#
…

2026-03-08

00:14ElijahBare[m] has joined #asahi-gpu
00:35[KongrooParadox] has joined #asahi-gpu
00:38KongrooParadox has quit [Ping timeout: 480 seconds]
00:50pb17 has quit [Ping timeout: 480 seconds]
01:11pb17 has joined #asahi-gpu
01:17chadmed has quit [Quit: Konversation terminated!]
01:18chadmed has joined #asahi-gpu
02:21gavbog[m] has joined #asahi-gpu
02:42pb17 has quit [Ping timeout: 480 seconds]
03:03pb17 has joined #asahi-gpu
03:08KongrooParadox has joined #asahi-gpu
03:10[KongrooParadox] has quit [Ping timeout: 480 seconds]
03:10wildblue has joined #asahi-gpu
03:15rainbyte has joined #asahi-gpu
03:35[KongrooParadox] has joined #asahi-gpu
03:38KongrooParadox has quit [Ping timeout: 480 seconds]
04:11minecrell1 has joined #asahi-gpu
04:13ziad[m] has quit [Ping timeout: 480 seconds]
04:13nel[m] has quit [Ping timeout: 480 seconds]
04:13FG[m] has quit [Ping timeout: 480 seconds]
04:13gavbog[m] has quit [Ping timeout: 480 seconds]
04:13WindDogStudio[m] has quit [Ping timeout: 480 seconds]
04:13_master_sword_[m] has quit [Ping timeout: 480 seconds]
04:13svenpeter[m] has quit [Ping timeout: 480 seconds]
04:13weypro[m] has quit [Ping timeout: 480 seconds]
04:13yuri[m]1 has quit [Ping timeout: 480 seconds]
04:13Bjarne[m] has quit [Ping timeout: 480 seconds]
04:13Darkodayday[m] has quit [Ping timeout: 480 seconds]
04:13Tieppo[m] has quit [Ping timeout: 480 seconds]
04:13adambenmoussa[m] has quit [Ping timeout: 480 seconds]
04:13tabbytab10[m] has quit [Ping timeout: 480 seconds]
04:13Yo_man[m] has quit [Ping timeout: 480 seconds]
04:13prike2830 has joined #asahi-gpu
04:13SlShahafLevi[m] has quit [Ping timeout: 480 seconds]
04:13Moderation[m] has quit [Ping timeout: 480 seconds]
04:13nullbert[m] has quit [Ping timeout: 480 seconds]
04:13ElijahBare[m] has quit [Ping timeout: 480 seconds]
04:13JamesCalligeros[m] has quit [Ping timeout: 480 seconds]
04:13xXxMeowUwUKittenxXx[m] has quit [Ping timeout: 480 seconds]
04:14pas[m] has quit [Ping timeout: 480 seconds]
04:14Felix[m]1 has quit [Ping timeout: 480 seconds]
04:14ants[m] has quit [Ping timeout: 480 seconds]
04:14lebakassemmerl[m] has quit [Ping timeout: 480 seconds]
04:14jmax[m] has quit [Ping timeout: 480 seconds]
04:14Dcow[m] has quit [Ping timeout: 480 seconds]
04:14FiniteCode[m] has quit [Ping timeout: 480 seconds]
04:14pulvite[m] has quit [Ping timeout: 480 seconds]
04:14lacamar[m] has quit [Ping timeout: 480 seconds]
04:14zackng[m] has quit [Ping timeout: 480 seconds]
04:14neobrain[m] has quit [Ping timeout: 480 seconds]
04:14minecrell has quit [Ping timeout: 480 seconds]
04:14BriannaBabiack[m] has quit [Ping timeout: 480 seconds]
04:14CoZo[m] has quit [Ping timeout: 480 seconds]
04:19prike283 has quit [Ping timeout: 480 seconds]
04:19prike2830 is now known as prike283
05:08ziad[m] has joined #asahi-gpu
05:09WindDogStudio[m] has joined #asahi-gpu
05:18gavbog[m] has joined #asahi-gpu
05:41yuri[m]1 has joined #asahi-gpu
05:58KongrooParadox has joined #asahi-gpu
06:00pb17 has quit [Ping timeout: 480 seconds]
06:01[KongrooParadox] has quit [Ping timeout: 480 seconds]
06:18Darkodayday[m] has joined #asahi-gpu
06:29pb17 has joined #asahi-gpu
06:36oliverAlyssa comments on the tiled modifiers: These layouts fundamentally do not have meaningful strides. No matter how we
06:36oliver * specify strides for these layouts, userspace unaware of Apple image layouts
06:36oliver * will be unable to use correctly the specified stride for any purpose.
