00:36 koz_: When I try to build nouveau from git, I get this: http://paste.rel4tion.org/111
00:36 koz_: I'm really confused - I have no idea what I'm doing wrong, because AFAIK, that directory doesn't contain a Makefile.
00:59 pmoreau: koz_: Is `lib/modules/4.4.0-rc1-g81051f9/build` the folder containing the kernel source code?
01:00 koz_: pmoreau: That dir only has one file in it.
01:00 koz_: I'm looking into what the hell happened with that - I suspect I did something wrong in the kernel build.
01:00 pmoreau: Where do you have the kernel source code?
01:02 pmoreau: Try to run from the `drm` folder in Nouveau Git: `LINUX_DIR=/path/to/kernel/code make`
01:03 koz_: pmoreau: Alrighty, I'll try.
01:36 koz_: karolherbst: I apparently suck at building kernels. Gonna harass my distro maintainers for a build of 4.3.
01:37 karolherbst: koz_: good :D
01:38 karolherbst: koz_: usually the dev trees should have it already
01:38 koz_: Hopefully the patch will get my card reclocking.
01:38 koz_: Well, fully-reclocking.
02:54 Tom^: imirkin: well if you tell me how to get it i can probably do it for you. :p
02:54 Tom^: or karolherbst
03:03 pmoreau: Tom^: To get your card's VBIOS?
03:03 Tom^: yea
03:04 pmoreau: You should be able to use `nvagetbios > vbios.rom` to get it (as root of course)
03:05 pmoreau: If it complains, you can try instead `nvagetbios -s PRAMIN > vbios.rom` or `nvagetbios -s PROM > vbios.rom`, depending on the one which failed
03:05 Tom^: ok
03:06 Tom^: gonna watch some gold rush, drink a cup of coffe and eat a cinnamon bun then il do it. :p
03:07 pmoreau: Since none of them is here, you can take your time. ;-)
03:07 pmoreau: (Well, even if they were here. ;-D )
03:08 karolherbst: pmoreau: well I prever the vbios.rom file through debugfs ;)
03:10 karolherbst: pmoreau: just because this is what nouveau sees and it doesn't require envytools installed, though we also want nvapeek 101000 so this is kind of useless anyway :D
03:10 karolherbst: Tom^: if you get the vbios, please also do nvapeek 101000
03:10 pmoreau: :-D
03:11 karolherbst: this values is sometimes usefull for calculation some timings? or something else
03:11 Tom^: karolherbst: ok il try.
03:11 pmoreau: How do you get it through debugfs? Can't seem to find it…
03:12 pmoreau: Ha, in the dri folder
03:12 karolherbst: pmoreau: yeah
03:12 pmoreau: Just need to find the correct number
03:12 karolherbst: pmoreau: well, that's usually easy
03:13 pmoreau: Seems to be the same mapping as otherwise used, cool
03:15 karolherbst: what card does Tom^ have?
03:15 Tom^: 780ti
03:15 karolherbst: ahhhh
03:15 karolherbst: now it makes sense :D
03:15 Tom^: =D
03:16 karolherbst: I still need to learn the real name => IRC names mapping
03:16 karolherbst: it is a gk110 card, isn't it?
03:17 Tom^: NVF1 (GK110B)
03:18 karolherbst: I hope it's the same issue some others encountered, because for that I would have a fix already, but maybe something funky is going on with the gk110 cards, we will see
03:19 karolherbst: Tom^: but I guess dmesg showed something like "voltage failed: -22"?
03:19 Tom^: ¯\_(ツ)_/¯ i havent used nouveau in over a year.
03:20 Tom^: i just saw imirkin highlighting me about 780ti vbios :p
03:20 karolherbst: ohhh okay
03:21 karolherbst: Tom^: maybe it just works for you though :/ some keplers have a pain in the ass voltage table in their vbios
03:51 Tom^: karolherbst: any particular kernel or other version of software you want me to fetch it on?
03:57 karolherbst: doesn't matter
04:46 Tom^: karolherbst: nop sorry, either my ram or usb stick is about to die. building envytools on the liveiso yields I/O errors and a unresponsive system. and dont have the time to install linux on a proper hdd now
04:48 pmoreau: Tom^: These images [here](https://nouveau.pmoreau.org/) should have envytools included.
04:48 Tom^: hm ok
05:04 Tom^: pmoreau: yea that worked, only method that didnt complain tho was -s PROM . where should i upload these for you then. dropbox? and you want the .rom's that nvagetbios complained on too or just the successful one?
05:05 pmoreau: The successful one only :-) (Except if karolherbst says otherwise)
05:05 pmoreau: Upload it anywhere, I'll add it to our VBIOS repo.
