├── _config.yml ├── requirements.txt ├── res ├── contribute.md ├── head.md ├── papers │ ├── 04-compiler.json │ ├── 01-survey.json │ ├── 09-training-accelerator.json │ ├── 03-kernel.json │ ├── 02-library.json │ ├── 07-quantization.json │ ├── 08-dataloader.json │ ├── 06-scaling-graph.json │ ├── 10-inference-accelerator.json │ └── 05-distributed-training.json └── library.json ├── src ├── rebuild_citation.py ├── update_citation.py └── update_readme.py ├── .github ├── workflows │ ├── citation.yml │ └── rebuild.yml └── citation │ └── citation.json └── README.md /_config.yml: -------------------------------------------------------------------------------- 1 | markdown: GFM 2 | -------------------------------------------------------------------------------- /requirements.txt: -------------------------------------------------------------------------------- 1 | bs4 2 | requests 3 | datetime 4 | -------------------------------------------------------------------------------- /res/contribute.md: -------------------------------------------------------------------------------- 1 | ## Contribute 2 | 3 | We welcome contributions to [this repository](https://github.com/chwan1016/awesome-gnn-systems). To add new papers to this list, please update JSON files under `./res/papers/`. Our bots will update the paper list in `README.md` automatically. The citations of newly added papers will be updated within one day. 4 | -------------------------------------------------------------------------------- /res/head.md: -------------------------------------------------------------------------------- 1 | # Awesome Graph Neural Network Systems [![Awesome](https://cdn.rawgit.com/sindresorhus/awesome/d7305f38d29fed78fa85652e3a63e154dd8e8829/media/badge.svg)](https://github.com/sindresorhus/awesome) 2 | 3 | A list of awesome systems for graph neural network (GNN). If you have any comment, please create an [issue](https://github.com/chwan1016/awesome-gnn-systems/issues) or [pull request](https://github.com/chwan1016/awesome-gnn-systems/pulls). 4 | -------------------------------------------------------------------------------- /src/rebuild_citation.py: -------------------------------------------------------------------------------- 1 | import json 2 | import os 3 | 4 | path = 'res/papers' 5 | full_papers = set() 6 | 7 | for file in os.listdir(path): 8 | if not file.endswith(".json"): 9 | continue 10 | with open(os.path.join(path, file), 'r') as f: 11 | papers = json.load(f)['paper'] 12 | full_papers = full_papers | {paper['name'] for paper in papers} 13 | 14 | with open('.github/citation/citation.json', 'r') as f: 15 | table = json.load(f) 16 | 17 | for key in list(table.keys()): 18 | if key not in full_papers: 19 | table.pop(key) 20 | 21 | for paper in full_papers: 22 | if paper not in table: 23 | table[paper] = {'citation': 0, 'last update': '2016-04-08'} 24 | 25 | with open('.github/citation/citation.json', 'w') as f: 26 | json.dump(table, f) 27 | -------------------------------------------------------------------------------- /res/papers/04-compiler.json: -------------------------------------------------------------------------------- 1 | { 2 | "title": "GNN Compilers", 3 | "paper": [ 4 | { 5 | "venue": "MLSys 2022", 6 | "name": "Graphiler: Optimizing Graph Neural Networks with Message Passing Data Flow Graph", 7 | "affiliation": "ShanghaiTech", 8 | "link": "https://proceedings.mlsys.org/paper/2022/file/a87ff679a2f3e71d9181a67b7542122c-Paper.pdf", 9 | "source": "https://github.com/xiezhq-hermann/graphiler" 10 | }, 11 | { 12 | "venue": "EuroSys 2021", 13 | "name": "Seastar: Vertex-Centric Programming for Graph Neural Networks", 14 | "affiliation": "CUHK", 15 | "link": "http://www.cse.cuhk.edu.hk/~jcheng/papers/seastar_eurosys21.pdf" 16 | }, 17 | { 18 | "venue": "SC 2020", 19 | "name": "FeatGraph: A Flexible and Efficient Backend for Graph Neural Network Systems", 20 | "affiliation": "Cornell", 21 | "link": "https://arxiv.org/pdf/2008.11359.pdf", 22 | "source": "https://github.com/dglai/FeatGraph" 23 | } 24 | ] 25 | } 26 | -------------------------------------------------------------------------------- /.github/workflows/citation.yml: -------------------------------------------------------------------------------- 1 | on: 2 | schedule: 3 | - cron: "0 0 * * *" 4 | workflow_dispatch: 5 | 6 | concurrency: 7 | group: ${{ github.ref }} 8 | 9 | jobs: 10 | build: 11 | runs-on: ubuntu-latest 12 | steps: 13 | - uses: actions/checkout@v3 14 | - name: Setup Python 15 | uses: actions/setup-python@v4 16 | with: 17 | python-version: '3.x' # Version range or exact version of a Python version to use, using SemVer's version range syntax 18 | - name: Cache pip 19 | uses: actions/cache@v4 20 | with: 21 | # This path is specific to Ubuntu 22 | path: ~/.cache/pip 23 | # Look to see if there is a cache hit for the corresponding requirements file 24 | key: ${{ runner.os }}-pip-${{ hashFiles('requirements.txt') }} 25 | restore-keys: | 26 | ${{ runner.os }}-pip- 27 | ${{ runner.os }}- 28 | - name: Install dependencies 29 | run: pip install -r requirements.txt 30 | - name: Update citation 31 | run: | 32 | python src/update_citation.py 33 | python src/update_readme.py 34 | - name: Commit and push if needed 35 | run: |- 36 | git diff 37 | git config --global user.email "citation-bot@example.com" 38 | git config --global user.name "citation-bot" 39 | git add -A 40 | git commit -m "[Citation-Bot] update citation automatically" || exit 0 41 | git push 42 | -------------------------------------------------------------------------------- /res/papers/01-survey.json: -------------------------------------------------------------------------------- 1 | { 2 | "title": "Survey Papers", 3 | "paper": [ 4 | { 5 | "name": "Distributed Graph Neural Network Training: A Survey", 6 | "venue": "CSUR 2024", 7 | "affiliation": "BUPT", 8 | "link": "https://dl.acm.org/doi/10.1145/3648358" 9 | }, 10 | { 11 | "name": "A Comprehensive Survey on Distributed Training of Graph Neural Networks", 12 | "venue": "Proceedings of the IEEE 2023", 13 | "affiliation": "Chinese Academy of Sciences", 14 | "link": "https://ieeexplore.ieee.org/abstract/document/10348966" 15 | }, 16 | { 17 | "name": "A Survey on Graph Neural Network Acceleration: Algorithms, Systems, and Customized Hardware", 18 | "venue": "arXiv 2023", 19 | "affiliation": "UCLA", 20 | "link": "https://arxiv.org/abs/2306.14052" 21 | }, 22 | { 23 | "name": "Parallel and Distributed Graph Neural Networks: An In-Depth Concurrency Analysis", 24 | "venue": "arXiv 2022", 25 | "affiliation": "ETHZ", 26 | "link": "https://arxiv.org/abs/2205.09702" 27 | }, 28 | { 29 | "name": "Computing Graph Neural Networks: A Survey from Algorithms to Accelerators", 30 | "venue": "CSUR 2022", 31 | "affiliation": "UPC", 32 | "link": "https://dl.acm.org/doi/10.1145/3477141" 33 | } 34 | ] 35 | } 36 | -------------------------------------------------------------------------------- /.github/workflows/rebuild.yml: -------------------------------------------------------------------------------- 1 | on: 2 | push: 3 | branches: [ main ] 4 | pull_request: 5 | branches: [ main ] 6 | 7 | concurrency: 8 | group: ${{ github.ref }} 9 | cancel-in-progress: true 10 | 11 | jobs: 12 | build: 13 | runs-on: ubuntu-latest 14 | steps: 15 | - uses: actions/checkout@v3 16 | - name: Setup Python 17 | uses: actions/setup-python@v4 18 | with: 19 | python-version: '3.x' # Version range or exact version of a Python version to use, using SemVer's version range syntax 20 | - name: Cache pip 21 | uses: actions/cache@v4 22 | with: 23 | # This path is specific to Ubuntu 24 | path: ~/.cache/pip 25 | # Look to see if there is a cache hit for the corresponding requirements file 26 | key: ${{ runner.os }}-pip-${{ hashFiles('requirements.txt') }} 27 | restore-keys: | 28 | ${{ runner.os }}-pip- 29 | ${{ runner.os }}- 30 | - name: Install dependencies 31 | run: pip install -r requirements.txt 32 | - name: Rebuild citation 33 | run: | 34 | python src/rebuild_citation.py 35 | python src/update_readme.py 36 | - name: Commit and push if needed 37 | if: ${{ github.event_name == 'push' }} 38 | run: |- 39 | git diff 40 | git config --global user.email "rebuild-bot@example.com" 41 | git config --global user.name "rebuild-bot" 42 | git add -A 43 | git commit -m "[Rebuild-Bot] rebuild citation automatically" || exit 0 44 | git push 45 | -------------------------------------------------------------------------------- /res/library.json: -------------------------------------------------------------------------------- 1 | [ 2 | { 3 | "name": "PyG: Graph Neural Network Library for PyTorch", 4 | "source": "https://github.com/rusty1s/pytorch_geometric" 5 | }, 6 | { 7 | "name": "DGL: Python Package Built to Ease Deep Learning on Graph", 8 | "source": "https://github.com/dmlc/dgl" 9 | }, 10 | { 11 | "name": "Graph Nets: Build Graph Nets in Tensorflow", 12 | "source": "https://github.com/deepmind/graph_nets" 13 | }, 14 | { 15 | "name": "Euler: A Distributed Graph Deep Learning Framework", 16 | "source": "https://github.com/alibaba/euler" 17 | }, 18 | { 19 | "name": "StellarGraph: Machine Learning on Graphs", 20 | "source": "https://github.com/stellargraph/stellargraph" 21 | }, 22 | { 23 | "name": "Spektral: Graph Neural Networks with Keras and Tensorflow 2", 24 | "source": "https://github.com/danielegrattarola/spektral" 25 | }, 26 | { 27 | "name": "PGL: An Efficient and Flexible Graph Learning Framework Based on PaddlePaddle", 28 | "source": "https://github.com/PaddlePaddle/PGL" 29 | }, 30 | { 31 | "name": "CogDL: An Extensive Toolkit for Deep Learning on Graphs", 32 | "source": "https://github.com/THUDM/cogdl" 33 | }, 34 | { 35 | "name": "DIG: A Turnkey Library for Diving into Graph Deep Learning Research", 36 | "source": "https://github.com/divelab/DIG" 37 | }, 38 | { 39 | "name": "Jraph: A Graph Neural Network Library in Jax", 40 | "source": "https://github.com/deepmind/jraph" 41 | }, 42 | { 43 | "name": "Graph-Learn: An Industrial Graph Neural Network Framework", 44 | "source": "https://github.com/alibaba/graph-learn" 45 | }, 46 | { 47 | "name": "DeepGNN: a Framework for Training Machine Learning Models on Large Scale Graph Data", 48 | "source": "https://github.com/microsoft/DeepGNN" 49 | } 50 | ] 51 | -------------------------------------------------------------------------------- /src/update_citation.py: -------------------------------------------------------------------------------- 1 | import json 2 | import requests 3 | from bs4 import BeautifulSoup 4 | import datetime 5 | import random 6 | import time 7 | 8 | def get_citation(name): 9 | prefix = "https://scholar.google.com/scholar?q=allintitle:" 10 | name = name.replace('^', '') 11 | name = name.replace('$', '') 12 | scholar_link = prefix + name 13 | 14 | page = requests.get(scholar_link).text 15 | soup = BeautifulSoup(page, "html.parser") 16 | 17 | if "robot" in page: 18 | raise ConnectionRefusedError 19 | 20 | cite = soup.find_all("a", href=lambda value: value and value.startswith("/scholar?cites=")) 21 | s = cite[0].text if len(cite) >= 1 else "0" 22 | s = int(''.join(c for c in s if c.isdigit())) 23 | 24 | return s 25 | 26 | today = datetime.date.today() 27 | n_upd = random.randint(20, 30) 28 | print('n_upd =', n_upd) 29 | 30 | with open('.github/citation/citation.json', 'r') as f: 31 | table = json.load(f) 32 | 33 | sorted_table = sorted(table.items(), key=lambda item: item[1]['last update']) 34 | 35 | for i in range(n_upd): 36 | name = sorted_table[i][0] 37 | item = sorted_table[i][1] 38 | try: 39 | cite = get_citation(name) 40 | item['citation'] = cite 41 | item['last update'] = today.strftime("%Y-%m-%d") 42 | print('"' + name + '" updated,', 'citation =', cite) 43 | if i < n_upd - 1: 44 | time.sleep(random.randint(10, 80)) 45 | except ConnectionRefusedError: 46 | print('updating "' + name + '" failed') 47 | break 48 | 49 | table = dict(sorted_table) 50 | table = dict(sorted(table.items(), key=lambda item: item[1]['last update'])) 51 | 52 | with open('.github/citation/citation.json', 'w') as f: 53 | json.dump(table, f) 54 | -------------------------------------------------------------------------------- /res/papers/09-training-accelerator.json: -------------------------------------------------------------------------------- 1 | { 2 | "title": "GNN Training Accelerators", 3 | "paper": [ 4 | { 5 | "venue": "ISCA 2022", 6 | "name": "Graphite: Optimizing Graph Neural Networks on CPUs Through Cooperative Software-Hardware Techniques", 7 | "affiliation": "UIUC", 8 | "link": "http://iacoma.cs.uiuc.edu/iacoma-papers/isca22.pdf" 9 | }, 10 | { 11 | "venue": "ISCA 2022", 12 | "name": "Hyperscale FPGA-as-a-service architecture for large-scale distributed graph neural network", 13 | "affiliation": "Alibaba", 14 | "link": "https://dl.acm.org/doi/abs/10.1145/3470496.3527439" 15 | }, 16 | { 17 | "venue": "arXiv 2021", 18 | "name": "GCNear: A Hybrid Architecture for Efficient GCN Training with Near-Memory Processing", 19 | "affiliation": "PKU", 20 | "link": "https://arxiv.org/abs/2111.00680" 21 | }, 22 | { 23 | "venue": "DATE 2021", 24 | "name": "ReGraphX: NoC-enabled 3D Heterogeneous ReRAM Architecture for Training Graph Neural Networks", 25 | "affiliation": "WSU", 26 | "link": "https://arxiv.org/abs/2102.07959" 27 | }, 28 | { 29 | "venue": "TCAD 2021", 30 | "name": "Rubik: A Hierarchical Architecture for Efficient Graph Learning", 31 | "affiliation": "Chinese Academy of Sciences", 32 | "link": "https://arxiv.org/pdf/2009.12495.pdf" 33 | }, 34 | { 35 | "venue": "FPGA 2020", 36 | "name": "GraphACT: Accelerating GCN Training on CPU-FPGA Heterogeneous Platforms", 37 | "affiliation": "USC", 38 | "link": "https://dl.acm.org/doi/pdf/10.1145/3373087.3375312", 39 | "source": "https://github.com/GraphSAINT/GraphACT" 40 | } 41 | ] 42 | } 43 | -------------------------------------------------------------------------------- /res/papers/03-kernel.json: -------------------------------------------------------------------------------- 1 | { 2 | "title": "GNN Kernels", 3 | "paper": [ 4 | { 5 | "venue": "MLSys 2022", 6 | "name": "Understanding GNN Computational Graph: A Coordinated Computation, IO, and Memory Perspective", 7 | "affiliation": "THU", 8 | "link": "https://proceedings.mlsys.org/paper/2022/file/9a1158154dfa42caddbd0694a4e9bdc8-Paper.pdf", 9 | "source": "https://github.com/dgSPARSE/dgNN" 10 | }, 11 | { 12 | "venue": "HPDC 2022", 13 | "name": "TLPGNN: A Lightweight Two-Level Parallelism Paradigm for Graph Neural Network Computation on GPU", 14 | "affiliation": "GW", 15 | "link": "https://dl.acm.org/doi/abs/10.1145/3502181.3531467" 16 | }, 17 | { 18 | "venue": "IPDPS 2021", 19 | "name": "FusedMM: A Unified SDDMM-SpMM Kernel for Graph Embedding and Graph Neural Networks", 20 | "affiliation": "Indiana University Bloomington", 21 | "link": "https://arxiv.org/abs/2011.06391", 22 | "source": "https://github.com/HipGraph/FusedMM" 23 | }, 24 | { 25 | "venue": "SC 2020", 26 | "name": "GE-SpMM: General-purpose Sparse Matrix-Matrix Multiplication on GPUs for Graph Neural Networks", 27 | "affiliation": "THU", 28 | "link": "https://arxiv.org/abs/2007.03179", 29 | "source": "https://github.com/hgyhungry/ge-spmm" 30 | }, 31 | { 32 | "venue": "ICCAD 2020", 33 | "name": "fuseGNN: Accelerating Graph Convolutional Neural Network Training on GPGPU", 34 | "affiliation": "UCSB", 35 | "link": "https://seal.ece.ucsb.edu/sites/default/files/publications/fusegcn_camera_ready_.pdf", 36 | "source": "https://github.com/apuaaChen/gcnLib" 37 | }, 38 | { 39 | "venue": "IPDPS 2020", 40 | "name": "PCGCN: Partition-Centric Processing for Accelerating Graph Convolutional Network", 41 | "affiliation": "PKU", 42 | "link": "https://ieeexplore.ieee.org/document/9139807" 43 | } 44 | ] 45 | } 46 | -------------------------------------------------------------------------------- /res/papers/02-library.json: -------------------------------------------------------------------------------- 1 | { 2 | "title": "GNN Libraries", 3 | "paper": [ 4 | { 5 | "venue": "JMLR 2021", 6 | "name": "DIG: A Turnkey Library for Diving into Graph Deep Learning Research", 7 | "affiliation": "TAMU", 8 | "link": "https://arxiv.org/abs/2103.12608", 9 | "source": "https://github.com/divelab/DIG" 10 | }, 11 | { 12 | "venue": "arXiv 2021", 13 | "name": "CogDL: A Toolkit for Deep Learning on Graphs", 14 | "affiliation": "THU", 15 | "link": "https://arxiv.org/abs/2103.00959", 16 | "source": "https://github.com/THUDM/cogdl" 17 | }, 18 | { 19 | "venue": "CIM 2021", 20 | "name": "Graph Neural Networks in TensorFlow and Keras with Spektral", 21 | "affiliation": "Università della Svizzera italiana", 22 | "link": "https://arxiv.org/abs/2006.12138", 23 | "source": "https://github.com/danielegrattarola/spektral" 24 | }, 25 | { 26 | "venue": "arXiv 2019", 27 | "name": "Deep Graph Library: A Graph-Centric, Highly-Performant Package for Graph Neural Networks", 28 | "affiliation": "AWS", 29 | "link": "https://arxiv.org/abs/1909.01315", 30 | "source": "https://github.com/dmlc/dgl" 31 | }, 32 | { 33 | "venue": "VLDB 2019", 34 | "name": "AliGraph: A Comprehensive Graph Neural Network Platform", 35 | "affiliation": "Alibaba", 36 | "link": "https://dl.acm.org/doi/pdf/10.14778/3352063.3352127", 37 | "source": "https://github.com/alibaba/graph-learn" 38 | }, 39 | { 40 | "venue": "arXiv 2019", 41 | "name": "Fast Graph Representation Learning with PyTorch Geometric", 42 | "affiliation": "TU Dortmund University", 43 | "link": "https://arxiv.org/abs/1903.02428", 44 | "source": "https://github.com/rusty1s/pytorch_geometric" 45 | }, 46 | { 47 | "venue": "arXiv 2018", 48 | "name": "Relational Inductive Biases, Deep Learning, and Graph Networks", 49 | "affiliation": "DeepMind", 50 | "link": "https://arxiv.org/abs/1806.01261", 51 | "source": "https://github.com/deepmind/graph_nets" 52 | } 53 | ] 54 | } 55 | -------------------------------------------------------------------------------- /res/papers/07-quantization.json: -------------------------------------------------------------------------------- 1 | { 2 | "title": "Quantized GNNs", 3 | "paper": [ 4 | { 5 | "venue": "Neurocomputing 2022", 6 | "name": "EPQuant: A Graph Neural Network Compression Approach Based on Product Quantization", 7 | "affiliation": "ZJU", 8 | "link": "https://www.sciencedirect.com/science/article/abs/pii/S0925231222008293", 9 | "source": "https://github.com/Lyun-Huang/EPQuant" 10 | }, 11 | { 12 | "venue": "ICLR 2022", 13 | "name": "EXACT: Scalable Graph Neural Networks Training via Extreme Activation Compression", 14 | "affiliation": "Rice", 15 | "link": "https://openreview.net/pdf?id=vkaMaq95_rX", 16 | "source": "https://github.com/warai-0toko/Exact" 17 | }, 18 | { 19 | "venue": "PPoPP 2022", 20 | "name": "QGTC: Accelerating Quantized Graph Neural Networks via GPU Tensor Core", 21 | "affiliation": "UCSB", 22 | "link": "https://arxiv.org/abs/2111.09547", 23 | "source": "https://github.com/YukeWang96/PPoPP22_QGTC" 24 | }, 25 | { 26 | "venue": "CVPR 2021", 27 | "name": "Binary Graph Neural Networks", 28 | "affiliation": "ICL", 29 | "link": "https://arxiv.org/abs/2012.15823", 30 | "source": "https://github.com/mbahri/binary_gnn" 31 | }, 32 | { 33 | "venue": "CVPR 2021", 34 | "name": "Bi-GCN: Binary Graph Convolutional Network", 35 | "affiliation": "Beihang University", 36 | "link": "https://openaccess.thecvf.com/content/CVPR2021/html/Wang_Bi-GCN_Binary_Graph_Convolutional_Network_CVPR_2021_paper.html", 37 | "source": "https://github.com/bywmm/Bi-GCN" 38 | }, 39 | { 40 | "venue": "EuroMLSys 2021", 41 | "name": "Learned Low Precision Graph Neural Networks", 42 | "affiliation": "Cambridge", 43 | "link": "https://arxiv.org/abs/2009.09232" 44 | }, 45 | { 46 | "venue": "World Wide Web 2021", 47 | "name": "Binarized Graph Neural Network", 48 | "affiliation": "UTS", 49 | "link": "https://arxiv.org/pdf/2004.11147.pdf" 50 | }, 51 | { 52 | "venue": "ICLR 2021", 53 | "name": "Degree-Quant: Quantization-Aware Training for Graph Neural Networks", 54 | "affiliation": "Cambridge", 55 | "link": "https://arxiv.org/abs/2008.05000", 56 | "source": "https://github.com/camlsys/degree-quant" 57 | }, 58 | { 59 | "venue": "ICTAI 2020", 60 | "name": "SGQuant: Squeezing the Last Bit on Graph Neural Networks with Specialized Quantization", 61 | "affiliation": "UCSB", 62 | "link": "https://ieeexplore.ieee.org/abstract/document/9288186/", 63 | "source": "https://github.com/YukeWang96/SGQuant" 64 | } 65 | ] 66 | } 67 | -------------------------------------------------------------------------------- /res/papers/08-dataloader.json: -------------------------------------------------------------------------------- 1 | { 2 | "title": "GNN Dataloaders", 3 | "paper": [ 4 | { 5 | "venue": "VLDB 2024", 6 | "name": "Accelerating Sampling and Aggregation Operations in GNN Frameworks with GPU Initiated Direct Storage Accesses", 7 | "affiliation": "UIUC", 8 | "link": "https://arxiv.org/abs/2306.16384v2", 9 | "source": "https://github.com/jeongminpark417/GIDS" 10 | }, 11 | { 12 | "venue": "NSDI 2023", 13 | "name": "BGL: GPU-Efficient GNN Training by Optimizing Graph Data I/O and Preprocessing", 14 | "affiliation": "ByteDance", 15 | "link": "https://arxiv.org/abs/2112.08541" 16 | }, 17 | { 18 | "venue": "MLSys 2022", 19 | "name": "Accelerating Training and Inference of Graph Neural Networks with Fast Sampling and Pipelining", 20 | "affiliation": "MIT", 21 | "link": "https://proceedings.mlsys.org/paper/2022/file/35f4a8d465e6e1edc05f3d8ab658c551-Paper.pdf", 22 | "source": "https://github.com/MITIBMxGraph/SALIENT" 23 | }, 24 | { 25 | "venue": "EuroSys 2022", 26 | "name": "GNNLab: A Factored System for Sample-based GNN Training over GPUs", 27 | "affiliation": "SJTU", 28 | "link": "https://dl.acm.org/doi/abs/10.1145/3492321.3519557", 29 | "source": "https://github.com/SJTU-IPADS/gnnlab" 30 | }, 31 | { 32 | "venue": "KDD 2021", 33 | "name": "Global Neighbor Sampling for Mixed CPU-GPU Training on Giant Graphs", 34 | "affiliation": "UCLA", 35 | "link": "https://arxiv.org/pdf/2106.06150.pdf" 36 | }, 37 | { 38 | "venue": "PPoPP 2021", 39 | "name": "Understanding and Bridging the Gaps in Current GNN Performance Optimizations", 40 | "affiliation": "THU", 41 | "link": "https://dl.acm.org/doi/pdf/10.1145/3437801.3441585", 42 | "source": "https://github.com/xxcclong/GNN-Computing" 43 | }, 44 | { 45 | "venue": "VLDB 2021", 46 | "name": "Large Graph Convolutional Network Training with GPU-Oriented Data Communication Architecture", 47 | "affiliation": "UIUC", 48 | "link": "https://arxiv.org/abs/2103.03330", 49 | "source": "https://github.com/K-Wu/pytorch-direct_dgl" 50 | }, 51 | { 52 | "venue": "TPDS 2021", 53 | "name": "Efficient Data Loader for Fast Sampling-Based GNN Training on Large Graphs", 54 | "affiliation": "USTC", 55 | "link": "https://gnnsys.github.io/papers/GNNSys21_paper_8.pdf", 56 | "source": "https://github.com/zhiqi-0/PaGraph" 57 | }, 58 | { 59 | "venue": "SoCC 2020", 60 | "name": "PaGraph: Scaling GNN Training on Large Graphs via Computation-aware Caching", 61 | "affiliation": "USTC", 62 | "link": "https://dl.acm.org/doi/pdf/10.1145/3419111.3421281", 63 | "source": "https://github.com/zhiqi-0/PaGraph" 64 | }, 65 | { 66 | "venue": "arXiv 2019", 67 | "name": "TigerGraph: A Native MPP Graph Database", 68 | "affiliation": "UCSD", 69 | "link": "https://arxiv.org/pdf/1901.08248.pdf" 70 | } 71 | ] 72 | } 73 | -------------------------------------------------------------------------------- /res/papers/06-scaling-graph.json: -------------------------------------------------------------------------------- 1 | { 2 | "title": "Training Systems for Scaling Graphs", 3 | "paper": [ 4 | { 5 | "venue": "ICDE 2025", 6 | "name": "Hyperion: Optimizing SSD Access is All You Need to Enable Cost-efficient Out-of-core GNN Training", 7 | "affiliation":"ZJU", 8 | "link": "https://wangzeke.github.io/doc/Hyperion-ICDE25.pdf", 9 | "source":"https://github.com/RC4ML/Hyperion" 10 | }, 11 | { 12 | "venue": "DaMoN 2024", 13 | "name": "In situ neighborhood sampling for large-scale GNN training", 14 | "affiliation":"Boston University", 15 | "link": "https://dl.acm.org/doi/10.1145/3662010.3663443", 16 | "source":"https://github.com/CASP-Systems-BU/damon24-gnn-in-situ-sampling/" 17 | }, 18 | { 19 | "venue": "HPCA 2024", 20 | "name": "BeaconGNN: Large-Scale GNN Acceleration with Out-of-Order Streaming In-Storage Computing", 21 | "affiliation": "UCLA", 22 | "link": "https://ieeexplore.ieee.org/abstract/document/10476427" 23 | }, 24 | { 25 | "venue": "EuroSys 2023", 26 | "name": "MariusGNN: Resource-Efficient Out-of-Core Training of Graph Neural Networks", 27 | "affiliation": "UW–Madison", 28 | "link": "https://arxiv.org/abs/2202.02365", 29 | "source": "https://github.com/marius-team/marius" 30 | }, 31 | { 32 | "venue": "VLDB 2022", 33 | "name": "ByteGNN: Efficient Graph Neural Network Training at Large Scale", 34 | "affiliation": "ByteDance", 35 | "link": "https://dl.acm.org/doi/abs/10.14778/3514061.3514069" 36 | }, 37 | { 38 | "venue": "VLDB 2022", 39 | "name": "Ginex: SSD-enabled Billion-scale Graph Neural Network Training on a Single Machine via Provably Optimal In-memory Caching", 40 | "affiliation": "Seoul National University", 41 | "link": "https://dl.acm.org/doi/10.14778/3551793.3551819", 42 | "source": "https://github.com/SNU-ARC/Ginex" 43 | }, 44 | { 45 | "venue": "ISCA 2022", 46 | "name":"SmartSAGE: Training Large-scale Graph Neural Networks using In-Storage Processing Architectures", 47 | "affiliation": "KAIST", 48 | "link":"https://dl.acm.org/doi/10.1145/3470496.3527391" 49 | }, 50 | { 51 | "venue": "ICML 2022", 52 | "name": "GraphFM: Improving Large-Scale GNN Training via Feature Momentum", 53 | "affiliation": "TAMU", 54 | "link": "https://arxiv.org/abs/2206.07161", 55 | "source": "https://github.com/divelab/DIG/tree/dig-stable/dig/lsgraph" 56 | }, 57 | { 58 | "venue": "ICML 2021", 59 | "name": "GNNAutoScale: Scalable and Expressive Graph Neural Networks via Historical Embeddings", 60 | "affiliation": "TU Dortmund University", 61 | "link": "https://arxiv.org/abs/2106.05609", 62 | "source": "https://github.com/rusty1s/pyg_autoscale" 63 | }, 64 | { 65 | "venue": "OSDI 2021", 66 | "name": "GNNAdvisor: An Adaptive and Efficient Runtime System for GNN Acceleration on GPUs", 67 | "affiliation": "UCSB", 68 | "link": "https://www.usenix.org/system/files/osdi21-wang-yuke.pdf", 69 | "source": "https://github.com/YukeWang96/OSDI21_AE" 70 | } 71 | ] 72 | } 73 | -------------------------------------------------------------------------------- /src/update_readme.py: -------------------------------------------------------------------------------- 1 | import json 2 | import os 3 | 4 | def load_head(file): 5 | with open("res/head.md", 'r') as f: 6 | for line in f: 7 | file.write(line) 8 | 9 | def load_content(file): 10 | file.write("## Contents\n") 11 | file.write("- [Open Source Libraries](#open-source-libraries)\n") 12 | file.write("- [Papers](#papers)\n") 13 | 14 | path = 'res/papers' 15 | 16 | for pf in sorted(os.listdir(path)): 17 | if not pf.endswith(".json"): 18 | continue 19 | with open(os.path.join(path, pf), 'r') as f: 20 | title = json.load(f)['title'] 21 | file.write(" - [{}](#{})\n".format(title, title.lower().replace(' ', '-'))) 22 | 23 | file.write("- [Contribute](#contribute)\n") 24 | 25 | def load_lib(file): 26 | file.write("## Open Source Libraries\n") 27 | with open("res/library.json", 'r') as f: 28 | libraries = json.load(f) 29 | for lib in libraries: 30 | name = lib['name'] 31 | source = lib['source'] 32 | repo = source.split('github.com/', 1)[1] 33 | file.write("- [{}]({}) ![GitHub stars](https://img.shields.io/github/stars/{}.svg?logo=github&label=Stars)\n".format(name, source, repo)) 34 | 35 | def load_paper(file): 36 | file.write("## Papers\n") 37 | 38 | with open('.github/citation/citation.json', 'r') as f: 39 | citation = json.load(f) 40 | 41 | path = 'res/papers' 42 | for pf in sorted(os.listdir(path)): 43 | if not pf.endswith(".json"): 44 | continue 45 | with open(os.path.join(path, pf), 'r') as f: 46 | table = json.load(f) 47 | title = table['title'] 48 | file.write("### {}\n".format(title)) 49 | file.write("| Venue | Title | Affiliation |       Link       |   Source   |\n") 50 | file.write("| :---: | :---: | :---------: | :---: | :----: |\n") 51 | for paper in table['paper']: 52 | file.write("|{}|".format(paper['venue'])) 53 | file.write("{}|".format(paper['name'])) 54 | file.write("{}|".format(paper['affiliation'])) 55 | if 'link' in paper: 56 | file.write(" [[paper]]({})".format(paper['link'])) 57 | cite = citation[paper['name']]['citation'] 58 | if cite > 10000: 59 | cite = int(cite / 1000 + 0.5) 60 | cite = '{}k'.format(cite) 61 | elif cite > 1000: 62 | cite = int(cite / 100 + 0.5) / 10 63 | cite = '{}k'.format(cite) 64 | file.write("![Scholar citations](https://img.shields.io/badge/Citations-{}-_.svg?logo=google-scholar&labelColor=4f4f4f&color=3388ee)".format(cite)) 65 | file.write("|") 66 | if 'source' in paper: 67 | source = paper['source'] 68 | file.write(" [[code]]({})".format(source)) 69 | repo = source.split('github.com/', 1)[1] 70 | if repo.count('/') == 1: 71 | file.write("![GitHub stars](https://img.shields.io/github/stars/{}.svg?logo=github&label=Stars)".format(repo)) 72 | file.write("|\n") 73 | 74 | def load_contribute(file): 75 | with open("res/contribute.md", 'r') as f: 76 | for line in f: 77 | file.write(line) 78 | 79 | 80 | file = open("README.md", 'w') 81 | 82 | load_head(file) 83 | load_content(file) 84 | load_lib(file) 85 | load_paper(file) 86 | load_contribute(file) 87 | 88 | file.close() 89 | -------------------------------------------------------------------------------- /res/papers/10-inference-accelerator.json: -------------------------------------------------------------------------------- 1 | { 2 | "title": "GNN Inference Accelerators", 3 | "paper": [ 4 | { 5 | "venue": "JAIHC 2022", 6 | "name": "DRGN: a dynamically reconfigurable accelerator for graph neural networks", 7 | "affiliation": "XJTU", 8 | "link": "https://link.springer.com/article/10.1007/s12652-022-04402-x" 9 | }, 10 | { 11 | "venue": "DAC 2022", 12 | "name": "GNNIE: GNN Inference Engine with Load-balancing and Graph-specific Caching", 13 | "affiliation": "UMN", 14 | "link": "https://arxiv.org/abs/2105.10554" 15 | }, 16 | { 17 | "venue": "IPDPS 2022", 18 | "name": "Understanding the Design Space of Sparse/Dense Multiphase Dataflows for Mapping Graph Neural Networks on Spatial Accelerators", 19 | "affiliation": "GaTech", 20 | "link": "https://arxiv.org/abs/2103.07977", 21 | "source": "https://github.com/stonne-simulator/omega" 22 | }, 23 | { 24 | "venue": "arXiv 2022", 25 | "name": "FlowGNN: A Dataflow Architecture for Universal Graph Neural Network Inference via Multi-Queue Streaming", 26 | "affiliation": "GaTech", 27 | "link": "https://arxiv.org/abs/2204.13103" 28 | }, 29 | { 30 | "venue": "CICC 2022", 31 | "name": "StreamGCN: Accelerating Graph Convolutional Networks with Streaming Processing", 32 | "affiliation": "UCLA", 33 | "link": "https://web.cs.ucla.edu/~atefehsz/publication/StreamGCN-CICC22.pdf" 34 | }, 35 | { 36 | "venue": "HPCA 2022", 37 | "name": "Accelerating Graph Convolutional Networks Using Crossbar-based Processing-In-Memory Architectures", 38 | "affiliation": "HUST", 39 | "link": "https://ieeexplore.ieee.org/document/9773267" 40 | }, 41 | { 42 | "venue": "HPCA 2022", 43 | "name": "GCoD: Graph Convolutional Network Acceleration via Dedicated Algorithm and Accelerator Co-Design", 44 | "affiliation": "Rice, PNNL", 45 | "link": "https://arxiv.org/abs/2112.11594", 46 | "source": "https://github.com/RICE-EIC/GCoD" 47 | }, 48 | { 49 | "venue": "arXiv 2022", 50 | "name": "GenGNN: A Generic FPGA Framework for Graph Neural Network Acceleration", 51 | "affiliation": "GaTech", 52 | "link": "https://arxiv.org/abs/2201.08475" 53 | }, 54 | { 55 | "venue": "DAC 2021", 56 | "name": "DyGNN: Algorithm and Architecture Support of vertex Dynamic Pruning for Graph Neural Networks", 57 | "affiliation": "Hunan University", 58 | "link": "https://ieeexplore.ieee.org/document/9586298" 59 | }, 60 | { 61 | "venue": "DAC 2021", 62 | "name": "BlockGNN: Towards Efficient GNN Acceleration Using Block-Circulant Weight Matrices", 63 | "affiliation": "PKU", 64 | "link": "https://arxiv.org/abs/2104.06214" 65 | }, 66 | { 67 | "venue": "DAC 2021", 68 | "name": "TARe: Task-Adaptive in-situ ReRAM Computing for Graph Learning", 69 | "affiliation": "Chinese Academy of Sciences", 70 | "link": "https://ieeexplore.ieee.org/abstract/document/9586193" 71 | }, 72 | { 73 | "venue": "ICCAD 2021", 74 | "name": "G-CoS: GNN-Accelerator Co-Search Towards Both Better Accuracy and Efficiency", 75 | "affiliation": "Rice", 76 | "link": "https://arxiv.org/abs/2109.08983" 77 | }, 78 | { 79 | "venue": "MICRO 2021", 80 | "name": "I-GCN: A Graph Convolutional Network Accelerator with Runtime Locality Enhancement through Islandization", 81 | "affiliation": "PNNL", 82 | "link": "https://dl.acm.org/doi/pdf/10.1145/3466752.3480113" 83 | }, 84 | { 85 | "venue": "arXiv 2021", 86 | "name": "ZIPPER: Exploiting Tile- and Operator-level Parallelism for General and Scalable Graph Neural Network Acceleration", 87 | "affiliation": "SJTU", 88 | "link": "https://arxiv.org/abs/2107.08709" 89 | }, 90 | { 91 | "venue": "TComp 2021", 92 | "name": "EnGN: A High-Throughput and Energy-Efficient Accelerator for Large Graph Neural Networks", 93 | "affiliation": "Chinese Academy of Sciences", 94 | "link": "https://arxiv.org/abs/1909.00155" 95 | }, 96 | { 97 | "venue": "HPCA 2021", 98 | "name": "GCNAX: A Flexible and Energy-efficient Accelerator for Graph Convolutional Neural Networks", 99 | "affiliation": "GWU", 100 | "link": "https://ieeexplore.ieee.org/abstract/document/9407104" 101 | }, 102 | { 103 | "venue": "APA 2020", 104 | "name": "GNN-PIM: A Processing-in-Memory Architecture for Graph Neural Networks", 105 | "affiliation": "PKU", 106 | "link": "http://115.27.240.201/docs/20200915165942122459.pdf" 107 | }, 108 | { 109 | "venue": "ASAP 2020", 110 | "name": "Hardware Acceleration of Large Scale GCN Inference", 111 | "affiliation": "USC", 112 | "link": "https://ieeexplore.ieee.org/document/9153263" 113 | }, 114 | { 115 | "venue": "DAC 2020", 116 | "name": "Hardware Acceleration of Graph Neural Networks", 117 | "affiliation": "UIUC", 118 | "link": "http://rakeshk.web.engr.illinois.edu/dac20.pdf" 119 | }, 120 | { 121 | "venue": "MICRO 2020", 122 | "name": "AWB-GCN: A Graph Convolutional Network Accelerator with Runtime Workload Rebalancing", 123 | "affiliation": "PNNL", 124 | "link": "https://ieeexplore.ieee.org/abstract/document/9252000" 125 | }, 126 | { 127 | "venue": "arXiv 2020", 128 | "name": "GRIP: A Graph Neural Network Accelerator Architecture", 129 | "affiliation": "Stanford", 130 | "link": "https://arxiv.org/pdf/2007.13828.pdf" 131 | }, 132 | { 133 | "venue": "HPCA 2020", 134 | "name": "HyGCN: A GCN Accelerator with Hybrid Architecture", 135 | "affiliation": "UCSB", 136 | "link": "https://arxiv.org/pdf/2001.02514.pdf" 137 | } 138 | ] 139 | } 140 | -------------------------------------------------------------------------------- /res/papers/05-distributed-training.json: -------------------------------------------------------------------------------- 1 | { 2 | "title": "Distributed GNN Training Systems", 3 | "paper": [ 4 | { 5 | "venue": "FAST 2025", 6 | "name": "LeapGNN: Accelerating Distributed GNN Training Leveraging Feature-Centric Model Migration", 7 | "affiliation": "ZJU", 8 | "link": "https://www.usenix.org/system/files/fast25-chen-weijian-leap.pdf", 9 | "source": "https://github.com/ISCS-ZJU/LeapGNN-AE" 10 | }, 11 | { 12 | "venue": "VLDB 2025", 13 | "name": "NeutronTP: Load-Balanced Distributed Full-Graph GNN Training with Tensor Parallelism ", 14 | "affiliation": "NEU", 15 | "link": "https://arxiv.org/abs/2412.20379", 16 | "source": "https://github.com/iDC-NEU/NeutronTP" 17 | }, 18 | { 19 | "venue": "arXiv 2023", 20 | "name": "Communication-Free Distributed GNN Training with Vertex Cut", 21 | "affiliation": "Stanford", 22 | "link": "https://arxiv.org/pdf/2308.03209.pdf" 23 | }, 24 | { 25 | "venue": "arXiv 2023", 26 | "name": "GNNPipe: Accelerating Distributed Full-Graph GNN Training with Pipelined Model Parallelism", 27 | "affiliation": "Purdue", 28 | "link": "https://arxiv.org/pdf/2308.10087.pdf" 29 | }, 30 | { 31 | "venue": "OSDI 2023", 32 | "name": "MGG: Accelerating Graph Neural Networks with Fine-Grained Intra-Kernel Communication-Computation Pipelining on Multi-GPU Platforms", 33 | "affiliation": "UCSB", 34 | "link": "https://www.usenix.org/system/files/osdi23-wang-yuke.pdf", 35 | "source": "https://github.com/YukeWang96/MGG_OSDI23" 36 | }, 37 | { 38 | "venue": "SIGMOD 2022", 39 | "name": "NeutronStar: Distributed GNN Training with Hybrid Management ", 40 | "affiliation": "NEU", 41 | "link": "https://dl.acm.org/doi/pdf/10.1145/3514221.3526134", 42 | "source": "https://github.com/iDC-NEU/NeutronStarLite" 43 | }, 44 | { 45 | "venue": "VLDB 2022", 46 | "name": "Sancus: Staleness-Aware Communication-Avoiding Full-Graph Decentralized Training in Large-Scale Graph Neural Networks", 47 | "affiliation": "HKUST", 48 | "link": "https://www.vldb.org/pvldb/vol15/p1937-peng.pdf", 49 | "source": "https://github.com/chenzhao/light-dist-gnn" 50 | }, 51 | { 52 | "venue": "MLSys 2022", 53 | "name": "BNS-GCN: Efficient Full-Graph Training of Graph Convolutional Networks with Partition-Parallelism and Random Boundary Node Sampling", 54 | "affiliation": "Rice, UIUC", 55 | "link": "https://proceedings.mlsys.org/paper/2022/file/d1fe173d08e959397adf34b1d77e88d7-Paper.pdf", 56 | "source": "https://github.com/RICE-EIC/BNS-GCN" 57 | }, 58 | { 59 | "venue": "MLSys 2022", 60 | "name": "Sequential Aggregation and Rematerialization: Distributed Full-batch Training of Graph Neural Networks on Large Graphs", 61 | "affiliation": "Intel", 62 | "link": "https://arxiv.org/abs/2111.06483", 63 | "source": "https://github.com/IntelLabs/SAR" 64 | }, 65 | { 66 | "venue": "WWW 2022", 67 | "name": "PaSca: A Graph Neural Architecture Search System under the Scalable Paradigm", 68 | "affiliation": "PKU", 69 | "link": "https://dl.acm.org/doi/abs/10.1145/3485447.3511986" 70 | }, 71 | { 72 | "venue": "ICLR 2022", 73 | "name": "PipeGCN: Efficient Full-Graph Training of Graph Convolutional Networks with Pipelined Feature Communication", 74 | "affiliation": "Rice", 75 | "link": "https://openreview.net/pdf?id=kSwqMH0zn1F", 76 | "source": "https://github.com/RICE-EIC/PipeGCN" 77 | }, 78 | { 79 | "venue": "ICLR 2022", 80 | "name": "Learn Locally, Correct Globally: A Distributed Algorithm for Training Graph Neural Networks", 81 | "affiliation": "PSU", 82 | "link": "https://openreview.net/pdf?id=FndDxSz3LxQ", 83 | "source": "https://github.com/MortezaRamezani/llcg" 84 | }, 85 | { 86 | "venue": "arXiv 2021", 87 | "name": "Distributed Hybrid CPU and GPU training for Graph Neural Networks on Billion-Scale Graphs", 88 | "affiliation": "AWS", 89 | "link": "https://arxiv.org/pdf/2112.15345.pdf" 90 | }, 91 | { 92 | "venue": "SC 2021", 93 | "name": "DistGNN: Scalable Distributed Training for Large-Scale Graph Neural Networks", 94 | "affiliation": "Intel", 95 | "link": "https://dl.acm.org/doi/pdf/10.1145/3458817.3480856", 96 | "source": "https://github.com/dmlc/dgl/pull/3024" 97 | }, 98 | { 99 | "venue": "SC 2021", 100 | "name": "Efficient Scaling of Dynamic Graph Neural Networks", 101 | "affiliation": "IBM", 102 | "link": "https://dl.acm.org/doi/pdf/10.1145/3458817.3480858" 103 | }, 104 | { 105 | "venue": "CLUSTER 2021", 106 | "name": "2PGraph: Accelerating GNN Training over Large Graphs on GPU Clusters", 107 | "affiliation": "NUDT", 108 | "link": "https://ieeexplore.ieee.org/abstract/document/9556026" 109 | }, 110 | { 111 | "venue": "OSDI 2021", 112 | "name": "$P^3$: Distributed Deep Graph Learning at Scale", 113 | "affiliation": "MSR", 114 | "link": "https://www.usenix.org/system/files/osdi21-gandhi.pdf" 115 | }, 116 | { 117 | "venue": "OSDI 2021", 118 | "name": "Dorylus: Affordable, Scalable, and Accurate GNN Training with Distributed CPU Servers and Serverless Threads", 119 | "affiliation": "UCLA", 120 | "link": "http://web.cs.ucla.edu/~harryxu/papers/dorylus-osdi21.pdf", 121 | "source": "https://github.com/uclasystem/dorylus" 122 | }, 123 | { 124 | "venue": "arXiv 2021", 125 | "name": "GIST: Distributed Training for Large-Scale Graph Convolutional Networks", 126 | "affiliation": "Rice", 127 | "link": "https://arxiv.org/abs/2102.10424" 128 | }, 129 | { 130 | "venue": "EuroSys 2021", 131 | "name": "FlexGraph: A Flexible and Efficient Distributed Framework for GNN Training", 132 | "affiliation": "Alibaba", 133 | "link": "https://dl.acm.org/doi/pdf/10.1145/3447786.3456229" 134 | }, 135 | { 136 | "venue": "EuroSys 2021", 137 | "name": "DGCL: An Efficient Communication Library for Distributed GNN Training", 138 | "affiliation": "CUHK", 139 | "link": "https://dl.acm.org/doi/abs/10.1145/3447786.3456233", 140 | "source": "https://github.com/czkkkkkk/ragdoll" 141 | }, 142 | { 143 | "venue": "SC 2020", 144 | "name": "Reducing Communication in Graph Neural Network Training", 145 | "affiliation": "UC Berkeley", 146 | "link": "https://arxiv.org/pdf/2005.03300.pdf", 147 | "source": "https://github.com/PASSIONLab/CAGNET" 148 | }, 149 | { 150 | "venue": "VLDB 2020", 151 | "name": "G$^3$: When Graph Neural Networks Meet Parallel Graph Processing Systems on GPUs", 152 | "affiliation": "NUS", 153 | "link": "http://www.vldb.org/pvldb/vol13/p2813-liu.pdf", 154 | "source": "https://github.com/Xtra-Computing/G3" 155 | }, 156 | { 157 | "venue": "IA3 2020", 158 | "name": "DistDGL: Distributed Graph Neural Network Training for Billion-Scale Graphs", 159 | "affiliation": "AWS", 160 | "link": "https://arxiv.org/pdf/2010.05337.pdf", 161 | "source": "https://github.com/dmlc/dgl/tree/master/python/dgl/distributed" 162 | }, 163 | { 164 | "venue": "MLSys 2020", 165 | "name": "Improving the Accuracy, Scalability, and Performance of Graph Neural Networks with Roc", 166 | "affiliation": "Stanford", 167 | "link": "https://proceedings.mlsys.org/paper/2020/file/fe9fc289c3ff0af142b6d3bead98a923-Paper.pdf", 168 | "source": "https://github.com/jiazhihao/ROC" 169 | }, 170 | { 171 | "venue": "arXiv 2020", 172 | "name": "AGL: A Scalable System for Industrial-purpose Graph Machine Learning", 173 | "affiliation": "Ant Financial Services Group", 174 | "link": "https://arxiv.org/abs/2003.02454" 175 | }, 176 | { 177 | "venue": "ATC 2019", 178 | "name": "NeuGraph: Parallel Deep Neural Network Computation on Large Graphs", 179 | "affiliation": "PKU", 180 | "link": "https://www.usenix.org/system/files/atc19-ma_0.pdf" 181 | } 182 | ] 183 | } 184 | -------------------------------------------------------------------------------- /.github/citation/citation.json: -------------------------------------------------------------------------------- 1 | {"G-CoS: GNN-Accelerator Co-Search Towards Both Better Accuracy and Efficiency": {"citation": 0, "last update": "2025-12-15"}, "GRIP: A Graph Neural Network Accelerator Architecture": {"citation": 0, "last update": "2025-12-15"}, "GCNear: A Hybrid Architecture for Efficient GCN Training with Near-Memory Processing": {"citation": 0, "last update": "2025-12-15"}, "FlowGNN: A Dataflow Architecture for Universal Graph Neural Network Inference via Multi-Queue Streaming": {"citation": 0, "last update": "2025-12-15"}, "ReGraphX: NoC-enabled 3D Heterogeneous ReRAM Architecture for Training Graph Neural Networks": {"citation": 0, "last update": "2025-12-15"}, "GIST: Distributed Training for Large-Scale Graph Convolutional Networks": {"citation": 0, "last update": "2025-12-15"}, "Binary Graph Neural Networks": {"citation": 0, "last update": "2025-12-15"}, "Parallel and Distributed Graph Neural Networks: An In-Depth Concurrency Analysis": {"citation": 0, "last update": "2025-12-15"}, "GraphACT: Accelerating GCN Training on CPU-FPGA Heterogeneous Platforms": {"citation": 0, "last update": "2025-12-15"}, "Seastar: Vertex-Centric Programming for Graph Neural Networks": {"citation": 0, "last update": "2025-12-15"}, "PaSca: A Graph Neural Architecture Search System under the Scalable Paradigm": {"citation": 0, "last update": "2025-12-15"}, "AGL: A Scalable System for Industrial-purpose Graph Machine Learning": {"citation": 0, "last update": "2025-12-15"}, "GE-SpMM: General-purpose Sparse Matrix-Matrix Multiplication on GPUs for Graph Neural Networks": {"citation": 0, "last update": "2025-12-15"}, "Graphiler: Optimizing Graph Neural Networks with Message Passing Data Flow Graph": {"citation": 0, "last update": "2025-12-15"}, "DistGNN: Scalable Distributed Training for Large-Scale Graph Neural Networks": {"citation": 0, "last update": "2025-12-17"}, "DIG: A Turnkey Library for Diving into Graph Deep Learning Research": {"citation": 0, "last update": "2025-12-17"}, "Accelerating Graph Convolutional Networks Using Crossbar-based Processing-In-Memory Architectures": {"citation": 0, "last update": "2025-12-17"}, "Understanding the Design Space of Sparse/Dense Multiphase Dataflows for Mapping Graph Neural Networks on Spatial Accelerators": {"citation": 0, "last update": "2025-12-17"}, "Understanding and Bridging the Gaps in Current GNN Performance Optimizations": {"citation": 0, "last update": "2025-12-17"}, "MGG: Accelerating Graph Neural Networks with Fine-Grained Intra-Kernel Communication-Computation Pipelining on Multi-GPU Platforms": {"citation": 0, "last update": "2025-12-17"}, "GNNPipe: Accelerating Distributed Full-Graph GNN Training with Pipelined Model Parallelism": {"citation": 0, "last update": "2025-12-17"}, "GNN-PIM: A Processing-in-Memory Architecture for Graph Neural Networks": {"citation": 0, "last update": "2025-12-17"}, "Sancus: Staleness-Aware Communication-Avoiding Full-Graph Decentralized Training in Large-Scale Graph Neural Networks": {"citation": 0, "last update": "2025-12-17"}, "Distributed Graph Neural Network Training: A Survey": {"citation": 0, "last update": "2025-12-17"}, "Bi-GCN: Binary Graph Convolutional Network": {"citation": 0, "last update": "2025-12-17"}, "Sequential Aggregation and Rematerialization: Distributed Full-batch Training of Graph Neural Networks on Large Graphs": {"citation": 0, "last update": "2025-12-17"}, "PCGCN: Partition-Centric Processing for Accelerating Graph Convolutional Network": {"citation": 0, "last update": "2025-12-17"}, "Hardware Acceleration of Graph Neural Networks": {"citation": 0, "last update": "2025-12-17"}, "Distributed Hybrid CPU and GPU training for Graph Neural Networks on Billion-Scale Graphs": {"citation": 0, "last update": "2025-12-17"}, "DGCL: An Efficient Communication Library for Distributed GNN Training": {"citation": 0, "last update": "2025-12-17"}, "TLPGNN: A Lightweight Two-Level Parallelism Paradigm for Graph Neural Network Computation on GPU": {"citation": 0, "last update": "2025-12-17"}, "GNNAutoScale: Scalable and Expressive Graph Neural Networks via Historical Embeddings": {"citation": 0, "last update": "2025-12-17"}, "I-GCN: A Graph Convolutional Network Accelerator with Runtime Locality Enhancement through Islandization": {"citation": 0, "last update": "2025-12-17"}, "ByteGNN: Efficient Graph Neural Network Training at Large Scale": {"citation": 0, "last update": "2025-12-17"}, "AWB-GCN: A Graph Convolutional Network Accelerator with Runtime Workload Rebalancing": {"citation": 0, "last update": "2025-12-17"}, "GNNAdvisor: An Adaptive and Efficient Runtime System for GNN Acceleration on GPUs": {"citation": 0, "last update": "2025-12-18"}, "DyGNN: Algorithm and Architecture Support of vertex Dynamic Pruning for Graph Neural Networks": {"citation": 0, "last update": "2025-12-18"}, "BGL: GPU-Efficient GNN Training by Optimizing Graph Data I/O and Preprocessing": {"citation": 0, "last update": "2025-12-18"}, "EnGN: A High-Throughput and Energy-Efficient Accelerator for Large Graph Neural Networks": {"citation": 0, "last update": "2025-12-18"}, "Reducing Communication in Graph Neural Network Training": {"citation": 0, "last update": "2025-12-18"}, "fuseGNN: Accelerating Graph Convolutional Neural Network Training on GPGPU": {"citation": 0, "last update": "2025-12-18"}, "A Survey on Graph Neural Network Acceleration: Algorithms, Systems, and Customized Hardware": {"citation": 0, "last update": "2025-12-18"}, "A Comprehensive Survey on Distributed Training of Graph Neural Networks": {"citation": 0, "last update": "2025-12-18"}, "QGTC: Accelerating Quantized Graph Neural Networks via GPU Tensor Core": {"citation": 0, "last update": "2025-12-18"}, "BlockGNN: Towards Efficient GNN Acceleration Using Block-Circulant Weight Matrices": {"citation": 0, "last update": "2025-12-18"}, "Learned Low Precision Graph Neural Networks": {"citation": 0, "last update": "2025-12-18"}, "2PGraph: Accelerating GNN Training over Large Graphs on GPU Clusters": {"citation": 0, "last update": "2025-12-18"}, "Communication-Free Distributed GNN Training with Vertex Cut": {"citation": 0, "last update": "2025-12-18"}, "MariusGNN: Resource-Efficient Out-of-Core Training of Graph Neural Networks": {"citation": 0, "last update": "2025-12-18"}, "Improving the Accuracy, Scalability, and Performance of Graph Neural Networks with Roc": {"citation": 0, "last update": "2025-12-18"}, "Large Graph Convolutional Network Training with GPU-Oriented Data Communication Architecture": {"citation": 0, "last update": "2025-12-18"}, "GraphFM: Improving Large-Scale GNN Training via Feature Momentum": {"citation": 0, "last update": "2025-12-18"}, "Dorylus: Affordable, Scalable, and