00:21 mark4: need to ask some questions about nouveau with regard to hybrid gfx
00:21 mark4: will it work? is a good place to start
00:23 imirkin: will what work?
00:23 mark4: hybrid graphics with nouveau
00:23 mark4: my nvidia card has ZERO access to the displayn
00:23 mark4: NONE
00:24 mark4: is not connected to the display at all
00:24 imirkin: can you list some of your expectations wrt "working"?
00:24 mark4: only the integrated intel gpu is connected to the display
00:24 imirkin: [short version is yes - it should basically work]
00:24 mark4: how does nouveau handle hybrid gfx?
00:25 imirkin: PRIME
00:25 mark4: dont know what prime is
00:25 imirkin: the counterpart to optimus, if you watch transformers
00:25 mark4: i am familiar with the moive
00:26 imirkin: this may have some info useful to you: https://nouveau.freedesktop.org/wiki/Optimus/
00:26 mark4: there are two kinds of hybrid system
00:26 mark4: where GPU A and GPU B both have direct access to the display
00:26 mark4: turn one or other GPU off
00:26 mark4: cannot do that here
00:26 imirkin: right, that's the pre-optimus setup
00:26 mark4: NVIDIA -> INTEL -> display
00:26 imirkin: with a soft or hard mux
00:27 imirkin: the idea is that any rendering done on the nvidia gpu is handed to the intel ddx, which then displays it
00:27 mark4: right - and that will work with nouveau?
00:27 mark4: with prime?
00:28 imirkin: everything basically runs on the intel chip, except for the applications you explicitly select
00:28 imirkin: by running them with DRI_PRIME=1 in the environment
00:28 imirkin: yes, that works
00:28 imirkin: or at least is supposed to.
00:28 mark4: thats not encouraging
00:28 imirkin: ok
00:29 imirkin: what GPU do you have, btw, and what kinds of things do you expect to run on it?
00:29 mark4: i have a $2000 BRICK or i have a fred flintstone $2000 posing as a $900 laptop
00:29 mark4: 965 m
00:29 mark4: which nvidia claims is supported in linux
00:29 imirkin: lspci -nn -d 10de:
00:29 mark4: it has not worked on any kernel past 4.3.x
00:29 imirkin: nvidia's claims will be based on its proprietary driver
00:30 mark4: which is what that machine is using
00:30 mark4: im bootting it right now
00:30 mark4: but its a 965m i believe
00:30 imirkin: anyways, if it's a maxwell chip, while nouveau "supports" it, it won't work too well compared to what you might expect
00:30 imirkin: if it's a kepler chip, it should be reasonably well-supported
00:30 mark4: how do i know if its a maxwell
00:30 mark4: lspci does not show maxwell i dont think
00:30 imirkin: run that command, see what it outputs
00:30 imirkin: if it returns GKxxx it's kepler
00:31 imirkin: if it's GMxxx then it's maxwell
00:31 mark4: GM
00:31 imirkin: which one
00:31 mark4: 01:00.0 VGA compatible controller [0300]: NVIDIA Corporation GM204M [GeForce GTX 965M] [10de:13d9] (rev a1)
00:32 mark4: which means what for using nouveau?
00:32 imirkin: ok yeah. so that requires a signed blob in order to operate the PMU, which means ... no reclocking.
00:32 imirkin: nouveau should work reasonably well on it, but i'm not sure it'll beat out the intel gpu
00:32 imirkin: [in terms of maor fps]
00:32 mark4: the intel GPU is complete and UTTER garbage for 3d
00:32 mark4: garbage with a capital G
00:33 mark4: in fact, i cant even run glxgears
00:33 mark4: no opengl detected
00:33 imirkin: that means you have some kind of major configuration issue
00:33 imirkin: which is entirely unrelated to the intel driver
00:33 imirkin: and probably related to some incomplete attempts to operate the nvidia driver.
