00:26 mrnuke: imirkin: so, 4.3.0-rc1 is not packaged in Fedora. However, I can now get to the login screen :D
00:26 mrnuke: KDE login screen that is, at glorious 4K resolution
00:26 mrnuke: can't complain thusfar
00:26 imirkin: cool
00:33 mrnuke: so segfault in OsLookupColor when I try to log into KDE
00:34 mrnuke: but at least I get the to the darn login screen
00:34 mrnuke: and I get shiny 4k VTs
00:34 mrnuke: something there must be working
00:36 imirkin: mrnuke: that's a core X issue
00:37 imirkin: at least... i think
00:37 imirkin: errr.... hrm
00:38 imirkin: not the issue i was thinking of
00:38 imirkin: pastebin xorg log?
00:40 pmoreau: mrnuke: Do you get a libsane.so.0 not found in dmesg?
00:48 mrnuke: imirkin: some other day man. I need to catch some zzzZZs
00:48 mrnuke: pmoreau: I haven't checked
00:48 mrnuke: but I don't remember seeing one
00:49 pmoreau: Someone yesterday had a bt from OsLookupColor because libsane.so wasn't found. Thought it could be your case as well.
00:49 imirkin: i think i've seen that crash a bunch actually
00:52 pmoreau: dmesg and Xorg.log from yesterday: http://paste2.org/Z78CWC7U and http://paste2.org/fJg88ABb
00:55 imirkin: it'll require someone who understands X to debug
00:55 imirkin: it's going into present_event_notify and then somehow ending up in OsLookupColor? seems dodty.
00:55 imirkin: dodgy*
00:55 imirkin: airlied: --^
00:57 imirkin: feels like a race between client shutdown and usage... but what do i know
02:00 karolherbst: mupuf_: do you know if any of your cards have voltage issues?
02:00 karolherbst: I want to debug those, but luckily with your patches I get the right voltage values. so its hard to debug for me here locally
02:51 karolherbst: pecisk: are you there?
02:52 pecisk: karolherbst, yes
02:52 karolherbst: I have a little patch ready in a few seconds you should try out
02:54 pecisk: karolherbst, I will be able to test it in 8 hours, but sure, give me details, I will try it when I get home
02:55 karolherbst: is the kernel C99 already by the way?
02:56 njin: Hi, is anyone aware of problems with nvidia fx 370?
02:57 karolherbst: njin: which problems?
02:57 karolherbst: pecisk: update karlton_test branch and try to set voltage with nouveau.debug=debug
02:57 karolherbst: this should print a lot of stuff
02:57 njin: freeze, bug1487573
02:58 njin: bug#1 487573
02:58 njin: bug#1487573
02:58 karolherbst: which bugtracker?
02:58 njin: Launchpad
02:58 pecisk: karolherbst, roger
02:59 karolherbst: njin: and what does this bug have to do with nouveau?
02:59 njin: [ 8958.546987] nouveau E[compiz[2121]] fail set_domain
02:59 njin: [ 8958.547001] nouveau E[compiz[2121]] validating bo list
02:59 njin: [ 8958.547014] nouveau E[compiz[2121]] validate: -22
03:00 karolherbst: maybe I found the wrong bug then
03:00 njin: I am trying ti understand
03:00 karolherbst: njin: do you have the link?
03:00 njin: https://bugs.launchpad.net/ubuntu/+source/compiz/+bug/1487553
03:01 njin: karolherbst, it is opened against compiz and I am trying to triage it
03:02 karolherbst: njin: do you have a tesla card?
03:03 karolherbst: anyway, the Xorg error is harmless
03:03 karolherbst: "fail set_domain" seems to be the more important one
03:04 njin: karolherbst: is not me the reporter, NVIDIA Corporation G84GL [Quadro FX 370] [10de:040a] (rev a1) (prog-if 00 [VGA controller])
03:04 karolherbst: okay, tesla
03:05 karolherbst: mhh, nouveau_gem_set_domain fails
03:05 njin: seems a tesla, yes
03:05 karolherbst: with -EINVAL
03:05 karolherbst: imirkin: any idea?
03:05 karolherbst: ohh wait
03:06 karolherbst: njin: but do you have the same issue?
03:06 njin: no, it affect only the reporter
03:06 karolherbst: I see
03:06 karolherbst: I assume there is something stupid in the kernel cmdline
03:06 njin: it is running wily
03:06 njin: it/she
03:06 karolherbst: mhh
03:06 karolherbst: doesn't seem the case
03:07 karolherbst: strange
03:08 karolherbst: I assume the card is just really badly supported
03:08 karolherbst: and something trivial is missing
03:09 karolherbst: no clue though what it could be
03:10 njin: ok, no matter, can I assign the bug to nouveau then??
03:11 karolherbst: I would say it is something in the module
03:11 karolherbst: don't think userspace should be able to mess the module that hard up
03:16 njin: the is more a linux problem than nouveau ?
03:19 karolherbst: it should be nouveau, but without know what exactly is causing it
03:20 karolherbst: mwk: I think bit 2 is pcie v2+ support, but not really sure
03:20 karolherbst: its only not set for cards with SYSTEM_SPEED=2.5
03:21 karolherbst: is there a way to disable pcie v2 in linux?
03:22 njin: karolherbst, thanks for your help
03:30 mwk: no idea
03:39 karolherbst: mhh maybe my bios has something, will check it
04:03 karolherbst: mwk: okay, I can't do anything with my bios, so I can't really confirm this
04:08 karolherbst: haha nice, messed up my bios and now my gpu fan won't turn off
04:11 karolherbst: wait a second
04:12 karolherbst: nouveau got loaded without skeggsbs hack again
04:56 bozhan: hi
04:57 bozhan: does this "volt/gk104: add support for pwm and gpio modes" affect my gk107 gar
04:57 bozhan: s/gar/card/
05:10 karolherbst: bozhan: maybe, most likely not
05:11 karolherbst: bozhan: why do you ask?
05:14 bozhan: just checking when my card will start to work with nouveau
05:14 karolherbst: bozhan: what issues do you have?
05:14 karolherbst: gk107 should work already
05:14 bozhan: https://bugs.freedesktop.org/show_bug.cgi?id=70354
05:15 karolherbst: bozhan: did you test nouveau yourself?
