A novel approach toward parallel implementation of BFS algorithm using graphic processor unit

Abstract

Graph related algorithms are significant to many of the research areas and disciplines. A very big graph with millions of vertices is common in scientific research works and in the implementations of engineering tasks. Many researcher have tried to implement graph algorithms in parallel architectures, where in this paper, authors have tried to accelerate this implementation in an efficient way. In this paper, a GPU implementation of breadth-first search (BFS) is introduced to accelerate graph algorithm implementation. First, a BFS algorithm is implemented in a sequential environment and then on GPU. Experimental results show that the GPU-based approach of BFS outperforms the same as sequential.