Energy Aware Task Scheduling for Soft Real Time Systems using an Analytical Approach for Energy Estimation

Embedded systems have pervaded all walks of our life. With the increasing importance of mobile embedded systems and flexible applications, considerable progress in research has been made for power management. Power constraints are increasingly becoming the critical component of the design specifications of these systems. It helps in pre-determining the suitable hardware architecture for the target application. Very Large Instruction Word (VLIW) processors provide a means to efficiently exploit the instruction level parallelism (ILP) exhibited by a significant segment of embedded applications. Circuit level or gate level power analysis techniques prove to be impractical for the power cost estimation of the software component of the system. The aim of this paper is to present a technique to estimate ‘pre-run time’ and ‘power’ of a software mapped onto a hardware system; guaranteeing the compliance of temporal constraints while generating a schedule of tasks of software. Real time systems must handle several independent macro-tasks, each represented by a task graph, which includes communications and precedence constraints. We propose a novel approach for power estimation of embedded software using the Control Data Flow Graph (CDFG) or task graph model. This methodology uses an existing Hierarchical Concurrent Flow Graph (HCFG) technique for the power analysis of the CDFGs. We have evaluated our technique for energy efficient scheduling over various task graph benchmarks using Trimaran, an environment for software characterization and PrimePower from Synopsys has been used to obtain power estimates for the elementary functional units of datapath. The results obtained prove the utility and efficacy of our proposed approach for power analysis of embedded software. We present a methodology to obtain an energy optimal voltage assignment and perform scheduling by taking advantage of the relaxation in execution time of tasks.