00:00japele: yea, I must send it at mmio dot dumps at gmail dot com or somewhere else?
00:00imirkin_: that works
00:16oghr: are there any plans on making nv40 working with recent desktop environments like kde or gnome? Nouveau is completely broken on such usecases
00:20imirkin_: patches welcome?
00:21imirkin_: someone with the relevant hw has to debug the various issues...
00:21imirkin_: [and i have absolutely no interest in installing a "recent desktop environment" on my system]
00:23oghr: imirkin_: patches are not been merged when send! https://lists.freedesktop.org/archives/nouveau/2016-October/026105.html
00:24imirkin_: firstly, that's a kernel patch (not my department...)
00:24imirkin_: secondly, i have no clue why it wasn't merged.
00:25imirkin_: but patches can easily fall through the cracks
00:25imirkin_: ping it, etc
00:25oghr: i am pinging it now...
00:26imirkin_: via email.
00:27oghr: have to be send to email@example.com ?
00:30oghr: could you send it to the list? Maybe then it would be finally merged
00:33oghr: thanks for https://lists.freedesktop.org/archives/nouveau/2017-February/027290.html
00:36skeggsb: imirkin_: pushed
00:38imirkin_: thanks :)
00:45oghr: skeggsb: thanks
01:06oghr: imirkin: what priority does this bug here have? https://www.phoronix.com/scan.php?page=news_item&px=Nouveau-Civ6-Mesa-17.1
01:11oghr_: sorry, lost connection. imirkin, have i missed some answer?
01:18oghr: what is the difference in code/functionality inside of a family? Is there any? For example geforce 8400 vs gt240. When something (opengl, ...) didnt work on 8400 then this mean automaticly that it also didnt work on gt240?
01:20imirkin: gt240 is a DX10.1 GPU, while 8400 is either a DX10 or DX10.1 GPU, depending on specifics
01:21imirkin: they are largely identical, although there are some differences
01:22imirkin: oghr: civ6 has 0 priority. first off, i don't have a copy of the game or a trace that leads to the problem, so it's entirely inactionable. secondly, i'm almost certain that the issue is due to multithreading deficiencies in nouveau, which i am not working on addressing.
01:23imirkin: that said, i'd be happy to help someone debug it if they were to become interested in it.
01:24oghr: so first wait for the multithreading fix in mesa, then try to run the game again and then if still there run advanced debugging?
01:24imirkin: oghr: not necessarily in that order. one could debug the game and determine that it is, indeed, the multithreading thing that's taking it down
01:25imirkin: or, conversely, determine that it's a totally separate issue
01:25oghr: imirkin: would you like to help fixing nv40 when debugging logs are been delivered?
01:26imirkin: oghr: i'd be happy to perform some analysis, and assist an interested party. however the heavy lifting would most likely have to be done by someone other than me
01:26imirkin: right now i don't have a nv4x plugged in, only a nv3x
01:26imirkin: (and some later gpu's)
01:27imirkin: and i got outbid for a cheap pci nv4x. o well.
01:27imirkin: (i have some pcie ones, but pcie slots are a precious commodity)
01:27oghr: does nv3x work with modern desktop environment?
01:27imirkin: thoroughly doubtful
01:27imirkin: esp as it doesn't expose GL 2.x or even GLES 2
01:28imirkin: further frustrating matters is that i have absolutely no interest in such environments
01:30oghr: in what environment should bugs happen to wakeup your interests?
01:30imirkin: apitrace ;)
01:31imirkin: if you can find me a trace that triggers the problem, then i'm *much* more likely to debug the problem than under any other conditions
01:33oghr: so anything i find using nv40 and i could trace with https://github.com/apitrace/apitrace would wakeup your interest? :)
01:34imirkin: well, if the issue reproduces when you replay the trace.
01:34imirkin: that might even be enough for me to plug a nv4x in
01:35dboyan: hello again, I just wonder if there is some "starter project" to get familiar with nouveau, or learn some basic approaches?
01:35imirkin: dboyan: what GPU do you have?
01:36imirkin: and how long would you hope a starter project to take?
01:36oghr: of course, if the issue is reproducible. Not some strange problem that appears once and i never get it again... :)
01:36imirkin: oghr: or some issue that always happens, but never shows up in traces :) been known to happen too
01:37oghr: dboyan: if you can build linux live images: there is a need to take mmio-traces of binary drivers just by running a live-linux. At the moment its extremely frustrating to run a mmio-trace for a "normal" user
01:39imirkin: oghr: http://nouveau.pmoreau.org/ might help with the live image situation... i forget if it includes blob though. i think it does.
01:39dboyan: imirkin, A 740M (GK208), and a new 1050 mobile. I still have a week before the school season begin, but I think I will have some spare time even after that
01:40imirkin: hmmmm let's see...
01:41oghr: dboyan: forget the crappy 1050. if you can return the device, return it quickly
01:41imirkin: you won't be able to do much with the 1050 until the firmware situation gets sorted
01:42oghr: dboyan: and after the firmware is been released, there would still be a huge force for requiring closed source binary blobs that cant be optimized/touched by the developers here because of signature verification
01:43oghr: dboyan: so this hardware would always be a freedom-crap if nothing something really unexpected happens like a error in the signature verfification of the hardware or something else like that
01:43dboyan: Yeah, I guess the 740M will be a more helpful one
01:44oghr: dboyan: https://www.phoronix.com/scan.php?page=news_item&px=NVIDIA-Unfriendly-OSS-Hardware
01:45skeggsb: oghr: as i told you in PM already, NVIDIA is no more "evil" than AMD there.. both require fw blobs
01:45skeggsb: the *only* difference is AMD release them a lot better...