00:33 mark4: well im going to bite the bullet and install nouvau because beint stuck on a 4.3 kernel is unacceptable
00:34 mark4: no the nvidia driver works fine with this kernel
00:34 imirkin: yeah, but you're using the nvidia libGL and nvidia libglx etc
00:34 imirkin: so the intel driver's GL won't work
00:34 mark4: yes
00:34 mark4: no eselect opengl set xorg-x11
00:34 mark4: that deselects the nvidia opengl drivers
00:34 mark4: and... no glxgears
00:35 imirkin: did you restart X after that?
00:35 mark4: dont need to that i know of but yes ive tried everything
00:35 imirkin: well, you definitely need to do that in order to pick up the proper libglx.so to load into the X server
00:35 imirkin: anyways, if the intel GL doesn't work, nouveau's won't either. you'll have to fix your configuration.
00:36 mark4: get it installed first
00:36 mark4: fix later
00:36 mark4: ty
00:36 imirkin: :)
00:36 mark4: fix after install i mean
00:37 imirkin: sure. just add it to VIDEO_CARDS and rebuild mesa. i'd also recommend adding +dri3 to xf86-video-intel
00:37 imirkin: and then wipe your xorg.conf, and things should Just Work (tm)
00:38 imirkin: note that in order to get any accel from nouveau on there you'll need at least kernel 4.6, and the nvidia signed firmware from linux-firmware
00:39 mark4: lol
00:39 mark4: it just works (tm) rarely does
00:39 imirkin: well, if it doesn't, grab dmesg and xorg logs, and we can go from there
00:40 mark4: ya
00:40 imirkin: also - word of warning - nvidia did something funky on GTX 970's with 4GB of VRAM. if they did something equally funky with your gpu, then nouveau might not work at all.
00:41 mark4: were gona find out...
00:41 imirkin: although atm, the GTX 970 with 4GB vram is the only board with the funkiness
00:41 imirkin: [that i'm aware of]
00:44 skeggsb: there's some laptop boards i've had bug reports for with the same issue
00:44 imirkin: goodie
00:45 imirkin: skeggsb: any word on info to get that fixed properly?
00:45 skeggsb: pft
00:45 skeggsb: no
00:45 imirkin: is it easy to detect and just avoid the load entirely for now?
00:46 mark4: step 1 is to emerge gentoo-sources for kernel 4.9 and build that with nouveau support
00:46 mark4: skeggsb, what boards?
00:47 skeggsb: i don't recall off-hand
06:22 TvD: I am using Tails and getting a nouveau error message upon startup.
06:23 TvD: Yet the Gnome GUI works. No kernel driver is in use for my Nvidia Quadro according to `lspci -v`.
06:26 TvD: Can someone explain to me why?
06:27 skeggsb: knowing what this error message is would likely help there..
09:10 mupuf_: karolherbst_work: Thorsten liked the LED driver :D https://plus.google.com/+ThorstenLeemhuis/posts/aL8hhjdPT7o?sfc=true
09:12 karolherbst_work: :D
10:19 PaulePanter: Hmm. I am lost.
10:19 PaulePanter: > lspci -nn | grep VGA
10:19 PaulePanter: 01:00.0 VGA compatible controller [0300]: NVIDIA Corporation G98 [Quadro NVS 295] [10de:06fd] (rev a1)
10:19 PaulePanter: 18:00.0 VGA compatible controller [0300]: NVIDIA Corporation G98 [Quadro NVS 295] [10de:06fd] (rev a1)
10:19 PaulePanter: That’s NV98 according to the code names.
10:19 PaulePanter: But that is not listed in the feature matrix.
10:19 PaulePanter: https://nouveau.freedesktop.org/wiki/FeatureMatrix/
10:20 PaulePanter: Currently, the proprietary Nvidia driver 340.98 (legacy) is used for these.
10:20 PaulePanter: Where can I find out, how Nouveau from Linux 4.9 compares?
