07:36 danvet: mripard, )[PATCH] drm/sun4i: Constify static structs <- in case you've missed
10:38 shadeslayer: Hi, could someone assign https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6125 to marge bot?
10:42 MrCooper: done
10:42 shadeslayer: Thanks!
12:47 pq: Should userspace prepare for dynamically appearing and disappearing writeback connectors, just like it has to prepare for normal connectors?
15:29 ajax: hmph, no idr
15:30 ajax: anyone remember a benchmark from back in the day that would measure GL API dispatch speed?
15:31 ajax: i feel like i remember seeing this around like '05 but my archives are failing me
15:46 MrCooper: ajax: FWIW there's a drawoverhead test in piglit, probably not exactly what you're looking for though
15:50 ajax: yeah, not quite the thing. i'm looking for more like how many calls to glNoOpEXT can you retire in one second
16:52 Lyude: Hey airlied this might be a bit too far back to remember, but do you have any idea why we opted to only cache EDIDs for SST sinks on MST topologies (in drm_dp_mst_detect_port() ) and not for both SST sinks and legacy converters?
17:47 karolherbst: Lyude: btw, are you aware of the kexec regression from the backport?
17:50 karolherbst: ehh.. should be more precise.. I'd like to look into the nouveau kexec regression we have a bug for, but I was just wondering if anybody already knew about it
17:50 karolherbst: https://bugzilla.redhat.com/show_bug.cgi?id=1856149
17:51 karolherbst: I doubt it's just a ppc64le problem :)
17:55 Lyude: karolherbst: I was not, thanks for letting me know!
17:55 karolherbst: I think I'll wait until the drm fixes are in to rested
17:55 karolherbst: or do you have a build handy?
17:56 karolherbst: ohhhhh
17:56 karolherbst: wrong channel :D
18:08 danvet: Lyude, lack of load detect for vga most likely
18:08 danvet: we might miss the hotunplug
18:08 danvet: with DP we're supposed to not miss these
18:08 danvet: same kinda applies to hdmi and dvi too, hotplug on those is wobbly in the best of times
18:08 Lyude: danvet: jfyi, we can add load detect support for dp vga adapters (yes-it's a thing)
18:09 Lyude: danvet: so, and maybe you'll have some interesting input on this, the reason I was asking is because now that I've got connector epochs hooked up in nouveau I was wondering if it would be worth it to make sure that the DP MST helpers make use of them and handle hotplug reprobing themselves as well
18:10 danvet: Lyude, yeah I know, iirc airlied even typed a patch once for the apple special ones
18:10 Lyude: as well, that way we can ensure on suspend/resume we only start doing a proper reprobe of mst connectors once we've finished reprobing link addresses
18:11 danvet: Lyude, I think in general pulling both the direct sst and mst tree connector probe handling into helpers would be really awesome
18:11 danvet: so that we dont' have to duplicate all the stuff around what is ok to cache and what not
18:11 danvet: and also have like load detect on all sst connectors
18:11 Lyude: danvet: you mean direct sst as in not using MST at all?
18:11 danvet: (5 years after it stopped being relevant maybe, but hey)
18:11 danvet: yeah
18:11 danvet: I mean it's pretty much the same
18:12 danvet: at least for probe
18:12 danvet: link training and modeset is all different ofc
18:12 Lyude: that it is! in fact...
18:12 danvet: but it's also a todo for about as long as mst helpers exist, so no rush :-)
18:12 Lyude: https://gitlab.freedesktop.org/lyudess/linux/-/commits/wip/nouveau-dp-work i've only moved out the downstream info retrieval from i915, but I wanted to pull more code out of i915 and into helpers anyway
18:16 Lyude: and afaict yeah - the code between nouveau and i915 is nearly identical (note that I did just use i915 as a reference for what we were missing in nouveau, so that definitely contributed...)
18:20 danvet: yeah I think the big challenge with sst helpers is that we have an entire garden of different dp helper approaches in sst-only dp drivers
18:20 danvet: mst supporting ones are a bit more uniform I think
18:21 Lyude: for the most part yeah, other then the dsc stuff i need to strip out of amdgpu and make into helpers
18:23 Lyude: anyway-since moving everything into a set of helpers doesn't actually seem to be overkill like i thought it might be, I'm totally fine with doing that :)