On Tuesday, 11 January 2022 07:26:32 CET H. Nikolaus Schaller wrote: > > Am 10.01.2022 um 22:51 schrieb Paul Boddie <[email protected]>: > > > > I wouldn't know without really taking a look at it. However, it sounds > > like the Linux support would be comparable to what has been done (or > > attempted) with the Ingenic auxiliary cores (AUX and MCU), > > I just want to add that the OMAP processors have similar facilities. > All of them have some sort of DSP and the OMAP5 comes with > "two Cortex-M4 processors for low-power offload and real-time control".
Yes, in the remoteproc driver collection, OMAP/AM processors seem to be represented, as well as some STM, Qualcomm, MediaTek and (as noted) Ingenic devices: https://elixir.bootlin.com/linux/latest/source/drivers/remoteproc [...] > AFAIR there were attempts (or success) to run uCLinux on the OMAP5 > subcores. Since they lack MMU it is a little limited. The lack of an MMU obviously limits the kind of system that can run on these things, although they might have some limited memory protection arrangement. Obviously, given the privileges these auxiliary cores have, one would want some kind of stable, well-understood program running on them. > But do these nice things solve any real world problem? > > If we can't get a nice and working and well accepted stack on the main > processor - how would a SoC specific subcore help? It adds fragmentation > is not available everywhere, difficult to debug etc. If they can remedy some kind of problem or weakness with the main processor, then they could be helpful. One of the comments I noted from the PinePhone survey was the slow response to calls: "microphone not working on receiving calls, too slow, I cannot pick up the call in time in time" https://www.pine64.org/2021/12/29/pinephone-community-poll/#comment-5519 As I noted before, it reminded me of those remarks made once upon a time that the Openmoko FreeRunner was slow at responding to calls, although I understand that there were also issues with the modem. Now, if something running on a "real-time" core could help with that, it might be interesting. A self-contained telephony application, for example. The featurephone in your smartphone. ;-) Nevertheless, I accept that the fragmentation such cores bring makes them unattractive for any broad effort to develop a Free Software phone architecture. Still, sometimes the lower-level software projects get the most enthusiastic support: I believe Rockbox got some use on the Ben NanoNote, for instance, and that doesn't seem so different from the kind of software these cores might support. Meanwhile, dual personality systems can also use software-based techniques and be realised running on the main processor cores, and I believe that such techniques are already widely used to implement various security-related features. I'm sure I mentioned that a while ago. Paul _______________________________________________ Community mailing list [email protected] http://lists.goldelico.com/mailman/listinfo.cgi/community http://www.tinkerphones.org
