05:35 tannerb3: imirkin, interestingly, now that I have switched back to X and Awesome from Sway, I get perfectly fine framerate on my 4k monitor, unlike with Sway
05:35 tannerb3: so sounds like nouveau does fine, but sway is doing something wrong?
05:45 gnarface: could it be the difference between whether compositing is on or off?
09:37 nk_: Hello, I am new to open source and wish to start contributing. I do programming in C,C++,Python and Java. Have very basic knowledge about coding for drivers.Can someone please suggest as to how I can start contributing to Nouveau?
09:38 karolherbst: nk_: are you a student?
09:39 nk_: karolherbst: Yes
09:39 karolherbst: nk_: uhh nice
09:39 karolherbst: nk_: then you could also do a GSoC or EVoC later. Just wanted to know to keep that in mind
09:40 nk_: karolherbst: Oh Yes, I have read about GSoC.
09:40 karolherbst: nk_: normally you should start with easy tasks and what interest you. The nouveau project covers many areas, not just OpenGL stuff.
09:40 karolherbst: nk_: we have a list of project ideas here: https://www.x.org/wiki/SummerOfCodeIdeas/
09:40 karolherbst: search for nouveau
09:40 karolherbst: there is also the link to the trello board
09:41 karolherbst: and there is also the nouveau wiki (with a few outdated pages): https://nouveau.freedesktop.org/wiki/
09:41 karolherbst: but I myself to prefer to find an interesting topic first and then try to figure out how to work on that
09:42 nk_: karolherbst: What kind of hardware is essential to start with?
09:43 karolherbst: nk_: depends on what you want to do, but a Nvidia GPU is kind of a must, doesn't really matter which one, because we try to support all of them
09:43 karolherbst: if you have issues with yours, fine, you can try to start with solving those ;0
09:43 karolherbst: * ;)
09:47 nk_: karolherbst: I checked out the problems put up on trello, in case I want to start contributing to a particular task how should I go about it? As in how will I get references regarding the topic or what has been already done?
09:47 karolherbst: nk_: ask here
09:47 nk_: Oh Okay!
09:50 nk_: And I have read on the internet that usually for developing drivers and other kernel modules people use a spare computer, is it essential? Since developing on my laptop I will have to change core configurations. What do you suggest?
09:50 karolherbst: nk_: well, you can work on your laptop. I do the same, but you can easliy mess up your X server and stuff like that
09:51 karolherbst: there are some ways to be able to unload nouveau while X is running (when using intel to display/render everything), but then you can't use external screens if they are connected to the nvidia GPU
09:51 karolherbst: usually having a second machine is less painfull
09:52 nk_: karolherbst: Oh alright. Thanks karolherbst!
10:28 pmoreau: I would say it also depends on what kind of development you are doing: if working on the kernel side, a second machine can be really helpful, but if it is within Mesa, one is enough as you can have two versions of Mesa living side by side.
10:31 karolherbst: pmoreau: well you still have fun with piglit in that case
11:19 sBBooky__sBammer: Yo_y got hacked
11:19 sBBooky__sBammer: Sl0w mode anti-sigyn fag: OFF: no sygin fag
11:19 sBBooky__sBammer: _ohm luvs pedo
11:19 sBBooky__sBammer: android is the bestest
11:19 sBBooky__sBammer: you eat too many dix eh ;)
11:19 sBBooky__sBammer: so, suck less cocks, plz
11:19 sBBooky__sBammer: TUU NE PEUX PASSSSS
11:19 sBBooky__sBammer: sBooky sBam!1!1!1!1!1!1!
11:19 sBBooky__sBammer: _Ω:
11:19 sBBooky__sBammer: hid:
11:19 sBBooky__sBammer: Nkk7_1 p0lice hacker
11:19 sBBooky__sBammer: eat thaat
11:19 sBBooky__sBammer: u suckk all
11:19 sBBooky__sBammer: pork_god!
11:19 sBBooky__sBammer: eat thaat
11:19 sBBooky__sBammer: u suckk all
11:19 sBBooky__sBammer: Nkk7_1 p0lice hacker
11:19 sBBooky__sBammer: eat thaat
11:19 sBBooky__sBammer: u suckk all
11:19 sBBooky__sBammer: pork_god!
11:19 sBBooky__sBammer: eat thaat
11:19 sBBooky__sBammer: u suckk all
11:19 sBBooky__sBammer: pork_god!
