00:43bartm0n: i've had this issue for a while now - my 2nd monitor will not work when using the nouveau driver. It is detected by xrandr, it's possible to set mirroring but the monitor remains in standby. It works when using nvidia's proprietary driver...
00:43bartm0n: currently using Fedora 23: 4.2.7-300.fc23.x86_64, card's family is NV50, chip G94
00:46bartm0n: also xorg-x11-drv-nouveau-1.0.12-1.fc23.x86_64 :)
03:17RSpliet: bartm0n: mind pasting your dmesg to a paste website of choice and sharing the URL here?
03:17RSpliet: and your xorg.0.log
07:26mupuf: imirkin: I was close then :D It is just an off-by-one in the name :D
08:35innervision: hello, i have serious xserver crashes every 2 hours with the message [ 3549.739575] nouveau E[gnome-shell] failed to idle channel 0xcccc0000 [gnome-shell]
08:36innervision: can you help me with that please?
08:38innervision: I use a GeForce GTX 660 together with debian 8.2 (kernel 3.16.0-4-amd64)
08:39RSpliet: innervision: although I'm not sure whether there is a solution for this, first step should always be running modern software
08:39Tom^: for starters get a kernel that isnt soon 2 years old :/
08:40innervision: but I did a fresh debian install some days ago
08:40RSpliet: innervision: we're not debian. The only kernel we can attempt to support is 4.3
08:41innervision: ok so switching to the newest ubuntu could solve my problem?
08:41RSpliet: innervision: Ubuntu and Debian are notorious for shipping old software under the flag "stable"
08:41RSpliet: ask #debian to see what they can suggest wrt. installing a more recent kernel and userspace
08:41innervision: ok I didn't know that
08:42RSpliet: (kernel is one, but you probably want the libdrm, ddx and mesa updated as well)
08:42innervision: but it's everything else than 'stable'
08:42innervision: ok, thank you very much
08:42RSpliet: that can easily end up being either a philisophical discussion or one that ends up bashing distributions ;-)
08:42huehner: innervision: check debian stable-backports repo, in there you can i.e. find 4.2.6 kernel packaged
08:43huehner: innervision: https://tracker.debian.org/pkg/linux
08:43Tom^: 4.2.6 is still old imo :P
08:44huehner: ;) just wait some short time and it will get 4.3 as current debian sid has now
08:44huehner: just easier for most people compared to compiling themselves
08:45RSpliet: Tom^: let's not go there
08:46innervision: maybe I should just switch to sid or the debian testing release
08:46RSpliet: innervision: that 4.2 kernel might already bring a few important fixes in (weren't there some playlist fixes along the time? can't remember)
08:47innervision: the strange thing is, I had no problems with debian wheezy just a while ago and the exact same hardware
08:47imirkin: mupuf: -ENOCONTEXT
08:47RSpliet: innervision: with older kernels you were most likely not having any acceleration at all with nouveau and instead falling back to software rendering and such
08:48RSpliet: in my experience cards like yours run fairly stable using recent versions of the various nouveau components
08:48imirkin: innervision: make sure to update kernel and mesa. if you're using a distro that doesn't make that easy, try to get help from a distro support channel. we can't support anything but the latest versions.
08:50innervision: ok but that info helps me a lot - thanks!
09:18mupuf: imirkin: GF106 vs GF116
09:18imirkin: ah right
09:29karolherbst: I am always suprised how much cleaning the fan and cooling pipes help..
09:48karolherbst: mupuf: new version with clamping: https://github.com/karolherbst/nouveau/commit/175d8c2521bcacd8780d6ce15a022b4ce5e99e0a
10:09karolherbst: I think 125 is too much for fermi/keplers
10:09karolherbst: never saw anything above 115 on them, so I change that to 115
10:11mupuf: I am pretty sure the condition on the chipset will not be accepted upstream
10:11karolherbst: but I also saw 115 only on a nva0 card
10:11karolherbst: but I think this should be fine anyway
10:11karolherbst: there is a deeper issue anyway I think
10:11karolherbst: I found like three different "kinds" of fsrm configuration types and didn't found a way to differ between them yet
10:12mupuf: have to go, tell me more about it in the mean time
10:12mupuf: maybe I can help
10:12karolherbst: yeah, I will try to create a nice overview
10:18karolherbst: well for nvc0-nve7 there seems to be only two types, but they are not seperated by chipset :/
10:18karolherbst: mupuf: https://gist.github.com/karolherbst/1854da8068b510d4cc14#file-gistfile1-txt-L33-L40
10:18karolherbst: there seems to be a three and a two stage fsrm somehow
10:18karolherbst: maybe it is the same, but can be configured to be either one
10:18karolherbst: but I didn't manage to get my "two stage" on to behave like a "three stage" one
10:19karolherbst: the only difference I found so far is, that bits 3-5 are set in 0x02010c for the lowest threshold, but setting it on my gpu doesn't change a thing
10:20karolherbst: and I would need to play on a gpu with 0x02010c set to 0x00000049 to figure out the third type
10:21karolherbst: mlankhorst: your nvc8 has 0x02010c set to 0x00000049 right?
