07:49 MrCooper: anonymix007[m]: yes, tiling is supposed to be converted on the fly (note that nvidia is broken in this regard AFAIK)
08:01 mlankhorst: airlied: Your patches to implement memcg accounting interact with the dmem patches, and should likely attempt to use the same cgroup for both.
08:21 airlied: mlankhorst: indeed, I just want to keep the first round as simple as possible though and iterate towards joining things
08:21 airlied: but we should use the same objcg for both memcg and dmem
08:24 airlied: mlankhorst: also how it interacts with the ttm/xe swap code is probably in need of investigating
10:30 mlankhorst: I have no idea, just compiled it again together with the dmem patch
10:37 mlankhorst: My local kernel is now a mixture of xe preempt-rt patches and the memcg series
11:49 mlankhorst: airlied: it doesn't really like a null memcg, so it probably breaks all non-cgroup aware drivers
11:50 mlankhorst: presumably just init with root memcg in default case
11:51 mlankhorst: problem I encountered was mem_cgroup_uncharge_gpu_page() not handling null memcg when calling refill_stock()
12:02 mlankhorst: Though fixing it caused a percpu underflow, so that's likely still a bug somewhere..
12:05 mlankhorst: https://gitlab.freedesktop.org/mlankhorst/kernel/-/commits/ttm-cgroup-xe
13:23 zmike: once again asking literally anyone to ack https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/36416
13:29 karolherbst: you did the same thing I did :) so I guess that's implicitly an ack
13:29 zmike: yeah for unmaintained drivers it's the safest choice
13:30 zmike: well whatever
13:31 zmike: maybe we'll get to delete stuff before I go on vacation after all
13:33 dj-death: zmike: can give you a ack
13:33 dj-death: well for crocus
13:33 dj-death: i915 is ...
13:34 dj-death: was I of drinking age when that driver started?
13:34 zmike: haha
13:34 dj-death: french drinking age, which must have been 16 at the time ;)
13:35 zmike: lul
13:35 dj-death: "While it was previously legal for those 16 and older to consume fermented beverages like wine and beer, the law was changed in 2009 to raise the minimum age to 18"
13:35 dj-death: woah, I'm not even joking
13:35 daniels: iris, crocus, pastis
14:05 onefreakosoid: So the path to collecting the rembrants to be elliminated comes with minor tweak on second barrel and has fairly natural path otherwise besides this. 371-141=230 where the other items at non-focal get 371-256=115 it can also start with inverse index. hence then remove -230 it yields -115 and from 371-115=256 where other elements stay at 371 but 512 subtract causes them to yield -141 and
14:05 onefreakosoid: once they combine on another entry with coalesced removals at runtime or adds at compile time then at any point of execution hashes elmination values can be derived. All in all the code is very efficient but not dramatically complicated, hence the work on such paradigm is quite enjoyable, and if you mess up with bringing drivers up i get a call to do that also on my own, which I am very
14:05 onefreakosoid: capable of but isn't at all that interesting or enjoyable or so to speak fairly boring job especially with people behaving so in the mix.
14:56 mupuf: dj-death: lol, i had missed that too! Heck, i even remembered 14 a the age to drink light beers 🤣
15:09 mlankhorst: mupuf: at least you were old enough to have a limit on alcohol ;)
15:10 mupuf: ha ha, yeah! And no wine served in primary schools too ;)
15:15 mlankhorst: That campaign encouraged the French to "drink less than a litre of alcohol per meal".
15:16 mlankhorst: Seen as moderation in France
16:56 dj-death: mupuf: I heard you still can in the UK, a pint of cider with a meal under adult supervision
17:02 mupuf: ha ha
17:12 glehmann: anholt: does your rb in https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/36468 still stand?
22:00 zzyiwei: curious...are gfxstream devs allowed to land changes like this: https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/36459
22:02 airlied: probably not without asking for acks from others
22:05 zzyiwei: yeah...that aligns with my expectations
22:11 llyyr: is there any point to even having gfxstream in mesa? each time i've been reminded that it exists in tree in mesa as a linux user is when they make some changes that break other things
23:18 robclark: zzyiwei: yeah, we kinda haven't formalized _who_ needs to ack changes, kinda relying on good behavior.. this perhaps exposes a gap
23:19 robclark: zzyiwei: anyways, if in doubt revert.. they can always reland if it turns out to be the right thing
23:41 zzyiwei: robclark: I've only noticed it while rebasing my stuff. I'll send a revert for now, AND let them know about this.
23:41 robclark: yeah, that is a sign of the process going wrong ;-)
23:42 zzyiwei: Here we go: https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/36532