Comparison of Thread Execution Methods for GPU-oriented OpenCL Programs on Multicore Processors
Graduate School of Science and Engineering, Ritsumeikan University, 1-1-1 Noji-Higashi, Kusatsu, Shiga 525-8577, Japan
Embedded operating system workshop (EWiLi), 2015
@article{hojo2015comparison,
title={Comparison of Thread Execution Methods for GPU-oriented OpenCL Programs on Multicore Processors},
author={Hojo, Naohisa and Taniguchi, Ittetsu and Tomiyama, Hiroyuki},
year={2015}
}
With the broad deployment of multicore processors, there are increasing demands to port OpenCL programs written for GPUs onto the multicore processors. However, OpenCL programs written for GPUs cannot run efficiently on multicore processors since GPU-oriented OpenCL programs generally consist of a huge number of threads. This paper presents experimental comparisons of three thread execution methods for GPU-oriented OpenCL programs on multicore processors using a set of industry-oriented OpenCL benchmark programs.
October 16, 2015 by hgpu
