Resource Partitioning Approach for Task Allocation in Dynamically Reconfigurable Computing Systems

Dynamic reconfiguration feature in recent embedded systems provides flexibility in the execution of applications. Using this feature, several tasks of an application can be mapped simultaneously into a limited area of resources in several Execution Cycles (ECs). The complexity of resource management and efficient scheduling of tasks in such systems require more consideration of the task’s attributes. In this paper, we proposed a new model for execution configuration which satisfies task constraints such as routing and resource wasting to derive multiple configurations for each task. In order to reduce the surface fragmentation and efficient resource utilization, one partition-based scheduling algorithm with a scheduled time metric has been proposed to properly select among possible configurations and partitions. Several experiments with different scenarios such as heavy and light workloads have been conducted and the results show that rejection ratio is reduced 42.10% and 24.14% in compared with two First and Best Fit algorithms, respectively. Some improvements in other evaluation metrics have been obtained, as well.