05:15 bozhan: no it's not
05:15 bozhan: once a month ago
05:15 mupuf_: karolherbst: a wrong voltage is one voltage does differs from nvidia
05:15 mupuf_: so yeah, both my nve6 and gm107 qualify
05:15 mupuf_: anyway, have anyone seen this? http://blogs.igalia.com/apinheiro/2015/09/14/optimizing-shader-assembly-instruction-on-mesa-using-shader-db/
05:16 karolherbst: mupuf_: I was aware of the fact there is something like that
05:16 karolherbst: sadly there are more challanging issues in nouveau left yet :/
05:17 bozhan: karolherbst: do you think that is worth to try to compile noveau myself again?
05:17 karolherbst: so the best nouveau could do is to try to reduce reg usage (or just eliminate some instructions)
05:17 karolherbst: bozhan: there is a hack for that kind of issue
05:17 karolherbst: what kernel are you running?
05:18 bozhan: Linux bozhan 4.1.0-2-amd64 #1 SMP Debian 4.1.6-1 (2015-08-23) x86_64 GNU/Linux
05:18 bozhan: but i think i tried 4.2.0 once with same problems
05:18 karolherbst: bozhan: on 4.2 apply this: https://github.com/karolherbst/nouveau/commit/25b7dd9ed0caa4b1513843835223189df0cb100a
05:18 karolherbst: ohh wait
05:19 karolherbst: the reworks were done in 4.3
05:19 karolherbst: mhhh
05:19 karolherbst: then use my master branch: https://github.com/karolherbst/nouveau/commits/master
05:19 bozhan: k
05:19 karolherbst: bozhan: hybrid graphics?
05:19 bozhan: later , now i'm going to kill some bad guys in eve online :)
05:20 bozhan: karolherbst: yes
05:20 karolherbst: okay
05:20 karolherbst: mupuf_: there are some cases, where on high cstates nouveau fails to set voltage
05:20 karolherbst: with -EINVAL
05:21 karolherbst: and I want to debug it with this patch: https://github.com/karolherbst/nouveau/commit/c0532d21f36635ff049d79aa280359f10ffea709
05:21 karolherbst: usually nouveau applies a different voltage than listed in the voltage map table
05:22 karolherbst: usually 125000uv too high
05:22 mupuf_: ack
05:22 mupuf_: then it should be easy to fix :)
05:22 mupuf_: have to go, just woke up in toronto
05:22 karolherbst: I see
05:22 karolherbst: I wanted to ask you if I could play with reator with one of these cards :D
05:23 karolherbst: guess this has to wait then
05:55 karolherbst: okay, I think I got it, now checking why this is bad
05:56 karolherbst: okay, mhhh
05:57 karolherbst: voltage is voltMapTable[id].voltage_min + voltMapTable[voltMapTable[id].link].voltage_min for voltMapTable[id].link != 0xff
05:58 karolherbst: okay, that may make sense
06:08 karolherbst: found the issue
06:08 karolherbst: nice
06:09 karolherbst: mupuf_: voltage table accept voltages from 825000 to 1212500, but for two last cstates nouveau tries to set it to 1212500 + 12500 and 1225000 + 12500
06:10 karolherbst: so it fails, because nouveau doesn't find a matching "voltabe table" entry for the two last ones
06:10 mupuf_: right
06:11 mupuf_: check what the blob does
06:11 mupuf_: have fun with reator
06:11 karolherbst: how can I read the voltage out on blob?
06:11 karolherbst: :D
06:12 karolherbst: I think I will figure that out
06:14 karolherbst: ohh on gpio you only set the vid and then that's it?
06:17 bozhan: karolherbst: http://pastebin.com/m24zZMYG
06:17 bozhan: karolherbst: that is with you treee
06:17 bozhan: s/you/your/
06:17 karolherbst: seems to be a totally different issue then
06:18 karolherbst: you should mmiotrace the blob and file a bug with it then
06:18 karolherbst: or find the right issue
06:19 karolherbst: chances are, nobody will be able to debug/test it with hardware which shows the same issue as yours, so maybe you want to give it a try yourself ;)
06:19 bozhan: i've tried several times to make mmiotrace:) and send themto mmio.dukms;asdfasdf@gmail.com
06:19 karolherbst: I see
06:20 karolherbst: but you could try to figure it out yourself. trace the blob, trace nouveau
06:20 karolherbst: and check what gr related nouveau does wrong or different compared to the blob
06:20 karolherbst: ohh wait
06:20 karolherbst: this looks odd "nouveau 0000:01:00.0: priv: GPC0: 419eb4 badf1000 (3800820c)"
06:40 bozhan: ppl what do you think about muslim invasion in west europe?
06:40 pecisk: bozhan, it's offtopic for channel?
06:41 karolherbst: despite that, your wording makes it impossible to speak with you about it
06:43 bozhan: i think the same, just read some news and asked myself what do think peoples which aren't politics
06:44 bozhan: sorry
07:14 karolherbst: ohh man, somehow reading voltage out of the pwm is much easier than from the gpios :/
07:14 karolherbst: what a mess
07:39 karolherbst: mupuf_: new blob is needed
07:39 karolherbst: flush_workqueue is GPL only with 4.2 and the blob uses this
07:40 karolherbst: http://lkml.iu.edu/hypermail/linux/kernel/1508.2/02297.html
07:41 karolherbst: oh man
07:47 karolherbst: okay, seems to be fixed for 4.2
07:47 karolherbst: mupuf_: how to update or change kernel on reator?
09:50 karolherbst: new pcie speed change API and base code design: https://github.com/karolherbst/nouveau/commit/5be678fccadbe682e9e1d23dabd445b237642d26
09:54 karolherbst: mhh I am not sure about nvkm_pci_raise_pcie_version though
09:54 karolherbst: the return value
09:55 karolherbst: currently it returns success whenever the version is higher than 1
09:56 karolherbst: nvkm_pci_set_max_pcie_version?
09:56 karolherbst: maybe this is a better name for it
09:57 karolherbst: don't know
09:57 karolherbst: any ideas?
09:58 mlankhorst: _init(). :P
09:59 karolherbst: there is a _init function already
09:59 karolherbst: this is part of init
09:59 karolherbst: but chipset independent
09:59 karolherbst: it is actually called inside init
10:00 karolherbst: maybe I should explain what I want to do inside the init phase
10:00 karolherbst: 1. max out pcie version, because anything else doesn't make much sense (we get 2.5 speed also in v2 mode)
10:01 karolherbst: 2. run chipset specific init steps (kepler raises lnk_cap and lnk_ctl2 to max, while earlier chipset doesn't do that)
10:02 karolherbst: for the first one, I want to print an error message when something happens you actually don't want to happen
10:02 karolherbst: like v1 => v2 transition fails, allthough the card and the bus supports v2
10:03 Karlton: [5~:m #fsf can't just indorse people who don't agree with your cause and actively go against it
10:03 karolherbst: Karlton: :p
10:04 Karlton: stupid client
10:04 karolherbst: change it
10:05 Karlton: was trying out sic but I think I borked it
10:05 Karlton: /quit
10:05 karolherbst: :D
10:05 karolherbst: well
10:34 imirkin: what was the issue with metro redux + nouveau? crash due to VM exhaustion right?