11:19 sBBooky__sBammer: Youu got hacked
11:19 sBBooky__sBammer: I am t00 pro
11:19 sBBooky__sBammer: android is th3 bestestest
11:19 sBBooky__sBammer: Nkk7_1 is the best developer
11:19 sBBooky__sBammer: Nkk7_1 is the proest
11:19 sBBooky__sBammer: Nkk7_1 is the p0lice hacker
11:19 sBBooky__sBammer: eat thaaat
11:19 sBBooky__sBammer: you suckk all
11:19 sBBooky__sBammer: eat thaaat
11:20 sBBooky__sBammer: Yo_y got hacked
11:20 sBBooky__sBammer: Sl0w mode anti-sigyn fag: OFF: no sygin fag
11:20 sBBooky__sBammer: _ohm luvs pedo
11:20 sBBooky__sBammer: android is the bestest
11:20 sBBooky__sBammer: you eat too many dix eh ;)
11:20 sBBooky__sBammer: so, suck less cocks, plz
11:20 sBBooky__sBammer: TUU NE PEUX PASSSSS
11:20 sBBooky__sBammer: sBooky sBam!1!1!1!1!1!1!
11:20 sBBooky__sBammer: _Ω:
11:20 sBBooky__sBammer: hid:
11:20 sBBooky__sBammer: Nkk7_1 p0lice hacker
11:20 sBBooky__sBammer: eat thaat
11:20 sBBooky__sBammer: u suckk all
11:20 sBBooky__sBammer: pork_god!
11:20 sBBooky__sBammer: eat thaat
11:20 Asu: there are people entertained doing that?
11:20 Asu: oh well...
11:21 karolherbst: Asu: bots?
11:21 Asu: spambots
11:22 spooky__sBammer: Yo_y got hacked
11:22 spooky__sBammer: Sl0w mode anti-sigyn fag: OFF: no sygin fag
11:22 spooky__sBammer: _ohm luvs pedo
11:22 spooky__sBammer: android is the bestest
11:23 spooky__sBammer: Yo_y got hacked
11:23 spooky__sBammer: Sl0w mode anti-sigyn fag: ON: no sygin fag
11:24 spooky__sBammer: _ohm luvs pedo
11:24 spooky__sBammer: android is the bestest
11:25 pmoreau: sigh
11:26 spooky__sBammer: Yo_y got hhacked
11:26 spooky__sBammer: Sl0w mode anti-sigyn fag: ON: no sygin fag
11:30 spooky__sBammer: Yo_y got hhacked
11:31 bruzzo: mhh
11:33 spooky__sBammer: Yo_y got hhacked
11:34 spooky__sBammer: Yo_y got hhacked
11:36 Fuchs: mwk: we are k-lining these as well, so feel free to remove the most recent ones that were on IPs to keep your banlist small
11:36 mwk: ack
11:36 Fuchs: in addition to that, the one actual account monitoring the spam got k-lined along, keep an eye out for new ones of these
11:36 Fuchs: (hi, sorry for jumping in)
11:37 mwk: yeah, figured that
11:37 mwk: thanks
11:37 Fuchs: welcome, sorry for the noise
15:41 karolherbst1: anybody tested nouveau on 4.14-rc7? it seems to fail on gr init for gm107
15:45 tobijk: karolherbst1: that is a non deterministic problem, it happens very rarely for me, mostly it works
15:46 karolherbst1: tobijk: well, so it doesn't work
15:46 tobijk: but i have a gp106, so maybe i should consider myself lucky :D
15:47 tobijk: karolherbst: btw do you have recent piglit results around? i wanted to check my own results again some pther generation
15:48 karolherbst: tobijk: sadly no. I can give them to you later
15:48 tobijk: karolherbst: np, no hurry :)
15:50 tobijk: karolherbst: coming back to the gr init problem: does it happen after a cold start? it mostly happens when i reboot my system, so it seems we are driving the card a bit wrong
15:51 karolherbst: tobijk: mhh interesting
15:51 karolherbst: I will try that
15:52 karolherbst: will need to reboot quite often now, so I won't be online, I will read messages later
20:07 Lyude: Do the tegra codenames line up with the full-gpu generation they're based on? (e.g., would gk20a be based off of kepler because of the gk)
20:08 imirkin_: yes
20:08 imirkin_: GK20A = kepler(2), GM20B = maxwell(2)
20:08 Lyude: sweet, that explains this register then
20:08 imirkin_: i think there's a pascal one in theory? don't know the chip name though
20:08 Lyude: i think there is, it's come up on my register scans a couple times
20:09 imirkin_: usually when the last position is a letter, it's an IGP
20:09 imirkin_: e.g. NV1A/NV1F/NV2A/NV4B/NV4C/NV4E/NVAA/NVAC/NVAF
20:10 imirkin_: [that might be an exhaustive list... i think there might be a NV0A, but i don't think there are still any in existence, if there ever were]
20:11 imirkin_: i guess NV4A isn't an IGP. but it's a weird chip.