12:46 imirkin: PaulePanter: NV98 is one of the many chips in the NV50 family
12:46 imirkin: the families are what's described in the feature matrix
13:50 PaulePanter: imirkin: Ah, stupid me.
13:51 PaulePanter: imirkin: So besides power management, everything seems to be supported with the extracted firmware. Awesome.
14:06 imirkin_: PaulePanter: actually, you should be able to reclock it... only lightly tested though
14:12 PaulePanter: Even better. Thanks.
14:39 NanoSector: karolherbst_work: were your power management changes in 4.9 or were they staged for 4.10?
14:40 imirkin_: 4.10
14:40 NanoSector: dammit
14:40 thican: indeed :'(
14:41 NanoSector:wonders if it's safe to compile 4.10 from git now
14:42 thican: You might try to patch your 4.9 kernel with the recent modification from nouveau, no?
14:42 thican: *modifications
14:42 imirkin_: to compile it? sure ;)
14:42 NanoSector: oh I can just compile karolherbst's branch, that's right
14:42 NanoSector: imirkin_: and eventually run it :P
14:42 thican: imirkin_: well, isn't the window still open until December the 23rd?
14:42 imirkin_: something like that. but compiling it won't kill you
14:42 karolherbst_work: thican: for 4.10 yes
14:43 thican: oh, I understand the joke, ok :-)
14:43 NanoSector: heh
14:43 NanoSector: but will it kill my laptop? :>
14:43 thican: Foul me once, … ;-)
14:43 imirkin_: thican: "fool", usually...
14:44 imirkin_: NanoSector: i'd imagine drm-next is a lot safer, since it's 4.9 + just the drm stuff
14:44 NanoSector: https://teleirc.nanosector.nl/LrmhFqNg/file_619.jpg < I was bored during physics class.
14:44 imirkin_: whoa, there's a ti-84 now?!
14:44 NanoSector: that's from 2001 or something :P
14:44 imirkin_: definitely not from 2001
14:44 imirkin_: iirc the 83-plus only came out in like 2000 or 2001
14:45 NanoSector: oh, TI 84 is from 2004
14:45 NanoSector: these are cool: https://en.wikipedia.org/wiki/TI-Nspire_series
14:46 imirkin_:had a ti-89 in HS. it was awesome. it could do integrals.
14:47 NanoSector: this one does math...stuff
14:47 NanoSector: and pictures
14:49 NanoSector: anyway, is drm-next a branch of the kernel repo?
14:49 thican: NanoSector: I was using mine (83+ IIRC) for chemistry too! :-þ
14:50 NanoSector: :P
14:51 imirkin_: NanoSector: try airlied's repo
14:52 NanoSector: git://people.freedesktop.org/~airlied/linux ?
14:52 imirkin_: yes
14:52 imirkin_: it has many branches, one of them is called drm-next
14:52 NanoSector: fancy
14:54 NanoSector:mods an Arch PKGBUILD to compile that
14:54 NanoSector: heck, why not mainline
14:55 NanoSector: ohhhh that uses .tar.xz packages, right
15:00 NanoSector: ...this is going to take a while
15:03 NanoSector: oh, thought it'd be worse, it's advancing, gradually
15:27 NanoSector: distcc is so awesome with this, i've 3 systems compiling this package now and it's racing
15:28 imirkin_: multiple computers are faster than one computer... who knew
15:28 NanoSector: yeah, it's black magic at work
16:17 xen: hi! i've set foot on a very long road with my Xorg setup process. i'm trying to use 6 displays from 2 nvidia VGAs [770GTX, 710GT] as one Xorg Screen. nv's blob driver faild me yesterday. it has a limitation regarding the number of displays i can use in BaseMosaic mode on consumer cards.
16:18 xen: is this setup even possible with the open source driver?
16:19 imirkin_: yeah
16:20 imirkin_: depends on your specific GPUs, but in theory there's nothing stopping it
16:20 xen: that's really good news. could you tell me how?
