On Windows, we can choose between developing with Visual Studio (Visual C++) or with the Windows port of GCC (MinGW). To stay consistent with the Linux version, we will be using MinGW along with MSYS2. This means that we'll have the same compiler toolchain and same Bash shell and utilities, along with the Pacman package manager.

After installing the vendor's GPGPU SDK, as described previously, simply execute the following command line in an MSYS2 shell in order to install the OpenCL headers:

    $ pacman -S mingw64/mingw-w64-x86_64-opencl-headers

Or, execute the following command line when using the 32-bit version of MinGW:

    mingw32/mingw-w64-i686-opencl-headers 

With this, the OpenCL headers are in place. We now just have to make sure that the MinGW linker can find OpenCL library. With the NVidia CUDA SDK, you can use the CUDA_PATH environment variable for this, or browse the install location of the SDK and copy the appropriate OpenCL LIB file from there to the MinGW lib folder, making sure not to mix the 32-bit and 64-bit files.

With the shared library now also in place, we can compile the OpenCL applications.