AI 中文总结
针对现实网络中边插入或删除时维护有根生成森林的挑战，提出四种全动态并行算法，经 GPU 环境实验，其每秒插入和删除吞吐量远超现有并行静态算法。
AI 中文摘要
生成树是图论中的基本结构，在网络维护、路由调整等各种应用中至关重要。现实世界网络的动态特性要求随着基础图的演变对这些结构进行高效更新。动态维护有根生成树对于处理 2 - 连通分量和最小加权生成树的算法尤为关键。本文应对一批边插入或删除时维护有根生成森林的挑战，提出四种全新的全动态并行算法来更新生成森林而无需从头重建。实验表明在 GPU 环境下每秒有 200 万次插入和 140 万次删除的吞吐量，显著优于现有并行静态算法。
英文摘要
Spanning trees are fundamental structures in graph theory, essential for various applications such as network maintenance, routing adjustments, and many more. The dynamic nature of real-world networks requires efficient updates to these structures as the underlying graph evolves. Maintaining rooted spanning trees dynamically is particularly crucial for algorithms addressing 2-connected components and minimum-weighted spanning trees. In this paper, we address the challenge of maintaining a rooted spanning forest when a batch of edges are inserted or deleted. We present four novel fully dynamic parallel algorithms to update the spanning forest without reconstructing it from scratch. To the best of our knowledge, parallel algorithms for this problem remain largely unexplored. Our experiments on a diverse collection of real-world graphs using a GPU environment demonstrate a throughput of 2 million insertions and 1.4 million deletions per second, significantly outperforming state-of-the-art parallel static algorithms.
CommentsAccepted in IPDPSW(APDCM 2026)
