For the last months I’ve been working on a GPU accelerated version of the wavelet-based Dirac Video Codec. I am happy to announce that my work is finally in a state that it can be tested by others.
It shows how CUDA can be used to significantly accelerate video codecs in a platform independent way, without any shady “purevideo” madness. It is also a nice example of integrating CUDA into an existing library.
The library also takes advantage of asynchronous features to overlap GPU and CPU computing, as introduced in CUDA 1.1.
The rest can be found here: http://www.cs.rug.nl/~wladimir/sc-cuda/