Using graphical processing units to solve the multidimensional GinzburgLandau equation

Branislav Aleksić; Najdan Aleksić; Vladimir Skarka; Milivoj Belić

doi:10.1088/0031-8949/2012/T149/014036

Using graphical processing units to solve the multidimensional GinzburgLandau equation

Branislav Aleksić^*, Najdan Aleksić, Vladimir Skarka, Milivoj Belić

^*Corresponding author for this work

Research output: Contribution to journal › Article › peer-review

8 Citations (Scopus)

Abstract

We demonstrate the use of ultrafast hardware, based on a graphical processing unit (GPU), to solve the complex GinzburgLandau equation. We implement an improved finite-difference time-domain method. We utilize parallel processing of our numerical procedure, resulting in tremendous acceleration in the execution of routines and substantial reduction of cost. Simulations are performed in two and three spatial dimensions with time as the marching variable. The numerical algorithm is also implemented on a CPU to make a comparison with the GPU. An acceleration of about 95 times is achieved. The benefits are discussed in detail and the results are presented visually to achieve the best solution strategy for the given problem.

Original language	English
Article number	014036
Journal	Physica Scripta
Issue number	T149
DOIs	https://doi.org/10.1088/0031-8949/2012/T149/014036
Publication status	Published - Apr 2012
Externally published	Yes

Access to Document

10.1088/0031-8949/2012/T149/014036

Cite this

@article{f91e541e910240f6a63d46e1eca62712,

title = "Using graphical processing units to solve the multidimensional GinzburgLandau equation",

abstract = "We demonstrate the use of ultrafast hardware, based on a graphical processing unit (GPU), to solve the complex GinzburgLandau equation. We implement an improved finite-difference time-domain method. We utilize parallel processing of our numerical procedure, resulting in tremendous acceleration in the execution of routines and substantial reduction of cost. Simulations are performed in two and three spatial dimensions with time as the marching variable. The numerical algorithm is also implemented on a CPU to make a comparison with the GPU. An acceleration of about 95 times is achieved. The benefits are discussed in detail and the results are presented visually to achieve the best solution strategy for the given problem.",

author = "Branislav Aleksi{\'c} and Najdan Aleksi{\'c} and Vladimir Skarka and Milivoj Beli{\'c}",

year = "2012",

month = apr,

doi = "10.1088/0031-8949/2012/T149/014036",

language = "English",

journal = "Physica Scripta",

issn = "0031-8949",

publisher = "Institute of Physics",

number = "T149",

}

TY - JOUR

T1 - Using graphical processing units to solve the multidimensional GinzburgLandau equation

AU - Aleksić, Branislav

AU - Aleksić, Najdan

AU - Skarka, Vladimir

AU - Belić, Milivoj

PY - 2012/4

Y1 - 2012/4

N2 - We demonstrate the use of ultrafast hardware, based on a graphical processing unit (GPU), to solve the complex GinzburgLandau equation. We implement an improved finite-difference time-domain method. We utilize parallel processing of our numerical procedure, resulting in tremendous acceleration in the execution of routines and substantial reduction of cost. Simulations are performed in two and three spatial dimensions with time as the marching variable. The numerical algorithm is also implemented on a CPU to make a comparison with the GPU. An acceleration of about 95 times is achieved. The benefits are discussed in detail and the results are presented visually to achieve the best solution strategy for the given problem.

AB - We demonstrate the use of ultrafast hardware, based on a graphical processing unit (GPU), to solve the complex GinzburgLandau equation. We implement an improved finite-difference time-domain method. We utilize parallel processing of our numerical procedure, resulting in tremendous acceleration in the execution of routines and substantial reduction of cost. Simulations are performed in two and three spatial dimensions with time as the marching variable. The numerical algorithm is also implemented on a CPU to make a comparison with the GPU. An acceleration of about 95 times is achieved. The benefits are discussed in detail and the results are presented visually to achieve the best solution strategy for the given problem.

UR - http://www.scopus.com/inward/record.url?scp=84860451624&partnerID=8YFLogxK

U2 - 10.1088/0031-8949/2012/T149/014036

DO - 10.1088/0031-8949/2012/T149/014036

M3 - Article

AN - SCOPUS:84860451624

SN - 0031-8949

JO - Physica Scripta

JF - Physica Scripta

IS - T149

M1 - 014036

ER -

Using graphical processing units to solve the multidimensional GinzburgLandau equation

Abstract

Access to Document

Other files and links

Fingerprint

Cite this