Regardless of which platform and GPU you have, the most important part of doing OpenCL development is to obtain the OpenCL runtime for one's GPU from its manufacturer. Here, AMD, Intel, and NVidia all provide an SDK for all mainstream platforms. For NVidia, OpenCL support is included in the CUDA SDK.
Along with the GPU vendor's SDK, one can also find details on their website on which GPUs are supported by this SDK.
After installing the vendor's GPGPU SDK using the provided instructions, we still need to download the OpenCL headers. Unlike the shared library and runtime file provided by the vendor, these headers are generic and will work with any OpenCL implementation.
For Debian-based distributions, simply execute the following command line:
$ sudo apt-get install opencl-headers
For other distributions, the package may be called the same, or something different. Consult the manual for one's distribution on how to find out the package name.
After installing...