16:20 imirkin_: how many outputs do you have on each one?
16:20 imirkin_: or rather, plan to use
16:20 xen: 4 on the 770GTX, and 3 on the 710
16:21 xen: this setup is fully functional under windows with nv's blob
16:21 imirkin_: my *guess* is that the 710 isn't wired up to use all 3 at once
16:21 imirkin_: ok
16:21 xen: they removed the limitation on windows
16:21 xen: for "feature parity", it remains in unix
16:21 xen: :)
16:21 imirkin_: in that case it should Just Work (tm). you'll need to use reverse-prime to slave the 710's outputs onto the 770
16:21 imirkin_: look at the relevant section here: https://nouveau.freedesktop.org/wiki/Optimus/
16:22 imirkin_: ("Using outputs on discrete GPU")
16:22 xen: wow
16:22 imirkin_: for the purposes of this discussion, the 770GTX is your "integrated" gpu, and the 710 is your "discrete" gpu
16:22 xen: these are unknown waters
16:22 imirkin_: and those commands won't work quite right... when you have it booted, let me know, and i can provide further guidance.
16:23 xen: if i understand you right, i might be able to use 3d acceleration this way, right?
16:23 imirkin_: yep
16:23 xen: oh my.. :3
16:23 imirkin_: note that nouveau is nowhere as advanced as nvidia's driver, along many axes
16:24 xen: thank you very much kind sir, i'm starting on the guide
16:24 imirkin_: xrandr --setprovideroffloadsink nouveau Intel
16:24 imirkin_: er
16:24 imirkin_: xrandr --setprovideroutputsource nouveau Intel
16:24 imirkin_: that needs to become
16:24 imirkin_: xrandr --setprovideroutputsource 1 0
16:24 imirkin_: since both your providers will be nouveau :)
16:25 xen: it doesn't have to be, i just need my desktop running with all the displays. if i can play some videos, it will be all right
16:25 imirkin_: some DE's will autoconfigure that stuff.
16:25 xen: it's a developer config after all
16:25 xen: let's hope so
16:25 xen: i'm using KDE plasma 5 with i3 as WM
16:25 imirkin_: oh, yeah, that won't work
16:25 imirkin_: tons of people have reported issues with plasma5
16:25 xen: it's gonna be just i3 then :D
16:26 imirkin_: my advice thus far has been to either not use it, or not use nouveau accel
16:26 xen: i got it to work with nouveau + gtx770[4displays] for now
16:26 imirkin_: [all the various kde applications have issues with nouveau right now... they changed something without ensuring that it'll work on the underlying libraries]
16:26 xen: do you mean that it wont work in this dual gpu setup?
16:27 imirkin_: nothing to do with dual-gpu
16:27 imirkin_: it's just unstable. according to reports. i have no idea, personally - i don't use any applications with a K in them
16:27 xen: youtube 4k is playing just now
16:27 xen: i'll stick with it until it breaks then :)
16:27 imirkin_: if it works for you - great. you've been duly warned.
16:28 imirkin_: perhaps they detect nouveau now and turn something off. dunno.
16:28 xen: ACK
16:29 xen: thank you again, i'll report if i got somewhere wtih it :)
16:29 imirkin_: and feel free to ask when you're trying to get it to work
16:39 karolherbst_work: *sigh* guess I won't find time this weekend as well
16:50 karolherbst_work: mhh crap, nothing really interests me at fosdem this year :O this will gonna be fun
18:56 philhug_: any developer interested in a GTX650? I have one I don't need anymore.
18:58 imirkin_: hakzsam: --^ you still want a kepler?
19:04 karolherbst: philhug_: you know if this one has any reclocking issues with the recent drm-next tree?
19:09 philhug_: karolherbst: I don't know, I'd have to install it again to try.
