05:05mlankhorst: where's everyone?
05:41gnurou: mlankhorst: we are in the graphics devroom with tagr_, last row
06:13mlankhorst: a lesson about colorspaces, would have been useful 3 years ago when I first encountered it :P
06:17mlankhorst: I learned about it first when toying with wine's VMR implementation (similar to VDPAU's PAU part), learned even more when working on vdpau..
06:20RSpliet: hehehe, it's a good lesson nonetheless
06:21mlankhorst: without it the presentation will be useless
06:28mlankhorst: and if you mix up your color spaces you're going to have a bad time
06:31mlankhorst: it's not true about not knowing what colorspace is used for SDTV, it's supposed to be a bit set in the transport stream, unfortunately most media players ignore it :p
06:35RSpliet: (ugh, Dutch engineers playing to be angry and frustrated, toe-curlingly bad :p)
06:35mlankhorst: 16-240 is stupid though..
06:35RSpliet: oh I believe you
06:38mlankhorst: but color spaces are completely messed up
06:46RSpliet: I think the projectos interprets the data in the wrong colour space
06:46mlankhorst: no the ugly green color looks correct
06:47RSpliet: is that NVIDIA-green?
06:47mlankhorst: it predates nvidia
06:48mlankhorst: hm it looks like I forgot my charger..
06:49RSpliet: if it's ASUS, you can borrow mine
06:49mlankhorst: it's a macbook
06:49RSpliet: oh... I'm fairly sure that wouldn't fit
06:51mlankhorst: looks like I'll have to use my pocket computer then :P
09:51tobijk: hey somebody wants to review my second patch for arb_cull_distance? that'd make a basic but complete test set for this extension :)
09:54imirkin: i think you're looking for #dri-devel
09:54tobijk: yeah wrong channel, thats right
09:54tobijk: maybe you could have a look at it as well? :)
09:55tobijk: welcome back btw
09:55imirkin: i don't have a sufficiently strong handle on all that stuff. it generally makes sense, but... who knows.
10:07tobijk: you could review the last iteration of my folding folding patch, if you are bored :P
14:04koz_desktop: Is this supposed to fail to build with nouveau: https://github.com/CPFL/gdev
14:05imirkin: no, gdev is supposed to work with nouveau
14:05imirkin: (some bits of it, anyways)
14:05tobijk: waht does fail or how
14:06koz_desktop: tobijk: It fails to build with the version of boost that I have. The full error dump is *huge*, but I can paste it to you if you like.
14:07tobijk: some failing bits would suffice i guess :)
14:07tobijk: *for now
14:08koz_desktop: First few: http://dpaste.com/2EJGC1Q
14:09imirkin: koz_desktop: you may have better luck asking the gdev guys...
14:09imirkin: this seems like a straight-up issue in gdev and not something nouveau-related
14:09tobijk: well we can at least look at the errors :>
14:11koz_desktop: imirkin: So I should put up an issue on their Github?
14:11koz_desktop: I just wanna make sure it's not me being a retard.
14:12imirkin: koz_desktop: probably not retard... just using an incompatible boost version or something
14:12tobijk: mh looks like it
14:12tobijk: koz_desktop: you could see if these function(es) were removed for cleanup (remove of legacy code)
14:13tobijk: if thats the case, they'd need to adapt anyway at some point
14:14koz_desktop: I'll put an issue on their page and see what they say.
14:18koz_desktop: I found it when looking for instructions on how to reclock my card.
14:20imirkin: koz_desktop: which card do you have?
14:20koz_desktop: Let me double-check.
14:20imirkin: lspci -nn -d 10de:
14:20koz_desktop: GTX 650
14:21imirkin: boot with nouveau.pstate=1
14:21koz_desktop: Where do I find my grub config file again?
14:21imirkin: and you should be able to change clocks by echo'ing to /sys/class/drm/card0/device/pstate
14:21imirkin: dunno. mine's in /boot/grub/menu.lst
14:21koz_desktop: imirkin: Someone on here told me before that I can do it through GRUB parameters exclusively.
14:21tobijk: mine is in /boot/grub2/
14:22imirkin: koz_desktop: yeah. you can also add nouveau.config=NvClkMode=0f or something
14:22imirkin: might need to force post too... i.e. nouveau.config=NvClkMode=0f,NvForcePost=1
14:23koz_desktop: So the whole thing would be?
