03:13 FrankD_: allo
07:52 gnurou: just noticed that this went online: ftp://download.nvidia.com/open-gpu-doc/virtual-p-state-table/1/virtual-P-state-table.html - not sure how much is new though
08:13 skeggsb: gnurou: ah, i was wondering when that was going to happen, it has been mentioned before
08:13 skeggsb:takes a look
08:15 skeggsb: it's funny how much stuff is "reserved" ;)
08:45 gnurou: skeggsb: it's because we are planning for the future! :]
09:46 RSpliet: "this code is based on nouveau driver which is capable of reading various entries from BIT."
09:47 RSpliet: should this still be carried into envytools?
10:25 RSpliet: mmmeh, I guess I should just look that up if I really want to know :-P good to know the real name of the table and fields though :-)
16:27 imirkin_: hrm, sounds like recent mesa changes killed glReadPixels? :(
16:31 imirkin_: someone's gonna have to take a look at what's wrong with the readpixels cache =/
16:31 imirkin_: somehow i'm guessing that someone will be me
17:12 zeq: RSpliet: I just tried using your experimental fermi memory reclocking patches, I know it's not expected to work but if it's useful it failed with "fb: missing link training data", then "fb: init failed, -22". (GF108 - GDDR5)
17:36 mvaenskae: imirkin_: cheers, you helped me with the nouveau issue on my macbook 2010 the past days; out of interest, should i report my issues to some bugtracker?
17:37 imirkin_: mvaenskae: https://bugs.freedesktop.org/show_bug.cgi?id=83957 and https://bugs.freedesktop.org/show_bug.cgi?id=77574
17:37 imirkin_: do those sound like what you were having?
17:37 mvaenskae: imirkin_: i have a feeling that some registers are not set up correctly as upon init i get already two nouveau errors when accell is enabled, just one when noaccell is running
17:38 imirkin_: if not, file a new bug
17:38 imirkin_: please prefix it with [NVAF]
17:38 mvaenskae: imirkin_: lately my device just crashes the kernel horribly, frozen X
17:38 mvaenskae: but i can turn it off and it then recovers the screen for shutdown
17:38 imirkin_: remind me if you're using fancy new GL ui?
17:39 mvaenskae: i am using i3 and when running nouveau.noaccell=1 i have no issues except higher cpu usage and my fans becoming really cheerful when doing video decoding
17:39 imirkin_: are you using vdpau?
17:39 imirkin_: (when using acceleration)
17:40 mvaenskae: i use vaapi and vdpau (vaapi being emulated through vdpau)
17:40 imirkin_: what player do you use?
17:40 mvaenskae: at least i used it, now i cannot as i disabled it
17:40 mvaenskae: mpv that is
17:40 imirkin_: hm ok.
17:40 imirkin_: well mpv is known to have huge issues with nouveau
17:40 orbea: heh, mpv + hwdec = no
17:40 imirkin_: i'd recommend trying with mplayer.
17:40 orbea: +nouveau*
17:41 mvaenskae: and i use the nvidia-firmware as well but it seems to load nouveau only (i mean by nvidia-firmware this https://nouveau.freedesktop.org/wiki/VideoAcceleration/)
17:41 mvaenskae: the issue is, it just hardlocks at random, even without mpv
17:41 mvaenskae: like randomly updating my htop view :)
17:42 imirkin_: ok
17:42 imirkin_: then it's not that particular issue
17:42 mvaenskae: interested in some errors?
17:43 mvaenskae: i have 50k lines of errors in my kernel, most of them likely due to nouveau :)
17:43 mvaenskae: kernel-log that is
17:43 imirkin_: not particularly tbh. but file a bug, it should be useful to (a) others experiencing the same issues and (b) for a nouveau developer who's interested in investigating
17:44 imirkin_: be sure to include the full kernel log since boot, not just a few selected lines.
17:45 mvaenskae: i should file it on bugs.freedesktop.org, correct?
17:46 mvaenskae: also i do believe a kernel log without nouveau.noaccell=1 is required, right? :)
17:46 imirkin_: yes, under xorg -> Driver/nouveau
17:46 imirkin_: right. and with the errors.
17:46 mvaenskae: i just wonder what causes X to just hang itself
17:47 mvaenskae: meaning not even trashing the framebuffer anymore
17:47 imirkin_: GPU hang of sorts probably?
17:47 mvaenskae: it never did that in the past that's why it's strange
17:47 mvaenskae: should i also supply some information of my installed packages?
17:48 mvaenskae: it would be good to know which ones you think i should post
17:48 imirkin_: well ... what changed "since the past"?
17:50 mvaenskae: generally new install and i did change my pdf renderer to not use opengl (mupdf has opengl support, i disabled it, i am not 100% certain what it did on the lower levels though)
17:51 imirkin_: mvaenskae: i mean if it's a kernel change, you could bisect
17:51 imirkin_: i forget - did i tell you to remove nouveau_dri.so?
17:51 imirkin_: or was that someone else?
17:52 imirkin_: anyways, if that wasn't you, that's something to try - it's a step between disabling acceleration entirely and leaving it fully on.
