07:20 karolherbst: mupuf: yay, now we even have support for the hwmon stuff within mesa :)
08:14 netz: whassup. any news on nouveau for gtx 750 ti?
08:23 haypo: hi. nouveau failed with "fifo: CACHE_ERROR" and "fb: trapped write" http://paste.alacon.org/42405 it seems like i'm not the first one having the error (i found similar recent bug reports), what do you need for an useful bug report?
08:23 haypo: i still have access to the computer through SSH, can i run some debug program?
08:26 haypo: it's Fedora 24, VGA compatible controller: NVIDIA Corporation G96 [GeForce 9400 GT] (rev a1) (prog-if 00 [VGA controller]), xorg-x11-drv-nouveau-1.0.12-4.fc24.x86_64, mesa-dri-drivers-12.0.3-1.fc24.x86_64
08:26 haypo: marcheu: ^^
08:34 haypo: sorry, i need this computer, i have to reboot it
09:08 karolherbst: netz: with 4.9 there will be reclocking support
09:09 netz: nice. there was something else missing, right?
09:09 netz: 4.9 kernel?
09:15 karolherbst: netz: yes
09:16 netz: well shite :P
09:16 netz: I may have jumped ship to the rx480 before then :P
09:16 karolherbst: well, you can always compile nouveau yourself :p
09:17 netz: too much other stuff to be done for that sort of geeky masturbation :P
13:13 pmoreau: imirkin_: Another option could be the wiki from Phabricator, as freedesktop hosts an instance of it https://phabricator.freedesktop.org. But it seems to suffer from the same problem as the current one: let's only use 1/3rd of your screen, because.
13:15 karolherbst: pmoreau: .....
13:17 karolherbst: pmoreau: I doubt that the phabricator wiki will handle multiple projects that well though
13:17 pmoreau: But otherwise the design is quite nice.
13:17 pmoreau: Why so?
13:18 karolherbst: isn't there one wiki for all freedesktop?
13:18 imirkin: appears to require to log in up-front
13:18 karolherbst: and that as well
13:18 karolherbst: ohh wait
13:18 ajax: which can be literally anything
13:18 karolherbst: imirkin: https://phabricator.freedesktop.org/w/
13:18 karolherbst: this is fine
13:18 karolherbst: but yeah
13:19 karolherbst: I doubt we can do multiple wikis
13:19 karolherbst: and then we also have account management
13:19 karolherbst: like I can't edit anything
13:20 pmoreau: I think we can talk with Daniel Stone about that, since he seems to be managing it
13:21 karolherbst: I don't want to depend on others though to enable wiki access…
13:21 pmoreau: Well, then, sure
13:22 pmoreau: You can do multiple wikis: GStreamer and Wayland have "some" kind of wiki there.
13:23 imirkin: ideally we'd be a real project with real resources and real time to put towards these things
13:23 imirkin: sadly we live in the real world
13:23 karolherbst: pmoreau: it is one wiki basically
13:24 pmoreau: karolherbst: It has one root, but you can set different write/read access per subtree.
13:26 karolherbst: yeah okay, maybe we could do it this way then. But for this, we need a Nouveau project on it as well
13:28 pmoreau: Daniel Stone offered me to have my SPIR-V work hosted there, I assume he would be ok with creating a Nouveau project, given that every Nouveau dev agrees (and that I haven’t scared off imirkin already :-D)
13:28 karolherbst: well, one thing though, the wiki isn't stored in a format we could just download, is it?
13:29 pmoreau: Let me check
13:34 pmoreau: Probably not, as you would need a separate Git repo for each subtree, but as you said, this is a single hierarchy, so most likely with a single Git repo underneath
13:35 karolherbst: pmoreau: well, we should know for sure, because before we set for something, we should make sure to not vendor lock in ourselves
13:36 pmoreau: yup
13:37 pmoreau: I’ll ask on the IRC channel and check on the website.
13:41 karolherbst: uhh SQL stuff
13:42 pmoreau: Looks like not: https://secure.phabricator.com/T8594 ; there is some redesign going on: https://secure.phabricator.com/M1466
13:42 karolherbst: well the question is rather if we have access at all? is that phabricator hosted by freedesktop or hosted by those?
