I'm very happy to see this. Besides using less hardware it simplifies
the process of matching the data from different implementations. There
are just too many ways that two floating point computations can get
slightly different results. This may open several doors.
Don Reid - W7DMR
On 9/4/26 10:42, Bruce Perens via Freetel-codec2 wrote:
I am experimenting with a re-implementation of codec2-mod done by Ai
at my direction. I had the AI characterize the mathematical domains,
there were three, and completely re-implement Codec2 in Riust, using
fixed--point. This is an independent implementation observing only the
algorithms of the original. The intent is to support M17 on hardware
it would not run upon at present.
The codec is at
https://github.com/BrucePerens/hams_open/tree/main/daemons/ham_digital_modes/src/codec2_3200
The C-language-compatibility layer is at
https://github.com/BrucePerens/hams_open/tree/main/daemons/codec2_3200_capi
This should make it drop-in to M17. Licensing is LGPL-3-and-later.
Obviously, copyright means nothing any longer and people like myself
with inadequate competence in a field can now direct a machine to do
the work anyway.
Thanks
Bruce
--
Bruce Perens K6BP
_______________________________________________
Freetel-codec2 mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/freetel-codec2
_______________________________________________
Freetel-codec2 mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/freetel-codec2
