Task-aware power control method for multi-mode video compression system

Dynamic voltage and frequency scaling (DVFS) technique, to some extent, controls the energy consumption for embedded multi-mode real-time video compression system. Yet, continuously decreasing voltage and frequency might affect interface resource transfers and even the normal system operations. To solve the problem, we proposed a task-aware power control method in response to task variations. We first analyzed the relationship between the workloads and processor resources, and then established a task-aware resource configuration model. Based on this model, we designed an adaptive power controller, which dynamically adjusts processor frequency and the number of cores. The experiments indicate that our method fulfills function and performance requirements of multi-mode video encoding system, with power consumption for encoding a frame reduced by 11.4% compared with traditional DVFS technique.