/*
    FFT32
    fft16(z);       24 MUL, 144 ADD
    fft8(z+n4*2);   4 MUL, 52 ADD
    fft8(z+n4*3);   4 MUL, 52 ADD
    pass(z,ff_cos_##n,n4/2);\   56 MUL 137 ADD
    ==============
    88 MUL, 385 ADD
*/

/*
    FFT64
    fft32(z);                   88 MUL, 385 ADD
    fft16(z+n4*2);              24 MUL, 144 ADD
    fft16(z+n4*3);              24 MUL, 144 ADD
    pass(z,ff_cos_##n,n4/2);\   120 MUL, 280 ADD
    ===============================
    256 MUL, 953 ADD

*/

/*
    FFT128
    fft64(z);                   256 MUL, 953 ADD
    fft32(z+n4*2);              88 MUL, 385 ADD
    fft32(z+n4*3);              88 MUL, 385 ADD
    pass(z,ff_cos_##n,n4/2);\   248 MUL, 568 ADD
    ===============================
    680 MUL, 2291 ADD

*/

/*
    FFT256
    fft128(z);                  680 MUL, 2291 ADD
    fft64(z+n4*2);              256 MUL, 953 ADD
    fft64(z+n4*3);              256 MUL, 953 ADD
    pass(z,ff_cos_##n,n4/2);\   504 MUL, 1144 ADD
    ===============================
    1696 MUL, 5341 ADD

*/

/*
    FFT512
    fft256(z);                   1696 MUL, 5341 ADD
    fft128(z+n4*2);              680 MUL, 2291 ADD
    fft128(z+n4*3);              680 MUL, 2291 ADD
    pass(z,ff_cos_##n,n4/2);\    1016 MUL, 2296 ADD
    ===============================
    4072 MUL, 12219 ADD

*/