10:34 karolherbst: yes
10:35 karolherbst: bo related I think
10:35 imirkin: njin: "fail set_domain" usually means "you have run out of vram, have a nice day"
10:38 karolherbst: imirkin: anything I could try out?
10:38 imirkin: karolherbst: nope, just wanted to confirm that that was the issue.
10:38 imirkin: karolherbst: my memory is basically mush when it comes to these things
10:39 imirkin: so many bugs
10:39 imirkin: so many different causes
10:39 imirkin: so many failed debugging sessions
10:39 imirkin: but a few fixes here and there, which is nice too
10:39 karolherbst: I see
10:39 imirkin: the texture buffer one i just sent is a fun one
10:39 imirkin: although i doubt it comes up much outside of piglit tests
10:40 karolherbst: who knows
10:40 njin: imirkin, thanks then is nouveau related?
10:40 karolherbst: maybe its better
10:40 imirkin: njin: the error? the error is printed by nouveau...
10:40 njin: ok, i get
10:40 imirkin: njin: i didn't fully understand what you were asking tbh, just saw that bit of log
10:41 imirkin: thought i'd clarify what it meant
10:41 imirkin: basically each render has certain buffers it needs to render, and some of those are required to be in vram
10:41 imirkin: if you don't have enough vram to hold it all, the submit fails
10:42 karolherbst: I bet compiz is just using to much vram :p
10:42 njin: lol
10:42 imirkin: now, we require more stuff to be in vram than absolutely necessary
10:43 imirkin: and i bet that the submit management could be improved so that there was a higher chance of it all fitting
10:43 karolherbst: he has a quadro card though
10:43 imirkin: but i suspect the GPU in question has a piddly 256MB of VRAM in the first place or something
10:43 karolherbst: I thought these have tons of vram?
10:43 imirkin: fx 370 = "look ma, i got a quadro for $10"
10:43 karolherbst: :D
10:44 karolherbst: wow
10:44 karolherbst: 256MB indeed
10:44 imirkin: i have a fx 370lp which is a G98, which is the weakest GPU of the pre-GT215's... GT218 is about the same as G98.
10:45 karolherbst: I just look at those and I really think: "what are these even good for?"
10:45 imirkin: well, that one was good for me to test VP3 video decoding accel ;)
10:45 karolherbst: :D
10:45 imirkin: and i did get it for literally $10
10:46 imirkin: although admittedly somewhat after its best-by date
10:47 njin: so u think that the card can't run compiz - unity ?
10:48 njin: well I am out of range in this things, i am just trying to help to triage this issue
10:49 imirkin: njin: dunno about "can't", but certainly sounds like it's having issues :)
10:49 njin: ok, I understand
10:51 njin: have a nice day you too
10:51 imirkin: njin: what buckets are you selecting between for the triage?
11:04 karolherbst: maybe disabling all compiz plugins may help :D
11:08 karolherbst: ohh I should check witcher 2
11:09 imirkin: i pushed a patch which fixed some issues with it
11:09 karolherbst: yeah, saw your comment
11:09 imirkin: should be in 11.0
11:09 karolherbst: I use mesa master anyway
11:25 tobijk: karolherbst: to your changes and in particular to nvkm_pci_raise_pcie_version(): i would expect this to return the pcie lvl somehow, maybe 1=2_5, 2=5_0...
11:25 tobijk: just some idea :)
11:25 karolherbst: tobijk: pcie version isn't tied to speed
11:26 tobijk: s/pcie version/speed/
11:26 karolherbst: yeah, got what you mean
11:26 karolherbst: so nvkm_pci_set_pcie_link should return the actual speed then?
11:26 karolherbst: for example
11:27 karolherbst: and nvkm_pci_raise_pcie_version should return the pcie version which the card was set to
11:27 tobijk: thats my idea, yep
11:28 karolherbst: mhh now that I think of it, I never got much into these int to bool conversion thing...
11:29 tobijk: i had it upside down, but return specific values :)
11:29 karolherbst: if I return -ESOMETHING will this evaluate to true or false?
11:29 karolherbst: mhhh
11:42 karolherbst: imirkin: seems good so far
11:43 karolherbst: testing maxed out settings now :)
11:44 karolherbst: imirkin: it looks better but there is still something left
11:44 imirkin: karolherbst: well, i'll need a trace. please try to take one at a reasonable resolution
11:45 imirkin: so that my piddly GF108 can run it during my lifetime
11:45 karolherbst: ohh, and a crash
11:45 imirkin: ooh, i like crashes!
11:45 imirkin: much easier to debug
11:45 karolherbst: with eon?
11:45 karolherbst: I doubt that
11:46 imirkin: if it's in an apitrace...
11:47 karolherbst: ohh right
11:47 karolherbst: seems like I just set too high settings
11:47 karolherbst: will be fun for ya to debug that :p
11:47 imirkin: well, that shouldn't cause *crashes*
11:47 karolherbst: ohh yes, it does
11:47 imirkin: but please do the trace with 640x480 or something
11:47 imirkin: rather than 2560x1440
11:47 imirkin: heh
11:48 karolherbst: it only crashes while loading a game
11:48 karolherbst: hey
11:48 karolherbst: I have only full hd
11:48 karolherbst: I am not insane
11:48 karolherbst: ohh wait I am, I joined this IRc channel today, should leave it because the only seem to rant in there, will wait a few days though
11:55 karolherbst1: okay, that went bad
11:56 karolherbst: a lot of fun coming towards me, I can clearly see that
12:02 pecisk: karolherbst: don't leave yet, I have test to do for you :)
12:02 karolherbst: imirkin: pstore, last message block is at the top: https://gist.github.com/karolherbst/924156ad095ad8da186c
12:05 karolherbst: pecisk: witcher 2 just messed up system
12:06 karolherbst: imirkin: I still get those issues
12:08 pecisk: karolherbst: blame Virtual Programming :)
12:08 karolherbst: ohhh
12:08 karolherbst: 600 fps in witcher 2 menu?
