00:07mwk: mooch: alright, I think I'm done
00:13mwk: mooch: you changed 32 to 8 on a reg instead of on the space :)
00:15mooch: wait, there we go
00:15mooch: k i think it's ready for further review
00:18mwk: mooch: ok, just a few syntax nits now
00:18mwk: I really need to finish that sphinx plugin for register doc generation some day :(
00:20mooch: k that's fixed
00:22mwk: mooch: alright, 2 more blank lines missing and it should be good to go
00:22mwk: please fix that & squash that big stack of commits, and I'll merge that
00:23mooch: sorry, i didn't know the syntax lol
00:25mwk: I hope I'm right about it, I haven't used it for some time myself...
00:26mwk: mooch: still missing one in pramdac.rst
00:26mooch: whoops, sorry
00:27mooch: turns out i hadn't saved one of the files lol
00:27mooch: there we go
00:29mwk: alright, looks good
00:30mwk: the commits aren't well-split though, the second one contains fixes for the first
00:31mwk: mooch: also FYI, your patch introduces so-called whitespace errors (ie. spaces at the end of line), I don't care much about them but people tend to complain about those in patches, you might want to check your editor settings
00:38mwk: mooch: so, should I merge that?
00:38imirkin: mwk: when you get a chance, have a look at my envydis patch too
00:39mwk: imirkin: which one is that?
00:39imirkin: i sent a patch - gm107: separate out sched decoding from regular ops
00:40mwk: ah, so not a PR
00:40imirkin: no. github PR's are a pain to create, review, and more generally - use.
00:41mwk: this is horrible
00:41imirkin: figured you wouldn't like it. perhaps you have good ideas for doing it differently :)
00:42mwk: for one, don't use the ?: operator, it's an abomination
00:42mwk: as in, the a ?: b operator; a ? b : c is fine
00:42imirkin: heh ok
00:43mooch: well, i didn't spot any whitespace errors in pramdac.rst
00:45mooch: no whitespace errors spotted in pcrtc.rst either, so i dunno what's going on
00:45mooch: it might be git or github
00:45mwk: mooch: https://github.com/envytools/envytools/pull/61/files#diff-9a2a5ed4ad62905b9e81c6c7949be544R31
00:46mooch: oh yeah, thanks
00:47mwk: imirkin: alright, it's horrible, but unfortunately necessary
00:48mwk: so... go for it
00:48mooch: i've never seen a ?: b
00:48mooch: wtf does it do
00:48imirkin: i'll flip the ?: thing
00:48mwk: mooch: a ? a : b
00:48mooch: oh god
00:48mwk: except that a is only evalueated once
00:48mwk: it's a gnu extension
00:49mwk: imirkin: oh, and a nit that changes nothing
00:49mwk: + if (deco->isa->tsched && (cur*stride % deco->isa->schedpos) == 0)
00:49mwk: I'd say just use (cur % deco->schedpos)
00:50mwk: it makes sense to count schedpos in ISA bytes, not octets
00:50imirkin: i wasn't sure what the units were
00:51mwk: cur is in ISA code bytes, stride == octets per ISA code byte
00:51mwk: btw, is ".i_need_g80as_hack = 1" still true for GM107?
00:52imirkin: no clue what that does
00:52mwk: I thought they got rid of short instructions already...
00:52imirkin: yeah, those are gone a long time ago
00:52mwk: oh, it's another horrible hack related to alignment
00:54mwk: if it's set, and the assembler is about to assemble an 8-byte instruction that would come out unaligned, the preceding instruction (which must be a 4-byte one) is forced to an 8-byte form
00:55mwk: or something like that... I don't remember how exactly it worked, but it's supposed to avoid unaligned insns by converting some short ops to their long forms
00:57imirkin: i see
00:57imirkin: that's definitely unnecessary for gk110+
00:57imirkin: i think fermi allegedly has short instructions, but they're never used
00:57imirkin: definitely never used on gk104 either, dunno if it supports them at all or not
00:58mwk: Fermi has short insns, but we had lots of trouble getting them to work and eventually dropped the whole thing
00:59mwk: mooch: umm, why did you change your name?
00:59imirkin: i've never seen the blob emit them either
01:00mwk: me neither
01:05mooch: mwk: i didn't?
01:05mooch: or are you talking about the whole darius/melissa thing?
01:06mooch: if you're talking about that, it's because i'm a trans girl
01:06mwk: I just saw a mismatch between your github page and the git commit author
01:06mwk: thought there was some misconfiguration going on
01:07mwk: so, should I merge it as Darius?
01:07mooch: merge it as melissa please
01:07mwk: ok, can you submit a fixed version then?
01:08mooch: k gimme a sec
01:08mooch: wait, how do i change my name on commits i've already made?
01:08mwk: git commit --amend --reset-author should do the trick for the most recent commit
01:09mooch: how do i do it for both commits?
01:09mwk: assuming you've reconfigured it properly, which you probably want to do
01:09mwk: and to do it for both commits, just use git rebase --intractive, change both commits to edit, then do the amend as above
01:10mwk: there are probably better ways to automate that, but not worth it for two commits
01:12mooch: is it fixed now?
