#
#asahi-gpu · 2026-02-08 13:14
jannaulet's do that. any specific reason for the non-pot surface width/height? CI config in mesa uses 256/256 consistently
#asahi-gpu · 2026-02-08 13:08
alyssa/dev/shm/caselist.txt
#asahi-gpu · 2026-02-08 13:08
alyssawith manual caselist - cat gles/aosp_mustpass/main/gles2-main.txt gles/aosp_mustpass/main/gles3-main.txt gles/aosp_mustpass/main/gles31-main.txt gles/khronos_mustpass/main/gles2-khr-main.txt gles/khronos_mustpass/main/gles3-khr-main.txt gles/khronos_mustpass/main/gles31-khr-main.txt gl/khronos_mustpass/4.6.1.x/gl46-main.txt gl/khronos_mustpass_single/4.6.1.x/gl46-khr-single.txt >
#asahi-gpu · 2026-02-08 13:08
alyssa--deqp-log-shader-sources=disable
#asahi-gpu · 2026-02-08 13:07
alyssaMESA_SHADER_CACHE_DIR=/dev/shm EGL_PLATFORM=surfaceless deqp-runner run --deqp ~/GLCTS/glcts --output output --timeout 60 --caselist /dev/shm/caselist.txt --skips /dev/shm/skips.txt --tests-per-group=1000 -- --deqp-surface-type=pbuffer --deqp-gl-config-name=rgba8888d24s8ms0 --deqp-surface-width=400 --deqp-surface-height=300 --deqp-log-images=disable
#asahi-gpu · 2026-02-08 13:07
alyssaperso I use one big merged caselist and a single invocation for all of gles3.2+gl4.6 cts (except for WSI)
#asahi-gpu · 2026-02-08 13:06
alyssajannau: you can make the toml a lot more compact because `glcts` secretly contains all the deqp-gles tests too. whether that's actually better is unclear. see https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/34960
#asahi-gpu · 2026-02-08 13:04
alyssajannau: serves me right for not running dEQP-EGL in years. oops :p
#asahi-gpu · 2026-02-08 13:03
jannauoliver: https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/39763 and update lamci.sh. test_gl will then execute the OpenGL CTS
#asahi-gpu · 2026-02-08 12:47
oliverhttps://files.narf.zone/42504d42556d0ff9/failures.csv
#asahi-gpu · 2026-02-08 12:27
oliver(or all non 32bbp formats with a non one block size)
#asahi-gpu · 2026-02-08 12:27
oliverthat one leads to more failures, it will crash for yuv formats.
#asahi-gpu · 2026-02-08 12:27
oliverhere is vulkan: https://files.narf.zone/43fb7f449ad6d6a0/0001-hk-add-support-for-interchange_compressed.patch
#asahi-gpu · 2026-02-08 12:27
oliverbut, like i said, the changes do not affect vulkan at all, it is just gl.
#asahi-gpu · 2026-02-08 12:11
jannauit finished on g14s without unexpected failures
#asahi-gpu · 2026-02-08 12:10
oliver(on mesa main, not rc)
#asahi-gpu · 2026-02-08 12:08
oliverhttps://files.narf.zone/27f02c5aa3e17d58/failures.csv
#asahi-gpu · 2026-02-08 12:07
olivertest_vk finished with some 42 failures on mesa 26.0.0-rc3. i have a g13s according to glxinfo. also my changes did not affect vulkan at all. i'll run it again with the changes applied to the vk driver too.
#asahi-gpu · 2026-02-08 10:54
oliverbest_modifier seems to be only used in the context of the "screen". when checking with ASAHI_MESA_DEBUG i see it only used for the surface, all the other textures allocated are either LINEAR or GPU_TILED ... which is fine imho
#asahi-gpu · 2026-02-08 10:47
jannauduh. I guess I'll finish the asahi OpenGL CI expectations now
#asahi-gpu · 2026-02-08 10:46
jannaudue to the placement in the best_modifier array I expect interchange to be used even when it shouldn't be. but yes when CTS passes it'll a good idea to break interchange on purpose to verify that it is exercised
#asahi-gpu · 2026-02-08 10:46
noopwafel(certainly not with vulkan CTS I guess, since this seems gallium-only?)
