01:03 mishehu: greets folks, running kernel 4.6.5 on slackware64 14.2. when the kernel first starts to load, all is fine, but it looks like about the time that nvidiafb is loaded, I lose the console and get no signal on the hdmi even though fbcon is loaded.
01:03 mishehu: I have an atom-ion, MCP79.
06:08 pmoreau: mishehu: If you want to use Nouveau, you should not be using nvidiafb.
06:08 mishehu: pmoreau: oh, they conflict?
06:09 pmoreau: They do
06:09 mishehu: pmoreau: is there something instead of nvidiafb that I should use from the framebuffer drivers? or none at all?
06:10 pmoreau: IIRC, nvidiafb is an old and unmaintained driver
06:11 mishehu: I though that was the rivafb
06:11 mishehu: or was taht just that rivafb was for older nvidia cards?
06:11 pmoreau: Nouveau should automatically use nouveaufb. I don’t think you need to enable it separately
06:11 pmoreau: Most likely
06:18 pmoreau: Most of the changes for nvidiafb were made between 2005 and 2007, with the latest in 2010 and 2012.
06:41 mishehu: pmoreau: rebuilding my kernel, removed nvidiafb altogether. I do see that nouveau does set fbcon automatically, and a few others. I explicitly set them in case I had accidentally overidden them.
06:41 pmoreau: You can also blacklist nvidiafb/rivafb as an easy "fix"
06:58 mishehu: yeah I had tried that and it didn't work
06:59 mishehu: and even now still not getting anything. as soon as it loads nouveau, poof, no display.
06:59 mishehu: I'm not setting any non-default values to nouveau.
07:12 pmoreau: mishehu: Could you blindly get the output of dmesg and give a link to it?
07:16 mishehu: yeah I can pb it
07:16 mishehu: hang on a sec. you want my kernel's .config as well?
07:17 pmoreau: I don’t think it is needed
07:17 mishehu: ok hold a sec please
07:21 mishehu: pmoreau: http://pastebin.ca/3670141
07:23 pmoreau: Thanks
07:23 mishehu: no, thank you :-)
07:24 pmoreau: Nouveau seems happy
07:25 pmoreau: Could you link the Xorg.0.log, and try booting with systemd.unit=multi-user.target?
07:25 mishehu: no systemd on slackware (thank gawd)
07:25 mishehu: and I've not even gotten as far as trying to launch X.
07:25 pmoreau: Ah… hum…
07:26 mishehu: it's at runlevel 3 in /etc/inittab. which is the default for slackware, and is multiuser.
07:26 mishehu: 4 would be to launch direect to X
07:27 mishehu: that is indeed damn weird
07:27 mishehu: I log in blindly
07:28 mishehu: and run startx, I do get signal on the tv again
07:28 pmoreau: Nouveau does not seem to create any fb
07:28 mishehu: even though lsmod shows it connected to fbdev
07:29 mishehu: err fb I mean -> fb 52470 5 fbcon,nouveau,softcursor,tileblit,bitblit
07:31 pmoreau: You should have something like "nouveau 0000:03:00.0: DRM: allocated 1920x1080 fb: 0x60000, bo ffff8802b9ff2c00" and "nouveau 0000:03:00.0: fb0: nouveaufb frame buffer device", if the NVIDIA card is driving a screen.
07:34 pmoreau: You could add a `nouveau.debug=debug` and `drm.debug=0xf` to the kernel command line to get more debug info
07:34 mishehu: well this is an htpc, so I think I'll just set runlevel 4.
07:51 pmoreau: gtg
07:51 mishehu: might not make much of an htpc.. vdpauinfo is reporting to me that it has no support for anything under decoder capabilities... not mpeg2's, not h264's, and not vc1's.
07:51 pmoreau: Do you have the nouveau-fw?
07:51 mishehu: pmoreau: let me check
07:52 pmoreau: https://nouveau.freedesktop.org/wiki/VideoAcceleration/
07:52 mishehu: yeah slack doesn't seem to have that package.
