more tasks can run and or resident on the reconfigurable hardware concurrently.In order to utilize the reconfigurable hardware efficiently,the reconfigurable resource management and hardware task placement are very important.A task-top based keep all maximal empty rectangles(TT-KAMER)algorithm is presented in this paper.Maximal empty rectangles can efficiently represent all the empty reconfigurable resources on the reconfigurable hardware.Based on the TT-KAMER algorithm,hardware task placement can also be implemented by the first fit algorithm and the heuristic best fit algorithm.The results indicate that the algorithms can implement resource allocation and on-line task placement efficiently,and high reconfigurable resource utilization can be obtained.