#asahi-gpu · 2026-11-07 05:17
chadmedhttps://asahilinux.org/2026/11/progress-report-7-3/#asahi-gpu · 2026-08-29 04:18
chadmed(for interchange)#asahi-gpu · 2026-08-29 04:18
chadmedohh the luma plane has 32 bytes of metadata per tile because there's 4 16x16 compression subtiles lol#asahi-gpu · 2026-08-25 11:32
chadmedmy only thinking is that its a remnant of some sort of apple tv specific format that predates both AGX and Interchange#asahi-gpu · 2026-08-25 11:31
chadmedand then i remembered i own 4 apple tvs :P #asahi-gpu · 2026-08-25 11:31
chadmedthat was my thinking, and i thought "no that sounds stupid when have apple ever done anything like an htpc" #asahi-gpu · 2026-08-25 11:28
chadmedand i still have no idea what an "HTPC Surface" is#asahi-gpu · 2026-08-25 11:28
chadmeddoes seem like it would Just Work(tm) with the correct surface metadata though #asahi-gpu · 2026-08-25 11:28
chadmedah well interchange will suffice for now i suppose, not like macos ever does this anyway#asahi-gpu · 2026-08-25 11:27
chadmedi dont think im setting up the gpu tiling layout properly, no matter what i do iomfb complains about the plane size being too small for whatever sanity check it's calculated itself#asahi-gpu · 2026-08-23 06:22
chadmedoh i think i got the description of interchange wrong in the dcp branch. are the tiles _and_ subtiles/pixel data are both raster order? the difference from AGX tiling confused me#asahi-gpu · 2026-08-22 12:32
chadmedbut yes there is no enablement work for new hardware here, this is for direct scanout support on machines we already support#asahi-gpu · 2026-08-22 12:32
chadmedi presume this will continue working on m3 and up#asahi-gpu · 2026-08-22 11:57
chadmed(fyi for oliver too)#asahi-gpu · 2026-08-22 11:57
chadmedthis branch doesnt have the hacks in it and will only use interchange where we expect it to be used#asahi-gpu · 2026-08-22 11:56
chadmedim not 100% if the vulkan stuff is done correctly or if there's a better way to make that sane#asahi-gpu · 2026-08-22 11:56
chadmedhttps://gitlab.freedesktop.org/chadmed/mesa/-/tree/asahi/interchange #asahi-gpu · 2026-08-22 11:56
chadmedwith that, there are no regressions in vk or gles/gl cts vs main#asahi-gpu · 2026-08-22 11:55
chadmedokay with interchange hacked in as the driver's default/preferred format when twiddling and compression are allowed, the only thing that seems to be 100% broken are formats with blocksize > 4 and z/s buffers but i've banned array and compressed images too for safety#asahi-gpu · 2026-08-20 10:08
chadmedso the condition is ail_can_compress && !util_is_depth_or_stencil#asahi-gpu · 2026-08-20 10:08
chadmedi should clarify that the above means in addition to the limitations already in place for compressible/twiddled images#asahi-gpu · 2026-08-20 10:02
chadmedquestion is do we still want to limit it to _just_ scanout/shared buffers (i think we do/should)#asahi-gpu · 2026-08-20 10:01
chadmedseems like the only thing interchange cant do is depth/stencil#asahi-gpu · 2026-07-01 08:59
chadmedM3 progress, Apple bugs, and more! https://asahilinux.org/2026/06/progress-report-7-1/#asahi-gpu · 2026-04-26 09:22
chadmedNew progress report: https://asahilinux.org/2026/04/progress-report-7-0/#asahi-gpu · 2026-03-08 11:17
chadmedthis is stupid#asahi-gpu · 2026-03-08 11:16
chadmedits just when both are set to overlay#asahi-gpu · 2026-03-08 11:16
chadmedthe exact same behaviour with one plane exposed as cursor is totally fine, even with mutable zpos#asahi-gpu · 2026-03-08 11:16
chadmedthis might actually be more kwin than dcp, idk#asahi-gpu · 2026-03-08 10:55
chadmedyeah#asahi-gpu · 2026-03-08 10:51
chadmedkwin does the former when it hides the cursor #asahi-gpu · 2026-03-08 10:50
chadmedwould 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#asahi-gpu · 2026-03-08 10:45
chadmedlike 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#asahi-gpu · 2026-03-08 10:44
chadmedi am reasonably sure that some of this is kwin being pathological wrt overlays given that it is very new code and still WIP#asahi-gpu · 2026-03-08 10:44
chadmedthis is the problem#asahi-gpu · 2026-03-08 10:43
chadmedtell that to kwin :P #asahi-gpu · 2026-03-08 10:42
chadmedyes and this also causes issues when the cursor is not occluded#asahi-gpu · 2026-03-08 10:38
chadmedbecause overlays are supposed to be fungible? #asahi-gpu · 2026-03-08 10:36
chadmedno i know _this_ is not an x11 issue, the x11 issue was dcp crashing immediately with a plane set to cursor#asahi-gpu · 2026-03-08 10:35
chadmedkwin kept trying to put window surfaces at z=3 and move the cursor down to z=2 for some reason for example#asahi-gpu · 2026-03-08 10:34
chadmedthat 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#asahi-gpu · 2026-03-08 10:33
chadmedif 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#asahi-gpu · 2026-03-08 10:32
chadmedwhen i added the cursor plane the first time, xorg would cause dcp to crash immediately#asahi-gpu · 2026-03-08 10:31
chadmedthis would also break other wayland compositors that dont do sw cursor fallback anyway#asahi-gpu · 2026-03-08 10:31
chadmedyeah#asahi-gpu · 2026-03-08 10:09
chadmeddo we still care somewhat about x11 compat?#asahi-gpu · 2026-03-08 10:07
chadmedhm okay, surface 0 as primary, surface 1 as overlay and surface 3 as _cursor_ works #asahi-gpu · 2026-03-08 09:58
chadmedmight be a leftover from 12.x#asahi-gpu · 2026-03-08 09:58
chadmedyeah ik about the comments/commit message, but ive never actually noticed it being a problem #asahi-gpu · 2026-03-08 09:54
chadmedmacos uses it for the desktop/fullscreen stuff