07:53 mishehu: I see it as an add-on
07:53 mishehu: thanks for pointing me to the correct location
07:53 pmoreau: You’re welcome
07:54 pmoreau: gtg, I’ll be back in a few hours
07:55 mishehu: looks like it grabs the firmware from a pretty old version of the closed source driver heh.
08:01 karolherbst: doesn't matter though
08:14 mishehu: karolherbst: I was figuring not.
08:14 mishehu: hopefully all's fine now...
08:15 drathir: Lekensteyn: thanks a lot didnt know about not workin autosuspending when blacklisted...
08:30 Lekensteyn: drathir: well, did you just notice that or have you just read it :P
08:33 karolherbst: got a " gr: GPC0/TPC0/MP trap: global 00000004 [MULTIPLE_WARP_ERRORS] warp 3e0009 [ILLEGAL_INSTR_ENCODING]" today
08:33 karolherbst: 5 in total
08:34 karolherbst: no idea what though :/
09:37 karolherbst: skeggsb: thanks for merging
09:39 karolherbst: the iccsense patch is (not that anybody thinks here the reclocking stuff is merged yet)
09:49 karolherbst: updateing my stuff to 4.7 today
10:21 dcomp: https://home.dcomp.co.uk/piglit/latest/changes.html
11:16 pmoreau: karolherbst: Yeah, I would have thought you were talking about the reclocking stuff. :-D
11:16 karolherbst: :)
11:17 karolherbst: 75% of the patches are ready to merge though
11:17 karolherbst: but I don'T think it makes sense to merge them partially
11:18 karolherbst: mhh, allthough the update on temperature change could be split out
11:18 karolherbst: yeah, actually makes kind of sense
11:19 karolherbst: not completly though, because we risk overheating the gpu more this way
11:19 karolherbst: well "risk", as in we could drive the gpu into the hw overheating protection
11:29 pmoreau: :-)
14:02 karolherbst:
14:09 karolherbst: ahh crap, I messed up that runpm thing on my branch again...
14:09 karolherbst: I guess I will make it awesome now and support reclocking whily gpu is off :)
14:14 SaveTheRobots: karolherbst: hey, i noticed you moved branches around in your nouveau repo, is stable_reclocking_kepler_v5 still the best branch to use with my 780 Ti? or maybe master_4.7, or dynamic_reclocking? (is this working? :O)
15:14 karolherbst: SaveTheRobots: I keep the branch names as they are
15:15 karolherbst: master_work usually contains more stuff, which I update rarely, but should be stable enough for general use, otherwise stick to stable_reclocking_kepler_v5 for reclocking only things
15:15 karolherbst: I've also updated to 4.7
15:16 orbea: so that branch should work for both 4.6 and 4.7 kernels?
15:16 karolherbst: nope, only for 4.7
15:16 orbea: ah
15:16 karolherbst: you can always rebase on master_4.6 but then you miss updates from the master branch
15:16 orbea: i'll have to update my kernel and try it out
15:16 karolherbst: usually I only update master_$newest_kernel_version
15:22 SaveTheRobots: karolherbst: awesome, thanks dude... also, how is dynamic reclocking coming along? is it in a useable state?
15:22 karolherbst: well, it is useable, but it ain't work that well for desktops
15:23 karolherbst: also there are some stuff we have to take care before
15:23 karolherbst: it was just prototyping
15:23 SaveTheRobots: ah ok, cool, i'll steer clear of it for now then
15:24 SaveTheRobots: another random question... what pstate would you recommend for general desktop use (GNOME/KDE) ?
15:24 SaveTheRobots: i'm on 0a with GNOME 3.20 (arch) atm and some of the workspace transitions can be a bit janky
15:26 orbea: SaveTheRobots: check your sensors and see what looks good
15:27 orbea: i usually stick to the lowest pstate and clock up as needed when gaming
15:28 SaveTheRobots: orbea: yeah, i just clocked down to 07 and it's no more janky than before, i guess it's GNOME's fault
15:29 SaveTheRobots: guess i'll stick to lowest, as well
15:29 orbea: could very well be, I use a minimal wm and have none of that (spectrwm)
15:30 SaveTheRobots: same, i normally use bspwm on Void but i fancied giving GNOME a try, it was unusable with the nvidia drivers because of lag/tearing/etc
15:55 karolherbst: just got this after loading nouveau: https://gist.github.com/karolherbst/16644414dbcb1900aef9d59f3e831edc
15:56 karolherbst: and why the hell did my computer go into suspend after I removed nouveau
15:57 karolherbst: skeggsb: any idea about the paste?
