06:06 gnurou: skeggsb: seems you haven't applied http://comments.gmane.org/gmane.comp.freedesktop.xorg.nouveau/25119 yet, any reason for this? PCI devices on ARM64 will be broken without this...
06:07 skeggsb: gnurou: sorry, slipped past, i'll grab it now
06:08 gnurou: skeggsb: naisu. please don't forget the secure boot fix series too ;)
06:09 skeggsb: i've got that one locally already
06:09 gnurou: skeggsb: I think Robin's patch should go into 4.6 and 4.7 as well, both are broken by my carelessness
06:17 skeggsb: gnurou: sorted, and sent to airlied
06:17 gnurou: skeggsb: great! thanks a lot
10:15 tacchinotacchi: hey, I'm running the linux-ck kernel on arch and if I try to compile nouveau out of the tree it blames about missing headers
10:16 tacchinotacchi: how do I actually build it against my kernel? ( the headers are in /usr/lib/modules/4.5.7-1-ck/
10:18 mupuf: tacchinotacchi: the out of the tree nouveau requires a very specific version of linux to work
10:19 mupuf: *very specific*
10:19 mupuf: unless you are a dev, OOT nouveau is a waste of your time
10:19 tacchinotacchi: I'm trying to be one
10:19 mupuf: ok, cool
10:20 tacchinotacchi: mupuf: it's kind of tedious to recompile the whole tree everytime I refactor if I do
10:20 mupuf: https://github.com/skeggsb/nouveau/commit/0e782017dc057db6cf058eec80646bb6652c8c3a <-- here is the commit you need
10:20 tacchinotacchi: so it won't compile against 4.5 ck?
10:20 mupuf: nope
10:20 mupuf: compile this particular commit, install it and have fun until ben rebases
10:21 mupuf: which does happen at least at the same frequency as the kernel releases
10:22 tacchinotacchi: how about I apply ck patches to the last commit of the kernel?
10:22 tacchinotacchi: Will it compile OOT then?
10:22 tacchinotacchi: I mean without rebasing every now and then
10:22 mupuf: no, it won't, you would need to port them to drm-next@92181d47ee74749bdf902f080dab9fa9ef44b37f
10:22 mupuf: just ignore ck when you are working on nouveau
10:23 tacchinotacchi: them? who?
10:23 tacchinotacchi: anyway, ok, bye bye rt audio
10:24 mupuf: them == ck patches
10:26 tacchinotacchi: the kernel tree from skeggsb took like 3 minutes to compile instead of 15, something's telling me I'll have to go over the config and enable some stuff :D
15:19 orbea: is DRI3 at all useful for nouveau? Or should I just stick with DRI2?
15:31 imirkin: useful? dunno. it's certainly usable, but afaik you do get funny artifacts in some situations
15:36 orbea: trying it now, nothing really broken looking yet
15:37 orbea: comparing DRI2 and DRI3 seems to make a speed improvement? http://dpaste.com/3RJSMPA http://dpaste.com/2KF4EEP
15:38 imirkin: slightly less overhead for displaying frames
15:38 imirkin: but that's never the bottleneck
15:40 orbea: seems to improved the laggy moments in PCSX2 and Xenosaga Episode 1 by 2-3 fps... :P
15:41 orbea: instead of ~11, get ~14
15:41 orbea: still not playable, but its something
15:41 imirkin: what mesa version are you on?
15:42 imirkin: gregory pointed out some idiocy in st/mesa that i fixed, should speed things up a bit if you're cpu limited. fix is in 12.0-rc3 as well as mesa-git.
15:43 orbea: Mesa 12.1.0-devel (git-7d7e015)
15:44 orbea: seems I should have the fix then :)
15:45 imirkin: likely, yeah
17:09 Watball: Nouveau randomly stopped working on boot for me today, giving no output to screen. I managed to boot by putting nomodeset in kernel parameters. I found this in /var/log/messages: https://bpaste.net/show/f8b3bfa83a06 . I'm on debian stretch, I did an apt upgrade yesterday, that may have something to do with it. Not really sure what to do next, would apt purging and then reinstalling nouveau help? Any ideas?
17:28 imirkin_: Watball: you have 2x GF104 in there right? looks like the second one comes up ok :)
17:29 Watball: Yeah I have two video cards
17:29 Watball: Though not SLI'd I think
17:30 imirkin_: Watball: unfortunately i have no idea what could be wrong. can you figure out what kernel worked and which one doesn't? (i.e. which kernel versions)
17:30 imirkin_: SLI doesn't (well - shouldn't) matter
17:30 Watball: I don't think the kernel changed? Everything worked fine yesterday, today I boot and I just get a black screen
17:30 imirkin_: well it should say in your messages log - you didn't include the whole thing
17:31 imirkin_: but you're right in that it could have "randomly" stopped working
17:31 imirkin_: in such a case i'd recommend doing a cold boot and seeing if that helps
17:32 Watball: I've turned it off and on a couple of times
17:32 imirkin_: disconnect power?
