Utilize este link para identificar ou citar este item:
https://bdm.unb.br/handle/10483/15791
Título: | A study of replacing CUDA by OpenCL in KGPU |
Autor(es): | Melo, Rodrigo Siqueira de |
Orientador(es): | Costa Júnior, Edson Alves da |
Assunto: | Computação gráfica Unidades de Processamento Gráfico (GPUs) |
Assunto: | Linguagem de programação (Computadores) C (linguagem de programação) OpenCL (Linguagem de programação de computador) Arquitetura de computador |
Data de apresentação: | 10-Jul-2015 |
Data de publicação: | 18-Jan-2017 |
Referência: | MELO, Rodrigo Siqueira de. A study of replacing CUDA by OpenCL in KGPU. 2015. 68 f., il. Monografia (Bacharelado em Engenharia de Software)—Universidade de Brasília, Brasília, 2015. |
Abstract: | GPU is a very high parallel device which became popular. Nowadays, many processors
already coming with a minimal GPU in the same die, this characteristic creates a new
and unexplored application area for this device. CUDA and OpenCL are two non-graphics
libraries commonly used for take advantages of GPU. CUDA was created by NVidia, and
it was designed to run on NVidia’s GPUs. On the other hand, OpenCL was created to
run on many different devices. Those libraries, interacts with the operating system by
using device drivers, and usually this is the unique connection between them. A group of
researchers from Utah proposed the use of GPU as a coprocessor, they developed a device
driver based on CUDA for achieving this goal (they called it as KGPU). In this work we
improved KGPU’s code, added support to OpenCL, and we analyzed the possibility of
use this project as a mature solution. |
Informações adicionais: | Monografia (graduação)—Universidade de Brasília, Faculdade UnB Gama, 2015. |
Aparece na Coleção: | Engenharia de Software
|
Este item está licenciado na Licença Creative Commons