Accelerating Low-End Edge Computing with Cross-Kernel Functionality Abstraction

Chao Wu, Yaoxue Zhang, Yuezhi Zhou, Qiushi Li
Department of Computer Science and Technology, TNList, Tsinghua University
18th International Conference on Algorithms and Architectures for Parallel Processing, 2018


   title={Accelerating Low-End Edge Computing with Cross-Kernel Functionality Abstraction},

   author={Wu, Chao and Zhang, Yaoxue and Zhou, Yuezhi and Li, Qiushi},



Download Download (PDF)   View View   Source Source   



This paper envisions a future in which high performance and energy-modest parallel computing on low-end edge devices were achieved through cross-device functionality abstraction to make them interactive to cloud machines. Rather, there has been little exploration of the overall optimization into kernel processing can deliver for increasingly popular but heavy burden on low-end edge devices. Our idea here is to extend the capability of functionality abstraction across edge clients and cloud servers to identify the computation-intensive code regions automatically and execute the instantiation on the server at runtime. This paper is an attempt to explore this vision, ponder on the principle, and take the first steps towards addressing some of the challenges with edgeBoost. As a kernel-level solution, edgeBoost enables edge devices to abstract not only application layer but also system layer functionalities, as if they were to instantiate the abstracted function inside the same C kernel programming. Experimental results demonstrate that edgeBoost makes cross-kernel functionality abstraction efficient for low-end edge devices and benefits them significant performance optimization than the default scheme unless in a constraint of low transmission bandwidth.
No votes yet.
Please wait...

* * *

* * *

HGPU group © 2010-2021 hgpu.org

All rights belong to the respective authors

Contact us: