02:52 karolherbst: imirkin: okay, poking the same regs on nouveau doesn't change the performance
02:53 karolherbst: but I am sure when nouveau does this zcull thing right, we can expect around 0-10% more perf
03:11 RSpliet: karolherbst: depends a lot on the workload of course :-)
03:12 karolherbst: yeah I know
03:12 karolherbst: mhh
03:12 karolherbst: I just enabled zcull in mesa
03:12 karolherbst: and heaven is just white now
03:12 karolherbst: :D
03:12 RSpliet: yes
03:13 RSpliet: I figured it wasn't as simple as flicking the magical zculling switch
03:13 karolherbst: simple benchmark still work though
03:14 RSpliet: I'm kind of curious how this early-z thing would work... since you haven't done your viewport transformation yet, right?
03:14 RSpliet: I... damn it, back to MIAOW
03:15 karolherbst: RSpliet: well earlyz is done earlier then all the other Z optimisations
03:15 karolherbst: I think earlyz checks pixels directly or something
03:16 RSpliet: early-z doesn't run *before* vertex shaders right?
03:17 karolherbst: it runs before pixel shaders, that's I am sure off
03:17 karolherbst: maybe even before fragment
03:17 RSpliet: fragment shader == pixel shader
03:17 karolherbst: ohh okay
03:17 karolherbst: then before fragment
03:17 karolherbst: zcull runs 100% before fragment
03:17 RSpliet: yeah, that makes sense, the point is reducing fragments to process in your shaders :-)
03:18 karolherbst: right
03:18 karolherbst: mhhh
03:18 karolherbst: antichamber won't start now :/
03:18 RSpliet: but you don't know the location of a vertex (or fragment for that matter) until you've done your vertex transformations, so logically I'd apply them afterwards
03:19 RSpliet: the name early-z implies otherwise though :-P
03:19 karolherbst: ohh right, maybe I shouldn't play with LD_LIBRARY_PATH inside steam
03:19 karolherbst: RSpliet: I think eary-z means just earlier than any other z optimisation
03:19 RSpliet: that would make sense :-)
03:20 karolherbst: "GeForce 6 Series GPUs are able to cull nonvisible primitives before shading at a high rate and clip partially visible primitives at full speed. Previous NVIDIA products would cull nonvisible primitives at primitive-setup rates, and clip all partially visible primitives at full speed."
03:20 karolherbst: that's under Early Culling/Clipping
03:20 RSpliet: "before shading" :-D
03:21 karolherbst: but the dev docs are a bit, well
03:21 karolherbst: RSpliet: yeah well, GeForce 6000 was new then
03:21 karolherbst: :D
03:21 RSpliet: I know, I have one of those (well, technically a 7600, but meh)
03:21 RSpliet: it does do vertex and fragment shading
03:22 karolherbst: https://gist.github.com/karolherbst/fd2df7783807f252d6a2
03:23 karolherbst: best parts: "Coarse Z/Stencil culling (also known as ZCULL) " and "Similarly, fine-grained Z/Stencil culling (also known as EarlyZ)"
03:23 karolherbst: and the exampls when it doesn't drive good
03:23 karolherbst: because that might tell us how it works
03:24 karolherbst: "2. If the pixel shader writes depth."
03:24 karolherbst: that's a good point
03:24 karolherbst: that basically means: if one of the pixel shaders does that, we just disable zcull for that "area"
03:25 RSpliet: yes, that makes sense
03:25 karolherbst: so zcull is just a "stupid" algorithm basically
03:25 karolherbst: which just eliminates the most obvious things
03:25 karolherbst: at least it sounds like it
03:26 RSpliet: "stupid" algorithms are often the most effective
03:27 karolherbst: yes
03:28 RSpliet: that depth test doesn't sound very trivial :-D
03:29 karolherbst: I have no clue about gl anyway
03:29 karolherbst: meh :/
03:29 RSpliet: I'm learning...
03:30 karolherbst: I was just thinking if I push some bs zcull commands to the gpu, some stuff should be there and some not
03:30 karolherbst: and then I could play around a little
03:31 RSpliet: zculling is in my lecture for tomorrow :-P
03:31 karolherbst: :D
03:31 karolherbst: seems like you will be the man for that
03:32 RSpliet: can you download http://www.cl.cam.ac.uk/teaching/1516/CompGraph/CGIP.pdf?
03:32 RSpliet: http://www.cl.cam.ac.uk/teaching/1516/CompGraph/CGIP.pdf ?
03:33 karolherbst: yeah
03:33 karolherbst: well don't know if the second one
03:33 karolherbst: still waiting
03:33 RSpliet: page 42 (40) :-P
03:33 RSpliet: second one is the correct link for the first one
03:33 karolherbst: ohh
03:33 RSpliet: accidentally added a question mark, shouldn't matter though
03:33 karolherbst: yeah, saw it
03:33 karolherbst: no
03:33 karolherbst: my irc client is smart
03:34 karolherbst: mhhh
03:34 karolherbst: but that's about a specific culling thing, right?
03:35 karolherbst: like if you look at a house and you don't see the back wall, you don't draw it
03:35 RSpliet: that's one form of culling yes
03:36 RSpliet: but it also explains the zbuffer :-P
03:36 karolherbst: ahh okay
03:36 RSpliet: slide 240 explains the problem
03:39 karolherbst: yes, makes sense somehow
03:40 karolherbst: maybe I should stay with the stuff I can do best: messing up the kernel
03:40 karolherbst: :D
03:42 karolherbst: lol
03:42 karolherbst: I found the problem with fear
03:43 karolherbst: prg: I think the problem is, that fear just stress bechmarks the gpu
03:43 karolherbst: this even got my intel gpu into a hang state
04:24 sillysausage: hmm, my workstation sort of froze suddenly. I could hear audio in the background of a video i was playing.
04:25 sillysausage: The mouse came back briefly
04:25 sillysausage: keyboard PS/2 stopped responding
04:25 sillysausage: nouveau E[ PFIFO][0000:04:00.0] write fault at 0x000a300000 [PAGE_NOT_PRESENT] from PGRAPH/DISPATCH on channel 0x005fb74000 [systemd-logind[610]]
04:25 sillysausage: nouveau E[ PFIFO][0000:04:00.0] PGRAPH engine fault on channel 2, recovering...
04:25 sillysausage: i saw this in the log file i wonder if that is an indication
04:26 sillysausage: i've never debugged graphics related issues before, so i'm not sure if this sounds like it
04:26 sillysausage: i'm using i3wm with compton
04:26 sillysausage: all i had opened was another session of urxvt and firefox
04:31 sillysausage: mm fuck, had it happened again
04:31 sillysausage: nouveau E[ PFIFO][0000:04:00.0] write fault at 0x00089c0000 [PAGE_NOT_PRESENT] from PGRAPH/DISPATCH on channel 0x005fb74000 [systemd-logind[627]]
04:31 sillysausage: nouveau E[ PFIFO][0000:04:00.0] PGRAPH engine fault on channel 2, recovering...
04:31 sillysausage: only way to fix it seems to be to reset my computer hmm, and no it's not a laptop
04:31 sillysausage: plain old workstation with a NVIDIA 580 GTX
05:13 RSpliet: sillysausage: upgrade your kernel
05:13 imirkin: sillysausage: anything in the logs before that?
05:14 sillysausage: nothing else no unfortunately and the kernel is 4.2.5
05:14 sillysausage: which is fairly recent isn't it
05:15 imirkin: yeah i'm not aware of anything in 4.3 that'd help you
05:16 imirkin: i'm guessing "systemd-logind" is actually "X" and it forgets to update its process name before exec'ing?
05:17 sillysausage: im just having a check
05:17 sillysausage: nope nothing really there besides that
05:18 sillysausage: but it happened twice and that error occurred at exactly that time
05:18 sillysausage: it's an archlinux machine btw
05:19 imirkin: unfortunately that error doesn't really say why it felt the need to write to that address
05:19 RSpliet: imirkin: are the addresses plausible in the first place?
05:19 sillysausage: yeah what i mainly came in here to ask is what to do in order to find out more
05:19 sillysausage: so next time it happens
05:19 sillysausage: i can get something meaningful
05:19 imirkin: skeggsb: how feasible would it be to dump the pushbufs when that happens?
05:19 sillysausage: and is there a document that talks about this
05:19 imirkin: sillysausage: as a user, nothing you can do... debugging nouveau kinda sucks =/
05:20 sillysausage: im guessing you need a special kernel?
05:20 imirkin: oh, debugging as a developer sucks too, don't worry :)
05:21 imirkin: RSpliet: sure, the addresses are plausible
05:26 simonpatapon: imirkin
05:26 RSpliet: imirkin: hmm okay, well, I find the perfect alignment a bit curious (use-after-free on pushbuf? trying to update a barrier after free'ing the pushbuf?) but I'll leave it up to the expert :-P
05:27 imirkin: RSpliet: the alignment is expected... it's probably a RT address or... something
05:27 RSpliet: RT?
05:27 imirkin: RSpliet: otoh i have no idea what /DISPATCH is
05:27 imirkin: RSpliet: render target
05:28 RSpliet: ah... dispatch sounds like the warp/wavefront scheduler
05:28 RSpliet: but does the slash imply a hierarchy or an "or"? :-P
05:28 imirkin: hierarchy
05:28 simonpatapon: imirkin: did you got my msg yesterday night?
05:29 imirkin: simonpatapon: you wrote many messages... sounds like you got reverse prime working in the end though?
05:31 simonpatapon: yes! solution was no xorg at all + xrandr --setprovideroutputsource 0 1
05:31 simonpatapon: i wanted to thank you again for your help
05:32 imirkin: np, glad it worked out
05:32 simonpatapon: xorg.xonf*
05:50 simonpatapon: see you guys
05:50 mlankhorst: tagr: do I need to do anything to enable usb 3.0 on the nouveau rootfs? seems in tegra you need to set some usb option in syslinux which I couldn't find back
06:41 karolherbst: I think my mice breaks /
06:52 karolherbst: imirkin: okay, running my dyn reclock stuff for a few days now and nouveau didn't crash because of that until now :)
06:54 karolherbst: I only get those LB_ERRORs and FB_FLUSH_TIMEOUTs :/
06:54 imirkin_: that's pretty good still
06:55 karolherbst: well I need to test something with pretty unstable load :/
07:02 karolherbst: ahh crap
07:03 karolherbst: imirkin_: sometimes this happens: https://gist.github.com/karolherbst/8c3fe923c89b6c78cc45
07:05 imirkin_: that sounds unfortunate.
07:05 imirkin_: but i can offer little beyond sympathy :)
07:16 tagr: mlankhorst: there is no support yet for USB 3.0
07:16 tagr: mlankhorst: unless you're using gnurou's tree, which in turn is based on mine and might actually support it
07:17 tagr: but then I'm not aware of any USB option that needs to be specifically set
07:17 imirkin_: i'm guessing he means the l4t thing
07:22 karolherbst: all this fun :/
07:55 mlankhorst: tagr: ok
07:56 mlankhorst: tagr: whats your tree? it would be worth a shot
08:00 tagr: github.com/thierryreding/linux branch staging/work
08:00 tagr: I rebase it all the time, so watch out for that
08:00 tagr: mlankhorst: I also break things sometimes =)
09:15 mlankhorst: tagr: I'm getting 10s latency now when enabling my usb webcam, cant be worse..
10:11 siro: imirkin: for https://bugs.freedesktop.org/show_bug.cgi?id=93004, the apitrace file is on sarnex ftp. is it ok for you to ask him for access ?
10:11 kubast2: Ok but I would had to get my whole setup on arch linux + I will have to take a shower eat dinner etc.
10:12 karolherbst: kubast2: no worries
10:12 karolherbst: we can do it tomorrow
10:12 siro: I could upload the file somewhere else as it's to big for the bugtracker
10:12 imirkin_: siro: how big is it? can you put it up on google drive? don't assume i'll be the only one looking at this
10:12 karolherbst: kubast2: there are still other issue I try to track down, but some of them just happen like only when you don't wait for them to appear
10:16 kubast2: I will be tomorrow then. Through can you hand me the github repos with nvidia debug tools and nouveau drm driver ,so I have everything that we will need tomorrow.
10:17 siro: imirkin: 86MByte
10:18 imirkin_: siro: ok, just stick it on gdrive or something
10:19 kubast2: ok got the tools
10:49 karolherbst: https://gist.github.com/karolherbst/22799d6be4db1732c661 :/
10:49 karolherbst: this one is annoying
10:49 karolherbst: I get a kworker looping there
10:50 kubast2: karolherbst ,which Branch should I clone ?
10:51 karolherbst: ohhh crap :/ now I know why my workaround doesn't work :(
10:51 karolherbst: crappy crap
10:51 karolherbst: kubast2: https://github.com/karolherbst/nouveau
10:51 karolherbst: branch: master_karol_no_touchy
10:51 karolherbst: of course
10:51 karolherbst: when the timer value for nvkm_msec doesn't change
10:51 kubast2: "/home/kubast2/nouveau/nouveau/drm/nouveau/include/nvif/os.h:34:28: fatal error: soc/tegra/fuse.h: No such file or directory"
10:52 karolherbst: nouveau loops there forever
10:52 karolherbst: kubast2: yeah, arch
10:52 karolherbst: don't know why that only happens there
10:52 karolherbst: but modify the os,h file
10:52 karolherbst: and remove those tegra files
10:54 karolherbst: skeggsb: we have a situation here. There are cases where nouveau turns off the card, allthough it calls timer->read later
10:55 karolherbst: I don't know why that happens though
10:59 karolherbst: how can I translate nouveau_pmops_runtime_suspend+0x0/0xd0 to the position inside nouveau_pmops_runtime_suspend? :/
11:01 kubast2: Ok I got the openssh server set up in case something goes wrong
11:02 karolherbst: kubast2: it will :D
11:02 kubast2: http://dpaste.com/353PESC
11:03 karolherbst: kubast2: do you have a 4.3 kernel?
11:03 kubast2: 4.2.5
11:03 karolherbst: mhhh
11:03 karolherbst: do you want to upgrade that?
11:03 kubast2: kernel that is ,sure why not
11:04 kubast2: gonna unlock testint in arch
11:08 karolherbst: kubast2: okay, after you compiled and isntalled nouveau (don't forget to update initramfs and stuff), just use nouveau as usual
11:12 mlankhorst: tagr: your tree doesn't select MAILBOX + XHCI mbox driver so I was getting a compiler error
11:19 kubast2: So I installed 4.3 ,but it seems like x loads all avalible gpu drivers
11:25 kubast2: Reverted to 4.2.5
11:26 kubast2: Default make installs nouveau driver?
11:28 kubast2: Ok it doesn't
11:29 kubast2: Oh lovelly arch didn't installed 4.2.5 back
11:34 kubast2: Well I think I'm gonna install fedora
11:34 kubast2: It will be fater this way
11:40 karolherbst: kubast2: mmhh?
11:40 kubast2: I know what happend ,but have no idea how to solve it other than just reinstall[linux(testing) owns files of linux(not-testing) and those pacman says "nope you can't downgrade"]
11:41 kubast2: X server doesn't start xD
11:41 kubast2: I installed linux-4.3
11:41 kubast2: +other -testing arch packages
11:41 kubast2: Now x doesn't start
11:43 kubast2: *nouveau-drm.c didn't compiled on 4.2.5 and i through you told me to install a newer one
11:44 karolherbst: yeah, but mhhh, it shouldn't affect X though
11:45 kubast2: X doesn't start because it loads nvidia ,nouveau ,nv ,vesa ,fbdev ,modsetting. But it never did threw an error about it
11:45 karolherbst: wut...
11:46 kubast2: Most of them aren't even installed
11:46 karolherbst: well you could remove nvidia
11:46 karolherbst: then it should work
11:46 kubast2: Well I need to install all nouveau stuff first
11:47 kubast2: Change mesa ,install nouveau ,then compile your nouveau.
11:50 kubast2: And to do that
11:50 kubast2: I need to uninstall shit load of dependencies
11:51 karolherbst: mhh?
11:51 karolherbst: what do you hae to change about mesa?
11:51 kubast2: https://wiki.archlinux.org/index.php/Xorg
11:51 kubast2: Nvidia have it's own mesa driver
11:51 kubast2: On arch
11:52 karolherbst: nvidia doesn't have any mesa driver
11:52 kubast2: Well in arch one gets to install mesa or mesa-nvidia
11:52 karolherbst: mesa-nvidia? serisouly?
11:52 pmoreau: mesa-nvidia ?!
11:52 kubast2: Yeah
11:52 pmoreau: I don't have that one
11:52 karolherbst: do you mean lib32-nvidia-libgl?
11:52 kubast2: And mesa-nvidia-340xx for 340xx
11:53 kubast2: it's multilib
11:53 pmoreau: nvidia-libgl I agree, but not mesa-nvidia
11:53 karolherbst: then nvidia
11:53 karolherbst: there is no such thing as mesa-nvidia
11:53 pmoreau: pacman -Ss mesa-nvidia doesn't return anything
11:53 pmoreau: Or maybe you have some strange repo :-)
11:53 karolherbst: :D
11:54 karolherbst: I hope not
11:54 karolherbst: they mess up stuff
11:54 kubast2: Yeah
11:54 kubast2: libgl
11:54 kubast2: Not mesa
11:54 karolherbst: kubast2: then jsut remove the nvidia and the multilib nvidia package
11:54 pmoreau: You don't need to uninstall everything
11:54 karolherbst: that's all you ahve to do
11:54 karolherbst: or you just change all that libgl entry stuff
11:55 pmoreau: kubast2: Just run `pacman -S mesa-libgl` and confirm that you want to remove nvidia-libgl, and you should be good to go
12:01 kubast2: ok got it now I'm making karols nouveau driver
12:01 kubast2: the xserver runs fine
12:02 kubast2: kde5 looks better on nouveau[fonts]
12:03 karolherbst: kubast2: then go into /sys/kernel/debug/dri/0
12:03 karolherbst: as root
12:03 karolherbst: after you installed nouveau
12:03 karolherbst: from my repository
12:03 kubast2: gotta reboot
12:03 karolherbst: nobody has any clue what is going on here? https://gist.github.com/karolherbst/22799d6be4db1732c661
12:04 karolherbst: this bothers me a bit, because nouveau decides to turn off the gpu, while the gpu or the driver got into some crazy state, while a game is still running
12:05 kubast2: I'm there
12:06 kubast2: "DVI-D-1 HDMI-A-1 VGA-1 bufs clients gem_names name vbios.rom vm vma"
12:06 karolherbst: well
12:06 karolherbst: that doesn't look like my branch
12:06 kubast2: I updated mkinitcpio
12:06 karolherbst: did you also copy the nouveau.ko file?
12:07 karolherbst: best way is to just overwrite the already installed one inside/lib/modules/
12:07 karolherbst: and rename the old one
12:08 hakzsam: imirkin, no regressions with xonotic-glx and with the doom3 trace you sent me yesterday
12:08 imirkin_: hakzsam: perfect
12:09 hakzsam: I'm going to push and to rebase my edgeflag changes on top of it
12:09 imirkin_: sgtm
12:10 imirkin_: man these talos guys sure like to write shaders. 23M of glsl source and counting.
12:12 kubast2: Tbh I have no idea where it is according to make[1]: it's in /usr/lib/modules/4.3.0-1-ARCH/build ,but it's not really there ,and /usr/lib/modules/4.3.0-1-ARCH/build/drivers/gpu/nouveau only contains Kconfig file
12:13 kubast2: wait
12:14 kubast2: /usr/lib/modules/4.3.0-1-ARCH/kernel/drivers/gpu/drm/nouveau/
12:15 kubast2: definietlly not here the md5sums are the same as the module in /lib64/modules/4.3.0-1-ARCH/kernel/drivers/gpu/drm/nouveau/
12:15 kubast2: and the same in /lib/(...)/nouveau
12:16 kubast2: oh
12:16 kubast2: "Module.symvers and .tmp_versions" the only files that clean deleted
12:18 kubast2: nvm
12:20 prg_: karolherbst, wrt fear, so nothing to worry about as long as gameplay works fine, right?
