05:25lsd_: can someone explain me shortly what it does mean? MMIO write of 8000001f FAULT at 6013d4 , ignore ?
17:07YuGiOhJCJ: hello, I already asked the question yesterday and I got no answers, but I try again today who knows... I am using a GeFroce GTX 1060 with nouveau 1.0.15 on linux 4.19.9 and I have some tearing problems with this video https://youtu.be/ceX18O9pvLs is there an option to fix that in /etc/X11/xorg.conf.d/20-nouveau.conf please?
17:08imirkin_: [i haven't looked at your video, but tearing is hard to fix]
17:14YuGiOhJCJ: oh... but with the Intel driver I was able to fix tearing with these two options: Option "AccelMethod" "sna" and Option "TearFree" "true", nothing similar for nouveau?
17:15imirkin_: you can use modesetting, which maybe has a TearFree option?
17:15imirkin_: but then you're using GL-based accel so you just get misrendering ;)
17:15imirkin_: [and increased likelihood of hangs]
17:23YuGiOhJCJ: that's maybe what the Intel driver options did too because I had some misrenderings when I enabled them, but having no tearing is so good for videos
17:25YuGiOhJCJ: I will take a look to modesetting
17:29imirkin_: how is the video being displayed?
17:29imirkin_: i suspect if you use xv, you'll have less tearing
17:30YuGiOhJCJ: the video is displayed through Firefox
17:30imirkin_: yeah dunno
17:31YuGiOhJCJ: xv? you mean the interactive image display for the X Window System?
17:31imirkin_: i mean XVideo
17:31imirkin_: which is an extension to the original X11 Window System
17:38YuGiOhJCJ: I read that https://en.wikipedia.org/wiki/X_video_extension and there is a code at the bottom of the page, it compiles and run fine on my machine so I suppose that xv is installed, do I need to load this extension when I start X? or maybe just applications like Firefox need to be linked to this library?
18:25YuGiOhJCJ: it seems that modesetting does not have a "TearFree" option https://manpages.debian.org/stretch/xserver-xorg-core/modesetting.4.en.html anyway if I use this driver, I suppose that I will not have 3D acceleration and that could be a problem for video games, so I must keep nouveau and unfortunately I will get tearings :\
18:28imirkin_: you should still have 3d accel
18:28imirkin_: in fact it uses opengl to provide you with 2d accel
18:28imirkin_: so you're relying on the nouveau 3d component even more! yay!
18:31YuGiOhJCJ: oh nice but yes, no TearFree option with modesetting
18:32imirkin_: (as opposed to using a much simpler and way-more-tested and way-more-error-resilient impl in xf86-video-nouveau)
18:32YuGiOhJCJ: I also read something interesting https://wiki.archlinux.org/index.php/Nouveau#Vertical_Sync using Compton could be a solution
18:33imirkin_: yeah, a compositor can do the trick
18:33imirkin_: i should look into doing TearFree for nouveau
18:33imirkin_: should be relatively straightforward, or so i've been led to believe
20:51Lyude: no promises, but I'm gonna try to get nouveau a christmas present this year if I get the time
20:52imirkin_: fermi reclocking?
20:52Lyude: hehe, nah
20:52Lyude: just some stuff that has been waiting on me for too long :)
20:57imirkin_: not sure what that would be, but happy to take anything!
21:22huehner: Hello. trying to understand issue here with Pascal GTX 1050 (4.18.20, mesa 18.3) starting a game (minecraft java/ogl) get X locking up with following error here: nouveau 0000:01:00.0: fifo: fault 00 [READ] at 00000000035f9000 engine 15 [CE0] client 01 [GPC0/T1_0] reason 02 [PTE] on channel 3 [007fb84000 Xorg]
21:23imirkin_: huehner: iirc minecraft does multi-threaded GL access?
