calling CUDA from C++ in MSVC

Accelerated Computing CUDA CUDA Programming and Performance
1

Good Afternoon,

I’m trying to slowly port a homebrew c++ code of mine to CUDA and I’d like to move one piece at a time. This means I’d have to call CUDA from C++.

I know it’s possible, I’ve seen the example in the SDK, and I know how to do extern “C”, etc. However how do I set up a MSVC solution to accomodate this?

I’ve been using the CUDA VS Wizard to create new project and it works great. I’m not sure what I need to do to tweak the build order, etc.

If I try right now to compile a mixed CUDA/C++ project, nvcc tells me it’s missing main, and it never gets to compiling the C++ code.

Thank you,

Philip

2

There are many ways to do this. One of the simpler ones is to just compile .cu files into object files and let VS do all the linking. Add a file to the project, go to its properties, and configure the custom build step with a command-line “nvcc -c …” (no quotes, … is your file and whatever other options you need).

Paulius

3

Thanks. I wound up getting it working.

Appreciate the advice.

philip