05:06 pmoreau: (I'll also need the result of `nvapeek 101000` please)
05:06 Tom^: oh right forgot that one xD . brb.
05:14 Tom^: pmoreau: 00101000: 8040689a
05:14 Tom^: and https://www.dropbox.com/s/22liacwxst4xzpl/vbiosprom.rom?dl=0
05:16 pmoreau: Thanks!
05:21 ata2001: Hello! Sorry for this dumb question, but can i have OpenGl 4 with NVE7?
05:22 karolherbst: ata2001: yes
05:22 pmoreau: IIRC, you should be able to have OpenGL 4.1 on Kepler with a recent Mesa (prboably 11.0+)
05:23 karolherbst: mine glxinfo: https://gist.github.com/karolherbst/1ed69e1ead35202c135b
05:23 karolherbst: I have a nve6
05:23 karolherbst: *my
05:24 karolherbst: ata2001: did you checked glxinfo on your system or are you asking because you might get such a gpu?
05:25 ata2001: My glxinfo says, i have 2.1 "OpenGL version string: 2.1 Mesa 11.0.5"
05:25 karolherbst: ata2001: istall this dxtn library
05:26 karolherbst: or how this is called
05:26 karolherbst: ata2001: does "glxinfo | grep s3tc" output anything?
05:26 karolherbst: like GL_EXT_texture_compression_s3tc
05:26 imirkin: ata2001: don't forget --enable-texture-float when building mesa
05:26 pmoreau: libtxc_dxtn on Arch
05:27 ata2001: yes
05:27 ata2001: GL_EXT_texture_compression_s3tc, GL_EXT_texture_cube_map,
05:27 karolherbst: ata2001: and OpenGL renderer string: Gallium 0.4 on NVE7?
05:27 karolherbst: in glxinfo
05:28 imirkin: s3tc is not required for any GL version, fyi
05:28 ata2001: OpenGL renderer string: Gallium 0.4 on NVE7
05:28 imirkin: although many games do implicitly depend on it
05:28 karolherbst: imirkin: I thought it is for 3.0?
05:28 pmoreau: karolherbst: Did you manage to download Tom^ VBIOS? Nothing happens when I click on the download button…
05:28 imirkin: karolherbst: you were mistaken.
05:28 Tom^: pmoreau: uhm should i upload it somewhere else then? :P
05:29 imirkin: the reason why people see GL 2.1 on otherwise GL3+ driver/hw combos is that they're missing --enable-float-texture when building mesa
05:29 imirkin: ARB_texture_float is, in fact, part of GL 3.0
05:29 pmoreau: Tom^: Could be an issue on my end. If he manages to download it, then everything's fine.
05:29 karolherbst: imirkin: ohh okay, I thoguht s3tc is required too
05:29 karolherbst: ata2001: then check "glxinfo | grep ARB_texture_float"
05:29 imirkin: it's clearly not there... heh.
05:30 karolherbst: :D
05:30 karolherbst: I want to be sure
05:30 imirkin: pmoreau: worksforme
05:30 karolherbst: experience told me, that even if it has to be there, sometimes it's not
05:30 karolherbst: yeah works for me too
05:30 ata2001: nothing
05:30 pmoreau: Then I let you add it to the VBIOS repo. :-)
05:30 Tom^: pmoreau: http://s000.tinyupload.com/index.php?file_id=51684184658217937529 ( only for you <3 )
05:31 karolherbst: meh imirkin is right :P
05:31 imirkin: who knew :p
05:31 pmoreau: Tom^: Thanks! :-D
05:31 Tom^: also i have no idea if that site is trustable but i googled . "adfree file hosting" :P
05:31 karolherbst: :D
05:31 karolherbst: isn't every site implicitly "adfree" usually? :D
05:32 karolherbst: once I was browsing without addblocker.... never again
05:32 karolherbst: meh, this vbios just looks too ordinary
05:32 karolherbst: except the clocks are a bit funny
05:32 karolherbst: and voltage step -12500
05:33 karolherbst: but I thought the kernel module handles that already
05:33 karolherbst: otherwise a lot more would have complained already
05:33 karolherbst: pmoreau: check the PM_Mode table
05:33 karolherbst: the domain => clocks mapping is a bit odd in nvbios
05:33 karolherbst: memclk freq = 988 MHz at 0f?
05:33 karolherbst: and vdec freq = 3500 MHz
05:34 karolherbst: these should be swapped I assume
05:34 pmoreau: I'll do
05:34 karolherbst: Tom^: could you test with upstream nouveau or my branch with 4.3?
