Combined Dynamic Voltage Scaling and Adaptive Body Biasing for Heterogeneous Distributed Real-time Embedded Systems

  • Authors:
  • Le Yan;Jiong Luo;Niraj K. Jha

  • Affiliations:
  • Princeton University, NJ;Princeton University, NJ;Princeton University, NJ

  • Venue:
  • Proceedings of the 2003 IEEE/ACM international conference on Computer-aided design
  • Year:
  • 2003

Quantified Score

Hi-index 0.00

Visualization

Abstract

Dynamic voltage scaling (DVS) is a powerful technique for reducingdynamic power consumption in a computing system. However, astechnology feature size continues to scale, leakage power is increasing and willlimit power savings obtained by DVS alone. Previous system-level real-timescheduling approaches use DVS alone to optimize power consumption withoutconsidering leakage power. To overcome this limitation, we propose a newscheduling algorithm that combines DVS and adaptive body biasing (ABB)to simultaneously optimize both dynamic power consumption and leakagepower consumption for real-time distributed embedded systems. First, wederive an analytical expression to determine the optimal supply voltage andbody bias voltage under a given clock frequency. Based on this expression, wecompute the optimal energy consumption at a given clock frequency and analyzethe tradeoff between energy consumption and execution time for a set oftasks with precedence relationships and real-time constraints. We then proposea scheduling algorithm to reduce total power consumption under givenreal-time constraints. This algorithm also considers variations in power consumptionof different tasks and characteristics of different voltage-scalableprocessing elements (PEs) to maximize power reduction. Experimental resultsshow that the average power reduction of our technique with respect toDVS alone is 34.7%, while the average saving compared to no voltage scalingis 68.3% for the 0.07µm technology.