13:43 karolherbst: pmoreau: I checked the sources
13:43 karolherbst: well, if we want to export it, we also want the history for obvious resons
13:45 karolherbst: I would rather go for a git based wiki or something similiar, because then we can pretty trivially edit those files without having to mess with the websites, or even autogenerate a few things
13:45 pmoreau: It is hosted by freedesktop
13:45 pmoreau: Like I am hosting my own
13:45 karolherbst: I see
14:02 RSpliet: karolherbst: the freedesktop wiki *is* based on git
14:09 karolherbst: RSpliet: I know
14:29 mwk: "simple" is, of course, relative...
14:51 E14n: I am having problems getting 3x4k displays working. I have had the working with the proprietary nvidia drivers but had some issues. The troubleshooting guide recommended I ask here about raising the "card performance mode"
14:51 E14n: Can anyone help with that
14:52 Calinou: E14n: that requires a recent kernel + enabling a boot time option IIRC
14:52 Calinou: karolherbst should know more about this
14:52 imirkin_: E14n: what gpu do you have?
14:53 imirkin_: E14n: and what issues are you running into with nouveau?
14:58 E14n: imirkin_: I have a gtx980ti http://www.gigabyte.com/products/product-page.aspx?pid=5472#ov
14:59 E14n: imirkin_: I am having problems getting it running on more then two displays.
14:59 imirkin_: E14n: that's a GM200 i believe. i kinda doubt nouveau will be able to drive 3x 4k screens with it, but worth giving it a shot i suppose
14:59 waltercool: yeah, 980ti is GM2XX
14:59 imirkin_: E14n: how are you connecting the screens?
15:00 imirkin_: [note that due to nvidia requiring signed firmware, nouveau is no longer able to really operate new GPUs]
15:01 E14n: imirkin_: 3 display port cables.
15:01 imirkin_: and are the monitors MST?
15:01 imirkin_: i.e. are they the type of monitor that secretly has 2 panels inside that have to be driven individually?
15:01 imirkin_: if so, that's why you're maxing out at 2
15:01 imirkin_: kepler+ gpu's only have 4 CRTC's
15:02 waltercool: imirkin_: I can use dual screen with GM204, and I can use 4 monitors with nouveau. I just have tested with one external video only btw on my laptop
15:02 imirkin_: which means that you can have, basically, 4 separate "picture" streams. if each monitor is taking up 2, then you max out at with 2 monitors (2*2=4)
15:02 imirkin_: waltercool: thank you. i'm well aware of the various limitations :)
15:03 waltercool: imirkin_: not a limitation, my laptop doesn't have more than 3 external interfaces :)
15:04 waltercool: E14n: xrandr does display all those 3 display interfaces?
15:04 karolherbst: E14n: are all displays connected since boot?
15:04 karolherbst: E14n: also, what exactly are your issues?
15:05 imirkin_: [of course, if they are those DP-MST panels, they won't work too well with nouveau, as it doesn't support MST, only SST...]
15:05 E14n: imirkin_: I don't think they are MST. Two are DELL P2715Q and the second is a Acer XB280HK bprz https://www.dell.com/content/products/productdetails.aspx/dell-p2715q-monitor
15:06 karolherbst: E14n: if your issue is, that you have a lot of flickering, thank Nvidia for that, cause we can't do shit about that until we get PMU firmwares
15:06 karolherbst: E14n: does it work with 1 or 2 displays? (kind of)?
15:06 E14n: waltercool: yes they all display, but one appears disconnected.
15:06 karolherbst: ohh I see
15:07 karolherbst: I except really bad perf though
15:07 E14n: karolherbst: yes two work reasonably well. I am just using it for coding, but the third shows as disconnected.
15:08 karolherbst: I see
15:08 E14n: At the moment I have all showing as connected but none they are showing blank screens.
15:08 karolherbst: well, there are some issues with a third display for sure
15:09 imirkin_: with nouveau, or with nvidia?
15:09 E14n: nvidia works fine. I was changing because I couldn't get a mono program to work with the nvidia drivers.
15:10 waltercool: E14n: Just an stupid question, if you disconect one of those 'already connected monitors', and replug the one with problem, will work? (I mean, understading if is a limitation of more than 2 monitor, or a port issue)
15:11 E14n: imirkin_: When I say works fine, it displayed full screen video almost everything else.
15:11 imirkin_: E14n: are you on up-to-date software? iirc we had a (wrong) limitation around maximum fb size at one point
15:12 E14n: waltercool: Yes I switched the port the cable was on and the other monitor worked, but only after a reboot.
15:13 E14n: imirkin_: I believe I have the latest release versions. Let me check the exact versions
15:14 imirkin_: E14n: this would be linux kernel and xf86-video-nouveau potentially
15:14 imirkin_: although... you need a pretty new kernel for GM200 support...
15:14 waltercool: after... 4.6, right?
15:14 imirkin_: no, i think GM200 was only added in 4.7
15:14 E14n: xf86-video-nouveau-1.0.12, Linux version 4.8.0-rc8
15:15 karolherbst: well
15:15 karolherbst: modesetting ddx for that gpu anyway
15:15 imirkin_: right, whereas the change i was thinking of (5102ec3e99e271b3934a1e00bb52a2545eb7dbc5) went into kernel v4.3
15:16 waltercool: Yeah, saw it
15:16 E14n: karolherbst: I was using Linux Kernel 4.7.5 before but no difference
15:17 waltercool: E14n: Just for extra info which may worth to analyze, can you pastebin your Xorg and dmesg?
15:17 karolherbst: E14n: well I am sure it is simply nouveau doing silly things. But because no nouveau dev has access to such a setup, we can't really do much about it except trying to guess where the issue is
15:18 waltercool: I have tested a bit nouveau with my GM20X, but I don't have too much monitor to do that kind of testing sadly
15:18 waltercool: too many*
15:20 E14n: dumb question is there a good way to dump dmesg to postbin.
15:21 waltercool: dmesg > dmesg.txt, then upload dmesg.txt
15:22 waltercool: Time ago there it was a good CLI app for pastebin and nopaste, but I can't found those apps anymore on my repo
15:22 RSpliet: Fedora still has fpaste, those kind of things are a bit distro-dependent
15:22 E14n: waltercool: OK I am remote logged in because the screens aren't working.
15:24 waltercool: E14n: When I connect an external monitor in my laptop, all monitors goes f*cked up, so I need to run xrandr --auto and fix theirselves
15:26 waltercool: not automatically goes on that way, but when I enable it with KDE
15:26 waltercool: RSpliet: at least, isn't on Gentoo :'(
15:30 waltercool: Hey karolherbst, btw, just in case you need to do some testing with my nasty GM204, I can do it
15:30 waltercool: I mean, with some development
15:35 karolherbst: waltercool: I am fine :) just need those silly pmu firmwares
15:36 karolherbst: if you don't mind your GPU being burned, we can fully reclock it already
15:36 waltercool: karolherbst: lol yeah
15:36 waltercool: karolherbst: how do you reclock it?
15:36 karolherbst: just controlling the fan is no thing we can do without the firmware
15:36 waltercool: I have full manual fan for that kind of tests
15:36 karolherbst: waltercool: same as kepler
15:36 karolherbst: waltercool: as in hw control?
15:36 karolherbst: sysfs won't work
15:37 waltercool: just combo keys
15:37 waltercool: is hardware fan
15:37 waltercool: I meant, a hardware combo keys
15:37 karolherbst: ahh I see
15:37 karolherbst: so bypassing OS
15:37 waltercool: yep
15:37 karolherbst: EC controlled fan?
15:38 karolherbst: anyway, gotta go. Sadly we can't be sure if the fan is there or not, which makes the situation even uglier
15:38 waltercool: I'm not sure, but works also at BIOS
15:38 karolherbst: because on laptops, we could just recock those cards
15:38 karolherbst: …
15:38 karolherbst: well, gotta go anyway
15:38 RSpliet: freudian typo?
15:38 waltercool: np ;)
16:03 waltercool: imirkin_: How do you reclock maxwell on git nouveau? There is some exposed functionality to userspace or do I need to recompile modifying something?
