Bottleneck Analysis of Dynamic Graph Neural Network Inference on CPU and GPU
Not on the map. Why?
Abstract (arXiv)
Dynamic graph neural network (DGNN) is becoming increasingly popular because of its widespread use in capturing dynamic features in the real world. A variety of dynamic graph neural networks designed from algorithmic perspectives have succeeded in incorporating temporal information into graph processing. Despite the promising algorithmic performance, deploying DGNNs on hardware presents additional challenges due to the model complexity, diversity, and the nature of the time dependency. Meanwhile, the differences between DGNNs and static graph neural networks make hardware-related optimizations for static graph neural networks unsuitable for DGNNs. In this paper, we select eight prevailing DGNNs with different characteristics and profile them on both CPU and GPU. The profiling results are summarized and analyzed, providing in-depth insights into the bottlenecks of DGNNs on hardware and identifying potential optimization opportunities for future DGNN acceleration. Followed by a comprehensive survey, we provide a detailed analysis of DGNN performance bottlenecks on hardware, including temporal data dependency, workload imbalance, data movement, and GPU warm-up. We suggest several optimizations from both software and hardware perspectives. This paper is the first to provide an in-depth analysis of the hardware performance of DGNN Code is available at https://github.com/sharc-lab/DGNN_analysis.
Key facts. Cited by 2 papers in the map (data through 14 Oct 2026). Builds on 26 of its 54 references that are in the map. First posted on arXiv 8 Oct 2022.
Builds on
26 of 54 references.
References · 26
- 2019A Comprehensive Survey on Graph Neural Networkscited by 2,919
- 2018Graph Neural Networks: A Review of Methods and Applicationscited by 1,576
- 2014On the Properties of Neural Machine Translation: Encoder-Decoder Approachescited by 1,304
- 2020Image Segmentation Using Deep Learning: A Surveycited by 660
- 2020Temporal Graph Networks for Deep Learning on Dynamic Graphscited by 530
- 2020Inductive Representation Learning on Temporal Graphscited by 462
- 2019Predicting Dynamic Embedding Trajectory in Temporal Interaction Networkscited by 448
- 2019EvolveGCN: Evolving Graph Convolutional Networks for Dynamic Graphscited by 447
- 2018Neural Relational Inference for Interacting Systemscited by 339
- 2017Recent Trends in Deep Learning Based Natural Language Processingcited by 263
- 2019Time2Vec: Learning a Vector Representation of Timecited by 213
- 2020Distilling Knowledge from Graph Convolutional Networkscited by 129
14 more
- 2019Deep Learning on Image Denoising: An overviewcited by 128
- 2020Foundations and modelling of dynamic networks using Dynamic Graph Neural Networks: A surveycited by 124
- 2020HyGCN: A GCN Accelerator with Hybrid Architecturecited by 74
- 2022TGL: A General Framework for Temporal GNN Training on Billion-Scale Graphscited by 61
- 2020Binary Graph Neural Networkscited by 34
- 2019GraphACT: Accelerating GCN Training on CPU-FPGA Heterogeneous Platformscited by 31
- 2020Accelerating Graph Sampling for Graph Machine Learning using GPUscited by 29
- 2022I-GCN: A Graph Convolutional Network Accelerator with Runtime Locality Enhancement through Islandizationcited by 21
- 2022Model-Architecture Co-Design for High Performance Temporal GNN Inference on FPGAcited by 12
- 2019Learning Temporal Attention in Dynamic Graphs with Bilinear Interactionscited by 11
- 2022Low-latency Mini-batch GNN Inference on CPU-FPGA Heterogeneous Platformcited by 7
- 2021GNNSampler: Bridging the Gap between Sampling Algorithms of GNN and Hardwarecited by 6
- 2021VersaGNN: a Versatile accelerator for Graph neural networkscited by 3
- 2021Geometric learning of the conformational dynamics of molecules using dynamic graph neural networkscited by 2
Built on by
Citing papers per year: 2023: 1; 2024: 1.
2 papers.