19:10 karolherbst: philhug_: would be really awesome if you would test it against the upcoming 4.10 kernel, because we know there are some issues on some cards, but currently I have no access to any of those
19:30 xen: i'm back
19:30 xen: xrandr --listproviders
19:30 xen: Providers: number : 3
19:30 xen: Provider 0: id: 0x10d cap: 0x7, Source Output, Sink Output, Source Offload crtcs: 4 outputs: 4 associated providers: 0 name:nouveau
19:30 xen: Provider 1: id: 0xa4 cap: 0xb, Source Output, Sink Output, Sink Offload crtcs: 4 outputs: 4 associated providers: 0 name:Intel
19:30 xen: Provider 2: id: 0x66 cap: 0x7, Source Output, Sink Output, Source Offload crtcs: 4 outputs: 3 associated providers: 0 name:nouveau
19:32 xen: xrandr --setprovideroutputsource 2 0 gives me a nice black screen for all the outputs connected to the 770, however, the cursor changes if i hover it over the -used to be- windows
19:38 xen: i also see the outputs of the 710 after this command in the xrandr output list, 2 of them appear connected, which is correct
19:39 xen: i hope i'm on the right track :)
20:18 philhug_: karolherbst: what are the issues I should experience?
20:19 philhug_: I just ran glmark2 on it and it went through without issues.
20:19 karolherbst: a freeze would be nice
20:20 karolherbst: did you set to the highest clocks?
20:20 philhug_: I set it to 0f
20:20 philhug_: fan switches on now from time to tiem
20:20 karolherbst: could you give me the content of pstate?
20:27 karolherbst: imirkin_: know anything of an issue like this? http://www.hug.cx/tmp/IMG_20161216_212453.jpg
20:28 karolherbst: philhug_: I assume this may me due to outdated userspace library versions, but imirkin_ might know more about it
20:33 philhug_: karolherbst: if there's anything else you want me to test on the card let me now.. I'll run glmark another few times to see if it remains stable
20:36 NanoSector: imirkin_: I installed the kernel you said but it didn't change anything for me
20:38 imirkin_: karolherbst: let me guess - xf86-video-modesetting?
20:38 karolherbst: philhug_: usually it should freeze real fast, everything which happens later than 30 minutes is usually due to other things or less important than immediaty hangs
20:38 karolherbst: philhug_: xorg log file pls
20:38 imirkin_: xen: sounds like everything's working
20:39 xen: except everything is black, except my cursor
20:39 imirkin_: NanoSector: can you remind me what your situation was?
20:39 xen: this might be the issue you were mentioning about KDE
20:39 imirkin_: xen: should there be anything displayed?
20:40 xen: yes, the desktop which i was issued the xrandr command from :)
20:40 imirkin_: heh
20:40 imirkin_: it could be that you're running into some dumb limitation
20:40 xen: the 4 displays are working flawlessly from the 770, until i add the outputs from the 710
20:40 imirkin_: can you pastebin xorg log?
20:40 imirkin_: [esp *after* the issue happens]
20:40 xen: yes, i think so
20:42 NanoSector: imirkin_: my laptop wouldn't resume from suspension
20:42 NanoSector: imirkin_: I also tried the kernel on my desktop and it broke everything
20:42 imirkin_: =/
20:43 NanoSector: amdgpu wouldn't load for whatever reason and i had radeon blacklisted to use amdgpu
20:45 NanoSector: reverted back to stock arch kernel and radeon on that machine though as I was getting pretty terrible lag in games even though it was 60 FPS
20:46 imirkin_: you realize this is #nouveau, right?
20:52 karolherbst: imirkin_: looks like it is nouveau indeed
20:52 xen: imirkin_: http://pastebin.com/pro
20:52 xen: http://pastebin.com/Fc9KG7m1
20:53 karolherbst: 1.0.13 even
20:53 xen: http://pastebin.com/xtD1EB31
20:53 philhug_: imirkin_: I have the same black screen with cursor visible issue. it happens after screen is waking up from sleep
20:55 philhug_: imirkin_: http://pastebin.com/8BLRb9Mq
20:55 imirkin_: xen: the first paste indicates that only the first 4 screens are turned on. is that the case?
