00:33 imirkin: hakzsam: if you want a bug to look into - have a look at the updated QBO tests + 64-bit results -- my macro is apparently bogus - i don't think query_sequence can be used like that, and it can't be used at all to write a 64-bit value based on my experiments. not entirely sure how to fix it... probably p2mf.
00:34 imirkin: it may make sense to create a second macro that handles the 64-bit version of it. right now the macro i wrote figures it out based on whether there's a limit or not, but that's a bit dumb.
00:35 imirkin: and it has no way of dealing with the limit (like for occlusion predicate) + 64-bit write.
00:36 imirkin: i believe the current situation is that it writes both 32-bit words to the same location
02:02 imirkin: Riastradh: ping on the netbsd + nouveau thing :) [i'm gonna keep pinging you until you either tell me to stop, or produce the doc]
02:11 iterati: hi, so I am building https://github.com/karolherbst/nouveau/tree/stable_reclocking_kepler_v6 for 4.8.2
02:11 iterati: I get: nouveau: loading out-of-tree module taints kernel.
02:11 iterati: and screen blinking... I wait. Then I can manually load it
02:12 iterati: also nouveau: Unknown symbol led_classdev_register (err 0)
02:12 iterati: how to fix that ?
02:12 imirkin: did you use 'insmod' or 'modprobe'?
02:12 iterati: imirkin: modprobe after it failed to load automatically
02:12 imirkin: well, sounds like you built it against a .config that differs from your running kernel's
02:13 Riastradh: imirkin: No problem. Remind me what format you want for the document? A nouveau.freedesktop.org wiki page?
02:13 imirkin: or perhaps you forgot to run 'depmod -a'?
02:13 imirkin: Riastradh: any format that i can use to create a wiki page out of
02:13 Riastradh: OK.
02:13 imirkin: i'm happy to reformat. i just can't create the content :)
02:14 iterati: imirkin: I think that's it. Thanks... let's try and reboot then...
02:14 imirkin: Riastradh: and the intention is that this should take you no longer than 15 mins. if it's taking you longer, you're probably going into too much detail
02:15 imirkin: Riastradh: the audience should be competent netbsd users who aren't necessarily up on all the latest hotness wrt kms, etc
02:15 Riastradh: imirkin: I'm not writing a long document. I just need to remind myself what the state of affairs is, for each relevant release branch.
02:16 Riastradh: I think it is included in the source but disabled by default in all NetBSD 7 branches and enabled by default in HEAD, but I need to double-check.
02:16 imirkin: k. for the record, i am not a competent netbsd user :) but i know one!
02:16 Riastradh: So it just takes a little focussed time to write down for myself all the questions I intend to answer, and then find their answers.
02:16 Riastradh: s/find/confirm/1
02:17 imirkin: sounds reasonable
02:38 Riastradh: imirkin: How shall I send it to you?
02:50 imirkin: Riastradh: email is fine, or pastebin or whatever
02:50 Riastradh: imirkin: Address?
02:50 imirkin: see PM
03:25 Riastradh: *poof*
03:27 imirkin: added a sentence to the first paragraph
04:05 imirkin: the troll doesn't like getting called out on his trolling...
04:08 skeggsb: i particularly love the innocent "whatever do you mean, i wasn't being a wanker *winkwinknudgenudge*"
04:27 airlied: imirkin: GL_RGBA4 FBO incomplete in one CTS
04:27 airlied: but it expects it to complete
04:28 imirkin: airlied: if it's a texture attachment, CTS is wrong, at least according to some comments on the list
04:29 imirkin: airlied: if it's a renderbuffer attachment, i'll eat my proverbial hat - i think that should work ;)
04:30 airlied: imirkin: cool, I've no idea, just trawling the fail logs
04:31 airlied: just going to compare radeonsi/nve7 once they both finish
04:32 imirkin: cool thanks
07:18 karolherbst: silly mobile mail client :/ sorry for those HTML emails...
07:25 skeggsb: karolherbst: hehe, i got into trouble by my bf the other day.. he asked "why do you reply to emails the way you do [responsing inline instead of top-posting]? and your email program strips out the html"...
07:25 skeggsb: karolherbst: my only response was "... because you do it wrong" :P
07:28 karolherbst: :D
07:28 karolherbst: it is crazy, but the mobile gmail client has no setting to switch the mail to plain text...
07:29 karolherbst: I broke gmail on my old phone, no idea how I did that, so I never used it
07:30 karolherbst: skeggsb: by the way, did you saw my mail about those unknown memory reclocking bits?