16:10 imirkin_: no clue
16:10 imirkin_: first you need to enable it
16:11 imirkin_: it's only flipped on for GM10x, not GM20x
16:19 E14n: waltercool: dmesg is at http://pastebin.com/aLvFuZYF I am using the builtin xorg.conf and /etc/X11/xorg.conf.d/10-monitor.conf http://pastebin.com/M1MXZZJE
16:20 imirkin_: skeggsb: nouveau 0000:60:00.0: disp: 0x5d18[0]: INIT_GENERIC_CONDITON: unknown 0x07
16:20 imirkin_: we've seen that before, right?
16:20 waltercool: hmmm yeah, I was looking that also
16:20 imirkin_: it's esp relevant as it used to be called "DP_CONDITION" :)
16:21 iftum: can i reclock fermi?
16:21 imirkin_: iftum: no.
16:22 waltercool: E14n: I can't help after that because I have no idea about it, but I'm pretty sure that error is related with the problem
16:32 waltercool: I found a reference about that here: https://people.freedesktop.org/~cbrill/dri-log/?channel=nouveau&date=2016-06-01 , but I didn't found any solution there
16:37 karolherbst: waltercool: 1. it is messy, 2. it is even more messy. best is to not give any instructions, cause somebody will burn their card for sure
16:37 waltercool: I see
16:38 waltercool: karolherbst: so OK, better to wait for Nvidia to provide the signed blob to do it in a better way
16:38 karolherbst: blame nvidia, it is their fault. or blame us if we are supposed to just exploit any security issue
16:39 waltercool: Jajaja I know, I always blame nvidia lol
16:40 imirkin_: or blame yourself, for buying nvidia
16:40 karolherbst: no idea what those nvidia managers are thinking, because writing exploits will be the only thing we might be able to do in the end
16:40 Wonka: "FUCK YOU, nVidia!"
16:40 waltercool: imirkin_: nah, I want to use an AMD, but there is no enough high end amd videocard for laptops on market
16:41 Wonka: waltercool: well, there's rumours the next MacBook will use an AMD GPU...
16:41 imirkin_: waltercool: just about any AMD gpu will beat out the nvidia gpu you have...
16:41 waltercool: Wonka: I prefer to keep with my maxwell2 to move on macbooks lol
16:41 imirkin_: [unless you're interested in running random code downloaded off the interwebs at ring0]
16:42 karolherbst: the gpu can access full sysram anyway :D
16:42 waltercool: imirkin_: that's sadly true, I switch from nouveau to nvidia when I want to game, but not the idea neither
16:43 Wonka: waltercool: I meant, they can't be that bad...
16:43 karolherbst: pls could we just not have the same discussion each week?
16:43 waltercool: lol
16:43 Wonka: "Polaris", they claim in this (german) article: https://www.heise.de/newsticker/meldung/Bericht-Neues-MacBook-Pro-wohl-Ende-Oktober-3336405.html
16:44 Wonka: I'm mostly here because of this MacBook Pro 3,1 I have, which has an nvidia gpu...
16:45 Wonka: otoh, it's nearly unusable now, so why bother.
16:45 waltercool: Wonka: but almost all Apple hardware uses very weird components, and are very unfriendly to modify them (ram or hdd)
16:45 Calinou: the fact Apple hardware is technically outdated doesn't incite me to get some :(
16:45 waltercool: specially the webcam on macbooks is nasty
16:45 Calinou: (like, they'll stick with Skylake, no Kaby Lake)
16:47 karolherbst: Calinou: because apple selling their software and design, not hardware....
16:47 waltercool: Calinou: yeah, that was a very bad idea from them, usually their late-<year> models are always using new intel gen, I don't know why they didn't wanted to move
16:47 waltercool: I may assume they don't want the same problems of SkyLake 1st gen
16:48 Wonka: waltercool: that's why I probably won't buy any newer Apple hardware. My current MacBook Pro 8,2 (early 2011) at least has a real SATA HDD and exchangeable RAM modules...
16:48 Wonka: waltercool: the webcam is UVC here
16:49 waltercool: waltercool: yeah, their hardware is great, but I used to work with a MacBook 2014 and the hardware is awful, blobed wireless card, the hardware is not upgradeable, the webcam doesn't work (but Thunderbolt one does)... etc
16:49 waltercool: I meant, the hardware qualty is great
16:49 imirkin_: dunno. i can't stand their keyboard, and hate the trackpad.
