02:07 luc: Hi, I wonder if GPUOpen-Drivers/pal is only AMD internal library, is there any drivers in the public domain that is using it?
02:17 mareko: xgl
02:24 luc: thanks
09:03 MrCooper: MESA: warning: INTEL_HWCONFIG_TOTAL_PS_THREADS (128) != devinfo->max_threads_per_psd (64)
09:03 MrCooper: should I be worried about this?
11:43 mlankhorst: sima, airlied: https://patchwork.freedesktop.org/series/138088/ for drm-next
11:53 sima: mlankhorst, can you add dim cite for the offending merge commit?
11:54 sima: mlankhorst, also I wonder how that happened ...
11:55 mlankhorst: drm-tip merge conflict being resolved wrongly
11:55 mlankhorst: 8bdb468dd7a5 ("Merge tag 'drm-xe-next-2024-08-28' of https://gitlab.freedesktop.org/drm/xe/kernel into drm-next")
11:55 sima: oh it was airlied not me who's done the xe merge
11:56 mlankhorst: It was already resolved wrongly in drm-tip I think, so resolution carried over
11:57 sima: nah, there was no conflict on these at all
11:57 sima: at least per git show --remerge-diff
12:03 sima: nah, I looked wrong, there was a conflict
12:11 mlankhorst: It's there in drm-rerere under fixups, so something went wrong at least
12:12 sima: yeah old drm-tip was right, or at least had the 2 first two hunks correct
12:12 sima: the second one seems new
12:13 sima: mlankhorst, the 3rd hunk looks like a bugfix, I think that should go in as a separate patch with fixes: e7b180b22022f?
12:13 sima: that way drm-next will match what was in drm-tip
12:15 sima: mlankhorst, ack https://paste.debian.net/hidden/3d0f26a0/ ?
12:15 mlankhorst: Or fixes: for both?
12:15 mlankhorst: ack anyway
12:16 sima: eh don't want to jump the process too much, plus rodrigo might want to do the review for learning, imo that's best
12:16 mlankhorst: Yeah
12:17 sima: plus with this it matches drm-tip exactly from last week, so I'm much more comfy that this was tested
12:17 sima: don't want yet another fixup :-)
13:02 mlankhorst: The other change is not from may btw, let me find the commit id
13:04 mlankhorst: Fixes: 66a0f6b9f5fc ("drm/xe/display: handle HPD polling in display runtime suspend/resume")
13:04 mlankhorst: Cc: Arun R Murthy <arun.r.murthy@intel.com>
13:04 mlankhorst: Cc: Vinod Govindapillai <vinod.govindapillai@intel.com>
13:12 mlankhorst: It's a recent change regarding runtime pm, I'll put it in my patch series
13:54 sima: mlankhorst, yeah that's the one I've meant needs to be a separate fixup
17:39 Company: is there any method to get per-process VRAM usage on Intel?
17:40 Company: my codevelopers all have Intel laptops and can't see what I can see on my AMD
17:40 sima: fdinfo should I think
17:41 Company: what's fdinfo?
17:41 sima: /proc/$pid/fdinfo/$fd_you_want
17:42 sima: amd has it too, but iirc there was some confusion about how exactly to print stuff
17:42 sima: but I think that's aligned now
17:42 sima: tursulin_ will know more
17:42 Company: that just means someone needs to write a tool for it I guess
17:43 Company: because nvtop just works on AMD but doesn't on Intel
17:54 HdkR: Company: nvtop is supposed to read GPU memory on Intel. Sounds like a bug if it doesn't
17:55 HdkR: Although maybe it is xe only?
17:56 Company: nvtop hasn't had a release since that was added
17:56 Company: so Fedora's version can't do it
17:56 psykose: if its in fdinfo it should work but it's 6.8+
17:56 psykose: i think only xe support is not in a release
17:56 psykose: can always manually build master and compare i guess
17:56 Company: I just compiled git main and it's displaying something, but that's definitely not correct
17:57 psykose: rip
17:58 Company: because picking the first render fd of the app is wrong, as apparently they all print something different?
17:59 Company: apparently GTK apps have 3 different client ids with 5 different fds
18:00 Company: I suppose that's GL + Vulkan + someone else?
19:08 airlied: Company: intel laptops don't have vram, so I assume you mean ram
19:10 Company: airlied: ultimately I mean what Vulkan calls "device-local memory"
19:10 Company: no idea what that means for Intel laptops
19:32 DemiMarie: Company: for iGPUs device and host memory are one and the same
19:33 Company: that's not what Vulkan tells me
19:44 robert_mader: Hi, can anyone with marge permissions quickly reassign https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30728? Previous run just timed out. That would be super nice :)
19:55 austriancoder: robert_mader: done
19:55 robert_mader: Thanks!