00:04 airlied: do we have a way to "profile" a non-game/non-rendering compute shader on AMD GPUs?
01:50 mareko: airlied: what do you need?
01:51 mareko: RADV can produce an RGP file, then open it in the Radeon GPU Profiler
01:55 airlied: mareko: trying to work out where the bottleneck in a very large shader is in llama.cpp, but not sure how best to capture a trace
01:57 airlied: ah I find MESA_VK_TRACE
02:06 airlied: now I just have to learn how to interpret rgp
04:57 mareko: the issue with blits is that the perf decision is no longer quads vs rectangles, it's draw vs compute
05:08 mareko: or rather draw-rectangles vs compute
07:03 valentine: sghuge: Those are piglit tests running with rusticl. Just make sure to set RUSTICL_ENABLE=iris, otherwise it should be the same as running any other piglit test
11:17 mlankhorst: pixelcluster: hey I'm about to review, let me find your latest series
12:47 mlankhorst: pixelcluster: series no longer applies, likely due to 591711b32681 ("drm/ttm: Convert -EAGAIN from dmem_cgroup_try_charge to -ENOSPC")
12:47 mlankhorst: Can you rebase and resubmit?
13:09 mairacanal: mripard, tzimmermann: sorry to ask, but which branch should mailmap patches go to: -next or -fixes?
13:12 mripard: I'd put it in next, but I'm not sure it really matters
13:12 tzimmermann: i'd put it in -fixes, so it goes upstream ASAP
13:20 mripard: fair enough :)
19:51 pixelcluster: mlankhorst: resent, patch 2/6 had a nontrivial fixup so i dropped your R-b there
19:52 mlankhorst: Thank you
19:57 mlankhorst: https://sashiko.dev/#/message/20260721195543.C41021F00A3D%40smtp.kernel.org
19:57 mlankhorst: hmm
19:59 mlankhorst: css_tryget issue seems real
20:00 mlankhorst: and 3/6, I believe we no longer return -EAGAIN but -ENOSPC
20:01 mlankhorst: See before mentioned commit, will comment on the patch
20:15 pixelcluster: mlankhorst: oh yeah i guess i will rewrite that fix
20:16 pixelcluster: also, i don't think i can push anything as i don't have commit rights
20:16 pixelcluster: my mailmap patch is iirc also hanging around because of that :P
20:16 pixelcluster: i should probably request access but also i'm slightly scared as i've never pushed anything before :D
20:17 pixelcluster: i.e. never pushed any commits to kernel repos
20:17 pixelcluster: i did, of course, use the git push command lots of times
20:17 pixelcluster: :P
20:23 mlankhorst: clone maintainer-tools
20:24 mlankhorst: and then push to drm-misc-next
20:24 mlankhorst: using dim push
20:25 mlankhorst: https://gitlab.freedesktop.org/drm/maintainer-tools/
20:44 pixelcluster: okay will try tomorrow, mailmap patch should be a nice trivial exercise to get into things :)
21:08 rpavlik: can't help but think a mailmap patch means celebration, though that might just be due to why I personally made mailmap patches ;-) (so congrats on new name or new job or...)
21:27 glehmann: or the worst option: just more responsibilities