TODO

* simd

* threads/mpi

* DCT/DST codelets?  which kinds?

* user interface (& manual)

* implement fftw2 buffering: copy to buffers, transform in-place, copy back

* --enable-long-double fails in verify().  The problem appears
  to be a bug in gcc-2.95 -malign-double.  The problem disappears
  with gcc-3.1, icc, and without -malign-double.  I couldn't figure
  out exactly what the problem is. 

* still too much code duplication in rader

* rdft/rank-geq2 (and per-dimension kind's) needs to be tested

* cleanup reodft/verify.c using kernel/verify-lib.c

* 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.

* better test program (need syntax for real even/odd transforms, too)

* $prefix/bin/fftw_wisdom to generate/update /etc/fftw/wisdom and wisdom_conf.c

TODO before/at release:

* remember to remove -no-randomized-cse from rdft codelets before
  final release.  This flag speeds up the generator but produces worse
  codelets

* remember to thank/acknowledge: Franz and Stefan, Aurora project
  in Vienna, MIT Lincoln Labs, Compaq/HP testdrive program,
  sourceforge, AMD for the athlon box.

TODO before FFTW-$2\pi$:

* 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.

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