#
#asahi-gpu · 2023-08-09 20:27
alyssaer, this test is ETC2 EAC actually. but probably that too.
#asahi-gpu · 2023-08-09 20:26
alyssa132x132 ASTC 4x4 cube map with a full mipmap
#asahi-gpu · 2023-08-09 20:26
alyssaI think we have an ail bug O_O
#asahi-gpu · 2023-08-09 12:38
alyssanice work :)
#asahi-gpu · 2023-08-09 11:49
linaCool! ^^
#asahi-gpu · 2023-08-09 11:47
jannauand seems to work when booting with the display connected
#asahi-gpu · 2023-08-09 11:47
jannauflaky in the sense that only every 2nd connect works but is stable once a display was detected
#asahi-gpu · 2023-08-09 11:46
jannaulina: I should be able to get branch with dcpext and flaky usb-c dp-altmode running on the m2 pro mac mini
#asahi-gpu · 2023-08-09 11:44
linaOnly problem is I don't know how to run the "real" CTS without a video output for X/Wayland ^^;;
#asahi-gpu · 2023-08-09 11:40
linaalyssa: G14D passes the surfaceless dEQP/KHR tests up to GLES31 with the MR I sent, except KHR-GLES31.core.texture_buffer.texture_buffer_atomic_functions is flaky. This is on asahi/mesa with the deqp debug flag.
#asahi-gpu · 2023-08-09 11:27
jannauthat required addtional bits in VDM_BARRIER
#asahi-gpu · 2023-08-09 11:27
jannaulina: failing transform feedback tests were KHR-GLES31.core.draw_indirect.advanced-twoPass-transformFeedback*
#asahi-gpu · 2023-08-09 11:25
jannaulina: cluster coherency is based on macOS tracing on M1 Max https://oftc.irclog.whitequark.org/asahi-gpu/2023-07-05#32280709
#asahi-gpu · 2023-08-04 03:11
alyssaregardless I've scribbled enough math in my notebook that I now understand how coefficient registers work on AGX :~)
#asahi-gpu · 2023-08-03 22:30
alyssabut it's still not... quite right.
#asahi-gpu · 2023-08-03 22:26
alyssaseemingly without the swap. possibly I just misread the metal assembly
#asahi-gpu · 2023-08-03 22:26
alyssaok, yeah, that does it. mostly.
#asahi-gpu · 2023-08-03 22:15
alyssabut otherwise the same
#asahi-gpu · 2023-08-03 22:15
alyssaand you divide by W at the end as expected
#asahi-gpu · 2023-08-03 22:15
alyssadidn't comb through everything in full, but seemingly x/y are swapped in the perspective case (why?)
#asahi-gpu · 2023-08-03 22:01
alyssaoh what a delight, the coeffs are different if perspective is set. wheeee.
#asahi-gpu · 2023-08-03 21:42
alyssaeverything else is shader local changes
#asahi-gpu · 2023-08-03 21:42
alyssabindings unchanged
#asahi-gpu · 2023-08-03 21:42
alyssaunknown 3:0 set
#asahi-gpu · 2023-08-03 21:42
alyssaunk 2 : 0 -> 3
#asahi-gpu · 2023-08-03 21:42
alyssalooking at interpolateAtOffset again
#asahi-gpu · 2023-08-02 05:49
linaSo it just needs a m1n1 fix to ignore it in this case instead of fail
#asahi-gpu · 2023-08-02 05:48
linajannau: Ohh ok, looks like the Mac Studio doesn't actually use any of this for the GPU, I guess iBoot just throws that info in but the driver doesn't use it.
#asahi-gpu · 2023-08-02 01:39
linaNot just that one I guess, probably everything t600x?
#asahi-gpu · 2023-08-02 01:39
linajannau: Oof, that means I need to work out the fuses for that one too ;;
#asahi-gpu · 2023-08-02 01:27
alyssajannau: classic lol
#asahi-gpu · 2023-08-01 23:38
jannaulina: mac studio m1 ultra has 'afr-perf-states' in the 13.5 ADT but no "apple,afr-opp" property in the linux devicetree
#asahi-gpu · 2023-08-01 22:17
jannaulol, macOS just flashed a full magenta display on shutdown
#asahi-gpu · 2023-08-01 13:41
alyssaso https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/21685 is what we need after all
#asahi-gpu · 2023-08-01 13:41
alyssajannau: by decree of mareko, this is a mesa/st bug and not a driver bug
#asahi-gpu · 2023-08-01 13:22
alyssa:P
#asahi-gpu · 2023-08-01 13:22
jannauI've only noticed firefox and kate crashes
#asahi-gpu · 2023-08-01 13:22
alyssa=D
#asahi-gpu · 2023-08-01 13:22
alyssasweet!
#asahi-gpu · 2023-08-01 13:22
alyssanow whenever I have memory corruption in Mesa, now I know it can't possibly be the fault of my C code and I can blame the kernel
#asahi-gpu · 2023-08-01 13:21
alyssaomg, thanks guys
#asahi-gpu · 2023-08-01 13:19
marcanthat could explain basically any random problem, including the canberra stuff
#asahi-gpu · 2023-08-01 13:19
marcanoof, heh
#asahi-gpu · 2023-08-01 13:12
jannaumy understandig was that the buggy PER_VMA_LOCK causes random user space memory corruption
#asahi-gpu · 2023-08-01 13:11
jannaunot necessarily use-after-free but random memory corruption
#asahi-gpu · 2023-08-01 13:07
marcanthat would be fun
#asahi-gpu · 2023-08-01 13:06
marcanwait, you think that option triggered the Mesa UAF from the kernel?
#asahi-gpu · 2023-08-01 12:28
jannauat least that means we (most likely) do not have a use-after-free in mesa
#asahi-gpu · 2023-08-01 12:27
jannauargh, I can't read. CONFIG_PER_VMA_LOCK is not user selectable. I must have mixed it up with CONFIG_PER_VMA_LOCK_STATS
#asahi-gpu · 2023-08-01 11:56
alyssa_jannau__: fwiw https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24412

plus ›