05:35 karolherbst: it should work afaik, but maybe there is something funky
05:35 Tom^: do you have an .iso ready for booting with that? :P
05:35 karolherbst: 4.3 kernel?
05:35 karolherbst: you can compile nouveau out of tree and install the module :)
05:36 karolherbst: imirkin: I am pretty sure it was just the 0ed voltage table header issue he got
05:37 karolherbst: the vbios looks okayish to me except this domain clock mapping
05:37 Tom^: karolherbst: well im not so sure i can, only OS atm on this desktop is NSA10. sort of made it into my gaming station and left linux on the t60p :p
05:38 karolherbst: :D I see
05:38 Tom^: so yea, i fetched the vbios from a "live iso" on a usb stick.
05:38 karolherbst: pmoreau: what nouveau version does the live iso have?
05:38 pmoreau: Let's see what the description on the webpage says… :-)
05:38 karolherbst: Tom^: thin is, pre 4.4 nouveua will mess up the nvidia card at 0d and 0f pstate
05:39 pmoreau: karolherbst: Nouveau module (branch master from Ben Skeggs' repo): commit 4b7af6c
05:39 karolherbst: pmoreau: mhhhh
05:39 karolherbst: we would really need the newet stuff for kepler cards :/
05:40 karolherbst: ohh wait
05:40 karolherbst: I think this is new enough
05:40 karolherbst: yeah
05:40 karolherbst: the pwm fix is also there
05:40 karolherbst: okay nice
05:40 karolherbst: Tom^: seems like the live iso is good
05:40 karolherbst: Tom^: just boot with nouveau.pstate=1
05:40 Tom^: sure
05:41 Tom^: and then? set it to 0d or 0f ?
05:41 Tom^: or you want the vbios after
05:41 karolherbst: first try 0a
05:41 karolherbst: I do care more about the core clock
05:41 karolherbst: it should change while going to 0a
05:42 karolherbst: Tom^: do you know why we wanted to check the 780 ti at all? :D
05:42 Tom^: not the slightest.
05:43 karolherbst: Tom^: the phoroix guy did some benchmark on the 780 ti with recent nouveau, but his core clock doesn't change at all, so we tried to find out if that's gk110 specifc or just his card
05:43 karolherbst: maybe all gk110 cards are messy
05:43 Tom^: ah
05:44 pmoreau: Tom^: The second link worked better (though I had some ads :-D)
05:44 Tom^: haha
05:44 pmoreau: karolherbst: Pushed the VBIOS to the repo
05:45 karolherbst: nice
05:45 karolherbst: first nvf1 vbios
05:45 karolherbst: :D
05:45 ata2001: karolherbst: I have this USE flag(Gentoo) enabled, do i need to disable it? bindist: Disable patent-encumbered ARB_texture_float, EXT_texture_shared_exponent, and EXT_packed_float extensions.
05:45 karolherbst: pmoreau: by the way: I was able to use those vbios files from the website
05:45 karolherbst: ata2001: disable bindist
05:45 karolherbst: always
05:45 karolherbst: for everything
05:45 imirkin: ata2001: yes, you have to disable it if you want those extensions, and GL 4.1 on your GPU.
05:45 karolherbst: except you give the package to others
05:45 pmoreau: karolherbst: Which website?
05:46 karolherbst: pmoreau: https://www.techpowerup.com/vgabios/index.php
05:46 karolherbst: it's quite easy, they just have a stupid header in the first 0x550 bits or something
05:46 karolherbst: after you remove them, you are good to go
05:47 pmoreau: Ok
05:47 ata2001: karolherbst: Thank you! Free softwares have very helpful community.
05:47 karolherbst: pmoreau: if you compare their vbios files with one of the from the repo you see it
05:47 Tom^: ok there we go right now booted on 4.3.0-rc5-lts+ and pstate set to 0a
05:47 karolherbst: Tom^: check in pstate if the core clock is right
05:48 pmoreau: I could try to play with the Titan X VBIOS to get it correctly parsed by NVBIOS, but… that's way to far away on my todo list.
05:48 Tom^: "0a: core 324-692 MHz memory 1620 MHz AC DC *"
05:48 karolherbst: Tom^: last line
05:48 Tom^: core is 691 MHz
05:49 karolherbst: last line is the current stuff
05:49 karolherbst: mhh looks good
05:49 karolherbst: now you can try 0d
05:49 karolherbst: it may hang your machine though :D
05:49 Tom^: clk: failed to raise voltage: -22 and clk: error setting pstate 2: -22
05:50 karolherbst: but memory is okay?
