Javascript must be enabled for the correct page display

A GPU implementation of the 1D and 2D Conservation Element / Solution Element scheme

Westers, F.M. (2017) A GPU implementation of the 1D and 2D Conservation Element / Solution Element scheme. Bachelor's Thesis, Computing Science.

[img]
Preview
Text
BachelorThesisCS.pdf - Published Version

Download (2MB) | Preview
[img] Text
Toestemming.pdf - Other
Restricted to Backend only

Download (78kB)

Abstract

The demand for faster and more efficient algorithms in fluid dynamics is high. The CE/SE scheme is a scheme for simulating flows, which is designed in such a way that many spatial elements can be calculated in parallel. This characteristic makes it very suitable for implementation using parallelism on GPUs. The aim of this thesis was to implement the 1D and 2D CE/SE Euler scheme on a GPU using CUDA. Next, the speed-up compared to the existing CPU algorithms was measured to determine whether it is worth converting more complicated versions of the scheme to GPU algorithms. In this study is was found that the algorithm is considerably faster on GPU than on CPU, respectively 23 times in double precision 1D and 1.5 times faster in 2D. Using single precision, GPU is 4.2 times faster in 2D. Still, by using multiple GPUs, the performance is most likely to increase even more. Therefore, the work presented in this thesis a starting point for further implementations of the CE/SE scheme on GPUs

Item Type: Thesis (Bachelor's Thesis)
Degree programme: Computing Science
Thesis type: Bachelor's Thesis
Language: English
Date Deposited: 15 Feb 2018 08:27
Last Modified: 15 Feb 2018 08:27
URI: http://fse.studenttheses.ub.rug.nl/id/eprint/15035

Actions (login required)

View Item View Item