17:47 NanoSector: hey, I'm not sure if this is because reclocking is still unstable, but whenever I try reclocking my GT750M (DDR3 edition) the command (echo 0f or 0a > /sys/kernel/debug/dri/1/pstate) hangs, yet the card seems to have changed power states if i 'cat' the same location
17:48 NanoSector: however when I llaunch a program using DRI_PRIME=1 (because Optimus), whenever OpenGL is in use, it shows a blank screen and if I launch it a second time, my whole system hangs
17:49 NanoSector: however the main window which doesn't use OpenGL works
17:54 karolherbst: NanoSector: you should try out my branch
17:55 NanoSector: karolherbst, your branch?
17:55 karolherbst: https://github.com/karolherbst/nouveau/tree/stable_reclocking_kepler_v5 branch stable_reclocking_kepler_v5
17:55 karolherbst: compiles against 4.7 kernel
17:55 NanoSector: oh, cheers
17:55 karolherbst: cd drm; make
17:55 NanoSector: I'm still on 4.6 though, I use Arch
17:55 karolherbst: it is in review phase so it might be merged in 4.9
17:56 karolherbst: NanoSector: mhhh, today I rebased it on 4.7 :/
17:56 NanoSector: :(
17:56 NanoSector: can I download an earlier commit?
17:56 karolherbst: wait a second
17:56 karolherbst: I still have 4.6 sources installed
17:56 NanoSector: (and is cd drm; make all I need to do?)
17:58 karolherbst: nope, you have to install it too. but wait until I have that 4.6 branch updated
17:58 NanoSector: aight :)
18:01 NanoSector: also how does Optimus in Nouveau really work? like Bumblebee (e.g. render offscreen -> copy to memory -> display on Intel) or more like rendering directly into Intel's framebuffer, if I name that correctly?
18:02 karolherbst: branch: stable_reclocking_kepler_v5_4.6
18:03 karolherbst: NanoSector: no clue how the internals work, but it has lower overhead than bumblebee
18:03 NanoSector: yeah main reason I'm trying to use this is to see if I can get this to work better than Bumblebee :)
18:03 NanoSector: When maximizing a window with BB it starts lagging real bad
18:03 karolherbst: which resolution?
18:04 NanoSector: 1920x1080
18:04 karolherbst: mhh
18:04 karolherbst: should be fine with bumblebee actually
18:04 karolherbst: except if you only have pcie 1.0
18:04 NanoSector: well, less horizontal because dock and top bar (GNOME3)
18:04 NanoSector: I think my CPU is not powerful enough or something
18:04 NanoSector: i5 4200U + 8GB DDR3 + GT750M here
18:04 NanoSector: anyway, building now :) thanks!
18:05 karolherbst: you have to install the module after that
18:05 NanoSector: make install?
18:05 karolherbst: mhh
18:05 karolherbst: causes problems usually
18:05 NanoSector: like?
18:05 karolherbst: you should have a nouveau.ko file inside /lib/modules/$kernel_version somewhere
18:05 karolherbst: NanoSector: like the original nouveau is loaded instead
18:05 NanoSector: oh someone just told me there's linux-mainline in the AUR for Arch
18:06 karolherbst: NanoSector: in /lib/modules/4.6.3-$tag/kernel/drivers/gpu/drm/nouveau
18:06 karolherbst: NanoSector: well, won't help much
18:06 karolherbst: if that stuff would be mainlined, I would have told you to upgrade your kernel
18:07 NanoSector: yeah :)
18:07 NanoSector: does your branch replace the original nouveau module?
18:08 karolherbst: you should remove the original one, and place the new compiled one there
18:08 karolherbst: and then regenerate initramfs
18:08 NanoSector: ahh i see
18:08 NanoSector: sorry I'm not good with all the compiling and installing business :)
18:09 NanoSector: so i just copy the nouveau.ko from the drm/nouveau directory to /lib/modules/kernelthing/kernel/drivers/gpu/drm/nouveau?
