#asahi-dev · 2026-11-15 18:21
sofusyeah in 90 deg steps#asahi-dev · 2026-11-15 18:18
leiosofus: does the MSR thing also do rotation?#asahi-dev · 2026-11-15 15:13
jannausingle full framebuffer layer doesn't appear to do blending. so that explains why XRGB works show_notch=1#asahi-dev · 2026-11-15 14:27
jannauyes, it's identical if I use surface 1,2 instead of 0,1#asahi-dev · 2026-11-15 13:55
jannauit is not, surface 0 has a noticeable different color but iirc we already arrived at the conclusion that surface 0 has special color space conversion properties (maybe just something set by iboot)#asahi-dev · 2026-11-15 13:08
jannau14.x or M3 specific#asahi-dev · 2026-11-15 13:07
chadmedoh ok this is also 14.x specific, sorry i misunderstood this morning #asahi-dev · 2026-11-15 12:59
jannauthis is j514s/t6030 with 14.8.3, I don't see how this could be display related. dcp produces everything the display gets. I'll test on t6020 but I'd be surprised if it shows the same behavior. at least the single premultiplied surface over opaque bg works as expected#asahi-dev · 2026-11-15 12:41
chadmedi tested alpha stuff on a dell 2721dgf and j415 and didnt notice any of this#asahi-dev · 2026-11-15 12:41
chadmedcan you repro on another machine or with an external display? is your screen busted?#asahi-dev · 2026-11-15 12:40
chadmedlol wtf#asahi-dev · 2026-11-15 12:37
jannauwtf, this is depedent on the dst rectangle size or layer? I have two identical surfaces (premultiplied, same format, color space, xfer func), one for top 74 pixels (notch) and one for everything below. the same pixel values in the buffer are rendered differently#asahi-dev · 2026-11-15 12:30
chadmedtheres no rush #asahi-dev · 2026-11-15 12:29
MichaelReeves[m]eh im tired ill finish tomorrow#asahi-dev · 2026-11-15 12:21
MichaelReeves[m]ok so turns out I had some time before bed so think I've finished getting rid of Frontend class now (and did it a nicer way), just testing then will push#asahi-dev · 2026-11-15 12:15
jannaumostly annoyed that it almost works and I fear it might be firmware "bug" which could cause trouble with premult alpha in general#asahi-dev · 2026-11-15 12:10
sofusno worries, i was just wondering#asahi-dev · 2026-11-15 12:09
chadmedi think part of the problem is that the gpu driver has busted yuv support in general so id like that to get fixed first #asahi-dev · 2026-11-15 12:09
chadmedi havent had a chance to look at it sorry. i have been extremely unwell and very busy at work lately#asahi-dev · 2026-11-15 12:08
sofuswhich i find kinda confusing, why even do interchange if its just converted to normal rgba?#asahi-dev · 2026-11-15 12:07
sofusi think macos might convert surfaces using the scaler#asahi-dev · 2026-11-15 12:06
sofuschadmed: how did it go with gpu interchange#asahi-dev · 2026-11-15 11:18
chadmedre that, do we actually need to support it? was there something about the patch to work around xrgb being required that you didnt like?#asahi-dev · 2026-11-15 11:15
jannauI'd love to use it in the display driver to support XRGB (not really) #asahi-dev · 2026-11-15 11:14
jannaumain use case is video / display related so VA VPP is a good fit#asahi-dev · 2026-11-15 11:13
jannauyeah, probably nothing in the direction of desktop software which is written with x86 in mind anyway#asahi-dev · 2026-11-15 11:06
sofusyup. Even there, i dont think most vaapi-clients use vpp#asahi-dev · 2026-11-15 11:05
chaos_princessso i guess this can only be used under the hood in our vaapi driver#asahi-dev · 2026-11-15 11:04
sofusgstreamer as always, nothing else i think#asahi-dev · 2026-11-15 11:03
chadmednothing desktop-grade that i know of #asahi-dev · 2026-11-15 11:03
chaos_princessis there software that actually uses those kinds of v4l devices?#asahi-dev · 2026-11-15 10:13
sofusR noted that the m1 used different registers so i dont expect that to work. I tested on t60{2,3}0#asahi-dev · 2026-11-15 10:12
sofusi pushed some changes to the asahi-va-driver in case anyone wants to test it#asahi-dev · 2026-11-15 10:10
sofusiirc M2Scaler#asahi-dev · 2026-11-15 10:08
jannauisn't apple's kernel driver also named *Scaler*#asahi-dev · 2026-11-15 10:08
JamesCalligeros[m]Thats what i thought too :p#asahi-dev · 2026-11-15 10:08
MichaelReeves[m]maybe not the best acronym to use for a peripheral on an arm device#asahi-dev · 2026-11-15 10:08
sofusfair, the 'R' is the most confusing for me#asahi-dev · 2026-11-15 10:08
MichaelReeves[m]it makes me think of MSR instruction in arm#asahi-dev · 2026-11-15 10:07
JamesCalligeros[m]Sometimes their naming convention is just nonsense#asahi-dev · 2026-11-15 10:06
sofusMichaelReeves[m]: yeah, its kinda weird that the adt's have scaler as the node name instead then right?#asahi-dev · 2026-11-15 10:06
sofusoh nice! That should not be to hard to do#asahi-dev · 2026-11-15 10:04
jannausofus: could useful for vaapi video processing. vlc currently breaks with vaapi because it expects to be able to do nv12 to planar yuv#asahi-dev · 2026-11-15 10:04
MichaelReeves[m]guessing here from skimming but [Memory, Media, Matrix] Scaler & [Rotater, Remapper] seem like the possibilities #asahi-dev · 2026-11-15 09:59
sofusbtw does anyone know what MSR stands for?#asahi-dev · 2026-11-15 09:59
sofusnot sure how usefull it is on linux. But very cool hw and fun to re :)#asahi-dev · 2026-11-15 09:58
sofusmy latest obsession https://github.com/sofus13/linux/tree/msr-wip#asahi-dev · 2026-11-15 09:36
MichaelReeves[m]sorry, I got busy, I'll have to finish getting rid of it tmrw#asahi-dev · 2026-11-15 06:59
MichaelReeves[m]yeah true, I'll get rid of it, thanks#asahi-dev · 2026-11-15 06:22
chadmedthe web one for example is going to already insert another middle layer, being the python-side "server" between the core and the web interface