High energy consumption has become a limiting factor for battery-operated embedded systems. Most low-power compiler optimization techniques take the approach of minimizing the energy consumption while meeting small performance loss. In addition, it is possible that the available energy budget is not sufficient to meet the optimal performance objective. In such situation, energy-constrained optimization is more significant. In this paper, we explore two kinds of energy-aware prefetching optimizations: prefetching optimization with minimizing energy consumption and energy-constrained prefetching optimization. We exploit energy saving opportunities through reducing memory stalls and CPU stalls caused by too early or too late prefetching. We build models for these two kinds of energy-aware prefetching optimization approaches and use a group of array-dominated applications to validate our approach.
展开▼