14:24koz_desktop: And do I need to alter the GRUB_CMDLINE_LINUX entry or the GRUB_CMDLINE_LINUX_DEFAULT entry?
14:25imirkin: dunno... that sounds like grub2-related insanity
14:25yurikoles: what is channel for mesa?
14:25imirkin: or perhaps something distro-specific
14:25imirkin: yurikoles: #dri-devel
14:27yurikoles: thank you
14:27koz_desktop: Well, here goes nothing. Hope this works...
14:28yurikoles: by the way I switched my card to r9 27x
14:28yurikoles: but I have gt240 ddr3 also
14:28yurikoles: how to enable the full power of gt240?
14:29yurikoles: I mean reclocking and so on
14:29tobijk: you mean higher clocks?
14:29tobijk: first step: boot with nouveau.pstate=1
14:29yurikoles: do we have dynamic reclocking now?
14:30tobijk: you need to reclock with /sys/class/drm/cardX/device/pstate file
14:32imirkin: yurikoles: not until reclocking is anywhere in the area of reliable
14:39koz_desktop: And it works! My card can now run in 0f.
14:40tobijk: yurikoles: for your new hw the r9, something radeon would benefit from hugely i guess is : https://www.opengl.org/registry/specs/ARB/viewport_array.txt
14:40imirkin: yeah, but that one won't be easy
14:41tobijk: it seems the infrastructure is there, as the older radeon cards support it
14:43imirkin: yeah, but it's going to mean a bunch of plumbing in the driver
14:44imirkin: and also enabling support for it will force-enable AMD_vertex_shader_viewport_index as well
14:44imirkin: and ARB_fragment_layer_viewport
14:45imirkin: the latter of which will need special care iirc
14:45imirkin: as well as interactions with llvm, which is enough to make me not want to look at it :)
14:46tobijk: heh ok :)
14:48tobijk: anyway, yurikoles, do soemething for the open drivers, we _need_ developers (you cant hide now, we'll get you :P)
14:49yurikoles: ok, guys
14:50yurikoles: for the very beginning I need to set-up some good GUI IDE that supports Makefile projects
14:51yurikoles: as I know qtcreator can handle autocrap?
14:51imirkin: dunno. i use emacs.
14:51tobijk: still eclipse :D (now i get a bat on the head)
14:52yurikoles: is there netbeans guys?
14:52tobijk: use wahtever you like and works with c/c++ projects
14:53yurikoles: as I know git plugin for eclipse was very buggy, is it true?
14:53tobijk: i use console git
14:53yurikoles: problem is not in the language but infrastructure
14:53yurikoles: me prefer to use CLion, but it lacks support of Makefile projects, only CMake
14:54koz_desktop: Is there a way to confirm that I'm running at 0f?
14:54imirkin: i'm sure i'll regret asking, but why would an editor care about the makefile system one uses
14:54koz_desktop: I'm not seeing any performance improvement...
14:54imirkin: koz_desktop: cat /sys/class/drm/card0/device/pstate
14:54tobijk: just cat pstate
14:55imirkin: koz_desktop: assuming you booted with nouveau.pstate=1
14:55koz_desktop: imirkin: Yes, I booted with nouveau.pstate=1
14:55koz_desktop: I just get a list of all the options...
14:55tobijk: yurikoles: for the infrastructure/build, make up a small bash script running the build?
14:56imirkin: koz_desktop: which line has the * on it? and/or what's under the -- thing
14:56imirkin: koz_desktop: actually just pastebin that file and also your dmesg
14:56koz_desktop: imirkin: Never mind, forgot a comma in my config file for GRUB. I'll do a reboot and see if it still works.
14:57yurikoles: the main issue is code analysing/completion
14:58imirkin: yurikoles: in emacs, there's a key-combo that completes based on tokens encountered in open buffers... Alt-/ iirc
14:58imirkin: that accounts for like 99.75% of use-cases
14:58imirkin: and for the rest, i use 'git grep' in a shell
14:58imirkin: a fancier user would have tags set up, but it's just not sufficiently necessary
14:59imirkin: i've used it in larger codebases though
14:59yurikoles: for now please vote for https://youtrack.jetbrains.com/issue/CPP-274 !