16:50 Wonka: I wish I had hardware trackpad buttons.
16:50 Wonka: but well, who still has?
16:50 imirkin_: (well, i don't just hate *their* trackpad... all trackpads...]
16:50 waltercool: I like my trackpad lol, it has external buttons, and not clickable trackpad only
16:51 waltercool: clickable trackpad always gave me problems
16:52 Calinou: I am considering either Dell XPS or ThinkPad as next laptop
16:52 Calinou: decent battery life is important to me though
16:54 waltercool: XPS 15 + Razer Core = much win
16:54 waltercool: but I read at reddit it that combo will not work for XPS 13
17:44 karolherbst: if somebody wants to look into CPU bottleneck issues, civ V is a really good example where it really hurts
17:45 imirkin_: karolherbst: where is the CPU going?
17:45 imirkin_: shader building, or "other"?
17:46 karolherbst: no clue, game engine I guess
17:46 karolherbst: it gets worse the longer the game is
17:46 imirkin_: super.
17:46 karolherbst: first rounds, everything completly smooth, later: not so much anymore
17:46 imirkin_: check if nouveau overhead increases
17:46 karolherbst: it doesn't depend on how long the game runs
17:46 imirkin_: we had some "leaks" stuffing buffers into bufctx without ever clearing
17:46 karolherbst: but how far you suceeded with the save
17:46 imirkin_: weird.
17:47 karolherbst: the same with nvidia though
17:47 karolherbst: but not as bad
17:47 karolherbst: might be that gl overhead is rather slim, but I guess I try to take a look
17:47 karolherbst: cpu usage by the game is around 150%
17:48 karolherbst: it might do some texture things though
17:48 karolherbst: cause you see the game reloading textures in areas you didn't looked at for a long time
17:49 karolherbst: and cpu usage jumps by 40%
17:53 imirkin_: weird.
17:53 karolherbst: it is the same with nvidia actually
17:55 karolherbst: would be just nice to reduce cpu overhead for this game, because it is really annoying, especially because the game runs that well from the start
18:30 booti386: Hello!
18:31 waltercool: karolherbst: I think that problem happens with every videocard, I been playing that game for 4 hrs with both AMD (radeonsi) and Nvidia driver, and the game gets super slow
18:31 karolherbst: waltercool: yeah, I figured already
18:33 waltercool: no idea why, but the games get very slow, not only loading the game turns (cause' more units, more things to process), but also graphics performances goes wrongly also
18:33 waltercool: game*
18:33 karolherbst: yeah
18:33 karolherbst: the latter is a shame
18:33 karolherbst: because there isn't _that_ much :D
18:33 karolherbst: except all the improvements and stuff
18:33 booti386: I'm having a display hang after some time of use
18:33 booti386: Dmesg says:
18:34 booti386: [19493.494499] nouveau 0000:01:00.0: fifo: SCHED_ERROR 0a [CTXSW_TIMEOUT]
18:34 booti386: [19493.494506] nouveau 0000:01:00.0: fifo: gr engine fault on channel 2, recovering...
18:34 waltercool: yeah
18:34 booti386: 01:00.0 VGA compatible controller [0300]: NVIDIA Corporation GK106 [GeForce GTX 650 Ti] [10de:11c6] (rev a1)
18:34 waltercool: booti386: kernel version?
18:34 imirkin_: booti386: i think that's common
18:34 imirkin_: booti386: a bunch of people have had success in using the blob ctxsw firmware
18:35 booti386: [ 0.000000] Linux version 4.8.0-040800rc5-generic (kernel@tangerine) (gcc version 6.2.0 20160830 (Ubuntu 6.2.0-2ubuntu11) ) #201609041832 SMP Sun Sep 4 22:34:01 UTC 2016
18:35 imirkin_: my extract_firmware.py script is able to grab it from some blob version
18:35 booti386: Oh, so it's a bug into your reverse-engineered firmwares ? :(
18:36 imirkin_: i dunno if "reverse-engineered firmware" is the proper way to qualify it
18:36 imirkin_: it's a (undiagnosed) bug in the firmware we have written
18:36 waltercool: I think is more test/trial, isn't?
