I learned a lot from the article which shows and discusses detailed Kernel code. However, there is no listing for the Host Code that illustrates how the Kernels are actually used. I can see that some of the Kernels are straightforward to implement, but implementing the various butterfly Kernels are not trivial. Having the host code is essential to really understanding how to implement the Kernels and their profiling as discussed in the article.
Is the Host code from the article available anywhere?
Has anyone else had experience getting these particular Kernels to work for a complete FFT calculation?
I sent this question directly to "AMD Support" back on July 5, 2012, but never really got an answer.