IEEE Transactions on Computer-Aided Design of Integrated Circuits and Systems · 2019 · 85 citations · 34 references
EngineeringComputer EngineeringComputer ScienceCompilersComputer ArchitectureField-programmable Gate ArraysAccelerator ArchitectureDeep LearningField-programmable Gate Array ArchitectureFpga ArchitectureHardware AccelerationInstruction Set ArchitectureAcceleratorsAcyclic Directed Graph RepresentationsCompiler SupportData Layout OptimizationDeep Learning AcceleratorsDirected Acyclic Graph PipelinesField-programmable Gate Array AcceleratorsSubgraph Isomorphism
state-of-the-art method for several artificial intelligence domains in recent years. The increasingly complex CNN models are both computation-bound and I/O-bound. Fieldprogrammable gate array-based accelerators driven by custom instruction set architecture (ISA) achieve a balance between generality and efficiency, but there is much on them left to be optimized. We propose the full-stack compiler deep neural network virtual machine (DNNVM), which is an integration of optimizers for graphs, loops and data layouts, an assembler, a runtime supporter, and a validation environment. The DNNVM works in the context of deep learning frameworks and transforms CNN models into the directed acyclic graph: XGraph. Based on XGraph, we transform the optimization challenges for both data layout and pipeline into graph-level problems. DNNVM enumerates all potentially profitable fusion opportunities by a heuristic subgraph isomorphism algorithm to leverage pipeline and data layout optimizations, and searches for the best choice of execution strategies of the whole computing graph. On the Xilinx ZU2@330 MHz and ZU9@330 MHz, we achieve equivalently state-of-the-art performance on our benchmarks by naïve implementations without optimizations, and the throughput is further improved up to 1.26× by leveraging heterogeneous optimizations in DNNVM. Finally, with ZU9@330 MHz, we achieve state-of-the-art performance for VGG and ResNet50. We achieve a throughput of 2.82 TOPs/s and an energy efficiency of 123.7 GOPs/s/W for VGG. Additionally, we achieve 1.38 TOPs/s for ResNet50 and 1.41 TOPs/s for GoogleNet.
34
Deep Residual Learning for Image Recognition
Kaiming He, Xiangyu Zhang, Shaoqing Ren et al. · 2016 · 222.1K citations · Full text
Going deeper with convolutions
Christian Szegedy, Wei Liu, Yangqing Jia et al. · 2015 · 46.8K citations
MobileNets: Efficient Convolutional Neural Networks for Mobile Vision Applications
Andrew Howard, Menglong Zhu, Bo Chen et al. · arXiv (Cornell University) · 2017 · 9.9K citations · Full text