05:50 Tom^: core still the same but memory at 6999 MHz
05:50 karolherbst: the memory clock
05:50 karolherbst: okay
05:50 karolherbst: looks good then
05:50 karolherbst: I mean, yeah well, good enough
05:50 karolherbst: will check it
05:50 imirkin: karolherbst: do you know why it fails to raise voltage?
05:51 karolherbst: imirkin: yes
05:51 karolherbst: imirkin: the voltage map entry wants 1225000uv
05:51 karolherbst: but: -- Mode GPIO, Base voltage 825000 µV, voltage step -12500 µV, acceptable range [825000, 1212500] µV --
05:51 imirkin: sad :(
05:52 karolherbst: Tom^: do you want to clock to 0f
05:52 karolherbst: and run vblank_mode=1 glxgears
05:53 Tom^: if i want to? hell yes.
05:53 karolherbst: pmoreau: is glxspheres on the livecd
05:53 pmoreau: karolherbst: I doubt that
05:53 karolherbst: :/
05:54 pmoreau: I will add it and start generating a new one
05:54 karolherbst: Tom^: everything below 20k fps is disappointing :D
05:54 pmoreau: karolherbst: Anything else that could be useful to add?
05:54 Tom^: hm Error: couldn't get an RGB, Double-buffered visual
05:55 karolherbst: :O
05:55 karolherbst: Tom^: then plain glxinfo
05:55 karolherbst: pmoreau: mhhh, benchmarks, but no big ones
05:55 karolherbst: pmoreau: maybe gputest?
05:55 Tom^: Error: couldn't find RGB GLX visual or fbconfig
05:55 karolherbst: it needs around 32MB
05:56 kubast2: http://askubuntu.com/questions/475972/error-couldnt-find-rgb-glx-visual-or-fbconfig-ubuntu-12-04 ubuntu live cd?
05:56 karolherbst: mhh LIBGL_DEBUG=verbose glxinfo
05:56 Tom^: haha no this is pmoreau livecd
05:57 Tom^: startx launched openbox and it has firefox, xterm etc :p
05:57 Tom^: karolherbst: same thing
05:57 karolherbst: Tom^: did LIBGL_DEBUG=verbose print anything usefull?
05:57 karolherbst: it should print more though
05:57 Tom^: nop error with visual fbconfig :p
05:58 pmoreau: Tom^: Hum… I'll need to investigate that. :-/
05:59 Tom^: https://imgur.com/nK0vSp5
06:01 pmoreau: Tom^: Could you upload the Xorg.log somewhere please?
06:01 imirkin: Tom^: pastebin your xorg log?
06:01 Tom^: sure
06:01 Tom^: was just skimming it through :p
06:04 Tom^: http://ix.io/moC
06:05 Tom^: libnettle.so.4 is missing it seems
06:05 pmoreau: AIGLX error: dlopen of /usr/lib/xorg/modules/dri/nouveau_dri.so failed (libnettle.so.4: cannot open shared object file: No such file or directory)
06:05 pmoreau: Yeah
06:05 imirkin: bad pmoreau
06:05 pmoreau: :-(
06:05 imirkin: pmoreau: i'd also drop xf86-video-vesa into your distro... just in case :)
06:06 imirkin: nice to have that hanging around when debugging graphics things
06:06 pmoreau: Why is it looking for libnettle.so.4 and not for libnettle.so.6?!
06:07 imirkin: because you used arch
06:07 Tom^: did you compile something towards the old one? before libnettle got upped to .6 ?
06:07 pmoreau: imirkin: And? :p It correctly uses libnettle.so.6 on my laptop running Arch
06:08 imirkin: i bet you're not cross-compiling
06:08 imirkin: and so it's building against whatever's on the build system
06:08 imirkin: rather than against whatever should be in the target environment
06:08 pmoreau: Could be there has been some packages which failed to compile without notifying me…
06:09 Tom^: or did you compile xf86-video-nouvau before updating libnettle ?
06:09 pmoreau: imirkin: The system also has libnettle.so.6 since it also uses Arch and is up-to-date
06:11 Tom^: pmoreau: symlinking libnettle makes X complain about missing libLLVM-3.6.so too
06:11 Tom^: D:
06:11 pmoreau: --"
06:11 Tom^: your distro is giving me flashbacks of dll hell.