10:22karolherbst: mupuf: and your nvc4 too as it seems
10:52mlankhorst: karolherbst: no idea about that one, only have whats in the archive
12:28RSpliet: I'm seeing exactly this
12:29RSpliet: apparently mesa uses sha1 support somewhere since relatively recently, depending on openssl;
12:32karolherbst: ohhh wait
12:33karolherbst: RSpliet: you know that shader-cache compile flag?
12:33karolherbst: maybe it is related?
12:34karolherbst: RSpliet: but without a call stack, it can be whatever
12:38RSpliet: karolherbst: I know...
12:39RSpliet: it's more a general frustration that every time I try to rebuild Mesa, new stuff goes wrong
12:39RSpliet: I like code sharing, but having an OpenSSL-kitchen-island cluncked on feels a bit of overkill
12:40RSpliet: if it's not an SELinux problem, I might just rebuild with a different sha1 lib
13:14prototron: I have a video card problem that I am trying to solve for days now... so after reading FAQ and troubleshooting guide, I ended up here
13:16prototron: So I have a Lenovo Y400 notebook which has a Geforce GT 750M GPU. So Debian 8 (Jessie) installed nouveau driver by default (which they highly recommend anyway over the proprietary drivers from NVidia)
13:18prototron: Then at boot time I get this funky colorful stripe near the top of the screen, and it hangs forever
13:20RSpliet: prototron: are you running a 4.3 kernel?
13:21karolherbst: prototron: but this is a dual gpu laptop, right?
13:21bartmon: is there any known archive of this channel?
13:21karolherbst: bartmon: read the topic
13:25prototron: kernel 3.16.0-4-amd64
13:25prototron: the laptop has the intel HD 4000 chip + the Nvidia 750m GPU
13:26karolherbst: right, so the intel gpu will drive your desktop
13:26karolherbst: most likely
13:26karolherbst: could you do a screenshot and check if the line is in there?
13:26karolherbst: prototron: and when did the line appear the first time?
13:27prototron: well it is a fresh install of Debian
13:27karolherbst: it may be, that the display itself is a bit bricked, that's why I asked since when the line appears
13:28prototron: if that can help, I am dual booting with Windows 8 and the GPU works flawlessly
13:28karolherbst: ahh okay, so this issue only occurs on the linux side
13:29prototron: I also got it working on Ubuntu 14.04 using the proprietary drivers from NVidia
13:29karolherbst: yeah, but the intel gpu is displaying your desktop
13:29karolherbst: not the nvidia one
13:29karolherbst: except you have a fancy bios and disabled intel output
13:30bartmon: karolherbst, well i feel embarrassed. thank you. :9
13:31prototron: and right now I booted Debian using nomdoeset
13:31karolherbst: remove that
13:31karolherbst: well you can leave it if the display stays black without it
13:31karolherbst: but I doubt that
13:32prototron: this is the only way to boot linux
13:32prototron: otherwise, funky stripe + hangs
13:32karolherbst: then something is really broken
13:33karolherbst: prototron: could you get us a dmesg and xorg log without the nomodeset option?
13:33prototron: I have the kern.log file, is that ok?
13:34karolherbst: you could show it though
13:34karolherbst: I will complain if it isn't
13:35karolherbst: RSpliet: but why did you ask for 4.3?
13:36imirkin_: RSpliet: there's a way to get it to not use libssl
13:36RSpliet: karolherbst: because I refuse to debug set-ups that are not up to date
13:36RSpliet: debian is a trigger word for "update your kernel, then we'll talk" ;-)
13:37RSpliet: imirkin_: looking into it yes
13:37karolherbst: RSpliet: :D
13:38karolherbst: RSpliet: yeah but this sounds like an intel issue or bricked system configuration anyway
13:41prototron: karolherbst, ok i'm sending you the log file
13:41karolherbst: :O you can send file through IRC
13:43karolherbst: well I accepted it, but it timed out
13:43prototron: hum shit
13:44karolherbst: I think your sender IP was fulty
13:44karolherbst: cause I got 0.0.0.190
13:44karolherbst: well you could also pastebin it somewhere
13:46imirkin_: skeggsb: not sure when i'll have time to look over your patches in depth again... but iirc i was happy with your older version, so don't block on me
13:47prototron: karolherbst, http://paste.ubuntu.com/14120955/
13:50karolherbst: prototron: well I would need a log from a boot without nomodeset
13:50prototron: thought that was it
13:51karolherbst: but this is interessting: "[Firmware Bug]: Duplicate ACPI video bus devices for the same VGA controller, please try module parameter "video.allow_duplicates=1"if the current driver doesn't work."
