TODO before FFTW-$2\pi$:

* MPI version

* DCT/DST codelets?  which kinds?

* investigate the addition-chain trig computation

* I can't believe that there isn't a closed form for the omega
  array in Rader.

* merge genfft-k7 generator with the main genfft branch.
  genfft-k7 was written by Stefan Kral based on the fftw-2.1 genfft.
  (The k7 stuff is becoming obsolete because it is not 64-bit clean.
  We should phase it out in the next release.)

* implement rdft/problem2 for even radices other than 2.

* convolution problem type(s)

* Explore the idea of having n < 0 in tensors, possibly to mean
  inverse DFT.

* better estimator: possibly, let "other" cost be coef * n, where
  coef is a per-solver constant determined via some big numerical
  optimization/fit.

* vector radix, multidimensional codelets

* it may be a good idea to unify all those little loops that do
  copying, (X[i], X[n-i]) <- (X[i] + X[n-i], X[i] - X[n-i]),
  and multiplication of vectors by twiddle factors.

* Pruned FFTs (basically, a vecloop that skips zeros).

* Try FFTPACK-style back-and-forth (Stockham) FFT.  (We tried this a
  few years ago and it was slower, but perhaps matters have changed.)

* dif, difsq simd codelets

* Generate assembly directly for more processors, or maybe fork gcc.  =)

* ensure that threaded solvers generate (block_size % 4 == 0)
  to allow SIMD to be used.

* consider whether it would be simpler to have a uniform description
  (IO, m0, m1, W) for twiddle problems, where IO always points
  to the beginning of the array, W always points to the beginning
  of the twiddle table, and the problem is solved in the
  range m \in [m0, m1).  Currently we have a messy situation in which
  we sometimes use [m0, m1), sometimes we use [mstart, mstart+mcount),
  and we have to adjust W with the X(twiddle_shift)() hack.  Codelets
  should obey the uniform protocol as well.

* memoize triggen.

* orb problems with FFTW_PRESERVE_INPUT ought to use SIMD somehow.
  (Currently they reduce to an r2r problem that does not use SIMD.)

* eliminate alignment hacks, which ought to be obsolete by now.