18:36 imirkin_: (and i use the term 'we' quite liberally... i was in no way involved in writing that firmware]
18:37 karolherbst: maybe we should write one from scratch and the issue will just disappear :D
18:37 imirkin_: karolherbst: what do you think the current one is?
18:37 karolherbst: no idea?
18:37 imirkin_: it's written from scratch, afaik
18:38 booti386: imirkin_: Don't your extract_firmware.py only extracts the video decoding firmwares?
18:38 karolherbst: yeah I know, but there is somewhere an issue we just can't see
18:38 karolherbst: so it may be something super silly :/
18:38 karolherbst: especially because it doesn't happen for everybody
18:38 imirkin_: booti386: it also extracts ctxsw fw if you pick the right blob version
18:39 karolherbst: booti386: did you change the clocks?
18:39 karolherbst: booti386: which gpu by the way?
18:39 booti386: imirkin_: Oh, ok :)
18:40 booti386: karolherbst: Nope, but I know it crashes very early if I set it to: 0f: core 1032 MHz memory 5400 MHz
18:41 booti386: karolherbst: NVE6 (01:00.0 VGA compatible controller [0300]: NVIDIA Corporation GK106 [GeForce GTX 650 Ti] [10de:11c6] (rev a1))
18:43 booti386: karolherbst: pstate: http://hastebin.com/agavebulaz.pl
18:44 hermier: hi, I was forced to update to a gk208 (gt 730) after a 630 card failure, and now I see "kernel rejected pushbuf" errors
18:45 karolherbst: imirkin_: or maybe this graph issue us just the gpu being unstable on high engine clocks, and some cards just have high gr multipliers...
18:45 hermier: is there any way I can fix that issue, or help to fix it ?
18:45 imirkin_: hermier: there should be errors in dmesg
18:46 hermier: [ 1883.800338] nouveau 0000:01:00.0: akregator[2762]: Unknown handle 0x00000035
18:46 hermier: [ 1883.800341] nouveau 0000:01:00.0: akregator[2762]: validate_init                                                                                                                                                                            [ 1883.800342] nouveau 0000:01:00.0: akregato
18:46 imirkin_: odd.
18:46 imirkin_: do you have a weird libdrm version?
18:46 hermier: followed by: Unknown handle 0x0000002b
18:46 hermier: distro provided
18:47 imirkin_: which version? :)
18:47 hermier: 2.4.70, I'm looking if there is a patch in it
18:47 imirkin_: that's fine
18:47 imirkin_: i think some distro decided 2.4.60 was a good idea
18:47 imirkin_: which was basically the one release that was bad
18:48 booti386: imirkin_: Which firmware filename is the ctxsw?
18:48 imirkin_: booti386: sadly the names have changed
18:49 booti386: Oh :(
18:49 imirkin_: originally they were like nve6_4019[cd] etc
18:49 imirkin_: but now they have to be in some weird directory, and have to be called fecs/gpccs or so
18:49 imirkin_: i forget the details
18:49 imirkin_: i noted it down in some bug somewhere
18:49 hermier: libdrm is not patched
18:50 imirkin_: hermier: did you change what software you used? the other way of getting such errors is if your software uses concurrent GL contexts
18:50 booti386: Ok, and out of curiosity, where the rewritten fimwares are? :)
18:50 imirkin_: the nouveau ones?
18:50 hermier: I'm not sure, I received the card yesterday, and there was a massive distro update today
18:50 booti386: Yes
18:51 imirkin_: booti386: https://github.com/skeggsb/nouveau/tree/master/drm/nouveau/nvkm/engine/gr/fuc
18:51 hermier: but looking at the things, I think I have all new stable important packages
18:51 imirkin_: .fuc is the source, .fuc.h are the "compiled" versions
18:51 booti386: Oh, thank you :)
18:51 imirkin_: hermier: are you using KDE?
