00:00 nyef`: Yeah, that I can't help you with (though someone else here might be able to). Most advanced thing I have locally is a Maxwell.
00:01 ashkitten: oh it might be that i have nvidia-libgl installed instead of mesa
00:02 ashkitten: hmm
00:02 ashkitten: i have both actually
00:02 karolherbst: pmoreau: anything new :O
00:02 nyef`: So you could be picking up the "wrong" one.
00:12 ashkitten: ok so whenever the gpu gets overloaded everything starts to turn black
00:14 ashkitten: this is a new one...
00:14 ashkitten:uploaded an image: file1501892048028.jpg (107KB) <https://matrix.org/_matrix/media/v1/download/matrix.org/DKaueQzDygjRXOOoQEjCqxlp>
01:39 glisse: what is maxwell limit on virtual address, 40bit ?
01:40 mwk: yeah
01:40 glisse: oh i thought it was higher on maxwell already
01:40 mwk: hmm
01:41 glisse: likely me having bad memory/impression :)
01:41 mwk: changing it would be quite involved, but let me double-check...
01:41 glisse: i ask from hardware perspective
01:41 glisse: not from a software one
01:42 mwk: yeah, same here
01:42 glisse: the kernel has couple issue to support anything more
01:42 glisse: i mean nouveau
01:42 mwk: there's lots of places where the bitfields are tucked tightly
01:42 mwk: they'd have to move to support >40-bit addresses and we'd notice
01:43 glisse: they have a new mmu class afaict
01:43 mwk: nope
01:43 mwk: seems to be same old 40-bit
01:43 glisse: pascal is 49bit
01:44 mwk: oh?
01:45 glisse: yeah they have a new class on pascal from what i was told but the old one keep working
01:46 mwk: class of what?
01:46 mwk: MMU doesn't involve classes
01:46 glisse: well mmu if that's the right name
01:46 mwk: hmm
01:46 mwk: could be
01:46 mwk: it's not like I know much about Pascal
01:47 mwk: nouveau just uses same MMU driver as Fermi
01:47 glisse: it might be part of the channel class
01:47 glisse: if channel are class
01:50 glisse: https://www.google.com/url?sa=t&rct=j&q=&esrc=s&source=web&cd=1&cad=rja&uact=8&ved=0ahUKEwisosWt_b7VAhXk7oMKHXm-ADIQFggmMAA&url=http%3A%2F%2Fdownload.nvidia.com%2Fopen-gpu-doc%2Fpascal%2F1%2Fgp100-mmu-format.pdf&usg=AFQjCNElLvn_WgVuUtyGOeQCDPhVXQq7-A
01:50 glisse: http://download.nvidia.com/open-gpu-doc/pascal/1/gp100-mmu-format.pdf
01:57 mwk: huh, that's something new
08:06 karolherbst: pq: the funny thing is, that the vbios of that GPU simply looks to be broken
08:06 karolherbst: or "buggy"
08:07 karolherbst: if nvidia doesn't select the highest pstate, then all we have to check is, if every cstate and the base state of a pstate can be reached, otherwise disable that pstate
08:13 karolherbst: or we simply miss some knowledge
08:16 JustOneIt: tty1 gives me this error: libGL error: failed to load driver: nouveau
08:17 JustOneIt: and i got tearing (google-chrome playing netflix)
08:17 JustOneIt: any idea?
09:22 JustOneIt: which is the best configuraion to have a fast and smooth video (netflix) playing? Without tear effects if possible...
09:23 JustOneIt: maybe switching to compton (is it possible with nouveau)?
09:25 JustOneIt: or the best is to keep a plain nouveau install?
09:35 JustOneIt: is it possible to use opengl with nouveau?
21:47 mwk:slowly becoming mad and obsessed with patches, swatches, curves, and transitions
21:50 mwk: I'm not even sure I know what patches/swatches/curves/transitions *are*
21:51 RSpliet: If I can recommend a good therapist, mine is called Jose Cuervo
21:51 mwk: aaaaaaaaaaaaaaaaaaaa
21:54 mwk: the fact that constructing FD context badly blows shit up before I even submit a command doesn't help matters :(
22:16 skeggsb: glisse: yeah, pascal has a new mmu layout (49-bit), but supports the old style too. i have code to switch to the new one, been trying to clean it up and push it for you along with the other patches i sent you recently
22:29 glisse: skeggsb: yeah i was bite by that, still trying to find my heading in the code
23:14 karolherbst: skeggsb: what were those mmiotraces issues you had tracing a gp108?
23:14 skeggsb: karolherbst: the machine just locked up for unknown reasons
23:14 skeggsb: i tried a few times
23:14 karolherbst: :(
23:14 skeggsb: the driver worked fine without mmiotrace running, so, it's not that
23:15 karolherbst: I think there might be a bug somewhere within the mmiotracer, but I wasn't able to track it down yet
23:16 karolherbst: skeggsb: does pstore works on your machine?
23:16 skeggsb: i don't even know what that is :P
23:17 skeggsb:found out
23:20 karolherbst: if you have a reliable way of getting kernel logs you could try out a patch I have, which also prints some stuff into the log
23:20 karolherbst: skeggsb: https://github.com/karolherbst/linux/commit/92aea447a776f10aad0a2e971b5f2b208a1161d2.patch
23:20 skeggsb: i can probably use netconsole if pstore doesn't work i reckon
23:20 karolherbst: hopefully
23:21 skeggsb: i'll give it another try tomorrow
23:21 karolherbst: but this is page fault handling we are talking about, in doubt, nothing works
23:21 skeggsb: with your patch too if that'll help
23:21 karolherbst: I fear that some pages are armed twice or more, but don't know for sure
23:23 karolherbst: uhh, there is a minor mistake in that patch...
23:24 karolherbst: skeggsb: updated version: https://github.com/karolherbst/linux/commit/6ba23f59f142b73888716c547fcdce32194f28ac.patch
23:51 mwk: mmmm, secret blend equations
23:51 mwk: or a secret blend of 7 equations?
23:57 skeggsb: mwk: https://github.com/blueshogun96/Cxbx/blob/master/src/CxbxKrnl/EmuD3D8/Convert.h << potentially indicates what those *might* be
23:59 mwk: well that came out of nowhere