Matrix algorithms using CUDA technology

U ovom radu su korištenjem CUDA tehnologije implementirani algoritmi za zbrajanje, množenje i inverz matrice. U početku je objašnjenja CUDA arhitektura i navedeni su neki primjeri primjene CUDA tehnologije u industriji. Zatim su algoritmi objašnjeni teorijski i prikazan je način njihove programske implementacije. Na svim implementiranim algoritmima je provedeno testiranje performansi i točnosti koristeći tri modela grafičkih kartica. Zbog bolje mogućnosti usporedbe korištena je i serijska verzija svih algoritama koja se izvodi na CPU-u. Izmjereni rezultati su grafički prikazani kako bi se dobio bolji uvid u vremensku složenost implementiranih algoritama.In this paper, algorithms for addition, multiplication and matrix inverse are implemented using CUDA technology. In the beginning, the CUDA architecture is explained and some examples of the application of CUDA technology in industry are given. Then the algorithms are explained theoretically and the method of their programmatic implementation is shown. All implemented algorithms were tested for performance and accuracy using three models of graphics cards. Due to the better possibility of comparison, the serial version of all algorithms, which is executed on the CPU, was also used. The measured results are graphically presented in order to get a better insight into the time complexity of the implemented algorithms