Some more very fine work from Danilo and the mcHF team, here's an email from Danilo:
"I think we're done now with FreeDV :-) :-) We achieved our goal: The first clean QSO has been taking place between 2 mcHF using FreeDV yesterday. We found a few performance hotspots and could eliminate some of these. We lost about 3ms more per 20ms in total which now allows us to run the mcHF normal code plus FreeDV 1600 RX and TX in real-time." So they have a fully embedded SDR/FreeDV system up and running. Well done guys :-) These guys have taught me a few tricks about STM32 optimisation! Mostly through C code changes - the latest patches have no ARM library calls. The changes are documented and now checked into codec2-dev. They have worked hard to build the tfdmdv framework for their target platform, which allowed us to carefully check all changes against the reference Octave modem. This is a crucial step for any ARM-specific optimisations. It's been a pleasure to work with Danilo and the mcHF team. Thanks, David On 09/09/16 08:34, David Rowe wrote: > Further to the email below, Danilo and the mCHF team have submitted the > first of some ARM-specific patches to the fdmdv modem used from FreeDV 1600. > > They worked hard to get a version of the tfdmdv framework running on a > STM32 target platform. This means they can carefully check nothing has > broken as they try other changes. > > The changes have been checked into SVN. The x86 code checks out. I > have built, but not tried the resulting sm1000 binary. If someone would > like to try just email me and I will send you the binary, or feel free > to build and try it yourself. > > Thanks, > > David > > On 31/08/16 10:25, David Rowe wrote: >> Hello List, >> >> Recently I've been working off-list with the mcHF team, who are >> integrating FreeDV with their STM32F4 based radio. >> >> They have sent me some very nice generic C patches which I have checked >> into SVN, and are now looking at using platform specific optimisations >> such as ARM library calls. These are tricker to tests, as they need to >> run on target harwdare, then dump files to a PC host for evaluation. >> >> To test the patches I have resurrected the tfdmdv Octave code that I >> broke last year while working on the cohpsk modem. This is a C/octave >> framework for verifying the operation of the fdmdv modem against the >> reference Octave code, more in this blog post: >> >> http://www.rowetel.com/blog/?p=3427 >> >> It's also a useful test framework for any ports of the fdmdv code to >> other languages. >> >> I still need to fix the other Octave fdmdv utilties. >> >> Cheers, >> >> David >> >> >> ------------------------------------------------------------------------------ >> _______________________________________________ >> 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 > ------------------------------------------------------------------------------ _______________________________________________ Freetel-codec2 mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/freetel-codec2