Accurate GNN Training with Distributed CPU Servers and Serverless Threads": {"citation": 0, "last update": "2025-12-18"}, "PaGraph: Scaling GNN Training on Large Graphs via Computation-aware Caching": {"citation": 0, "last update": "2025-12-18"}, "NeuGraph: Parallel Deep Neural Network Computation on Large Graphs": {"citation": 0, "last update": "2025-12-18"}, "Computing Graph Neural Networks: A Survey from Algorithms to Accelerators": {"citation": 0, "last update": "2025-12-18"}, "ZIPPER: Exploiting Tile- and Operator-level Parallelism for General and Scalable Graph Neural Network Acceleration": {"citation": 0, "last update": "2025-12-19"}, "GCNAX: A Flexible and Energy-efficient Accelerator for Graph Convolutional Neural Networks": {"citation": 0, "last update": "2025-12-19"}, "DistDGL: Distributed Graph Neural Network Training for Billion-Scale Graphs": {"citation": 0, "last update": "2025-12-19"}, "SGQuant: Squeezing the Last Bit on Graph Neural Networks with Specialized Quantization": {"citation": 0, "last update": "2025-12-19"}, "CogDL: A Toolkit for Deep Learning on Graphs": {"citation": 0, "last update": "2025-12-19"}, "AliGraph: A Comprehensive Graph Neural Network Platform": {"citation": 0, "last update": "2025-12-19"}, "BeaconGNN: Large-Scale GNN Acceleration with Out-of-Order Streaming In-Storage Computing": {"citation": 0, "last update": "2025-12-19"}, "In situ neighborhood sampling for large-scale GNN training": {"citation": 0, "last update": "2025-12-19"}, "Ginex: SSD-enabled Billion-scale Graph Neural Network Training on a Single Machine via Provably Optimal In-memory Caching": {"citation": 0, "last update": "2025-12-19"}, "SmartSAGE: Training Large-scale Graph Neural Networks using In-Storage Processing Architectures": {"citation": 0, "last update": "2025-12-19"}, "Deep Graph Library: A Graph-Centric, Highly-Performant Package for Graph Neural Networks": {"citation": 0, "last update": "2025-12-19"}, "Fast Graph Representation Learning with PyTorch Geometric": {"citation": 0, "last update": "2025-12-19"}, "Accelerating Sampling and Aggregation Operations in GNN Frameworks with GPU Initiated Direct Storage Accesses": {"citation": 0, "last update": "2025-12-19"}, "NeutronStar: Distributed GNN Training with Hybrid Management ": {"citation": 0, "last update": "2025-12-19"}, "LeapGNN: Accelerating Distributed GNN Training Leveraging Feature-Centric Model Migration": {"citation": 0, "last update": "2025-12-19"}, "NeutronTP: Load-Balanced Distributed Full-Graph GNN Training with Tensor Parallelism ": {"citation": 0, "last update": "2025-12-19"}, "Hyperion: Optimizing SSD Access is All You Need to Enable Cost-efficient Out-of-core GNN Training": {"citation": 0, "last update": "2025-12-19"}, "Relational Inductive Biases, Deep Learning, and Graph Networks": {"citation": 0, "last update": "2025-12-19"}, "FlexGraph: A Flexible and Efficient Distributed Framework for GNN Training": {"citation": 0, "last update": "2025-12-19"}, "TigerGraph: A Native MPP Graph Database": {"citation": 0, "last update": "2025-12-19"}, "Degree-Quant: Quantization-Aware Training for Graph Neural Networks": {"citation": 0, "last update": "2025-12-19"}, "BNS-GCN: Efficient Full-Graph Training of Graph Convolutional Networks with Partition-Parallelism and Random Boundary Node Sampling": {"citation": 0, "last update": "2025-12-19"}, "Binarized Graph Neural Network": {"citation": 0, "last update": "2025-12-19"}, "GenGNN: A Generic FPGA Framework for Graph Neural Network Acceleration": {"citation": 0, "last update": "2025-12-19"}, "GNNLab: A Factored System for Sample-based GNN Training over GPUs": {"citation": 0, "last update": "2025-12-19"}, "StreamGCN: Accelerating Graph Convolutional Networks with Streaming Processing": {"citation": 0, "last update": "2025-12-19"}, "FusedMM: A Unified SDDMM-SpMM Kernel for Graph Embedding and Graph Neural Networks": {"citation": 0, "last update": "2025-12-19"}, "$P^3$: Distributed Deep Graph Learning at Scale": {"citation": 0, "last update": "2025-12-19"}, "G$^3$: When Graph Neural Networks Meet Parallel Graph Processing Systems on GPUs": {"citation": 0, "last update": "2025-12-20"}, "PipeGCN: Efficient Full-Graph Training of Graph Convolutional Networks with Pipelined Feature Communication": {"citation": 0, "last update": "2025-12-20"}, "EPQuant: A Graph Neural Network Compression Approach Based on Product Quantization": {"citation": 0, "last update": "2025-12-20"}, "Graph Neural Networks in TensorFlow and Keras with Spektral": {"citation": 0, "last update": "2025-12-20"}, "Efficient Scaling of Dynamic Graph Neural Networks": {"citation": 0, "last update": "2025-12-20"}, "Accelerating Training and Inference of Graph Neural Networks with Fast Sampling and Pipelining": {"citation": 0, "last update": "2025-12-20"}, "Hardware Acceleration of Large Scale GCN Inference": {"citation": 0, "last update": "2025-12-20"}, "Learn Locally, Correct Globally: A Distributed Algorithm for Training Graph Neural Networks": {"citation": 0, "last update": "2025-12-20"}, "TARe: Task-Adaptive in-situ ReRAM Computing for Graph Learning": {"citation": 0, "last update": "2025-12-20"}, "GCoD: Graph Convolutional Network Acceleration via Dedicated Algorithm and Accelerator Co-Design": {"citation": 0, "last update": "2025-12-20"}, "Hyperscale FPGA-as-a-service architecture for large-scale distributed graph neural network": {"citation": 0, "last update": "2025-12-20"}, "Understanding GNN Computational Graph: A Coordinated Computation, IO, and Memory Perspective": {"citation": 0, "last update": "2025-12-20"}, "DRGN: a dynamically reconfigurable accelerator for graph neural networks": {"citation": 0, "last update": "2025-12-20"}, "Global Neighbor Sampling for Mixed CPU-GPU Training on Giant Graphs": {"citation": 0, "last update": "2025-12-20"}, "Efficient Data Loader for Fast Sampling-Based GNN Training on Large Graphs": {"citation": 0, "last update": "2025-12-20"}, "Graphite: Optimizing Graph Neural Networks on CPUs Through Cooperative Software-Hardware Techniques": {"citation": 0, "last update": "2025-12-20"}, "Rubik: A Hierarchical Architecture for Efficient Graph Learning": {"citation": 0, "last update": "2025-12-20"}, "HyGCN: A GCN Accelerator with Hybrid Architecture": {"citation": 0, "last update": "2025-12-20"}, "GNNIE: GNN Inference Engine with Load-balancing and Graph-specific Caching": {"citation": 0, "last update": "2025-12-20"}, "FeatGraph: A Flexible and Efficient Backend for Graph Neural Network Systems": {"citation": 0, "last update": "2025-12-20"}, "EXACT: Scalable Graph Neural Networks Training via Extreme Activation Compression": {"citation": 0, "last update": "2025-12-20"}} -------------------------------------------------------------------------------- /README.md: -------------------------------------------------------------------------------- 1 | # Awesome Graph Neural Network Systems [![Awesome](https://cdn.rawgit.com/sindresorhus/awesome/d7305f38d29fed78fa85652e3a63e154dd8e8829/media/badge.svg)](https://github.com/sindresorhus/awesome) 2 | 3 | A list of awesome systems for graph neural network (GNN). If you have any comment, please create an [issue](https://github.com/chwan1016/awesome-gnn-systems/issues) or [pull request](https://github.com/chwan1016/awesome-gnn-systems/pulls). 4 | ## Contents 5 | - [Open Source Libraries](#open-source-libraries) 6 | - [Papers](#papers) 7 | - [Survey Papers](#survey-papers) 8 | - [GNN Libraries](#gnn-libraries) 9 | - [GNN Kernels](#gnn-kernels) 10 | - [GNN Compilers](#gnn-compilers) 11 | - [Distributed GNN Training Systems](#distributed-gnn-training-systems) 12 | - [Training Systems for Scaling Graphs](#training-systems-for-scaling-graphs) 13 | - [Quantized GNNs](#quantized-gnns) 14 | - [GNN Dataloaders](#gnn-dataloaders) 15 | - [GNN Training Accelerators](#gnn-training-accelerators) 16 | - [GNN Inference Accelerators](#gnn-inference-accelerators) 17 | - [Contribute](#contribute) 18 | ## Open Source Libraries 19 | - [PyG: Graph Neural Network Library for PyTorch](https://github.com/rusty1s/pytorch_geometric) ![GitHub stars](https://img.shields.io/github/stars/rusty1s/pytorch_geometric.svg?logo=github&label=Stars) 20 | - [DGL: Python Package Built to Ease Deep Learning on Graph](https://github.com/dmlc/dgl) ![GitHub stars](https://img.shields.io/github/stars/dmlc/dgl.svg?logo=github&label=Stars) 21 | - [Graph Nets: Build Graph Nets in Tensorflow](https://github.com/deepmind/graph_nets) ![GitHub stars](https://img.shields.io/github/stars/deepmind/graph_nets.svg?logo=github&label=Stars) 22 | - [Euler: A Distributed Graph Deep Learning Framework](https://github.com/alibaba/euler) ![GitHub stars](https://img.shields.io/github/stars/alibaba/euler.svg?logo=github&label=Stars) 23 | - [StellarGraph: Machine Learning on Graphs](https://github.com/stellargraph/stellargraph) ![GitHub stars](https://img.shields.io/github/stars/stellargraph/stellargraph.svg?logo=github&label=Stars) 24 | - [Spektral: Graph Neural Networks with Keras and Tensorflow 2](https://github.com/danielegrattarola/spektral) ![GitHub stars](https://img.shields.io/github/stars/danielegrattarola/spektral.svg?logo=github&label=Stars) 25 | - [PGL: An Efficient and Flexible Graph Learning Framework Based on PaddlePaddle](https://github.com/PaddlePaddle/PGL) ![GitHub stars](https://img.shields.io/github/stars/PaddlePaddle/PGL.svg?logo=github&label=Stars) 26 | - [CogDL: An Extensive Toolkit for Deep Learning on Graphs](https://github.com/THUDM/cogdl) ![GitHub stars](https://img.shields.io/github/stars/THUDM/cogdl.svg?logo=github&label=Stars) 27 | - [DIG: A Turnkey Library for Diving into Graph Deep Learning Research](https://github.com/divelab/DIG) ![GitHub stars](https://img.shields.io/github/stars/divelab/DIG.svg?logo=github&label=Stars) 28 | - [Jraph: A Graph Neural Network Library in Jax](https://github.com/deepmind/jraph) ![GitHub stars](https://img.shields.io/github/stars/deepmind/jraph.svg?logo=github&label=Stars) 29 | - [Graph-Learn: An Industrial Graph Neural Network Framework](https://github.com/alibaba/graph-learn) ![GitHub stars](https://img.shields.io/github/stars/alibaba/graph-learn.svg?logo=github&label=Stars) 30 | - [DeepGNN: a Framework for Training Machine Learning Models on Large Scale Graph Data](https://github.com/microsoft/DeepGNN) ![GitHub stars](https://img.shields.io/github/stars/microsoft/DeepGNN.svg?logo=github&label=Stars) 31 | ## Papers 32 | ### Survey Papers 33 | | Venue | Title | Affiliation |       Link       |   Source   | 34 | | :---: | :---: | :---------: | :---: | :----: | 35 | |CSUR 2024|Distributed Graph Neural Network Training: A Survey|BUPT| [[paper]](https://dl.acm.org/doi/10.1145/3648358)![Scholar citations](https://img.shields.io/badge/Citations-0-_.svg?logo=google-scholar&labelColor=4f4f4f&color=3388ee)|| 36 | |Proceedings of the IEEE 2023|A Comprehensive Survey on Distributed Training of Graph Neural Networks|Chinese Academy of Sciences| [[paper]](https://ieeexplore.ieee.org/abstract/document/10348966)![Scholar citations](https://img.shields.io/badge/Citations-0-_.svg?logo=google-scholar&labelColor=4f4f4f&color=3388ee)|| 37 | |arXiv 2023|A Survey on Graph Neural Network Acceleration: Algorithms, Systems, and Customized Hardware|UCLA| [[paper]](https://arxiv.org/abs/2306.14052)![Scholar citations](https://img.shields.io/badge/Citations-0-_.svg?logo=google-scholar&labelColor=4f4f4f&color=3388ee)|| 38 | |arXiv 2022|Parallel and Distributed Graph Neural Networks: An In-Depth Concurrency Analysis|ETHZ| [[paper]](https://arxiv.org/abs/2205.09702)![Scholar citations](https://img.shields.io/badge/Citations-0-_.svg?logo=google-scholar&labelColor=4f4f4f&color=3388ee)|| 39 | |CSUR 2022|Computing Graph Neural Networks: A Survey from Algorithms to Accelerators|UPC| [[paper]](https://dl.acm.org/doi/10.1145/3477141)![Scholar citations](https://img.shields.io/badge/Citations-0-_.svg?logo=google-scholar&labelColor=4f4f4f&color=3388ee)|| 40 | ### GNN Libraries 41 | | Venue | Title | Affiliation |       Link       |   Source   | 42 | | :---: | :---: | :---------: | :---: | :----: | 43 | |JMLR 2021|DIG: A Turnkey Library for Diving into Graph Deep Learning Research|TAMU| [[paper]](https://arxiv.org/abs/2103.12608)![Scholar citations](https://img.shields.io/badge/Citations-0-_.svg?logo=google-scholar&labelColor=4f4f4f&color=3388ee)| [[code]](https://github.com/divelab/DIG)![GitHub