Research on a Real-Time Task Scheduling Algorithm for Hybrid Reconfigurable System-on-Chip

Todays reconfigurable hardware devices have huge densities and partially dynamically reconfigurable, allowing for the configuration and execution of hardware tasks in a real multitasking manner. This makes reconfigurable platforms an ideal target for many special application areas, such as embedded system that combines high computation demands with dynamic task sets. The key issues of software running on this platform are online task scheduling and hardware resource management. Finding the available start time and empty space for arrival tasks on FPGAs with runtime partially reconfigurable abilities is the most important phase in on-line scheduling algorithm. The scheduling of hardware task has the highest impact on the performance of the reconfigurable computing system. Presented in this paper is a group-contiguous algorithm which evaluates placement position based on task group information, and the notion of task contiguous for scheduling algorithm is introduced. By utilizing the temporal information and optimized place strategy, the proposed algorithm achieves high scheduling performance and reduces the waste of reconfigurable resources. Simulation experiments conducted with synthetic workloads evaluate the performance and the runtime efficiency of the proposed schedulers. The simulation results show that using the GC algorithm, higher task accept ratio can be achieved than using other existent scheduling algorithms.