TY - GEN
T1 - Rotation scheduling and voltage assignment to minimize energy for SoC
AU - Qiu, Meikang
AU - Yang, Laurence T.
AU - Sha, Edwin H.M.
PY - 2009
Y1 - 2009
N2 - Low energy consumption is a critical issue in embedded systems design. As the technology feature sizes of SoC (Systems on Chip) become smaller and smaller, the percentage of leakage power consumption in total power consumption continues to grow. Traditional Dynamic Voltage Scaling (DVS) fail to accurately address the impact of scaling on system power consumption as the leakage power increases exponentially. The combination of DVS and Adaptive Body Biasing (ABB) is an effective technique to jointly optimize dynamic and leakage energy dissipation. In this paper, we propose an optimal soft real-time Rotation scheduling and voltage assignment algorithm, RSVAME (Rotation Scheduling and Voltage Assignment to Minimize Energy), to minimize both dynamic and leakage energy via DVS and ABB. Voltage transition overhead has been considered in our approach. We conduct experiments on a set of DSP benchmarks based on the power model of 70 nm technology. The experimental results show that our approach achieves significant energy saving than that of ILP (Integer Linear Programming) approach.
AB - Low energy consumption is a critical issue in embedded systems design. As the technology feature sizes of SoC (Systems on Chip) become smaller and smaller, the percentage of leakage power consumption in total power consumption continues to grow. Traditional Dynamic Voltage Scaling (DVS) fail to accurately address the impact of scaling on system power consumption as the leakage power increases exponentially. The combination of DVS and Adaptive Body Biasing (ABB) is an effective technique to jointly optimize dynamic and leakage energy dissipation. In this paper, we propose an optimal soft real-time Rotation scheduling and voltage assignment algorithm, RSVAME (Rotation Scheduling and Voltage Assignment to Minimize Energy), to minimize both dynamic and leakage energy via DVS and ABB. Voltage transition overhead has been considered in our approach. We conduct experiments on a set of DSP benchmarks based on the power model of 70 nm technology. The experimental results show that our approach achieves significant energy saving than that of ILP (Integer Linear Programming) approach.
UR - https://www.scopus.com/pages/publications/70749141579
U2 - 10.1109/CSE.2009.153
DO - 10.1109/CSE.2009.153
M3 - 会议稿件
AN - SCOPUS:70749141579
SN - 9780769538235
T3 - Proceedings - 12th IEEE International Conference on Computational Science and Engineering, CSE 2009
SP - 48
EP - 55
BT - Proceedings - 12th IEEE International Conference on Computational Science and Engineering, CSE 2009 - 7th IEEE/IFIP International Conference on Embedded and Ubiquitous Computing, EUC 2009
T2 - 7th IEEE/IFIP International Conference on Embedded and Ubiquitous Computing, EUC 2009
Y2 - 29 August 2009 through 31 August 2009
ER -