01:46imirkin: skeggsb: s/a lot better//
01:46imirkin: as far as i'm concerned, NVIDIA's still not done releasing them for GM20x
01:47skeggsb: yes, well, that's true enough i suppose
01:48oghr: skeggsb: yes. i know. i would personaly recommend at least geforce 780 ti / Titan Z and tell anyone to not buy the new hardware that require signed software - because i dont like beeing forced to use some unknown signed software
01:48imirkin: i'm not pissed at NVIDIA for forcing us to use their firmware. i'm pissed coz they won't release the firmware in a redistributable fashion.
01:48skeggsb: imirkin: agreed
01:54dboyan: imirkin, I once tried to "optimize" atomic operation handling on kepler, but after some careful thinking, some ideas were wrong, some others I don't think I can implement...
01:58imirkin: dboyan: you could debug the issue you were having ;)
01:59imirkin: i'm not seeing any *great* starter-type projects for kepler...
01:59imirkin: you could ... i dunno. i'm part-way done implementing ARB_shader_ballot and ARB_shader_clock for gallium. you could take that over if you like
02:03dboyan: imirkin, I personally think compiler work may be easier.
02:03imirkin: ok, well in that case, there are probably a few items
02:03imirkin: i gtg, but i'll get back to you with some options
02:03dboyan: I tried to debug the bug I found. I was somewhat overwhelmed by the mmt trace.
02:05dboyan: also I'm not that familiar with the handling of nouveau, especially when the handling is not same with the blob driver (the nouveau mmt trace is not annotated)
03:24nv40-bugs: this was easy. How to crash the whole system because nv40 card is been inserted without the need of using gnome or plasma:
03:25nv40-bugs: download and install latest lubuntu with latest kernel: http://cdimage.ubuntu.com/lubuntu/daily-live/current/zesty-desktop-amd64.iso
03:25nv40-bugs: run apt-get install supertuxkart
03:25nv40-bugs: start supertuxkart
03:26nv40-bugs: done. system crashes completely. it have to be killed by turning elecrtricity off.
03:26nv40-bugs: when nv40 card is been removed and some other card is been inserted supertuxkart starts just fine
03:43nv40-bugs: imirkin: https://pastebin.mozilla.org/8978789
03:44nv40-bugs: should i still make a huge tracefile with apitrace?
17:17imirkin: pmoreau: i think your auto-builder's broken again
18:55pmoreau: imirkin: Yeah, some mismath between Ben's out-of-tree and a version of drm-next
21:55karolherbst: mupuf: do you think it is a good or bad idea to tell the PMU about the currently set clocks?
21:58Echelon9: @karolherbst: I updated the continuous integration PR on GitHub - some new developments commented within for you
22:01imirkin: Echelon9: fyi the whole "@" thing reads very jarringly over irc. definitely not the convention.
22:07karolherbst: Echelon9: I doubt the python3/cython3 thing works this way
22:07karolherbst: Echelon9: I have no find cython3 or python3 cmake module
22:07karolherbst: and they aren't part of the official cmake distribution afaik
22:09karolherbst: mhh od
22:09karolherbst: I was mistaken
22:09karolherbst: I have no cython3 binary
22:09Echelon9: I'm fairly sure cython / cython3 is being brought down as a direct Debian package
22:10karolherbst: there is no cython3 really
22:10karolherbst: it's just cython
22:10Echelon9: Perhaps put another way, are you locally able to use nvapy without cython3 (and cython only)?
22:10karolherbst: and it supports python2 and python3
22:11Echelon9: Yes, given envytools requires python3+ I thought better to be explicit that cython3 is the dependency
22:11karolherbst: there is no cython3
22:12karolherbst: we don't do whatever debian does, for upstream, there is no cython3
22:12Echelon9: Okay, I'll amend the commit so it just refers to cython
22:13karolherbst: maybe there is a way to check if cython supports python3
22:13karolherbst: Echelon9: is /usr/bin/cython3 by any chance a wrapper around /usr/bin/cython -3 ?
22:16Echelon9: This is what it is for me locally: http://pastebin.com/hxc73eSx
22:17karolherbst: I see
22:19Echelon9: I presume you locally just have /usr/bin/cython?
22:19Echelon9: And that nvapy works for you with cython and python3+
22:19karolherbst: which either runs under python2 or python3 depending on what python is my default one
22:20karolherbst: and compiles against python2 or python3 with -2 or -3
22:20karolherbst: so I could use cython running under python3 to compile against python2
22:20karolherbst: and with your change, cython isn't found anymore for me
22:21karolherbst: cmake is just silly if it comes to python though :/
22:21Echelon9: OK, I'll revert the change from cython -> cython3 as it's all a bit abstract trying to target Travis-CI images when cython-whatever is still blacklisted
22:21Echelon9: And i don't want to make life hard for you and other devs
22:21karolherbst: I would only add the travis file within this PR
22:21karolherbst: without changing any cmake things
22:21karolherbst: because otherwise we will again run into discussions like this