dc.contributor.advisor | Natvig, Lasse | nb_NO |
dc.contributor.author | Bakke, Glenn Ruben Årthun | nb_NO |
dc.date.accessioned | 2014-12-19T13:31:23Z | |
dc.date.available | 2014-12-19T13:31:23Z | |
dc.date.created | 2010-09-02 | nb_NO |
dc.date.issued | 2009 | nb_NO |
dc.identifier | 347109 | nb_NO |
dc.identifier | ntnudaim:4301 | nb_NO |
dc.identifier.uri | http://hdl.handle.net/11250/250325 | |
dc.description.abstract | In this thesis we do a broad study of the languages, libraries and frameworks for general purpose computations on graphics processors. We have also studied the different graphics processor architectures that has been developed the last decade. 8 example programs in OpenGL, CUDA, MPI and OpenMP has been made to emphasize the mechanisms for parallelization and memory managment. The example programs have been benchmarked and the source lines are been counted. We found out that programs for the GPU outperforms the CPU multi-core programs in some cases. This depends on the algorithm and problem size. Also we found out that the new languages created for GPGPU are more suited for programming the GPU than with graphics libraries. | nb_NO |
dc.language | eng | nb_NO |
dc.publisher | Institutt for datateknikk og informasjonsvitenskap | nb_NO |
dc.subject | ntnudaim | no_NO |
dc.subject | MIT informatikk | no_NO |
dc.subject | Komplekse datasystemer | no_NO |
dc.title | Programming Graphic Processing Units (GPUs) | nb_NO |
dc.type | Master thesis | nb_NO |
dc.source.pagenumber | 165 | nb_NO |
dc.contributor.department | Norges teknisk-naturvitenskapelige universitet, Fakultet for informasjonsteknologi, matematikk og elektroteknikk, Institutt for datateknikk og informasjonsvitenskap | nb_NO |