12:21 prg_: it's just being weird during benchmarking
12:31 kubast2: DVI-D-1 HDMI-A-1 VGA-1 bufs clients cstate current_load gem_names name pstate vbios.rom vm vma
12:32 kubast2: forget the modules is made in nouveau/
12:50 karolherbst: prg_: well it crashed for me with gallium nine :/
12:50 karolherbst: but that was inside d3dadapter.so
12:50 karolherbst: kubast2: looks better now
12:51 karolherbst: you could keep an eye on the last line of pstate and the content of current_load
12:51 karolherbst: it should run stable for games not reaching vsync mark
12:51 karolherbst: or not vsynced stuff
12:51 karolherbst: but if the load fluctuate a lot it may crash the gpu
12:52 karolherbst: or nouveau
12:52 karolherbst: but because you don't have many states anyway, it shall not happen
12:54 kubast2: Well I have checked quake3 webgl demo
12:56 karolherbst: and?
12:57 kubast2: I will need to run through ssh the cat commands
12:57 kubast2: the gpu load is smaller when the game window isn't in focus
12:57 karolherbst: ohh k
12:59 kubast2: so the ssh works only for localhost
13:02 kubast2: made an quick bash script
13:04 kubast2: everything seems about right
13:04 kubast2: core 254 pstate 0f
13:04 kubast2: 1058 mhz 500 mhz
13:05 karolherbst: kubast2: yeah, figures
13:05 karolherbst: kubast2: the thing I am more interested in is, how does it change on loads capped by vsync
13:06 karolherbst: normal desktop usage or some lightwight games, stuff like that
13:06 karolherbst: you could cat pstate and current_load every second or something and log that into a file or something
13:06 kubast2: I have every 5 secounds
13:06 karolherbst: mhhh
13:06 karolherbst: that's too long
13:06 kubast2: but the script can be easilly modified
13:07 kubast2: sleep 1
13:07 karolherbst: the gpu will downclock after 3 seconds when there is no load
13:07 kubast2: tbh
13:07 kubast2: it doesn't downlock
13:08 kubast2: it's at 0f
13:08 kubast2: on idle
13:08 karolherbst: mhhh
13:08 karolherbst: then give me current_load
13:08 kubast2: core:0 mem:76 video:0 pcie:0
13:08 karolherbst: okay, it should downclock then
13:08 karolherbst: still at 0f?
13:08 kubast2: yeah
13:09 karolherbst: did you put anything into pstate?
13:09 kubast2: no
13:09 karolherbst: mhh okay
13:09 karolherbst: reboot with nouveau.debug=clk=trace
13:10 karolherbst: I bet the 0a pstate is just too slow and my algorithm rather uses 0f
13:12 kubast2: ok got it
13:16 kubast2: ok so now there's none of the pstates choosed
13:17 kubast2: ok so the change works now
13:17 kubast2: the screen turns half black and the pstate is changed
13:18 kubast2: 07
13:18 kubast2: 0a
13:18 kubast2: with web browser
13:18 kubast2: 07 at idle ,0a when scrolling on image heavy website
13:20 kubast2: 87-152 core[firefox] 0a-07
13:21 kubast2: 180+[cs:go] 0f
13:21 kubast2: 0[idle= of
13:21 kubast2: steam closed of core:0 mem:69 video:0 pcie:0
13:21 kubast2: but still 0f
13:23 kubast2: perhaps steam changes pstate manually
13:23 kubast2: nah it would need root
13:24 kubast2: ok launched webgl app
13:24 kubast2: and then turned it off
13:24 kubast2: which seems to fix the pstates
13:25 kubast2: pstates work normally now
13:28 karolherbst: kubast2: I need your dmesg now
13:28 kubast2: hmm
13:28 kubast2: 0f again
13:28 kubast2: ok
13:29 kubast2: http://dpaste.com/0P83FP5
13:30 karolherbst: kubast2: see those nouveau 0000:01:00.0: clk: thingies?
