#
…

2025-10-05

00:02anticataclyst has quit [Ping timeout: 480 seconds]
00:10anticataclyst has joined #asahi
00:18anticataclyst has quit [Ping timeout: 480 seconds]
00:26anticataclyst has joined #asahi
00:26omidmash3 has joined #asahi
00:32omidmash has quit [Ping timeout: 480 seconds]
00:32omidmash3 is now known as omidmash
00:34anticataclyst has quit [Ping timeout: 480 seconds]
00:37anticataclyst has joined #asahi
00:37ipatch has joined #asahi
00:41HitokiriKincho has quit [Remote host closed the connection]
00:45anticataclyst has quit [Ping timeout: 480 seconds]
00:46diwi_ has joined #asahi
00:47diwi has quit [Read error: Connection reset by peer]
00:53anticataclyst has joined #asahi
01:10anticataclyst has quit [Ping timeout: 480 seconds]
01:15pb17 has quit [Ping timeout: 480 seconds]
01:16pb17 has joined #asahi
01:17anticataclyst has joined #asahi
01:20os has quit [Ping timeout: 480 seconds]
01:27anticataclyst has quit [Ping timeout: 480 seconds]
01:28nyx_o has quit []
01:30Brainium has quit []
01:31Brainium has joined #asahi
01:31nyx_o has joined #asahi
01:34HitokiriKincho has joined #asahi
01:34HitokiriKincho has quit [Remote host closed the connection]
01:34HitokiriKincho has joined #asahi
01:35anticataclyst has joined #asahi
01:38andymandias has quit [Read error: Connection reset by peer]
01:38andymandias has joined #asahi
01:42Brainium has quit []
01:51HitokiriKincho has quit [Remote host closed the connection]
01:52HitokiriKincho has joined #asahi
02:07HitokiriKincho has quit [Ping timeout: 480 seconds]
02:08tobhe_ has joined #asahi
02:08HitokiriKincho has joined #asahi
02:15tobhe has quit [Ping timeout: 480 seconds]
02:31hexdump02 has quit [Ping timeout: 480 seconds]
02:34ihaveamac has quit [Quit: fail]
02:35ihaveamac has joined #asahi
02:42cylm_ has quit [Quit: WeeChat 4.7.1]
02:47asahi has joined #asahi
02:48asahi has quit []
02:49asahi has joined #asahi
02:51asahi has quit [Remote host closed the connection]
02:51underhood31 has joined #asahi
03:09rizztty has quit [Ping timeout: 480 seconds]
03:09rizztty has joined #asahi
04:01nnym has quit [Quit: quitting out | tjmnkrajyej@gmail.com]
04:02nnym has joined #asahi
04:54andymandias has quit [Read error: Connection reset by peer]
04:54andymandias has joined #asahi
05:42HitokiriKincho has quit [Remote host closed the connection]
05:45HitokiriKincho has joined #asahi
05:46HitokiriKincho has quit [Remote host closed the connection]
05:46HitokiriKincho has joined #asahi
06:22nnym has quit [Quit: quitting out | tjmnkrajyej@gmail.com]
06:25nnym has joined #asahi
07:03HitokiriKincho has quit [Remote host closed the connection]
07:13lukeleppan has quit [Read error: Connection reset by peer]
07:15lukeleppan has joined #asahi
07:20HitokiriKincho has joined #asahi
07:30HitokiriKincho has quit [Ping timeout: 480 seconds]
07:48HitokiriKincho has joined #asahi
07:56HitokiriKincho has quit [Ping timeout: 480 seconds]
08:21HitokiriKincho has joined #asahi
08:29HitokiriKincho has quit [Ping timeout: 480 seconds]
08:38seb_ has quit [Quit: Konversation terminated!]
08:38seb_ has joined #asahi
09:03HitokiriKincho has joined #asahi
09:08seb_ has quit [Quit: Konversation terminated!]
09:12seb_ has joined #asahi
09:12HitokiriKincho has quit [Ping timeout: 480 seconds]
09:27roxfan2 is now known as roxfan
09:35___nick___ has joined #asahi
09:54mattia013 has quit [Remote host closed the connection]
09:55mattia013 has joined #asahi
09:56mattia013 has quit []
09:56mattia013 has joined #asahi
10:22linkmauveHi, is this a relevant dump of the latest supported DRM driver features? https://drmdb.emersion.fr/snapshots/205637cf682c
10:23linkmauveOne primary plane, one overlay plane, one eDP connector, and that’s it?