07:31 skeggsb: i did, but i haven't had a chance to look at the vbios or anything
07:31 karolherbst: I see
09:51 karolherbst: nice, 33c3 ticket sale date is public: 7th, 19th and 25th of november :) I will try to get a ticket, if anybody of you want to come, remember those dates :p
09:52 Jaga: you might want to be fast too ;)
09:52 Jaga: lots of tickets already went in presales to former angels and associations/group-orders
10:00 karolherbst: uhh I live in hamburg
10:00 karolherbst: I would get additional days off just for that
10:02 Jaga: tickets are sold only online afaik
10:02 Jaga: presales were online too
10:02 Jaga: one of the reasons why they go fast, the event is international afterall
10:29 karolherbst: Jaga: yeah, I chcked like daily if the dates are online :D
10:30 Jaga: good ;)
10:31 karolherbst: the thing is, because my company is in hamburg, I get additional days off besides my days guaranteed through law and contract :) which is somewhat nice, because that means 5 more paid days off :D
10:31 karolherbst: (law says 10 days in two years for educational stuff)
10:31 karolherbst: (in addition)
10:34 Jaga: free paid holiday days are always nice
10:34 karolherbst: uhh, size of company seems important though
10:34 karolherbst: .. we are only.. 7 :D
10:34 karolherbst: well, maybe I am lucky, maybe not
10:34 karolherbst: just saw today, that there is the possibility
10:35 Jaga: :)
10:35 karolherbst: still have to get that nouveau shirt for 33c3 :O
10:35 karolherbst: or I make one there, should be possible
10:37 karolherbst: Jaga: by the way, did you see the "reing tutorial" I began to write? Didn't finish it yet, but I think it is already somewhat usefull
10:37 karolherbst: would be nice to get some nouveau unrelated people to take a look and point to non understandable parts or where more details are needed
10:38 Jaga: no, I haven't noticed
10:38 Jaga: can you throw me a link
10:41 karolherbst: Jaga: https://gist.github.com/karolherbst/4341e3c33b85640eaaa56ff69a094713
10:41 karolherbst: the vbios part is the one I cared most for now
10:46 Jaga: karolherbst: so you run nvbios giving it a file vbios.rom as argument? where is that file from? is that what you're referring to in the INFO above?
10:46 Jaga: karolherbst: or is it reading from your graphics card?
10:48 Jaga: what does POST mean in nvafakebios? it is kinda explained how to do it, but it's not explained what it really does, so it sounds like black magic to me
10:48 Jaga: you don't specify what 0x619f04 is - is it an address on the card??
10:49 Jaga: also, the formula (no r at the end in english) splits into 2 lines on my display, would be maybe better in its own separate line
10:49 karolherbst: you should just leave those comment at the gist directly, so that they are there and I can work on those later
10:50 karolherbst: they are all good, don't want to forget about these
10:50 Jaga: eh, my github account situation is a bit messy, so it's actually easier for me to type in here
10:51 karolherbst: k
10:52 karolherbst: funny, didn't write how to extract the vbios
10:52 karolherbst: so obvious
10:52 Jaga: also, you might want to put commands that other commands refer to earlier, so modprobe earlier than nvafakebios for example
10:52 karolherbst: POST is like the bios POST
10:53 Jaga: I never heard that graphic cards did their own POST, you might want to add a note on that
10:53 karolherbst: yeah, I am pretty much unsure about the right order anyway, so I didn't put enough thoughts into this. The text is more important for now
10:53 karolherbst: well
10:53 karolherbst: the bios usually does
10:53 karolherbst: but the driver also can run those POST scripts, which are in the vbios
10:54 karolherbst: it's a messy situation, cause for optimus the gpu needs to be posted whenever it is resumed…, at least I think we do it or have to do it
10:55 Jaga: btw you might want to keep strictly to calling graphic card bios "vbios" always, you seem to use them interchangeably so it might be confusing
10:56 Jaga: also "Nouveau Userspace Library" is in the "Tools" section which till then listed just shell commands, might be also confusing
10:56 karolherbst: it's a different thing
10:57 karolherbst: the nouveau userpsace library is building the kernel module as a .so file and let's you write applications against this
10:57 karolherbst: the nva* tools don't use it
10:58 karolherbst: anyway, my idea was to split between tools you always use and tools you only use for that specific task
10:58 karolherbst: no idea if that makes sense
10:58 karolherbst: as I said, didn't thought much about the order
10:59 mezo: hey, is there way to fix the "overbrightness" in rocket league? http://imgur.com/a/MisRf
10:59 Jaga: karolherbst: btw "adviced" could be better replaced with "recommended"
11:00 Jaga: karolherbst: also, do you have an example of a script that should be written with envytools?