stars](https://img.shields.io/github/stars/divelab/DIG.svg?logo=github&label=Stars)| 44 | |arXiv 2021|CogDL: A Toolkit for Deep Learning on Graphs|THU| [[paper]](https://arxiv.org/abs/2103.00959)![Scholar citations](https://img.shields.io/badge/Citations-0-_.svg?logo=google-scholar&labelColor=4f4f4f&color=3388ee)| [[code]](https://github.com/THUDM/cogdl)![GitHub stars](https://img.shields.io/github/stars/THUDM/cogdl.svg?logo=github&label=Stars)| 45 | |CIM 2021|Graph Neural Networks in TensorFlow and Keras with Spektral|Università della Svizzera italiana| [[paper]](https://arxiv.org/abs/2006.12138)![Scholar citations](https://img.shields.io/badge/Citations-0-_.svg?logo=google-scholar&labelColor=4f4f4f&color=3388ee)| [[code]](https://github.com/danielegrattarola/spektral)![GitHub stars](https://img.shields.io/github/stars/danielegrattarola/spektral.svg?logo=github&label=Stars)| 46 | |arXiv 2019|Deep Graph Library: A Graph-Centric, Highly-Performant Package for Graph Neural Networks|AWS| [[paper]](https://arxiv.org/abs/1909.01315)![Scholar citations](https://img.shields.io/badge/Citations-0-_.svg?logo=google-scholar&labelColor=4f4f4f&color=3388ee)| [[code]](https://github.com/dmlc/dgl)![GitHub stars](https://img.shields.io/github/stars/dmlc/dgl.svg?logo=github&label=Stars)| 47 | |VLDB 2019|AliGraph: A Comprehensive Graph Neural Network Platform|Alibaba| [[paper]](https://dl.acm.org/doi/pdf/10.14778/3352063.3352127)![Scholar citations](https://img.shields.io/badge/Citations-0-_.svg?logo=google-scholar&labelColor=4f4f4f&color=3388ee)| [[code]](https://github.com/alibaba/graph-learn)![GitHub stars](https://img.shields.io/github/stars/alibaba/graph-learn.svg?logo=github&label=Stars)| 48 | |arXiv 2019|Fast Graph Representation Learning with PyTorch Geometric|TU Dortmund University| [[paper]](https://arxiv.org/abs/1903.02428)![Scholar citations](https://img.shields.io/badge/Citations-0-_.svg?logo=google-scholar&labelColor=4f4f4f&color=3388ee)| [[code]](https://github.com/rusty1s/pytorch_geometric)![GitHub stars](https://img.shields.io/github/stars/rusty1s/pytorch_geometric.svg?logo=github&label=Stars)| 49 | |arXiv 2018|Relational Inductive Biases, Deep Learning, and Graph Networks|DeepMind| [[paper]](https://arxiv.org/abs/1806.01261)![Scholar citations](https://img.shields.io/badge/Citations-0-_.svg?logo=google-scholar&labelColor=4f4f4f&color=3388ee)| [[code]](https://github.com/deepmind/graph_nets)![GitHub stars](https://img.shields.io/github/stars/deepmind/graph_nets.svg?logo=github&label=Stars)| 50 | ### GNN Kernels 51 | | Venue | Title | Affiliation |       Link       |   Source   | 52 | | :---: | :---: | :---------: | :---: | :----: | 53 | |MLSys 2022|Understanding GNN Computational Graph: A Coordinated Computation, IO, and Memory Perspective|THU| [[paper]](https://proceedings.mlsys.org/paper/2022/file/9a1158154dfa42caddbd0694a4e9bdc8-Paper.pdf)![Scholar citations](https://img.shields.io/badge/Citations-0-_.svg?logo=google-scholar&labelColor=4f4f4f&color=3388ee)| [[code]](https://github.com/dgSPARSE/dgNN)![GitHub stars](https://img.shields.io/github/stars/dgSPARSE/dgNN.svg?logo=github&label=Stars)| 54 | |HPDC 2022|TLPGNN: A Lightweight Two-Level Parallelism Paradigm for Graph Neural Network Computation on GPU|GW| [[paper]](https://dl.acm.org/doi/abs/10.1145/3502181.3531467)![Scholar citations](https://img.shields.io/badge/Citations-0-_.svg?logo=google-scholar&labelColor=4f4f4f&color=3388ee)|| 55 | |IPDPS 2021|FusedMM: A Unified SDDMM-SpMM Kernel for Graph Embedding and Graph Neural Networks|Indiana University Bloomington| [[paper]](https://arxiv.org/abs/2011.06391)![Scholar citations](https://img.shields.io/badge/Citations-0-_.svg?logo=google-scholar&labelColor=4f4f4f&color=3388ee)| [[code]](https://github.com/HipGraph/FusedMM)![GitHub stars](https://img.shields.io/github/stars/HipGraph/FusedMM.svg?logo=github&label=Stars)| 56 | |SC 2020|GE-SpMM: General-purpose Sparse Matrix-Matrix Multiplication on GPUs for Graph Neural Networks|THU| [[paper]](https://arxiv.org/abs/2007.03179)![Scholar citations](https://img.shields.io/badge/Citations-0-_.svg?logo=google-scholar&labelColor=4f4f4f&color=3388ee)| [[code]](https://github.com/hgyhungry/ge-spmm)![GitHub stars](https://img.shields.io/github/stars/hgyhungry/ge-spmm.svg?logo=github&label=Stars)| 57 | |ICCAD 2020|fuseGNN: Accelerating Graph Convolutional Neural Network Training on GPGPU|UCSB| [[paper]](https://seal.ece.ucsb.edu/sites/default/files/publications/fusegcn_camera_ready_.pdf)![Scholar citations](https://img.shields.io/badge/Citations-0-_.svg?logo=google-scholar&labelColor=4f4f4f&color=3388ee)| [[code]](https://github.com/apuaaChen/gcnLib)![GitHub stars](https://img.shields.io/github/stars/apuaaChen/gcnLib.svg?logo=github&label=Stars)| 58 | |IPDPS 2020|PCGCN: Partition-Centric Processing for Accelerating Graph Convolutional Network|PKU| [[paper]](https://ieeexplore.ieee.org/document/9139807)![Scholar citations](https://img.shields.io/badge/Citations-0-_.svg?logo=google-scholar&labelColor=4f4f4f&color=3388ee)|| 59 | ### GNN Compilers 60 | | Venue | Title | Affiliation |       Link       |   Source   | 61 | | :---: | :---: | :---------: | :---: | :----: | 62 | |MLSys 2022|Graphiler: Optimizing Graph Neural Networks with Message Passing Data Flow Graph|ShanghaiTech| [[paper]](https://proceedings.mlsys.org/paper/2022/file/a87ff679a2f3e71d9181a67b7542122c-Paper.pdf)![Scholar citations](https://img.shields.io/badge/Citations-0-_.svg?logo=google-scholar&labelColor=4f4f4f&color=3388ee)| [[code]](https://github.com/xiezhq-hermann/graphiler)![GitHub stars](https://img.shields.io/github/stars/xiezhq-hermann/graphiler.svg?logo=github&label=Stars)| 63 | |EuroSys 2021|Seastar: Vertex-Centric Programming for Graph Neural Networks|CUHK| [[paper]](http://www.cse.cuhk.edu.hk/~jcheng/papers/seastar_eurosys21.pdf)![Scholar citations](https://img.shields.io/badge/Citations-0-_.svg?logo=google-scholar&labelColor=4f4f4f&color=3388ee)|| 64 | |SC 2020|FeatGraph: A Flexible and Efficient Backend for Graph Neural Network Systems|Cornell| [[paper]](https://arxiv.org/pdf/2008.11359.pdf)![Scholar citations](https://img.shields.io/badge/Citations-0-_.svg?logo=google-scholar&labelColor=4f4f4f&color=3388ee)| [[code]](https://github.com/dglai/FeatGraph)![GitHub stars](https://img.shields.io/github/stars/dglai/FeatGraph.svg?logo=github&label=Stars)| 65 | ### Distributed GNN Training Systems 66 | | Venue | Title | Affiliation |       Link       |   Source   | 67 | | :---: | :---: | :---------: | :---: | :----: | 68 | |FAST 2025|LeapGNN: Accelerating Distributed GNN Training Leveraging Feature-Centric Model Migration|ZJU| [[paper]](https://www.usenix.org/system/files/fast25-chen-weijian-leap.pdf)![Scholar citations](https://img.shields.io/badge/Citations-0-_.svg?logo=google-scholar&labelColor=4f4f4f&color=3388ee)| [[code]](https://github.com/ISCS-ZJU/LeapGNN-AE)![GitHub stars](https://img.shields.io/github/stars/ISCS-ZJU/LeapGNN-AE.svg?logo=github&label=Stars)| 69 | |VLDB 2025|NeutronTP: Load-Balanced Distributed Full-Graph GNN Training with Tensor Parallelism |NEU| [[paper]](https://arxiv.org/abs/2412.20379)![Scholar citations](https://img.shields.io/badge/Citations-0-_.svg?logo=google-scholar&labelColor=4f4f4f&color=3388ee)| [[code]](https://github.com/iDC-NEU/NeutronTP)![GitHub stars](https://img.shields.io/github/stars/iDC-NEU/NeutronTP.svg?logo=github&label=Stars)| 70 | |arXiv 2023|Communication-Free Distributed GNN Training with Vertex Cut|Stanford| [[paper]](https://arxiv.org/pdf/2308.03209.pdf)![Scholar citations](https://img.shields.io/badge/Citations-0-_.svg?logo=google-scholar&labelColor=4f4f4f&color=3388ee)|| 71 | |arXiv 2023|GNNPipe: Accelerating Distributed Full-Graph GNN Training with Pipelined Model Parallelism|Purdue| [[paper]](https://arxiv.org/pdf/2308.10087.pdf)![Scholar citations](https://img.shields.io/badge/Citations-0-_.svg?logo=google-scholar&labelColor=4f4f4f&color=3388ee)|| 72 | |OSDI 2023|MGG: Accelerating Graph Neural Networks with Fine-Grained Intra-Kernel Communication-Computation Pipelining on Multi-GPU Platforms|UCSB| [[paper]](https://www.usenix.org/system/files/osdi23-wang-yuke.pdf)![Scholar citations](https://img.shields.io/badge/Citations-0-_.svg?logo=google-scholar&labelColor=4f4f4f&color=3388ee)| [[code]](https://github.com/YukeWang96/MGG_OSDI23)![GitHub stars](https://img.shields.io/github/stars/YukeWang96/MGG_OSDI23.svg?logo=github&label=Stars)| 73 | |SIGMOD 2022|NeutronStar: Distributed GNN Training with Hybrid Management |NEU| [[paper]](https://dl.acm.org/doi/pdf/10.1145/3514221.3526134)![Scholar citations](https://img.shields.io/badge/Citations-0-_.svg?logo=google-scholar&labelColor=4f4f4f&color=3388ee)| [[code]](https://github.com/iDC-NEU/NeutronStarLite)![GitHub stars](https://img.shields.io/github/stars/iDC-NEU/NeutronStarLite.svg?logo=github&label=Stars)| 74 | |VLDB 2022|Sancus: Staleness-Aware Communication-Avoiding Full-Graph Decentralized Training in Large-Scale Graph Neural Networks|HKUST| [[paper]](https://www.vldb.org/pvldb/vol15/p1937-peng.pdf)![Scholar citations](https://img.shields.io/badge/Citations-0-_.svg?logo=google-scholar&labelColor=4f4f4f&color=3388ee)| [[code]](https://github.com/chenzhao/light-dist-gnn)![GitHub stars](https://img.shields.io/github/stars/chenzhao/light-dist-gnn.svg?logo=github&label=Stars)| 75 | |MLSys 2022|BNS-GCN: Efficient Full-Graph Training of Graph Convolutional Networks with Partition-Parallelism and Random Boundary Node Sampling|Rice, UIUC| [[paper]](https://proceedings.mlsys.org/paper/2022/file/d1fe173d08e959397adf34b1d77e88d7-Paper.pdf)![Scholar citations](https://img.shields.io/badge/Citations-0-_.svg?logo=google-scholar&labelColor=4f4f4f&color=3388ee)| [[code]](https://github.com/RICE-EIC/BNS-GCN)![GitHub stars](https://img.shields.io/github/stars/RICE-EIC/BNS-GCN.svg?logo=github&label=Stars)| 76 | |MLSys 2022|Sequential Aggregation and Rematerialization: Distributed Full-batch Training of Graph Neural Networks on Large Graphs|Intel| [[paper]](https://arxiv.org/abs/2111.06483)![Scholar citations](https://img.shields.io/badge/Citations-0-_.svg?logo=google-scholar&labelColor=4f4f4f&color=3388ee)| [[code]](https://github.com/IntelLabs/SAR)![GitHub stars](https://img.shields.io/github/stars/IntelLabs/SAR.svg?logo=github&label=Stars)| 77 | |WWW 2022|PaSca: A Graph Neural Architecture Search System under the Scalable Paradigm|PKU| [[paper]](https://dl.acm.org/doi/abs/10.1145/3485447.3511986)![Scholar citations](https://img.shields.io/badge/Citations-0-_.svg?logo=google-scholar&labelColor=4f4f4f&color=3388ee)|| 78 | |ICLR 2022|PipeGCN: Efficient Full-Graph Training of Graph Convolutional Networks with Pipelined Feature Communication|Rice| [[paper]](https://openreview.net/pdf?id=kSwqMH0zn1F)![Scholar citations](https://img.shields.io/badge/Citations-0-_.svg?logo=google-scholar&labelColor=4f4f4f&color=3388ee)| [[code]](https://github.com/RICE-EIC/PipeGCN)![GitHub stars](https://img.shields.io/github/stars/RICE-EIC/PipeGCN.svg?logo=github&label=Stars)| 79 | |ICLR 2022|Learn Locally, Correct Globally: A Distributed Algorithm for Training Graph Neural Networks|PSU| [[paper]](https://openreview.net/pdf?id=FndDxSz3LxQ)![Scholar citations](https://img.shields.io/badge/Citations-0-_.svg?logo=google-scholar&labelColor=4f4f4f&color=3388ee)| [[code]](https://github.com/MortezaRamezani/llcg)![GitHub stars](https://img.shields.io/github/stars/MortezaRamezani/llcg.svg?logo=github&label=Stars)| 80 | |arXiv 2021|Distributed Hybrid CPU and GPU training for Graph Neural Networks on Billion-Scale Graphs|AWS| [[paper]](https://arxiv.org/pdf/2112.15345.pdf)![Scholar citations](https://img.shields.io/badge/Citations-0-_.svg?logo=google-scholar&labelColor=4f4f4f&color=3388ee)|| 81 | |SC 2021|DistGNN: Scalable Distributed Training for Large-Scale Graph Neural Networks|Intel| [[paper]](https://dl.acm.org/doi/pdf/10.1145/3458817.3480856)![Scholar citations](https://img.shields.io/badge/Citations-0-_.svg?logo=google-scholar&labelColor=4f4f4f&color=3388ee)| [[code]](https://github.com/dmlc/dgl/pull/3024)| 82 | |SC 2021|Efficient Scaling of Dynamic Graph Neural Networks|IBM| [[paper]](https://dl.acm.org/doi/pdf/10.1145/3458817.3480858)![Scholar citations](https://img.shields.io/badge/Citations-0-_.svg?logo=google-scholar&labelColor=4f4f4f&color=3388ee)|| 83 | |CLUSTER 2021|2PGraph: