Over the last two decades, it has been widely accepted that dynamic voltage/frequency scaling (DVS) is one of the most effective techniques of minimizing the energy consumption of the embedded systems. One common assumption of almost all of the existing DVS algorithms is that the value of power consumption monotonically increases as the level of applied voltage to the system increases, and the power is a convex function of the voltage. Theoretically, under that assumption, previous works have shown that the DVS problem for a set of tasks with continuous or discrete convex power function is energy-optimally solvable in polynomial time. However, recently it is observed that some of the DC-DC converters do not follow the convexity. Thus, the power function of the whole system including a DC-DC converter will not be convex any more. In this context, we want to answer the following question: is there an energy-optimal polynomial-time algorithm that is able to solve the DVS problem with any non-convex power function? The work answers `yes', and proposes an energy-optimal polynomial-time algorithm for the new problem.
[1]
Elad Alon,et al.
Design Techniques for Fully Integrated Switched-Capacitor DC-DC Converters
,
2011,
IEEE Journal of Solid-State Circuits.
[2]
Hiroto Yasuura,et al.
Voltage scheduling problem for dynamically variable voltage processors
,
1998,
Proceedings. 1998 International Symposium on Low Power Electronics and Design (IEEE Cat. No.98TH8379).
[3]
F. Frances Yao,et al.
A scheduling model for reduced CPU energy
,
1995,
Proceedings of IEEE 36th Annual Foundations of Computer Science.
[4]
Xin He,et al.
SuperRange: Wide operational range power delivery design for both STV and NTV computing
,
2014,
2014 Design, Automation & Test in Europe Conference & Exhibition (DATE).
[5]
Taewhan Kim,et al.
DC–DC Converter-Aware Power Management for Low-Power Embedded Systems
,
2007,
IEEE Transactions on Computer-Aided Design of Integrated Circuits and Systems.