20:56 xen: by default when i boot up, yes
20:56 xen: the second paste is after i issue the xrandr command
20:56 xen: then the xorg log after the crash
20:56 imirkin_: xen: ok, but the first xrandr was BEFORE you turn on the other screens right?
20:56 xen: yes
20:56 imirkin_: ah ok
20:57 imirkin_: xen: what is the command that you issue?
20:57 imirkin_: [ 5333.221] resize called 13440 2544
20:57 imirkin_: [ 5333.419] resize called 13440 1440
20:57 imirkin_: [ 5333.913] (EE) NOUVEAU(0): [DRI2] DRI2SwapBuffers: drawable has no back or front?
20:57 imirkin_: this seems ... unfortunate.
20:57 xen:  xrandr --setprovideroutputsource 2 0; xrandr > xrandr_proba
20:57 xen: this ^
20:57 imirkin_: no, i meant the command that breaks everything
20:57 xen: yes, this is it
20:58 xen: hence the logging part :)
20:58 imirkin_: errrrrr
20:58 imirkin_: the first xrandr command suggests that everything's already hooked up
20:59 xen: the first xrandr output shows that on the 770GTX, the 4 displays are connected.
20:59 xen: then i add the 2nd VGA with the 2 connected displays, 1 second later, crash
21:00 imirkin_: xen: the first xrandr shows that the second GPU is already slaved off of the first one
21:00 imirkin_: otherwise you wouldn't see those extra outputs
21:00 xen: if you are referring to this: http://pastebin.com/Fc9KG7m1, then yes
21:01 imirkin_: so my guess is that slaving it a *second* time isn't such a hot idea
21:01 imirkin_: also, note that the max fb width is 16384
21:01 xen: wait
21:01 xen: i only did that command once
21:02 imirkin_: then a robber came and did it while your head was turned
21:02 imirkin_: and then vanished into the darkness
21:02 xen: at boot, only this is shown: http://pastebin.com/eTYsEUbC
21:02 imirkin_: ah ok
21:02 imirkin_: so the xrandr you gave me was AFTER you ran the command
21:02 xen: i give the command, then i have ~ 1-2secs before everything turns black
21:02 xen: yes
21:02 imirkin_: i see.
21:02 xen: i thought i linked them in order
21:02 xen: sorry :\
21:03 imirkin_: you must be using some helpful piece of software which breaks everything for you.
21:03 xen: oh nice...
21:03 xen: what do you suggest for testing?
21:03 imirkin_: i'd start looking for things that have K in their names
21:03 xen: is there a way to issue this command after X inits, but before K* does?
21:03 imirkin_: something is noticing the extra outputs
21:03 imirkin_: and then tries to auto-configure them
21:03 imirkin_: and fails horribly
21:03 xen: yes, that might be the case
21:04 xen: KDE tends to do things on it's own
21:04 xen: :\
21:04 xen: a bit like M$
21:05 xen: if what i'm suggesting isn' possible, K* will be wiped, but i really like it aside from it's crappy default WM :\
21:06 imirkin_: i don't use KDE, so i really can't help with that
21:06 imirkin_: IME these systems are all quite configurable
21:07 imirkin_: however in the amount of time you spend figuring that configuration out ....
21:07 xen: i've spent ~3days to figure out how to replace the WM
21:07 xen: that's quite true :)
21:08 xen: for testing, i'm going to try i3 on it's own i think
21:08 xen: if that works, taming KDE would be a nice task for the weekend
21:09 xen: also, i have one more question lingering
21:09 xen: xrandr gave my my IGP as an output device
21:09 imirkin_: yeah, note the intel(G1) messages in your xorg log
21:09 xen: would it be possible to use that as well? [or it would be too greedy?]
21:09 imirkin_: absolutely
21:10 imirkin_: but again, note that kepler can't scan out from a framebuffer > 16384x16384
21:10 imirkin_: so ... arrange your screens carefully.