06:12 pmoreau: It should be using LLVM 3.7
06:19 Tom^: pmoreau: if i were to assume your xf86-video-nouveau got built/packaged before you upped libnettle and libllvm
06:20 pmoreau: Tom^: That might be a correct assumption, as currently it only rebuilds packages which got new commits…
06:20 karolherbst: Tom^: on my system only curl, gnutls and libarchive needs nettle
06:21 pmoreau: And given that xf86-video-nouveau's activity is quite low, that might be what happened
06:21 karolherbst: ehhm /usr/lib/xorg/modules/dri/nouveau_dri.so is mesa
06:21 karolherbst: not the ddx
06:22 Tom^: indeed
06:25 pmoreau: Got to get a lunch before it's dinner time
06:26 pmoreau: I started a recompilation of all git packages, but it stucked on libdrm tests…
06:26 pmoreau: bbl
06:26 imirkin: Tom^: thanks for testing :)
06:26 Tom^: happy to help :p
06:27 pmoreau: Tom^: You increased the pool of testers for that iso by almost an order of magnitude!
06:27 Tom^: xD
06:51 karolherbst: imirkin: mhh, when nouveau was loaded before X starts, it is less messy to remove the module later when something happens :/
06:51 karolherbst: I think the X server does something differently, when the module gets loaded later
09:53 karolherbst: fun times: https://gist.github.com/karolherbst/c1f4bcd9a081dcf93ca1
09:54 karolherbst: gpu is off, but in use :/
09:54 karolherbst: meh
09:55 karolherbst: and then this: pci_pm_runtime_suspend(): nouveau_pmops_runtime_suspend+0x0/0xd0 [nouveau] returns -16
09:59 karolherbst: I think I know what happens
09:59 karolherbst: userspace already closed all fds, but the kernel module is still waiting for the gpu to unmess itself (cause of fifo errors or whatever)
10:00 karolherbst: but
10:00 karolherbst: runpm has a use count of 0, because there is nothing using the gpu
10:00 karolherbst: now the gpu is turned off and nouveau loops somewhere waiting for the gpu
10:45 kisak: hi, I happened to notice the 840M is double listed on the decoder ring http://nouveau.freedesktop.org/wiki/CodeNames/ , is that accurate?
10:46 kisak: NV117 and NV 118
10:46 glennk: marketing names can map to several different chips
10:47 kisak: -_- that's not nice
10:48 kisak: in any case, thanks for the factoid
13:11 pmoreau: Tom^: It had some fails on building Mesa due to not properly detecting LLVM, and some fails on building the kernel due to the dependency cycles between lnet and libcfs.
13:11 Tom^: O_o
13:11 pmoreau: Tom^: First issue has been fixed (and I enabled OpenCL support in Mesa while I was at it), and the second one is being fixed.
13:12 pmoreau: So, hopefully the new ISO (probably ready in ~1h I guess) should work.
13:12 Tom^: cool
13:12 Tom^: il test it when its ready for prime time
13:12 pmoreau: :-)
13:12 Tom^: see if i get those 20k fps in glxgears ;)
13:12 pmoreau: Eh eh eh !
13:13 pmoreau: And it should have glxspheres as well now
13:13 Tom^: nice
13:13 pmoreau: I'll test it before pinging you (if I haven't gone to sleep in the mean time :-D)
13:13 Tom^: oki
13:15 xae: hi, just wondering, how is nouveau with nvidia gtx 850m (maxwell)? i've read that maxwell isn't very well supported at all yet with nouveau due to nvidia being asshats
13:15 imirkin: xae: which maxwell is this... GM108?
13:16 imirkin: xae: if so, it should be moderately supported with a simple kernel patch. but i'm waiting for GM20x to become usable before spending any serious time fixing the 3d driver for maxwell issues.
13:17 xae: gm10* so first gen maxwell as far as im aware
13:17 xae: The GM10x GPUs are also used in the GeForce 800M series and the Quadro Kxxx series.
13:17 xae: according to wikipedia ^
13:18 imirkin: yeah, unfortunately nvidia likes to mix it up
13:18 xae: anyway i can check specifically which it is?
13:18 imirkin: lspci -nn -d 10de:
13:21 xae: imirkin: GM107M
13:21 imirkin: ok, so that should be supported out of the box
13:21 imirkin: on kernel 4.1+
13:21 xae: :D good news
13:21 imirkin: however like i said, accel def has some issues
13:21 imirkin: also i was lazy and didn't finish tess on it, so ... no tess.
13:21 xae: this is my first time having used nvidia in all my years of linux and it's proving somewhat.... irritating
13:22 imirkin: also no reclocking, so it'll be slower than the intel gpu you have inside
13:22 imirkin: [not to mention buggier]
13:22 xae: what's the consensus with the nouveau prime vs bumblebee?
13:22 imirkin: you can't mix bumblebee and nouveau unless you *really* know what you're doing
13:22 xae: hmm
13:22 imirkin: nouveau will auto-suspend the gpu though
13:23 imirkin: (which is why you can't mix it with bumblebee... don't want them fighting)
13:23 xae: alright
13:23 imirkin: it's possible though... karolherbst does it afaik. you can ask him the specifics.
