CUDA is a parallel programming environment for NVidia graphics cards [1]. It has found good acceptance for games, scientific computing and with the increasing acceptance of volunteer computing with BOINC [2] or distributed.net [3] it has a steadily growing user base.

For intructions towards an installation of CUDA drivers for Debian refer to NvidiaGraphicsDrivers. The CUDA-capable graphics cards also understand OpenCL, i.e. the respective standard for such application accelerations. But small differences prevail between AMD's and NVidia's implementations and binaries need to be explicitly prepared for either.

References