04:07 dikiy_: Hi all. I try to downgrade to mesa 10.0
04:07 dikiy_: Have an arch distro. I have downloaded the mesa-10.0, mesa-libgl-10.0 and nouveao-dri packages.
04:08 dikiy_: I try to install them with pacman -U, but it says: xf86-video-nouveau: need mesa-dri. The question is, can I downgrade xf86-video-nouveau too, without downgrading a whole Xorg srver?
04:20 dikiy_: another question, how can I enable software OpenGL, as a workaround of not working hardware 3D on my NV46?
04:24 dikiy_: prg: thx! it works.
04:26 dikiy_: slow, but it works :)
06:22 dikiy_: does Mesa build with llvm 3.6?
06:23 dikiy_: I try do build Mesa 10.4.x with llvm 3.6 and it fails...
08:59 zezic: Hi! I want to use Xinerama with 1 monitor on integrated Intel videocard and 1 monitor on NVidia PCI-E card with nouveau. It is possible?
09:00 zezic: I read that nouveau fully supports xinerama, but can i mix different GPU drivers to show one screen over multiple cards?
09:00 dardevelin: hi folks
09:01 dardevelin: anyone knows the state of the art regarding NVIDIA Geforce GTX850 4GB
09:01 dardevelin: before I dive into google foo for information
09:01 dardevelin: cheers
09:02 dardevelin: I thin the laptop in question has options in uefi to disable the nvidia one and just go with the intel one: https://www.worten.pt/portatil-avancado-15-6-asus-n551jk-cn220h-i7-4710hq-12-1-128-silve.html
09:02 dardevelin: s/thin/think
09:04 dardevelin: another place with more info on the laptop: http://www.fnac.pt/Asus-K550JK-XO172H-Computador-Portatil-Computador-Portatil/a853855#ficheDt
09:04 dardevelin: woops wrong link
09:05 dardevelin: this is the other place (correct link) : http://www.fnac.pt/Asus-N551JK-CN220H-Computador-Portatil-Computador-Portatil/a853856#ficheDt
09:10 imirkin: zezic: that's what xinerama was designed for :)
09:11 imirkin: dardevelin: is that a GM107 or GM108?
09:11 dardevelin: GM ?
09:11 imirkin: dardevelin: lspci -nn -d 10de:
09:12 dardevelin: I don't own the computer yet, sorry. I am just trying to access before buying
09:12 dardevelin: sorry
09:12 imirkin: oh
09:12 dardevelin: I know it's probably not going the most friendly, but at the current moment it really only is 'worth' buying a _high end_ like this
09:13 imirkin: according to pci.ids, it's probably a GM107, which starting with kernel 4.1 should have acceleration support
09:13 dardevelin: :)
09:13 imirkin: but it also likely won't be too useful if you're looking to actually use it for gaming
09:13 dardevelin: no not all
09:13 imirkin: since there won't be any reclocking
09:14 tobijk: imirkin: a question to your patch12/12 of the tess series...
09:14 imirkin: tobijk: sure
09:14 dardevelin: I am not a gamer. I just want to at least be able to use it somewhat properly
09:14 imirkin: dardevelin: then the intel should be more than sufficient
09:14 tobijk: shouldnt you add the removed NV50_SEMANTIC_CLIPDISTANCE as a TGSI_* one again?
09:15 dardevelin: yeah, nonetheless trying to get something that at least does some work would be good
09:15 imirkin: tobijk: that's what TGSI_SEMANTIC_CLIPDIST is
09:15 dardevelin: I don't buy computers often
09:15 dardevelin:tries to not enter into consumer fest mode
09:15 imirkin: :)
09:15 zezic: imirkin: thank you!
09:15 tobijk: there is no diff between the 0x04 and 0x10? :O
09:15 imirkin: my current box is a i7-920. the one before that was an athlon xp 1700+. i don't buy computers often either ;)
09:16 imirkin: tobijk: no, there is. and 4 is wrong for how TGSI_SEMANTIC_CLIPDIST is used
09:16 imirkin: 4 would be right if only the .x component of clipdist were used
09:16 tobijk: ah ok
09:16 dardevelin: example my current laptop has like 5y
09:16 tobijk: just making sure that isnt missing something obviously :)
09:17 dardevelin: and age is hitting on it, but for my usage (the only handy cap now and then is really ram) which at the current moment due to the age of the computer and price it would cost to upgrade that would make no sense
09:17 imirkin: did it work well 5y ago?
09:18 dardevelin: Nah. :) it's only running decently with nouveau somewhat recently
09:18 imirkin: are you doing anything on it now that you weren't doing 5y ago? or is the argument that the ram gets slower/smaller with age?