13:23 xae: i'll do that and have a play around with it later, going afk soon
13:52 dcomp: can bumblebee do prime offload?
13:59 imirkin: iirc bumblebee has some crazy second X server thing... prime is just a plain thing that just works. you don't use "bumblebee" to get it... in fact, removing bumblebee is the best course of action if you're looking to use open drivers.
14:07 Tom^: isnt bumblebee just a glorified script that launches X offscreen on the nvidia gpu and then uses prime to pipe it back ?
14:11 karolherbst: imirkin: what do I do :O
14:12 imirkin: Tom^: not prime... its own thing. also it has a kernel module to facilitate turning gpu on and off in laptops.
14:12 imirkin: karolherbst: bumblebee + nouveau
14:12 karolherbst: ohh right
14:13 karolherbst: imirkin: the only thing I noticed is, that you can turn on the gpu through bbswitch while nouveau turned it off
14:13 karolherbst: this messes things up big time
14:13 imirkin: right.
14:13 karolherbst: but other than that
14:13 karolherbst: no problem
14:13 karolherbst: I think bumblebee actually removes the nouveau module too
14:14 karolherbst: yeah
14:14 karolherbst: but nvidia can't handly it
14:14 karolherbst: *handle
14:14 karolherbst: xae: basically you can have bumblebeed running and bsswitch loaded while nouveau is loaded
14:15 karolherbst: but you shouldn't run anything through bumblebee as long as the gpu wasn't turned off after nouveau was unloaded
14:15 karolherbst: if I want to run something on the nvdia driver I do this: unload nouveau, echo OFF > bbswtich, optirun something
14:16 Tom^: imirkin: right thought of primus.
14:17 karolherbst: imirkin: also I get this error when the card is turned of the second time through bbswitch: nouveau 0000:01:00.0: unknown chipset (ffffffff)
14:17 karolherbst: so nouveau won't load at all when the gpu is off
14:17 imirkin: not entirely sure what primus is... just a wrapper that sets DRI_PRIME=1?
14:17 karolherbst: works the first time though
14:17 karolherbst: imirkin: primusrun?
14:18 karolherbst: primusrun is in-application texture copies from one glx context (nvidia X server) to another glx context (intel X server)
14:19 karolherbst: bascially it implements the libGl.so ABI and provides the entire symbols and intercepts every gl(x) call
14:19 pmoreau: Damn, the ISO jumped to 930M… :-( I really need to investigate that.
14:21 karolherbst: imirkin: like on every glxSwapBuffers call the nvidia X server window content is read and inserted into a intel gl context via a texture, which get's displayed
14:22 Tom^: pmoreau: is that a bad thing? :p
14:22 karolherbst: pm -Os :p
14:23 karolherbst: pmoreau:
14:23 pmoreau: Tom^: Yeah! It's way to heavy! The first ISO created were at least 700M
14:23 pmoreau: Maybe even more around 500M
14:45 karolherbst: imirkin: I am getting a bit paranoid: what if in other situations nouveau also doesn't get some IRQs and that's why nouveau _thinks_ the gpu is messed but, but it's really not?
15:07 pmoreau: Tom^: The latest ISO is working, if you want to give it another try. :-)
15:07 Tom^: sure
15:19 Tom^: pmoreau: sure works
15:19 pmoreau: Great!
15:21 karolherbst: Tom^: nice :)
15:21 karolherbst: now 0f and vblank_mode=1 glxgears :D
15:21 karolherbst: pmoreau: did you add glxspheres too?
15:21 imirkin: his cstate doesn't change right?
15:21 pmoreau: karolherbst: I did :-)
15:21 karolherbst: imirkin: it did
15:22 karolherbst: imirkin: not for the highest one
15:22 imirkin: i thought the voltage table was insufficient
15:22 imirkin: right
15:22 karolherbst: imirkin: but that's because request voltage >= highest voltage
15:22 karolherbst: >
15:22 karolherbst: not >=
15:22 imirkin: perhaps we should just switch to a cstate with a voltage table entry?
15:22 karolherbst: imirkin: dunno
15:23 Tom^: karolherbst: http://i.imgur.com/jjPVO10.png
15:23 imirkin: the api doesn't really facilitate that
15:23 Tom^: karolherbst: that is with vblank_mode=0 :p
15:23 karolherbst: Tom^: meh, as fast as mine
15:23 karolherbst: Tom^: cat pstate
15:23 karolherbst: it should still ahve the 07 core clock, right?
