CUDA link error with CMake

The following error message may be encountered when building CUDA code with CMake:

error LNK2019: unresolved external symbol __cudaRegisterLinkedBinary

It is caused by generated glue code not being linked in. The glue code is generated into files named with the project and a suffix:

myprojectname.device-link.obj

Ensure CUDA is enabled as a language for the project:

PROJECT(GPUProcessing LANGUAGES CUDA CXX)

Assuming you want separable compilation, you must enable resolving device symbols:

set_target_properties(${PROJECT_NAME}
   PROPERTIES
       CUDA_SEPARABLE_COMPILATION ON
       CUDA_RESOLVE_DEVICE_SYMBOLS ON
)

Note that the CUDA C++ language will follow the default C++ language level, so if you enable C++17 you will need to specify C++14 for CUDA (as of 9.2) as it does not yet support C++17:

set(CMAKE_CXX_STANDARD 17)
set(CMAKE_CUDA_STANDARD 14)

Also the CUDAToolkit package is the preferred and modern way to build with CUDA, thus:

find_package(CUDAToolkit)
add_executable(
    binary_linking_to_cudart 
    my_cpp_file_using_cudart.cpp
)
target_link_libraries(binary_linking_to_cudart PRIVATE CUDA::cudart)

Useful links:

gavinb/cuda_device_link_error.md