Saved in:
Bibliographic Details
Main Authors: Wu, Qizhe, Zhao, Letian, Gui, Yuchen, Wang, Huawen Liang Xiaotian
Format: Preprint
Published: 2024
Subjects:
Online Access:https://arxiv.org/abs/2411.03857
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866912107292786688
author Wu, Qizhe
Zhao, Letian
Gui, Yuchen
Wang, Huawen Liang Xiaotian
author_facet Wu, Qizhe
Zhao, Letian
Gui, Yuchen
Wang, Huawen Liang Xiaotian
contents Graph Convolutional Networks (GCNs) are state-of-the-art deep learning models for representation learning on graphs. However, the efficient training of GCNs is hampered by constraints in memory capacity and bandwidth, compounded by the irregular data flow that results in communication bottlenecks. To address these challenges, we propose a message-passing architecture that leverages NUMA-based memory access properties and employs a parallel multicast routing algorithm based on a 4-D hypercube network within the accelerator for efficient message passing in graphs. Additionally, we have re-engineered the backpropagation algorithm specific to GCNs within our proposed accelerator. This redesign strategically mitigates the memory demands prevalent during the training phase and diminishes the computational overhead associated with the transposition of extensive matrices. Compared to the state-of-the-art HP-GNN architecture we achieved a performance improvement of $1.03\times \sim 1.81\times$.
format Preprint
id arxiv_https___arxiv_org_abs_2411_03857
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Efficient Message Passing Architecture for GCN Training on HBM-based FPGAs with Orthogonal Topology On-Chip Networks
Wu, Qizhe
Zhao, Letian
Gui, Yuchen
Wang, Huawen Liang Xiaotian
Hardware Architecture
Machine Learning
Graph Convolutional Networks (GCNs) are state-of-the-art deep learning models for representation learning on graphs. However, the efficient training of GCNs is hampered by constraints in memory capacity and bandwidth, compounded by the irregular data flow that results in communication bottlenecks. To address these challenges, we propose a message-passing architecture that leverages NUMA-based memory access properties and employs a parallel multicast routing algorithm based on a 4-D hypercube network within the accelerator for efficient message passing in graphs. Additionally, we have re-engineered the backpropagation algorithm specific to GCNs within our proposed accelerator. This redesign strategically mitigates the memory demands prevalent during the training phase and diminishes the computational overhead associated with the transposition of extensive matrices. Compared to the state-of-the-art HP-GNN architecture we achieved a performance improvement of $1.03\times \sim 1.81\times$.
title Efficient Message Passing Architecture for GCN Training on HBM-based FPGAs with Orthogonal Topology On-Chip Networks
topic Hardware Architecture
Machine Learning
url https://arxiv.org/abs/2411.03857