Performance analysis of deep learning workloads using roofline trajectories
Performance analysis of deep learning workloads using roofline trajectories
复制标题
DOI：
10.1007/s42514-019-00018-4
复制
发表时间：
2019-11
影响因子：
0.9
通讯作者：
M. Javed;K. Ibrahim;Xiaoyi Lu
中科院分区：
文献类型：
--
作者：
M. Javed;K. Ibrahim;Xiaoyi Lu
在过去的十年中，源自卷积神经网络 (CNN) 的技术（称为深度学习应用）彻底改变了癌症检测、自动驾驶汽车、虚拟助手等多个领域。然而，此类应用的许多用户本身并不是机器学习方面的专家。因此，社区对以优化方式运行此类应用程序的了解有限。深度学习应用程序的性能问题通常通过采用更适合此类计算密集型操作的定制硬件（例如 GPU）来解决。然而，这种程度的绩效只有在财务成本越来越高的情况下才能实现，只有大公司和政府才有足够的资源来大规模使用它们。因此，普通用户只能访问商品集群，在许多情况下，只有 CPU 作为唯一的处理元素。为了让这些用户有效地利用他们可支配的资源，需要共同努力找出最佳的硬件和软件配置。这项研究就是朝这个方向迈出的一步，我们使用 Roofline 模型对代表性 CNN 模型进行系统分析，并确定黑盒和应用感知优化的机会。根据我们的研究结果，与默认配置的普通 TensorFlow 相比，我们能够获得高达 3.5 倍的加速。
Over the last decade, technologies derived from convolutional neural networks (CNNs) called Deep Learning applications, have revolutionized fields as diverse as cancer detection, self-driving cars, virtual assistants, etc. However, many users of such applications are not experts in Machine Learning itself. Consequently, there is limited knowledge among the community to run such applications in an optimized manner. The performance question for Deep Learning applications has typically been addressed by employing bespoke hardware (e.g., GPUs) better suited for such compute-intensive operations. However, such a degree of performance is only accessibly at increasingly high financial costs leaving only big corporations and governments with resources sufficient enough to employ them at a large scale. As a result, an average user is only left with access to commodity clusters with, in many cases, only CPUs as the sole processing element. For such users to make effective use of resources at their disposal, concerted efforts are necessary to figure out optimal hardware and software configurations. This study is one such step in this direction as we use the Roofline model to perform a systematic analysis of representative CNN models and identify opportunities for black box and application-aware optimizations. Using the findings from our study, we are able to obtain up to 3.5speedup compared to vanilla TensorFlow with default configurations.