11:01 Jaga: also, in sentence "Look there for information related to power management, connectors, clocks, memory, init scripts, gpios, external devices and many other things." - you might want to add what tool can be used for finding that info
11:01 karolherbst: Jaga: I once wrote a script which parses the entire PLL configuration of the memory clock.. just a bunch of nvapeeks and python stuff
11:01 karolherbst: so, dunno, was a bad choice anyway
11:02 Jaga: karolherbst: I'm asking cause when you ask somebody to write a script in a new tool, it makes it much easier for them to either link to an example or to provide one
11:02 karolherbst: the point is more, that you write a script instead of executing commands manually
11:02 karolherbst: ohh no, I meant bash scripts
11:02 Jaga: well mentioning some tool sounds like "here, have this tool you're never heard of before, start coding a script in it now"
11:02 karolherbst: instead of executing commands yourself, write a script ;)
11:02 karolherbst: more like this
11:02 Jaga: then that imo could use clarifying
11:03 karolherbst: k
11:03 Jaga: what script, what should be in it etc
11:03 Jaga: it's good to put examples :)
11:03 karolherbst: well, the script is free to contain everything, that's a little problem
11:03 karolherbst: cause it highly depends on what you RE
11:03 karolherbst: but yeah, will try to write a simple example
11:04 Jaga: great :)
11:04 karolherbst: thanks for the comments
11:05 Jaga: np :)
11:05 Jaga: it's the least I can do
12:52 karolherbst: anyway, if somebody decides to also come to 33c3, I would even think about sleeping in the same hostel/hotel, cause I would skip my 1h per direction train ride :D
12:52 karolherbst: but then again, who needs sleep anyway!
12:53 karolherbst: I expect that there will be something going on 24h a day!
13:34 Jaga: karolherbst: official program ends 24:00-ish, so what you can do is hang out with people, go to small parties, and stuff ran by groups
13:34 Jaga: well, there's these official parties for 2 days iirc
13:42 karolherbst: uhh yeah, 24 is quite late for me, my last chance to get home is like 1am … and most of the time I wouldn't want to get home so early
13:43 karolherbst: mupuf: as you see, even if you would come, I couldn't bring it over my heart to provide you with such a crappy hosting service :p
13:43 karolherbst: not after what you have done for us
13:44 karolherbst: mupuf: by the way, do you still plan to come or was it just a nice idea and you wouldn't have time anymore?
13:45 mezo: karolherbst: is dynamic reclocking safe to use? so far im rly suprised by it :)
13:48 karolherbst: well
13:48 karolherbst: memory doesn't reclock, but generally, yes
13:49 mezo: hm, but performance is fine, what i can tell so far.
13:49 karolherbst: well, memory is max clock
13:49 karolherbst: I somehow plan to land the stuff soon enough and have a switch to turn it on
13:49 karolherbst: need to work on the firmware more though
13:50 mezo: would be great
13:53 karolherbst: there are also some perf problems if the load is changing too fast, but in avarage it is only slow for 0.05seconds
13:53 karolherbst: so you wouldn't really notice
13:53 mezo: how much does the power consumption increase, when the memory is always max?
13:55 mupuf: karolherbst: well, it still is a viable option for me
14:00 karolherbst: k
14:00 karolherbst: note the dates, cause it is the only chance to get tickets
14:00 karolherbst: mupuf: https://events.ccc.de/2016/10/16/33c3-tickets-status-und-freier-verkauf/
14:00 karolherbst: english below
14:01 mupuf: let's cross fingers
14:01 karolherbst: I will so miss all three of those...
14:02 mupuf: loool, well, we can both try and buy two tickets
14:02 mupuf: first one who gets through ... tells the other one
14:24 karolherbst: :D
14:24 karolherbst: and if we buy two at the same time?
14:25 karolherbst: well, then we have 4! and are able to bring more people to the party!
14:25 mupuf: hehe
14:25 karolherbst: anyway, it isn't like nobody else wants them
14:25 karolherbst: so we would always find others
14:25 mupuf: is it selling *that* fast thugh?
14:25 karolherbst: yeah
14:25 karolherbst: pretty
14:26 karolherbst: like nothing else
14:26 karolherbst: they even have the problem finding a bigger location
14:26 karolherbst: cause they would get even more visitors
14:29 karolherbst: but I don't think it will be that bed actually. I checked, the last time it was sold out after three weeks
14:29 mupuf: then no worries
14:30 karolherbst: I think it will be worse this year
14:30 karolherbst: cause next year they have to move to a smaller locaton
14:35 karolherbst: well, I am sure we will be fine though
15:45 karolherbst: mupuf: well if it is just the two of us, I might be able to get a place to sleep within hamburg somewhere, it is christmas, so some will rather stay at their parents.
