Today I pushed two relatively simple changes that I wanted to get out of the way.
The first changeset adds NEON intrinsics to FFTW (although some testing / tuning still needs to be done). The end product will not be relying on NEON intrinsics but I thought it would be a good thing to benchmark against in any case. By specifically not using intrinsics (in the simd/ directory of FFTW), a lot more will have to be rewritten by hand in .S file format and I'm corresponding with Matteo about which routines to give special attention to.
The second changeset adds support for the arm7 cycle counter which will be used during the benchmarking stages.