13:30 kubast2: yeah
13:30 kubast2: *actually no
13:30 karolherbst: ohhh wait
13:31 karolherbst: your voltage stuff fails
13:31 kubast2: 0000:01:00.0: clk:
13:31 kubast2: yeah now I see them
13:32 kubast2: I'm guesing it's not good
13:32 karolherbst: then I need your pstate file
13:34 karolherbst: kubast2: ohhhhhhhh
13:34 karolherbst: I have a fix for your card already
13:34 karolherbst: :O
13:34 kubast2: really
13:34 karolherbst: yeah
13:34 karolherbst: there was another with the same issue
13:34 karolherbst: that's what you need: https://github.com/karolherbst/nouveau/commit/a7c625c6be45bc790d47e58c3d403a3725fa6f43
13:35 karolherbst: kubast2: okay, go to the nouveau repository
13:35 karolherbst: and update the branch
13:35 karolherbst: I pushed the commit on the branch
13:35 kubast2: git pull
13:35 kubast2: yup
13:36 kubast2: hmm
13:36 kubast2: bunch of build errors
13:37 kubast2: http://dpaste.com/2ZAYJJY
13:37 karolherbst: build inside drm
13:38 kubast2: right
13:42 kubast2: ok moved module and updated mkinitcpio
13:46 < AndChat|341361> Something crashed
13:47 < AndChat|341361> The screen froozed ,when I launched quake3 webgl demo
13:48 < AndChat|341361> No reaction from ctrl+alt+f1
13:48 < AndChat|341361> Only hard reset
13:49 < AndChat|341361> After testing out nouveau with the voltage fix
13:53 kubast2: The gpu and pc boots back to system just fine. Through I have school tomorrow and it's 23:00 ,so I kinda gotta go.
13:54 kubast2: I will try to give you dmesg tomorrow
15:05 imirkin_: skeggsb: bleh, we still get the vram config wrong?
15:05 skeggsb: imirkin_: ... where?
15:05 imirkin_: i thought that's what your patchw as about
15:05 mark4_: do nouveau drivers work with hybrid graphics ?
15:05 imirkin_: i guess it's about the number of ppc or whatever
15:05 skeggsb: no, for some reason GPC5 only has 1 PPC, all the others have 2
15:05 imirkin_: mark4_: http://nouveau.freedesktop.org/wiki/Optimus/ probably has the info you want
15:06 imirkin_: skeggsb: weird.
15:06 skeggsb: we assume all GPCs have the same number of those
15:06 skeggsb: as does nvgpu
15:06 skeggsb: resman, apparently, knows better
15:06 imirkin_: hehe
15:06 imirkin_: skeggsb: any progress on parallel piglit stability? were you able to repro my fail at least?
15:07 skeggsb: i haven't tried to repro yours yet
15:08 mark4_: imirkin_, i read that and it has not given me a definitive answer
15:09 imirkin_: mark4_: ok, then ask a definitive question :)
15:09 mark4_: does nouveau work with hybrid graphics systems where there is NO mux
15:09 imirkin_: define 'work'
15:09 mark4_: can i get to an X desktop
15:09 imirkin_: sure
15:10 imirkin_: although there's no nouveau involvement required for that
15:10 imirkin_: in a hybrid graphics set up your display will be attached to the IGP in all likelihood
15:10 mark4_: yes there is if i want to use the nvidia card for graphics
15:10 imirkin_: so you clearly have something very specific in mind for 'working'
15:11 mark4_: actually the only reason to switch to nouveau from nvidia-drivers is because im not currently playing any games and do not need real 3d support
15:11 imirkin_: in general it should be possible both to offload rendering via "prime" and to address remote crtc's via "reverse prime"
15:12 mark4_: in that case i might as well just keep the nvidia card powered down and just use the integrated for everything
15:12 imirkin_: of course various things could go wrong that will make nouveau not work for you
15:12 imirkin_: yes, that's probably the best -- if you can disable it in your bios
15:12 imirkin_: then it'll just be powered down and all connectors should be available via your IGP
15:13 mark4_: it took 3 weeks to get X up and running with the nvidia card when i got this machine, it worked fine with the intel but that does not have real 3d support, probably even nouveau's accellerated 3d is better
15:13 mark4_: but thats a non issue right now
15:14 mark4_: but ty. i can try nouveau and or just the integrated :)