21:10 xen: xrandr --listproviders
21:10 xen: Providers: number : 3
21:10 xen: Provider 0: id: 0x10d cap: 0x7, Source Output, Sink Output, Source Offload crtcs: 4 outputs: 4 associated providers: 0 name:nouveau
21:10 xen: Provider 1: id: 0xa4 cap: 0xb, Source Output, Sink Output, Sink Offload crtcs: 4 outputs: 4 associated providers: 0 name:Intel
21:10 xen: Provider 2: id: 0x66 cap: 0x7, Source Output, Sink Output, Source Offload crtcs: 4 outputs: 3 associated providers: 0 name:nouveau
21:10 xen: yep, it shows
21:11 xen: that's a lot of pixels :)
21:11 xen: i thought IGP's are disabled when something is plugged in in PCI-E
21:12 xen: this is surprising
21:13 xen: just out of curiousity, how many slave devices can i have?
21:14 imirkin_: as many as you want
21:15 imirkin_: note that all rendering will be done on whatever you choose to be the primary gpu
21:15 imirkin_: and then the image will be shipped off to the slave gpu's for scanout
21:15 xen: i'm beginning to like this thing :)
21:15 xen: for what i'll be using it for, the 770 has to be strong enough
21:16 xen: on the other hand, if i make the 710 the primary one, and render all the 2D stuff with that, would i be able to use the 770 for ~dedicated openCL computing?
21:17 imirkin_: can't have nouveau and nvidia loaded at the same time on the same gpu
21:17 imirkin_: and nouveau doesn't support opencl
21:18 xen: oh :\
21:18 imirkin_: it does support opengl compute stuff though
21:19 philhug_: karolherbst: I'mm switch back to the GTX1060 then..
21:19 xen: i've never written a line in compute before
21:20 imirkin_: xen: it's a lot like opencl... but you use glsl instead of opencl c.
21:20 xen: this project drives me to interesting places :)
21:20 imirkin_: but you still have buffers, images, atomics, local memory, etc
21:20 imirkin_: (opencl local mem is called "shared" in glsl btw)
21:21 xen: that's all i need
21:21 imirkin_: no goto, or other unstructured control flow
21:21 imirkin_: [which is bad news for the gpu anyways, so you shouldn't be doing it]
21:22 xen: goto is bad everywhere
21:22 imirkin_: and you're stuck writing a bit of GL to make it all go, but it's not a big deal once you get the hang of it
21:23 xen: i'm in the image processing business [sort of] so i don't lack motivation :)
21:24 imirkin_: well, if you need *perf*, you'll find nouveau lags the blob driver by quite a bit
21:24 imirkin_: but if you don't mind a 30% hit, it should work ok
21:25 imirkin_: [the more memory bound you are, the less the hit]
21:25 karolherbst: philhug_: ohh, could you do am mmiotrace on it and retrieve the vbios through nouveau as well?
21:25 karolherbst: mainly the vbios, but I also want to have a good trace, but that's less important for now
21:26 xen: livestream processing is not a must right now
21:27 xen: i'm working on computation intensive stuff, but time is not that big of an issue, since it does not have to be ~realtime
21:27 imirkin_: ok
21:28 philhug_: karolherbst: sure
21:28 xen: this workstation will be just a playground anyways :) we have a server cluster for the heavy stuff
21:30 philhug_: karolherbst: what should I trace?
21:30 imirkin_: xen: i think xorg 1.19 integrates some important work wrt prime screen offload, so might want to upgrade to that
21:31 karolherbst: philhug_: first the vbios though and for tracing just normal desktop and at least one full reclock (you can check with nvidia-settings)
21:33 philhug_: how do I retrieve vbios through nouveau? can't find anything in /sys/module/nouveau
21:33 imirkin_: philhug_: /sys/kernel/debug/dri/0/vbios.rom
21:33 philhug_: ah, I already sent you
21:33 xen: imirkin_: do you mean the dev version from git?