13:51karolherbst: what is this?
13:51karolherbst: prototron: ohh wait
13:52karolherbst: there is aboot without nomodeset in the log
13:52karolherbst: my mistake
13:52karolherbst: prototron: did you add "video=1366x768@60" ?
13:52josla972: RSpliet: Hi, did you read my feedback about the reclocking patches?
13:53RSpliet: josla972: no I didn't sorry
13:53josla972: RSpliet: ok
13:53josla972: I have a G84 card
13:53josla972: so I changed 0x94 -> 0x84 in your last patch
13:53RSpliet: oh yes, never looked at those, can't imane it works
13:53prototron: karolherbst, I did try it. First time I booted, it had the bug I am talking about. So I checked the logs and saw that it tried to boot with a 1080 resolution, which is not supported by my laptop display
13:54karolherbst: wow strange
13:54josla972: RSpliet: ok, I managed to reclock everything but memory
13:54prototron: karlmag, then I tried video=1366x768@60
13:54prototron: i mean karolherbst
13:54RSpliet: josla972: hmm yes, but the perf difference is probably minimal
13:54josla972: RSpliet: I see
13:54karolherbst: "fbcon: nouveaufb (fb0) is primary device" :O
13:54josla972: would being able to reclock memory involve a lot of work?
13:55karolherbst: that shouldn't happen
13:55karolherbst: yeah I think I got the issue now
13:56karolherbst: prototron: could you give me the output of "lspci | grep VGA"?
13:56prototron: karolherbst, 01:00.0 VGA compatible controller: NVIDIA Corporation GK107M [GeForce GT 750M] (rev a1)
13:57karolherbst: prototron: is there some option in the vbios to disable the intel gpu or to set the "main" gpu to either intel or nvidia?
13:58prototron: karolherbst, hmmm i don't know I would have to verify
13:58karolherbst: yeah please do, because the kernel doesn't seem to see your intel gpu
13:58prototron: karolherbst, I THINK no, but I am not 100% sure, not even 90% :P
13:58karolherbst: the issue is this
13:58karolherbst: you need something from a newer kernel
13:58karolherbst: 3.16 is too old
13:59karolherbst: so if you plan to use nouveau as your primary driver, you need to update your kernel
13:59karolherbst: I don't understand why the intel gpu doesn't even appear
13:59karolherbst: you should check your bios
14:00prototron: karolherbst, Ubuntu 14.04 uses kernel 4.3 right?
14:00RSpliet: josla972: quite a bit yes, most importantly I simply don't have a card like that
14:01karolherbst: prototron: well check your bios first, because you want to use the intel gpu as your main one
14:01josla972: RSpliet: ok
14:01karolherbst: because of battery life
14:02imirkin_: RSpliet: perhaps you could help him collect the info you need?
14:03prototron: (Ubuntu 14.04 uses 3.19)
14:04karolherbst: imirkin_: when was the c800 workaround added through an option? 4.3?
14:04prototron: ok hang on, rebooting to check my BIOS
14:05imirkin_: karolherbst: yes
14:05imirkin_: karolherbst: default to on in 4.4-rc5
14:06karolherbst: default on for all keplers?
14:06josla972: yes, if you need me to extract info about my card I can maybe be of help
14:07karolherbst: ahh PGOB fuse
14:15prototron: karolherbst, so no option in the BIOS to choose between intel or nVidia
14:17karolherbst: prototron: k, then your entire lspci output
14:18karolherbst: pls pastebin it somewhere
14:19karolherbst: yeah, your intel gpu is totally missing
14:20imirkin_: karolherbst: i think i've seen that on those lenovo Y series... some of them are even dual-nvidia gpu
14:20karolherbst: prototron: then boot with "video.allow_duplicates=1"
14:21karolherbst: imirkin_: check his cpu
14:21karolherbst: it is a 3830MQ intel one
14:21karolherbst: I meant 3630
14:21imirkin_: the one i saw it on *should* have had an intel gpu, but it was fused off
14:21karolherbst: oh god why
14:21imirkin_: i try not to ask such questions
14:22karolherbst: seriously, somehow I fell like never buying me a lenovo laptop :D
14:22imirkin_: buy the T series.
14:22prototron: they run pretty good on Windows, but nothing but problems on linux
14:22prototron: specially the gpu
14:22karolherbst: no, lenovo died for me and not only because of their laptops
14:23karolherbst: I won't forget that root certificate thing they did :D
14:23Wonka: .oO( lenovo died for your sins? )
14:23karolherbst: prototron: do you know if the intel gpu is "there" with windows?
