Here is an organized folder to develop CUDA applications which should work unchanged under Mac OS X, Linux or Windows.

Install CMake via the installer or your package manger, e.g. on OS X you can do:

$ brew install cmake

Install CUDA 4.x from NVIDIA's CUDA Zone.

Finally you will also need to make sure that GLEW is installed.

In the src/ create a project folder e.g. matrixMul/ with its source files and create a new CMakeLists.txt listing the CUDA and C/C++ files and the libraries to link against e.g.:

INCLUDE_DIRECTORIES(
  .
)

CUDA_ADD_EXECUTABLE (matrixMul
  matrixMul.cu
  matrixMul_gold.cpp
)
  
TARGET_LINK_LIBRARIES (matrixMul
  cutil
  shrutil
  ${CUDA_CUBLAS_LIBRARIES}
)

Finally add the new project folder, for example matrixMul to the src/CMakeLists.txt file:

add_subdirectory (matrixMul)

Open a Visual Studio command shell or a terminal in Mac OS X or Linux and go to the top directory and do:

$ cmake CMakeLists.txt

Under Windows this should create a Visual Studio project file which you can open, and under Mac OS X or Linux do:

$ make

and hopefully the CUDA executable will be compiled in the bin/ folder.

Please refer to the NVIDIA end user license agreement (EULA) associated with this source code for terms and conditions that govern your use of this software.