作者: Yahia Benmoussa , Eric Senn , Nicolas Derouineau , Nicolas Tizon , Jalil Boukhobza
DOI: 10.1109/PATMOS.2016.7833693
关键词: Adaptive filter 、 Code (cryptography) 、 Energy (signal processing) 、 Efficient energy use 、 Computer science 、 Metadata 、 Frame (networking) 、 Metadata standard 、 Decoding methods 、 Real-time computing
摘要: We present in this paper GM-DVFS, an adaptive DVFS scheme for energy efficient decoding of H.264 videos. GM-DVFS uses metadata (normalized by the MPEG Green Metadata standard) providing information about upcoming workload. In solution we propose, these are processed within filter to build dynamically accurate video complexity model. This model serves calculate minimal required processor frequency a frame while guaranteeing real time constraints. Our performance evaluations showed that proposed algorithm is able converge (4%) less than 1 second (in worst case). Moreover, it simple implement (250 lines C code) and induces very low overhead (1400 cycles/frames). On other hand, allows achieve up 46% saving as compared on demand Linux governor.