10:24chaos_princesshw can do a bit more, but that is about right
10:28jannaudepending on the device there might be an additional hdmi connector with primary and overlay plane
10:29jannaudesktop devices have the hdmi connector only
10:30linkmauveThere is no dedicated compositing hardware in the display controller, to be exposed as overlay/cursor planes? And the supported formats are only XRGB2101010 or X/ARGB8?
10:31linkmauveIs there a different 2D block used for compositing on macOS?
10:32jannauthere is a overlay plane. it has some quirks which makes it unusable as cursor plane
10:32chaos_princessmultiple, right? iirc 4?
10:32jannaumacos switches between hw blending and sw blending to work around this issue
10:32jannauit can only blend 2 layers/planes though
10:34jannaucurrently only XRGB2101010 and X/ARGB8 are exposed. hw supports more but needs more work
10:43leioI still don't understand how macos uses the extra planes then if can only blend 2 layers; like what's it used for, why do they exist
10:43leiopreparing the frame after or something?
10:43chaos_princessfor safari to directly feed dcp from avd, bypassing the gpu
10:45linkmauveSo it would support some YUV format(s) as well?
10:45jannauI think the 1st layer supports only a limited set of formats but is more power efficient
10:45jannauyes, it has YUV support
10:53__nick__ has joined #asahi
10:56___nick___ has quit [Ping timeout: 480 seconds]
11:00Moprius has joined #asahi
11:01yuka has quit [Remote host closed the connection]
11:01Moprius has quit []
11:05Moprius has joined #asahi
11:05yuka has joined #asahi
11:06Moprius has quit []
11:06Moprius has joined #asahi
12:03chadmedlinkmauve: the overlay plane is only useful as a cursor plane if there is compositor support for it (i.e. it will _immediately_ fall back to sw compositing if a cursor swap fails, like kwin does)
12:04chadmedfor video scanout, i want to (eventually, when i figure out how and get the time to do so) wire up all the YUV and HDR formats supported by the primary plane, however the hardware only seems to support framebuffers compressed with apple-proprietary formats
12:04chadmedso not really going to be useful until avd is working
12:05chadmeddcp _seems_ to have been designed partially with the apple tv in mind. the overlay plane is almost tailor made for displaying an OSD/media playback controls and not a whole lot else
12:07chadmedleio: macos never uses more than two planes afaik. additionally the cursor used to fall back to software near the edge of the screen just like kwin, but iirc in macos 13 they started just mangling the cursor framebuffer
12:07chadmedwhich i could not do in linux without making dcp crash. its also just a really disgusting hack and i hated it
12:10linkmauvechadmed, mpv --vo=dmabuf-wayland supports that TV use case, it uses a wl_subsurface for the OSD when displayed on top of the video (which includes the UI as well as subtitles).
12:10linkmauveOnce you want to play with that, it could be a good testcase.
12:11linkmauveGTK also supports that usecase using this widget: https://docs.gtk.org/gtk4/class.GraphicsOffload.html
12:11linkmauveIt’s used by GNOME’s new video player, Showtime.
12:11jannauI had a version which copied the pointer framebuffer to internal frame buffer with appropriate padding. We would need user space to allocate the frame buffer with enough padding
12:13chadmedlinkmauve: yeah id expect compositors/mpv to handle that stuff no worries. it was more just that the constraints imposed by only considering that one use case made it really difficult to get cursors to work properly
12:13chadmedfor context, the plane does not allow your destination rectangle to ever clip screen space. if you try to clip screen space, the firmware will crash
12:14chadmedyeah hence "this seems to only really be designed for OSDs"
12:14cylm has joined #asahi
12:14svenweren't there some rumors that dcp was originally introduced to reduce power consumption on the apple watch?
12:15svenespecially for that always-on display
12:15chadmedthat would make sense too
12:15chadmedprobably makes more sense than apple tv considering audio seems to just be bolted on as an afterthought
12:16chadmedi would expect a "designed for apple tv" thing to properly handle bitstream audio, and i would then expect the apple tv to support bitstream formats
12:16chadmedas it is, however, neither are true
12:18chadmedit actually really annoyed me that we couldnt just plumb avservice into the existing hdmi audio stuff
12:33linkmauveHow was their display controller before these devices? Like on earlier iPhone and such?
12:34linkmauveDid they composite every frame on the GPU?
12:36svenif you look at the dcp <-> xnu firmware interface it very much looks like they cut their previous display controller driver in half
12:36svenwe can still access the display controller mmio from the main core as well, at least on m1/m2
12:36km has joined #asahi
12:37chaos_princessthe pre-dcp controller that is used for the touchbar has (at least) two planes.
