Design and Test Based on Stream Processor Programmable Cluster Architecture

DOI

10.2991/amcce-15.2015.69How to use a DOI?

Keywords

stream processor; GPU programming model; CUDA technique; stream processor cluster

Abstract

As the representative of common programmable stream processor, the performance of GPU develops rapidly, which has broken Moore's law which is obeyed by CPU. GPU applies the performance of programmability and functional expansibility to support complicated computation and process, and the feature has been acknowledged in the industry. The paper deeply researches programmable model CUDA based on NVIDIA GPU and analyzes CUDA technique. And the paper applies MPI+CUDA hybrid programming model and common stream processor to load the model, which not only separates the control of stream processor cluster from computation, and optimizes multi-data stream processing strategy, but also promotes the performance of stream processor cluster system compared with traditional x86 cluster system.

Copyright

© 2015, the Authors. Published by Atlantis Press.

Open Access

This is an open access article distributed under the CC BY-NC license (http://creativecommons.org/licenses/by-nc/4.0/).

Download article (PDF)

TY  - CONF
AU  - Wen Xu
AU  - Dongyan Wu
AU  - Lijun Qi
AU  - Mingge Wang
AU  - Jingxin Xiao
PY  - 2015/04
DA  - 2015/04
TI  - Design and Test Based on Stream Processor Programmable Cluster Architecture
BT  - Proceedings of the 2015 International Conference on Automation, Mechanical Control and Computational Engineering
PB  - Atlantis Press
SP  - 360
EP  - 364
SN  - 1951-6851
UR  - https://doi.org/10.2991/amcce-15.2015.69
DO  - 10.2991/amcce-15.2015.69
ID  - Xu2015/04
ER  -