Dualna hitra gradienta metoda na grafičnih procesnih enotah

diplomsko delo

Iztok Ramovš (Author), Uroš Lotrič (Mentor)

Abstract

V napravah tipa tokamak je eden izmed večjih izzivov proces vodenja oblike in toka plazme, ki je pod konstantnim vplivom motenj. Tega izziva se je bilo možno lotiti z uporabo metode prediktivnega vodenja. Reševanje optimizacijskega problema prediktivnega vodenja je bilo narejeno z algoritmom z dualno hitro gradientno metodo. Vendar pa se je izkazalo, da se narejena implementacija izvaja prepočasi za hitro dinamiko v napravah tokamak. Cilj te diplomske naloge je ponovna implementacija te metode z uporabo paralelnih tehnologij z namenom hitrejšega izvajanja. Pri tem moramo paziti, da paralelna implementacija probleme rešuje pravilno in z enako natančnostjo kot njena sekvenčna izvedba. Pohitritev poizkušamo doseči s paralelnim izvajanjem algoritma na splošno namenskih grafičnih procesnih enotah z izbrano tehnologijo OpenCL.

Keywords

Tokamak;MPC;OpenCL;dualna hitra gradientna metoda;paralelna implementacija;GPE;računalništvo in informatika;univerzitetni študij;diplomske naloge;

Data

Language:	Slovenian
Year of publishing:	2018
Typology:	2.11 - Undergraduate Thesis
Organization:	UL FRI - Faculty of Computer and Information Science
Publisher:	[I. Ramovš]
UDC:	004(043.2)
COBISS:	1537944259
Views:	1092
Downloads:	375
Average score:	0 (0 votes)
Metadata:

Other data

Secondary language:	English
Secondary title:	Dual fast gradient method on Graphical Processing Units
Secondary abstract:	One of the biggest challenges in tokamak devices is to maintain plasma shape which is affected by constant disturbances. One of the possible solutions for this challenge is usage of model predictive control methodology. Optimization problem which represents used methodology was solved with dual fast gradient method. However, described implementation was too slow. The goal of this thesis is to re-implement the existing implementation using parallel technologies. New implementation must be able to provide solutions with the same level of accuracy as current serial implementation. We attempt to achieve faster execution with parallel program running on general purpose graphics processing units. Chosen technology is OpenCL framework.
Secondary keywords:	Tokamak;MPC;OpenCL;dual fast gradient method;parallel implementation;GPU;computer and information science;diploma;
File type:	application/pdf
Type (COBISS):	Bachelor thesis/paper
Study programme:	1000468
Embargo end date (OpenAIRE):	1970-01-01
Thesis comment:	Univ. v Ljubljani, Fak. za računalništvo in informatiko
Pages:	53 str.
ID:	10958792