14:59yurikoles: all of you saw idea at least once
15:00yurikoles: now they have wrote ide for c and ++
15:00tobijk: that still makes a small problem
15:00tobijk: just use a bash script :)
15:01yurikoles: is there github mirror for mesa?
15:01yurikoles: nouveau kernel
15:01yurikoles: amd kernel?
15:02imirkin: not an official one... why would you want that?
15:02imirkin: linux kernel has one iirc
15:02tobijk: yeah but is outdated? now
15:02tobijk: was used after the kernel.org hack thingy
15:03yurikoles: ok, no problem
15:03imirkin: nah, github has some official mirrors set up that track other repos
15:04yurikoles: I don't like to wait for slow git servers on my gygabit link
15:04imirkin: you can clone mine from github and then fiddle with the .config to point it back at the fdo ones
15:04tobijk: fdo is fast enough
15:05yurikoles: but nouveau and radeon doing their developent in separate repos, yes?
15:05imirkin: there's a lot of diff things
15:05tobijk: mesa is one big repo
15:05imirkin: shared dev in mesa
15:05imirkin: for kernel everyone has their own trees
15:06imirkin: xf86-video-nouveau is just a single tree, which is tiny
15:06imirkin: libdrm is shared too
15:06imirkin: (and also tiny)
15:06yurikoles: https://github.com/imirkin/nouveau ?
15:06yurikoles: is this cut of kernel module?
15:06imirkin: that's a clone of ben's nouveau repo for the stand-alone driver
15:07imirkin: you want https://github.com/imirkin/mesa/
15:07imirkin: [note that i have some commits in my master branch that aren't upstream]
15:07imirkin: [and a bunch of random branches for various features in varying levels of completeness]
15:20koz_desktop: imirkin: File looks like this - http://dpaste.com/385B8VN
15:20koz_desktop: This is the exact result of cat /sys/class/drm/card0/device/pstate
15:21imirkin: koz_desktop: you're on the lowest level though... that's the "AC" line (you must be running on a super-new kernel)
15:21tobijk: try 0a or something first
15:22tobijk: echo 0a > pstate
15:22koz_desktop: tobijik: I thought I set it in the kernel at boot time.
15:22koz_desktop: But I'll try.
15:22imirkin: oh wait
15:22imirkin: not 0f
15:22koz_desktop: OK, correcting.
15:22imirkin: that one takes it as decimal, that's really confusing. skeggsb_ -- think that's worth fixing?
15:22tobijk: yeah thats worth a fix :)
15:23koz_desktop: OK, take two.
15:23tobijk: maybe he should have tried first on the console ;-)
15:29koz_desktop: OK, now I have 0f mode working.
15:29koz_desktop: However, I still get like, 10fps on Minetest. :(
15:29koz_desktop: (with max settings and 512 textures, sure, but still)
15:30tobijk: maybe your test is crap? :D
15:30koz_desktop: tobijk: Minetest is a pretty fun game.
15:30tobijk: bit is it good for testing?
15:30imirkin: koz_desktop: are you sure it's working? pastebin pstate output?
15:31koz_desktop: imirkin: http://dpaste.com/1K78QFT
15:31koz_desktop: That looks like working to me.
15:31imirkin: AC: core 324 MHz memory 5022 MHz
15:31imirkin: memory clocked up but core didn't
15:31imirkin: that's... unfortuante
15:32tobijk: core should be ~1070
15:32imirkin: koz_desktop: what happens if you echo 0f to pstate (as root)?
15:33koz_desktop: No change.
15:33tobijk: can you do 0a > pstate
15:34koz_desktop: Memory gets lower, core stays put.
15:35imirkin: koz_desktop: pastebin dmesg?
15:35koz_desktop: Where is dmesg?
15:35tobijk: dmesg > yourfile.extension
15:36koz_desktop: imirkin: http://dpaste.com/1ABNVAC
15:37tobijk: [ 408.574746] nouveau E[ CLK][0000:01:00.0] failed to raise voltage: -22
15:37imirkin: first off -- nouveau.config=NvClkMode=15, NvForcePost=1
15:37imirkin: no space after the ,
15:38imirkin: not sure if that will help the voltage raising failure though
15:38tobijk: i dont think that will make a difference here?!