15:24 Tom^: well i cat'ed 0a first to get 691 :p
15:24 Tom^: so according to the pstate output i doubled the 07 core clock
15:24 karolherbst: imirkin: https://github.com/karolherbst/nouveau/commit/4fbc9434ba85e2e01ef8bcfd0cf8764f2967a881
15:24 karolherbst: Tom^: ohh okay
15:25 imirkin: karolherbst: yeah i remember those patches
15:26 karolherbst: okay, I will go to bed for now. Tom^ we can take one fo my branches tomorow and manucally select a core clock
15:26 Tom^: sure
15:26 karolherbst: pmoreau: is git and stuff on the live cd? :p
15:26 karolherbst: nvm
15:26 karolherbst: cu
15:26 Tom^: i have pacman at my hand so i can just get those np
15:27 pmoreau: karolherbst: I think Git should be installed
15:27 pmoreau: Yep, it is
17:52 orbea: what would I need to get at least opengl 3.1 using nouveau and a GeForce GTX 780 Ti? Im at OpenGl 3.0 and Mesa 11.0.5 now.
17:54 sarnex: orbea: are you looking at the core profile?
17:54 orbea: Uhh, no? Could you elaborate?
17:55 sarnex: as far as i know the compatibility context is generally seen as a mistake and can't have new features added
17:55 sarnex: someone who knows more feel free to jump in
17:55 sarnex: orbea: can you pastebin glxinfo
17:57 orbea: http://dpaste.com/0Y8BCCY
17:57 sarnex: orbea: did you remove anything from the output?
17:57 orbea: sorry, didn't paste right
18:00 orbea: better http://dpaste.com/3K7W2FP
18:08 sarnex: orbea: OpenGL core profile version string: 4.1 (Core Profile) Mesa 11.0.5
18:08 sarnex: if a game is saying you have 3.0 then it's using the compat profile and i dont think there are any changes to support more extensions with that profile
18:08 sarnex: s/changes/plans
18:10 orbea: OpenGL version string: 3.0 Mesa 11.0.5
18:12 orbea: annoying cause my older cards had a compat profile of 3.3...
18:13 koz_: I'm trying to build nouveau from the branch karol gave me, but I get this error: http://paste.rel4tion.org/112
18:15 sarnex: orbea: are you sure? mine is 3.0 on radeonsi. 4.1 is new to nouveau so maybe that was the core context before it jumped to 4.1
18:16 orbea: sarnex: maybe im misremembering
18:16 orbea: its been a while
18:24 koz_: imirkin: I'm trying to build nouveau from the branch karol gave me, but I get this error: http://paste.rel4tion.org/112
18:24 koz_: Did I miss something?
18:36 Tom^: sarnex: you might just need to build mesa from git.
18:36 sarnex: Tom^: i use mesa git on gentoo
18:36 Tom^: because i for sure remember it reporting higher then 3.0 on my 780ti
18:37 sarnex: really? the compatibility profile?
18:37 imirkin: you won't get higher than 3.0 out of mesa for compat profiles
18:37 Tom^: are you sure?
18:37 Tom^: i recall i did before mesa 11 was out when i compiled it from git. hm
18:37 sarnex: that likely the core profile
18:43 Tom^: yeh perhaps
18:46 koz_: I'm trying to build nouveau from source, but I get this error: http://paste.rel4tion.org/112
18:46 koz_: Could anyone please give me a clue as to what on earth I'm missing?
19:29 imirkin: koz_: just comment it out
20:43 koz_: I'm building nouveau from source (this branch: https://github.com/karolherbst/nouveau/tree/koz_) against the 4.3 kernel, and I get this error: http://paste.rel4tion.org/114 .
20:45 imirkin: koz_: revert the commit that says "drm-next bla bla bla"
20:46 koz_: imirkin: I only cloned --depth 1.
20:46 koz_: So do I need to do the whole clone and then backtrack a commit?
20:46 imirkin: why would you do that?
20:47 koz_: At karol's suggestion.
20:47 koz_: You were here when he made it IIRC.
20:47 imirkin: ah. well, i've never actually checked anything out with --depth 1, so i dunno how that works
20:47 imirkin: but if it's what i think it is, it's just a totally useless checkout that you can do absolutely nothing with
20:47 koz_: Well, I'll reclone, revert and try again.
20:48 imirkin: koz_: ftr, this is the commit you want to revert: https://github.com/karolherbst/nouveau/commit/d55c3200d876032f6149c83045eea96c05472dc9
20:48 imirkin: you could grab that as a patch (stick .patch to the end of that url) and apply it in reverse (patch -R)
20:49 koz_: imirkin: I'm checking the history, and I don't see any DRM-next commit.
20:50 imirkin: did you not see the link above?
20:52 koz_: Ah, I see
20:52 koz_: OK, gonna try that now.
