Hash match on GPU

Author

Keh Kok Yong ; Karuppiah, E.K.

Author_Institution

Accelerative Technol. Lab., MIMOS Berhad, Kuala Lumpur, Malaysia

fYear

2013

fDate

2-4 Dec. 2013

Firstpage

150

Lastpage

155

Abstract

Information is one of the most influential forces transforming the growth of businesses, and its amount is ever growing exponentially. There is a significant challenge to have an efficient matching tool to search for a required piece of information. String matching poses a computationally intensive challenge for massive data. In this paper, we present a comparison of an exact string matching mechanism using a Graphic Processing Unit (GPU). We progressively design the mechanism and data structure to fit on this parallel processing architecture. We then evaluate our proposed Hash Match implementation by comparing two other different mechanisms, Column Search (Brute Force) and Boyer-Moore-Horspool in two different NVIDIA cards, based on the “Fermi” architecture on C2075 and “Kepler” architecture on K20c.

Keywords

data structures; graphics processing units; information retrieval; parallel architectures; string matching; Boyer-Moore-Horspool; Brute Force; Fermi architecture; GPU; Hash matching tool; K20c; Kepler architecture; NVIDIA cards; business growth; column search; data structure; graphic processing unit; massive data; parallel processing architecture; string matching mechanism; Acceleration; Computer architecture; Graphics processing units; Instruction sets; Random access memory; Big Data; CUDA; FERMI; GPU; Hash; KEPLER; Matching; NVIDIA; String;

fLanguage

English

Publisher

ieee

Conference_Titel

Open Systems (ICOS), 2013 IEEE Conference on

Conference_Location

Kuching

Print_ISBN

978-1-4799-3152-1

Type

conf

DOI

10.1109/ICOS.2013.6735065

Filename

6735065