Combining admission tests for heuristic partitioning of real-time tasks on ARM big.LITTLE multi-processor architectures

This paper tackles the problem of admitting real-time tasks on non-symmetric multiprocessor platforms, where partitioned EDF-FF is used. We extend the already obtained results for symmetric multiprocessor platforms and provide an improved equation based on the number of tasks in the taskset that considers the first few heaviest tasks in the taskset. This equation has been enveloped in three effective and efficient admission tests that are suitable for being used online and in dynamic environments to partition and admit tasks on NUMP platforms and, specifically, on ARM big.LITTLE architectures. This results in an improvement of the state of the art in terms of the considered architectures and looks promising in terms of admitted tasksets as shown by an extensive number of tests performed on tasksets generated randomly with the widely adopted randfixedsum algorithm.