CUDA compiler not invoked (nvcc)

asked 2013-03-05 08:01:26 -0600

anauta gravatar image

Hi,

I cannot compile OpenCV (latest GIT sources) with CUDA 5.0 support. I have tried MANY things, yet the error is always the same:

For instance, when I want to build opencv_core, I get:

error LNK1181: cannot open input file 'build\modules\core\CMakeFiles\cuda_compile.dir\src\cuda\Debug\cuda_compile_generated_matrix_operations.cu.obj'

The cuda code is not being built. Also I do not see any reference of the nvcc compiler in the Output windows.

My system works fine otherwise: Clearing 'WITH_CUDA' makes everything build fine. Also, I can build and run the examples from CUDA 5.0 (both in 64 bit as well as 32 bit) without problems.

Can anyone help?

My system details: * Windows 7, core i7, 64 bit, 16GB RAM. * Visual Studio 2010 Premium * cMake 2.8.10.2

Note that in cMake, CUDA_SDK_ROOT_DIR is not found. However, when I tried to build with CUDA 4.2 (all cMake info appeared correct), the same error occurred...

edit retag flag offensive close merge delete