20:55 koz_: How do I apply patches in reverse? I have the patch file in the nouveau directory, but I'm not sure how to call patch -R properly.
20:58 imirkin: patch -p1 -R < foo.patch
20:58 koz_: Ah, thanks.
21:05 koz_: Didn't work. Still same core clock problem. :(
21:06 imirkin: well hold on... are you sure that you loaded the right thing?
21:06 imirkin: i assume it built :)
21:06 koz_: I did this: cd drm
21:06 koz_: Then make
21:06 imirkin: can you provide a dmesg? i think it should be apparent
21:06 koz_: Then sudo make install.
21:06 imirkin: errrr
21:06 imirkin: what did that do?
21:06 imirkin: i can't imagine anything good =/
21:07 koz_: Let me check exactly.
21:07 imirkin: anyways... the idea is that running make builds a new nouveau.ko module
21:07 imirkin: you can then either load it yourself instead of the one that came with your kernel, or replace your system one with it
21:07 koz_: I assumed sudo make install did the latter.
21:07 koz_: I can show you the messages it gave me.
21:07 imirkin: pastebin dmesg
21:08 imirkin: actually.... that might not make it apparent
21:08 koz_: Well, do you want it anyway?
21:09 imirkin: sure :)
21:09 koz_: How would I normally replace my system nouveau with this one?
21:10 imirkin: locate the existing nouveau.ko and replace it with the new one (in drm/nouveau/nouveau.ko)
21:10 imirkin: if your modules are loaded from a initrd, you need to also update your initrd
21:10 imirkin: i know little of such things though... i like to keep it simple
21:11 koz_: imirkin: dmesg: http://paste.rel4tion.org/116
21:11 koz_: Near as I can tell, sudo make install did exactly this.
21:12 imirkin: Initialized nouveau 1.3.0 -- hmmm... didn't we bump it up to 1.3.1?
21:13 koz_: Where in the source would I find this?
21:13 imirkin: sec
21:14 imirkin: trying to remember ;)
21:14 koz_: No rush. :)
21:16 imirkin: can't find it... i'm tired
21:16 imirkin: well, you need to make sure that the actual built module is loaded. you need to be able to do something like that ;)
21:16 imirkin: good luck
21:17 koz_: Well, the install recipe in the Makefile does this: $(MAKE) -C $(LINUXDIR) $(KBUILD_OPTS) modules_install
21:17 koz_: That, as far as I know, is exactly the whole 'replace kernel module' thing.
21:29 koz_: Nope, it's definitely that one. I checked.
21:31 imirkin: ok cool
21:31 imirkin: and the module gets loaded from your fs, not an initrd?
21:31 koz_: Yes.
21:32 imirkin: ok cool
21:35 koz_: imirkin: OK, this is odd - you said the version was 1.3? The package for my distro looks like this: xf86-video-nouveau-1.0.11-3
21:35 koz_: Is it just me or is this *really* behind?
21:35 imirkin: that's the DDX
21:35 imirkin: i was talking about the version printed in the kernel
21:35 koz_: DDX?
21:35 koz_: Ah.
21:36 imirkin: which isn't connected to anything at all, really
21:36 imirkin: aha found it, actually. yeah 1.3.0 is the latest... i thought we had bumped it since then but i guess not.
21:36 imirkin: it's used by userspace to know what abi is available
21:37 koz_: OK, so I'm not insane. :P
21:37 imirkin: DDX = device-dependent X driver
21:38 koz_: So should I assume that my reclocking problem lies with something else?
21:38 imirkin: dunno, check with karol
21:39 koz_: imirkin: Will do. Thanks for all your help!
21:39 koz_: Do we have a messaging bot or something here?
21:40 imirkin: not sure what that is
21:40 imirkin: so i guess "no"
21:41 imirkin: this channel is logged though
21:41 imirkin: and there's this cool new thing called "email" -- check it out :)
21:41 koz_: Lol...
21:41 koz_: Fair enough.
21:42 koz_: Alrighty, I'll contact karol and ask.
23:45 jdb: Hello everyone. I'm checking out the feature matrix for Nouveau and had a quick question. It says that SLI is currently a Todo item, but does Nouveau support multi-card setups that aren't in SLI? For example, if I have two identical NV124 cards, with one and two DVI connections respectively, is it possible for me to run that as one X screen, or would I need to seat all the connections to just one of the cards
23:46 jdb: ?
23:46 jdb: (Oh that line-broke automatically at a bad place. I didn't mean the passive-aggressive post-question-mark, I swear).
23:57 jdb: Hm I should clarify, I meant more "Should I expect that to work out of the box, or will I have to go down the Zaphod/Xinerama path?"