author: | Antoine Veenstra |
title: | Edit Distance on GPU clusters using MPI |
keywords: | GPU Clusters, MPI, GPU programming |
topics: | Algorithms and Data Structures |
committee: | Marieke Huisman |
started: | July 2017 |
end: | July 2017 |
Abstract
In this paper, we describe a verified implementation of the Levenshtein distance problem on a GPU cluster using MPI and OpenCL. The implementation is based on an existing verified single GPU implementation. The speed of the implementation is higher on a cluster, but the efficiency is affected by the overhead, which is caused by the extra communication between nodes.