Energy estimation models for video decoders: reconfigurable video coding‐CAL case‐study

Abstract

In this study, a platform-independent energy estimation methodology is proposed to estimate the energy consumption of reconfigurable video coding (RVC)-CAL video codec specifications. This methodology is based on the performance monitoring counters (PMCs) of embedded platforms and demonstrates its portability, simplicity and accuracy for on-line estimation. It has two off-line procedure stages: the former, which automatically identifies the most appropriate PMCs with no specific detailed knowledge of the employed platform, and the latter, which trains the model using either a linear regression or a multivariable adaptive regression splines (MARS) method. Experimenting on an RVC-CAL decoder, the proposed PMC-driven model can achieve an average estimation error <10%. In addition, the maximal model computation overhead is 4.04%. The results show that the training video sequence has significant influence on the model accuracy. An experimental metric is introduced to achieve more stable accurate models based on a combination of training sequences. Furthermore, a comparison demonstrates better predictive ability of MARS techniques in scenarios with multi-core platforms. Finally, the experimental results show a good potential of energy efficiency improvement when the estimation model is combined into the RVC framework. In two different scenarios, the battery lifetime is increased 5.16% and 20.9%, respectively.

Power Modeling for Video Streaming Applications on Mobile Devices

“…In the literature, many models were proposed for estimating the decoding energy and the decoding power [10], [11], [22], [23], [28], [29]. As most modern smartphones provide hardware video decoder modules that are more power-efficient than software video decoders [15], we only consider a hardware decoder model proposed in [25].…”

“…To have a representative set of video sequences, we take 16 sequences from the HEVC common test conditions [43]. Each sequence is coded with four quality levels (x264 and x265 encoding, medium preset for both H.264/AVC and HEVC video codecs, constant rate factors 18,23,28,33). Hence, we measure a large number of different video frame rates, resolutions, and bitrates.…”

Automatic instrumentation of dataflow applications using PAPI

“…The resulting extra abstraction layer aims at unifying the procedure of monitoring each actor independently, hence, to configure the instrumentation in terms of the hardware resource that is executing the actor (the so-called PAPI components) and/or the specific events being monitored. This is a preliminary step where both heterogeneous platforms (including several PAPI components) and some Key Performance Indicators (KPI), such as energy consumption estimation, will be supported [9].…”

“…Using PAPI, the PMCs can be accessed to profile information such as memory usage, code parallelization, workload distribution, I/O utilization, etc. Additionally, some other parameters can be estimated combining this information, such as power or energy [9,10]. Having these performance indicators would contribute not only to designers' productivity but also to achieve an iterative design flow.…”

PAPIFY: Automatic Instrumentation and Monitoring of Dynamic Dataflow Applications Based on PAPI

“…This information can, in turn, be used to infer information of a higher level, such as memory usage, code parallelization, workload distribution, I/O utilization, etc. Additionally, other parameters, from an even higher level of abstraction, can be estimated combining this information, e.g., power or energy [7], [8]. The availability of these performance indicators contributes not only to increase designers' productivity, but also enables iterative design flows.…”