14:24karolherbst: anyway, laptop with 750M gpu is a bit instane for battery life
14:24prototron: The dedicated GPU (750M) is disabled when not plugged in with DC adapter
14:24imirkin_: i think the Y series isn't so much laptop as luggable :)
14:25imirkin_: prototron: errr... really? you don't appear to have any other gpu in the system
14:25karolherbst: prototron: sooooo
14:25karolherbst: prototron: what happens when you boot linux without the adapter plugged in?
14:25karolherbst: does the intel gpu magically appears then?
14:25karolherbst: in lspci
14:26prototron: never tried this
14:27karolherbst: why use prime when you can hack something together yourself :D
14:27karolherbst: ohh wait
14:27karolherbst: it is called optimus or something
14:31prototron: lspci results here https://paste.fedoraproject.org/303594/
14:31prototron: doesnt seem to be there
14:32prototron: maybe I can boot windows to see if it's "there" ?
14:32karolherbst: the hardware manager should list it
14:32prototron: yeah. brb
14:44prototron: karolherbst, didn't see the HD 4000 chip in the display adapter menu, or anywhere
14:45karolherbst: okay, so it seems like you really only have the 750M then
14:45karolherbst: so, now you need a 4.3 kernel
14:45prototron: weird, because according to Intel's website, the 3630QM is supposed to have that chip
14:46karolherbst: debian sid has it I think
14:46karolherbst: prototron: yeah, but imirkin_ already said he saw some laptops where the intel gpu was just disabled
14:46imirkin_: specifically in a lenovo Y-series laptop
14:47imirkin_: but that one had dual GTX 650's or so
14:47prototron: ok. My laptop is SLI capable, but I don't have a second gpu
14:49karolherbst: xenial has 4.3
14:49karolherbst: willy just 4.2 :/
14:49karolherbst: ohh that's ubuntu
14:50karolherbst: prototron: do you have any preferences regarding distributions?
14:50prototron: not ubuntu
14:51prototron: I wanted something reliable and stable (this is why i chose Debian)
14:51karolherbst: k, I have no clue what is fancy these days with modern kernel, so others have to suggest something
14:51karolherbst: well, debian isn't that stable on modern hardware
14:51imirkin_: prototron: so did debian achieve your reliability and stability goals?
14:51prototron: imirkin, No.
14:51imirkin_: prototron: so... perhaps time to reevaluate those goals :)
14:52prototron: well, I don't know yet, cause I just struggle with the GPU
14:52josla972: what are good dists these days?
14:52karolherbst: prototron: you need a 4.3 kernel with an option
14:52josla972: i have been running gentoo for more than 10 years
14:52josla972: too lazy to switch
14:52imirkin_: GAH!!! grrr.... my xkb settings got lost :(
14:52karolherbst: prototron: you need to boot then with nouveau.config=War00C800_0=1
14:52imirkin_: josla972: me too :)
14:53karolherbst: well I run gentoo only for 3 years :/
14:53karolherbst: same installation though
14:53imirkin_: karolherbst: some day you'll catch up!
14:53josla972: compiling from source nowadays is quite fast
14:53karolherbst: well gentoo on my powermac g4 867MHz dual was fun
14:53josla972: but then came thunderbird, libreoffice, firefox etc
14:53imirkin_: i used slackware before and built everything from source
14:53imirkin_: gentoo just systematized that
14:53imirkin_: so it wasn't an exercise in pain when i wanted to update, say, evoluation
14:53karolherbst: I always wanted to try out LFS
14:54josla972: portage is very nice
14:54josla972: i build everything in a tmpfs
14:54karolherbst: prototron: so as you see, nobody has any clues here :D
14:54karolherbst: josla972: zram°
14:54karolherbst: compressed ram disc
14:54josla972: never did that
14:55karolherbst: I have a 16GB swap on zram
14:55karolherbst: I think you could even do a your system ram x 3 swap with zram inside ram
14:55karolherbst: it basically is like tmpfs
14:55karolherbst: just that it is compressed via lz4? or something
14:55karolherbst: and you create a real fs into that
14:55karolherbst: or swap
14:55josla972: ok kind of like ZFS with compression then I guess
14:55karolherbst: so you basically have a full ext4 fs compressed into ram
14:56karolherbst: and I do all my compilation into that
14:56karolherbst: 11GB ram disc
14:56karolherbst: usually only uses up to 3.5GB of real RAM
14:56josla972: i will look into that
14:56karolherbst: sys-block/zram-init is the package
14:56karolherbst: josla972: and this: https://wiki.gentoo.org/wiki/Zram
14:57karolherbst: this can be used to boost system ram on a low RAM environment, it is really nice
14:58josla972: never head of it
14:58josla972: experimental though
14:58josla972: had to set ~amd64
14:58karolherbst: zram-init are just the openrc/system files
14:58karolherbst: you still need it from in the kernel
14:58karolherbst: as a module
14:59karolherbst: but the wiki page will tell you everything
14:59josla972: i got screwed by poettering and changed to systemd after upgrading gnome
14:59karolherbst: yeah I also use systemd
14:59josla972: not because I wanted to
14:59karolherbst: it's not _that_ bad, but some things are just terrible
14:59karolherbst: well I choose mainly because of kwin_wayland
14:59josla972: it's new
14:59josla972: for me as a user and owner of my computer it brings me nothing of value that I can see directly
15:00Tom^: there comes a point when the tech savvy linux user realize it doesnt matter wich distro he uses. its just a matter of which package manager you prefer and enough time and patience.
