Workload- and User-aware Battery Lifetime Management for Mobile SoCs

Chetoui, Sofiane; Reda, Sherief

doi:10.23919/date51398.2021.9474111

Cited by 3 publications

(1 citation statement)

References 16 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Firstly, mobile GPUs are mainly designed for rendering instead of general-purpose computing. Their computing power is highly constrained due to the battery life consideration [25]. Secondly, the DL libs are not as well optimized for GPUs as CPUs.…”

Section: Impacts Of Hardwarementioning

confidence: 99%

Benchmarking of DL Libraries and Models on Mobile Devices

Zhang¹,

Li²,

Che³

et al. 2022

Preprint

View full text Add to dashboard Cite

Deploying deep learning (DL) on mobile devices has been a notable trend in recent years. To support fast inference of on-device DL, DL libraries play a critical role as algorithms and hardware do. Unfortunately, no prior work ever dives deep into the ecosystem of modern DL libs and provides quantitative results on their performance. In this paper, we first build a comprehensive benchmark that includes 6 representative DL libs and 15 diversified DL models. We then perform extensive experiments on 10 mobile devices, which help reveal a complete landscape of the current mobile DL libs ecosystem. For example, we find that the best-performing DL lib is severely fragmented across different models and hardware, and the gap between those DL libs can be rather huge. In fact, the impacts of DL libs can overwhelm the optimizations from algorithms or hardware, e.g., model quantization and GPU/DSP-based heterogeneous computing. Finally, atop the observations, we summarize practical implications to different roles in the DL lib ecosystem. CCS Concepts• General and reference → Measurement; • Human-centered computing → Ubiquitous and mobile devices.

show abstract

Section: Impacts Of Hardwarementioning

confidence: 99%

Benchmarking of DL Libraries and Models on Mobile Devices

Zhang¹,

Li²,

Che³

et al. 2022

Preprint

View full text Add to dashboard Cite

show abstract

A Comprehensive Deep Learning Library Benchmark and Optimal Library Selection

Zhang

Che

Chen

et al. 2024

IEEE Trans. on Mobile Comput.

View full text Add to dashboard Cite

An Evaluation Method of an Energy Consumption as an Operation Parameter in a Cyber-Physical System

Nurgaliyev,

Tokhmetov,

Tanchenko

2024

sjaitu

View full text Add to dashboard Cite

The research of energy consumption in an Internet of Things network and its analytical evaluation is the goal of this work. The authors of this work concentrate on developing a model for calculating the actual gain in power consumption in order to estimate the actual energy required. The method suggests measuring the difference in energy usage under three primary battery-powered working modes to maximize a device's lifetime. Due to the fact that each CPS device state has its own energy metrics, it is feasible to choose the best operation course for entire network. The presented technique is certainly viable, as demonstrated by the experimental examination of Zigbee and BLE devices. The comparison of power levels using a temperature sensor in three basic scenarios (power modes) dictates how the CPS device lifetime can be optimized. Multi-regime consumption models, in which the rates of charging and discharging are dependent upon the energy level, are analyzed in this paper. This work aimed to state an optimal energy consumption by finding the right balance between operational power and battery lifetime through mathematical modeling. Therefore, it is easy to determine the energy cost of power stage, for instance, to send data by setting the minimal duration of each working condition in terms of power consumption. Moreover, a reasonable balance of power consumption and battery lifetime which impacts the data collection from sensors is vital to the development of data extraction algorithms. The practical results depict how device should be accessible to be able to lose less power even during switching on/off or how operate more effective if it used for a short period of time. A long-term network could become a reality once battery life is optimized enough to not disturb a user.

show abstract

Workload- and User-aware Battery Lifetime Management for Mobile SoCs

Cited by 3 publications

References 16 publications

Benchmarking of DL Libraries and Models on Mobile Devices

Benchmarking of DL Libraries and Models on Mobile Devices

A Comprehensive Deep Learning Library Benchmark and Optimal Library Selection

An Evaluation Method of an Energy Consumption as an Operation Parameter in a Cyber-Physical System

Contact Info

Product

Resources

About