#
#asahi-dev · 2026-01-07 21:21
leiovulkan video does not have an interface to support stateful firmware based codecs
#asahi-dev · 2026-01-07 21:21
leiowe should probably do vulkan video driver IF the hardware can do stateless
#asahi-dev · 2026-01-07 21:21
svenyeah, iirc she let the domain expire (maybe by accident) after she moved on and went to university
#asahi-dev · 2026-01-07 21:19
fl0_idleio somehow I just thought it was a simple style...
#asahi-dev · 2026-01-07 21:17
leio(and https://web.archive.org/web/20240105073841/https://eiln.net/avd-notes.html )
#asahi-dev · 2026-01-07 21:16
leioah, nice, the content is up on alternative site. https://web.archive.org/web/20240325152417/https://eiln.net/avd-notes2.html is not lacking CSS though
#asahi-dev · 2026-01-07 21:15
Glanzmanncoughs xournalpp in svens direction.
#asahi-dev · 2026-01-07 21:05
fl0_idnice
#asahi-dev · 2026-01-07 20:49
svenipads with their pencil are *really* nice for that
#asahi-dev · 2026-01-07 20:49
sveni still have some old notes like that i think where i write register names from atcphy next to a MMIO trace after getting them from xnu's debug output
#asahi-dev · 2026-01-07 20:48
fl0_idwow i like this kind of markup :) https://eiln.github.io/posts/10-05-2023-avd-notes/draft.png
#asahi-dev · 2026-01-07 20:36
vupoh nice
#asahi-dev · 2026-01-07 20:36
svenhttps://eiln.github.io
#asahi-dev · 2026-01-07 20:32
vupalso, is there any more documentation of avd beyond this repo: https://github.com/eiln/avd ?
#asahi-dev · 2026-01-07 20:27
leiojannau: not beyond vkv research yet, no :(
#asahi-dev · 2026-01-07 20:14
svennickchan: okay, gonna pick it up probably this weekend then
#asahi-dev · 2026-01-07 19:58
jannaualso there will be large overlap. the code will essentially do the same
#asahi-dev · 2026-01-07 19:57
jannauva-api or vk video doesn't really makes much of difference. Having a working driver certainly beats discussions about which api is the correct one
#asahi-dev · 2026-01-07 19:56
vupI have written some very simple linux drivers, but nothing beyond that
#asahi-dev · 2026-01-07 19:54
jannauvup: the driver for the video decoding hardware (avd) will be more on the complex side of drivers. do you have any relevant background for that or would you start completely from scratch
#asahi-dev · 2026-01-07 19:54
chaos_princessva-api or vk video, stateless v4l is not a good fit for how the hw works
#asahi-dev · 2026-01-07 19:53
vupbut not a lot of software support afaict. Does anybody have a strong opinion about this?
#asahi-dev · 2026-01-07 19:53
vupthe next question would be, where should this driver be situated :P From what I can tell there are atleast three ways userspace talks to hardware acceleration for video. VA-API, vulkan video and v4l m2m. From what I can tell the stateless v4l m2m path might make the most sense, although I don't have a good overview of what software supports that. I *think* atleast firefox does, which would be good enough for starters. Vulkan video would be nice,
#asahi-dev · 2026-01-07 19:52
chaos_princessi don't think apple themselves do
#asahi-dev · 2026-01-07 19:52
chaos_princessdon't bother with the jpeg decoder
#asahi-dev · 2026-01-07 19:51
integralpilotokay I did research and apparently if we made a JPEG decoder not much software would use it, but video decoders would be used by a LOT of software
#asahi-dev · 2026-01-07 19:49
jannauleio: did you manage to do any work on avd?
#asahi-dev · 2026-01-07 19:45
integralpilotI think a driver for that would be really cool!
#asahi-dev · 2026-01-07 19:44
integralpilotoh interesting!
#asahi-dev · 2026-01-07 19:40
chaos_princessall of them have h254/h265/vp9, but av1 is m3+
#asahi-dev · 2026-01-07 19:40
integralpilotI know M3+ has a AVD hardware decoder, not sure about earlier
#asahi-dev · 2026-01-07 19:40
vupwhere would linux userspace even use hardware JPEG decoding?
#asahi-dev · 2026-01-07 19:40
integralpilotoh cool!
#asahi-dev · 2026-01-07 19:39
vupwell the advantage of video would be that it would be actually interesting to me :P
#asahi-dev · 2026-01-07 19:37
integralpilotit might be easier to start with static image decoding first, IIRC there’s a hardware JPEG decoder unit but I don’t think there’s a driver yet
#asahi-dev · 2026-01-07 19:32
vupis there already a driver for hardware accelerated video decoding? I saw most of the functionality was already reverse engineered. If there is no driver, would this be a suitable driver to get started with?
#asahi-dev · 2026-01-07 19:31
integralpilotbut doesn’t cause anything besides the extra logs and (maybe) power status taking longer than it should to update when you plug stuff in/out
#asahi-dev · 2026-01-07 19:28
integralpilotand yeah unknown charger event is unrelated, its a thing with newer SMC firmware and currently fixed on the “asahi” branch
#asahi-dev · 2026-01-07 19:27
integralpilotapparently according to Apple a reboot resets the SMC, if you’re plugged in: https://support.apple.com/en-au/102605
#asahi-dev · 2026-01-07 15:01
chaos_princessthere is pmu nvram, there is spi flash that stores some key value variables, there are coprocs that turn on as soon as power is applied and do not reset unless power is fully lost
#asahi-dev · 2026-01-07 15:00
fl0_idchaos_princess I assume that is similar to some nvram things on x86? Which is why some x86 with built in battery have a like a little extra hole to "cut power" but I had f.e. an example with a hackintosh where even that didn't work. reflashing the bios chip fixed it...
#asahi-dev · 2026-01-07 14:55
sklugCool. Thanks. And otherwise thanks for the fantastic work!
#asahi-dev · 2026-01-07 14:53
chaos_princessit is a laptop, so who knows, kinda hard to do a completely cold boot, i don't think a reboot resets the smc for example
#asahi-dev · 2026-01-07 14:52
chaos_princessok, we have seen that one, and fix is queued
#asahi-dev · 2026-01-07 14:52
sklugWhat scared me was that the issue survived a cold boot. Is there anything retained in the bootloader?
#asahi-dev · 2026-01-07 14:51
sklugI have a kernel oops that falls into that time: https://paste.debian.net/hidden/3488aa76
#asahi-dev · 2026-01-07 14:49
sklugA M2 MBP 14" 2023
#asahi-dev · 2026-01-07 14:49
chaos_princessthat is certainly something. what device is it?
#asahi-dev · 2026-01-07 14:49
sklugFrom that on the USB-C display would not be recognized anymore.
#asahi-dev · 2026-01-07 14:48
sklugWithout dissplay

plus ›