12:09 pecisk: ok, rebuilt karlton_test branch, reboot
12:09 pecisk: much performance. wow :)
12:09 karolherbst: I want 60, but well
12:09 pecisk: karolherbst: what card do you have?
12:09 karolherbst: 770M
12:09 karolherbst: ohh vsync was disabled on low
12:11 pecisk: ok, bb to test update
12:12 karolherbst: imirkin_: bout those traces, it seems like I won't get any
12:12 karolherbst: "87355: message: major api error 1: GL_INVALID_VALUE in glTexImage2DMultisample(samples < 1)"
12:12 karolherbst: maybe I need to find a way to create better ones
12:14 karolherbst: ohh right apitrace: warning: glMapBufferRange: MAP_COHERENT_BIT unsupported (https://github.com/apitrace/apitrace/issues/232)
12:15 karolherbst: wait, it does use glMapBufferRange in GL2.1 mode?
12:17 imirkin_: MAP_COHERENT_BIT = ARB_buffer_storage iirc
12:17 pecisk: karolherbst: now I have newest karlton_test, turn on 0f and try to set 50?
12:18 karolherbst: pecisk: I alredy know what the issue is, have to RE the blob now :/
12:18 njin: imirkin, i was blaming nouveau but now after your triage i have to better evaluate
12:18 imirkin_: karolherbst: that's not great... looks like some issue in nv84_fence_read? looks like it gets a null pointer of some sort
12:18 imirkin_: on which it then does poniter math
12:19 imirkin_: njin: well, certainly nouveau should be more resistant to these sorts of things
12:19 karolherbst: my entire drm got messed up
12:19 karolherbst: and kernel crash
12:19 imirkin_: njin: and if you want to file it that way, that's fine... but probably won't get fixed in any sort of near term
12:19 pecisk: why sad face? that's good news right?
12:20 karolherbst: imirkin_: disabling buffer_storage doesn't work :/
12:20 karolherbst: pecisk: no :D
12:20 karolherbst: it would be much easier if the issue itself would be easy to fix
12:20 pecisk: ohh
12:20 njin: I unerstand
12:20 karolherbst: RE the blob is tricky, because you need hardware
12:20 imirkin_: karolherbst: how did you disable it?
12:20 karolherbst: imirkin_: MESA_EXTENSION_OVERRIDE=-GL_ARB_buffer_storage
12:20 karolherbst: the game crashes with it
12:20 imirkin_: hm ok
12:20 karolherbst: so its really disabled
12:21 karolherbst: maybe eon has a nice flag for it
12:21 pecisk: I think some of the eon flags were mentioned in phoronix articles
12:21 karolherbst: this was for bioshock
12:22 pecisk: right, Witcher 2 is their first one
12:23 karolherbst: right
12:23 karolherbst: ohh maybe here is something https://github.com/virtual-programming/witcher2-linux/issues
12:25 karolherbst: will file a bug there, maybe they respond
12:26 pecisk: actually they do a lot
12:27 pecisk: after getting battered on Steam about releasing broken port, they started to working with community
12:27 pecisk: in result their later OGL 4 games are much better ports
12:27 Yoshimo: official bugtracker on github for a linux port? rather exotic
12:28 karolherbst: they have it for all their games
12:28 karolherbst: https://github.com/virtual-programming/witcher2-linux/issues/153
12:28 karolherbst: I just opend that
12:28 karolherbst: I think they saw it kind of works for vavle
12:28 karolherbst: so why should they do anything else? :D
12:28 pecisk: when witcher 2 issues raised their porter actually started to read forums and created bug tracker by suggestions of community
12:29 pecisk: so they were kinda pushed in that direction, but it turned out well for both company and community
12:29 karolherbst: yeah, its good to have bug trackers for stuff like that
12:30 pecisk: it's a no brainer, especially if your speciality is ports...there will be bugs and you will need good way to report those
12:30 karolherbst: yes
12:31 karolherbst: at least we now that this is a nouvau issue, because it works with intel :D
12:32 karolherbst: imirkin_: maybe shaders are messed up?
12:32 imirkin_: hahaha
12:32 imirkin_: that's like saying "does driver X not support high level shader technique Y"
12:33 karolherbst: the thing is though, it only happens some times
12:33 imirkin_: karolherbst: get me a trace that repros and i can look at it
12:33 imirkin_: until then you're on your own
12:33 karolherbst: ...
12:33 karolherbst: okay
12:33 imirkin_: [or convince the publisher to start giving the game out for free to mesa developers, like valve does]
12:35 karolherbst: mhhh
12:35 karolherbst: on gog you could get money back for 30 days :D
12:35 imirkin_: it's not about the cost
12:35 karolherbst: I know
12:35 karolherbst: I was searching for a demo though
12:36 imirkin_: it's about supporting open-source graphics development
12:36 imirkin_: i don't exactly hold a *lot* of leverage in that arena, but i'll use all the leverage i can
12:37 imirkin_: perhaps soon they'll feel the squeeze on their business, from their linux port (1% of users) not running well on nouveau (1% of that 1%)
12:37 pecisk: well, witcher is on sale constantly for 3€
12:37 pecisk: witcher 2 that is
12:37 imirkin_: pecisk: again, it's not about the money :)
12:38 imirkin_: it's about getting these game publishers to support open source development
12:38 imirkin_: the cost to them would be negligible
12:38 imirkin_: it's not like i'd buy the game anyways (i don't play games), and there are probably a max of 50 people who'd be eligible in the first place
12:39 imirkin_: [at least going by collabora's criteria for handing out the valve package]
12:40 imirkin_: but... i don't have any contacts at any of these, so i just mention it over and over again, hoping that someone will hear.
12:40 glennk: imirkin_, oth one more driver = more QA combinations for them to support
12:41 imirkin_: glennk: so... you're in favor of them *not* providing free access to mesa devs?
12:42 pecisk: I get your point :)
12:42 imirkin_: anyways, for now i say good things about valve and bad things about everyone else
12:42 glennk: imirkin_, no, i'm just saying there's a real cost for them involved
12:42 imirkin_: glennk: there's a real cost anyways
12:43 pecisk: imirkin: as far as I know VP is cool bunch of people it is worth to email them about this :)
12:43 imirkin_: glennk: e.g. someone who guys the game and fixes the driver, same deal
12:43 imirkin_: pecisk: i mentioned it one of the bugs about running on mesa
12:43 imirkin_: pecisk: https://github.com/virtual-programming/bioshockinfinite-linux/issues/27#issuecomment-133496116
12:43 pecisk: I don't think guy who does bug vetting can make that kind of decision :)
12:44 imirkin_: i know
12:44 imirkin_: like i said, i mention it early and often
12:44 imirkin_: the thing is that VP isn't able to really do antyhing about it
12:44 imirkin_: they don't sell games
12:44 imirkin_: they just port 'em
12:44 pecisk: good point
12:44 pecisk: we don't know what kind of deals they have
12:44 imirkin_: it's the game publishers that have to make the program available
12:45 karolherbst: at least I found a workaround for my bug :/
12:45 imirkin_: "don't use nouveau"?
