GPU OpenCV and TBB support

I am about to train my Haar classifier for flowers (which I am very skeptical about). I follow the CodingRobin Tut for everything.
http://coding-robin.de/2013/07/22/train-your-own-opencv-haar-classifier.html

It has now been emphasized that I am using GPU support, multithreading, etc., otherwise it will take several days to learn. I'm going to use the prebuilt libraries and hence the built-in opencv_traincascade utility .

I want to ask in advance if I will be able to use GPU support if I am using prebuilt libraries, given that I am installing CUDA?
Where does TBB fit the whole picture?
Do you recommend that I build the entire library from scratch with TBB and CUDA validation, or would that be a waste?

Note . I am using OpenCV 2.4.11 . And I am a complete newbie to OpenCV.