06:36oliver * Userspace aware of the image layouts do not use strides. The most "correct"
06:36oliver * convention would be setting the image stride to 0. Unfortunately, some
06:36oliver * software assumes the stride is at least (width * bytes per pixel). We
06:36oliver * therefore require that stride equals (width * bytes per pixel). Since the
06:36oliver * stride is arbitrary here, we pick the simplest convention.
06:37oliverThis then fails in atomic_check for textured that do not have a stride of a multiple of 64 bytes.
06:39oliverWhat do you guys thing is the best way to go here? Only check the stride for the linear modifier?
06:40chadmedi actually dont think dcp cares too much. the stride check in plane_atomic_check is relatively new, and linear worked fine before we added it
06:40jannaurounding the stride up to the next multiple of 64
06:43chadmedi commented out the check and cursors work again. if this is a limitation purely on the agx side then it should not spill into the dcp driver and mesa or the asahi driver needs to check for this
06:47jannausince the tiles are 16x16 with a 4-byte element size the pitch should be 64-byte aligned
06:48jannaudcp very much needs (or needed) 64-byte aligned pitch, see 0f27ba328875947a513
06:54Tieppo[m] has joined #asahi-gpu
06:57oliverThe pitch is actually 64 bytes aligned, it just does not match the comment.
07:03oliverAnd due to that userspace sets pitch based on the width, not rounded up like jannau suggested.
07:04balrog_ has joined #asahi-gpu
07:08balrog has quit [Ping timeout: 480 seconds]
07:10FG[m] has joined #asahi-gpu
07:10WindowPain has quit [Quit: Ping timeout (120 seconds)]
07:11WindowPain has joined #asahi-gpu
07:13Yo_man[m] has joined #asahi-gpu
07:27oliverWe can now either "break" userspace and change the comment (and adjust mesa too) or update the atomic_check and only verify stride for linear, because we know tiled will have proper stride. Which one should it be?
07:29chadmedwe cannot break userspace, the uapi has been upstreamed
07:31jannauI doubt there is user space (ignoring mesa) which has code specific to apple modifiers and much less so code written to explicitly considering that comment
07:35nel[m] has joined #asahi-gpu
07:38chadmedwhy would swap_enabled ever be set to >0xf when there is a surface wtf
07:40_master_sword_[m] has joined #asahi-gpu
07:48pb17 has quit [Ping timeout: 480 seconds]
07:48pb17 has joined #asahi-gpu
07:56jannauI think we should change the comment but changing dcp isn't out of question
07:56jannaudoes dcp actually care about the stride for tiled formats?
07:58chadmedi tried rounding them up just before and it exploded pretty badly
07:58chadmedlots of magenta and black
08:00chadmedit doesnt seem to care at all if theyre not multiples of 64 though. currently have a cursor displays just fine with stride = 132b
08:00chadmedunless drm_info is lying
08:01svenpeter[m] has joined #asahi-gpu
08:02svenpeter[m] is now known as Guest4594
08:02jannaumacos sets stride to width (already a multiple of 16) * 16 * blocksize for &b38
08:04Moderation[m] has joined #asahi-gpu
08:13weypro[m] has joined #asahi-gpu
08:14jannauwe can't really break user-space with the interchange modifier
08:15jannauwe can just document a different stride for interchange
08:17nullbert[m] has joined #asahi-gpu
08:22SlShahafLevi[m] has joined #asahi-gpu
08:38skipwich has quit [Ping timeout: 480 seconds]
08:44ElijahBare[m] has joined #asahi-gpu
08:49skipwich has joined #asahi-gpu
08:52Guest4594 is now known as svenpeter[m]
08:59chadmedim giving up on >2 planes again for now. im not sure what it is we could possibly be doing so differently to macos that causes this to not work
08:59chadmedreproducing the behaviour in traces still doesnt work
09:00oliverjannau: I imagine that your commit would help, yes.
09:01xXxMeowUwUKittenxXx[m] has joined #asahi-gpu
09:07oliveryes, confirmed, works as expected.
