Como puede imaginar, la arquitectura de hardware de las GPU puede variar significativamente. Hay diferentes fabricantes, e incluso diferentes generaciones de GPU del mismo fabricante pueden no ser comparables. Al mismo tiempo, los fabricantes de GPU normalmente no divulgan todos los detalles del hardware de sus productos al público. Una de las consecuencias de esto es la necesidad de usar controladores propietarios bajo Linux, si desea aprovechar al máximo su tarjeta gráfica.
Fortunately an industry consortium lead by The Khronos Group has developed an open, standardized interface called OpenCL, which allows your GPU to be used as a numerical processing device. OpenCL offers a C99-like programming language with a strong focus on parallel computing. An application that wants to use OpenCL will need OpenCL source code that it hands over to a hardware-specific compiler at run-time. This way applications can use OpenCL on different GPU architectures (even at the same time). All of the hardware “secrets” are hidden in this compiler and are normally not visible to the user (or the application). The compiled OpenCL code is loaded onto your GPU and – with certain API calls – it is ready to perform calculations for you.