6670 floating point library

ilke ALTIN

Hey,

I am quite new to the 66x family and have been implementing fixed FFT and FIR functions using 6670. I have been able to use the dsp library for fft and fir filtering, however I could not find the floating point versions for fft and fir calculations. Is a floating point library available for 6670 or should I try using the library for 674x? Thanks in advance.

Regards,

ilke

over 14 years ago

0 Benoit Dribault over 14 years ago

Intellectual 440 points

Hi Ilke,

I have the latest version of the DSPLIB for C66x, which is 3.1.0.0

In the dsplib_c66x_3_1_0_0\packages\ti\dsplib\src folder, you should see the folders :

- DSPF_dp_fftDPxDP and DSPF_sp_fftSPxSP which is the fft for respectively double precision (double) and simple precision (float) FFT.

- DSPF_sp_fir_cplx DSPF_sp_fir_gen DSPF_sp_fir_r2 DSPF_sp_fircirc for simple precision FIR.

In those folders, you should see a folder called c66 which contains the sources of the library. In this folder, there are two implementations for the operation.

A natural c implementation (easy to read and understand) ex : DSPF_sp_fftSPxSP_cn.c and DSPF_sp_fftSPxSP_cn.h

An optimized version ex DSPF_sp_fir_gen.c and DSPF_sp_fir_gen.h

Take a look at those files to have a little description. You can add the source and the corresponding header to your project (I use copy-paste) tu use those functions.

Be careful, sometime, you must respect some assumptions needed by an optimized version. (refere to those files)

I hope it will be helpful.

Regards,

Benoit