User Interaction Aware Reinforcement Learning for Power and Thermal Efficiency of CPU-GPU Mobile MPSoCs

  title={User Interaction Aware Reinforcement Learning for Power and Thermal Efficiency of CPU-GPU Mobile MPSoCs},
  author={Somdip Dey and Amit Kumar Singh and Xiaohang Wang and Klaus Dieter Mcdonald-Maier},
  journal={2020 Design, Automation \& Test in Europe Conference \& Exhibition (DATE)},
Mobile user’s usage behaviour changes throughout the day and the desirable Quality of Service (QoS) could thus change for each session. In this paper, we propose a QoS aware agent to monitor mobile user’s usage behaviour to find the target frame rate, which satisfies the desired user’s QoS, and applies reinforcement learning based DVFS on a CPU-GPU MPSoC to satisfy the frame rate requirement. Experimental study on a real Exynos hardware platform shows that our proposed agent is able to achieve… 

Figures from this paper

QUAREM: Maximising QoE Through Adaptive Resource Management in Mobile MPSoC Platforms

Heterogeneous multi-processor system-on-chip (MPSoC) smartphones are required to offer increasing performance and user quality-of-experience (QoE), despite comparatively slow advances in battery

CPU-GPU-Memory DVFS for Power-Efficient MPSoC in Mobile Cyber Physical Systems

This paper proposes a novel approach to dynamic voltage and frequency scaling (DVFS) on CPU, GPU and RAM in a mobile MPSoC, which caters to the performance requirements of the executing application while consuming low power.

New approaches to DVFS in mobile MPSoC for power-, thermal-efficiency and reliability

This poster presentation introduces three novel approaches to DVFS in mobile MPSoCs such that the executing applications can meet the performance requirement while consuming less power, reducing peak temperature and improving thermal side-channel security and reliability in terms of lifespan of the device.

Learning-Oriented QoS- and Drop-Aware Task Scheduling for Mixed-Criticality Systems

A novel optimistic scheme is proposed, that introduces a learning-based drop-aware task scheduling mechanism, which carefully monitors the alterations in the behaviour of the MC system at run-time, to exploit the generated dynamic slacks for reducing the LC tasks penalty and preventing frequent drops of LC tasks in the future.

Asynchronous Hybrid Deep Learning (AHDL): A Deep Learning Based Resource Mapping in DVFS Enabled Mobile MPSoCs

A novel methodology, Asynchronous Hybrid Deep Learning (AHDL), which sets a new paradigm of using Deep Learning approaches to map resources to application (tasks) at the same time without the need of manual feature extraction by domain experts is proposed.

Quality Optimization of Adaptive Applications via Deep Reinforcement Learning in Energy Harvesting Edge Devices

This article is the first to propose a deep reinforcement learning-based dynamic voltage frequency scaling (DVFS) method that optimizes the application execution quality of energy harvesting edge devices to mitigate the variations of quality modeling and harvested energy simultaneously.

Low-Complexity Runtime Management of Concurrent Workloads for Energy-efficient Multi-core Systems

This paper proposes a novel runtime optimization approach aiming for maximum power normalized performance through multivariate linear regression (MLR) and demonstrates low-cost and low-complexity runtime algorithms that continuously adapt system configuration to improve the IPS/Watt by up to 139% compared to existing approaches.

Low-Complexity Run-time Management of Concurrent Workloads for Energy-Efficient Multi-Core Systems

Low-cost and low-complexity run-time algorithms that continuously adapt system configuration to improve the IPS/Watt by up to 139% compared to existing approaches are demonstrated.

ThermalAttackNet: Are CNNs Making It Easy to Perform Temperature Side-Channel Attack in Mobile Edge Devices?

A power- and memory-efficient CNN model is designed that is capable of performing thermal side-channel attack on the MPSoC and can be used by industry practitioners and academics as a benchmark to design methodologies to secure against such an attack in MP soC.

FruitVegCNN: Power- and Memory-Efficient Classification of Fruits & Vegetables Using CNN in Mobile MPSoC

This paper proposes a power- and memory-efficient CNN model, FruitVegCNN, to perform classification of fruits and vegetables in a mobile multi-processor system-on-a-chip (MPSoC).



On the Impacts of Greedy Thermal Management in Mobile Devices

Through experiments on a commercial smartphone, the impact of application duration on throttling-induced performance loss is characterized and quality-of-service (QoS) tuning is proposed as an effective way of providing the mobile system user with consistent performance levels over extended application durations.

DeadPool: Performance Deadline Based Frequency Pooling and Thermal Management Agent in DVFS Enabled MPSoCs

  • Somdip DeyA. SinghXiaohang WangK. Mcdonald-Maier
  • Computer Science
    2019 6th IEEE International Conference on Cyber Security and Cloud Computing (CSCloud)/ 2019 5th IEEE International Conference on Edge Computing and Scalable Cloud (EdgeCom)
  • 2019
This paper proposes an intelligent software agent that works alongside other resource mapping and partitioning mechanism in order to monitor and reduce the operating temperature of the system by regulating the operating frequency of the CPU cores while catering for performance constraint at the same time.

Frame-based and thread-based power management for mobile games on HMP platforms

This paper proposes a frame- and thread-based MPSoC power management strategy for games based on the fact that gaming workload has high temporal correlation among frames and evaluates selected workload predictors on a per-frame basis to use a hybrid workload predictor.

Learning Transfer-Based Adaptive Energy Minimization in Embedded Systems

The proposed approach is implemented as a power governor in Linux and extensively validated on an ARM Cortex-A8 running different benchmark applications, showing that with intra- and inter-application variations, it can effectively minimize energy consumption by up to 33% compared to the existing approaches.

TEEM: Online Thermal- and Energy-Efficiency Management on CPU-GPU MPSoCs

A thermal and energy management mechanism which achieves reduction in thermal gradient as well as energy-efficiency through resource mapping and thread-partitioning of applications with online optimization in heterogeneous MPSoCs is proposed.

P-EdgeCoolingMode: an agent-based performance aware thermal management unit for DVFS enabled heterogeneous MPSoCs

P-EdgeCoolingMode is capable of pro-actively monitoring performance and based on the user's demand the agent takes necessary action, making the proposed methodology highly suitable for implementation on existing as well as conceptual Edge devices utilising heterogeneous MPSoCs with dynamic voltage and frequency scaling (DVFS) capabilities.

EdgeCoolingMode: An Agent Based Thermal Management Mechanism for DVFS Enabled Heterogeneous MPSoCs

A light-weight novel thermal management mechanism in the form of intelligent software agent, which monitors and regulates the operating temperature of the CPU cores to improve reliability of the system is proposed.

Online Learning for Adaptive Optimization of Heterogeneous SoCs

This paper presents an online learning framework to construct adaptive analytical models for modeling GPU frame processing time, GPU power consumption and SoC power-temperature dynamics and demonstrates that the proposed approach achieves less than 6% error under dynamically varying workloads.

Integrated CPU-GPU power management for 3D mobile games

A power management approach is proposed that takes a unified view of the CPU-GPU DVFS, resulting in reduced power consumption for latest 3D mobile games compared to an independent CPU- GPU power management approaches.

RewardProfiler: A Reward Based Design Space Profiler on DVFS Enabled MPSoCs

This paper proposes a hybrid approach of resource mapping technique on DVFS enabled MPSoC, which is suitable for IDE integration due to the reduced design points in the methodology resulting in significant reduction in profiling time.