02:43 benjaminl: is a correct reading of the view-local dependencies section of the vulkan spec that a view-local subpass dependency actually defines separate execution/memory dependencies for each view
02:43 benjaminl: and so sequences with viewmasks like 0b101 -> 0b010 -> 0b101 wouldn't create a dependency chain between the first subpass and the last subpass
02:44 benjaminl: (while there would be a dependency chain without VIEW_LOCAL_BIT)
05:32 mareko: jenatali: any idea why d3d12 has assertion failures with this MR? https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/31942 - https://gitlab.freedesktop.org/mesa/mesa/-/jobs/67264066
05:35 jenatali: mareko: Not offhand, and I won't be able to take a look this week since it's Thanksgiving :(
05:35 jenatali: The I/O is the messiest part of that compiler for sure
05:35 jenatali: I assume https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/32047 doesn't fix it?
09:47 tursulin: airlied, sima: fyi there will be no drm-intel-next-fixes pull request this week, simply due absence of fixes
09:49 airlied: tursulin: thanks for letting me know
09:50 airlied: also I'll be doing any final rc1 pulls on my Friday morning (everyone's Thursday evening) and if anyone misses it'll be rc2
10:45 tarceri: dcbaker: yes please ignore that patch for backporting
12:47 Company: dj-death: random thing I noticed: on my Intel, GL_UNIFORM_BUFFER_OFFSET_ALIGNMENT is 32 and VkPhysicalDeviceLimits.minUniformBufferOffsetAlignment is 64. I was assuming those 2 numbers would be the same?
12:47 Company: I just did a cross-check to make sure I'm querying the same limit in both and now I'm confused
13:29 dj-death: Company: yeah, that could likely be reduced
13:29 Company: want me to file an issue about it?
13:31 dj-death: Company: sure
13:31 dj-death: Company: a0de2e0090535bd49b70c52917e7bdab628d354a is the change, it doesn't really explain why
13:31 dj-death: I mean cache issues
13:31 dj-death: but those caching issues would more likely relate to buffer allocation
13:32 dj-death: having them aligned to 64B is importatnt
13:32 dj-death: otherwise different fixed function can step over each other's data
13:58 davestevenson: Asking again. Quick query for drm maintainers. I'm about to merge the vc4 series https://lore.kernel.org/linux-arm-kernel/20241025-drm-vc4-2712-support-v2-0-35efa83c8fc0@raspberrypi.com/. Do the dtbinding changes go through drm or dt tree? I'm still learning the processes.
14:04 Company: dj-death: it's https://gitlab.freedesktop.org/mesa/mesa/-/issues/12210 now
14:06 pinchartl: davestevenson: DT bindings changes that are bundled with DRM drivers can go through the DRM tree, once you get an ack from the DT maintainers
14:10 davestevenson: Thanks Laurent. I have the acks from DT maintainers, so I believe I'm good to go.
14:54 dj-death: Company: saw it
16:28 dcbaker: tarceri: thanks, will do