Fast Triangle Counting

TL;DR

This paper introduces a new serial algorithm for triangle counting in graphs, achieving competitive performance on real and synthetic datasets using recent high-performance processors, advancing network analysis techniques.

Contribution

The paper presents a novel serial triangle counting algorithm that outperforms previous methods in speed and efficiency on benchmark datasets.

Findings

Competitive performance on Graph500 and Graph Challenge datasets

Effective utilization of Intel Xeon Platinum 8480+ and CPU Max 9480 processors

Outperforms existing triangle counting algorithms in speed

Abstract

Listing and counting triangles in graphs is a key algorithmic kernel for network analyses including community detection, clustering coefficients, k-trusses, and triangle centrality. We design and implement a new serial algorithm for triangle counting that performs competitively with the fastest previous approaches on both real and synthetic graphs, such as those from the Graph500 Benchmark and the MIT/Amazon/IEEE Graph Challenge. The experimental results use the recently-launched Intel Xeon Platinum 8480+ and CPU Max 9480 processors.