18:10 Tom^: indeed, and remove the old one which is called nouveau.ko.gz , run depmod -A and reboot.
18:10 NanoSector: cheers :)
18:10 Tom^: this will be sort of reversed on your next kernel update tho :p
18:10 NanoSector: i'll take a backup of it just in case
18:11 NanoSector: rebooting :)
18:14 NanoSector: karolherbst, it reclocked without a hitch
18:14 karolherbst: :)
18:14 NanoSector: that is, if the OpenGL app didn't freeze :P
18:15 NanoSector: am I supposed to reclock without an Opengl app running?
18:15 karolherbst: doesn't matter
18:15 karolherbst: some say, when the gpu is under load, it may crash more likely
18:15 karolherbst: but I think that's not true :p
18:15 karolherbst: :D
18:15 karolherbst: but I also fixed a bunch of issues in the reclocking process for kerpler gpus and newer
18:15 karolherbst: it is mainly policy stuff though
18:16 NanoSector: then my whole DE froze
18:17 RSpliet: karolherbst: I think we pause the graphics engine during reclock, but the behaviour of the context switching engine during memory clock changing is unknown to me
18:19 karolherbst: RSpliet: no clue either, it didn't crash for me for a while now, so I really have no idea
18:20 karolherbst: NanoSector: what do you mean with "then"?
18:20 NanoSector: karolherbst, when I tried quitting the app
18:20 NanoSector: though I think I know what the issue is
18:20 NanoSector: not related to nouveau
18:22 karolherbst: NanoSector: I see
18:22 NanoSector: ok when I issued "echo 0f > /sys/kernel/debug/dri/1/pstate", for about 3 seconds it was butter smooth (compared to choppy before reclock) and then it got completely stuck
18:23 NanoSector: like, just the OpenGL app
18:23 NanoSector: not my computer
18:23 karolherbst: mhhh
18:23 NanoSector: should I try 0a instead?
18:23 karolherbst: sure you use my branch?
18:23 NanoSector: yeah
18:24 karolherbst: is there a boost file besides pstate?
18:24 NanoSector: yes
18:24 karolherbst: k
18:24 NanoSector: good or bad?
18:24 karolherbst: mhh, it shouldn't really matter which perf level you use, it shouldn't hang
18:24 karolherbst: but I know, that for a few cards it isn't enough and something is still missing :/
18:24 karolherbst: no clue what
18:24 NanoSector: [ 282.686104] nouveau 0000:01:00.0: gr: TRAP ch 4 [00ff7d3000 Emuthread - Sta[1446]]
18:25 NanoSector: [ 282.686117] nouveau 0000:01:00.0: gr: ROP0 80000000 80000001
18:25 NanoSector: [ 282.686122] nouveau 0000:01:00.0: gr: ROP1 80000000 80000001
18:25 karolherbst: could mean anything if memory crashes
18:25 NanoSector: it throws a LOT of that in dmesg
18:25 karolherbst: yeah, normal
18:25 karolherbst: just means, that the gpu is messed up
18:25 NanoSector: :(
18:25 NanoSector: oh the app quit
18:25 karolherbst: you could leave us your vbios and if you got time you can even mmiotrace nvidia
18:25 karolherbst: NanoSector: yeah, after some time it might recover
18:25 karolherbst: maybe it is indeed the application
18:26 NanoSector: just tell me what to do :)
18:26 karolherbst: well, you could test other applications first
18:27 NanoSector: OK from my phone, tried launching the same thing again and it froze like before
18:27 NanoSector: Like whole system freezes
18:27 karolherbst: yeah, that's dri2 fault :D
18:27 karolherbst: with dri3 only the application should freeze and not X
18:28 karolherbst: that's purely X fault to hang your entire session
18:28 NanoSector: Oh did i not enable DRI3
18:28 karolherbst: well, you would have to do even more than that
18:28 karolherbst: you have to prevent X from loading the nouveau ddx, too
18:28 NanoSector: No but i thought i enabled that
18:29 karolherbst: anyhow, you might want to try out different applications, but if the gpu recovered, it usually tells us, that the mesa did something wrong with the gpu
18:29 karolherbst: you could try something else, like vblank_mode=1 glxspheres64
18:29 Tom^: NanoSector: which or what opengl app is this?