12:45 karolherbst: nope, go into arena, load game
12:45 karolherbst: then its okay
12:46 imirkin_: problem solved!
12:46 imirkin_: do you get messages about shaders being evicted?
12:46 imirkin_: there's sort of a bug relating to that
12:46 imirkin_: but... it'd be *very* rare that you'd hit it
12:46 karolherbst: imirkin_: like that? https://gist.github.com/karolherbst/119c19f6397c317c6a8e
12:47 imirkin_: no, not like that
12:47 imirkin_: FTR, i think those "not uniquely defined" messages are BS
12:47 karolherbst: anyway, that's the output
12:48 imirkin_: what's the rejected pushbuf?
12:48 imirkin_: what's in dmesg?
12:48 imirkin_: that's weird -- no pushbufs in the submit...
12:48 karolherbst: Unknown handle 0x00000420
12:48 karolherbst: validate_init
12:48 karolherbst: validate: -2
12:48 imirkin_: hrm ok
12:53 pecisk: karolherbst: so you need any other testing from me? Also good question is when do you think first iteration of your current reclocking code will land in Nouveau? I think is it worth to hype it a bit
12:53 karolherbst: 4.4
12:53 karolherbst: got a kind of confirmation from skeggsb
12:54 karolherbst: I try to finish my pcie work for 4.5
12:56 pecisk: karolherbst: first 2016 release? I kinda lost track how frequent releases are
12:57 karolherbst: every ten weeks in avarage
12:57 pecisk: ahh cool
12:59 pecisk: karolherbst: also how far reclocking is supported? I mean Kepler can be explained as something in progress. I guess rest of families are bit different?
13:00 karolherbst: for gddr5 there isn't that much besides kepler
13:00 karolherbst: maxwell is pain
13:00 karolherbst: fermi is pain too
13:00 karolherbst: so
13:00 karolherbst: that settles that
13:00 imirkin_: karolherbst: there's my GT215 with GDDR5 ;)
13:00 pecisk: karolherbst: you got answer on your ticket
13:00 karolherbst: I already answered
13:00 pecisk: quick response
13:01 pecisk: nice
13:02 mupuf_: karolherbst: why is maxwell a pain?
13:02 pecisk: karolherbst: is it worth to draw maxwell/fermi card owners into testing? Would that help to get it moving?
13:02 mupuf_:thought that no-one but me even tried it
13:02 karolherbst: ohh right, reclocking works on maxwell?
13:02 imirkin_: pecisk: fermi and maxwell aren't ready for any sort of testing of anything
13:03 pecisk: I see
13:03 karolherbst: mupuf_: I tried on reator
13:03 karolherbst: mhhh
13:03 mupuf_: well, the clock tree parsing may be wrong
13:03 karolherbst: I just pushed both cards to 0f and nothing happened
13:03 karolherbst: so
13:03 mupuf_: ah ah
13:03 imirkin_: lol
13:03 mupuf_: well, I did crash my GPU when upclocking once
13:03 karolherbst: and with nothing I meant no error
13:03 mupuf_: so it definitely did something
13:03 karolherbst: :D
13:03 mupuf_: probably because the voltage did not get raised
13:03 karolherbst: but I was using my tree
13:03 mupuf_: because of my silly tree
13:03 mupuf_: silly mistake*
13:04 karolherbst: mupuf_: did it crash even without X running?
13:04 mupuf_: yes
13:04 mupuf_: but hey, no-one had a look at maxwell reclocking AFAIK
13:04 karolherbst: mhhh
13:04 karolherbst: I mean I really pushed both cards to 0f
13:05 karolherbst: and no card crashed
13:05 karolherbst: I also did it more than once
13:05 mupuf_: oh, well
13:05 mupuf_: very good then :)
13:05 karolherbst: but I had my gddr5 fix in then
13:05 mupuf_: then maxwell may just work
13:05 karolherbst: hopefully
13:05 karolherbst: :D
13:05 mupuf_: and voltage management was the only hting missing
13:05 karolherbst: :)
13:06 mupuf_: the nve6 is not happy thoug
13:06 mupuf_: the screen gets corrupted when upclocking
13:06 karolherbst: yeah, there are issues like that
13:06 karolherbst: I am aware of them
13:06 karolherbst: but as long as the card does not crash, its fine
13:06 karolherbst: you can simply reclock until it works
13:06 karolherbst: :D
13:06 mupuf_: that means more PFB-related-black-magic is needed
13:06 mupuf_: lol
13:06 karolherbst: skeggsb said something bout isohub
13:06 mupuf_: oh, possible too
13:07 mupuf_: I will see ben tonight
13:07 karolherbst: Karlton had the same issue I think?
13:07 mupuf_: ben == skeggsb
13:07 karolherbst: sometimes screen messed up and sometimes it just worked
13:07 karolherbst: okay
13:07 mupuf_: we will likely talk about PM
13:08 mupuf_: he will likely be busy with maxwell enabling when the microcodes land
13:08 mupuf_: but in the mean time, he may want to work on reclocking.
13:08 mupuf_: I will give another shot at clock gating soon
13:09 mupuf_: and will need to have a look at the voltage mapping table too
13:09 pecisk: karolherbst: last question - is it worth to point people to your repo's direction if they want to try it out/test it for Kepler cards?
13:10 karolherbst: gddr5 branch then
13:10 karolherbst: but the more testing the better the patch :D
13:10 karolherbst: I think its in pretty good state now
13:10 pecisk: okey :)
13:10 karolherbst: even skeggsb gave his okay for it
13:11 pecisk: karolherbst: I will drop a thread in linux_gaming on reddit
13:11 karolherbst: mupuf_: I don't think the mapping table is the thing which goes wrong
13:12 mupuf_: what else then?
13:12 karolherbst: but rather the voltage table
13:12 mupuf_: really?
