5
Conclusions
To summarize, in this paper we provided an experimental methodology to extend
the DVFS operating points supported by an embedded CPU beyond the nominal
ones, and validating the new points. Moreover, our approach allows to measure
the overheads occurring in DVFS transitions, both in terms of performance and
energy consumption. The approach has been tested on a real embedded devel-
opment platform, using CPU stressing workloads.
Overall, the outcome of this work aims at being a support for hardware
designer and OS level developers, in order to optimize DVFS mechanisms and
run-time management policies, taking into account both its advantages and costs.
Do'stlaringiz bilan baham: |