arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2103.16234cs.DCcs.LG
收藏

cuConv:一种用于 CNN 推理的卷积 CUDA 实现

cuConv: A CUDA Implementation of Convolution for CNN Inference

Barcelona Supercomputing Center (BSC)(巴塞罗那超级计算中心(BSC))

Marc Jordà, Pedro Valero-Lara, Antonio J. Peña

AI总结:

针对 CNN 推理中常用卷积配置的效率问题,本文提出无需数据预处理、强调合并访存的 GPU 实现 cuConv,相对 cuDNN 最高实现 2.29 倍加速。

摘要:

卷积是基于卷积神经网络(CNN)的深度学习应用的核心运算。当前 GPU 架构在训练和部署深度 CNN 方面非常高效,因此被大量用于生产环境。然而,对于某些常用网络配置,现有最先进实现仍存在效率不足的问题。本文提出一种基于 GPU 的 CNN 推理卷积运算实现,其优先支持合并访存,且无需预先进行数据变换。实验表明,在一系列常见 CNN 前向传播卷积配置中,该方案带来了显著性能提升;相对于 cuDNN 中最佳的卷积实现,加速比最高可达 2.29 倍,因而覆盖了当前现有方法中的一个相关区域。

英文摘要:

Convolutions are the core operation of deep learning applications based on Convolutional Neural Networks (CNNs). Current GPU architectures are highly efficient for training and deploying deep CNNs, and hence, these are largely used in production for this purpose. State-of-the-art implementations, however, present a lack of efficiency for some commonly used network configurations. In this paper we propose a GPU-based implementation of the convolution operation for CNN inference that favors coalesced accesses, without requiring prior data transformations. Our experiments demonstrate that our proposal yields notable performance improvements in a range of common CNN forward propagation convolution configurations, with speedups of up to 2.29x with respect to the best implementation of convolution in cuDNN, hence covering a relevant region in currently existing approaches.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