PGI Accelerator Compiler
PGI Accelerator Compiler supports you to code applications for x64-Bit CUDA enabled NVIDIA Grafikkarten.
Actually three different approaches are available:
Whereas the PGI Accelerator Compiler delivers a programming model for High-Level Programmierung von NVIDIA GPUs by OpenMP like pragma statements, the PGI CUDA Fortran Compiler is an equivalent to the NVIDIA CUDA C Compiler and supports Low-Level coding.
OpenACC is a open parallel programming standard defined by PGI, Cray, NVIDIA and CAPS designed to process code on CUDA-enabled GPUs.
The CUDA C x86 Compiler target the parallel execution of CUDA C code on non-GPU machines.
All Workstation, Server or CDK Compilers and PGI Visual Fortran are available as Accelerator Version.
More information concerning the PGI Accelerator Compiles you'll find here