20:11 imirkin_: (it's NV44 but AGP/PCI)
20:24 Lyude: I or someone else should write a more proper register scanner sometime then these shell scripts i've been using
20:25 karolherbst: Lyude: write a trello card!
20:25 Lyude: oh right, the future is now, I forget that sometimes
20:25 karolherbst: but I also have some funky scripts, which would be better suited to be C programs in the end
20:25 Lyude: or python
20:25 karolherbst: well, nva is in C
20:26 Lyude: ah right
20:26 Lyude: would be kind of nice if we had a python frontend as well (which would be pretty easy to do)
20:26 Lyude: is that another card
20:26 karolherbst: Lyude: do you know swig?
20:26 karolherbst: Lyude: http://www.swig.org/
20:26 Lyude: i've seen it but never tried it
20:27 karolherbst: well I used it once
20:27 karolherbst: it's nice
20:27 imirkin_: there is a python frontend for rnn
20:27 imirkin_: i thought nva too? not sure
20:27 imirkin_: swig's pretty easy if you don't do anything complicated
20:27 karolherbst: Lyude: you basically scan your header files and then python modules pop out
20:27 karolherbst: imirkin_: I am sure it is fine for nva
20:27 imirkin_: yep
20:28 karolherbst: and swig just translates the API, so all the internal stuff can stay internal
20:31 karolherbst: ohh nice, there is nothing we would really need to share, just functions
20:31 karolherbst: we could leave everything out involving struct nva_regspace
20:31 karolherbst: for now
20:39 Lyude: https://trello.com/c/4xhrtohN/201-write-a-proper-mmio-and-probably-others-scanning-tool-for-searching-tracesavailable-source
20:41 karolherbst: Lyude: part two could become really difficult. In a few subdev we deal with base + offset values
20:42 karolherbst: there are also some macros within nvidia source to deal with offsets
20:45 Lyude: karolherbst: yeah I already figured since we don't really have anything we can use to get a consistent data source. that being said I should probably note on there it's really not expected to be 100% accurate at least in the sense that it might not notice something
20:45 Lyude: my highest expectations would just be text search, if we find a text match figure out what info we can get from the surrounding context and don't bother with anything we can't get
20:45 karolherbst: Lyude: I am thinking if maybe clang/llvm has good enough tools to find such occurances... most likely not, but maybe we could write a tool for this.
20:45 Lyude: i know at least llvm would most definitely be able to help with something like this
20:46 Lyude: ....actually
20:46 Lyude: karolherbst: that solves the macro problem entirely, we could just search through the AST for the register value instead of a text search
20:46 karolherbst: But it starts to sounds like an effort not really worth spending time on, because if it isn't perfect, you still search with the old methods. Allthough I am quite sure it will help within enough cases if we just cover the basics
20:47 karolherbst: Lyude: yeah, that would be fine, just thinking about how much of an effort that would be
20:47 Lyude: yeah, my main reason for suggesting it was that it gets kind of confusing juggling around 10 terminals, all of which are doing searches through repos and traces :P
20:47 Lyude: true
20:47 Lyude: another reeason why I figured it might be nice just to use python since it'd probably save a good bit of time
20:47 karolherbst: yeah
20:47 karolherbst: I am all for the first part
20:48 karolherbst: this would be a big help already and I think quite easy to do
20:48 karolherbst: thing is, for the demmio stuff, I think the bottlenecks are somewhere else, not because a script calls some binaries
20:49 karolherbst: but more like if we scan multiple times
20:49 karolherbst: so an application with a state saving all the results might make sense
20:49 karolherbst: so that we don't rescan all the time
20:49 Lyude: the main bottlenecks I see are just the fact it can't do scans multithreaded
20:50 karolherbst: maybe even something like scan the entire repository and provide real time search
20:50 karolherbst: maybe even with a disc cache
20:50 Lyude: there might also be another way to optimize further then that, but multithreading would be a great start
20:50 Lyude: (that is why I have so many terminals open, each one takes a core!)
