07:50 airlied: okay just workgroup scope matrixes left, definitely the big boss
09:51 jani: tzimmermann: mlankhorst: ack for merging this via drm-intel? https://lore.kernel.org/r/20250919143418.3102605-2-nemesa.garg@intel.com
09:52 jani: the series is mostly about i915, and sima has acked it in general
10:20 karolherbst: So... I've worked on some range analysis opts to be able to extract constant addresses in address calcs more easily, it makes use of uub, but also adds its own passes (to set nuw flags and other analysis). Feedback welcomed: https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/36113
10:31 tzimmermann: jani, sure ack
10:32 jani: tzimmermann: thanks
12:54 tzimmermann: jani, how urgent is that overflow bug? should we revert or wait for the submitter?
12:55 jani: tzimmermann: looks like it's hitting a handful of CI tests across a lot of machines
12:56 jani: jsa1: what do you think ^
12:57 jani: tzimmermann: I guess the fix is not complicated, my main worry is annotating whatever we end up doing in a way that the broken commit doesn't propagate to stable alone
12:57 tzimmermann: jani, looks easy to fix, i'll send you a patch in a bit
12:57 jani: tzimmermann: thanks
13:07 Venemo: hey guys
13:07 Venemo: what would be a good way to trigger scaling in a drm driver?
13:23 pq: Venemo, I think Weston with a suitable video player could do. But something a la IGT might be less of a hassle to get up - if it has a test like that.
13:31 Venemo: pq: what would be a suitable video player? and why only weston?
13:31 pq: GNOME or KDE Plasma should do as well, but a lot more to install I suppose.
13:32 pq: mpv maybe? I'm not sure. It needs to deliver scanout-able dmabuf to the compositor while using wp_viewport for scaling rather than doing it itself.
13:35 Venemo: hm.
13:35 sarbes: Kodi should be able to do it as well.
13:36 pq: hence, if IGT has a test for it, that would be easier to use and not need any of Weston/GNOME/KDE nor a player.
13:41 Venemo: I'm trying to diagnose an issue where a user has a weird register config as if scaling was enabled, and causes issues for the user
13:41 Venemo: I'm trying to reproduce that but not sure how to
13:46 pq: The said desktops will try to push app buffers straight to KMS when possible, and that may involve setting up scaling. So any hardware-decoding video player or openGL or Vulkan app might trigger it.
13:48 pq: The decision is likely made frame by frame, falling back to GPU composition as necessary.
13:48 Company: opengl and vulkan apps will adjust their buffer size
13:48 pq: usually they do
13:48 Company: video players are a better choice I think
13:48 Company: playing a 720p video for example
13:49 Company: though you need to make sure that the video player uses direct wayland output (GStreamer with wWaylandsin
13:49 Company: *waylandsink)
13:49 pq: they are candidates for direct scanout, and the DE could decide to do something like desktop overview or window thumbnails or such, where the window is shown forced scaled.
13:49 Company: gtk-demo's video player would also work with the right videos
13:54 Venemo: the user reports that his entire desktop looks messed up, which I haven't been able to reproduce
13:54 Venemo: so there was no video player app in the picture
13:56 pq: Nothing comes to mind.
14:07 Venemo: pq: as far as I understand he even gets the issue in just a console
14:09 Venemo: so it makes me wonder not only how to repro that but also whether scaling works correctly on this hw in the first place
14:09 MrCooper: then I doubt it's scaling via CRTC properties; maybe it's implicit scaling due to the mode not matching the display's native resolution?
14:11 tzimmermann: jani, i posted a possible fix BTW
14:11 Venemo: MrCooper: you can find all the details here, including drm_info and all dce registers from umr from the user's system: https://gitlab.freedesktop.org/drm/amd/-/issues/1776#note_3105926
14:12 Venemo: MrCooper: the SCL0_SCL_* registers look suspicious to me, especially SCL0_SCL_TAP_CONTROL
14:14 Venemo: MrCooper: looking at the drm_info it looks like the framebuffer size matches the monitor resolution and also matches the current mode
14:15 Venemo: but maybe you can see something in there which I haven't noticed?
14:15 MrCooper: yeah, drm_info shows SRC_W/H property values matching CRTC_W/H, i.e. there's no scaling as far as the KMS API is concerned
14:15 Venemo: yes, I would think that too
14:15 MrCooper: you'll need to keep digging in the driver instead
14:16 Venemo: I am
14:16 Venemo: SCL_TAP_CONTROL is set by dce60_setup_scaling_configuration and is only non-zero when there is scaling
14:23 Venemo: MrCooper: is it possible that the vbios sets up the default mode with scaling enabled and then DC just forgets to clear this register?
14:24 MrCooper: sounds plausible
23:42 zmike: mareko: https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/37420
23:42 zmike: last one for the year, I promise
23:49 HdkR: I can see that promise bending already when something comes up that is running too slow.