PGI User Forum

kehkok · Joined: 21 Apr 2012 Posts: 3

I successfully installed PGI Accelerator Compilers in my laptop. How do I start a HelloWorld project by using OpenAcc on Cuda?

I have Cuda 4.0, 4.1, 4.2 and 5 in my laptop too.

mkcolg · Posted: Fri Dec 07, 2012 10:14 am Post subject:

Hi kehkok,

I'm assuming you mean that you have Portland Visual Fortran (PVF) installed and during installation you selected 'yes' when asked if you wish to install the CUDA components? If so, I'll refer you to the PVF users guide (http://www.pgroup.com/doc/pvfug.pdf) on the basics of using PVF. There's also an Sample Project, AccelPM_Matmul, in the "PGI Visual Fortran -> Samples -> gpu" directory, which should help in getting you started. Matmul is essentially the 'Hello World' program for GPUs.

kehkok · Joined: 21 Apr 2012 Posts: 3

Hi Mat,

I would like to use VC++ and I am not Fortran developer. I am using Visual Studio 2010 for all my Cuda development. I can compile Cuda 4.0 to 5 VC++ solution. I read the documentation, there is step by step for Visual Studio in Fortran for PGI, but there isn't any steps for MS Visual Studio VC++. Can you point me to right direction?

mkcolg · Posted: Mon Dec 10, 2012 9:04 am Post subject:

Hi kehkok,

It sounds like you are looking for OpenACC support in C? Microsoft VC++ doesn't support OpenACC so you'll need to use PGI's C compiler, pgcc. While we haven't productised it, we do have a beta C plugin for VS. Though, I'm not positive on its current status. Please send a note to PGI Customer Support (trs@pgroup.com) for details.

If you're looking form CUDA C support from within MS VS, please contact NVIDIA and/or Microsoft.

Thanks,
Mat