new project

Hello, i want to ask questions about creating my own project.
For example, let’s consider the “template” project.
I would like to know what is the difference between “template.vc.proj” and “template.sol”. In fact what are their functions ? How are they linked ?
Besides i can’t change the name of the file “template.vc.proj”, as if i change its name, “template.sol” can’t be open.

Thanks for your help.

More precisely, what should i do in order to create a new project without copying the project “template”. I mean i want to start a new project from nothing (exept the sdk and the toolkit of course). Someone can help me ?

Maybe its a good idea to look at the variables that are set in the template. Like library path linker variables etc… I don’t know that for sure because it was a very long time ago that I did something with VS. But just copy all those paths then you know it will work. and you can start doing your thing.

julandcuda if you mean MS Visual Studio these topics can help you:
http://forums.nvidia.com/index.php?showtopic=51621
http://forums.nvidia.com/index.php?showtopic=44535
http://forums.nvidia.com/index.php?showtopic=47636
http://forums.nvidia.com/index.php?showtopic=48024

Thank you for your replies.

There are instructions in the SDK release notes for windows for creating a new project using template as a starting point.

Mark

How do you do to create a new project without using a previous project ?
Is it easy to create this project by starting from nothing ?

Just as easy as all the other projects you start from scratch in VS. Trial and error.

Or you can use the wonderful tool CMake (www.cmake.org) and add FindCUDA.cmake (http://forums.nvidia.com/index.php?showtopic=29482).

Its not very bad. But yes it can be frustrating if your code is not working because you missed some semi-colon or comma in the project-options. The cuda rule step in one of the posts on the Windows XP(?) forum could be helpful too.

Generally what I do is -

  1. Create a new VS project as usual.

  2. Add the code file for the cuda code.

  3. Rename it with the .cu extension

  4. Add the custom build step rules (I have copy-pasted the rule in a txt file where I copy it from into the new file).

  5. Change the name of the file in the rule. (I generally call my file cudaCode.cu, so I can skip this step)

  6. Write the cuda kernels in a separate file that I #include in the cudacode.cu

  7. Set the cuda kernel file to not be included in the build.

  8. Include the cuda libraries in the Additional Libraries in the Project Properties > Linker > Input

  9. Add paths to the Additional Libraries in Project Properties > Linker > General

  10. Add paths for Additional Include Directories in Project Properties > C/C++

  • Disclaimer:

Author has only worked with small-medium sized projects. Author is not sure if this would be a manageable way to work with huge projects with a large number of kernels and corresponding host code.