#
…

2025-02-25

00:27dapselt^ has quit [Remote host closed the connection]
00:37dapsayemolt^ has joined #asahi-gpu
00:59pb17 has quit [Ping timeout: 480 seconds]
00:59m5zs7k_ has quit []
01:03m5zs7k has joined #asahi-gpu
01:23pb17 has joined #asahi-gpu
01:45chrisl has quit [Ping timeout: 480 seconds]
02:35pb17 has quit [Ping timeout: 480 seconds]
02:35opticron has quit [Read error: Connection reset by peer]
02:35opticron has joined #asahi-gpu
02:44pb17 has joined #asahi-gpu
02:57cow123 has quit [Ping timeout: 480 seconds]
03:19chrisl has joined #asahi-gpu
03:27chrisl has quit [Ping timeout: 480 seconds]
04:12chrisl has joined #asahi-gpu
04:20chrisl has quit [Ping timeout: 480 seconds]
04:43pb17 has quit [Ping timeout: 480 seconds]
04:48marvin24 has joined #asahi-gpu
04:51marvin24_ has quit [Ping timeout: 480 seconds]
04:57pb17 has joined #asahi-gpu
05:13chrisl has joined #asahi-gpu
05:21chrisl has quit [Ping timeout: 480 seconds]
05:56r0ni has quit [Read error: Connection reset by peer]
06:21chrisl has joined #asahi-gpu
06:21pb17 has quit [Ping timeout: 480 seconds]
06:29chrisl has quit [Ping timeout: 480 seconds]
06:35pb17 has joined #asahi-gpu
06:38chadmedalyssa: is this the DCP fb format we will need for planar scanout?
06:53john-cabaj has quit [Ping timeout: 480 seconds]
07:28chrisl has joined #asahi-gpu
07:36chrisl has quit [Ping timeout: 480 seconds]
08:00ece964311 has joined #asahi-gpu
08:04ece96431 has quit [Ping timeout: 480 seconds]
08:15pb17 has quit [Ping timeout: 480 seconds]
08:34pb17 has joined #asahi-gpu
09:22chrisl has joined #asahi-gpu
09:25dapsayemolt^ has quit [Ping timeout: 480 seconds]
09:30chrisl has quit [Ping timeout: 480 seconds]
09:54jmax has joined #asahi-gpu
09:58pb17 has quit [Ping timeout: 480 seconds]
10:15pb17 has joined #asahi-gpu
10:29chrisl has joined #asahi-gpu
10:37chrisl has quit [Ping timeout: 480 seconds]
11:14chadmed has quit [Quit: Konversation terminated!]
11:14chadmed has joined #asahi-gpu
11:36pb17 has quit [Ping timeout: 480 seconds]
11:37chrisl has joined #asahi-gpu
11:45chrisl has quit [Ping timeout: 480 seconds]
11:48pb17 has joined #asahi-gpu
12:22chrisl has joined #asahi-gpu
12:26dapsayemolt^ has joined #asahi-gpu
12:31chrisl has quit [Ping timeout: 480 seconds]
13:19bass_exe has quit [Ping timeout: 480 seconds]
13:27chrisl has joined #asahi-gpu
13:29bass_exe has joined #asahi-gpu
13:35chrisl has quit [Ping timeout: 480 seconds]
13:42pb17 has quit [Ping timeout: 480 seconds]
14:04pb17 has joined #asahi-gpu
14:23chrisl has joined #asahi-gpu
14:31chrisl has quit [Ping timeout: 480 seconds]
14:48john-cabaj has joined #asahi-gpu
15:14r0ni has joined #asahi-gpu
15:21pb17 has quit [Ping timeout: 480 seconds]
15:24chrisl has joined #asahi-gpu
15:30pb17 has joined #asahi-gpu
15:32chrisl has quit [Ping timeout: 480 seconds]
15:32alyssachadmed: no, that one is infinitely more messed up
15:32alyssaApple calls that one interchange-tiled
15:32alyssaand I cannot figure out any way to use it on Linux without, like, rewriting kwin
15:33alyssaso... "will" is "will not" ;)
16:11cow123 has joined #asahi-gpu
17:03chrisl has joined #asahi-gpu
17:11chrisl has quit [Ping timeout: 480 seconds]
17:29pb17 has quit [Ping timeout: 480 seconds]
17:45alyssaZLS *does* support twiddled, the bits are where you'd expect them to be
17:46alyssaso when apple says sparse z/s starts with m2 they mean that specifically their weird version of sparse does
17:46alyssawe probably could do sparse z/s on m1 with some shenanigans if we were so motivated
17:47alyssa..yeah actually we definitely could
17:47alyssaooh dEQP-GLES3.functional.texture.filtering.3d.sizes.3x7x5_linear_mipmap_linear: failing
17:49pb17 has joined #asahi-gpu
17:59chrisl has joined #asahi-gpu
18:07chrisl has quit [Ping timeout: 480 seconds]
18:15alyssagathering data for this layout is... annoying
18:15alyssasince apple only seems to use it with sparse, and apple's idea of sparse is crazy.
