Методологічні особливості реалізації нейроподібних паралельно-ієрархічних систем на основі технологій GPGPU

Abstract

The methodological features of realization of neural-like parallel-hierarchical systems on the basis of technologies GPGPU, in a context of research of a hardware-software platform of GPGPU are considered in the work. The analysis of architecture of modern GPU and model of parallel programming GPU is carry out in researches. Also considered the methods of the GPU programs optimization for parallel computing of general purpose in parallel-hierarchical systems which are offered as network model of the neural-like scheme of information processing.