09:08WindowPain has quit [Quit: Ping timeout (120 seconds)]
09:08WindowPain has joined #asahi-gpu
09:24jannauit's currently not wrong but I think we want to align the stride to tile width * blocksize in the modifier documentation. Same for the height
09:25jannaudcp could still ignore it because as far as I see it wouldn;t use that value anyway
09:44chadmedhahahah oh my god i "fixed" it
09:44chadmedsurface 2 is just... broken?
09:44chadmednormalized_zpos == 2 ? normalized_zpos + 1 : normalized_zpos
09:44chadmedthen just expose three planes
09:44chadmedso how does macos use surface 2
09:44chadmedthere must be rules for it or something
09:46chadmedswaps with three surfaces in them also seem to work fine like this too
09:49pb17 has quit [Ping timeout: 480 seconds]
09:49jmax[m] has joined #asahi-gpu
09:50kaazoo has joined #asahi-gpu
09:52jannauyes, there some limitations on what surfaces can do. I thought so far that surface 0 was special and had only limted support. We weren't using it because it caused issues w.r.t. to color space handling
09:54chadmedwhat issues did you see with surface 0? i havent noticed anything
09:54chadmedmacos uses it for the desktop/fullscreen stuff
09:55kaazoo has quit [Quit: Leaving.]
09:55pb17 has joined #asahi-gpu
09:58jannauchadmed: not me but marcan, see 312dc146eb0c6cdc9 but that doesn't appear to be an issue anymore
09:58chadmedyeah ik about the comments/commit message, but ive never actually noticed it being a problem
09:58chadmedmight be a leftover from 12.x
10:03adambenmoussa[m] has joined #asahi-gpu
10:07chadmedhm okay, surface 0 as primary, surface 1 as overlay and surface 3 as _cursor_ works
10:09chadmeddo we still care somewhat about x11 compat?
10:10lebakassemmerl[m] has joined #asahi-gpu
10:24okb has joined #asahi-gpu
10:24Bjarne[m] has joined #asahi-gpu
10:26chaos_princesswe should care enough to not deprecate and remove it in a single cycle
10:30yurix has quit [Ping timeout: 480 seconds]
10:31jannauI care enough not to break it. "X11 is not supported" was always meant as we do not care about improving the the experience on X11 or catering to requierements only X11 has
10:31chadmedthis would also break other wayland compositors that dont do sw cursor fallback anyway
10:32jannauthat said I do not see what in dcp could break X11
10:32chadmedwhen i added the cursor plane the first time, xorg would cause dcp to crash immediately
10:33jannauare you thinking of adding a cursor plane? I don't think we should do that
10:33chadmedif we want this to work eventually, we dont have a choice unless we teach every single compositor to never ever ever try to put anything on top of the plane which has the cursor on it
10:34chadmedthat seems to be what was breaking a lot of stuff, seems like one of the rules is "the smallest rectangle goes on top" or some variant of that
10:35chadmedkwin kept trying to put window surfaces at z=3 and move the cursor down to z=2 for some reason for example
10:35jannauthis is not an x11 issue but doing so would make dcp a broken KMS driver
10:36chadmedno i know _this_ is not an x11 issue, the x11 issue was dcp crashing immediately with a plane set to cursor
10:36jannauif that's a dcp requirement it's incompatible with KMS planes
10:38chadmedbecause overlays are supposed to be fungible?