15:45 karolherbst: mupuf: the location I am thinking about would be 20 mins by foot to the CCH and 10 by bus
15:48 karolherbst: and also like 10 mins away from a part where we could get a drink quite easily :)
15:48 karolherbst: ohh I was wrong, it is 40 mins by foot, silly google maps
15:49 karolherbst: anyway, 15 by bus, every 7 minutes
15:50 karolherbst: I will see what I can do :)
18:38 tobijk: imirkin: suse has disabled the locking patches again, you may consider retaking bugs from their users :D
18:39 imirkin_: k, thanks for the heads up
18:49 uramekus: ive been running envytools and such and between reboots ive noticed this while using nv
18:49 uramekus: [drm] Memory usable by graphics device = 2048M
18:50 uramekus: but my card is 4096M, is this completely normal?
18:50 karolherbst: uramekus: how sure are you that your gpu is 4G?
18:50 karolherbst: *GB
18:51 uramekus: 100%, bought it being 4g and i can see the memory phisically
18:52 uramekus: also journalctl | grep nouveau reports it too
18:52 uramekus: Oct 17 16:31:50 arque kernel: nouveau 0000:01:00.0: fb: 4096 MiB GDDR5
18:52 karolherbst: I see
18:54 karolherbst: pmoreau: do you need subroutins for the spirv stuff?
18:54 karolherbst: *subroutines
18:55 imirkin_: he does, but it has nothing to do with glsl_to_tgsi.
18:55 karolherbst: ahh I see
18:55 karolherbst: but it sounded like that the entire subroutine stuff will be removed from tgsi alltogether
18:56 imirkin_: maybe. first off, he's generating nv50 ir directly
18:56 karolherbst: ohh, right
18:56 imirkin_: secondly, the necessary hunks can always be reinstated if necessary for hansg's work, for example
18:57 karolherbst: true
19:02 karolherbst: another question. Nouveau is _the_ upstream thing for tk1 and newer tegras, right? Not that I missed anything and start to post garbage on the ML...
19:09 imirkin_: karolherbst: codestr0m (the guy who replied, pathscale guy) has a bit of history wrt nouveau
19:09 karolherbst: I figured
19:09 imirkin_: tbh i don't entirely know what that history is
19:09 imirkin_: but he has some animosity towards the project
19:09 karolherbst: didn't they write their own userspace or something like that?
19:10 imirkin_: "something like that"
19:10 imirkin_: not worth rehashing, imo
19:12 karolherbst: yeah, but still, his answer was pretty much assholish
19:13 imirkin_: just a tad
19:13 imirkin_: i chose not to reply.
19:14 karolherbst: I think he won't replay anyway
19:15 karolherbst: I just feel sorry for the user, that he actually will read his reply
19:16 karolherbst: and you can't leave it as the reply anyway, because... you know, the user maybe won't know better or something like that
19:23 imirkin_: karolherbst: the guy replied directly to me after that with a follow-up question
19:40 karolherbst: imirkin_: I see :D
19:46 karolherbst: well I guess in the end mupuf most likely knows best
21:24 robclark: imirkin, re: tegra stuff.. pixel-c is using nouveau (well, plus some patches) rather than the "official" driver.. if that says anything about the "official" driver..
21:25 imirkin_: robclark: i'm aware
21:25 robclark: ok
21:25 imirkin_: robclark: but proprietary userspace
21:25 robclark: yes
21:26 pmoreau: imirkin_: If you were talking about Hans' LLVM -> TGSI work, he will most likely drop it. He hasn't work on it since FOSDEM, and has been assigned to new projects.
21:26 imirkin_: ah, i see
21:26 robclark: btw, re: X startup, I wonder if there are still some non-upstream patches needed for the whole separate tegra-drm (display) vs nouveau (gpu) thing.. one of nv arm folks would know better probably
21:27 imirkin_: gnurou has a branch, which that user was running
21:27 imirkin_: robclark: https://github.com/Gnurou/xserver/commits/gk20a
21:28 robclark: ahh, yeah, something like that
21:34 karolherbst: mhh, system crashes while qt5 update -> not good
21:35 annoyed: just had a nouveau GPU lockup... is there a way to save my session?
