Real-time track reconstruction in high energy physics imposes stringent latency constraints, hindering the deployment of graph neural networks (GNNs) on general-purpose platforms. We present TrackGNN11https//github.com/silvenachen/TrackGNN, an open-sourced GNN accelerator for track reconstruction. Using a dataflow architecture with multiple parallelism and a self-adaptive renaming mechanism, TrackGNN shows 27.6× speedup over CPUs, up to 101.1× over GPUs, and 5.7× over an FPGA overlay. Compared with FlowGNN, the renaming mechanism also reduces end-to-end latency by 1.12-1.16× with negligible resource overhead.
Li, S, Zhang, H, Chen, R, da Silva, B, Borca-Tasciuc, G, Yu, D & Hao, C 2025, TrackGNN: A Highly Parallelized and Self-Adaptive GNN Accelerator for Track Reconstruction on FPGAs. in Proceedings - 2025 IEEE 33rd Annual International Symposium on Field-Programmable Custom Computing Machines, FCCM 2025. Proceedings - 2025 IEEE 33rd Annual International Symposium on Field-Programmable Custom Computing Machines, FCCM 2025, IEEE. https://doi.org/10.1109/FCCM62733.2025.00029
Li, S., Zhang, H., Chen, R., da Silva, B., Borca-Tasciuc, G., Yu, D., & Hao, C. (2025). TrackGNN: A Highly Parallelized and Self-Adaptive GNN Accelerator for Track Reconstruction on FPGAs. In Proceedings - 2025 IEEE 33rd Annual International Symposium on Field-Programmable Custom Computing Machines, FCCM 2025 (Proceedings - 2025 IEEE 33rd Annual International Symposium on Field-Programmable Custom Computing Machines, FCCM 2025). IEEE. https://doi.org/10.1109/FCCM62733.2025.00029
@inproceedings{28d6fa66aeca455ebaf8ad154adbe590,
title = "TrackGNN: A Highly Parallelized and Self-Adaptive GNN Accelerator for Track Reconstruction on FPGAs",
abstract = "Real-time track reconstruction in high energy physics imposes stringent latency constraints, hindering the deployment of graph neural networks (GNNs) on general-purpose platforms. We present TrackGNN11https//github.com/silvenachen/TrackGNN, an open-sourced GNN accelerator for track reconstruction. Using a dataflow architecture with multiple parallelism and a self-adaptive renaming mechanism, TrackGNN shows 27.6× speedup over CPUs, up to 101.1× over GPUs, and 5.7× over an FPGA overlay. Compared with FlowGNN, the renaming mechanism also reduces end-to-end latency by 1.12-1.16× with negligible resource overhead.",
author = "Shuyang Li and Hanqing Zhang and Ruiqi Chen and {da Silva}, Bruno and Giorgian Borca-Tasciuc and Dantong Yu and Hao, {Cong (Callie)}",
note = "Publisher Copyright: {\textcopyright} 2025 IEEE.",
year = "2025",
month = may,
day = "28",
doi = "10.1109/FCCM62733.2025.00029",
language = "English",
series = "Proceedings - 2025 IEEE 33rd Annual International Symposium on Field-Programmable Custom Computing Machines, FCCM 2025",
publisher = "IEEE",
booktitle = "Proceedings - 2025 IEEE 33rd Annual International Symposium on Field-Programmable Custom Computing Machines, FCCM 2025",
}