Research And Implementation Of Real-time Task Assignment Algorithms On Heterogeneous Multicore Of Processors

Because of the diversity of processors integrated in the heterogeneous multicore processor, the heterogeneous multicore processor supports some applications well(e.g. signal processing, image processing and so on). Moreover, it masks the research of multi-core real-time assignment be a major trend in the area of real-time embedded system.At present, the reasearch of the multi-resource real-time task assignment algorithms based on the heterogeneous multicore architecture is in a fledging period. The current achievement all focus on special multicore platforms, so they can’t be applied to general case, such as the migratable tasksets. According to the mushroom of the hardware architecture and the wide application requirements of the heterogeneous multicore architecture, it is urgent to propose an algorithm which can efficiently assign the tasks with multi-resource on the heterogeneous multicore architecture. Combined with rational real-time scheduling algorithms, the assignment algorithm can assgin the certain task set and then schedule to execute, which is much more in demand in the heterogeneous multicore real-time system.In this thesis, we pay attention to the real-time assignment algorithms on the heterogeneous multicore architecture. The thesis classifies real-time tasks according to the migrating capability, and then designs an assginment algorithm for each respectively. At first, for the tasks which can’t migrate between processors, the thesis classifies the task set once more, whice achieves a good Speedup Ratio. For the tasks which can migrate between processors, the thesis demonstrates the serializability of the splitting subtask on different platform. Then the thesis improve Bfair(Boundary Fair algorithm) aiming at non-integer resource demands of the tasks. Moreover, based on the improved Bfair, the thesis design a new real-time assignment algorithm.In the end, the thesis improves the algorithm proposed in the front part of this thesis and then implements them. The experiment with random task set for the algorithms are conducted and the experimental results are compared with the exsiting SKB-algorithms. It turned out that our algorithms are better than the exsiting assignment algorithms. This thesis effectively offers two solutions for the multi-resource real-time assignment issue on the heterogeneous multicore architecture, and provides two kinds of fundamental of algorithmics for the application of task assignment in the multicore real-time system.