21:36 karolherbst: annoyed: in theory you should be able to ssh into your machine, maybe there is a way of saving your session through command line tools, but I doubt that
21:36 annoyed: karolherbst: indeed i can... not sure what to do once im there tho
21:37 karolherbst: exactly
21:37 karolherbst: well
21:37 karolherbst: maybe you could suspend your system
21:37 karolherbst: I doubt it would work though
21:37 karolherbst: if you run systemd, maybe loginctl might help
21:39 annoyed: karolherbst: thanks, ill probably just end up rebooting. this happens pretty regularly for me... is there a way i can use it to help the nouveau project? id prefer to improve it than succumb to nvidia's garbage agenda
21:40 karolherbst: annoyed: what gpu do you have?
21:40 karolherbst: and what desktop environment are you using
21:41 annoyed: karolherbst: honestly, im not sure what GPU, but im running GNOME 3 on Debian 8
21:41 karolherbst: check lspci
21:41 mupuf: skeggsb: ah ah, right :D Hackers definitely do not use emails in the same way than the rest of the population
21:41 karolherbst: the heck? :D
21:42 karolherbst: ohh right
21:42 annoyed: karolherbst: looks like GF108GL [Quadro 600] (rev a1)
21:42 karolherbst: didn't got at first to what it was refering too
21:43 karolherbst: annoyed: mhh, sadly gnome 3 uses a lot of opengl contexts at once, which nouveau isn't exactly good at handling at the same time
21:43 karolherbst: it might be worth a shot trying something more simple like xfce, other then that, you might be able to figure out what freezes the gpu
21:43 karolherbst: those things are rather hard to debug
21:44 annoyed: karolherbst: :/ ya, seems pretty foreign to me...
21:44 annoyed: karolherbst: theres no automatic crash reporting or anything like that tho right? im kinda suffering in silence?
21:45 karolherbst: bsically, yes
21:45 karolherbst: I mean, your gpu freezes
21:45 karolherbst: what do you expect
21:46 karolherbst: dmesg might give you some clues
21:48 annoyed: karolherbst: i would hope that there is something i can do besides install the proprietary drivers (or fix it all myself, since i have no idea what im doing), but guess not... oh well, guess nvidia wins this time. thanks for the help
21:54 RSpliet: oh... annoyed might simply benefit from a 4.7 kernel - did wonders for me
21:57 RSpliet: since the 2 firmware fixes, nouveau on the fermi lended to me and my NVE7 run more stable than i915 on Skylake
21:58 imirkin_: RSpliet: are those firmware fixes upstream?
21:58 RSpliet: imirkin_ 4.7
21:58 RSpliet: but they don't benefit that one NVE6 bug report that even gnurou looked into
21:58 imirkin_: sure
21:59 imirkin_: can't fix everything at once :)
22:00 RSpliet: (ftr 786656295b318b5894d71d9c0b153b6c17e26f15 and ca79e49d6add77b69be3362ddfe5b068f62bf1de )
22:38 imirkin_: mupuf: when you get a chance, could you play with my ddx patches on your GM206? https://github.com/imirkin/xf86-video-nouveau
22:38 imirkin_: mupuf: i think you might be the only person who has one
22:38 imirkin_: [and is semi-willing to test]
22:38 mupuf: sounds like a good plan!
22:38 mupuf: I was very pleased to see the patches
22:39 mupuf:has a guest at home and samuel is using reator
22:39 imirkin_: yeah, doesn't have to be right now
22:39 imirkin_: but some time this week would be super
22:39 imirkin_: bbl
23:00 mupuf: imirkin_: NOUVEAU(0): Failed to initialise context object: COPY_NVE0 (0)
23:00 mupuf: 4.7.0-rc5
23:03 imirkin: blah.
23:03 imirkin: i'll look into it
23:04 hakzsam: you don't need that for > GK110
23:04 hakzsam: [according to mesa]
23:14 imirkin: yeah, that copy thing is weird.
23:15 imirkin: mupuf: btw, is that error fatal?
23:15 mupuf: imirkin_: nope
23:16 mupuf: but it uses llvmpipe for rendering
23:16 mupuf: which is ... defeating the purpose
23:16 imirkin: er huh?
23:16 imirkin: how does that even happen?
23:17 imirkin: you mean nouveau fails to load and it pulls up modesetting?
23:17 imirkin: or nouveau flips into noaccel mode and so you lose dri?
23:17 imirkin: either way, i'll work it out
23:18 mupuf: the latter
23:18 mupuf: Error initialising acceleration. Falling back to NoAccel
23:19 mupuf: https://paste.pound-python.org/show/4VDzHDNXFTp31diZtaZY/