15:39imirkin: dunno, skeggsb_ said it made a diff for him to get reclocking working on some of his cards
15:39koz_desktop: I should still correct it, though, if it's not meant to be there.
15:39koz_desktop: OK, I'll reboot (again) and see if this does anything.
15:39imirkin: i tend to trust him on such issues :)
15:39tobijk: well we all rely on him :D
15:41koz_desktop: Well, removing that comma changed my memory clock to be higher.
15:41koz_desktop: Core clock is *still* exactly where I left it.
15:42koz_desktop: imirkin: Do you wanna see the new dmesg?
15:43imirkin: koz_desktop: sure
15:44imirkin: koz_desktop: remove the space, not the comma
15:44koz_desktop: Sorry, I meant space.
15:45koz_desktop: imirkin: http://dpaste.com/09SXREW
15:45tobijk: stil lthe same error :/
15:47koz_desktop: tobijk: What *is* the error, exactly? My dmesg is like, 800 lines long.
15:47tobijk: [ 3.240938] nouveau E[ CLK][0000:01:00.0] failed to raise voltage: -22
15:47koz_desktop: Well... that's lame.
15:48tobijk: that why reclokcing is _experimental_ ;-)
15:48imirkin: 22 = EINVAL
15:48koz_desktop: tobijk: Well, I knew *that*. :P
15:49koz_desktop: Mostly wanted to make sure I'm not missing something retarded.
15:49tobijk: 3.18 reclocking was working on other cards
15:50koz_desktop: Would giving my kernel version help, or does dmesg already give it to you?
15:50tobijk: we have it
15:52tobijk: after blinking over the 3.18 stable tree, this should not be a regression in the 3.18 series
15:52tobijk: just a buggy card :/
15:53koz_desktop: It's a non-reference ASUS card.
15:53koz_desktop: Could that have anything to do with it?
15:54imirkin: it could be a reason why it's trying to do something we haven't seen before
15:54imirkin: could you file a bug that includes your dmesg and your vbios (cat /sys/kernel/debug/dri/0/vbios.rom)
15:54tobijk: silly thing is that the gddr5 memory is actually reclocking just fine, that normally causes the problems :>
15:54koz_desktop: Where do I file the bug?
15:55imirkin: bugs.freedesktop.org under xorg Driver/nouveau
15:55koz_desktop: imirkin: On it.
15:58koz_desktop: How do I check my Xorg version?
15:59imirkin: don't worry about that
15:59imirkin: just put unknown or git or something
15:59imirkin: it's a historical accident that Driver/nouveau is under nouveau... or really we should create a new component, but... oh well
15:59imirkin: that Driver/nouveau is under *xorg*
16:01koz_desktop: Can I only add one attachment?
16:02tobijk: add one first
16:02tobijk: the other ones after that
16:04koz_desktop: Submitted the bug.
16:04koz_desktop: Hopefully this'll help.
16:04koz_desktop: But yeah, you guys are doing awesome work!
16:07tobijk: koz_desktop: thanks for reporting! :)
16:07koz_desktop: Happy to help. I just want to make Minetest look pretty at full power.
16:07koz_desktop: Because having ~10fps does not for a nice game make.
16:28imirkin: tobijk: please familiarize yourself with the various features supported by nouveau before pinging bugs to see if it's still an issue
16:28tobijk: well sorry, but at least im trying to clean up the buglist
16:29imirkin: tobijk: yeah, i know. but over-pinging people is silly too...
16:29imirkin: tobijk: i cleaned up the bug list a while ago too... sort of have been trying to keep it clean
16:30imirkin: tobijk: the vast majority of the bugs on there are things that i don't think have been fixed
16:31tobijk: was not aware of a cleanup effort on your side, good to know
16:32imirkin: well, it was over a year ago
16:33tobijk: that makes it ahead of my time being involved ;-)
20:34Tom^: holy moly im running cs:go native on my 780ti at higher fps with pstate set to 0a then i do with the nvidia blob O_o
21:15yurikoles: OH SHI~
21:15yurikoles:is looking for the way to replace back 270x by 750ti
21:30mgalgoci: what's the general problem when you see: "NOUVEAU(0): [drm] failed to set drm interface version."