12:38linkmauvesven, and are there missing features in dcp that could be unlocked by going that MMIO route?
12:38svenmaybe, but then we'd need someone to reverse engineer and re-implement all the stuff dcp does
12:39svenwhich i'd be happy to see ofc but i don't have high hopes someone signs up for that on the off-chance it unlocks something useful ;)
12:40leioand if dcp doesn't use it, how likely it is you'll find it?
12:42svenyeah, that too. not impossible but unlikely
12:47jannauI'd rather not deal with DP link training, that seems to be already cursed enough with dptx
12:49jannaudcp conveniently also programs the DP to HDMI converters which we would take over as well
12:49crabbedhaloablut has quit []
12:50crabbedhaloablut has joined #asahi
12:53___nick___ has joined #asahi
12:56__nick__ has quit [Ping timeout: 480 seconds]
13:02thansen is now known as Guest28400
13:02thansen has joined #asahi
13:07Guest28400 has quit [Ping timeout: 480 seconds]
13:08Moprius has quit []
13:11HitokiriKincho has joined #asahi
13:19mkurz_ has quit [Quit: Konversation terminated!]
13:19mkurz_ has joined #asahi
13:19HitokiriKincho has quit [Ping timeout: 480 seconds]
13:19mkurz_ has quit []
13:20mkurz_ has joined #asahi
13:28Michae|Long has quit []
13:32HitokiriKincho has joined #asahi
13:47MichaelLong has joined #asahi
13:54HitokiriKincho has quit [Ping timeout: 480 seconds]
14:11___nick___ has quit [Ping timeout: 480 seconds]
14:23chadmedit is really nice that we can essentially just ask dcp nicely to do everything hard via rtkit
14:23chadmediomfb is a little bit crap and slow but eh
14:28Moprius has joined #asahi
14:49svenas painful as that rpc interface is I feel like implementing all that crap like link training for the various modes would be worse
14:50chadmedyeah exactly
14:50chadmedid rather deal with clunky swaps than... *gestures at displayport spec*
14:51Juest has joined #asahi
14:52bgtlover[mds]ha! that sounds as if the displayport spec is worse than hdmi
14:53chaos_princesslink training is just what you have to do on every high speed link, it is more physics hating you than the spec being obtuse on purpose
14:53gruetzhaxe has joined #asahi
14:57jannauI think the HDMI spec is worse but since you can't access the HDMI (or implement it in open source (HDMI 2.1)) it can't hurt you
15:01bgtlover[mds]<chaos_princess> "link training is just what you..." <- link training? that sounds like something to do with llms :p I hope not
15:03rizztty_ has joined #asahi
15:04rizztty has quit [Read error: Connection reset by peer]
15:05Juest has quit []
15:05svenit's usually clock recovery followed by a whole bunch of magic and adjusting filters and drive strengths and timings and whatever so that both sides can actually communicate at multi-Gbps speeds
15:07svenusually they exchange some well-known pattern at the beginning and tune all their parameters until the other side can recover that pattern successfully
15:08bgtlover[mds]lol indeed, someone had to say that. After all, apparently ddr5 memory requires proprietary blobs to train and the training involves machine learning the same way the branch predictor does, so yeah, the pun had to be made
15:09svenno, it really didn't
15:09bgtlover[mds]the pun about llms :p
15:10jannauno, it really didn't
15:12svenML for both branch prediction and link training also makes perfect sense. it's a very well specified problem and you immediately know if you made a mistake and the cost of a mistake isn't all that high. on the other hand, getting something correct can be verified immediately and you directly gain something
15:12chadmedits not that the dp spec itself is extraordinarily bad or anything, but managing any high bandwidth data link is a PITA especially when we know the thing on the other end of the link is probably noncompliant in some way or another. virtually every monitor is broken
15:12bgtlover[mds]but yeah, joking aside, it's very interesting that this stuff needs calibration, although achieving those speeds over thin wire must come with some tradeoffs because electrons can only travel so fast and in only so large a quantity before we get interference, causing information to degrade
15:13chadmedeasier to just go "hello dcp what modes did you negotiate" and then forward that to userspace
15:13bgtlover[mds]although I thought that happens automatically without interference from the OS, I thought the gpu and the other side do the negotiations on a firmware level, but yeah, TIL
15:13chadmedeven with all the cursed firmware versioning :p
15:14chadmedbgtlover[mds]: well yes, but someone has to write the firmware
15:14svenelectrons travels very slow actually (something like milimeters per second iirc), what you care about are the fields and the interference caused by the high speed switching between 0/1
15:14chadmedwhat we're discussing here is bypassing the firmware and for lack of a better term bit banging everything
15:15bgtlover[mds]ahh, gotcha! so y'all can't use apple's firmware?
