00:01 TMM: I'm currently running a rocm workload on both gpus simultaneously, logged into X, with a 4K video playing as well... And it's fine too
00:01 TMM: I'm pretty confused :)
00:01 LaserEyess: a gpu reset is always a kernel bug, but it's probably worth reporting this to gnome, too
00:02 TMM: Okay, so redhat kernel bugzilla then?
00:02 LaserEyess: this is amdgpu's bug tracker https://gitlab.freedesktop.org/drm/amd/
00:04 TMM: Alright, thanks, I appreciate it
00:23 TMM: I filed this issue: https://gitlab.freedesktop.org/drm/amd/-/issues/2351
13:51 sravn:now understand why drm_crtc_helper did not have include files cleaden up. 214 files changed, 371 insertions(+), 18 deletions(-) - and not finished yet. It was one of these "less than an hour" cleaning up jobs that turns out to be much more
15:12 graphitemaster: Does anyone know why Vulkan uses half Z clip
15:12 graphitemaster: Like what rationale was there for that decision
15:16 jenatali: graphitemaster: From https://registry.khronos.org/OpenGL/extensions/ARB/ARB_clip_control.txt, looks like it's to match D3D and improve precision of depth values
15:23 graphitemaster: Thanks, makes sense.
15:36 emersion: danvet: how is atomic state restoration after VT switch supposed to work for DP-MST connectors?
15:37 emersion: specifically: user switches VT away, DP-MST connector appears, other DRM master sets KMS property i don't know about on that new connector, user switches VT back
15:37 emersion: and then i end up with garbage on that connector because it has a prop i don't understand
15:37 emersion: cc pq ^
16:58 danvet: emersion, hm
16:58 danvet: I guess it wont quite work so well in that case
17:00 danvet: emersion, otoh I think in practice it wont be a big problem
17:00 danvet: since most connector stuff is about configuring connector/sink things
17:00 danvet: so maybe you'll run in a more funny color space than expected
17:00 danvet: or with hdcp on
17:00 danvet: but none of these are an outright disaster
17:00 danvet: also, if they are technicolor lolz, then ... it's hotpluggable
17:01 danvet: user can just yank and replug to reset stuff :-)
17:01 danvet: real fix ofc would be logind keeping track of things a bit better I guess
17:02 danvet: or listen to hotplug events while you're not the active vt
17:02 danvet: but both of these are racy
22:34 DavidHeidelberg[m]: I seen that time to time fails tests outside hasvk, maybe just temporarily disable mesa:intel suite?