Accelerating GNN Training over Large Graphs on GPU Clusters|NUDT| [[paper]](https://ieeexplore.ieee.org/abstract/document/9556026)![Scholar citations](https://img.shields.io/badge/Citations-0-_.svg?logo=google-scholar&labelColor=4f4f4f&color=3388ee)|| 84 | |OSDI 2021|$P^3$: Distributed Deep Graph Learning at Scale|MSR| [[paper]](https://www.usenix.org/system/files/osdi21-gandhi.pdf)![Scholar citations](https://img.shields.io/badge/Citations-0-_.svg?logo=google-scholar&labelColor=4f4f4f&color=3388ee)|| 85 | |OSDI 2021|Dorylus: Affordable, Scalable, and Accurate GNN Training with Distributed CPU Servers and Serverless Threads|UCLA| [[paper]](http://web.cs.ucla.edu/~harryxu/papers/dorylus-osdi21.pdf)![Scholar citations](https://img.shields.io/badge/Citations-0-_.svg?logo=google-scholar&labelColor=4f4f4f&color=3388ee)| [[code]](https://github.com/uclasystem/dorylus)![GitHub stars](https://img.shields.io/github/stars/uclasystem/dorylus.svg?logo=github&label=Stars)| 86 | |arXiv 2021|GIST: Distributed Training for Large-Scale Graph Convolutional Networks|Rice| [[paper]](https://arxiv.org/abs/2102.10424)![Scholar citations](https://img.shields.io/badge/Citations-0-_.svg?logo=google-scholar&labelColor=4f4f4f&color=3388ee)|| 87 | |EuroSys 2021|FlexGraph: A Flexible and Efficient Distributed Framework for GNN Training|Alibaba| [[paper]](https://dl.acm.org/doi/pdf/10.1145/3447786.3456229)![Scholar citations](https://img.shields.io/badge/Citations-0-_.svg?logo=google-scholar&labelColor=4f4f4f&color=3388ee)|| 88 | |EuroSys 2021|DGCL: An Efficient Communication Library for Distributed GNN Training|CUHK| [[paper]](https://dl.acm.org/doi/abs/10.1145/3447786.3456233)![Scholar citations](https://img.shields.io/badge/Citations-0-_.svg?logo=google-scholar&labelColor=4f4f4f&color=3388ee)| [[code]](https://github.com/czkkkkkk/ragdoll)![GitHub stars](https://img.shields.io/github/stars/czkkkkkk/ragdoll.svg?logo=github&label=Stars)| 89 | |SC 2020|Reducing Communication in Graph Neural Network Training|UC Berkeley| [[paper]](https://arxiv.org/pdf/2005.03300.pdf)![Scholar citations](https://img.shields.io/badge/Citations-0-_.svg?logo=google-scholar&labelColor=4f4f4f&color=3388ee)| [[code]](https://github.com/PASSIONLab/CAGNET)![GitHub stars](https://img.shields.io/github/stars/PASSIONLab/CAGNET.svg?logo=github&label=Stars)| 90 | |VLDB 2020|G$^3$: When Graph Neural Networks Meet Parallel Graph Processing Systems on GPUs|NUS| [[paper]](http://www.vldb.org/pvldb/vol13/p2813-liu.pdf)![Scholar citations](https://img.shields.io/badge/Citations-0-_.svg?logo=google-scholar&labelColor=4f4f4f&color=3388ee)| [[code]](https://github.com/Xtra-Computing/G3)![GitHub stars](https://img.shields.io/github/stars/Xtra-Computing/G3.svg?logo=github&label=Stars)| 91 | |IA3 2020|DistDGL: Distributed Graph Neural Network Training for Billion-Scale Graphs|AWS| [[paper]](https://arxiv.org/pdf/2010.05337.pdf)![Scholar citations](https://img.shields.io/badge/Citations-0-_.svg?logo=google-scholar&labelColor=4f4f4f&color=3388ee)| [[code]](https://github.com/dmlc/dgl/tree/master/python/dgl/distributed)| 92 | |MLSys 2020|Improving the Accuracy, Scalability, and Performance of Graph Neural Networks with Roc|Stanford| [[paper]](https://proceedings.mlsys.org/paper/2020/file/fe9fc289c3ff0af142b6d3bead98a923-Paper.pdf)![Scholar citations](https://img.shields.io/badge/Citations-0-_.svg?logo=google-scholar&labelColor=4f4f4f&color=3388ee)| [[code]](https://github.com/jiazhihao/ROC)![GitHub stars](https://img.shields.io/github/stars/jiazhihao/ROC.svg?logo=github&label=Stars)| 93 | |arXiv 2020|AGL: A Scalable System for Industrial-purpose Graph Machine Learning|Ant Financial Services Group| [[paper]](https://arxiv.org/abs/2003.02454)![Scholar citations](https://img.shields.io/badge/Citations-0-_.svg?logo=google-scholar&labelColor=4f4f4f&color=3388ee)|| 94 | |ATC 2019|NeuGraph: Parallel Deep Neural Network Computation on Large Graphs|PKU| [[paper]](https://www.usenix.org/system/files/atc19-ma_0.pdf)![Scholar citations](https://img.shields.io/badge/Citations-0-_.svg?logo=google-scholar&labelColor=4f4f4f&color=3388ee)|| 95 | ### Training Systems for Scaling Graphs 96 | | Venue | Title | Affiliation |       Link       |   Source   | 97 | | :---: | :---: | :---------: | :---: | :----: | 98 | |ICDE 2025|Hyperion: Optimizing SSD Access is All You Need to Enable Cost-efficient Out-of-core GNN Training|ZJU| [[paper]](https://wangzeke.github.io/doc/Hyperion-ICDE25.pdf)![Scholar citations](https://img.shields.io/badge/Citations-0-_.svg?logo=google-scholar&labelColor=4f4f4f&color=3388ee)| [[code]](https://github.com/RC4ML/Hyperion)![GitHub stars](https://img.shields.io/github/stars/RC4ML/Hyperion.svg?logo=github&label=Stars)| 99 | |DaMoN 2024|In situ neighborhood sampling for large-scale GNN training|Boston University| [[paper]](https://dl.acm.org/doi/10.1145/3662010.3663443)![Scholar citations](https://img.shields.io/badge/Citations-0-_.svg?logo=google-scholar&labelColor=4f4f4f&color=3388ee)| [[code]](https://github.com/CASP-Systems-BU/damon24-gnn-in-situ-sampling/)| 100 | |HPCA 2024|BeaconGNN: Large-Scale GNN Acceleration with Out-of-Order Streaming In-Storage Computing|UCLA| [[paper]](https://ieeexplore.ieee.org/abstract/document/10476427)![Scholar citations](https://img.shields.io/badge/Citations-0-_.svg?logo=google-scholar&labelColor=4f4f4f&color=3388ee)|| 101 | |EuroSys 2023|MariusGNN: Resource-Efficient Out-of-Core Training of Graph Neural Networks|UW–Madison| [[paper]](https://arxiv.org/abs/2202.02365)![Scholar citations](https://img.shields.io/badge/Citations-0-_.svg?logo=google-scholar&labelColor=4f4f4f&color=3388ee)| [[code]](https://github.com/marius-team/marius)![GitHub stars](https://img.shields.io/github/stars/marius-team/marius.svg?logo=github&label=Stars)| 102 | |VLDB 2022|ByteGNN: Efficient Graph Neural Network Training at Large Scale|ByteDance| [[paper]](https://dl.acm.org/doi/abs/10.14778/3514061.3514069)![Scholar citations](https://img.shields.io/badge/Citations-0-_.svg?logo=google-scholar&labelColor=4f4f4f&color=3388ee)|| 103 | |VLDB 2022|Ginex: SSD-enabled Billion-scale Graph Neural Network Training on a Single Machine via Provably Optimal In-memory Caching|Seoul National University| [[paper]](https://dl.acm.org/doi/10.14778/3551793.3551819)![Scholar citations](https://img.shields.io/badge/Citations-0-_.svg?logo=google-scholar&labelColor=4f4f4f&color=3388ee)| [[code]](https://github.com/SNU-ARC/Ginex)![GitHub stars](https://img.shields.io/github/stars/SNU-ARC/Ginex.svg?logo=github&label=Stars)| 104 | |ISCA 2022|SmartSAGE: Training Large-scale Graph Neural Networks using In-Storage Processing Architectures|KAIST| [[paper]](https://dl.acm.org/doi/10.1145/3470496.3527391)![Scholar citations](https://img.shields.io/badge/Citations-0-_.svg?logo=google-scholar&labelColor=4f4f4f&color=3388ee)|| 105 | |ICML 2022|GraphFM: Improving Large-Scale GNN Training via Feature Momentum|TAMU| [[paper]](https://arxiv.org/abs/2206.07161)![Scholar citations](https://img.shields.io/badge/Citations-0-_.svg?logo=google-scholar&labelColor=4f4f4f&color=3388ee)| [[code]](https://github.com/divelab/DIG/tree/dig-stable/dig/lsgraph)| 106 | |ICML 2021|GNNAutoScale: Scalable and Expressive Graph Neural Networks via Historical Embeddings|TU Dortmund University| [[paper]](https://arxiv.org/abs/2106.05609)![Scholar citations](https://img.shields.io/badge/Citations-0-_.svg?logo=google-scholar&labelColor=4f4f4f&color=3388ee)| [[code]](https://github.com/rusty1s/pyg_autoscale)![GitHub stars](https://img.shields.io/github/stars/rusty1s/pyg_autoscale.svg?logo=github&label=Stars)| 107 | |OSDI 2021|GNNAdvisor: An Adaptive and Efficient Runtime System for GNN Acceleration on GPUs|UCSB| [[paper]](https://www.usenix.org/system/files/osdi21-wang-yuke.pdf)![Scholar citations](https://img.shields.io/badge/Citations-0-_.svg?logo=google-scholar&labelColor=4f4f4f&color=3388ee)| [[code]](https://github.com/YukeWang96/OSDI21_AE)![GitHub stars](https://img.shields.io/github/stars/YukeWang96/OSDI21_AE.svg?logo=github&label=Stars)| 108 | ### Quantized GNNs 109 | | Venue | Title | Affiliation |       Link       |   Source   | 110 | | :---: | :---: | :---------: | :---: | :----: | 111 | |Neurocomputing 2022|EPQuant: A Graph Neural Network Compression Approach Based on Product Quantization|ZJU| [[paper]](https://www.sciencedirect.com/science/article/abs/pii/S0925231222008293)![Scholar citations](https://img.shields.io/badge/Citations-0-_.svg?logo=google-scholar&labelColor=4f4f4f&color=3388ee)| [[code]](https://github.com/Lyun-Huang/EPQuant)![GitHub stars](https://img.shields.io/github/stars/Lyun-Huang/EPQuant.svg?logo=github&label=Stars)| 112 | |ICLR 2022|EXACT: Scalable Graph Neural Networks Training via Extreme Activation Compression|Rice| [[paper]](https://openreview.net/pdf?id=vkaMaq95_rX)![Scholar citations](https://img.shields.io/badge/Citations-0-_.svg?logo=google-scholar&labelColor=4f4f4f&color=3388ee)| [[code]](https://github.com/warai-0toko/Exact)![GitHub stars](https://img.shields.io/github/stars/warai-0toko/Exact.svg?logo=github&label=Stars)| 113 | |PPoPP 2022|QGTC: Accelerating Quantized Graph Neural Networks via GPU Tensor Core|UCSB| [[paper]](https://arxiv.org/abs/2111.09547)![Scholar citations](https://img.shields.io/badge/Citations-0-_.svg?logo=google-scholar&labelColor=4f4f4f&color=3388ee)| [[code]](https://github.com/YukeWang96/PPoPP22_QGTC)![GitHub stars](https://img.shields.io/github/stars/YukeWang96/PPoPP22_QGTC.svg?logo=github&label=Stars)| 114 | |CVPR 2021|Binary Graph Neural Networks|ICL| [[paper]](https://arxiv.org/abs/2012.15823)![Scholar citations](https://img.shields.io/badge/Citations-0-_.svg?logo=google-scholar&labelColor=4f4f4f&color=3388ee)| [[code]](https://github.com/mbahri/binary_gnn)![GitHub stars](https://img.shields.io/github/stars/mbahri/binary_gnn.svg?logo=github&label=Stars)| 115 | |CVPR 2021|Bi-GCN: Binary Graph Convolutional Network|Beihang University| [[paper]](https://openaccess.thecvf.com/content/CVPR2021/html/Wang_Bi-GCN_Binary_Graph_Convolutional_Network_CVPR_2021_paper.html)![Scholar citations](https://img.shields.io/badge/Citations-0-_.svg?logo=google-scholar&labelColor=4f4f4f&color=3388ee)| [[code]](https://github.com/bywmm/Bi-GCN)![GitHub stars](https://img.shields.io/github/stars/bywmm/Bi-GCN.svg?logo=github&label=Stars)| 116 | |EuroMLSys 2021|Learned Low Precision Graph Neural Networks|Cambridge| [[paper]](https://arxiv.org/abs/2009.09232)![Scholar citations](https://img.shields.io/badge/Citations-0-_.svg?logo=google-scholar&labelColor=4f4f4f&color=3388ee)|| 117 | |World Wide Web 2021|Binarized Graph Neural Network|UTS| [[paper]](https://arxiv.org/pdf/2004.11147.pdf)![Scholar citations](https://img.shields.io/badge/Citations-0-_.svg?logo=google-scholar&labelColor=4f4f4f&color=3388ee)|| 118 | |ICLR 2021|Degree-Quant: Quantization-Aware Training for Graph Neural Networks|Cambridge| [[paper]](https://arxiv.org/abs/2008.05000)![Scholar citations](https://img.shields.io/badge/Citations-0-_.svg?logo=google-scholar&labelColor=4f4f4f&color=3388ee)| [[code]](https://github.com/camlsys/degree-quant)![GitHub stars](https://img.shields.io/github/stars/camlsys/degree-quant.svg?logo=github&label=Stars)| 119 | |ICTAI 2020|SGQuant: Squeezing the Last Bit on Graph Neural Networks with Specialized Quantization|UCSB| [[paper]](https://ieeexplore.ieee.org/abstract/document/9288186/)![Scholar citations](https://img.shields.io/badge/Citations-0-_.svg?logo=google-scholar&labelColor=4f4f4f&color=3388ee)| [[code]](https://github.com/YukeWang96/SGQuant)![GitHub stars](https://img.shields.io/github/stars/YukeWang96/SGQuant.svg?logo=github&label=Stars)| 120 | ### GNN Dataloaders 121 | | Venue | Title | Affiliation |       Link       |   Source   | 122 | | :---: | :---: | :---------: | :---: | :----: | 123 | |VLDB 2024|Accelerating Sampling and Aggregation Operations in GNN Frameworks with GPU Initiated Direct Storage Accesses|UIUC| [[paper]](https://arxiv.org/abs/2306.16384v2)![Scholar citations](https://img.shields.io/badge/Citations-0-_.svg?logo=google-scholar&labelColor=4f4f4f&color=3388ee)| [[code]](https://github.com/jeongminpark417/GIDS)![GitHub stars](https://img.shields.io/github/stars/jeongminpark417/GIDS.svg?logo=github&label=Stars)| 124 | |NSDI 2023|BGL: GPU-Efficient GNN Training by Optimizing Graph Data I/O and