18:29 NanoSector: Tom^, I was trying a game in the Dolphin emulator
18:30 NanoSector: but I can try Transistor too, a Steam game
18:30 NanoSector: or glxspheres64
18:31 NanoSector: is it possible that the 0f power state is causing issues though? I've seen Phoronix having issues with the highest power state only
18:32 NanoSector: and it hangs when I try to reclock without having an opengl app open
18:32 NanoSector: the command, that is :x
18:32 karolherbst: NanoSector: yeah, because the gpu suspended, but that shouldn't happen with my branch though
18:33 NanoSector: Hmm how can I double check I'm using your branch?
18:33 karolherbst: the boost file is there.
18:34 karolherbst: well anyway, after the gpu hanged once, the driver is in a messy state most likely anyway
18:34 karolherbst: even nvidia can't deal with that sanely
18:34 NanoSector: :(
18:34 Tom^: yea try something besides this dolphin emu game, after a clean reboot or similiar.
18:35 Tom^: hopefully it is just that game in the emu that locks things, narrows things down a bit.
18:35 NanoSector: so "DRI_PRIME=1 vblank_mode=0 glxspheres64"?
18:35 karolherbst: yeah
18:35 NanoSector: and I don't need to do the whole sink xrandr deal with DRI2?
18:35 karolherbst: you mean dri3
18:35 karolherbst: right
18:36 NanoSector: I'm using DRI2 atm now I think
18:36 karolherbst: X isn't involved with dri3 anymore
18:36 karolherbst: just plain kernel calls
18:36 NanoSector: glxspheres64 also hangs when setting pstate
18:36 NanoSector: X doesn't
18:36 karolherbst: mhhh, I see
18:36 karolherbst: then you are a bit unlucky then
18:37 karolherbst: NanoSector: there is a vbios.rom file besides pstate
18:37 karolherbst: could you cat the file somewhere and upload it?
18:37 NanoSector: sure, hang on
18:38 NanoSector: any preferred file sharing services?
18:38 karolherbst: nope
18:39 NanoSector: https://drive.google.com/open?id=0B2F3P3d66S-1Y3Q1MmdpRnJMYkE
18:39 karolherbst: it is a GK107, right?
18:39 NanoSector: GT750M, I think that's a GK107
18:39 NanoSector: it ended in -107 for sure
18:39 karolherbst: k
18:39 NanoSector: oh
18:39 NanoSector: glxspheres64 still outputs things
18:39 NanoSector: 133.012497 frames/sec - 148.441947 Mpixels/sec
18:39 NanoSector: 0.624729 frames/sec - 0.697198 Mpixels/sec
18:40 NanoSector: and my fans just calmed down
18:41 karolherbst: mhh, k
18:41 karolherbst: you could try to poke 2 into boost first
18:41 NanoSector: on a clean boot?
18:42 karolherbst: yeah
18:42 karolherbst: mhhh
18:42 karolherbst: and could you boot with nouveau.debug=debug =
18:42 karolherbst: ?
18:43 NanoSector: Sure
18:43 NanoSector: Though now everything hangs
18:43 NanoSector: Hang on
18:44 karolherbst: then could you once reclock to 0f while the gpu is on and give me the output of dmesg?
18:45 karolherbst: nothing has to run, you could do a glxinfo and poke a second later or something
18:46 NanoSector: http://hastebin.com/sogiboresu.mel
18:46 NanoSector: pstate and boost
18:46 NanoSector: reclocking to 0f now and dmesg'ing with the debug thing
18:46 NanoSector: do i poke 2 to boost too?
18:47 NanoSector: oh that's a lot of output
18:47 NanoSector: suppose you just want t he output of it reclocking?
