#asahi-dev · 2024-05-12 09:52
chadmedi tamed the beast! (kinda) #asahi-dev · 2024-05-12 08:53
chadmedsimple enough to recompile and check...#asahi-dev · 2024-05-12 08:53
chadmedyeah thats why i asked about 16.16. the arguments to FRAC_16_16 look the wrong way around too?#asahi-dev · 2024-05-12 08:49
chadmedeither way apple_plane_atomic_check doesnt appear to be doing what it's supposed to #asahi-dev · 2024-05-12 08:44
chadmed1/2 in 16.16 is ((1 << 16) / 2) right#asahi-dev · 2024-05-12 08:31
chadmedalso not sure why it tolerates some of these but then crashes on others#asahi-dev · 2024-05-12 08:31
chadmedthe fun thing is i can freely wiggle the cursor around with no issues until one of these swaps fails, then the planes swap and dcp treats the cursor as the primary plane until the next failed swap. not sure what is trying to make the fb 64x9 either since the cursor is not actually changing #asahi-dev · 2024-05-12 08:29
chadmedi suspect theres some horsin around with the "legacy" ioctls not calling some atomic helper function because this _only_ happens when setting up the plane as a cursor plane#asahi-dev · 2024-05-12 08:29
chadmedso the check just... isnt checking? #asahi-dev · 2024-05-12 08:29
chadmedIOMFB verify_scaling: layer 1[ARGB]: (64x32) -> (64x9) (min_scale=1/2, max_scale=4): illegal scaling requested#asahi-dev · 2024-05-12 07:10
chadmedunless it's just crashing for some other reason now#asahi-dev · 2024-05-12 07:10
chadmedjannau: that was the first thing i tried since it is the "correct" way to do it, no dice :p#asahi-dev · 2024-05-12 06:31
chadmedand rejecting the framebuffer in iomfb_flush doesnt seem to do anything? e.g. the text entry beam/resize cursors cause dcp to crash due to being too narrow in one dimension#asahi-dev · 2024-05-12 06:29
chadmedjannau: unfortunately it looks like planes are unaffected by the mode min and max. drm_info reports that the src rectangle can still be [0, UINT32_MAX] in both x and y#asahi-dev · 2024-05-11 23:39
chadmedon the other hand, the odd provenance of dcp probably makes it unique in not having been designed with cursors in mind perhaps?#asahi-dev · 2024-05-11 23:39
chadmedi wonder if something like this warrants a drm api. we surely cant be the only hardware that has restrictions like this, and with planes now being fungible...#asahi-dev · 2024-05-11 23:37
chadmedohhh the minimum surface size is 32x32 thats less insurmountable than i thought#asahi-dev · 2024-05-11 10:04
chadmedwell thats no good for us #asahi-dev · 2024-05-11 10:04
chadmedeugh and kwin is actively trying to do away with the distinction to match the universal plane api anyway #asahi-dev · 2024-05-11 09:29
chadmedthe only reason i added them in the first place was because MPV can't do ARGB and all i needed was to see if applications would pick up and use the plane. but then it just ended up working as it should :p#asahi-dev · 2024-05-11 09:25
chadmedlike if other applications arent going to behave and we cant assume that they will initialise the X channel, then i will just remove the XRGB formats from the list #asahi-dev · 2024-05-11 09:24
chadmedwell yeah thats what i was sort of asking. mpv's initialises things properly but if we cant rely on other applications doing the same then i assume we're just going to get a framebuffer-sized black square?#asahi-dev · 2024-05-11 09:22
chadmedxorg, kwin and weston all behave correctly wrt cursor. the plane is not being instantiated as a cursor plane so no compositor should be trying to use it like that#asahi-dev · 2024-05-11 09:20
chadmedno theres another comment in apple_drv.c that says overlays "must not" advertise XRGB at all#asahi-dev · 2024-05-11 09:19
chadmedwhich compositors try to use DRM_PLANE_TYPE_OVERLAY for cursors?#asahi-dev · 2024-05-11 09:15
chadmedlike if its going to break some other use case it's no trouble to just get rid of it of course#asahi-dev · 2024-05-11 09:13
chadmedjannau, alyssa: was there a technical reason dcp overlay planes shouldnt advertise support for/implement premultiplied RGB formats? ive been dicking around with mpv's osd/subtitle overlay and it all seems to work as intended even when in XRGB8888#asahi-dev · 2024-05-11 00:33
chadmedhehe were plain ARGB/ABGR formats ever tested#asahi-dev · 2024-05-10 13:06
chadmedrather than mess about with trying to pipe gstreamer's buffers into dmabuf or whatever#asahi-dev · 2024-05-10 13:06
chadmedi think therefore the easiest option will be to teach mpv's drm backend about planar formats#asahi-dev · 2024-05-10 13:06
chadmedez clap overlay planes work#asahi-dev · 2024-05-10 12:51
chadmedlet me see how mpv does it. it has a drm backend and already supports a bunch of drm formats, just not planar yuv#asahi-dev · 2024-05-10 12:50
chadmedyeah figured that from the blog posts :c#asahi-dev · 2024-05-10 12:35
chadmedit really annoys me that nothing in plasma has any of this hooked up yet#asahi-dev · 2024-05-10 12:33
chadmedi assume gstreamer can grab a drm device directly? #asahi-dev · 2024-05-10 12:03
chadmedfor context ive hooked up surf.has_planes and family and it works for our single-plane fb formats but i cant test it for planar yuv since nothing seems to want to draw a yuv framebuffer#asahi-dev · 2024-05-10 12:02
chadmedunless they just took that away between last time i tried and now for some reason#asahi-dev · 2024-05-10 12:02
chadmedsmh i cant remeber how i got weston to output YUV video from vlc to an overlay plane#asahi-dev · 2024-05-09 10:26
chadmeda lot of whacky symbols with "fence" in them :p#asahi-dev · 2024-05-09 10:25
chadmedthat seems to line up with whats in the bunched up call graph#asahi-dev · 2024-05-09 10:22
chadmedi made it on sysprof 44 which is packaged in ::gentoo#asahi-dev · 2024-05-09 10:19
chadmedof course#asahi-dev · 2024-05-09 10:19
chadmedso dunno what route your isp is taking#asahi-dev · 2024-05-09 10:19
chadmedusing telstra's looking glass i get a clean trace from their frankfurt pop to the aussiebb segment im on#asahi-dev · 2024-05-09 10:11
chadmedi guarantee my website is not worth your time :p#asahi-dev · 2024-05-09 10:00
chadmedhmm thats correct and it works from my mobile over 5G...#asahi-dev · 2024-05-09 09:57
chadmedit should ACK pings#asahi-dev · 2024-05-09 09:57
chadmedwhat ipv6 address do you get#asahi-dev · 2024-05-09 08:55
chadmedhttps://chadmed.au/res/kwin.syscap#asahi-dev · 2024-05-09 08:54
chadmedif im reading this sysprof capture correctly, the stutters coincide with bunched up calls to the gpu driver