07:56RSpliet: pq; that's sort of true yes :-) I don't think the sporadic task model accomodates for "as late as possible"
07:58RSpliet: and determining its WCET (dependent on the number of windows, display resolution(s)) is far from trivial, hence a "best-effort" approach seems to be as good as we can do
08:03karolherbst: imirkin_: you can also ask Tom^ for heaven benchmarks too, but I have some time now
08:30karolherbst: imirkin_: doesn't seem to make any significant changes at highest settings with normal tess with 8xmsaa
08:34karolherbst: mhh well there is an improvement though
08:34karolherbst: imirkin_: https://gist.github.com/karolherbst/faa04dc00ea0346c15d037df502dfbee
08:35karolherbst: but this could also be due to other changes, anyway, it got faster
08:42karolherbst: skeggsb: the pmu interface on the tegra seems to be totally different
08:44karolherbst: but we have "documentation" for all registers on the pmu
09:20AndrewR: I run into wine regression.. on first scene of 3Dmark2001SE (car chase) fps goes down to 0-1 range for whole scene, with one core constantly at 100% load! on nv92. Reverting wine to ver. 1.9.9 restores performance. But I wonder if newer wine just exposes bug in mesa/nouveau? Will apitrace taken with apitrace from 2015 still useful?
09:20RSpliet: AndrewR: potentially... did you build wine yourself and double-check it's not a heavy debugging build?
09:22karolherbst: AndrewR: you could git bisect wine and find the commit causing this
09:23AndrewR: RSpliet, yes I build wine myself, but mostly in same manner, and two versions since 1.9.9 already affected. But I can doublecheck, just it will take some time
10:42AndrewR: wine still compiling...may be I run into instance of Bug 97879 ? ( https://lists.freedesktop.org/archives/mesa-dev/2016-September/129544.html )
14:50perimetr: I got messages "nouveau 0000:01:00.0: iccsense: Unknown sensor type 30, power reading disabled" during booting in arch linux. And my laptop fan is going crazy these days. Is there any connection between the two?
14:56imirkin_: that's just for reporting power in hwmon
14:57imirkin_: laptop fans are usually controlled by EC
15:00perimetr: My fan was no problem, but I think two phenomenons recently started together. And when I view fan control utilities, I find hwmon in the source of those.
15:00perimetr: Absolutely no connection?
15:02imirkin_: hwmon reports fan speeds
15:02imirkin_: and chances are not via the nvidia card
15:05perimetr: Ok...I know this is nouveau channel, but do you have some idea about why my fan is always level2(highest speed)?
15:06perimetr: if I turn off fan by utility like fancontrol or i8kctl, it goes silent. But after a few seconds, it goes crazy again.
15:07imirkin_: coz your EC decides your laptop's too hot? dunno, sorry
15:07imirkin_: (EC = embedded controller)
15:10imirkin_: could try cleaning the fans/etc
15:11perimetr: I cleaned it a week ago with powerful air compressor^^
15:11perimetr: my conky and i8kutils show same temperature and it's usually 45-60 Celcius. Laptop hardware is too complicated, I can't pinpoint the problem.
15:12perimetr: Anyway thanks. My fan is making noise like crazy cats now. I have to turn off my laptop. Bye
17:28karolherbst: mwk: if you use c++17 features inside envytools, you also should adjust the cxxflags ;)
17:53karolherbst: ping on this PR https://github.com/envytools/envytools/pull/54
17:53karolherbst: any suggestions on making it less messy to parse?
18:03Jaga: late introduction, since I've been lurking here
18:03Jaga: hello #nouveau, I'll be hanging around
18:04Jaga: been out with some of the local to here people at xdc and it was fun enough to join
18:15RSpliet: Jaga: *waves* a shame we couldn't take up on your sauna offer, but we're putting time to good use finding the power-on pins in the docking port of an old laptop to hook up remote reboot
18:16Jaga: good to hear, I'm doing my homework now
18:16karolherbst: ohh, frigging IRC client, didn
18:16karolherbst: 't see you were talking here
18:16Jaga: the sauna offer stands for next time you're in the town tho ;)
18:16karolherbst: any my keyboard is frigging silly having the ' above the return key
18:16karolherbst: well, as it seems live is returning back here!
18:17karolherbst: mhh *life actually
18:20Jaga: will probably improve when everybody gets back home and it'll be long distance communication or nothing
18:21karolherbst: or whenever I feel bored at work
18:22Jaga: same here :p
18:24karolherbst: good :D
18:27karolherbst: to everybody: I will try to get a REing workshop place ready for 33C3 and the idea would be to maybe teach others how they can RE their nvidia gpus and maybe also just work on important things while others jump in and help if they have general reing experience or things like that
18:36tobijk: karolherbst: how did you manage to get splits on U32, i can only write tgsi with split from U64 :/ , https://trello.com/c/TnbFrUmM/162-nv50-optimize-mul-with-splitted-immediate
18:46karolherbst: tobijk: I didn't
18:46karolherbst: I don't really remember where I had this one from, but I saw it
18:47tobijk: mh ok
18:47imirkin_: AndrewR: if you give me an apitrace that repros the issue, i may be able to look at what's going on. i don't have a G92, but i do have a GT215 plugged in.