15:15chadmedno, we can and we do
15:15chadmedits just a hypothetical
15:15chadmedand not a likely one, for the reasons discussed
15:16chadmedthe dcp firmware already does all of this and even notifies us _when_ it's done it
15:16chadmedall we need to do is ask it for the information
15:16bgtlover[mds]ahh ok, no wonder I was confused while passively reading chat
15:17chadmedthe annoying stuff is mostly just that the firmware kinda sucks and has an interface that is annoying to work with
15:17svenjust look at atcphy.c for how link training might otherwise looks like, and even that's already surprisingly simple and without much conditions or loops
15:18sventhough i guess at those speeds most of link training just has to happen in hardware :D
15:31leiochadmed: it's annoying when those dcp negotiated modes don't really match what your monitor can do because (apple) reasons :(
15:32okb has joined #asahi
15:49okb has quit [Read error: Connection reset by peer]
16:02HitokiriKincho has joined #asahi
16:09okb has joined #asahi
16:10HitokiriKincho has quit [Ping timeout: 480 seconds]
16:25KxCORP589400034558813182016 has quit [Ping timeout: 480 seconds]
17:00glem810054889940869 has quit [Quit: bye.]
17:05sonne has quit [Ping timeout: 480 seconds]
17:09KxCORP589400034558813182016 has joined #asahi
17:17okb has quit [Remote host closed the connection]
17:24glem8100548899408698 has joined #asahi
17:27okb has joined #asahi
17:36sonne has joined #asahi
17:42sonne has quit [Read error: No route to host]
17:43sonne has joined #asahi
17:51sonne has quit [Read error: No route to host]
17:51sonne has joined #asahi
17:52HitokiriKincho has joined #asahi
17:54HitokiriKincho has quit [Remote host closed the connection]
17:59mixi has quit [Ping timeout: 480 seconds]
18:02HitokiriKincho has joined #asahi
18:02mixi has joined #asahi
18:06okb has quit [Ping timeout: 480 seconds]
18:07HitokiriKincho has quit [Remote host closed the connection]
18:08HitokiriKincho has joined #asahi
18:19okb has joined #asahi
18:41okb has quit [Ping timeout: 480 seconds]
18:53___nick___ has joined #asahi
19:07Moprius has quit []
19:23___nick___ has quit [Ping timeout: 480 seconds]
19:28okb has joined #asahi
19:45okb has quit [Remote host closed the connection]
19:45HitokiriKincho has quit [Quit: Leaving...]
19:45okb has joined #asahi
19:49zopieux has quit [Quit: zopieux]
19:52zopieux has joined #asahi
20:22abbe has quit [Remote host closed the connection]
20:22abbe has joined #asahi
20:23abbe has quit [Remote host closed the connection]
20:23abbe has joined #asahi
20:55ddxtanx has quit [Remote host closed the connection]
21:03satomi has joined #asahi
21:05satomi has quit []
21:10okb has quit [Read error: Connection reset by peer]
21:10john-cabaj has joined #asahi
21:40Brainium has joined #asahi
21:56Extern has joined #asahi
22:00ddxtanx has joined #asahi
22:04cylm has quit [Quit: WeeChat 4.7.1]
22:23Brainium has quit []
22:34john-cabaj has quit [Ping timeout: 480 seconds]
22:45diwi_ has quit [Read error: Connection reset by peer]
22:47anticataclyst has quit [Ping timeout: 480 seconds]
22:47diwi has joined #asahi
22:54anticataclyst has joined #asahi
23:03anticataclyst has quit [Ping timeout: 480 seconds]
23:05km has quit [Ping timeout: 480 seconds]
23:11anticataclyst has joined #asahi
23:19diwi_ has joined #asahi
23:19diwi has quit [Read error: Connection reset by peer]
23:19anticataclyst has quit [Ping timeout: 480 seconds]
23:26theruran has joined #asahi
23:27anticataclyst has joined #asahi
23:34gilvbp has joined #asahi
23:35anticataclyst has quit [Ping timeout: 480 seconds]
23:43malfunction54 has quit [Remote host closed the connection]
23:43anticataclyst has joined #asahi
23:43malfunction54 has joined #asahi
23:45Extern has quit [Quit: WeeChat 4.7.1]
23:45malfunction54 has quit [Remote host closed the connection]
23:46malfunction54 has joined #asahi
23:52anticataclyst has quit [Ping timeout: 480 seconds]
23:59anticataclyst has joined #asahi
…