gpuroofline: a model for guiding performance optimizations on gpus | |
Jia Haipeng ; Zhang Yunquan ; Long Guoping ; Xu Jianliang ; Yan Shengen ; Li Yan | |
2012 | |
会议名称 | 18th International Conference on Parallel Processing, Euro-Par 2012 |
会议日期 | August 27, 2012 - August 31, 2012 |
会议地点 | Rhodes Island, Greece |
关键词 | Laplace transforms Optimization |
页码 | 920-932 |
中文摘要 | 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. |
英文摘要 | 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. |
收录类别 | EI |
会议录 | Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
![]() |
语种 | 英语 |
ISSN号 | 0302-9743 |
ISBN号 | 9783642328190 |
内容类型 | 会议论文 |
源URL | [http://ir.iscas.ac.cn/handle/311060/15892] ![]() |
专题 | 软件研究所_软件所图书馆_会议论文 |
推荐引用方式 GB/T 7714 | Jia Haipeng,Zhang Yunquan,Long Guoping,et al. gpuroofline: a model for guiding performance optimizations on gpus[C]. 见:18th International Conference on Parallel Processing, Euro-Par 2012. Rhodes Island, Greece. August 27, 2012 - August 31, 2012. |
个性服务 |
查看访问统计 |
相关权益政策 |
暂无数据 |
收藏/分享 |
除非特别说明,本系统中所有内容都受版权保护,并保留所有权利。
修改评论