15:00josla972: but perhaps systemd does cool stuff in the background :)
15:00imirkin_:misses slackware's pre-sysvinit startup scripts
15:00imirkin_: so simple... oh well.
15:01josla972: the parallel initialization also gave nothing of value - in fact, booting became slower...
15:01karolherbst: josla972: yeah, my parallel openrc was also faster
15:01josla972: and now lots of new commands to learn like systemctl, timedatectl etc
15:02josla972: at least I managed to skip the networkmanager :)
15:02imirkin_: i will write my own init scripts before touching systemd...
15:03RSpliet: ^ wouldn't touch it with a hammer
15:03josla972: imirkin_: openrc no good?
15:03imirkin_: josla972: openrc is fine
15:03karolherbst: I never get why socket activation couldn't be done inside a library or something, I mean the concept is pretty simple :/
15:03imirkin_: josla972: but i really miss the simplicity of just having a rc.S and rc.M
15:03josla972: freebsd is nice i think
15:04Tom^: i feel like an villain being on systemd in here :<
15:04imirkin_: Tom^: as you should
15:04karolherbst: journald is the worst :(
15:04josla972: I heard they managed to fix gnome3 compatibility with openrc
15:04josla972: so I am tempted to go back
15:05imirkin_: what does gnome3 have that WindowMaker doesn't?
15:05Tom^: wobbly windows.
15:05karolherbst: who does _not_ have wobbly windows
15:05airlied: gnome3 doesn't have wobbly windows, that was compiz
15:05imirkin_: Tom^: ok... so win for WindowMaker then? :)
15:05Tom^: airlied: oh right it might be kwin with its effects that has it too.
15:05josla972: never tried windowmaker
15:05karolherbst: course it has
15:06Tom^: or im just old and remembering things wrong
15:06imirkin_: josla972: it's a NeXT-y wm
15:06karolherbst: kwin can do everything anyway
15:06josla972: I am just lazy
15:06prototron: karolherbst, so no luck with nouveau.config=War00C800_0=1
15:06karolherbst: prototron: yeah, you need a 4.3 kernel for that
15:06josla972: I had gnome2 and upgraded to gnome3
15:06josla972: imirkin_: will look into it
15:06RSpliet:hides his SystemD knickers and Gnome Shell socks
15:06karolherbst: have fun with NeXT
15:06imirkin_: josla972: it's not particularly great. it's just perfectly sufficient for all my needs.
15:07prototron: karolherbst, nor video.allow_duplicates=1
15:07imirkin_: josla972: i think there are more modern wm's that achieve similar things that i haven't really investigated
15:07karolherbst: xfwm4 is nice too
15:07josla972: I see
15:07imirkin_: WindowMaker has done the trick for me for the past... 15? years
15:07josla972: the design kind of reminds me of amiga workbench 3 :P
15:07karolherbst: imirkin_: you need gnustep for that, right?
15:08imirkin_: karolherbst: sure, but that has no actual bearing on anything
15:08karolherbst: yeah I know
15:08imirkin_: karolherbst: just that its config files live in ~/GNUStep
15:08Tom^: imirkin_: scrot of your desktop!
15:08imirkin_: scrot? is that like scrotum?
15:08josla972: this "meta-key + search for any program"-thing is neat
15:09karolherbst: isn't that like a default feature of every dm?
15:09josla972: but I think even xfce has that
15:09Tom^: imirkin_: i thought everyone knew of scrot, the simple command line screenshot utility for x. :p
15:09imirkin_: Tom^: like xwd?
15:09imirkin_: anyways... that'll just be my background image
15:09imirkin_: not too exciting
15:09josla972: karolherbst: it was not available in gnome2 afaik
15:09karolherbst: Tom^: my desktop looks boring
15:11imirkin_: Tom^: http://i.imgur.com/R2R8lms.jpg
15:12karolherbst: oh that gnustep look
15:12karolherbst: well next
15:12prototron: karolherbst, how about the 4.2.6 kernel?