18:51 booti386: Too much subdirs :)
18:52 imirkin_: hermier: and/or QtWebEngine? (maybe all of qt5 is affected, i forget)
18:52 hermier: imirkin_: yes, this error is all new in akregator
18:52 hermier: yes, I think it uses QtWebEngine
18:52 imirkin_: yeah, a bunch of people have reported issues with that
18:52 hermier: is there a work around for it ?
18:52 imirkin_: i have a mesa branch
18:52 imirkin_: which may help
18:52 imirkin_: https://github.com/imirkin/mesa/commits/locking
18:53 imirkin_: but i won't be pushing that, i ended up not liking that approach, and it's incomplete. i need to rewrite like half of nouveau to deal with this properly =/
18:54 booti386: Err, don't I have already all of them? http://hastebin.com/xayahijuzu.coffeescript
18:54 hermier: imirkin_ ouch
18:55 hermier: this is odd because I never saw it on nv50, but it is visible with nvc0
18:55 imirkin_: booti386: yes. but they won't get loaded by newer versions of nouveau - like v4.3 and newer. (i think.)
18:55 booti386: Oh. Sad. :D
18:55 imirkin_: hermier: well, it's a timing issue. and new versions of things appear to trigger it a lot more.
18:56 hermier: ok
18:56 imirkin_: booti386: you need to rename them to the "new" names
18:56 imirkin_: booti386: and then boot with nouveau.config=NvGrUseFW=1
18:57 booti386: Sorry, but how do I determine the new names?
18:57 imirkin_: i left a comment about it in a bug somewhere
18:57 booti386: Yeah, thank you :D
18:57 imirkin_: you're welcome :)
18:58 imirkin_: https://bugs.freedesktop.org/show_bug.cgi?id=93629#c9
18:58 booti386: Found it, haha :D https://bugs.freedesktop.org/show_bug.cgi?id=93629
18:59 imirkin_: hermier: that said, people have reported that the branch in question should help in most situations
18:59 imirkin_: hermier: so if you need to get going right now, it's not such a bad place to start
19:00 hermier: imirkin_: I'm looking at the changes right now to see if we can do something more elegant for cheap
19:00 imirkin_: long story short - no.
19:00 imirkin_: at least not without *killing* perf
19:02 hermier: do you have some numbers on the impact ?
19:03 imirkin_: nope
19:04 imirkin_: anyways, the "simple" thing is to just create a locking midlayer that effectively prevents multiple access into the driver
19:05 hermier: simple but inefficient
19:06 hermier: I saw that Qt was disabling multithread on mesa. Is mesa suporting it officially?
19:11 hermier: imirkin_: forgot to say it but to add to the fun I have a dual head setup XD
19:17 imirkin_: hermier: afaik nouveau is the only driver with issues
19:27 booti386: imirkin_: Sadly, [ 1.732441] nouveau 0000:01:00.0: Direct firmware load for nvidia/gk106/fecs_inst.bin failed with error -2
19:27 imirkin_: did you put the file there?
19:27 imirkin_: (hint: you didn't)
19:45 hermier: imirkin_: when you say "i need to rewrite like half of nouveau to deal with this properly" you mean changing the way pushbuf is working right now ?
19:45 imirkin_: hermier: i mean not using libdrm_nouveau
19:45 imirkin_: and carefully thinking about how kicking works
19:46 imirkin_: and this would be a good opportunity to normalize the error handling situation in nouveau
19:46 imirkin_: (which is currently non-existent)
19:46 hermier: hmmm ok
19:46 hermier: and you lack motivation to do that, or man power ?
