The "fft" binary
----------------
1/ Is built by typing "make"

2/ The sinopsis is
   fft <an_unsigned_int> <an_unsigned_int>

    The first arg is an integer in the range 0..15
    if 0 the Modpn 1D FFT is used1
    if 1 the BPAS 1D FFT (whichever code generator is used) is used
         and the result is not checked 
    if 16 the BPAS 1D FFT (whichever code generator is used) is used
         and the result is checked (takes time ...)
    if 2  then 1+1 shuffles are done (no FFT is compuuted)
    if 3  then 2 shuffles are done in one pass
    if 4  then 1+1 shuffles are done (no  FFT is compuuted)
    if 5  then 1+2 shuffles are done 
    if 6  then 3 shuffles are done in one pass
    if 7  then just vectors allocation
    if 8 then the TFT (relax) parallel code generator is used, no verification
    if 9 then the TFT (relax) parallel code generator is used and the results
          are verified using FFT+zero_padding+shuffling
    if 10 then TFT (relax) benchmarking is done
    if 11 then the results of TFT_tree1::Shuffle_tft(K,A,SB1) are returned
    12 and 13 are removed because of useless
    if 14 then the ITFT (strict) parallel code generator is used and the
    results are used with forward TFT

    The second arg is a power of 2 (at least 2)
    is the size of the input vector

    The characteristic (prime numer) is hard coded. 

    IMPORTANT (TODO) make the number of CilkPlus workers an argument
     of this test binary