15:12karolherbst: that was before my time :D
15:12karolherbst: prototron: nope, 4.3
15:12Tom^: imirkin_: cool , feels retro =D
15:12prototron: so fuck debian
15:13imirkin_: Tom^: has looked that way for me since about 1999 or so
15:13imirkin_: Tom^: although i think i started using this bg image in 2001? give or take
15:13Tom^: imirkin_: sometimes change is a good thing
15:13karolherbst: yeah right
15:13karolherbst: try plasma5
15:13Tom^: nonono not that big of a change
15:13prototron: imirkin, you mentionned that the HD 4000 chip is disabled, so how would I enable it?
15:14karolherbst: Tom^: you have to think bout it
15:14karolherbst: he won't change it for another 15 years then
15:14imirkin_: prototron: don't think you can.
15:14Tom^: hm fair enough
15:14josla972: oh well, gtg sleep
15:14imirkin_: Tom^: if there's a concrete benefit i'll switch
15:14imirkin_: Tom^: but i've gotten pretty used to operating WM
15:14imirkin_: moving/resizing windows is really easy
15:15Tom^: besides looking cool? none.
15:15imirkin_: since with a modifier key you don't have to aim for the stupid borders
15:15imirkin_: what could be cooler than fake-transparent aterm windows? :)
15:16imirkin_: i switched to aterm after the much cooler-of-the-day Eterm brought my system to its knees if i had like 4 of them running
15:16Tom^: ive never even heard of either
15:16imirkin_: aterm is part of the afterstep thing i think
15:16imirkin_: which was another wm i used before windowmaker
15:17Tom^: "As of January 1, 2008 aterm is no longer maintained;" no wonder. i didnt even use linux until around there.
15:17imirkin_: but then its knew versions were *way* worse, and i couldn't figure out how to get the old one working
15:17karolherbst: imirkin_: well I don't have window borders either
15:17karolherbst: I hate them, so I just eliminated them :D
15:17imirkin_: Tom^: it's a little annoying coz it doesn't support utf8
15:18imirkin_: Tom^: but i've tried switching to other terminals and they all handle a little differently
15:18Tom^: heh you could try rxvt
15:18imirkin_: like lacking triple-click, or the way you select areas is diff, etc
15:18imirkin_: yeah, i tried it
15:18imirkin_: i use terminals a *lot* so the thing has to work *really well*
15:18karolherbst: I am happy with my yakuake
15:18imirkin_: all the ones i've found appear to miss features i've come to rely on
15:19imirkin_: and i don't really need utf8 in my terminals so... meh
15:21Tom^: imirkin_: http://i.imgur.com/9SKUR5v.png :>
15:22imirkin_: what wm is that?
15:22Tom^: awesomewm :p
15:22imirkin_: one of the tiling ones?
15:23Tom^: im all fancy with my tiling and libvte terminal termite. and colors everywhere! :P
15:23karolherbst: I tried tiling onece, it was weird
15:24karolherbst: ohh colors, really :D
15:24RSpliet: I once had a tiled bathroom
15:24Tom^: well i only have the dev tags as tiling for terminals. the rest is floating like any other wm
15:24karolherbst: RSpliet: so not anymore? What happend?
15:24RSpliet: had trouble with the smell, so now it's windowed
15:24karolherbst: RSpliet: why not both?
16:11prototron: Upgraded to kernel 4.3m no luck
16:12imirkin_: pastebin dmesg
16:12prototron: wait, i'll try the boot parameter that karolherbst suggested
16:13imirkin_: oh yeah, you almsot definitely need that
16:16prototron: nope. it boots, but in software rendering mode
16:16imirkin_: pastebin dmesg :)
16:16prototron: yeah yeah
16:20imirkin_: doesn't seem to have taken the param
16:21imirkin_: not sure why not....
16:22imirkin_: this is a GK107
16:22imirkin_: prototron: looks like you need 4.4-rc5
16:23prototron: lol :P
16:23imirkin_: specifically this one: http://cgit.freedesktop.org/~darktama/nouveau/commit/?id=70dd0fc6c8c2d5a8e2bd34895e5ab4fb90510a37
16:28prototron: woops, started the 4.4 upgrade already
16:28imirkin_: that's fine
16:28imirkin_: 4.4-rc5 (or -rc6) will do just fine
16:28imirkin_: the above patch was in 4.4-rc1 iirc
16:28prototron: installing rc5
16:36prototron: imirkin, i had no popup telling me I was in software rendering mode
16:37imirkin_: prototron: that looks *much* happier
16:38prototron: I did see a funky stripe, the screen went black, and it continued normally.
