Evaluation of Image Pixels Similarity Measurement Algorithm Accelerated on GPU with OpenACC
OpenACC is a directive based parallel programming library that allows for easy acceleration of existing C, C++ and Fortran based applications with minimal code modifications. By annotating the bottleneck causing section of the code with OpenACC directives, the acceleration of the code can be simplif...
Saved in:
| Main Authors: | , |
|---|---|
| Format: | Article |
| Language: | en |
| Published: |
Penerbit Universiti, UTeM
2017
|
| Subjects: | |
| Online Access: | http://eprints.utem.edu.my/id/eprint/20109/1/ICTEC2017.pdf http://eprints.utem.edu.my/id/eprint/20109/ http://journal.utem.edu.my/index.php/jtec |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | OpenACC is a directive based parallel programming library that allows for easy acceleration of existing C, C++ and Fortran based applications with minimal code modifications. By annotating the bottleneck causing section of the code with OpenACC directives, the acceleration of the code can be simplified, leading for high portability of performance across different target Graphic Processing Units (GPUs). In this work, the portability of an implemented parallelizable chi-square based pixel similarity measurement algorithm has been evaluated on two consumer and professional grade GPUs. To our best knowledge, this is the first performance evaluation report that utilizes the OpenACC optimization clauses (collapse and tile) on different GPUs to process a less workload (low resolution image of 581x429 pixels) and a heavy workload (high resolution image of 4500 x 3500 pixels) to demonstrate the effectiveness and high portability of OpenACC. |
|---|