Proceedings of the Twenty-Fourth ACM Symposium on Operating Systems Principles
ACM SIGOPS 24th Symposium on Operating Systems Principles
Dandelion: a compiler and runtime for heterogeneous systems
Proceedings of the Twenty-Fourth ACM Symposium on Operating Systems Principles
| Hi-index | 0.00 |
We investigate the use of graphics processing units (GPUs) in accelerating Page Rank computation. We first introduce a compact web graph representation which requires much less memory allocation than a well-known compressed sparse row format. The web graph is then simply partition into smaller chunks to fit the GPUs' device memory. We propose a fast Page Rank algorithm to run on the GPU cluster. The design of algorithm is general and does not constrain on any large web graph fitting to the limited size of device memory. In the experiments, we test our Page Rank algorithm on a small GPU cluster, using a set of real web data. We compare the parallel Page Rank computation utilizing GPUs with CPUs. The results show that the proposed Page Rank computation on GPUs gives promising result.