16:38prototron: though I saw an error message: Failed to Load/Save Screen Backlight Brightness of backlight:acpi_video0
16:39imirkin_: glxinfo | grep renderer
16:39imirkin_: should say Gallium 0.4 on NVE7
16:39prototron: it does
16:39imirkin_: you should have acceleration then
16:40karolherbst: prototron: I guess that stripe is realted to the non accelerated framebuffer
16:40karolherbst: and the kernel switched modes and then it is gone
16:40imirkin_: if you're looking for increased performance, you can boot with nouveau.pstate=1
16:40imirkin_: which should enable you to switch between performance modes manually
16:40imirkin_: otherwise you stay in the lowest one
16:41karolherbst: prototron: you could isntall envytools and then I could tell you a simple way to improve your battery life by a bit :)
16:41karolherbst: mupuf: seems like we should also tackle clock gating on PGRAPH for kepler
16:41karolherbst: for those rare nvidia only laptops
16:41mupuf: karolherbst: one thing at a time :D
16:41karolherbst: it is _one_ reg write
16:41karolherbst: at init
16:42karolherbst: ENG_PWR or ENG_CLK has to be set to AUTO instead of RUN
16:43mupuf: karolherbst: so is the FSRM :D close to it :D
16:43prototron: imirkin, so I have to boot all the time with nouveau.config=War00C800_0=1 ?
16:43imirkin_: prototron: no
16:44imirkin_: prototron: that should be on by default now
16:44karolherbst: the FSRM seems to be a lot more messy after looking into that :D
16:44karolherbst: the theory is easy yes
16:44karolherbst: but practically it is a mess
16:44prototron: imirkin, and for nouveau.pstate?
16:44imirkin_: prototron: that's not default, and probably not going to be for a while
16:44imirkin_: certainly not for the 4.4 release
16:45Tom^: are we entering debugfs then?
16:45karolherbst: mupuf: set bit 0 to 1 in 20200 and the clock gating is working
16:45karolherbst: that's it
16:45mupuf: nope, not for every card
16:46mupuf: and we sometimes need to disable it
16:48karolherbst: I don't think the blob ever disables it on PGRAPH
16:48karolherbst: at least for keplers
16:48karolherbst: but I can check into that tomorrow
16:52karolherbst: mupuf: you are right :O on mwks card the blob disables it again
16:52mupuf: yep... nothing comes easy in PM...
16:52karolherbst: but that's the only card
16:53karolherbst: mupuf: maybe on kepler that doesn't have to be disabled anymore?
16:53prototron: awesome, my second and third displays work!
16:53karolherbst: mupuf: we would need more trace though for that I guess
16:54karolherbst: anyway, I head to bed now :D
16:54karolherbst: mupuf: did you take a look at my paste?
16:54mupuf: karolherbst: which one?
16:54prototron: karolherbst, thank you. This is much appreciated
16:55prototron: and imirkin too!
16:55karolherbst: that's basically my fsrm paste
16:55karolherbst: but the last part is important
16:55karolherbst: prototron: no problem
16:56mupuf: karolherbst: right, did you find multiple cards with different behaviour but same chipset>
16:56karolherbst: there are nve7 with one type, then nve6 with different ones
16:56karolherbst: it is a mess
16:56karolherbst: and then some chipsets even have both (different cards)
16:58karolherbst: mupuf: for v in $(find nvc* nvd* nve* -type f -iname "*.rom"); do echo $v; ../envytools/build/nvbios/nvbios $v 2>/dev/null | grep 'R\[0x02010c\]'; done
16:59karolherbst: for v in $(find nvc* nvd* nve* -type f -iname "*.rom"); do echo $v; nvbios $v 2>/dev/null | grep 'R\[0x02010c\]'; done
16:59karolherbst: 0x00000002 indicates two_stage fsrm
16:59karolherbst: 0x00000016 three_stage
17:00karolherbst: I bet there is a way to enable the third threshold on my gpu somehow, but until now I didn't find a way
17:00karolherbst: so, now I am into bed for sure
17:01prototron: imirkin, So I don't have to use at all param nouveau.config=War00C800_0=1 with kernel 4.4?
17:01imirkin_: prototron: that's right
17:02prototron: ok good
17:03prototron: karolherbst, I don't get what EnvyTools does
17:04prototron: ah he left.
17:13Spencer_H: Hey does anyone know if any more work has been managed on getting an 840M (nv118) to work with Nouveau?
17:14imirkin_: Spencer_H: https://bugs.freedesktop.org/show_bug.cgi?id=89558
17:15Spencer_H: Yeah I saw that. Is the patch at the bottom complete?
17:15Spencer_H: Anything I can do to help?
17:15imirkin_: someone needs to go over the golden register values and compare if it's any different from GM107
17:15imirkin_: not without investing a lot of time
17:16Spencer_H: I have time. Since I can't do much with my PC until it works.