18:15alyssabut juryrigged something
18:15alyssafor 3x7x5, mipmapped 3D
18:16alyssaaccording to the sparse map that spans 5 pages, each specified at 0x200 offsets
18:17alyssain units of 4
18:17alyssawhich is what we expect
18:18alyssaPBE descriptors looks almost like we expect. only weird thing is page-aligned layers isn't set?!
18:18alyssaeven though obviously they are
18:19alyssaditto for the texture descriptor
18:19alyssafirst,last level = 0..2
18:19alyssaunk mipmappe dis set
18:27chrisl has joined #asahi-gpu
18:33alyssaoh dEQP-GLES3.functional.texture.filtering.2d.sizes.3x7_linear_mipmap_linear fails too
18:35chrisl has quit [Ping timeout: 480 seconds]
18:38alyssaoh, sure
18:38alyssamipmapping first divides, then rounds
18:38alyssainstead of rounding first then dividng
18:38alyssaso the hw is a bit more efficient than my original guess.
18:43dapsayemolt^ has quit [Remote host closed the connection]
18:45alyssathere we go
18:54alyssa...except for compressed
18:54alyssaso it shares the same weird special case as gpu-tiled
18:54alyssathat makes sense, I guess
18:55alyssawait what
18:55alyssathen why did we break dEQP-GLES31.functional.copy_image.non_compressed.viewclass_128_bits.rgba32f_rgba32ui.texture2d_to_cubemap
19:01alyssaserious wat
19:06alyssafound a fault for a 65x65x6 cubemap case with r32f .. ok
19:09alyssaso why is 65x65 different than 3x7?
19:11alyssaer, 65x63
19:12alyssai guess maybe major vs minor axis is the actual issue here
19:19alyssaI *think* it does round-then-divide to figure out where levels start
19:19alyssabut divide-then-round to figure out the actual tile sizes in the levels
19:20alyssathis is better than always doing round-then-divide, because at least within a level stuff is nice and packed so you access fewer cache lines --> faster perf
19:20alyssa(perf is the same as divide-then-round, the "correct" approach)
19:20alyssabut worse then divide-then-round for RAM usage since there's extra padding sometimes
19:20alyssabut that doesn't matter for perf since levels are cache line padded anyway
19:34alyssa...except for compressed?
19:34alyssaOK yes here we go
19:34alyssacts likes this now
19:35alyssa /* AGX: Even the exceptions have exceptions */
19:43alyssaok, gl cts passing now with twiddled images
19:43alyssavk, you're up next
19:48pb17 has quit [Ping timeout: 480 seconds]
19:53pb17 has joined #asahi-gpu
19:53WindowPain6 has joined #asahi-gpu
19:59WindowPain has quit [Ping timeout: 480 seconds]
19:59WindowPain6 is now known as WindowPain
20:18chrisl has joined #asahi-gpu
20:26chrisl has quit [Ping timeout: 480 seconds]
20:30jannauagx driver compiles again after replacing a ton stale rust dependencies with upstream changes or patches under active discussion
20:47chrisl has joined #asahi-gpu
20:55chrisl has quit [Ping timeout: 480 seconds]
21:08master91122 has quit [Quit: Connection closed for inactivity]
21:22chrisl has joined #asahi-gpu
21:30chrisl has quit [Ping timeout: 480 seconds]
21:30chrisl has joined #asahi-gpu
21:33pb17 has quit [Ping timeout: 480 seconds]
21:38chrisl has quit [Ping timeout: 480 seconds]
21:39alyssaok, after that... significant diversion
21:39alyssaI'm now confident in my understanding of the different tilings so https://lore.kernel.org/asahi/20250225-apple-twiddled-modifiers-v2-1-cf69729e87f6@rosenzweig.io/T/#u sent out
21:40pb17 has joined #asahi-gpu
21:42alyssajannau: amazing work :)
21:44jannaudoesn't work yet because reserved CPU memory shouldn't be handled in the same way as iomem
21:46svenjannau: \o/
21:47alyssajannau: hey progress is progress =D
21:47jannauin principle this should have all timed used of_reservedmem but that looks annoying to wrap for rust
21:50jannauI suppose this should be used by all other asc drivers as well instead of the the automatic mapping I implemented for dart
22:13cow123 has quit [Read error: Connection reset by peer]
22:22jannauand working with adding more hacks
22:28cow123 has joined #asahi-gpu
22:31chrisl has joined #asahi-gpu
22:39chrisl has quit [Ping timeout: 480 seconds]
23:04Calandracas_ has joined #asahi-gpu
23:07pb17 has quit [Ping timeout: 480 seconds]
23:07jannauless than 2.5k added rust abstractions left (excluding drm): https://paste.centos.org/view/325a8533
23:09jannauthis is after abstactions already under discussion on rust-for-linux
23:09pjakobsson_ has quit [Ping timeout: 480 seconds]
23:10Calandracas has quit [Ping timeout: 480 seconds]
23:10jannauI still have to replace the xarray abstractions
23:37pb17 has joined #asahi-gpu
23:38chadmed has quit [Quit: Konversation terminated!]
23:38chadmed has joined #asahi-gpu
23:46ypwn_ has joined #asahi-gpu
23:47ypwn has quit [Read error: Connection reset by peer]
23:52pjakobsson has joined #asahi-gpu
…