I’m a total newbie in gpgpu and I’m beginning to investigate how to translate my current code for my new GPU (8800 GTS). My code is relatively simple and it is mainly fft bottlenecked. So my first approach is to use the batch mode of the CUFFT library to do something like this:
fft -> some simple processing -> fft -> some simple processing … so on …
With the current API I just don’t see how to insert my “some simple processing” between each fft. Somebody have any clue how to do this ?