17:16imirkin_: wait, is this preventing some useful use-case for you?
17:16Spencer_H: I'd need to know how to do that though.
17:16imirkin_: i was under the impression that it was only used as an accelerator, never any outputs attached
17:17Spencer_H: Well it's preventing using video's and the like. Since the intel gpu in this PC can't manage that smoothly.
17:17imirkin_: see those tables? someone needs to make sure that they haven't changed for GM108
17:17imirkin_: mmmm... well there's no video accel with nouveau for maxwell
17:18Spencer_H: How should I check if they've changed?
17:18imirkin_: and the clock speeds are going to be the boot ones, so it's not likely to be any faster than the intel chip
17:18imirkin_: "check the mmiotrace"
17:18imirkin_: sorry, no additional info beyond that
17:18Spencer_H: and just go through one by one and see if they spat out the same thing?
17:18Spencer_H: Because the proprietary drivers currently don't function.
17:19imirkin_: yeah, just need to find the place in the trace
17:19imirkin_: oh, that's surprising
17:19imirkin_: there's a trace that was made available already though
17:19Spencer_H: Well they 'function' but they make any desktop environment unusable.
17:19imirkin_: ah, well nouveau is unlikely to help
17:19imirkin_: try the patch in that bug
17:19imirkin_: it should get you 99% of the way there
17:19imirkin_: it's just not upstreamable as-is
17:21Spencer_H: And I might as well work on something.
17:21imirkin_: if you have a desire to work on 3d driver features, there are a few maxwell-specific things that have been sitting around due to ... lack of my caring
17:22Spencer_H: sure. I'd like to help. I don't know much about driver programming though. I do know C though.
17:22imirkin_: do you know GL?
17:22Spencer_H: I'm relatively familiar with it. Never made a big project with it though.
17:22imirkin_: do you know how tess works?
17:23imirkin_: (i.e. ARB_tessellation_shader)
17:23Spencer_H: In use or internally?
17:23imirkin_: in use
17:23Spencer_H: I have no idea how it works on the inside.
17:24imirkin_: easy -- just write the right values to the hw, and magic, it tessellates :)
17:24Spencer_H: Alright then!
17:24imirkin_: so i implemented tess for nouveau
17:24imirkin_: and it works on fermi and kepler
17:24imirkin_: however since all hw engineers do is sit around and figure out ways to make my life harder
17:25imirkin_: they decided to change the way that tess control shader outputs are read/written
17:25Spencer_H: Completely change it or just move it around?
17:25imirkin_: so someone needs to trace the blob to figure out precisely how it should be done
17:25imirkin_: and then implement that
17:25imirkin_: (but don't worry, they also changed stuff on kepler... i just happened to have figured it out)
17:26imirkin_: i think completely
17:26Spencer_H: So write a gl program that uses it, and then trace it while using the proprietary driver to see what it does or...? Is there some other starting place.
17:26imirkin_: it used to be regular "output" writes
17:26imirkin_: but now you have to fetch some gmem address from somewhere
17:26imirkin_: and then do other crazy things
17:26imirkin_: i even have some traces... somewhere
17:26imirkin_: let me see
17:27imirkin_: you can use the 'demmt' tool in envytools to decode those
17:27imirkin_: (demmt -l foobar.mmt.xz should work)
17:27Spencer_H: Ok. I don't have permission to access the files though it seems. But that might just be konqueror being whiny.
17:27imirkin_: no, probably my bad
17:28imirkin_: try again
17:28imirkin_: anyways, i gtg
17:28imirkin_: i'll bb in 30m or so
17:28Spencer_H: Yep, works now.
17:29imirkin_: search for START_ID when you get demmt going
17:58imirkin: Spencer_H: did you have a look?
17:59imirkin: fyi sanity is in reference to http://cgit.freedesktop.org/piglit/tree/tests/spec/arb_tessellation_shader/execution/sanity.shader_test
18:02imirkin: i guess not.
18:10imirkin: some day i'll find someone to pawn the maxwell tess stuff onto...
18:12airlied: your future self?
18:12imirkin: i don't envy that guy!
18:13skeggsb: i'm dreading having to potentially one day do the maxwell sched stuff :P
18:14skeggsb: imirkin: i'll take tess if you take that :P
18:14imirkin: that being the vodka-mayo combo?
18:15imirkin: i'm not promising anything :p
18:15imirkin: for now i'm focusing on getting my atomic/ssbo work upstreamable
18:15imirkin: by which i mean... i *will* be focusing on that
18:15imirkin: some time in jan probably
18:31imirkin: skeggsb: i suspect i'll start focusing more on maxwell when GM20x is actually supported... the texture stuff at least...
18:32imirkin: GTX 950's are still like $150 =/
18:32imirkin: but... at least people will be able to test it