#asahi-dev · 2023-10-31 00:44
marcanwe're not releasing support for these until a new macOS with support for the whole lineup comes, I don't want another 12.4 situation#asahi-dev · 2023-10-31 00:44
marcanlol#asahi-dev · 2023-10-30 16:02
marcanwe could at least attempt to detect that (print something if the return address is after a wfi)#asahi-dev · 2023-10-30 16:01
marcanjanneg: `bt` won't work for CPUs in deep idle because the stack is in a funny state and there is no LR nor frame pointer after waking up from that#asahi-dev · 2023-10-29 17:03
marcan(and I still don't get all of it)#asahi-dev · 2023-10-29 17:03
marcanit's taken me a while to really wrap my head around this whole ASoC architecture thing#asahi-dev · 2023-10-29 17:02
marcanalso now I need a loopback widget or something for mca :p#asahi-dev · 2023-10-29 17:02
marcanI feel like the visense stuff is a bit hacky but works for us, but I'm not sure if that whole concept is easy to represent in ASoC at all#asahi-dev · 2023-10-29 17:01
marcanfeel free :)#asahi-dev · 2023-10-29 17:01
marcanlet me know if you spot anything horribly wrong with the stuff I've been doing! :)#asahi-dev · 2023-10-29 17:00
marcanI need to get some sleep, but glad to see you around :)#asahi-dev · 2023-10-29 16:43
marcanit only just hit me *what* I found literally 5 minutes ago :)#asahi-dev · 2023-10-29 16:41
marcanbit 16+ in the port select enables loopback from that cluster,tx[ab] in order#asahi-dev · 2023-10-29 16:40
marcanpovik: I just found it myself today lol#asahi-dev · 2023-10-28 20:06
marcan< Tramtrist> Silence is best folks. <- good point. I just deleted the tag, and the next tag will have the speakers only output silence.#asahi-dev · 2023-10-28 19:59
marcanfrom now on every claim that speakers are enabled/usable or attempt to use them publicly against our advice will result in an extra one week delay in the release of actual proper speaker support.#asahi-dev · 2023-10-27 18:08
marcanthat's just the iMacs I think though, hopefully ivsense is wired up on all the others including the mac mini/etc#asahi-dev · 2023-10-27 13:04
marcanpovik: do you think it's possible to implement a loopback pcm somehow in admac/mca? turns out non-ivsense machines have open loop speaker protection, and that would probably be the easiest way to safely implement that...#asahi-dev · 2023-10-27 07:28
marcanthere's an unknown 0xf00 in the bus config for in channels that apparently matters#asahi-dev · 2023-10-27 07:28
marcanfixed it, it was a bad register setting#asahi-dev · 2023-10-27 04:27
marcanhm no, I tried resetting the rings after resume but that just wedges the DMA :/#asahi-dev · 2023-10-27 04:11
marcannah but I think this is something else#asahi-dev · 2023-10-27 03:58
marcanwhat's that 0xf...#asahi-dev · 2023-10-27 03:58
marcanhmmm#asahi-dev · 2023-10-27 03:58
marcanCHAN_BUSWIDTH[1] = 0xf21 (WORD=1(W_16BIT), FRAME=2(F_4_WORDS))#asahi-dev · 2023-10-27 03:58
marcanCHAN_BUSWIDTH[0] = 0x22 (WORD=2(W_32BIT), FRAME=2(F_4_WORDS))#asahi-dev · 2023-10-27 03:58
marcanon j314:#asahi-dev · 2023-10-27 01:27
marcanif it doesn't we'll need some other way to maintain frame alignment#asahi-dev · 2023-10-27 01:27
marcanany idea if maybe macOS does something else there and it supports other window sizes? guess I need to trace it#asahi-dev · 2023-10-27 01:26
marcanpovik: looks like after suspend/resume sometimes ivsense data gets "torn". on this model it's 8 channels (4c x I/V), but I noticed admac only knows window sizes up to 4, so maybe that's why sometimes it gets split down the middle when we have partial reads due to a suspend/resume?#asahi-dev · 2023-10-23 20:09
marcandoesn't have the conclusion though#asahi-dev · 2023-10-23 20:08
marcanstory is in https://people.freedesktop.org/~cbrill/dri-log/?channel=freedesktop&highlight_names=&date=2023-10-23&show_html=true#asahi-dev · 2023-10-23 20:03
marcanso initially it appeared as if the data was stale#asahi-dev · 2023-10-23 20:03
marcanit was a migration so my guess is the mirroring process got stopped accidentally at some point, and they figured out how to sync the missing data from the old cluster#asahi-dev · 2023-10-23 20:02
marcanmaybe they fixed it#asahi-dev · 2023-10-23 20:01
marcannicolas17: was there a message or something?#asahi-dev · 2023-10-23 20:00
marcanI don't see any notifications?#asahi-dev · 2023-10-23 20:00
marcanthey were doing a migration, guess they screwed it up...#asahi-dev · 2023-10-22 08:47
marcancan't repro right now, but I've never been able to reliably repro... I might set the default to 0x0f for a while and see if it goes away#asahi-dev · 2023-10-22 08:44
marcansome of the apple quirks seem to just break audio, trying 0xf now#asahi-dev · 2023-10-22 08:34
marcangot one amp OCP wedged again. this time I got some logs; it seems it delivered a lot of OCP events, then started spamming TDM clock error events forever. speakersafetyd also went nuts but I think that was probably just because it was getting garbage data for ivsense.#asahi-dev · 2023-10-22 00:19
marcanpovik: pushed sound/speaker-safety with all the latest fixes. with that, suspend/resume while playing audio works properly. review pretty please? :>#asahi-dev · 2023-10-21 21:54
marcanI think there's still another bug though, since it shouldn't have wedged itself on sleep before, just restarted#asahi-dev · 2023-10-21 21:54
marcanok, and now fixed the speakersafetyd stuff (needed explicit suspend/resume handling)#asahi-dev · 2023-10-21 19:53
marcanokay yeah, it's a regmap bug. so that fixes the amps.#asahi-dev · 2023-10-21 17:55
marcanit's mistracking the page after a cache restore#asahi-dev · 2023-10-21 17:54
marcaninb4 this is a core regmap bug...#asahi-dev · 2023-10-21 17:53
marcan(stupid test is a sine on all speakers so...)#asahi-dev · 2023-10-21 17:53
marcanit wasn't "low volume", it's "tweeters only".#asahi-dev · 2023-10-21 17:53
marcanit's a paging problem