13:12 karolherbst: this check fails volt->vid[i].uv == uv
13:12 karolherbst: basically ouveau finds an entry with the requested voltage
13:12 karolherbst: if it does not find one
13:12 karolherbst: it fails
13:13 karolherbst: okay, second argument
13:13 karolherbst: "-- ID = 66, link: 6, voltage_min = 1225000, voltage_max = 1250000 [µV] --" these values seem to be decent enough
13:13 karolherbst: could be wrong, but I don't think so
13:13 karolherbst: voltage table range is only [825000, 1212500]
13:13 karolherbst: so 1212500 is the max nouveau will do at all
13:14 karolherbst: both cstates go higher then that
13:14 karolherbst: => fail
13:14 karolherbst: it could be that the entries in the voltage map table might be wrong or something, but mhhh
13:14 karolherbst: we should check what is the highest voltage used on blob on his card
13:14 karolherbst: if its above 1212500, then the voltage table is not right
13:15 karolherbst: pecisk: do you know how to read the voltage out with the blob driver?
13:15 pecisk: karolherbst: nope
13:15 pecisk: never done that
13:15 karolherbst: I am bad with the gpios, maybe mupuf_ can explain it in a simple way? :D
13:16 karolherbst: pecisk: basically we need the highest voltage the blob uses on highest clocks
13:16 mupuf_: karolherbst: I think the problem is in the voltage mapping table
13:16 mupuf_: or maybe the two highest cstates are simply to be discarded because we cannot set the voltage high-enough (more likely)
13:16 karolherbst: you know "-- Mode GPIO, Base voltage 825000 µV, voltage step -12500 µV, acceptable range [825000, 1212500] µV --"
13:16 karolherbst: ohh, could be
13:16 karolherbst: pecisk: please check the highest clock on blob :D
13:17 mupuf_: pecisk: what do you need to know?
13:17 karolherbst: mupuf_: how to read voltage on blob
13:17 mupuf_: ah
13:18 mupuf_: I did write the code to read the vid out of the gpio state, haven't i?
13:18 mupuf_: Maybe not, actually
13:19 mupuf_: you know what, it should be a tool in nouveau/bin
13:19 mupuf_: to read back both the clock and the voltage
13:19 mupuf_: and nothing else
13:33 karolherbst: what's libmenu.so.5 ?
13:34 karolherbst: mupuf_: do you mean nv_perfmon ?
13:34 karolherbst: maybe something else
13:36 karolherbst: mupuf_: I think your pwr tool read the VID out, but somehow the value wasn't stable
13:36 karolherbst: pecisk: anyway, could you check what the highest clock is used by the blob?
13:37 pecisk: karolherbst: how to do that :)
13:37 karolherbst: nvidia-settings
13:38 karolherbst: it will tell you
13:38 pecisk: ok
13:38 pecisk: reboot then
13:40 mupuf_: karolherbst: well, the read is not atomic
13:40 mupuf_: so if it changes while I read, it is going to end in tears
13:40 karolherbst: I meant it wasn't stable with nouveau either
13:44 pecisk: karolherbst, http://fpaste.org/267653/34986114/
13:44 pecisk: from PowerMizer screen
13:49 karolherbst: mupuf_okay, seems like the both highest cstates are getting used
13:53 karolherbst: pecisk: okay, idea
13:53 karolherbst: nvidia-settings --query all | grep -i voltage
13:55 pecisk: http://fpaste.org/267657/35052614/
13:56 pecisk: karolherbst ^^
13:56 karolherbst: ohh nice
13:56 karolherbst: then put some really load on that card
13:56 karolherbst: *real
13:56 karolherbst: like run glxgears without vsync :D
13:57 karolherbst: and make sure the core runs at max clock
13:57 karolherbst: and then read the voltage out again
13:57 pecisk: karolherbst, TF2 on max settings maybe? It always make my card go zummmmmmm
13:58 karolherbst: why not
14:01 pecisk: karolherbst, max usage for both core and memory same voltage
14:02 karolherbst: so it stucks with 1200000
14:02 karolherbst: strange
14:02 karolherbst: seems wrong though
14:02 karolherbst: it should use less on lower clocks
14:03 karolherbst: pecisk: does nvidia-smi print anything?
14:03 pecisk: don't have that
14:03 pecisk: yeah, it seems to stuck to it no matter what I do, that's certainly a bug
14:06 karolherbst: mhh
14:07 karolherbst: somehow you can read the VID out from the gpios but somehow I can't do that and I don't find the tool mupuf_ was talking about
14:13 dylanetaft: Hey - what does trapped write page not present mean? Running into a somewhat reproducable hard lock with nouveau outside of 3d rendering
14:13 karolherbst: ohh wait
14:14 pecisk: karolherbst, it's in xorg-x11-drv-nvidia-cuda package for Fedora
14:14 dylanetaft: on 4.3 RC
14:14 Boohbah: i bisected my fan speed regression
14:14 karolherbst: mupuf_: is this the same on all cards or does it have to be adjusted? https://github.com/mupuf/pdaemon_trace/blob/master/pwr_read/pwr_read.c#L466-L471
14:14 Boohbah: should i post on freedesktop.org bugzilla, or kernel.org?
14:14 imirkin_: Boohbah: fd.o probably
14:15 imirkin_: Boohbah: xorg -> Driver/nouveau (there's no dedicated section for the drm piece unfortunately)
14:15 karolherbst: mupuf_: https://github.com/virtual-programming/witcher2-linux/issues/153#issuecomment-140535458
14:15 karolherbst: :D
14:15 karolherbst: ..
14:15 karolherbst: imirkin_
14:15 karolherbst: mupuf_: sorry
14:15 karolherbst: yeah why not, just gave around drm free binaries :D
14:16 Boohbah: imirkin_: i was wondering about that, thanks
14:17 pecisk: karolherbst, it seems driver has broken voltage report, as nvidia-smi give me empty space where voltage should be
14:18 dylanetaft: and fail set_domain validating bo list validate: -22
14:26 karolherbst: smart ass comments are the best
14:30 karolherbst: pecisk: you could try the one tool out from mupuf_ though
14:30 karolherbst: maybe it works
14:30 karolherbst: pecisk: clone this one: https://github.com/mupuf/pdaemon_trace
14:31 Boohbah: dylanetaft: i had a bug like that too
14:32 karolherbst: Boohbah: compiz?
