LYGIAGRETŪS SKAIČIAVIMAI SU CUDA
PARALLEL COMPUTING WITH CUDA
Author(s): Julija Semenenko, Dmitrij ŠešokSubject(s): Information Architecture, ICT Information and Communications Technologies
Published by: Vilniaus Universiteto Leidykla
Keywords: CUDA; graphics processing unit; CPU; parallel computing; numerical experiment;
Summary/Abstract: The principles of computing with NVIDIA’s parallel computing platform CUDA are analysed in the paper. Two numerical experiments, array addition and matrix multiplication and optimizing matrix multiplication (shared memory, bank conflict solutions, instruction-level parallelism) with Geforce and Quadro graphics cards and the CPU were run. Time values for computing int, float, double data types are provided in the paper.
Journal: Jaunųjų mokslininkų darbai
- Issue Year: 2017
- Issue No: 1 (47)
- Page Range: 87-93
- Page Count: 7
- Language: Lithuanian