19:47 imirkin_: yes :)
19:47 hermier: wich one XD
19:47 imirkin_: the implication was that both
19:47 hermier: ok
19:47 imirkin_: thing is that it's kind of a big thing
19:47 imirkin_: which will require a concerted effort to complete
19:48 imirkin_: it's not some small incremental change
19:48 hermier: sure
19:48 imirkin_: also i don't have a super-clear picture yet of how everything is going to work
19:48 imirkin_: so i need to think about it some more
19:48 imirkin_: also the time i can put towards stuff like this is limited
19:49 imirkin_: and alllso i kinda want a ioctl to wait on a fence in the kernel
19:49 imirkin_: anyways, the upshot of all this is "if you want an open-source supported graphics driver, use amd"
19:49 imirkin_: [or intel]
19:50 hermier: well I never understood that libdrm, it is causing to much fracture in the code
19:50 imirkin_: the thing is that a lot of seemingly innocuous actions can cause a kick
19:50 imirkin_: (kick = submit pushbuf to kernel)
19:51 imirkin_: many of which aren't ones you want to put locks around
19:51 imirkin_: like nouveau_bo_wait()
19:51 hermier: imirkin_ I proposed some patches a while ago to nouveau (none where accepted thought), so I'm not a noob on how nvidia/nouveau works
19:52 imirkin_: yeah - i could tell that you have *some* concept of how things work
19:52 imirkin_: although i don't immediately recognize your handle
19:52 hermier: was more than 2 years ago
19:52 imirkin_: ah yeah, my memory really only goes back like 2 hours :)
19:53 hermier: bad things start to happens wiil reboot
19:53 imirkin_: wait - were you the one who sent patches for having > 165mhz tmds?
19:53 imirkin_: [on hdmi]
19:55 imirkin_: ah no. he was not.
19:57 hermier: imirkin_ I'm not sure, probably
19:57 hermier: but that card died last week XD
19:57 imirkin_: nah. that was someone else.
19:57 imirkin_: and much more recently.
19:57 imirkin_: looks like you were active in 2012
19:59 hermier: imirkin_ patch series are the 4 on the top only ?
20:00 imirkin_: hermier: yes. they probably won't apply on top of mesa-git HEAD
20:00 hermier: I'll put them on top of 12.0.3 so it should be good
20:00 imirkin_: if they apply, great. if not, i recommend just using my branch directly.
20:07 hermier: imirkin_: I'll go to console to apply the patch, I'll come back to report, probably tomorrow or so.
20:35 hermier: imirkin_: so far so good, will look for longer uptime ^^
20:35 imirkin_: cool
20:37 hermier: imirkin_: if you need me to test some code, I'm usually around (even if I'm not on that channel)
20:37 imirkin_: thanks
20:38 imirkin_: hermier: in case you want to get a bit more perf out of your GK208, reclocking should work now with a not-yet-in-linus-tree branch
20:38 imirkin_: (and if you have DDR3 it should work with upstream)
20:39 karolherbst: imirkin_: why is that DDR3 related?
20:40 imirkin_: karolherbst: dunno, coz my DDR3 GK208 works just fine? :)
20:40 imirkin_: while GDDR5 has more subtleties
20:40 karolherbst: :D
20:40 karolherbst: well, it isn't gddr5 related though
20:40 karolherbst: mainly just luck
20:42 karolherbst: but yeah, there are still some memory bits missing for reclocking, I doubt though it is gddr5 related at all
20:42 imirkin_: i guess you spoke too soon about stability?
20:42 karolherbst: isn't gddr5 related
20:43 hermier: imirkin_: it helped but it crashed even more harder XD
20:43 hermier: yes card as gddr5
20:43 karolherbst: hermier: do you reclock your card?
20:43 hermier: nope I don't use vanilla
20:43 hermier: I meant I use
20:44 hermier: I don't even know how to do that
20:48 hermier: but imirkin_ it helped in the sence that akregator didn't instantly crashed
20:50 imirkin_: just ... not enough :)
20:51 hermier: seems like I'll be forced to use nvidia, didn't had to since ~2010 XD
20:51 imirkin_: next time, buy amd
20:52 hermier: hmmmmm
20:52 hermier: no XD
22:24 dviola: hi
22:24 dviola: how do you know whether a program is using DRI2 or 3?
22:24 dviola: is it possible to know
22:24 karolherbst: dviola: LIBGL_DEBUG=verbose
22:24 dviola: thanks
22:25 karolherbst: it prints "libGL: Using DRI3 for screen 0" for me
22:27 dviola: that worked great, thanks
22:37 dviola: I get dri3 too
22:37 dviola: retroarch
22:41 dviola: I'm having a few issues with the intel drivers here
22:41 dviola: would nouveau work ok with glamor/modesetting?
22:42 karolherbst: dviola: it depends on the gpu, usually you should use modesetting only for maxwell based gpus
22:54 dviola: it would be slow on the older gpus?