20:50 karolherbst: ohh, I see
20:51 karolherbst: yeah, I think what we want is a parser which builds a databse/search cache/whatever and then a frontend to earch through that cache
20:51 karolherbst: *search
20:51 Lyude: yes, that would be perfect
20:51 karolherbst: mhh
20:51 Lyude: heck, this would also be a good project for someone with no hw knowledge
20:51 karolherbst: yeah
20:52 karolherbst: maybe we could just use lucene for stuff like that
20:53 karolherbst: just take the demmio output and create indecies for this
20:54 karolherbst: there is even a python frontend for lucene
20:54 karolherbst: but maybe there also exists non java software
21:10 Lyude: ah i see, so it looks like the slcg stuff is just for kepler2. either way, still found some new regs :)
22:03 Lyude: Hey, has anyone been seeing lockups on kepler1.5 (GTX 780)
22:03 karolherbst: Lyude: let me guess, 4.14?
22:04 Lyude: yes
22:04 imirkin_: that's kepler2 btw
22:04 karolherbst: well, not on kepler 1.5 but on maxwell1
22:04 Lyude: so it is kepler2, ok
22:04 karolherbst: Lyude: I will bisect that stuff tomorrow
22:04 imirkin_: (assuming it's a GK110)
22:04 Lyude: karolherbst: alright; I mean I can't really continue until it's fixed anyway so I can start bisecting. I'll leave you my bisect log before I go
22:04 karolherbst: Lyude: for example for me a rmmod or suspending basically kills the driver
22:05 Lyude: imirkin_: correct
22:05 karolherbst: Lyude: how is it caused for you?
22:05 Lyude: karolherbst: booting my system
22:05 karolherbst: Lyude: and I assume it works on 4.13?
22:05 Lyude: that's about it. it crashes as early as fbcon
22:05 Lyude: karolherbst: not sure about that, building a 4.13 kernel right now so I can figure that out
22:05 karolherbst: mhh okay, mine was simply on a optimus system
22:05 karolherbst: so I can't say much about fbcon
22:06 Lyude: it's definitely more then fbcon that's breaking things
22:06 karolherbst: I think it has something to do with gr init
22:06 Lyude: skeggsb: btw ^
22:08 karolherbst: Lyude: by any chance, did you have a GPU with memory reclocking issues on kepler2 or newer?
22:08 Lyude: I haven't tried reclocking on kepler2 at all
22:09 Lyude: was just about to start slcg work on k2 though so I probably will very soon
22:09 karolherbst: ohh, okay
22:09 karolherbst: I just pushed out a patch yesterday which may fix some weirdness
22:20 Lyude: ok yeah, this is definitely a regression
22:20 Lyude:begins bisection
22:20 karolherbst: thanks!
23:14 Lyude: ugh, damn. karolherbst: looks like I screwed up the bisect somehow, or something outside of nouveau broke it
23:15 Lyude: so no bisect log :(, feel free to give reproducing it a shot tomorrow and see if you can bisect it
23:15 imirkin_: did it land on the merge commit?
23:15 karolherbst: Lyude: that was my thought as well
23:16 karolherbst: I also tried to bisect and gave up at some point
23:19 karolherbst: Lyude: well I will give it another shot tomorrow and don't use drm-next for that
23:39 pmoreau: Duh, it looks like I am still missing some alignment cases. :-(
23:45 pmoreau: Ah, of course: built-in data types is aligned to the next power of 2. Should be manageable to add.
23:51 imirkin_: skeggsb: is this another way of saying chip != 0x34? https://hastebin.com/gigomebuma.lisp
23:51 skeggsb: *shrugs* i pretend that code doesn't exist
23:52 imirkin_: yeah, best not to think about it
23:52 skeggsb: i have basically zero idea about how pre-nv50 modesetting works, so it's best i don't touch it lest i screw something i can't fix
23:53 imirkin_: it's unfortunate that someone tried to "clean up" the timing/arb logic
23:53 imirkin_: coz it no longer matches xf86-video-nv
23:53 skeggsb: feel free to re-port it from nv, if you're brave enough :P
23:53 imirkin_: and there's no way to know if it's different for a reason, or coz "well, that code doesn't seem important"
23:53 imirkin_: or whatever was in the head of the person chopping the code away
23:54 skeggsb: might be an idea to figure out who that was, and contact them perhaps
23:54 imirkin_: ahuillet? :)
23:54 skeggsb: or, was it him?
23:54 imirkin_: or marcheu ... who knows
23:54 imirkin_: either way, i doubt they'd remember
23:55 imirkin_: and it's not like i could test any of it
23:55 imirkin_: beyond the small handful of devices i have on hand
23:55 imirkin_: and it's *so not worth it*