09:19 dardevelin: imirkin, smaller as consumption really is higher. blame web browser intensive behavior
09:19 tobijk: hehe
09:19 dardevelin: and the age is coming to terms. I had to change thermal paste
09:19 dardevelin: i notice some bits on ram that don't report as existing
09:19 dardevelin: hard drives were clearly changed
09:19 dardevelin: but for 'space increase'
09:20 dardevelin: it's the gfx card as ps3
09:20 dardevelin: nvidia 9300gs
09:20 imirkin: i think ps3 was closer to a G72 (NV49 or so)
09:21 imirkin: or was that ps2?
09:21 imirkin: no, ps3
09:21 dardevelin: I love this computer specially the keyboard is just a wonder
09:21 dardevelin: wander*
09:21 dardevelin: eh i had it right the first time
09:23 dardevelin: the other things that make me interested in that laptop is if I am not mistaken actually has support for addition of my own keys uefi vs only all on/off uefi mode
09:24 dardevelin: and a uefi option to actually disable the nvidia and just stay with the intel one active
09:24 dardevelin:still has to check the wifi card to see if would required a proprietary driver as I wish to avoid them
09:25 dardevelin: heck if i had some 'guidelines' I would not mind trying to start using my 'free' time to help on reverse engineering the existing drivers
09:25 dardevelin: proprietary ones I mean
09:25 imirkin: dardevelin: not much sense RE'ing nouveau ;)
09:25 dardevelin: imirkin, hence clarification (before I look like an idiot) :)
09:26 imirkin: dardevelin: which card do you have again? G98?
09:26 imirkin: stay silent, people will think you a fool. open your mouth, and remove all doubt :p
09:27 dardevelin: yes
09:28 dardevelin: 01:00.0 VGA compatible controller [0300]: NVIDIA Corporation G98M [GeForce 9300M GS] [10de:06e9] (rev a1)
09:28 dardevelin: not sure you remember me, I had that non-bootable kernel
09:28 dardevelin: and you helped me extract the firmware
09:28 dardevelin: ended up being an addition of an 'if statement'
09:29 dardevelin: imirkin, if you search for dardevelin: http://people.freedesktop.org/~cbrill/dri-log/?channel=nouveau&date=2013-12-17
09:39 imirkin: and do you have GDDR3 VRAM or DDR2?
09:46 dardevelin: imirkin, in the current graphics card ?
09:48 dardevelin: 256mb dedicated ddr2 if not mistaken
09:49 dardevelin: imirkin, the link i showed you was just reference in case you did not remind me. the issues has never reappeared
09:49 dardevelin: that was just one of the incidents i have experienced with this card
09:57 imirkin: yeah, the reason i asked about the ram type is that RSpliet has been working on reclocking for cards including yours
09:57 imirkin: but i think he's only done gddr3 for now
10:11 dardevelin: imirkin, I see
12:27 zezic: Hi! Just install 3rd monitor. Now i have 2 monitors on Intel card and 1 on GeForce GT 610 with nouveau.
12:28 zezic: I got image corruption like this: http://i.stack.imgur.com/L5G5Q.png
12:28 zezic: But when i'm screenshot it's okay.
12:29 zezic: All 3 screens with 2560x1440 resolution.
12:29 zezic: Anybody knows how can i fix it?
12:30 zezic: Also got trailing from mouse cursor
12:35 zezic: How exactly it looks http://cs621830.vk.me/v621830448/2ba06/O3J-p63ryM8.jpg
12:44 zezic: http://cs621830.vk.me/v621830448/2ba17/s6qiw15SklM.jpg
12:55 imirkin_: zezic: you're not using xinerama but rather reverse-prime i assume
12:55 imirkin_: zezic: nouveau just draws the image that intel tells it to draw
12:55 zezic: How to check it?
12:55 imirkin_: aka your problems are with the intel ddx, not nouveau ddx
12:55 imirkin_: zezic: did you manually set up xinerama?
12:56 zezic: I just set screens positions with xrandr
12:56 imirkin_: yeah, so that's "reverse prime"
12:56 zezic: http://cs621830.vk.me/v621830448/2ba31/iu2TcERMPis.jpg
12:56 imirkin_: pop over into #intel-gfx and complain there :)
12:56 zezic: Left screen plugged in nouveau driven card
12:56 imirkin_: i like it ;)
12:56 zezic: imirkin_, can i set up xinerama?
