Compression-Aware Energy Optimization for Video Decoding Systems With Passive Power
Akyol, E.
van der Schaar, M.
Electr. & Comput. Eng. Dept., California, Univ., Santa Barbara, CA;
This paper appears in: Circuits and Systems for Video Technology, IEEE Transactions on
Publication Date: Sept. 2008
Volume: 18,
Issue: 9
On page(s): 1300-1306
ISSN: 1051-8215
Digital Object Identifier: 10.1109/TCSVT.2008.928886
First Published: 2008-07-25
Current Version Published: 2008-10-07
Abstract The objective of dynamic voltage scaling (DVS) is to adapt the frequency and voltage for configurable platforms to obtain energy savings. DVS is especially attractive for video decoding systems due to their time-varying and highly complex workload and because the utility of decoding a frame is solely depending on the frame being decoded before its display deadline. Several DVS algorithms have been proposed for multimedia applications. However, the prior work did not take into account the video compression algorithm specifics, such as considering the temporal dependencies among frames and the required display buffer. Moreover, the effect of the passive (leakage) power when performing DVS for multimedia systems was not explicitly considered. In this paper, we determine the optimal scheduling of the active and passive states to minimize the total energy for video decoding systems. We pose our problem as a buffer-constrained optimization problem with a novel, compression-aware definition of processing jobs. We propose low-complexity algorithms to solve the optimization problem and show through simulations that significant improvements can be achieved over state-of-the art DVS algorithms that aim to minimize only the active power.
Index
Terms
Available to subscribers and IEEE members.
References
Available to subscribers and IEEE members.
Citing Documents
Available to subscribers and IEEE members.
Cart
