Change search
ReferencesLink to record
Permanent link

Direct link
GPGPU – Att utnyttja GPU istället för CPU.
KTH, School of Computer Science and Communication (CSC).
KTH, School of Computer Science and Communication (CSC).
2012 (Swedish)Independent thesis Advanced level (professional degree), 10 credits / 15 HE creditsStudent thesis
Abstract [en]

Abstract - GPGPU

To utilize the GPU instead of the CPU

Jimmy Larsson Oskar Bodemyr

The following report has been written by Jimmy Larsson and Oskar Bodemyr as a bachelor's thesis in order to reach a bachelor's degree in the subject of Computer Engineering. At the time of writing this report, the most common compute device is the CPU. This report looks at the possible usages of GPGPU and how one might instead use the GPU for General Purpose Computations. The main question that the report will face is the following: Is it possible, using a non-trivial parallel algorithm, to show that a GPU is better suited to perform certain calculations than a CPU? The implementation of two test algorithms in order to study the basic functions of GPGPU, followed by an implementation of a matrix- multiplier provided data supporting the question of the report. The data itself proves that certain algorithms such as the matrix-multiplier of large amounts of data might take about 30 seconds to calculate on a CPU, while the same calculations can be done on a GPU in less than half a second.

Abstract [sv]

Sammanfattning - GPGPU

Att utnyttja GPU istllet fr CPU

Jimmy Larsson Oskar Bodemyr

Fljande rapport har skrivits av Jimmy Larsson och Oskar Bodemyr som ett kandidatexamensarbete. I dagslget r den vanligaste berkningsenheten datorns CPU-enhet. Denna rapport tittar p mjliga anvndningar av GPGPU och hur man istllet kan anvnda GPUn istllet fr CPUn fr allmnna berkningar. Huvudfrgan i denna rapport lyder: Kan man med hjlp av en icke-triviell parallell algoritm visa att GPU r mer lmpad att utfra vissa berkningar n CPU? Tv testalgoritmer implementerades fr att studera de grundlggande funktionerna inom GPGPU. Detta fljdes av en huvudalgoritm i form av matrismultiplikation som tillsammans med de tv testalgoritmerna gav resultat som anvndes fr att besvara huvudfrgan. Datan visar att vissa algoritmer, s som matrismultiplikation, med stora mngder data kan ta upp till ca 30 sekunder p en CPU, samtidigt som det bara tar en halv sekund p en GPU.

Place, publisher, year, edition, pages
Kandidatexjobb CSC, K12013
National Category
Computer Science
URN: urn:nbn:se:kth:diva-131017OAI: diva2:654463
Educational program
Master of Science in Engineering - Computer Science and Technology
Available from: 2013-10-07 Created: 2013-10-07

Open Access in DiVA

No full text

Other links
By organisation
School of Computer Science and Communication (CSC)
Computer Science

Search outside of DiVA

GoogleGoogle Scholar
The number of downloads is the sum of all downloads of full texts. It may include eg previous versions that are now no longer available

Total: 178 hits
ReferencesLink to record
Permanent link

Direct link