11147

Adaptive Task Size Control on High Level Programming for GPU/CPU Work Sharing

Tetsuya Odajima, Taisuke Boku, Mitsuhisa Sato, Toshihiro Hanawa, Yuetsu Kodama, Raymond Namyst, Samuel Thibault, Olivier Aumage
University of Tsukuba
The 2013 International Symposium on Advances of Distributed and Parallel Computing (ADPC 2013), hal-00920915, 2013

@inproceedings{odajima:hal-00920915,

   hal_id={hal-00920915},

   url={http://hal.inria.fr/hal-00920915},

   title={Adaptive Task Size Control on High Level Programming for GPU/CPU Work Sharing},

   author={Odajima, Tetsuya and Boku, Taisuke and Sato, Mitsuhisa and Hanawa, Toshihiro and Kodama, Yuetsu and Namyst, Raymond and Thibault, Samuel and Aumage, Olivier},

   language={Anglais},

   affiliation={Graduate School of Systems and Information Engineering [Tsukuba] , Center for Computational Sciences [Tsukuba] – CCS , Graduate School for Systems and Information Engineering [Tsukuba] , Laboratoire Bordelais de Recherche en Informatique – LaBRI , RUNTIME – INRIA Bordeaux – Sud-Ouest},

   booktitle={The 2013 International Symposium on Advances of Distributed and Parallel Computing (ADPC 2013)},

   address={Vietri sul Mare, Italie},

   audience={internationale},

   year={2013},

   month={Dec},

   pdf={http://hal.inria.fr/hal-00920915/PDF/ADPC2013-117.pdf}

}

On the work sharing among GPUs and CPU cores on GPU equipped clusters, it is a critical issue to keep load balance among these heterogeneous computing resources. We have been developing a runtime system for this problem on PGAS language named XcalableMP-dev/StarPU [1]. Through the development, we found the necessity of adaptive load balancing for GPU/CPU work sharing to achieve the best performance for various application codes. In this paper, we enhance our language system XcalableMP-dev/StarPU to add a new feature which can control the task size to be assigned to these heterogeneous resources dynamically during application execution. As a result of performance evaluation on several benchmarks, we confirmed the proposed feature correctly works and the performance with heterogeneous work sharing provides up to about 40% higher performance than GPU-only utilization even for relatively small size of problems.
No votes yet.
Please wait...

* * *

* * *

HGPU group © 2010-2024 hgpu.org

All rights belong to the respective authors

Contact us: