CPU vs. GPU for Community Detection: Performance Insights from GVE-Louvain and $\nu$-Louvain

TL;DR

This paper compares CPU and GPU implementations of community detection algorithms, demonstrating that CPU-based GVE-Louvain significantly outperforms GPU-based $ u$-Louvain and other methods in speed and scalability, highlighting the advantages of CPUs for this task.

Contribution

The paper introduces GVE-Louvain, a highly efficient multicore CPU implementation, and evaluates its performance against GPU-based $ u$-Louvain, providing new insights into hardware suitability for community detection.

Findings

GVE-Louvain outperforms GPU implementations by up to 50x in speed.

GVE-Louvain scales efficiently with thread doubling, achieving 1.6x speedup per doubling.

CPU-based methods may be more suitable than GPU-based ones for community detection tasks.

Abstract

Community detection involves identifying natural divisions in networks, a crucial task for many large-scale applications. This report presents GVE-Louvain, one of the most efficient multicore implementations of the Louvain algorithm, a high-quality method for community detection. Running on a dual 16-core Intel Xeon Gold 6226R server, GVE-Louvain outperforms Vite, Grappolo, NetworKit Louvain, and cuGraph Louvain (on an NVIDIA A100 GPU) by factors of 50x, 22x, 20x, and 5.8x, respectively, achieving a processing rate of 560M edges per second on a 3.8B-edge graph. Additionally, it scales efficiently, improving performance by 1.6x for every thread doubling. The paper also presents $\nu$-Louvain, a GPU-based implementation. When evaluated on an NVIDIA A100 GPU, $\nu$-Louvain performs only on par with GVE-Louvain, largely due to reduced workload and parallelism in later algorithmic passes. These results suggest that CPUs, with their flexibility in handling irregular workloads, may be better suited for community detection tasks.