This
presentation gives an overview of programming NVIDIA GPUs using the PGI
Accelerator programming model in C and Fortran, and using PGI CUDA
Fortran. It introduces the compute-specific details of NVIDIA GPUs,
gives a basic example of programming in CUDA C, and provides a detailed
overview of the explicit CUDA Fortran programming model implemented in
the PGI 10.0 compilers. This provides the basis for an overview of PGI
Accelerator directive-based programming for x64+NVIDIA systems,
including motivations, goals, current status and capabilities, and
upcoming features. The PGI Accelerator programming model is a
high-level implicit model for offloading compute-intensive code regions
from a host CPU to an Accelerator via Fortran directives or C pragmas,
simplifying the coding process and ensuring that the resulting programs
are portable to other platforms and compilers.
Doug Miles has worked in the HPC
industry since 1985 at Floating Point Systems, Cray, The Portland Group
and STMicroelectronics, primarily in applications engineering and
management positions.
He has been the director of technical and business operations at PGI since 2003.
La présentation donnée lors de ce séminaire est disponible au
format pdf.
L'annonce de ce séminaire est disponible sous
forme d'un
fichier
pdf.