| Previous | [ 1] | [ 2] | [ 3] | [ 4] | [ 5] | [ 6] | [ 7] | [ 8] | [ 9] | [ 10] | [ 11] | [ 12] | [ 13] | [ 14] | [ 15] | [ 16] | [ 17] | [ 18] |
¡@
Chien-Chung Yang, Kuochen Wang, Ming-Ham Lin and Pochun Lin
Department of Computer Science
National Chiao Tung University
Hsinchu, 300 Taiwan
It is not energy efficient to run a CPU at full speed all the time for all kinds of tasks
in mobile devices. This paper proposes two energy efficient intra-task dynamic voltage
scaling (DVS) algorithms for CPUs. There are three main contributions in this paper.
Firstly, unlike the tedious derivation in PACE [2], we have derived the same optimal
speed schedule with minimal energy consumption in a discrete and elegant way by using
the Lagrange multiplier procedure. Secondly, the CPU model assumed in PACE is ideal,
meaning that such a CPU supports all possible frequencies/voltage levels. We call such
CPUs as ideal CPUs. In reality, CPUs only support a limited set of frequency/voltage
levels, and we call this kind of CPUs as realistic CPUs. Thirdly, since energy consumption
is not a simple function of frequency, it is more reasonable to transform the original
nonlinear programming problem to the Multiple-Choice Knapsack Problem (MCKP).
Since the problem can be described by a multistage graph, we used dynamic programming
to derive an Optimal Schedule for Realistic CPUs (OSRC) with minimal energy
consumption for realistic CPUs by using actual power consumption specifications of realistic
CPUs. Considering potential computation and transition overheads, we have also
proposed a low overhead OSRC (LO-OSRC), which restricts the change of CPU frequency/
voltage to only once in the speed schedule. By using actual data from the power
consumption specifications of two classical CPUs for evaluation, experimental results
have shown that the energy saving of the proposed OSRC (LO-OSRC) is up to 10.3%
(9.4%) better than that of PACE for realistic CPUs.
Received February 27, 2007; revised August 7, 2007 & February 27, 2008; accepted August 20, 2008.
Communicated by Yao-Wen Chang.
* The insightful comments of the reviewers help improve the quality of the paper. This work was supported by
the NCTU EECS-MediaTek Research Center under grant Q583 and the National Science Council under
grant No. NSC96-2628-E-009-140-MY3. The authors would like to thank Jheng-Ming Chen for his help on
the preparation of the paper.