is it possible to use both “pgi accelerator directives” & “CUDA fortran programming” in the same project
Yes. PGI Accelerator compute regions are able to use CUDA Fortran device variables.
You may also wish to look at using Kernel directives instead of the PGI Accelerator model. Essentially Kernel directives are a light-weight version of the PGI Accelerator model, where the user manages data movement and the CUF kernel directives manage the kernels. It also give the user a more CUDA like interface.
See the following article for an introduction to CUF kernel directives: Account Login | PGI
- Mat