#asahi-dev · 2026-08-16 09:25
chadmedi dont think theres a good solution to the codec shared gpio problem. the best(tm) solution would be to use the new-ish shared reset controller api which is specifically for this#asahi-dev · 2026-08-15 02:22
chadmedsince that was the original behaviour and it got changed before being merged#asahi-dev · 2026-08-15 02:21
chadmedim not sure what to do about the shared gpio. i dont think upstream will want the shared-gpio-proxy behaviour to change to only requiring a vote when trying to deassert the "default value" #asahi-dev · 2026-08-15 00:02
chadmedmaking the cursor plane linear-only fixes all three surfaces on t600x and t8112#asahi-dev · 2026-08-15 00:01
chadmedand once again as per usual the solution is "do what macos does" #asahi-dev · 2026-08-14 12:14
chadmedare swaps even slower with interchange? theres no way....#asahi-dev · 2026-08-14 12:14
chadmedwhat are we even getting wedged on here#asahi-dev · 2026-08-14 12:14
chadmedapple-dcp 231c00000.dcp: RTKit: Message (id=1): pcc firmware: mprintf_pcc2d_ltm.c:370 isr_inbox0_not_empty() -- #asahi-dev · 2026-08-14 10:41
chadmednothing else will make it crash#asahi-dev · 2026-08-14 10:41
chadmedon t8112 i can get it to reliably crash when glmark2-wayland is on surface 1 and kwin clears the cursor surface#asahi-dev · 2026-08-14 10:40
chadmedthis may still be something to do with mesa. the frequency of crashes can be influenced by when we allow interchange in agx_pipe #asahi-dev · 2026-08-14 10:15
chadmedwas anyone able to test the interchange stuff with mutter? im still getting occasional dcp crashes with certain things when all three surfaces are in use (they do not happen when interchange is disabled)#asahi-dev · 2026-08-14 07:58
chadmedor 1024x1024 as the png suggests#asahi-dev · 2026-08-14 07:58
chadmedwas the input buffer a 128x128 red blob for this image? #asahi-dev · 2026-08-14 07:43
chadmeddo you have the raw binary handy?#asahi-dev · 2026-08-14 07:42
chadmedbut also if the whole image is one solid colour i would also expect no pixel data and just metadata#asahi-dev · 2026-08-14 07:42
chadmed128 // 16 = 8 so i expect 8 obvious tile areas not 16 unless i misunderstand the format#asahi-dev · 2026-08-14 07:27
chadmedthe actual pixel data? i can certainly try#asahi-dev · 2026-08-13 12:08
chadmedsigh all consumers now have to agree on both assert and deassert#asahi-dev · 2026-08-13 11:31
chadmedthe i2c bus is unhappy#asahi-dev · 2026-08-13 11:30
chadmedhmm it doesnt quite work right on suspend/resume#asahi-dev · 2026-08-13 11:20
chadmedyou might even be able to see a tiling pattern#asahi-dev · 2026-08-13 11:20
chadmedrender it into linear space, if its mostly black/null with some coloured specks in there then it's a compressed fb#asahi-dev · 2026-08-13 11:19
chadmedyeah ive already sorted it, it was just something i didnt expect reading some of the commit descriptions#asahi-dev · 2026-08-13 11:18
chadmedother than hope and pray that theres a fourcc in there that matches a dcp one, but i doubt it#asahi-dev · 2026-08-13 11:18
chadmedmaybe? i dont know what id be looking at or for though :P#asahi-dev · 2026-08-13 11:07
chadmed-.-#asahi-dev · 2026-08-13 11:07
chadmed"Selected by: ARCH_QCOM"#asahi-dev · 2026-08-13 11:03
chadmedi need to see if the gpio-shared-proxy thing will work for us. looks like it should#asahi-dev · 2026-08-13 10:57
chadmedthe i2s stuff that got missed last time will be in 7.3 so that pretty much covers all the non-macaudio prerequisites#asahi-dev · 2026-08-13 10:57
chadmedactually 7.4 is unrealistic #asahi-dev · 2026-08-13 10:56
chadmedi might be able to get most of the remaining audio stuff upstream by 7.4 or 7.5#asahi-dev · 2026-08-12 11:50
chadmedinteresting#asahi-dev · 2026-08-12 11:49
chadmedoooh lets go#asahi-dev · 2026-08-12 09:47
chadmedwe'll see, but worst case scenario we just limit those to linear. not sure we know how to get avd to spit out interchange anyway #asahi-dev · 2026-08-12 09:46
chadmedhm, i thought it was just the luma tiles being bigger#asahi-dev · 2026-08-12 09:44
chadmedproblem for another day#asahi-dev · 2026-08-12 09:43
chadmedyeah#asahi-dev · 2026-08-12 09:40
chadmeddoes igt stress test modifiers?#asahi-dev · 2026-08-12 09:39
chadmedgstreamer kmssink always uses linear framebuffers and mpv's drm vo can only output rgb#asahi-dev · 2026-08-12 09:39
chadmedim pretty sure the kernel branch i linked should Just Work with ycc surfaces too but i've not got an easy way to test that#asahi-dev · 2026-08-12 09:37
chadmedwell at least interchange works on all supported iosurfaces so im happy with that#asahi-dev · 2026-08-12 09:37
chadmedi just saw that too :D #asahi-dev · 2026-08-12 09:28
chadmediirc mutter does#asahi-dev · 2026-08-12 09:28
chadmedkwin doesnt expose a linux-dmabuf tranche with dcp as the target#asahi-dev · 2026-08-12 09:06
chadmedive updated mesa to do the right thing#asahi-dev · 2026-08-12 08:47
chadmedok turns out it was neither of those things, it was allowing all scanout buffers to use interchange. i guess we have to ensure that the buffer has a valid pixel format before we allow it or something#asahi-dev · 2026-08-12 08:37
chadmed(cc oliver)#asahi-dev · 2026-08-12 08:37
chadmedif mutter also explodes horribly then its probably the former#asahi-dev · 2026-08-12 08:36
chadmedi suspect theres either something wrong with doing interchange on >2 surfaces, or kwin does weird things with the framebuffers that dcp does not like