01:24imirkin: skeggsb: no nouveau in the drm pull? :(
01:24airlied:wonders did I miss a pull
01:24airlied: I did think it was a bit light
01:25imirkin: i don't see a linux-4.16 branch in his public tree
01:25imirkin: so ... no pulll request i guess
03:03Jimi-James: so, any word on the nouveau thing? not to be impatient or anything; please take your time
03:58imirkin: Jimi-James: ok, so my xorg log also doesn't have any mention of the dri3 module
03:58imirkin: Jimi-James: i do have this line which you don't have though:
03:59imirkin: [ 9.320] (II) NOUVEAU(0): DRI3 on EXA enabled
04:00imirkin: Jimi-James: here's my full xorg log: https://hastebin.com/evopamimaw.sql
04:03Jimi-James: what's EXA?
04:04imirkin: EXA is the acceleration mechanism
04:04imirkin: perhaps xorg 1.19.6 disables DRI3 on EXA
04:04imirkin: it's a little buggy
04:05Jimi-James: yeah that's the weird part of this issue
04:06Jimi-James: i tried downgrading to xorg 1.19.3, which is the last version that i KNOW didn't have this problem
04:06Jimi-James: DRI3 still wouldn't enable
04:06imirkin: hm, nope, don't see anything like that
04:06Jimi-James: whatever changed about my system that made this stop working is something i dont even know about
04:06Jimi-James: it's something that isnt package versions, GPUs, or xorg conf files
04:06imirkin: phase of the moon? :)
04:06Jimi-James: er, at least not package versions of the kernel or xorg or nouveau
04:07Jimi-James: or even amdgpu
04:07Jimi-James: any ideas for what else it could be?
04:07Jimi-James: as far as xorg packages go, i havent tried downgrading bdftopcf, but i cant imagine that has anything to do with it
04:08imirkin: sorry, i don't really know. might be a better question for #xorg-users or #xorg-devel
04:08Jimi-James: could xf86-input- packages have anything to do with it?
04:08imirkin: i'm not enough of an Xorg expert (un)fortunately
09:38mupuf: imirkin, rhyskidd, RSpliet: how would you describe your work on nouveau?
09:59pmoreau: mupuf: Are you doing a quality questionnaire for Nouveau contributors? :-p (just kidding)
10:35RSpliet: mupuf: lacking ;-)
11:49imirkin: mupuf: occasional
11:49imirkin: (perhaps you might want to phrase the question differently?)
13:18karolherbst: mupuf: we have just 25 minutes, right?
13:19pmoreau: karolherbst: Yes
13:19pmoreau: Questions included (I assume)
13:19karolherbst: it might be too many slides already... not sure what you want to talk about though
13:19karolherbst: I had less slides for 53 minutes on devconf
13:19karolherbst: and I ended on time
13:20karolherbst: and we support 4.5 already ;)
13:20pmoreau: imirkin: I *think* that he wants to know what you have worked on in the past year/2 years, to put that in the slides for the Nouveau status update.
13:20imirkin_: mmm let's see...
13:21pmoreau: karolherbst: I think some of the slides can be removed, like the one talking about the d3d9 tracker
13:21karolherbst: yeah... I just fill in my stuff for now
13:21karolherbst: removing things is easier than adding
13:21imirkin_: in no particular order...
13:21imirkin_: advanced blend
13:22imirkin_: lots of various compiler fixes and small opts
13:23karolherbst: pmoreau: can you fill in the OpenCL bits?
13:23imirkin_: clocks (ARB_shader_clock)
13:23pmoreau: karolherbst: Nah
13:24karolherbst: I won't, because I can't talk about it
13:24pmoreau: karolherbst: I had already started yesterday, continuing right now
13:25imirkin_: group voting
13:25pmoreau: imirkin_: Cool, thanks! :-) Was there some xf86-video-nouveau patches as well, for DP-MST or Pascal support?
13:26imirkin_: bindless on kepler
13:27imirkin_: i did send patches for xf86-video-nouveau DP-MST (and related) support, but no one's been able to test
13:27imirkin_: and yeah, maxwell+ support in xf86-video-nouveau
13:27imirkin_: although that may have been pre-2017
13:27karolherbst: pmoreau: going through that stuff, I have tons of stuff in the pipeline.....
13:28imirkin_: oh, just kidding - group voting was 2016 too.
13:28karolherbst: imirkin_: do you know when we enabled OpenGL versions above 4.0?
13:28karolherbst: I guess most of the work was done over the last two years...
13:28karolherbst: thinking about it, it doesn't matter :D
13:29karolherbst: who cares
13:29karolherbst: ohh, that helps though
13:30imirkin_: 4.1 in mesa 11.0, 4.3 in mesa 12.0, 4.5 (effectively) in mesa 13.0
13:35imirkin_: wow, i really didn't do much this past year
13:35imirkin_: o well.
13:35imirkin_: this year's probably more of the same.
13:35imirkin_: you could mention that i've been looking into 30bpp
13:35imirkin_: (building on top of Mario Kleiner's work)
13:36pmoreau: Ah, true, there was the 30bpp stuff.
13:36karolherbst: I remove the d3d9 section
13:36karolherbst: or is it new?
13:37karolherbst: no idea when it landed
13:37pmoreau: Thanks for the summary, and don’t worry about not having done as much as previous years; you did quite a lot in previous years.
13:37karolherbst: uhh 2013
13:37pmoreau: I think it was part of the 2013 presentation already
13:37imirkin_: skeggsb: with the new vmm, can we get sparse on kepler+?
