ISCAS OpenIR
gpuroofline: a model for guiding performance optimizations on gpus
Jia Haipeng; Zhang Yunquan; Long Guoping; Xu Jianliang; Yan Shengen; Li Yan
2012
Conference Name18th International Conference on Parallel Processing, Euro-Par 2012
SourceLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Pages920-932
Conference DateAugust 27, 2012 - August 31, 2012
Conference PlaceRhodes Island, Greece
Indexed TypeEI
ISSN0302-9743
ISBN9783642328190
Department(1) Lab. of Parallel Software and Computational Science Institute of Software Chinese Academy of Sciences China; (2) College of Information Science and Engineering Ocean University of China China; (3) State Key Laboratory of Computing Science Chinese Academy of Sciences China; (4) Graduate University of Chinese Academy of Sciences China
English AbstractPerformance optimization on GPUs requires deep technical knowledge of the underlying hardware. Modern GPU architectures are becoming more and more diversified, which further exacerbates the already difficult problem. This paper presents GPURoofline, an empirical model for guiding optimizations on GPUs. The goal is to help non-expert programmers with limited knowledge of GPU architectures implement high performance GPU kernels. The model addresses this problem by exploring potential performance bottlenecks and evaluating whether specific optimization techniques bring any performance improvement. To demonstrate the usage of the model, we optimize four representative kernels with different computation densities, namely matrix transpose, Laplace transform, integral and face-dection, on both NVIDIA and AMD GPUs. Experimental results show that under the guidance of GPURoofline, performance of those kernels achieves 3.74~14.8 times speedup compared to their nai¨ve implementations on both NVIDIA and AMD GPU platforms. © 2012 Springer-Verlag.; Performance optimization on GPUs requires deep technical knowledge of the underlying hardware. Modern GPU architectures are becoming more and more diversified, which further exacerbates the already difficult problem. This paper presents GPURoofline, an empirical model for guiding optimizations on GPUs. The goal is to help non-expert programmers with limited knowledge of GPU architectures implement high performance GPU kernels. The model addresses this problem by exploring potential performance bottlenecks and evaluating whether specific optimization techniques bring any performance improvement. To demonstrate the usage of the model, we optimize four representative kernels with different computation densities, namely matrix transpose, Laplace transform, integral and face-dection, on both NVIDIA and AMD GPUs. Experimental results show that under the guidance of GPURoofline, performance of those kernels achieves 3.74~14.8 times speedup compared to their nai¨ve implementations on both NVIDIA and AMD GPU platforms. © 2012 Springer-Verlag.
KeywordLaplace Transforms Optimization
Language英语
Content Type会议论文
URIhttp://ir.iscas.ac.cn/handle/311060/15892
Collection中国科学院软件研究所
Recommended Citation
GB/T 7714
Jia Haipeng,Zhang Yunquan,Long Guoping,et al. gpuroofline: a model for guiding performance optimizations on gpus[C],2012:920-932.
Files in This Item:
There are no files associated with this item.
Related Services
Recommend this item
Bookmark
Usage statistics
Export to Endnote
Google Scholar
Similar articles in Google Scholar
[Jia Haipeng]'s Articles
[Zhang Yunquan]'s Articles
[Long Guoping]'s Articles
Baidu academic
Similar articles in Baidu academic
[Jia Haipeng]'s Articles
[Zhang Yunquan]'s Articles
[Long Guoping]'s Articles
Bing Scholar
Similar articles in Bing Scholar
[Jia Haipeng]'s Articles
[Zhang Yunquan]'s Articles
[Long Guoping]'s Articles
Terms of Use
No data!
Social Bookmark/Share
All comments (0)
No comment.
 

Items in the repository are protected by copyright, with all rights reserved, unless otherwise indicated.