21:34 imirkin_: xen: no, xorg 1.19 was released some time back
21:34 xen: oh
21:34 xen: i don't have testing repos enabled
21:35 imirkin_: make sure you also get xf86-video-nouveau 1.0.13 - that has a very important fix for output slaving
21:35 karolherbst: philhug_: uhh the old one was from the 650
21:35 karolherbst: philhug_: I meant from the 1060 ;)
21:35 xen: ok
21:35 philhug_: ah :)
21:36 philhug_: I'm still on 650. ok, will switch then
21:36 karolherbst: thanks for checking on the 650 though. I really need to get my hands on a not so well working one :(
21:40 xen: i've installed everything [according to pacman]
21:41 xen: see if it boots :)
21:46 xen: omg
21:46 xen: it works.
21:46 xen: :DDD
21:46 xen: with plain i3 though, no K*
21:47 xen: imirkin_: you deserve a barrel of beer
21:47 imirkin_: all 6 monitors light up?
21:47 xen: yep
21:47 imirkin_: very nice
21:47 xen: indeed it is
21:49 xen: oh
21:49 xen: i lost one
21:49 imirkin_: and like i said, adding on the IGP outputs should be no additional trouble... might only be able to do 2 of them, not 3. the CRTC limitations on intel chips are ... odd.
21:49 imirkin_: where'd it go?
21:49 imirkin_: fly out the window? :)
21:50 xen: i have no idea :)
21:50 xen: current resolution is 7680x2520
21:50 xen: it has to fit in somewhere
21:50 xen: however, it went dark :(
21:50 imirkin_: with older kernels, you might have had a 8k limit
21:50 imirkin_: but that was fixed ... some time ago
21:50 xen: i'm on 4.8.13
21:51 imirkin_: yeah ok, that's definitely recent enough
21:52 xen: wow
21:52 imirkin_: commit 5102ec3e99e, went into 4.3
21:52 xen: i changed the resolution from it's native
21:52 xen: and it came back
21:52 xen: odd...
21:52 imirkin_: anything in dmesg or xorg logs?
21:52 xen: switched back, and it's ok now
21:52 imirkin_: like a INVALID_STATE message of some sort?
21:52 xen: no errors
21:53 xen: just resize notifications
21:53 xen: and modeline settings
21:53 imirkin_: and nothing in dmesg?
21:53 xen: nope
21:53 imirkin_: =/
21:53 xen: i just have to find a way to persist this
21:54 xen: i presume by simply configuring with xrandr will reset when restarting X
21:54 imirkin_: yes.
21:54 imirkin_: and there's no way to do this slaving stuff in the xorg.conf i'm afraid
21:54 xen: that's not a problem, i can do it from a shellscript that autoruns @ login
21:55 xen: the painful part is done i think :)
21:56 xen: now i just have to figure out a way to make it work with KDE
21:56 karolherbst: mwk: are there no min/max instructions in the falcon ISA?
22:55 karolherbst: :O
22:55 karolherbst: crap, silly intel regression
23:27 trippeh_: I cant seem to get HDMI 2.0 working with GM206, is it supported in nouveau?
23:27 trippeh_: 24fps @ 4K is kind of annoying. ;)
23:34 imirkin_: trippeh_: mmmmaybe. you should be able to boot with nouveau.hdmimhz=533 and perhaps that'll make it go
23:34 imirkin_: i'm not aware of anyone who's tried it
23:37 imirkin_: trippeh_: it should definitely work over DisplayPort.
23:37 trippeh_: hm. I'll test hdmimhz in a few minutes.
23:38 trippeh_: I also may have a active DP to HDMI2 adapter that could work
23:39 trippeh_: ...somewhere
23:55 trippeh_: hehe, found the adapter, but plugging it in crashes the system.
23:56 trippeh_:tries the mhz hack