17:32 Watball: Log says the same thing as the failing one a couple of times, I just copy-pasted one to show
17:32 Watball: Didn't disconnect power, did power off completely
17:32 Watball: Should I try disconnecting power?
17:32 imirkin_: it's unlikely to help much, but i'd give it a shot
17:32 imirkin_: for at least like 1 minute
17:33 imirkin_: stupid motherboard capacitors are huge
17:34 Watball: ah wait shit I think you're on to something
17:34 Watball: Kernel version got upgraded
17:35 Watball: 4.4.0-1-amd64 to 4.6.0-1
17:35 imirkin_: aha
17:36 Watball: So I think I'll first try to boot with that old kernel version, I think it's still in the GRUB list
17:36 Watball: back in a bit
17:38 Watball: Yep, that worked
17:38 Watball: non-eye hurting resolution ftw
17:38 imirkin_: now just need to figure out what broke it
17:38 imirkin_: i just glanced at a selection of disp-related changes and nothing pops out as "i broke it" type of thing
17:39 Watball: Any logs I can paste for you?
17:39 imirkin_: the only thing i can think of is ff683df7bf34f90766a50c7e7454e219aef2710e - but it's a bit unlikely
17:43 Watball: kern.log: https://bpaste.net/show/838220d61d78 Not sure if this is of any use, but there's some errors in here
17:56 imirkin_: Watball: sorry, dunno
17:56 imirkin_: Watball: your best recourse is to do a bisect and see what that points to
17:58 Watball: How would I go about doing that?
17:58 imirkin_: search the googles
17:59 Watball: Yeah okay but what should I be bisecting? the kernel?
18:00 imirkin_: yes
18:00 imirkin_: you can probably limit the bisect to drivers/gpu/drm
18:00 imirkin_: not 100% sure that limiting to nouveau would be a good idea
18:06 RSpliet: imirkin_: aren't there ways to then exclude all the commits touching only drivers that aren't nouveau?
18:06 imirkin_: RSpliet: probably.
18:07 imirkin_: if only the bisect command were documented
18:10 Watball: https://wiki.ubuntu.com/Kernel/KernelBisection looks like a good resource but I don't really have time right now to do all that :/
18:10 mogorva: Hi-ho! I changed my gfx card NV92 (GTS 250) >> NV106 (GT 730), and noticed that many games drop fps dramatically when anti-aliasing is enabled. This didn't happen on my old NV50 card.
18:10 Watball: Thanks for the help anyway imirkin_, I can just use the old kernel version for now
18:10 mogorva: I'm using Mesa git on kernel 4.7-rc3. Any idea what could it be?
18:12 mogorva: I'm talking about 5-10 year old games, not the latest stuff.
18:12 mogorva: *years
18:12 RSpliet: mogorva: cat /sys/kernel/debug/dri/<number>/pstate could give a hint
18:13 imirkin_: mogorva: did you clock up the gpu? the G92 probably started at higher clocks than recent GPUs do.
18:13 mogorva: no, i didn't enable reclocking
18:13 imirkin_: the default clocks are shit on kepler gpu's
18:13 imirkin_: (and many fermi's)
18:15 mogorva: 07: core 405 MHz memory 810 MHz
18:15 mogorva: 0f: core 653-901 MHz memory 5010 MHz
18:15 mogorva: AC: core 405 MHz memory 810 MHz
18:20 imirkin_: mogorva: echo 0f > .../pstate
18:21 imirkin_: note that there's a certain chance your box will hang
18:21 imirkin_: so save your work first :)
18:21 imirkin_: looks like you have the GDDR5 version of that card... i have the GDDR3 one
18:28 mogorva: imirkin_, that did the trick and improved performance noticeably indeed :)
18:28 mogorva: 0f: core 653-901 MHz memory 5010 MHz AC DC *
18:28 imirkin_: apparently 5ghz is faster than 800mhz. who knew, right?
18:28 imirkin_: the AC line is the important one btw
18:28 mogorva: yup :)
18:29 imirkin_: the thing with the * is just there to confuse you
18:29 mogorva: AC: core 901 MHz memory 5009 MHz
18:29 imirkin_: yea
18:32 mogorva: lm-sensors shows GPU Core: +0.88V in the 07 state, and +0.94V in 0f state
18:33 imirkin_: yeah. need maor volts if you want moar ghz
18:35 mogorva: it's a fanless card, my last silent card was a 7600GT
18:36 imirkin_: this one should be faster
18:37 mogorva: imirkin_, in Unigine Heaven with the binary drivers i received almost twice the score than with my old GTS 250
18:38 imirkin_: mogorva: nouveau should get about 60% of the score in heaven iirc
18:38 karolherbst: fyi: I guess I won't be around between 7am and 6pm UTC+2
18:39 imirkin_: karolherbst: finally taking a nap?
