Real-Time Scheduling on Heterogeneous Multiprocessors
Gurulingesh Raravi · 2014
Embedded computing is one of the most important areas in computer science today, witnessed by the fact that 98 % of all computers are embedded. Given that many embedded systems have to interact “promptly ” with their physical environment, the scientific community has invested signifi-cant efforts in developing algorithms for scheduling the workload, which is generally implemented as a set of tasks, at the right time and in proving before run-time that all the timing requirements will be satisfied at run-time. This field of study is referred to as the real-time scheduling theory. The scheduling theory for a unicore processor is well-developed; the scientific results are taught at all major universities world-wide and the results are widely-used in industry. Scheduling theory for multicores is emerging but the focus so far has been for multicores with identical pro-cessing units. This is unfortunate because the computer industry is moving towards heterogeneous multicores with a constant number of distinct processor types — AMD Fusion, Intel Atom and NVIDIA Tegra are some of the examples of such multicores. This work deals with the problem of scheduling a set of tasks to meet their deadlines on het-erogeneous multiprocessors with a constant number of distinct processor types. On heterogeneous