17:52 mvaenskae: no, that is something that i still left out as i believe noaccel=1 will disable it from being loadable
17:53 imirkin_: right, so noaccel disables *everything*
17:53 imirkin_: whereas removing nouveau_dri.so just disables GL
17:53 mvaenskae: it makes it work though
17:53 mvaenskae: GL == ??
17:53 imirkin_: so you still get X11 (and Xv) acceleration
17:53 imirkin_: GL = OpenGL
17:54 mvaenskae: i will give it a whirl tomorrow with nouveau_dri.so being renamed
17:54 mvaenskae: and using accel
17:54 imirkin_: [and, amusingly, you'll still be able to use vdpau accel, at least with mplayer]
17:55 mvaenskae: i believe the "X[1309]" shows me that X running with pid 1309 caused this, right? "trapped write at 002038e000 on channel 3 [0fb1a000 X[1309]] engine 00 [PGRAPH] client 0b [PROP] subclient 00 [RT0] reason 0000000b [VRAM_LIMIT]"
17:56 imirkin_: that sounds like something dumb happened. generally the first error message is the most significant one.
17:56 imirkin_: the rest tend to be fallout
18:00 mvaenskae: https://bpaste.net/show/5b95666a53bc <--- 15:33:17 shows when the corresponding error happened
18:02 mvaenskae: imirkin_: oh, should i also make the logging level larger for nouveau when creating a kernel log? if so what parts should i set to debug?
18:02 imirkin_: nah
18:02 imirkin_: Jun 23 15:33:17 gejutel kernel: nouveau 0000:04:00.0: gr: 00100000 [] ch 3 [000fb1a000 X[1309]] subc 2 class 502d mthd 0210 data 00002000
18:03 imirkin_: interesting. we're doing something bad to the 2d engine?
18:03 imirkin_: Jun 23 15:33:17 gejutel kernel: nouveau 0000:04:00.0: gr: TRAP_TEXTURE - TP0: 00000009 [ LINEAR_MISMATCH]
18:03 imirkin_: oh my how it hates us
18:04 mvaenskae: it looks awesome right? :D
18:04 mvaenskae: could it be due to some faulty HW?
18:06 mvaenskae: i do believe i owe you guys a bug report :) i will have it done this weekend
18:19 unraised: Hi, I'm having problem with a multicard setup. I have 2 Quadro NVS 290 cards installed, which support dual-head. If I plug two monitors into one card (leaving the other card unused), everything is fine. If I plug one monitor in each card, dmesg streams output:
18:20 Yoshimo: i would use pastebin unraised
18:21 unraised: http://pastebin.com/L0MJj02u
18:23 unraised: According to Freedesktop, the NVS 290 is a NV50 (Tesla) and the FeatureMatrix says multicard is supported, so I'm guessing this is a bug.
18:24 unraised: The problem is I want to use all four ports, over both cards, but I get those errors in the pastebin link and performance seems to suffer.
18:55 RSpliet: zeq: remind me to demand your VBIOS in due time
18:56 RSpliet: (or send it to mmio.dumps at gmail)
18:56 imirkin: unraised: are those the very first errors you see?
18:59 imirkin: mvaenskae: hmmmm... we try to stick 0x2000 into the layer. that's obviously false. feels like some off-by-one issue
19:19 imirkin: karolherbst: so does tressfx now work in tomb raider with the patches ken pushed?
19:19 karolherbst: imirkin: didn't check yet
19:20 imirkin: karolherbst: also ... did you say witcher2 had issues?
19:20 karolherbst: imirkin: and usually I won't have time to do anything related in the week anyway
19:20 karolherbst: imirkin: yeah
19:20 imirkin: like ... rendering issues, or just perf?
19:20 hakzsam: witcher2 has serious issues
19:20 hakzsam: both
19:20 karolherbst: imirkin: related to that extension you can't apitrace
19:20 imirkin: buffer storage?
19:20 karolherbst: yeah
19:20 karolherbst: I got a version without buffer storage
19:20 karolherbst: no issues there
19:21 imirkin: unfortunate =/
19:21 karolherbst: yeah
19:23 zeq: RSpliet: I don't seem to be able to get a dump with nvagetbios, nor from the sysfs rom file. Probably because of the failure to init nouveau. I'll try again next time I reboot.
19:34 karolherbst: imirkin: what do you think is the impact on having divergent errors?
19:34 imirkin: karolherbst: i have no clue what they are
19:34 karolherbst: I see
19:34 imirkin: karolherbst: i first saw them in F1 2015
19:34 karolherbst: I saw those in payday 2
19:34 karolherbst: well there are a lot of those
19:35 imirkin: i dunno if it's just a normal condition complaining of divergence, or if it's an error in the shader
19:35 imirkin: *normally* those errors don't indicate "normal" situations, but who knows
19:35 karolherbst: I am sure those are errors, but mhh what might they mean
20:32 zeq: RSpliet: sent my VBIOS to mmio.dumps
21:22 imirkin: skeggsb: reminder about that no-fbdev crash - the guy sent a patch...
21:50 skeggsb: imirkin: thanks, done