Acercamiento a la programación paralela usando CUDA
Archivos
Fecha
2012-06-28
Autores
Triana Brito, Diana Isabel
Título de la revista
ISSN de la revista
Título del volumen
Editor
Universidad Central "Marta Abreu" de la Villas
Resumen
Debido al uso cada vez más extendido de aceleradores de hardware en entornos de computación de altas prestaciones ha surgido una nueva tecnología de la programación en paralelo: CUDA (Compute Unified Device Architecture) un nuevo hardware y arquitectura de software para la emisión y gestión de cálculos en la GPU (Graphics Processor Unit) como un dispositivo de computación de datos en paralelo. Las ventajas de CUDA frente al empleo de otras arquitecturas clásicas son numerosas, el bajo coste de los dispositivos para ejecutar estos programas, el relativo poco esfuerzo para adaptar los algoritmos a esta filosofía de trabajo y la alta rentabilidad en tiempo de ejecución que se consigue en numerosos casos. En este trabajo se presentan las características de la tecnología, aplicaciones que usan esta arquitectura, se establece un procedimiento de trabajo y se prueban ejemplos en la solución de problemas haciendo uso de la tecnología.
Due to the extended use of hardware accelerators in computer environments of high benefits, a new technology of the programming in parallel has arisen: CUDA (Computes Unified Device Architecture), a new hardware and software architecture for the emission and administration of calculations in the GPU (Graphics Processor Unit) as a computer device of data in parallel. The advantages of CUDA are numerous in front of the employment of other classic architectures. Between them we can find, the first floor cost of the devices to execute these programs, the relative little effort to adapt the algorithms to this working philosophy and the high profitability in run-time that is gotten in numerous cases. In this work the characteristics of the technology applications that use this architecture are presented. Therefore a work procedure and the troubleshooting making use of the technology are establish.
Due to the extended use of hardware accelerators in computer environments of high benefits, a new technology of the programming in parallel has arisen: CUDA (Computes Unified Device Architecture), a new hardware and software architecture for the emission and administration of calculations in the GPU (Graphics Processor Unit) as a computer device of data in parallel. The advantages of CUDA are numerous in front of the employment of other classic architectures. Between them we can find, the first floor cost of the devices to execute these programs, the relative little effort to adapt the algorithms to this working philosophy and the high profitability in run-time that is gotten in numerous cases. In this work the characteristics of the technology applications that use this architecture are presented. Therefore a work procedure and the troubleshooting making use of the technology are establish.
Descripción
Palabras clave
Compute Unified Device Architecture, Hardware, Arquitectura de Software, Graphics Processor Unit