Dynamic Supply Voltage Level Generation for Minimum Energy Real Time Tasks using Geometric Programming

Communication and computation are moving towards mobile platforms to address the demands of emerging applications. Despite advances in process and battery technologies that allow processors to provide much greater computation per unit of energy and longer life of battery, the fundamental tradeoff between performance and battery life continues to remain critical. To maximize energy efficiency of processors in mobile electronics, Dynamic Voltage Scaling (DVS) is conventionally deployed to dynamically vary supply voltage and hence speed, at run time. The nonlinear relationship between CPU speed and power consumption enables spread out of task execution in time domain by leveraging on the available slackness by reducing voltage, than to run the CPU at full speed for short bursts and then switch to idle state. The proposed work aims to minimize the energy consumption of each task of real time periodic task sets, in a uniprocessor environment, using task utilization factor as a control variable for generating the optimized supply voltage to every task of task sets. The energy minimization of a task is implemented using Geometric Programming (GP), by varying frequency over a range on fixed task sets and on randomly varying task set instances and hence generating supply voltage levels. Results demonstrate that energy savings vary between 18% to 34%, for standard task sets and an average of 77% for randomly generated task sets, depending on the power delay characteristics of task sets.