Analýza a optimalizace GPGPU akcelerovaných aplikací z pohledu spotřeby energie

Abstract

In the last twenty years, the technology of graphics processing units went through rapid development and the architecture has been adapted to support massively parallel computations. This thesis deals with the energy efficiency of the Nvidia accelerator using the architecture of CUDA graphics processing units when running a parallel application and its optimization of energy consumption. The goal of this work is the development of a custom library CUPTIINF, that provides an application, as well as its sub-parts, computational resources consumption measurement, and also tuning of the hardware parameters during the application execution to reach maximum energy efficiency.