Scalable Parallel Minimum Spanning Forest Computation
ACM SIGPLAN Notices(2012)
摘要
The proliferation of data in graph form calls for the development of scalable graph algorithms that exploit parallel processing environments. One such problem is the computation of a graph's minimum spanning forest (MSF). Past research has proposed several parallel algorithms for this problem, yet none of them scales to large, high-density graphs. In this paper we propose a novel, scalable, parallel MSF algorithm for undirected weighted graphs. Our algorithm leverages Prim's algorithm in a parallel fashion, concurrently expanding several subsets of the computed MSF. Our effort focuses on minimizing the communication among different processors without constraining the local growth of a processor's computed subtree. In effect, we achieve a scalability that previous approaches lacked. We implement our algorithm in CUDA, running on a GPU and study its performance using real and synthetic, sparse as well as dense, structured and unstructured graph data. Our experimental study demonstrates that our algorithm outperforms the previous state-of-the-art GPU-based MSF algorithm, while being several order of magnitude faster than sequential CPU-based algorithms.
更多查看译文
关键词
Algorithms,Experimentation,Performance,Parallel Graph Algorithms,Minimum Spanning Forest,GPU
AI 理解论文
溯源树
样例
生成溯源树,研究论文发展脉络
Chat Paper
正在生成论文摘要