文摘
For a wide variety of applications such as multimedia applications, the number of tasks running simultaneously on Multiprocessor Systems-on-Chip (MPSoCs) changes dynamically at runtime. An intelligent resource management is required to efficiently map these tasks onto the processing elements. In this paper, a new flexible software framework for allocating tasks dynamically onto a heterogeneous MPSoC is presented. This MPSoC contains an ARM processor and multiple MicroBlazes. The framework maps the tasks on the MicroBlazes based on various strategies which are selectable to meet the application-specific demands. The performance of the framework is evaluated by an automotive application. In this context, an electronic control unit uses the proposed framework to avoid collisions with other vehicles. Thereby, realistic data given by a car simulator is distributed to the processing elements of the MPSoC. Here, the distance of vehicles influences the scheduling. The evaluation shows that this approach exploits the MPSoC scalability and simplifies the programming and efficient use of such systems.