Performance Evaluation of Query Processing Algorithms on GPGPUs
School of Informatics, University of Edinburgh
University of Edinburgh, 2010
@article{hristov2010performance,
title={Performance Evaluation of Query Processing Algorithms on GPGPUs},
author={Hristov, V.},
year={2010}
}
Modern Graphical Processing Units (GPUs) can perform general purpose computing, next to standard graphical processing. Open frameworks, such as the OpenCL standard by the Khronos Group, enable developers to easily harness the computational power of GPUs. While in certain aspects, these are more powerful than standard CPUs, the latter are still a more suitable solution to many modern problems. In this dissertation we show the performance that can be achieved by evaluating database queries on GPUs using OpenCL. This is achieved by implementing a simple database model, capable of performing queries on any OpenCL-enabled device. The performance evaluation shows the strengths and shortcomings of this approach.
January 30, 2012 by hgpu