10:38jannaudcp could reject the commit in atomic_check but if user space intends to render the cursor below the overlay dcp can not revert the order
10:41tabbytab10[m] has joined #asahi-gpu
10:41jannauif the overlay is opaque we could just not render the cursor when it is hidden but we probably run into the same issues as with display edges if it is partially occluded
10:42chadmedyes and this also causes issues when the cursor is not occluded
10:42jannauwe need explicit tests with igt or my overlay sample app before I'm happy to enable overlays
10:43jannauif the cursor is not occluded it can be rendered above the overlay
10:43chadmedtell that to kwin :P
10:44chadmedthis is the problem
10:44chadmedi am reasonably sure that some of this is kwin being pathological wrt overlays given that it is very new code and still WIP
10:45chadmedlike if its picking up an overlay for use with the cursor it should know to keep it at the highest z in most circumstances, but it just doesnt
10:48jannaukwin can do what it likes. I meant dcp can reorder the planes/surfaces as it likes if they don't overlap
10:49jannauwhy would the cursor be on top most plane?. hiding the cursor over video is common
10:50chadmedwould you then not just drop the fb object/make the plane invisible? why would you drop it to a lower plane and keep rendering it underneath
10:51chadmedkwin does the former when it hides the cursor
10:54jannaufor the partially occluded case, to keep the user-space code simpler. you can't really argue against valid API use regardlessly of how stupid it looks
11:01yurix has joined #asahi-gpu
11:02pulvite[m] has joined #asahi-gpu
11:14lacamar[m] has joined #asahi-gpu
11:16chadmedthis might actually be more kwin than dcp, idk
11:16chadmedthe exact same behaviour with one plane exposed as cursor is totally fine, even with mutable zpos
11:16chadmedits just when both are set to overlay
11:17chadmedthis is stupid
11:58pb17 has quit [Ping timeout: 480 seconds]
12:00OliverBestmann[m] has joined #asahi-gpu
12:03BriannaBabiack[m] has joined #asahi-gpu
12:08pb17 has joined #asahi-gpu
12:11okb has quit [Ping timeout: 480 seconds]
12:27yurix has quit [Ping timeout: 480 seconds]
12:45CoZo[m] has joined #asahi-gpu
13:00flokli has quit [Quit: WeeChat 4.8.1]
13:01flokli has joined #asahi-gpu
13:05yurix has joined #asahi-gpu
13:15pb17 has quit [Ping timeout: 480 seconds]
13:29pb17 has joined #asahi-gpu
13:35JamesCalligeros[m] has joined #asahi-gpu
13:38Dcow[m] has joined #asahi-gpu
13:40Felix[m]1 has joined #asahi-gpu
13:44okb has joined #asahi-gpu
13:48pas[m] has joined #asahi-gpu
13:49ants[m] has joined #asahi-gpu
13:51zackng[m] has joined #asahi-gpu
13:57FiniteCode[m] has joined #asahi-gpu
13:59Mary has quit [Quit: .]
13:59Mary has joined #asahi-gpu
14:02Mary has quit []
14:02Mary has joined #asahi-gpu
14:04neobrain[m] has joined #asahi-gpu
14:20CME has quit [Ping timeout: 480 seconds]
14:21CME has joined #asahi-gpu
14:23c10l has quit [Quit: Bye o/]
14:23c10l has joined #asahi-gpu
15:07pb17 has quit [Ping timeout: 480 seconds]
15:27pb17 has joined #asahi-gpu
16:44rosalyntg_ has quit [Remote host closed the connection]
16:46rosalyntg has joined #asahi-gpu
16:51pb17 has quit [Ping timeout: 480 seconds]
16:54pb17 has joined #asahi-gpu
17:14okb has quit [Ping timeout: 480 seconds]
17:22c10l has quit [Quit: Bye o/]
17:22c10l has joined #asahi-gpu
18:18Mary has quit [Quit: .]
18:18Mary has joined #asahi-gpu
18:19balrog has joined #asahi-gpu
18:20Mary has quit []
18:20Mary has joined #asahi-gpu
18:23balrog_ has quit [Ping timeout: 480 seconds]
18:23Mary has quit []
18:23Mary has joined #asahi-gpu
18:28okb has joined #asahi-gpu
18:29c10l has quit [Quit: Bye o/]
18:51okb has quit [Ping timeout: 480 seconds]
19:54rivendell has quit [Remote host closed the connection]
20:01Sammy[m]1 has joined #asahi-gpu
20:17Mary has quit [Quit: .]
20:18Mary has joined #asahi-gpu
21:21okb has joined #asahi-gpu
21:27karolherbst has quit [Read error: Connection reset by peer]
21:57karolherbst has joined #asahi-gpu
22:13chadmed has quit [Quit: Konversation terminated!]
22:13chadmed has joined #asahi-gpu
22:37okb has quit [Read error: Connection reset by peer]
22:48karolherbst has quit [Remote host closed the connection]
22:48karolherbst has joined #asahi-gpu
23:07picnoir has quit [Ping timeout: 480 seconds]
23:08c10l has joined #asahi-gpu
23:11picnoir has joined #asahi-gpu
23:28pb17 has quit [Ping timeout: 480 seconds]
23:37meze_ has quit [Remote host closed the connection]
23:55pb17 has joined #asahi-gpu
…