Distributed OpenCL Distributing OpenCL Platform on Network Scale
Department of Computer Engineering, Istanbul Technical University, Istanbul, Turkey
International Journal of Computer Applications (Special Issue on Advanced Computing and Communication Technologies for HPC Applications – ACCTHPCA), 2012
@article{key:article,
author={Barhh Eskikaya and D Turgay Altilar},
title={Article: Distributed OpenCL Distributing OpenCL Platform on Network Scale},
journal={IJCA Special Issue on Advanced Computing and Communication Technologies for HPC Applications},
year={2012},
volume={ACCTHPCA},
number={2},
pages={26-30},
month={July},
note={Published by Foundation of Computer Science, New York, USA}
}
This paper presents a framework that extends OpenCL by distributing computing process to many computing resources connected via network and enables the computing resources to run in parallel. Using JSON RPC (Remote Procedure Call technique relying on JavaScript Object Notation) in communication layer, Distributed OpenCL framework provides platform and operating system independency. Using this framework, a host program executed on a computer that has no OpenCL support is able to use other computing resources distributed on network in parallel. Results show that OpenCL platform model can be extended to network scale to provide a vendor, architecture and operating system independent and a parallel computing environment with reasonable communication overhead.
July 15, 2012 by hgpu