Preprocessing|ByteDance| [[paper]](https://arxiv.org/abs/2112.08541)![Scholar citations](https://img.shields.io/badge/Citations-0-_.svg?logo=google-scholar&labelColor=4f4f4f&color=3388ee)|| 125 | |MLSys 2022|Accelerating Training and Inference of Graph Neural Networks with Fast Sampling and Pipelining|MIT| [[paper]](https://proceedings.mlsys.org/paper/2022/file/35f4a8d465e6e1edc05f3d8ab658c551-Paper.pdf)![Scholar citations](https://img.shields.io/badge/Citations-0-_.svg?logo=google-scholar&labelColor=4f4f4f&color=3388ee)| [[code]](https://github.com/MITIBMxGraph/SALIENT)![GitHub stars](https://img.shields.io/github/stars/MITIBMxGraph/SALIENT.svg?logo=github&label=Stars)| 126 | |EuroSys 2022|GNNLab: A Factored System for Sample-based GNN Training over GPUs|SJTU| [[paper]](https://dl.acm.org/doi/abs/10.1145/3492321.3519557)![Scholar citations](https://img.shields.io/badge/Citations-0-_.svg?logo=google-scholar&labelColor=4f4f4f&color=3388ee)| [[code]](https://github.com/SJTU-IPADS/gnnlab)![GitHub stars](https://img.shields.io/github/stars/SJTU-IPADS/gnnlab.svg?logo=github&label=Stars)| 127 | |KDD 2021|Global Neighbor Sampling for Mixed CPU-GPU Training on Giant Graphs|UCLA| [[paper]](https://arxiv.org/pdf/2106.06150.pdf)![Scholar citations](https://img.shields.io/badge/Citations-0-_.svg?logo=google-scholar&labelColor=4f4f4f&color=3388ee)|| 128 | |PPoPP 2021|Understanding and Bridging the Gaps in Current GNN Performance Optimizations|THU| [[paper]](https://dl.acm.org/doi/pdf/10.1145/3437801.3441585)![Scholar citations](https://img.shields.io/badge/Citations-0-_.svg?logo=google-scholar&labelColor=4f4f4f&color=3388ee)| [[code]](https://github.com/xxcclong/GNN-Computing)![GitHub stars](https://img.shields.io/github/stars/xxcclong/GNN-Computing.svg?logo=github&label=Stars)| 129 | |VLDB 2021|Large Graph Convolutional Network Training with GPU-Oriented Data Communication Architecture|UIUC| [[paper]](https://arxiv.org/abs/2103.03330)![Scholar citations](https://img.shields.io/badge/Citations-0-_.svg?logo=google-scholar&labelColor=4f4f4f&color=3388ee)| [[code]](https://github.com/K-Wu/pytorch-direct_dgl)![GitHub stars](https://img.shields.io/github/stars/K-Wu/pytorch-direct_dgl.svg?logo=github&label=Stars)| 130 | |TPDS 2021|Efficient Data Loader for Fast Sampling-Based GNN Training on Large Graphs|USTC| [[paper]](https://gnnsys.github.io/papers/GNNSys21_paper_8.pdf)![Scholar citations](https://img.shields.io/badge/Citations-0-_.svg?logo=google-scholar&labelColor=4f4f4f&color=3388ee)| [[code]](https://github.com/zhiqi-0/PaGraph)![GitHub stars](https://img.shields.io/github/stars/zhiqi-0/PaGraph.svg?logo=github&label=Stars)| 131 | |SoCC 2020|PaGraph: Scaling GNN Training on Large Graphs via Computation-aware Caching|USTC| [[paper]](https://dl.acm.org/doi/pdf/10.1145/3419111.3421281)![Scholar citations](https://img.shields.io/badge/Citations-0-_.svg?logo=google-scholar&labelColor=4f4f4f&color=3388ee)| [[code]](https://github.com/zhiqi-0/PaGraph)![GitHub stars](https://img.shields.io/github/stars/zhiqi-0/PaGraph.svg?logo=github&label=Stars)| 132 | |arXiv 2019|TigerGraph: A Native MPP Graph Database|UCSD| [[paper]](https://arxiv.org/pdf/1901.08248.pdf)![Scholar citations](https://img.shields.io/badge/Citations-0-_.svg?logo=google-scholar&labelColor=4f4f4f&color=3388ee)|| 133 | ### GNN Training Accelerators 134 | | Venue | Title | Affiliation |       Link       |   Source   | 135 | | :---: | :---: | :---------: | :---: | :----: | 136 | |ISCA 2022|Graphite: Optimizing Graph Neural Networks on CPUs Through Cooperative Software-Hardware Techniques|UIUC| [[paper]](http://iacoma.cs.uiuc.edu/iacoma-papers/isca22.pdf)![Scholar citations](https://img.shields.io/badge/Citations-0-_.svg?logo=google-scholar&labelColor=4f4f4f&color=3388ee)|| 137 | |ISCA 2022|Hyperscale FPGA-as-a-service architecture for large-scale distributed graph neural network|Alibaba| [[paper]](https://dl.acm.org/doi/abs/10.1145/3470496.3527439)![Scholar citations](https://img.shields.io/badge/Citations-0-_.svg?logo=google-scholar&labelColor=4f4f4f&color=3388ee)|| 138 | |arXiv 2021|GCNear: A Hybrid Architecture for Efficient GCN Training with Near-Memory Processing|PKU| [[paper]](https://arxiv.org/abs/2111.00680)![Scholar citations](https://img.shields.io/badge/Citations-0-_.svg?logo=google-scholar&labelColor=4f4f4f&color=3388ee)|| 139 | |DATE 2021|ReGraphX: NoC-enabled 3D Heterogeneous ReRAM Architecture for Training Graph Neural Networks|WSU| [[paper]](https://arxiv.org/abs/2102.07959)![Scholar citations](https://img.shields.io/badge/Citations-0-_.svg?logo=google-scholar&labelColor=4f4f4f&color=3388ee)|| 140 | |TCAD 2021|Rubik: A Hierarchical Architecture for Efficient Graph Learning|Chinese Academy of Sciences| [[paper]](https://arxiv.org/pdf/2009.12495.pdf)![Scholar citations](https://img.shields.io/badge/Citations-0-_.svg?logo=google-scholar&labelColor=4f4f4f&color=3388ee)|| 141 | |FPGA 2020|GraphACT: Accelerating GCN Training on CPU-FPGA Heterogeneous Platforms|USC| [[paper]](https://dl.acm.org/doi/pdf/10.1145/3373087.3375312)![Scholar citations](https://img.shields.io/badge/Citations-0-_.svg?logo=google-scholar&labelColor=4f4f4f&color=3388ee)| [[code]](https://github.com/GraphSAINT/GraphACT)![GitHub stars](https://img.shields.io/github/stars/GraphSAINT/GraphACT.svg?logo=github&label=Stars)| 142 | ### GNN Inference Accelerators 143 | | Venue | Title | Affiliation |       Link       |   Source   | 144 | | :---: | :---: | :---------: | :---: | :----: | 145 | |JAIHC 2022|DRGN: a dynamically reconfigurable accelerator for graph neural networks|XJTU| [[paper]](https://link.springer.com/article/10.1007/s12652-022-04402-x)![Scholar citations](https://img.shields.io/badge/Citations-0-_.svg?logo=google-scholar&labelColor=4f4f4f&color=3388ee)|| 146 | |DAC 2022|GNNIE: GNN Inference Engine with Load-balancing and Graph-specific Caching|UMN| [[paper]](https://arxiv.org/abs/2105.10554)![Scholar citations](https://img.shields.io/badge/Citations-0-_.svg?logo=google-scholar&labelColor=4f4f4f&color=3388ee)|| 147 | |IPDPS 2022|Understanding the Design Space of Sparse/Dense Multiphase Dataflows for Mapping Graph Neural Networks on Spatial Accelerators|GaTech| [[paper]](https://arxiv.org/abs/2103.07977)![Scholar citations](https://img.shields.io/badge/Citations-0-_.svg?logo=google-scholar&labelColor=4f4f4f&color=3388ee)| [[code]](https://github.com/stonne-simulator/omega)![GitHub stars](https://img.shields.io/github/stars/stonne-simulator/omega.svg?logo=github&label=Stars)| 148 | |arXiv 2022|FlowGNN: A Dataflow Architecture for Universal Graph Neural Network Inference via Multi-Queue Streaming|GaTech| [[paper]](https://arxiv.org/abs/2204.13103)![Scholar citations](https://img.shields.io/badge/Citations-0-_.svg?logo=google-scholar&labelColor=4f4f4f&color=3388ee)|| 149 | |CICC 2022|StreamGCN: Accelerating Graph Convolutional Networks with Streaming Processing|UCLA| [[paper]](https://web.cs.ucla.edu/~atefehsz/publication/StreamGCN-CICC22.pdf)![Scholar citations](https://img.shields.io/badge/Citations-0-_.svg?logo=google-scholar&labelColor=4f4f4f&color=3388ee)|| 150 | |HPCA 2022|Accelerating Graph Convolutional Networks Using Crossbar-based Processing-In-Memory Architectures|HUST| [[paper]](https://ieeexplore.ieee.org/document/9773267)![Scholar citations](https://img.shields.io/badge/Citations-0-_.svg?logo=google-scholar&labelColor=4f4f4f&color=3388ee)|| 151 | |HPCA 2022|GCoD: Graph Convolutional Network Acceleration via Dedicated Algorithm and Accelerator Co-Design|Rice, PNNL| [[paper]](https://arxiv.org/abs/2112.11594)![Scholar citations](https://img.shields.io/badge/Citations-0-_.svg?logo=google-scholar&labelColor=4f4f4f&color=3388ee)| [[code]](https://github.com/RICE-EIC/GCoD)![GitHub stars](https://img.shields.io/github/stars/RICE-EIC/GCoD.svg?logo=github&label=Stars)| 152 | |arXiv 2022|GenGNN: A Generic FPGA Framework for Graph Neural Network Acceleration|GaTech| [[paper]](https://arxiv.org/abs/2201.08475)![Scholar citations](https://img.shields.io/badge/Citations-0-_.svg?logo=google-scholar&labelColor=4f4f4f&color=3388ee)|| 153 | |DAC 2021|DyGNN: Algorithm and Architecture Support of vertex Dynamic Pruning for Graph Neural Networks|Hunan University| [[paper]](https://ieeexplore.ieee.org/document/9586298)![Scholar citations](https://img.shields.io/badge/Citations-0-_.svg?logo=google-scholar&labelColor=4f4f4f&color=3388ee)|| 154 | |DAC 2021|BlockGNN: Towards Efficient GNN Acceleration Using Block-Circulant Weight Matrices|PKU| [[paper]](https://arxiv.org/abs/2104.06214)![Scholar citations](https://img.shields.io/badge/Citations-0-_.svg?logo=google-scholar&labelColor=4f4f4f&color=3388ee)|| 155 | |DAC 2021|TARe: Task-Adaptive in-situ ReRAM Computing for Graph Learning|Chinese Academy of Sciences| [[paper]](https://ieeexplore.ieee.org/abstract/document/9586193)![Scholar citations](https://img.shields.io/badge/Citations-0-_.svg?logo=google-scholar&labelColor=4f4f4f&color=3388ee)|| 156 | |ICCAD 2021|G-CoS: GNN-Accelerator Co-Search Towards Both Better Accuracy and Efficiency|Rice| [[paper]](https://arxiv.org/abs/2109.08983)![Scholar citations](https://img.shields.io/badge/Citations-0-_.svg?logo=google-scholar&labelColor=4f4f4f&color=3388ee)|| 157 | |MICRO 2021|I-GCN: A Graph Convolutional Network Accelerator with Runtime Locality Enhancement through Islandization|PNNL| [[paper]](https://dl.acm.org/doi/pdf/10.1145/3466752.3480113)![Scholar citations](https://img.shields.io/badge/Citations-0-_.svg?logo=google-scholar&labelColor=4f4f4f&color=3388ee)|| 158 | |arXiv 2021|ZIPPER: Exploiting Tile- and Operator-level Parallelism for General and Scalable Graph Neural Network Acceleration|SJTU| [[paper]](https://arxiv.org/abs/2107.08709)![Scholar citations](https://img.shields.io/badge/Citations-0-_.svg?logo=google-scholar&labelColor=4f4f4f&color=3388ee)|| 159 | |TComp 2021|EnGN: A High-Throughput and Energy-Efficient Accelerator for Large Graph Neural Networks|Chinese Academy of Sciences| [[paper]](https://arxiv.org/abs/1909.00155)![Scholar citations](https://img.shields.io/badge/Citations-0-_.svg?logo=google-scholar&labelColor=4f4f4f&color=3388ee)|| 160 | |HPCA 2021|GCNAX: A Flexible and Energy-efficient Accelerator for Graph Convolutional Neural Networks|GWU| [[paper]](https://ieeexplore.ieee.org/abstract/document/9407104)![Scholar citations](https://img.shields.io/badge/Citations-0-_.svg?logo=google-scholar&labelColor=4f4f4f&color=3388ee)|| 161 | |APA 2020|GNN-PIM: A Processing-in-Memory Architecture for Graph Neural Networks|PKU| [[paper]](http://115.27.240.201/docs/20200915165942122459.pdf)![Scholar citations](https://img.shields.io/badge/Citations-0-_.svg?logo=google-scholar&labelColor=4f4f4f&color=3388ee)|| 162 | |ASAP 2020|Hardware Acceleration of Large Scale GCN Inference|USC| [[paper]](https://ieeexplore.ieee.org/document/9153263)![Scholar citations](https://img.shields.io/badge/Citations-0-_.svg?logo=google-scholar&labelColor=4f4f4f&color=3388ee)|| 163 | |DAC 2020|Hardware Acceleration of Graph Neural Networks|UIUC| [[paper]](http://rakeshk.web.engr.illinois.edu/dac20.pdf)![Scholar citations](https://img.shields.io/badge/Citations-0-_.svg?logo=google-scholar&labelColor=4f4f4f&color=3388ee)|| 164 | |MICRO 2020|AWB-GCN: A Graph Convolutional Network Accelerator with Runtime Workload Rebalancing|PNNL| [[paper]](https://ieeexplore.ieee.org/abstract/document/9252000)![Scholar citations](https://img.shields.io/badge/Citations-0-_.svg?logo=google-scholar&labelColor=4f4f4f&color=3388ee)|| 165 | |arXiv 2020|GRIP: A Graph Neural Network Accelerator Architecture|Stanford| [[paper]](https://arxiv.org/pdf/2007.13828.pdf)![Scholar citations](https://img.shields.io/badge/Citations-0-_.svg?logo=google-scholar&labelColor=4f4f4f&color=3388ee)|| 166 | |HPCA 2020|HyGCN: A GCN Accelerator with Hybrid Architecture|UCSB| [[paper]](https://arxiv.org/pdf/2001.02514.pdf)![Scholar citations](https://img.shields.io/badge/Citations-0-_.svg?logo=google-scholar&labelColor=4f4f4f&color=3388ee)|| 167 | ## Contribute 168 | 169 | We welcome contributions to [this repository](https://github.com/chwan1016/awesome-gnn-systems). To add new papers to this list, please update JSON files under `./res/papers/`. Our bots will update the paper list in `README.md` automatically. The citations of newly added papers will be updated within one day. 170 | --------------------------------------------------------------------------------