#
✕ nick:chadmed
#asahi-dev · 2026-07-14 13:36
chadmed15.x needs sptm on m3
#asahi-dev · 2026-07-14 13:35
chadmedsofus: thats nothing, apple's version of bankstown had an off-by-one bug in it that created some very interesting looking spectrograms :P
#asahi-dev · 2026-07-14 13:33
chadmedim happy that it changed less than i expected but all the stuff that changed is really annoying
#asahi-dev · 2026-07-14 13:25
chadmedok thats close enough for today
#asahi-dev · 2026-07-14 13:25
chadmed 3.024621] apple-dcp 289c00000.dcp: RTKit: syslog message: nifiedPipeline.cpp:2413: src/UnifiedPipeline.cpp ERROR: DRIVER DIDN'T LOAD (start_hardware() failed.)
#asahi-dev · 2026-07-14 06:37
chadmedonly a few more unknown callbacks to go :3
#asahi-dev · 2026-07-13 06:00
chadmedthere is a lump of ~500ish bytes that need to fit in somewhere though, and it's not an extra argument to the swap_submit_dcp cb
#asahi-dev · 2026-07-13 05:59
chadmedtheres nothing new in there i dont think, stuff's just moved around and been padded out
#asahi-dev · 2026-07-13 03:48
chadmedthe swap req struct layour has changed quite significantly :))))))
#asahi-dev · 2026-07-12 13:01
chadmedthe tracer Just Worked
#asahi-dev · 2026-07-12 13:00
chadmeddoesnt look like dcp changed much for 14.8.3 which is nice. some structs changed size and some epic tags moved around. the fw doesnt zero some stuff anymore because RTKSTACK is in the middle of a bunch of structs where there should be zeroes/nothing
#asahi-dev · 2026-07-12 11:33
chadmed12 million tokens to find 1 copy-paste error
#asahi-dev · 2026-07-12 11:29
chadmedsigh v2 coming later...
#asahi-dev · 2026-07-12 11:28
chadmedoh it did actually find something stupid
#asahi-dev · 2026-07-12 11:27
chadmedtbf this would be fine if i was submitting only the first few patches in that series, which are indeed empty smc_hwmon nodes, but the advice is just a truism "the driver for this node wont do anything if the node is empty" yeah.... i know....
#asahi-dev · 2026-07-12 11:25
chadmed"the hwmon driver will not probe because there are no sensor nodes" yeah, look at the next commit idiot
#asahi-dev · 2026-07-12 11:23
chadmedyeah god forbid the LLMs scrape g.k.o but it's ok for sashiko bot to burn tokens giving me useless nonsense reviews
#asahi-dev · 2026-07-12 01:47
chadmedoh no it wasnt
#asahi-dev · 2026-07-12 00:59
chadmedoh m3 was already upstreamed with the smc subdev lol
#asahi-dev · 2026-07-11 22:24
chadmedhas anyone tested on M3 devs? i might as well add those too if all the same keys work
#asahi-dev · 2026-07-11 22:23
chadmedno stress, ill split it out per-soc anyway
#asahi-dev · 2026-07-11 22:19
chadmedsven: i'll get that done later today. will you take the changes via soc?
#asahi-dev · 2026-07-11 10:40
chadmedyes, it is not bridged though and we dont know where the logs are
#asahi-dev · 2026-07-11 10:37
chadmedaustralien sounds nothing like osterreic
#asahi-dev · 2026-07-11 10:33
chadmedi wish it was here :P
#asahi-dev · 2026-07-11 10:33
chadmedand if they split qld up who will thrash nsw every year in state of origin
#asahi-dev · 2026-07-11 10:33
chadmedbut every proposal to subdivide our states ends up being stupid and politicised
#asahi-dev · 2026-07-11 10:33
chadmedtheres probably a sane middle ground for a landmass this size between our 7 and their 50
#asahi-dev · 2026-07-11 10:31
chadmedgreedy george thats what theyre calling him
#asahi-dev · 2026-07-07 11:44
chadmedyeah
#asahi-dev · 2026-07-07 11:37
chadmedthats basically what i was going for with the cm3 also dispatching jobs to "idle" vps but idk how much state is too much
#asahi-dev · 2026-07-07 11:37
chadmedfeed the cm3 a descriptor in shared mem and make it do the hard work?
#asahi-dev · 2026-07-07 11:28
chadmedi dont see a way to keep the whole block fully stateless with the current design. maybe exposing "avd" as a media controller and then each vp as its own decoder?
#asahi-dev · 2026-07-07 11:27
chadmedyeah but how do we want to handle it? my approach seemed a little too stateful to fit nicely with v4l2sl/vk vid/vaapi
#asahi-dev · 2026-07-06 07:46
chadmedthe mca cluster does matter... i wonder if we just guessed/trialled and errored the correct routings
#asahi-dev · 2026-07-06 07:40
chadmedcould it be the mclk multiplier? could it be TX/RX direction clock config setting? who knows :P
#asahi-dev · 2026-07-06 07:39
chadmedcase in point, how the driver manipulates the MCA per-cluster SYNCGEN_MCLK_SEL register makes little sense. The driver set it to 0b111 when the cluster is a clock consumer, but simply setting BIT(1) also works for that case. Setting BIT(1) also works when it is a clock provider, and also setting BIT(0)
#asahi-dev · 2026-07-06 07:31
chadmedactually i can check this right now
#asahi-dev · 2026-07-06 07:29
chadmedpovik did most of this stuff though and the documentation appears largely lost to time which makes things difficult
#asahi-dev · 2026-07-06 07:28
chadmedim not sure if it matters which MCA cluster is used for which set of codecs. as i understand the MCA driver, the cluster index is referential to the CPU side of the DAI and can take in any group of CODECs. the macaudio driver just ensures for the speaker DAI that there are an even number of CODECs per CPU
#asahi-dev · 2026-07-06 07:21
chadmedyeah neither. i'll have a play with it
#asahi-dev · 2026-07-06 07:17
chadmedif A7-A9 have EL3 can you not just put m1n1 in it and use the usual svc/hvc on those?
#asahi-dev · 2026-07-06 07:10
chadmedjannau: thanks, i vaguely remember seeing that. i will see if it suits our use case when i have time and then maybe we can drop the fake regulator altogether
#asahi-dev · 2026-07-05 13:14
chadmed\o/
#asahi-dev · 2026-07-05 12:53
chadmedand the refcounted thing i thought of had actually been tried before and not accepted :P
#asahi-dev · 2026-07-05 12:52
chadmedthe shared gpio problem is annoying though. we tried to upstream the regulator solution and got told to come back with a treewide shared gpio api :/
#asahi-dev · 2026-07-05 12:50
chadmedand hard coding macaudio specific stuff into the tas2764 visense code but i dont think that is actually entirely necessary
#asahi-dev · 2026-07-05 12:50
chadmedabusing the regulator api to get shared gpio on the codec reset lines
#asahi-dev · 2026-07-05 12:49
chadmedmost of which unfortunately cannot be upstreamed
#asahi-dev · 2026-07-05 12:49
chadmedmacaudio actually squashed cleanly and if we get it upstream there'll only be 5ish patches left downstream for audio

plus ›