12:56 zezic: lol
12:57 imirkin_: a lot of people do use reverse prime successfully
12:57 imirkin_: and a bunch end up with the issues you're seeing
12:57 imirkin_: i have no idea what separates one group from the other
12:57 imirkin_: some sort of desktop setup/version/etc situation
12:57 imirkin_: to set up xinerama, you'll want something like http://nouveau.freedesktop.org/wiki/MultiMonitorDesktop/
12:57 imirkin_: i assume that intel ddx supports zaphod heads
12:58 imirkin_: i guess it might not =/
12:58 buhman: heh, tiny desk; not enough room for monitors
12:58 zezic: =)
12:58 buhman: I too once hung my monitor stands over the edge of a desk
12:58 imirkin_: yeah, i was about to comment -- you might want to invest another $30 into the desk to hold your $1.5k in monitors
12:58 tobijk: lol, yeah nice picture ^^
13:01 buhman: imirkin_: wouldn't it make sense to have the nouveau gpu tell the intel gpu what to do instead?
13:01 buhman: isn't that just reversing the setprovidersource thing?
13:01 imirkin_: buhman: you could set it up that way, sure. unfortunately you're sort of at the whim of xorg as to which is primary, and it's ~impossible to change
13:02 buhman: O.o
13:02 imirkin_: i guess it goes in bus order, and intel tends to be first
13:06 airlied: whatver the bios uses is primary
13:06 airlied: nothing to do with bus ordering
13:06 imirkin_: ah. well i stand by the "~impossible to change" comment ;)
13:07 imirkin_: airlied: do you know what separates the two groups of people btw?
13:07 imirkin_: i.e. the people with crazy artifacts and the people without
13:07 airlied: my guess is usually intel ddx changes
13:07 airlied: but it could also be different intel gpus
13:07 imirkin_: in that newer versions are better, worse, or randomized?
13:10 airlied: random maybe
13:10 airlied: I've only got a sandybridge laptop to test offload on
13:10 airlied: and I don't get to it that often
13:11 imirkin_: i thought it might be compositor setup too
13:11 imirkin_: (i always try to blame the compositor boogeyman on all my troubles)
13:11 imirkin_: (or rather, other people's troubles, since i don't use one)
13:12 airlied: well if they are using kwin I generally just hide
13:13 imirkin_: from the looks of it, gnome-shell seems equally, uh, high-quality
13:13 imirkin_: e.g. if you have a nouveau_vieux card, it just dies
13:14 airlied: you think that is more likely the nouveau_vieux driver :)
13:14 imirkin_: [which i assume also goes for any pre-GL2 gpu]
13:14 imirkin_: nouveau_vieux actually works surprisingly well
13:14 imirkin_: piglit is definitely able to trip it up
13:14 airlied: yeah compositing on a 15 year old hw doesn't work so well
13:14 airlied: you kinda need vram
13:15 imirkin_: in units larger than bytes? :)
13:15 airlied: and then people plug 1080p monitors into them
13:15 imirkin_: the issue is that people come in here and are like "waaa, ubuntu 15.04 doesn't work"
13:15 airlied: when originally the cards would be lucky to see 1024x768
13:15 imirkin_: that's not true -- 1280x1024 was *quite* common
13:15 imirkin_: and i had a 1600x1200 screen at the time
13:16 imirkin_: one of those 19" sony trinitrons. *awesome* screen
13:16 airlied: I thought the only 1600 screens at the time were SGI ones
13:16 imirkin_: geforce4 wasn't *that* long ago
13:16 imirkin_: it was well into 2000
13:16 imirkin_: in 1995, sure, the only 1600x1200 were the sgi screens
13:17 imirkin_: with their crazy 13w3 connectors
13:17 imirkin_: which were different from the sun 13w3 connectors because... they could
13:18 imirkin_: [no, i'm not bitter at all that i couldn't plug an sgi monitor into a sparc]
13:23 zezic: Guys. Just switched MB BIOS to init with PEG and nouveau display works OK, but Intel displays are black.
13:23 zezic: xrandr shows this: http://pastebin.com/WiimGMr2
13:23 imirkin_: can't win 'em all
13:23 imirkin_: i think you can't have 2 screens reverse-primed
13:23 zezic: I know that it's not nouveau fault, but, may be someone knows where is my mistake.
13:24 imirkin_: although someone proved me wrong the other day
13:24 imirkin_: airlied: can you update on what the latest is there?
13:24 zezic: xrandr shows correct Screen 0 size
13:24 airlied: yes two screens should work
13:24 zezic: 7680 x 1440
13:24 airlied: but not sure howe ell nouveau->intel works
13:24 imirkin_: zezic: btw, should i ask why you use a 57Hz mode for the HDMI1 screen instead of 60Hz?
13:25 zezic: imirkin_: because this screen plugged in integrated HDMI port.
13:25 imirkin_: airlied: what about offsets for reverse prime screens?