14:32 Boohbah: karolherbst: kwin compositor
14:32 karolherbst: mhh
14:32 Boohbah: dylanetaft: https://bugs.freedesktop.org/show_bug.cgi?id=91598
14:34 Boohbah: karolherbst: i mean plasmashell
14:40 pecisk: bless Fedora new feature of showing long console tasks completed in notifications
14:41 karolherbst: :D
14:41 karolherbst: pecisk: then cd in nouveau
14:41 karolherbst: and run autogen.sh
14:42 karolherbst: and make
14:44 karolherbst: ohh how funny
14:44 karolherbst: imirkin_: without GL_ARB_buffer_storage I don't get the issue?
14:48 karolherbst: imirkin_: how likely is it, that GL_ARB_buffer_storage is broken?
14:52 pecisk: karolherbst, already on it but it requires envytools
14:54 glennk: karolherbst, the likeliness of a GL extension being broken goes up as piglit test coverage goes down
15:05 karolherbst: calim: hi, I took a look at the sched stuff and wanted to ask you two things if you are up to
15:07 karolherbst: imirkin: maybe its a application bug and they fixed that
15:09 karolherbst: who can change the optimus page on freedekstop by the way?
15:11 pecisk: karolherbst, ok, I compiled it, which tool I shall use?
15:13 karolherbst: okay, so witcher 2 is able to bring my entire system down
15:14 pecisk: karolherbst, I compiled pdaemon_trace, what's next? :)
15:14 karolherbst: pecisk: pwr_read
15:14 karolherbst: open pwr_read.c
15:14 karolherbst: search for nv_init
15:14 karolherbst: and change the path
15:14 karolherbst: and then compile that file like told to you in the header
15:14 karolherbst: + -lnvhw
15:15 karolherbst: imirkin: any idea? nouveau 0000:01:00.0: fifo: read fault at 000790f000 engine 1b [CE2] client 18 [GR_CE] reason 02 [PTE] on channel 2 [00bf890000 witcher2[6826]]
15:16 karolherbst: nouveau 0000:01:00.0: fifo: ce2 engine fault on channel 2, recovering...
15:16 karolherbst: after that my entire drm stack goes strange and upset
15:21 pecisk: karolherbst, it complains about missing stropts.h
15:21 karolherbst: ohhh, wait
15:22 karolherbst: do you have glibc dev installed?
15:22 zlice: why does nouveau take so long to start x? is there a config or something i'm missing
15:22 karolherbst: zlice: how long is "so long"?
15:23 zlice: karolherbst: minutes
15:23 karolherbst: and did you try something else out besides nouveau?
15:23 zlice: nvidia
15:23 zlice: but...nvidia does something when i start a VM and prevents me from rebooting, leading to crashes - and i had to rebuil today
15:25 karolherbst: pecisk: sadly I have to sleep now, too damn tired :/
15:25 karolherbst: but if you get this tool compiled you should be able to run it and check the VID coloumn
16:58 calim: where are the questions ?
17:00 imirkin_: where aren't they :)
17:00 imirkin_: i'm full of 'em
17:01 calim: if I'm rude it's because I'm rather frustrated atm
17:01 imirkin_: well, there was the RA thing
17:01 calim: with people, not code
17:01 imirkin_: there's also a depth thing
17:01 imirkin_: (or i think it's a depth thing)
17:02 calim: I wanted the sched questions
17:02 imirkin_: hm, karol doesn't seem to be present atm
17:02 imirkin_: but he was playing with a scheduling pass
17:03 calim: that should be frustrating, too xD
17:03 imirkin_: basically there were questions about wtf getLatency() and getThroughput() were
17:03 imirkin_: and how they should be factored into things
18:06 marcosps1: hi guys :)
18:06 marcosps1: imirkin: http://pastebin.com/NXK2F6Up
18:07 marcosps1: so, insn here is a ld, so I need to test src->insn->op to verify is this is a merge and this try to get the value from insn. This makes sense, right?
18:10 marcosps1: imirkin_: in this case this is crashing for me... =/ I'm checking right now why this happens. It seems asImm() is failing here.
18:41 mrnuke: imirkin: dear, sir, I come gifting bears: http://paste.fedoraproject.org/267731/23675581
19:10 marcosps1: imirkin: around..?
19:13 mrnuke: yeah imirkin! come out of hiding, ya' partying drunk hacker!
19:13 marcosps1: mrnuke: :)
19:16 marcosps1: mrnuke: You're trying to hack on mesa too?
19:16 marcosps1: mrnuke: Sorry, not trying, but actually hacking?
19:20 mrnuke: marcosps1: no, I just want to get my maxwell card to stop crashing when starting KDE :)
19:21 marcosps1: mrnuke: :)
19:21 mrnuke: CamelCase is not my thing honestly. I'm more of a kernel coding style person
19:22 marcosps1: mrnuke: Do you have some code in kernel too?
19:22 mrnuke: no. I gave up on trying to merge the patches. The emailing patches in plaintext thing is bullshit
19:23 mrnuke: we have really grea tools like git<sitename>.[org|com], but those bastards still use email
19:23 marcosps1: mrnuke: Not actually :P
19:23 mrnuke: it's such a pain in the arsenal
19:24 marcosps1: mrnuke: Yes, git you just need to configure your email to not change tabs to spaces...
19:24 mrnuke: I can use git-email and friends, but I don't like to
19:24 mrnuke: I find that too clumsy
19:24 mrnuke: it gets out of the git ecosystem, and that sucks
19:25 mrnuke: git push to public server , hey, here's my changes, bro
19:25 marcosps1: mrnuke: How, I like git-email :)
19:25 marcosps1: at least this works for me :)
19:25 mrnuke: works for every other free software project out there
19:25 marcosps1: mrnuke: of course, have a gerrit place is easier, but ... :P
19:25 mrnuke: I use a Model M, but I'm not _that_ old school
19:26 mrnuke: gerrit sucks, BTW
19:26 marcosps1: mrnuke: how, I love it :)
19:27 mrnuke: though github got a lot of hings right
19:27 marcosps1: mrnuke: LibreOffice works great with gerrit..
19:27 mrnuke: "you, mister, here's the github branch I'd like you to merge. kthxbye"
19:27 mrnuke: so freaking easy
19:27 marcosps1: mrnuke: sure, it's nice and light, but, TDF has their own infracstructure, so....
19:28 mrnuke: I don't mind git pushing to any public repo then asking for a merge. It doesn't have to be github
19:29 mrnuke: but you can't really git push to email. That shit sucks
19:31 marcosps1: mrnuke: I don't like send patches by email too, this type of things get missed easily. But, I think we could use something like gerrit to handle patches...
