Hello!
I need to implement complex floating point (single precision) FFT on C6670. I am considering DSPLIB v.3.1. for my case. It seems there is DSPF_sp_fftSPxSP() function. From documentation which comes with DSPLIB its hard for me to get usage example. Would appreciate some reference.
Another point is complex data packing. In other thread on e2e (http://e2e.ti.com/support/dsp/c6000_multi-core_dsps/f/639/p/273221/957418.aspx#957418) I saw clear statement, when it comes to intrinsics for complex data manipulation, real part goes to odd half, imaginary - to even half of the register or register pair. I have to use some more complex data manipulations, so I use __float2_t container. This container places imaginary part in even address, real - in odd address.
I had experience with integer FFT of DSPLIB on C64. That used other packing principle, when real part was in even address, imaginary - in odd.
So my question is what is right order of input data to DSPF_sp_fftSPxSP()? Is it consistent with intrinsics input?
Thanks in advance.