18:48 NanoSector: [ 140.461816] nouveau 0000:01:00.0: clk: setting performance state f
18:48 NanoSector: [ 140.461822] nouveau 0000:01:00.0: pci: 8.0GT/s not supported by bus or card, droppingrequested speed to 2.5GT/s
18:48 NanoSector: [ 140.461824] nouveau 0000:01:00.0: pci: set link to 2.5GT/s x255
18:48 NanoSector: that's all it says when reclocking
18:49 NanoSector: and when running DRI_PRIME=1 glxinfo, it hangs at "[ 243.829531] nouveau 0000:01:00.0: devinit: running init tables"
18:49 NanoSector: glxinfo hangs, that is
18:49 NanoSector: and the echo 0f > pstate command is also hanging
18:51 NanoSector: all nouveau dmesg output up to this point: http://hastebin.com/evofocuwow.sm
18:53 NanoSector: karolherbst, does this answer your questions?
18:55 karolherbst: you reclocked while the card suspended :/
19:05 NanoSector: while running a game and reclocking to 0f with DRI3: http://hastebin.com/sugecudiqi.sm
19:06 karolherbst: NanoSector: mhhh
19:06 karolherbst: no clue, mind trying out my xorg config?
19:07 karolherbst: https://gist.github.com/karolherbst/1f1bdd1a3822df74097f
19:07 karolherbst: and make sure you install the dummy driver
19:08 NanoSector: Will do, hang on while i reboot once more
19:11 NanoSector: ok running on your xorg config now
19:13 NanoSector: karolherbst, that works for 5 seconds before it hangs again :(
19:13 NanoSector: it didn't hang immediately after reclocking
19:13 karolherbst: :/
19:13 karolherbst: did you try boost set to 2?
19:14 NanoSector: first set boost to 2 and then 0f to pstate?
19:14 NanoSector: oh
19:14 NanoSector: it still 'works', just at 0,00001 FPS or something
19:15 NanoSector: because it just pushed a frame to the game
19:16 karolherbst: mhh
19:16 karolherbst: no idea what is wrong, could be anything
19:16 karolherbst: but it should be the memory going weird
19:16 karolherbst: skeggsb knows about some bits which are missing there
19:17 NanoSector: shall I try first setting boost to 2 and then reclocking?
19:17 NanoSector: just to try all combinations
19:19 karolherbst: yeah
19:20 NanoSector: Nope, game hangs
19:22 NanoSector: hmm
19:22 NanoSector: vgaswitcheroo says the card is now turned off
19:22 NanoSector: even though a game is running
19:22 NanoSector: or, hanging
19:22 NanoSector: v
19:22 NanoSector: 1:DIS: :DynOff:0000:01:00.0
19:28 NanoSector: Just tried, the original nouveau module causes all sorts of issues when reclocking not present with your module :) so there's improvement
19:29 NanoSector: Including my wifi adapter crashing
19:30 karolherbst: I see
19:30 NanoSector: karolherbst, would you want me to try your module on the mainline 4.7 kernel?
19:30 karolherbst: makes no difference
19:30 NanoSector: Oh
19:30 karolherbst: did you tried my config with the dummy thing?
19:30 karolherbst: ahh
19:30 NanoSector: yeah, I'm using it right now
19:30 karolherbst: you did
19:30 karolherbst: mhh
19:30 karolherbst: I figure your desktop is more solid with that, when nouveau hangs?
19:31 NanoSector: how do you mean?
19:31 NanoSector: my desktop still responds just fine
19:31 NanoSector: just the app running on nouveau doesn't
19:32 karolherbst: k
19:32 karolherbst: well I think it is something we totally miss now
19:32 karolherbst: would need deeper investigation
19:32 NanoSector: if there's anything you'd like me to try, just ask
19:33 karolherbst: mhh, you could get bumblebee working again and run a tool while nvidia is loaded
19:33 NanoSector: sure, what tool?
19:33 karolherbst: I doubt it helps though, because I am quite sure I got it right
19:33 NanoSector: better be safe than sorry
19:33 karolherbst: a tool I wrote to compare voltage set by nvidia with what nouveau would do
19:34 NanoSector: sure, i'll undo the Xorg config and set up bumblebee
19:36 NanoSector: karolherbst, do you have a link to said tool?