19:32 mrnuke: even gerrit is better than git-email
19:32 marcosps1: mrnuke: But, I'm not SO upset to send a patch by email. If they don't reply after some days, I ask them :)
19:32 mrnuke: I've given up on being ignored
19:33 mrnuke: I had this patch one. Everyone who needed that functionality loved it
19:33 mrnuke: fucking subsystem maintainer couldn't give a flying fuck
19:34 marcosps1: mrnuke: this is anothr problem of using ML to receive patches.
19:35 marcosps1: mrnuke: I gave u talking with Ilia today and I sent him an email. Maybe you need to do this too :)
19:36 mrnuke: I don't have the time to harass linux maintainers about merging my patches
19:36 mrnuke: In coreboot, I go drink beer with my buddies and tell them what I want merged
19:36 mrnuke: problem solved
19:38 marcosps1: mrnuke, imirkin is not your buddy? He's a nice guy, and pushes a lot of things related to nouveau :)
19:39 mrnuke: I just met him a few days ago (IRC, not in person)
19:39 mrnuke: he convinced me to NOT buy the GPU I bought
19:40 marcosps1: mrnuke: but, think about it, you can help to solve a lot of problem now that you have a problematic GPU :)
19:42 mrnuke: My goal is to build a blob-free workstation
19:42 mrnuke: if I can log into KDE with 3x4k screens, then I'm good
19:43 mrnuke: of course, getting coreboot to run on the hardware is the hard part, but the libreboot guys are already taking care of that and paying a developer to do the work
19:43 mrnuke: imirkin: BTW, just upgraded _all_ packages to rawhide, and I can log in KDE with 2x4K displays
19:44 mrnuke: imirkin: as you said, performance sucks more than my ex, but it works(ish)
19:44 mrnuke: marcosps1: I was foolish enough to spend $400 on a Quadro K1200
19:44 mrnuke: four DP outputs though
19:44 marcosps1: mrnuke: :O
19:44 mrnuke: that shit's cool
19:45 marcosps1: this is a lot of money...
19:45 mrnuke: I work for Intel. I get paid enough where I won't cry for more than a few weeks about the price
19:46 marcosps1: mrnuke: sorry mr money :)
19:46 marcosps1: mrnuke: Some time ago I had the dream to work for a company like Intel, IBM, Red Hat, or something related...
19:47 marcosps1: mrnuke: But now I'm quite happy where I'm working on, but, I really like the way helpes Linux and Open Source in general...
19:48 mrnuke: where do you work?
19:50 marcosps1: mrnuke: http://www.zpesystems.com/
19:51 mrnuke: I used to work for google
19:51 marcosps1: mrnuke: o, you work for Intel to improve coreboot?
19:51 mrnuke: but I was doing proprietary firmware for some shit that would be obsolete in a few years
19:52 mrnuke: I wanted something better
19:52 mrnuke: so I fooled Intel into letting me do coreboot work
19:53 mrnuke: but improving coreboot is a different matter. Intel has other ideas
19:53 mrnuke: at least I know what I'm talking about
19:54 marcosps1: mrnuke: You seems to have a lot of experience. How old are you? (I don't know if asking your age is polite :) )
19:54 mrnuke: officially 26
19:54 mrnuke: but I'm transylvanian. I'm actually over 2500 years old
19:55 marcosps1: :)
19:57 mrnuke: I don't think there's much I can do to improve coreboot, but I've had the chance to hack on some nice tools, like flashrom
19:57 marcosps1: mrnuke: Nice!
19:57 marcosps1: mrnuke: You have a nice job too :)
19:58 marcosps1: mrnuke: coreboot has a public git repo? I'm curious now :)
19:59 mrnuke: coreboot.org man
19:59 mrnuke: or you can google that shit yourself
19:59 mrnuke: git clone, then "make menuconfig" then "make"
20:00 marcosps1: mrnuke: I think you can polite man.
20:00 marcosps1: *you can be a little more polite
20:01 marcosps1: ;)
20:01 mrnuke: I know I can, but why? We're all hackers. We all like beer, all-night hacking and swearing :)
20:02 marcosps1: I tried coreboot cgit, but I didn't found nothing, so I asked you ;)
20:03 mrnuke: allow me to blow your mind: http://review.coreboot.org/
20:03 marcosps1: HUahUAHU
20:03 marcosps1: mrnuke: Yes, I already found it here :)
20:07 Karlton: the only work station supported by libreboot is that old AMD server board
20:08 mrnuke: Karlton: I have more intel on libreboot than most people
20:08 mrnuke: there's a shiny new dual G34 board in the works
20:09 mrnuke: and the guy behind libreboot is paying a commercial dev to do the coding for that
20:10 mrnuke: more than my yearly salary
20:11 mrnuke: so yeah, I put my K12oo with free microcode on that server board... evil plan complete
20:14 Karlton: yeah maybe some more AMD boards are possible, but I was told anything Intel past 2008 would be unlikely to get support in libreboot
20:15 mrnuke: intel is a whole different beast
20:16 mrnuke: too many matryoshka processors
20:25 Karlton: The hardware checks the firmware for a signature and modern Intel AMT/ME firmware is also encrypted and protected with patent laws that forbid REing in some counties :/
20:26 mrnuke: RE is free game in the US
20:26 mrnuke: but the crypto stuff is a deal breaker
20:27 Karlton: s/counties/countries/
20:27 mrnuke: hey, I need to sign my development imges for coreboot
20:27 mrnuke: in windows
20:28 mrnuke: that shit sucks more tha a hoover vac
20:33 fling: Hello.
20:35 fling: mrnuke: is userspace 1.0.11 new enough for GK208?
20:37 mrnuke: what's "userspace 1.0.11" mean?
20:37 fling: xorg nouveau driver
20:39 mrnuke: I have 1.0.12 here
20:39 mrnuke: enough to start KDE
20:40 fling: libdrm-2.4.64; mesa-10.3.7 ?
20:40 mrnuke: but performance is well, lacking
20:41 fling: I want to make it not to crash when I open an image in a browser this is why I'm upgrading ;P
20:41 fling: Should I try the latest versions?
20:41 mrnuke: libdrm 2.4.61 and mesa
20:41 mrnuke: 10.5.4
20:46 mrnuke: http://gtech.myftp.org/~mrnuke/yeah_right.png
20:46 mrnuke: classic :)
20:48 fling: mrnuke: haha
20:56 Karlton: use youtube-dl :)
20:57 fling: or youtube-viewer
21:03 fling: is d3d9 working again btw? :P