通过对单电压动态可重构可编程逻辑门阵列(FPGA)实时任务模型的学习研究,建立基于双电压动态可重构FPGA的非可抢占任务模型.该模型很好地描述了双电压动态可重构FPGA任务的特征,在单电压任务模型的基础上增加了双电压任务模型特有的属性.并基于降序首次自适应算法提出一种新的硬件任务调度算法,该算法在保证任务集限制时间内完成所有任务的同时,利用动态电压调节的方法大幅度减少了任务集完成所需要的能量.在SunSaloris工作站下对任务调度模型及其算法进行仿真和评估,实验数据表明,这种基于双电压的任务模型的调度算法,能够有效降低FPGA任务执行的能耗,对于大规模的任务,能耗最高可节省24.1%.
A non-preemptive task model based on dual-supply voltage (Vdd) dynamic reconfigurable field programmable gate array (FPGA) was established by analyzing single voltage reconfigurable FPGA.The model described how the tasks excused in dual-Vdd dynamic reconfigurable FPGA and added special attributes of dual-Vdd task model based on single voltage task model.A novel scheduling strategy was proposed based on the first-fit-decreasing algorithm.The strategy can guarantee the execution deadline of the task set and greatly reduce the execution energy using dynamic adjustment of dual-Vdd.The experiment simulated on Sun Saloris shows that the strategy can achieve the great energy saving,which is up to 24.1% of large task set.