Author

Sangeet Dahal

Date of Award

2014

Document Type

Thesis

Degree Name

Master of Science (MS)

Department

Computer Science

Committee Chair

Timothy S. Newman

Committee Member

Heggere S. Ranganath

Committee Member

Ramazan S. Aygun

Subject(s)

Finite element method, Computer graphics, Real-time programming

Abstract

We present new efficient parallel approaches for 2D mesh smoothing on a GPU for Zhou and Shimada's [12] and Xu and Newman's [13] 2D mesh smoothing algorithms. Our parallel approaches have two main processing phases. In the first processing phase, also called the Pre-processing Phase, we detect all the internal vertices of an input mesh. The Pre-processing Phase transforms an input mesh into a representation called a neighbor list, which is processed in parallel by the next processing phase, the GPU Processing Phase. Two parallel approaches for the Zhou and Shimada Smoothing Algorithm were attempted and are described here. One of these uses multiple threads per vertex to smooth by re-positioning internal vertices. The other approach smoothes using a single thread per internal vertex. The multiple threads per internal vertex approach exhibits good performance in experiments we describe here. Speedups of GPU over CPU performance of the multiple thread per internal vertex approach range from 18.06 to 98.98 for the environment available to us. The single thread per internal vertex approach exhibits even better performance in experiments we describe here. Speedups of GPU over CPU performance for that single thread per internal vertex approach range from 28.38 to 154.57 for the environment available to us. The Xu and Newman Mesh Smoothing Algorithm was parallelized using a single thread per internal vertex approach. That approach exhibits a performance improvement over CPU performance in experiments we describe here. Speedups of GPU over CPU performance of that single thread per internal vertex approach range from 8.59 to 38.93 for the environment available to us.

main_programs.zip (164 kB)
SupplementarySerial and CUDA C/C++ Programs

Share

COinS
 
 

To view the content in your browser, please download Adobe Reader or, alternately,
you may Download the file to your hard drive.

NOTE: The latest versions of Adobe Reader do not support viewing PDF files within Firefox on Mac OS and if you are using a modern (Intel) Mac, there is no official plugin for viewing PDF files within the browser window.