18:39 mogorva: but in Unigine Valley I received almost the same score
18:39 imirkin_: mogorva: between blob and nouveau? or GTS 250 and the the GT 730?
18:39 karolherbst: imirkin_: exactly :)
18:40 mogorva: imirkin_, no, I only tried Unigine with the blob
18:40 mogorva: between the 2 cards
18:40 karolherbst: imirkin_: no idea if I may be able to be around in th week in this time, have to discuss stuff I guess (or simply ask)
18:40 karolherbst: mogorva: uhh lucky
18:41 karolherbst: mogorva: I guess your 730 is a kepler?
18:41 mogorva: karolherbst, yes it is a GK208
18:41 karolherbst: some 730 are fermi
18:42 imirkin_: his is a GK208B, just like mine
18:42 karolherbst: those cards aren't just only 33% as fast as the kepler variants, novueau also can't reclock fermis
18:42 imirkin_: but he's got GDDR5, while i have stinky DDR3
18:42 karolherbst: :O
18:42 karolherbst: even the gddr5 730
18:42 karolherbst: what a luck
19:19 mogorva: I finished the benchmark in Unigine Heaven with nouveau so I can compare with the blob
19:20 mogorva: nvidia blob 364.19: 482 pts <-> nouveau with reclocking: 362 pts
19:21 imirkin_: 75%. not bad.
19:21 mogorva: um...it was running @1440x900 on medium detail with 2xFAA
19:21 mogorva: haven't tried high detail yet
19:21 imirkin_: looks fancier with tess, right? :)
19:21 imirkin_: [compared to what you saw with your GTS 250]
19:22 mogorva: no, with the same GK208
19:22 imirkin_: right, for the test
19:22 imirkin_: i just mean in general
19:22 imirkin_: heaven optionally uses tessellation when it's available
19:22 imirkin_: which it isn't on GTS 250
19:22 mogorva: ah ok
21:21 orbea: :) I can play xonotic with nouveau, ultra settings and 60 fps if I use the third pstate setting :)
21:25 karolherbst: orbea: currently using my branch I guess?
21:25 orbea: actually no, just vanilla 4.6.2
21:26 karolherbst: ahh
21:27 orbea: I was thinking about trying your branch, but everytime I procrastinate :P
21:27 karolherbst: :D
21:27 karolherbst: well
21:27 karolherbst: it kind of depends on your vbios
21:27 karolherbst: and chip quality
21:28 Calinou: karolherbst: I ordered my GTX 1080
21:28 karolherbst: Calinou: wuhu
21:28 Calinou: it's a Gigabyte Gaming G1, will arrive in 3+ weeks though
21:28 Calinou: (Amazon had a nice deal, but obviously out of stock)
21:28 karolherbst: ohh
21:28 karolherbst: yeah well :D
21:28 karolherbst: though you will have to wait a long time before you can use it with nouveau :p
21:28 Calinou: yeah, will use blob anyway
21:30 orbea: karolherbst: I have a gtx 780 Ti, not really sure what to do with a vbios.rom :P
21:30 karolherbst: orbea: upload and give me the link :p
21:30 orbea: hold on
21:30 karolherbst: mhh on a 780 ti I would have expected more troubles though
21:31 orbea: karolherbst: http://ks392457.kimsufi.com/orbea/stuff/nouveau/vbios.rom
21:32 orbea: I made it like: cat /sys/kernel/debug/dri/0/vbios.rom > vbios.rom
21:33 orbea:assumes that is good way to get it
21:33 karolherbst: orbea: I am pretty sure reclocking fails with stock nouveau on your gpu
21:33 orbea: in which does it fail?
21:33 karolherbst: 0d and 0f
21:33 karolherbst: you should get a volting error in dmesg
21:34 orbea: I can reclock to them fine and they do seem to have a perf improvement, just this in dmseg http://dpaste.com/066XGB9
21:34 karolherbst: yeah
21:34 karolherbst: it doesn't reclock the core
21:34 karolherbst: check the last line in pstate
21:35 orbea: oh, never realized the last line changed before...
21:35 imirkin_: last line is "current state"
21:35 orbea: yea, core gets stuck at 691
21:35 karolherbst: imirkin_: I think my iMac G5 has a nvidia FX 5200 Ultra. Do you know of any problems with those gpus on pcc?