19:36 NanoSector: dkms is taking its time to generate the modules so might as well download it
19:37 karolherbst: awesome_tool branch :p
19:37 karolherbst: run make in top dir
19:37 karolherbst: bin/nv_cmp_volt as root
19:39 NanoSector: it's compiling
19:39 NanoSector: fatal error: drm/drm.h: file or folder doesn't exist
19:40 NanoSector: In file included from lib/drm.c:37:0:
19:40 NanoSector: ./drm/nouveau/uapi/drm/nouveau_drm.h:30:21: fatale fout: drm/drm.h: Bestand of map bestaat niet
19:40 NanoSector: #include <drm/drm.h>
19:40 NanoSector: oh, that's longer than i expected, sorry
19:43 karolherbst: uhh, you can remove that include I think
19:44 NanoSector: That worked
19:46 NanoSector: ./nv_cmp_volt: error while loading shared libraries: libnvif.so: cannot open shared object file: No such file or directory
19:47 karolherbst: uhh right
19:47 karolherbst: LD_LIBRARY_DIR=lib
19:47 karolherbst: or was it PATH?
19:47 karolherbst: ...
19:48 karolherbst: I think DIR, not sure
19:48 karolherbst: nope, path
19:49 NanoSector: so I ran: sudo env LD_LIBRARY_PATH=lib ./bin/nv_cmp_volt
19:49 NanoSector: failed to create device, -12
19:49 karolherbst: uhhh
19:49 karolherbst: I guess that ioremap protection strikes again
19:52 NanoSector: how about we drop this for tonight and see if we get any further tomorrow or so? :)
19:52 karolherbst: we could, yeah
19:52 NanoSector: with a fresh mind, because i understand if I'm confusing you right now
19:53 NanoSector: I don't think it's an easy job, working on a GPU driver anyway; I really admire you guys for getting this far, anyway
19:54 NanoSector: I'll stay connected in this channel (with bouncer), if anything comes to mind or you'd like me to test something, you can either PM me (preferred as I won't lose those messages) or poke me here
19:54 NanoSector: Thanks for your help tonight at least :)
19:55 karolherbst: no problem
21:31 NanoSector: karolherbst, hey, guess what, i got it working
21:31 NanoSector: well
21:31 NanoSector: i said too much
21:31 NanoSector: but i got it working longer than it did :D
21:31 karolherbst: :D
21:31 karolherbst: yeah, it happens
21:32 NanoSector: I noticed that in the pstate, when running without manually setting one, it didn't have a flag next to any of the pstates
21:32 NanoSector: so I was like, lets start with the lowest (07) and work up to the highest, step by step
21:32 NanoSector: and it ran fine for like 30 seconds
21:35 NanoSector: karolherbst, so I also noticed that when my GPU hangs, my wifi sometimes also throws an error that the hardware needs to be restarted
21:36 NanoSector: so that got me thinking, what if the issue is not in my *GPU*, but rather in the PCI bus handling? Since that's the only thing that they have in common, they're on the PCI bus
21:37 karolherbst: well, sometimes the gpu just messes up big times and writes garbage into sys mem I think
21:37 NanoSector: which would make more sense to me because Nouveau also claims it is trying to set a speed but failing
21:38 NanoSector: <NanoSector> [ 140.461822] nouveau 0000:01:00.0: pci: 8.0GT/s not supported by bus or card, droppingrequested speed to 2.5GT/s
21:38 karolherbst: nope, thats becauise the gpu is off
21:38 NanoSector: oh
21:38 NanoSector: :(
21:38 NanoSector: wishful thinking
21:39 NanoSector: anyway I'll let it run in 0a state for a while, since that worked
21:39 NanoSector: then I ramped it up to 0f and that froze
21:41 NanoSector: still running...
21:41 NanoSector: oh, nope, there it went
21:41 NanoSector: :(
21:42 NanoSector: well, that was relatively long, anyway
21:50 NanoSector: karolherbst, echoing 2 to 'boost' and then 0a to 'pstate' causes my GPU to run at a core freq of 1137 MHz and 0 MHz memory
21:50 NanoSector: so i'm lost, will stop pinging you now, sorry :P
22:56 tobijk: hmm shader-db is "really flexible" in chosing the right device to use :/