13:25 imirkin_: (in general)
13:25 airlied: yup they should work at most place, though not sure when the X server support landed
13:26 airlied: if people aren't using Fedora I'm generally lost wrt prime
13:26 imirkin_: airlied: as well as multiple screens on the "remote" gpu?
13:26 airlied: yes,
13:26 imirkin_: awesome. i seem to recall that didn't work not too long ago
13:26 airlied: it's workedi n Fedora/RHEL for a long time
13:26 airlied: upstream was less successful
13:26 imirkin_: ah ok
13:27 zezic: I make screenshot with scrot and it shows me corrent picture with 3 screens, hmm.
13:28 buhman: lies
13:28 zezic: But 2nd and 3rd screens blank really.
13:29 tobijk: do these kind of things work with windows in general? heh
13:29 imirkin_: yea
13:29 imirkin_: optimus works
13:29 imirkin_: they expend time and resources to make sure it works
13:29 airlied: windows 7 I think is pretty good, anything before that was a problem
13:30 airlied: vista had some serious issues
13:30 imirkin_: whereas here it's a side project of airlied's basically
13:30 airlied: and they all look pretty ugly
13:30 airlied: OSX are the only ppl to do it righ
13:30 buhman: I can't wait until everyone talks about how ancient and terrible kernel 4.1 is
13:30 airlied: and they have the advantage of picking their hw
13:30 airlied: nobody else does reverse optimus though
13:30 airlied: they switch to using the nvidia as primary on the fly when it is driving outpts
13:31 imirkin_: depends on one's definition of 'optimus'
13:31 tobijk: ;-)
13:31 airlied: but doing that with X and GLX especially was hard
13:32 tobijk: will EGL help with that when the time comes?
13:32 airlied: no because its a legacy problem
13:32 tobijk: in the distant wayland time? ;-)
13:33 airlied: EGL at least had a bit more defined interface for losing contexts
13:33 airlied:has to revisit it all with wayland
13:33 tobijk: O.o
13:33 airlied: its per compositor to deal with it
13:34 airlied: but mutter doesn't even do half the optimus stuff yet
13:35 tobijk: yay, looks really bright
13:35 tobijk: thats a nice picture you are drawing for your future work
13:35 tobijk: (i guess you have to do it)
13:40 tobijk: airlied: btw, do i have the ok from you to add your sign-off to your lowering_cull_distance patch?
13:40 imirkin_: tobijk: never add tags without people explicitly telling you it's ok
13:40 tobijk: thats why i'm asking
13:40 imirkin_: :)
13:42 airlied: tobijk: if it works :)
13:42 tobijk: airlied: as far as i can see it... :)
13:45 imirkin_: airlied: so what's next on your "easy gl extension" front? ARB_clear_texture/ARB_copy_image?
13:46 airlied: imirkin_: spir-v :-P
13:47 imirkin_: yes, that sure seems easy!
13:47 imirkin_: heh
13:47 airlied: I also have half started bindless texture
13:47 imirkin_: ah cool
13:47 imirkin_: for radeon i assume?
13:48 imirkin_: if you have plans for bringing it to nouveau, let me know, happy to provide some pointers
13:48 airlied: for softpipe :)
13:48 imirkin_: kepler+ is all bindless
13:48 imirkin_: heh ok
13:49 airlied: though I expect I should spend some time on r600 now to try and get gl4 on it
13:49 imirkin_: fp64 and gs5 are pretty close i guess
13:49 airlied: thugh the clear/copy ones could be good for gallium to grow support for
13:49 imirkin_: just tess left
13:50 imirkin_: you should also add accelerated ARB_clear_buffer_subdata support to radeon
13:50 imirkin_: you have to provide a hook, otherwise it does a software clear
13:50 imirkin_: [dunno why it was added that way to mesa...]
14:00 tobijk: airlied: i have checked it yet again. i'm happy with it :D
15:20 imirkin_: hrmph. one of blob and nouveau has a lot to learn about calculating sched codes. sadly i'm betting on nouveau.
17:36 tobijk: imirkin_: we are not cleaning up properly after the tess bits which are already commited :>
17:44 tobijk: imirkin_: quick fix till you commit the tess series: http://hastebin.com/bepafuyure.vhdl
18:59 imirkin: skeggsb: mwk: any objections to making envydis/gk110 sched take 7 values like gf100 does instead of 1 big one?
18:59 imirkin: it'll mean we have to fix up the existing gk110 asm shaders that have been written, but i don't think it's such a huge deal
18:59 imirkin: esp since it's 8 its per value
19:02 skeggsb: that'd be easier to read, so no objection from me
19:12 imirkin: ok, i'll push something out tomorrowish
22:25 pmoreau: RSpliet: Cool!! :) I'll give it a try… some time… Hum… Maybe tonight if I have time