21:35 karolherbst: *ppc
21:36 karolherbst: should be nv34
21:36 karolherbst: orbea: well
21:36 karolherbst: orbea: your gpu is highle underclocked :p
21:36 karolherbst: orbea: you could switch to 0a before going to 0d/0f
21:36 imirkin_: karolherbst: plenty. i have a g5 with it too
21:36 karolherbst: orbea: but I expect that your gpu will hang that way
21:37 karolherbst: imirkin_: ohh nice :D so when I am super bored and find some time, I could try things out on that one too
21:37 orbea: karolherbst: I always do cahnge to 0a first before going to higher settings
21:37 karolherbst: imirkin_: chances are it has the ATI gpu, but I am not quite sure
21:37 karolherbst: orbea: ahh I see
21:37 karolherbst: orbea: ohh, your 0a isn't really high either
21:38 orbea: 0a: core 324-692 MHz memory 1620 MHz
21:38 karolherbst: orbea: mhh looks like that your gpu has some OC potential, causes it didn't crash so far
21:38 karolherbst: seems like it can be driven at -0.5V
21:38 orbea: It hasn't crashed with pstates since 4.3.3 for me
21:38 karolherbst: no idea how that goes on higher clocks though
21:39 karolherbst: orbea: yeah, some gpus have a really high process quality
21:39 karolherbst: orbea: which means the required voltage for a stable clock signal decreases
21:39 orbea: I think I will try your branch next time I compile a kernel
21:39 karolherbst: well you would get doubled clocks
21:39 orbea: see how much of a diff it makes :)
21:39 karolherbst: depends on the workload
21:40 karolherbst: in xonotic none most likely
21:40 orbea: yea, xonotic runs fairly well even with 30 fps
21:40 karolherbst: it shouldn't reach 30 fps with a 780 ti :p
21:40 orbea: I'm more interested in seeing the more intensive games in PCSX2 work honestly, like xenosaga, here is an issue with details https://github.com/PCSX2/pcsx2/issues/1403
21:41 orbea: I get around 30 fps with 07
21:41 orbea: for xonotic
21:41 orbea: at ultra settings
21:41 karolherbst: orbea: for that shader clocks are important, though we currently have some cpu overhead
21:41 karolherbst: orbea: ahh, that makes sense :D
21:42 karolherbst: orbea: usually in pcsx2 you _need_ a fast CPU
21:42 karolherbst: orbea: something like hsw i7 4.0GHz+
21:42 orbea: yea, I have a AMD FX-6350, it works for many games, but not all
21:43 karolherbst: yeah well
21:43 karolherbst: slow CPU :p
21:43 orbea: Was thinking about getting a nice intel cpu, but then I realized the chipset is not compatible with my mobo...
21:43 karolherbst: nope, they usually arne't
21:43 karolherbst: and those intel chips are really expensive
21:44 orbea: yea, the one I was looking at was $300 or so
21:44 karolherbst: yeah
21:44 imirkin_: karolherbst: well, i'm aware of the problems... fixing them requires clear-headed investigation. i hack-fixed some of them, but hardly all
21:44 karolherbst: imirkin_: I see
21:44 imirkin_: basically all (NV11+) gpu's have a way to switch into big-endian mode
21:44 imirkin_: but it's highly unclear what all that does
21:44 imirkin_: and some things still need LE <-> BE treatment, others don't
21:45 karolherbst: mhhh
21:45 karolherbst: in the kernel or mesa?
21:45 karolherbst: *userspace
21:45 imirkin_: emsa
21:45 karolherbst: I see
21:46 karolherbst: well
21:46 karolherbst: I guess with piglit we might be able to figure out, but... there is no nvidia driver we could use, right?
21:57 RSpliet: Actually FX6350 isn't so bad... I think I have the same CPU
21:58 orbea: its decent, not the best, not the worst
21:58 RSpliet: oh no I got the 6300... either way, pretty good value for money unless you need multithreaded FP
21:58 RSpliet: but for that, you got GPUs right?
22:00 orbea: im tempted to buy a new mobo and new cpu eventually, nice ones and then downgrade my current ones to an older empty case with my geforce 9800 (?) cards
22:00 orbea: forgetthe model on those cards, haven'tused them fora while
22:50 Doctors: Hello, I'm having trouble changing my backlight levels with the nouveau driver. I am unable to change the levels using my hardware brightness keys nor through xbacklight. I can however change the levels between 1-10 by echoing the value I want into /sys/class/backlight/acpi_video0/brightness. I should note that /sys/class/backlight/nv_backlight/ exsists, however changing vaules inside there doesn't have any effect on the screen its self.
22:52 RSpliet: then somehow I suspect your problem is not with nouveau
22:52 RSpliet: but rather with the ACPI driver for your laptop
22:53 RSpliet: on my laptop the buttons don't work, but Gnome Shell displays a brightness slider in the upper right menu that does the trick. Not sure if your window manager has that too
22:56 orbea: could always manually add the commands to acpid
22:57 orbea: i used to have to do that, but then the acpi driver was fixed and I was getting double the effect :P
22:57 Doctors: Sliders and what not don't work for me.