The data type of the DFT is controlled by the TT_DATA template parameter.
The DFT on AIE supports the following types for TT_DATA: cint16, cint32 or cfloat. On AIE-ML the data must be of an integer type as floats are not supported; TT_DATA can be cint16 or cint32. The data type of the output will be equal to that of TT_DATA.
The data type of the DFT coefficients is specified by the template parameter TT_TWIDDLE. However, this is currently determined by the input data type. If TT_DATA is an integer type (cint16 or cint32), TT_TWIDDLE must be set to cint16. If TT_DATA is set to cfloat then TT_TWIDDLE must also be cfloat.