Floating point vs Fixed point libraries

 Based on you end application, you can take advantage of the floating point DSP on OMAPL138 to calculate floating point FFTs.

There are libraries you can use that provide optimized implementation of FFT in both fixed and floating point processor:

Fixed point FFTs :

http://focus.ti.com/docs/toolsw/folders/print/sprc265.html

Floating point FFTs :

http://focus.ti.com/docs/toolsw/folders/print/sprc900.html

Thanks,
Gaurav