为什么FPGA-GPU异质结是嵌入式深度神经网络的最佳选择？
Why is FPGA-GPU Heterogeneity the Best Option for Embedded Deep Neural Networks?
摘要：
图形处理单元（GPU）目前是深度学习（DL）加速器的主要可编程架构。然而，在DL加速器中采用现场可编程门阵列（FPGA）的势头越来越猛。在本文中，我们证明了直接硬件映射（DHM）的卷积神经网络（CNN）在嵌入式FPGA上大大优于GPU实现的能源效率和执行时间。然而，DHM是高度资源密集型的，并且在实现最先进的CNN时不能完全替代GPU。因此，我们提出了一种混合的FPGA-GPU DL加速方法，并证明异构加速优于GPU加速，即使包括通信开销。实验结果在嵌入Nvidia（R）Jetson TX 2 CPU-GPU板和Intel（R）Cyclone 10 GX FPGA板的异构多平台设置上进行。实验了SqueezeNet、MobileNetv 2和ShuffleNetv 2面向移动的CNN。我们表明，异构FPG-AGPU加速优于GPU加速，用于MobileNetv 2（12%-30%的能量减少，4%至26%的延迟减少），SqueezeNet（21%-28%的能量减少，相同的延迟）和ShuffleNetv 2（25%的能量减少，21%的延迟减少）的分类推理任务。
英文摘要：
Graphics Processing Units (GPUs) are currently the dominating programmable architecture for Deep Learning (DL) accelerators. The adoption of Field Programmable Gate Arrays (FPGAs) in DL accelerators is however getting momentum. In this paper, we demonstrate that Direct Hardware Mapping (DHM) of a Convolutional Neural Network (CNN) on an embedded FPGA substantially outperforms a GPU implementation in terms of energy efficiency and execution time. However, DHM is highly resource intensive and cannot fully substitute the GPU when implementing a state-of-the-art CNN. We thus propose a hybrid FPGA-GPU DL acceleration method and demonstrate that heterogeneous acceleration outperforms GPU acceleration even including communication overheads. Experimental results are conducted on a heterogeneous multi-platform setup embedding an Nvidia(R) Jetson TX2 CPU-GPU board and an Intel(R) Cyclone10GX FPGA board. The SqueezeNet, MobileNetv2, and ShuffleNetv2 mobile-oriented CNNs are experimented. We show that heterogeneous FPG-AGPU acceleration outperforms GPU acceleration for classification inference task over MobileNetv2 (12%-30% energy reduction, 4% to 26% latency reduction), SqueezeNet (21%-28% energy reduction, same latency), and ShuffleNetv2 (25% energy reduction, 21% latency reduction).
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。
