The data type to the FFT is controlled by the template parameter TT_DATA. This may take one of 3 choices: cint16, cint32 or cfloat. This selection applies to both input data and output data.
The template parameter TT_TWIDDLE may take one of two values, cint16 or cfloat, but currently this value is forced by the choice of TT_DATA, so TT_TWIDDLE must be set to cint16 for TT_DATA = cint16 or cint32 and TT_TWIDDLE must be set to cfloat when TT_DATA=cfloat.