21:23huehner: full dmesg output around that: https://pastebin.com/Ez8JimJy, Xorg not updating or responding to events until app is killed with kill -9
21:24huehner: imirkin_: don't know but possible
21:24huehner: imirkin_: always happens on start at same point with app log showing [22:18:54] [Client thread/INFO]: Created: 4096x2048 textures/blocks-atlas
21:24huehner: so could me related to uploading a texture of huge size (but a bit guessing here)
21:25imirkin_: well, it's saying the copy engine is full of sadness
21:25imirkin_: because some texture is not mapped in
21:25imirkin_: (CE0 = copy engine 0... there are 3+ of them)
21:25imirkin_: iirc 2 is used by the kernel to do buffer moves to-and-from vram
21:26huehner: imirkin_: browsing in skeggsb repo i found this commit, https://github.com/skeggsb/nouveau/commit/18dbe3649de02d93e13f89ffb8016ec8ca0d78e7
21:26imirkin_: this is in reference to display things
21:26huehner: imirkin_: apart what would be a good way for me to dig out more info here?
21:26imirkin_: not render things
21:27imirkin_: huehner: i believe karolherbst has a branch somewhere with his attempt at fixing the multithreaded thing
21:27imirkin_: tbh i haven't had the opportunity to look at it
21:27imirkin_: but he claims it helps in some cases
21:28imirkin_: unfortunately we have little-to-nothing in the way of debugging
21:29huehner: imirkin_: could apitrace being an idea giving that it fails at same point consistenly apparentl all the time?
21:30imirkin_: huehner: sure.
21:30huehner: then i'll try to figure that out
21:30imirkin_: i wouldn't want you to have the wrong expectations though
21:31imirkin_: most likely you will give up before anything is resolved.
21:31huehner: 2nd topic: probably unrelated, i see single error (in dmesg) on each boot just after firmware loading:
21:31huehner: nouveau 0000:01:00.0: bus: MMIO read of 00000000 FAULT at 122124 [ IBUS ]
21:31huehner: but i don't have any noticable issue
21:31imirkin_: according to rnndb,a t least
21:32huehner: imirkin_: thx for the heads up... i remember my first (and only) nouveau kernel patch... took me a long time to write 2 line fix ;)
21:32imirkin_: would have to figure out where it's getting issued
21:32imirkin_: (somewhere in the kernel driver)
21:32imirkin_: and figure out why the gpu hates it
21:32imirkin_: could be that the firmware is doing it at the wrong time
21:32imirkin_: in which case there's nothing you can do about it
21:33huehner: ok... i'll start with the other issue then as more motivating for that one (as something doesn't work for me here) ;)
21:48huehner: imirkin_: having LD_PRELOAD set for api-trace and getting a trace i got different behavior (no dmesg fault) but assertion failure https://pastebin.com/LuLDCCL8
21:49huehner: imirkin_: does that add any useful info by chance?
21:53imirkin_: yeah, there's like a 99.8735% chance that it's due to the multithreading thing
21:53imirkin_: i believe this is karol's latest effort: https://github.com/karolherbst/mesa/tree/mt_fixes_take2
21:54huehner: imirkin_: thx, let me try to build that and give it a try then
22:24pmoreau: Phew, there we go: `clinfo` running properly on this laptop (MCP79+G96). Karol had forgotten to set up some plumbing for Tesla cards. :-)
22:26pmoreau: “<built-in>:1:10: fatal error: 'opencl-c.h' file not found” Grrrr, I was certain I had fixed that stuff a long time ago. :-(
22:32huehner: imirkin_: that branch from karol seems to have been good idea :)
22:33huehner: imirkin_: did not fail but went past without a hitch (on first run) will probably try that a bit more over next couple of days
22:33huehner: imirkin_: is he still around here sometimes to give him feedback?
22:33imirkin_: he's normally here
22:33imirkin_: but i think he's on vacation
22:33imirkin_: but i'm not sure
22:35huehner: ok ,thx
22:53imirkin_: skeggsb: pin