#asahi-gpu · 2026-02-08 10:43
jannaufor development it'll probably more comfortable to have a local CTS build so you can run single tests under a debugger
#asahi-gpu · 2026-02-08 10:42
noopwafelI mean, the question is, are you going to even hit this case with CTS
#asahi-gpu · 2026-02-08 10:41
jannauthe tests test everything. your mesa changes are responsible to allow only working use cases
#asahi-gpu · 2026-02-08 10:37
oliveri have one question: the way i understood mesa, INTERCHANGE should only be used for surface textures with my change, like with the "window" or the "desktop", everything else should still use GPU_TILED as before. is this covered when running those tests?
#asahi-gpu · 2026-02-08 10:37
oliverperfect, thank you
#asahi-gpu · 2026-02-08 10:31
jannauCTS runs have the potential to crash the GPU so do not run this with unsaved work in your desktop session
#asahi-gpu · 2026-02-08 10:30
jannauoliver: https://gitlab.freedesktop.org/jannau/lamci
#asahi-gpu · 2026-02-07 23:19
austriancodermaybe also use https://crates.io/crates/deqp-runner to make the run part easier
#asahi-gpu · 2026-02-07 22:07
jannauit is annoying to build and run. I have some scripts to use the container images mesa builds for their CI
#asahi-gpu · 2026-02-07 22:06
jannauvulkan and opengl CTS is at https://github.com/KhronosGroup/VK-GL-CTS/tree/main
#asahi-gpu · 2026-02-07 22:05
jannauthe hope is that CTS run will find if there are corner cases which can be triggered by the mesa drivers
#asahi-gpu · 2026-02-07 22:04
jannaualso possible that the implementation was simply not finished at the release deadline and so it's not used
#asahi-gpu · 2026-02-07 22:02
jannauit's possibly that that just affects some corner cases like specific GPU generations/revisions/firmware versions
#asahi-gpu · 2026-02-07 22:01
jannauputting DRM_FORMAT_MOD_APPLE_INTERCHANGE_COMPRESSED in front of agx_best_modifiers[] like in the MR might be enough. Based on tracing macos uses shaders when the GPU reads interchange textures. The obvious conclusion was that it can't read them directly
#asahi-gpu · 2026-02-07 21:55
chadmed_changes to mesa to get it to use interchange where possible
#asahi-gpu · 2026-02-07 21:54
oliverchanges to the cts?
#asahi-gpu · 2026-02-07 21:44
oliverWhat do you mean by the last part, i.e. what kind of changes would that be?
#asahi-gpu · 2026-02-07 21:44
oliverNo, sorry, I have no experience with any of this. If you can give me a few pointers, that would be great.
#asahi-gpu · 2026-02-07 21:25
jannauoliver: did you run OpenGL/Vulkan CTS on the mesa changes? Ideally with further changes to ensure the interchange format is used whenever it can be
#asahi-gpu · 2026-02-07 21:20
oliverI've created pull requests in mesa and asahi kernel for the compressed scanout.
#asahi-gpu · 2026-02-05 08:52
JamesCalligeros[m]on macos avd does output interchange format yuv framebuffers fwiw, im not at home rn but it was something like &4v0 for 4:2:0/nv12
#asahi-gpu · 2026-02-05 08:40
JamesCalligeros[m]We dont need to care about that until avd, and even then avd can just do normal nv* iiuc
#asahi-gpu · 2026-02-05 07:54
jannauwe currently do not have to care
#asahi-gpu · 2026-02-05 07:43
oliverwe would need to get a compressed yuv framebuffer from somewhere for it to matter. it is probably possible somehow, i bet macos does that with direct scanout of yuv video, maybe directly from some decoder coprocessor. but if we currently can not do it, we might not need to care right now.
#asahi-gpu · 2026-02-05 07:40
jannauI don't think mesa can but the firmware has yuv matrices. not sure if they are used for sampling or rendering to yuv
#asahi-gpu · 2026-02-05 07:33
olivercan we render to yuv on the gpu side of things?
#asahi-gpu · 2026-02-05 07:30
jannautile size depends on element size, 8-bit YUV has 32x32 tiles
#asahi-gpu · 2026-02-05 07:30
oliverthat is just an alias in the dcp firmware afaik

plus ›