Just for fun:
  Keccak-3200 by using int128_t and extending the
  round constants (of course this is problematic because
  the construction of the round constants are designed
  to for at most 64 bits so we will get that the higher
  halfs are equal to the next's lower half; hopefully
  this does not matter)

Test parallelisation and acceleration.

Measure cycles per byte.

Add tests for HMAC.