18:47tobijk: imirkin_: splits on U32 to U16? do you have somethin in mind for that? :>
18:48karolherbst: it might be that tesla does stuff like that, not sure though
18:48imirkin_: tobijk: just use shifts + ands
18:48imirkin_: tobijk: i have a thing which converts those into clever things when possible
18:49imirkin_: like INSBF + EXTBF type stuff
18:49imirkin_: tobijk: although you may be able to just use CVT for that actually
18:50tobijk: imirkin_: oh i was on replacing the "split into two regs" with split to two imms from an initiall imm
18:50imirkin_: oh ok
19:04AndrewR: imirkin_, still bisecting wine .. Initially I set good point wrong, now it compiles version between 1.9.14+ (good) and 1.9.15+ (bad)
19:08karolherbst: silly nvidia, crashed my kernel again :/
19:14karolherbst: seriously.... why isn'T nvidias vbios parser like bug free? :D
19:14karolherbst: have to reboot again
20:37mupuf: Oh yeah oh yeah! New machine in wtrpm!
20:37mupuf: my first nvidia laptop!
20:38mupuf: thanks to RSpliet for thinking about using the dock's interface to get to the power switch and LED
20:39mupuf: some fine soldering later, here we go, a nv86 is ready to be formated and put into use for testing mesa :)
20:40imirkin: RSpliet - planning on finishing up your G8x reclocking work? :)
21:22Hoolootwo: I've noticed on my thinkpad, the volume control buttons seem to have something to do with the video driver. when I installed the proprietary driver, they stopped working, and with the buggy 4.8 kernel I'm on right now they also don't work right
21:22Hoolootwo: is that a thing that is even possible?
21:35AndrewR: imirkin, and bisection come down to this commit in wine: wined3d: Swizzle D3DCOLOR attributes in the GLSL FFP replacement when necessary.
21:35AndrewR: imirkin, http://source.winehq.org/git/wine.git/commit/b62f4139eda58e8e25cbc58864fcea2599a20e97
21:36AndrewR: imirkin, https://www.winehq.org/pipermail/wine-patches/2016-July/152881.html
21:49mupuf: hakzsam: welcome back!
21:50hakzsam: mupuf, yep
21:50mupuf: my old g86 is now added to wtrpm and works!
21:50mupuf: now, going to bed, see you later!
21:50mupuf: was fun to have you here!
21:50hakzsam: yeah, it was really fun with you guys
21:50hakzsam: good night :)
21:50mupuf: you'll tell me how everything went, but given what time it is, seems like everything was smooth!
21:52hakzsam: we can talk in the next few days, because I'm going to sleep as well
22:00AndrewR: imirkin, strange, under apitrace it was much more smooth even at bad commit, while fps was still much lower (I traced good and bad versions of wine). Currently compressing traces with xz
22:08AndrewR: imirkin, https://cloud.mail.ru/public/75Mf/fFuSprBh9 - 58 and 61 mb compressed
22:26tobijk: imirkin: any reason we do not always propagate consts? i have tricked the compiler by chance: http://pastebin.com/ksaqZQ0B
23:12imirkin: tobijk: load propagation happens before the split logic
23:13imirkin: AndrewR: thanks, downloading
23:15tobijk: imirkin: it would split it right, but the next insn isnt set right somehow...
23:16tobijk: uhm forget that
23:19imirkin: it doesn't prop the 64-bit val
23:21tobijk: imirkin: they are already converted to u32 when we want to propagate
23:22tobijk: at least that is what my print says :>
23:25imirkin: 0: mov u64 %r34d 0x0000000400000000 (0)
23:25imirkin: it's a 64-bit value
23:25imirkin: the load prop thing doesn't know about splits iirc
23:26tobijk: mhm i have mobed it behind the split pass for now
23:28imirkin: AndrewR: good: Rendered 1045 frames in 33.7289 secs, average of 30.9823 fps
23:29imirkin: bad: Rendered 964 frames in 36.5877 secs, average of 26.3477 fps
23:33imirkin: AndrewR: looks like the bad one has a lot more data cycling into it
23:35imirkin: AndrewR: lots of shaders that look like http://paste.debian.net/plainh/a3c5c504
23:35imirkin: could just be diff playing games
23:35imirkin: i.e. the actual shaders may be quite different
23:38imirkin: yeah, pretty hard to tell what the difference is
23:42imirkin: AndrewR: perhaps report it to wine guys - i wouldn't be surprised if it affected more than nouveau