13:47karolherbst: pmoreau: I think I am done with my changes :)
13:48pmoreau: Cool, I’ll pull them then.
13:49pmoreau: Done. And I’ll be boarding soon, so good thing I pulled now.
13:50pmoreau: Oj, a basic Vulkan driver for this year? O.O
13:50karolherbst: why not?
13:51karolherbst: that's my hope :D:
13:51karolherbst: but maybe we shouldn't announce such things?
13:51karolherbst: I mean having the compiler ready for that is already good and I think skeggsb wanted to work on some memory stuff to support vulkan as well
13:51pmoreau: Ah, I think I should be boarding.
13:51karolherbst: have fun
13:52pmoreau: See you later!
13:52karolherbst: please push though
13:53pmoreau: Done, though there wasn’t much. Going to finish on the flight.
13:54karolherbst: ohh just comments, right
13:55karolherbst: well I need to get off my train now anyway :)
15:05mupuf: imirkin_, RSpliet: I meant, topic of interest
15:06mupuf: what are you working on, when you work on Nouveau
15:07imirkin_: mmm ... probably 30bpp i guess (kinda)
15:07imirkin_: but i haven't touched it in a bit
15:07imirkin_: i was hoping my patches would land in 4.16
15:08imirkin_: i've also been pondering a more substantial rewrite of nouveau
15:08imirkin_: to get off of libdrm_nouveau
15:09imirkin_: (rewrite of nouveau as in all the buffer/command submission handling and surrounding logic in the 3d driver)
15:09imirkin_: but ... it's a more substantial project, and i may never get time for it.
15:40mupuf: imirkin_: oh, that reminds me of the kde problem, Is that related?
15:58imirkin_: mupuf: i dunno... we like to blame it on the concurrency stuff, but i don't know if that's the case
15:58imirkin_: it definitely *was* the concurrency issue for a while, but afaik they've added logic to not do stuff concurrently on nouveau
15:59imirkin_: unfortunately (for the people that want to use kde + nouveau), i have no plans on installing plasmashell (or kde), so i'll never investigate the issue directly
15:59imirkin_: i've generally just been directing people to use blob driver or switch to more open-source-friendly hw
15:59imirkin_: [i have a pretty dim view of nouveau's prospects these days in case it's unclear]
16:11mupuf: imirkin_: the blob is also crashing for me on kde, with a big error in dmesg about a unhandled IQA
16:15karolherbst: mupuf: are you in brussels already?
16:16mupuf: karolherbst: yep, but finish your section before leaving your hotel "D
16:16karolherbst: huh? I am still in Prague :p
16:17karolherbst: and I already did
16:17karolherbst: I will arrive like super late, like I am sure I won't be able to join anything prior 10pm today
16:20RSpliet: mupuf: topics of interest are DVFS and compiler
16:21karolherbst: how so
16:21karolherbst: I am sure people are interested in DVFS and performance
16:21karolherbst: users don't want to hear boring compiler tech talks!
16:21RSpliet: karolherbst, I thought I was asked to explain *my* topics of interest...?
16:22karolherbst: ohhh, I thought you were talking about something more general
16:22karolherbst: or what user care about
16:22RSpliet: And given my whims on playing with RA under bank-conflicts and opcode scheduling, I guess I should consider those my interests ;-)
16:22karolherbst: sorry for that then
16:22karolherbst: yeah, that's fine
16:22karolherbst: next time I read the logs
16:23RSpliet: imho compiler talks can be very interesting. As long as slides don't contain code or APIs :-P
16:24mupuf: karolherbst: sorry, I thought all of them were Pierre's changes, my bad :D
16:24karolherbst: RSpliet: :D
16:24karolherbst: yeah, they can
16:24RSpliet: (exemplary 10 lines of assembly to demonstrate transformation results are excluded O:-) )
16:24karolherbst: and they are!
16:24karolherbst: I was just talking about the random user wanting to play games at awesome speeds
16:24RSpliet: hahaha true
16:24mupuf: RSpliet: I had written reclocking and compiler, so I wasn't far ;)
16:25RSpliet: although if you're thinking about the FOSDEM audience, based on previous years and ArsePull(tm) statistics, 87% is going to be DRM developers, 3% are NVIDIA blob developers :-P
16:36imirkin_: btw, can i look at the slides?
16:38mupuf: imirkin_: sure, will PM you a link
16:46pmoreau: Yummy, waffles! \o/
17:53pmoreau: I pushed some extra changes to the presentation.
17:59Lyude: Anyone ever seen something like this before? https://lyude.net/~lyudess/tmp/journalctl.txt
18:01imirkin_: Lyude: https://cgit.freedesktop.org/mesa/drm/tree/nouveau/pushbuf.c#n266
18:02Lyude: I was really just asking around to see if it was a bug someone had fixed yet
18:02imirkin_: if what was a bug?
18:02imirkin_: "gpu hang"? sure, there's plenty of bugs like that open
18:26sooda: mupuf: mikko and i found our hotel and need to eat now, would you happen to share the feel?
18:27sooda: pmoreau: ^
22:55rhyskidd: mupuf: to your question, i'd describe my recent work as "occasional" or hobbyist still for now
22:55rhyskidd: Pascal sensor support (temp)
22:56rhyskidd: (really trying to improve Pascal support given the XPS 9560 has been a popular laptop amongst linux users, and the experience so far is a bit sub-par...)
22:56rhyskidd: some preliminary work around improving testing and tool (igt, shader-db etc)
22:56rhyskidd: i hope the presso at FOSDEM goes well this weekend!