CUDA, Eclipse, and Windows

I am new to CUDA and I apologize if this thread topic has already been covered.

I am wondering if it is possible to make CUDA, Eclipse, and Windows play nice together. I am beginning to think perhaps not.

I have installed the CUDA drivers, toolkit and SDK. I was originally using Microsoft Visual Studio, but I’m a Linux person and was not having any luck. (I was mostly frustrated with the apparent inability to manage your own makefile with Visual Studio and therefore my inability to make Visual Studio use both gcc and nvcc compilers in my projects.)

My ultimate goal is to take existing code and “CUDA-ize” pieces of it. So starting with a template project is not an option for me. I should be able to set up an environment to do this if CUDA is any sort of useful framework.

I am most familiar with Eclipse as a development environment and have experience writing OpenGL programs in Eclipse. I believe I have set up all my environment variables correctly and I have Cygwin installed.

When I try to compile the convolution example, I get the following error:

nvcc -ccbin “C:\CUDA\bin” -c -D_DEBUG -DWIN32 -D_CONSOLE -D_MBCS -Xcompiler /EHsc,/W3,/nologo,/Wp64,/Od,/Zi,/RTC1,/MTd -I"C:\CUDA\include" -I./ -I…/…/common/inc -o convolutionSeparable
nvcc fatal : nvcc cannot find a supported cl version. Only MSVC 7.1 and MSVC 8.0 are supported
make: *** [convolutionSeparable] Error 255

I do have MSVC 8.0 installed so I’m boggled.

Any assistance on this would be greatly appreciated. CUDA is an interesting framework, but I am very close to abandoning it altogether because it’s absurd ease of use. I’m not a guru at getting C/C++ projects up and going because I prefer Java, but I’m a graduate student and consider myself quite savvy.

Thank you in advance.


Is cl.exe in your Eclipse path? I bet it’s not (it’s not added by default to the Windows path, this is why there’s the Visual Studio Command Prompt entry in the start menu), so nvcc isn’t finding it and giving you that error.

ALTERNATELY, is gcc in your path as well as cl? I have no idea (haven’t played with Cygwin and nvcc’s interaction with it), but I’m wondering if it finds gcc, realizes “hey this isn’t Linux,” and dies.

As far as I know, I have my environment set up correctly, but I might be missing something. I can compile and run normal (non-CUDA) C/C++ code in Eclipse.

Does anyone know where cl.exe lives so that I can double check to make sure it’s in the path?

Thanks for your response.



Can you tell me exactly how cl.exe can be added to the PATH, cuz I cant seem to find it.