I don’t think it’s a simple library/include path add. I have this big project and I need to add cuda code into it. Please help.
Make a new CUDA project with the templates, add your source code. You might have to change a few include headers around as well as adding the CUDA includes to the respective files… that’s how I did it with 2 separate projects with about 6 different C files. Basically anything with a .cu extension will be processed by nvcc.
Isn’t there a more elegant way?
Perhaps something like this, minus starting from a blank project:
How does that translate to VS2010 and later?
Figured that was general enough, here’s one specific for VS2010 (Tom’s answer, again minus creating a new project):
Yeah, it’s actually pretty obvious and easy. :) Simply go into Build and Build Customizations and check CUDA 5.0 or whatever version you want to use and is installed.