Parallel computing on geostatistical data using CUDA

Shan, Feng

Permalink

https://hdl.handle.net/2142/46862 Copy

Description

Title

Parallel computing on geostatistical data using CUDA

Author(s)

Shan, Feng

Issue Date

2014-01-16T18:19:00Z

Director of Research (if dissertation) or Advisor (if thesis)

Hart, John C.

Department of Study

Computer Science

Discipline

Computer Science

Degree Granting Institution

University of Illinois at Urbana-Champaign

Degree Name

M.S.

Degree Level

Thesis

Keyword(s)

• Compute Unified Device Architecture (CUDA)
• parallel computing
• average local variance
• angular second moment
• entropy
• inverse difference moment

Abstract

Data analysis is receiving considerable attention with the design of new graphics processing units (GPUs). Our study focuses on geostatistical data analysis, which is currently applied in diverse disciplines such as meteorology, oceanography, geography, forestry, environmental control, and agriculture. While geostatistical analysis algorithms are applied in varied branches, those analyses can be accelerated by applying parallel computing using modern GPUs. The highly parallel structure makes modern GPUs more effective than general-purpose CPUs for algorithms where processing of large blocks of data is done in parallel. In our study, we compared the performance between serial and parallel computation on four texture features, including average local variance (ALV), angular second moment (ASM), entropy, and inverse difference moment (IDM). The later three features (ASM, Entropy and IDM) are features obtained using Gray Level Coocurrence Matrices (GLCM). We parallelized the computation by using multiple sliding windows on two-dimensional data concurrently. Our approach also includes, in addition to comparing to serial implementation, measuring the parallelized performance under different data sizes. As a result, parallel computation on geostatistical analyses using GPU can significantly increase the performance and efficiency. It has also demonstrated the possibility to provide solutions for specific needs by reducing the time of computation.

Graduation Semester

2013-12

Permalink

http://hdl.handle.net/2142/46862

Copyright and License Information

Copyright 2013 Feng Shan

Owning Collections