00:57 DrNick: GPUs don't have unique identifiers
00:57 DrNick: and there's already a /dev/dri/by-path
06:03 daniels: GitLab will be going down shortly - but briefly - for an upgrade to 13.0 https://about.gitlab.com/releases/2020/05/22/gitlab-13-0-released/
09:42 EdB: pmoreau: which LLVM version do you use for clover on nouveau ?
09:43 EdB: I can see that with_gallium_radeonsi set min version to 8.0.0, if you use an newer version, I think we can get ride of some compat code
09:43 pmoreau_: EdB: The latest available on Arch, which is 10.0.0.
09:45 EdB: nice
09:47 EdB: this morning I tackle another CTS failure :)
09:54 EdB: pmoreau_: regarding !5038 maybe you can ping curro for a final review. It's been a long time since I code for clover, I don't know if I'm legitimate to merge stuff
15:48 udovdh: hello
15:48 udovdh: wgat happened here? https://pastebin.com/sc9Xievm
15:49 udovdh: played a 4K video with mpv
15:49 udovdh: hwdec=vdpau, vo=gpu
15:49 udovdh: played OK after the reboot
15:49 udovdh: kernel 5.6.14, git mesa etc on AMD Ryzen 5 3400G with Radeon Vega Graphics
15:54 udovdh: Even a a page fault : amdgpu 0000:0a:00.0: AMD-Vi: Event logged [IO_PAGE_FAULT domain=0x0002 address=0x800002300 flags=0x0000]
15:58 kisak: udovdh: maybe try https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4895 ?
16:00 udovdh: kisak, thanks for the tip, will this work on Vega?
16:00 udovdh: sdma timeouts do not happen often but still a bit too regularly
16:00 udovdh: if I would be able to reproduce the issue, would that help?
16:02 kisak: that hardware should be GFX9
16:02 bnieuwenhuizen: udovdh: yes, that should work on Vega
16:02 udovdh: thanks kisak , bnieuwenhuizen
16:02 udovdh: will test soon if that makes the situation solid
16:07 udovdh: (because I see no other issue currently)
16:09 udovdh: is there a bug I should add my log to?
16:10 udovdh: because https://gitlab.freedesktop.org/mesa/mesa/-/issues/2814 describes other situation than what I see
16:11 udovdh: Would https://gitlab.freedesktop.org/mesa/mesa/-/issues/2814#note_484837 do enough to test without patching?
16:13 udovdh: firmwarez https://pastebin.com/LZ5RAU0T
16:51 karolherbst: honestly.. there more I run into applications (games mainly) which change the resolution, I am wondering if we should just remove all resolution change logic from userspace and always go with the default mode or the highest available ones... and I am sure broken displays and adapters will foil this plan as I have a display doing 1680x1050 instead of 1920x1080 on a certain hub :/
16:52 karolherbst: or at least remove it from wayland/X and leave the kms bits
16:55 emersion: with wayland that's already the case
16:55 emersion: wayland clients can't change the resolution
16:55 karolherbst: cool
16:55 karolherbst: what if it's a X client though?
16:55 emersion: the buffers are scaled
16:56 karolherbst: ahh
16:56 emersion: ideally with the display scaler
16:56 karolherbst: maybe I need to move to wayland on my gaming system as well and see how it goes then
16:56 karolherbst: most annoying are the ones disabling displays
16:56 emersion: this part is pretty new, i think it's not in a released xwayland version yet
16:57 emersion: before that, the games were just centered on-screen
16:57 karolherbst: or.. even switching displays
16:57 emersion: oh yeah, that's so annoying
16:59 emersion: there are also games changing the display LUT…
17:01 karolherbst: ufff
17:02 karolherbst: yeah..
17:36 FurretUber: Hi, what does the dmesg error "i915 0000:00:02.0: [drm] *ERROR* Failed to disable qgv points (-6)
17:37 FurretUber: ...means? It is appearing around 6 times per second
17:40 ickle: it means that 20f505f2253106f695ba6fa0a415159145a8fb2a deserves a bug report
18:21 FurretUber: I did a report for this: https://gitlab.freedesktop.org/drm/intel/-/issues/1919
19:36 dinosomething: With something like let’s say a browser like chrome, and it’s got a bunch of graphical elements in it like say a video, a input box, images whatever... are all those things composited in memory? I’m asking because if in the context of an egpu, if it goes out to the egpu to render each element then it has to come back to chrome to be composited then go back out to the display
19:37 dinosomething: I’m trying to figure out if eg chrome is actually better to only use the integrated graphics since it has to composite in its own memory anyway
19:38 dinosomething: I’m sorry if that’s confusing lol. But I realized that a browser is also a compositor like eg mutter etc, and the same concerns about mutter using the egpu seem to also be applicable to a browser I would think