01:13mwk: yep, two commits by Melissa
01:14mooch: ah okay good
01:14mooch: it's good to merge then!
01:16mwk: no per-commit notifications, interesting...
01:16mwk: oh well
01:22mwk: mooch: alright, thanks for the PR, gnight
08:17redbeardt: I have a rather odd question. I've got a fairly new laptop with an Nvidia 940M on it (Maxwell NV118/GM108) and it worked with nouveau out of the box. I then switched to nvidia for a few months, which also worked. About a month back, it just stopped working, I went back to nouveau, also not working, and now there's all this info on the net about now my card was never even supported by nouveau. Bloody weird.
08:17redbeardt: What do you guys make of this? D:
08:21pmoreau: Weird… Which kernel version did you have initially?
08:25redbeardt: Well it was Ubuntu 16.04.01 LTS so 4.4-ish
08:26redbeardt: I got sick of trying to fix it and decided to jump over to Arch linux a few days ago, which is on 4.7+. Problems persist, but the error messages differ.
08:26pmoreau: At least 4.7 has GM108 support
08:27pmoreau: Could you paste the whole output of dmesg on the website of your choice please?
08:30redbeardt: .. Trying to figure out how to keep it all without the head of it getting truncated *cough*
08:30Tom^: dmesg > wat
08:31Tom^: direct it into a file :p
08:31redbeardt: Oh that's what I did. I just realised it wasn't being truncated.. Brainfart
08:32redbeardt: sorry about the elan device spam, still working on that one
08:33hackafromuk: in fact, i have enough extensions and ways to enter the channel infinitely, it would be wiser if that imirkin would not play that sort of god here, i understand that politics is not the things to be discussed here, but he threattened with a ban before that
08:33hackafromuk: and noone cares what and who that moran is ignoring
08:34Tom^: hackafromuk: welcome to ignorelist again.
08:34pmoreau: redbeardt: It doesn’t busted, what is not working?
08:37redbeardt: pmoreau, well that's good news at least. I've been trying to get it to play nice enough so that it appears in xrandr --listproviders, so that I can set it up with PRIME as described here https://wiki.archlinux.org/index.php/PRIME
08:37hackafromuk: all i say , is i don't in any way take those threattening personally or am in any way scared what sort of pressure those local russians here apply
08:38pmoreau: redbeardt: Do you have xf86-video-nouveau installed? If so, remove it
08:38redbeardt: pmoreau, removed
08:40pmoreau: Now, if you restart X, does it work better?
08:41hackafromuk: why is that would you not try to instead implement the scheduler, it's been some decades when users are wishing to see those, and couple years back all the neccessary things even have been there?
08:41redbeardt: pmoreau, i'll check, i'm running IRC inside my desktop environment so i guess i'll brb. Will start up irssi in tty2
08:43hackafromuk: i am 33years old they picket in front of my window every day, how much of a big fuck i give?
08:43redbeardt: no apparent change :(
08:44pmoreau: Can you paste the content of Xorg.0.log as well then?
08:44redbeardt: pmoreau: lspci even shows that the nouveau module is still in use. I feel that blacklisting it would go against my intention of being able to use the card with nouveau though. :P
08:44redbeardt: pmoreau: ok sec
08:46hackafromuk: Tom^: well let's just say that i am entirely dissapointed that you have ignored me ("yeah really i care a lot") what the hell dude is it my problem who you ignore and what the fuck not idiot?
08:46redbeardt_webcha: pmoreau: http://pastebin.com/cEW9LXNW
08:47pmoreau: redbeardt_webcha: "This page is no longer available. It has either expired, been removed by its creator, or removed by one of the Pastebin staff."
08:47Tom^: pmoreau: works here
08:47Tom^: pmoreau: :o
08:47redbeardt_webcha: pmoreau: oh .. what the. Hold on.
08:47pmoreau: Must have failed to copy/paste
08:47pmoreau: working now
08:47pmoreau: Yep, missed the last W… --"
08:47redbeardt_webcha: pmoreau: ah ok
08:49pmoreau: Why is not trying to load modesetting for the NVIDIA card…
08:50redbeardt_webcha: pmoreau: Perhaps due to my attempts at xorg configuration
08:50pmoreau: Possibly :-)
08:50hackafromuk: anyhow i've gotten sober in the meanwhile, my impressions are still the same
08:51pmoreau: If you don’t need them, try removing them. Otherwise, do not forget to change the driver for the NVIDIA card from nouveau to modesetting
08:51hackafromuk: i.e calim did the work on shaders, he does not ban anyone
08:52redbeardt_webcha: pmoreau: I'm gonna try changing the driver as you say
08:52hackafromuk: so let's not try to give false positives for the people influencing them as being heros, that i only talk about shit here
08:53hackafromuk: i concur that talking about this sllah stuff diserves a ban here, but i've kept it reasonable most the time, talking about technology every so often
08:54redbeardt_w2: pmoreau: new Xorg.0.log w/ modesetting http://pastebin.com/XGtQVg2z
08:55redbeardt_w2: "Screen 1 deleted because of no matching config section." HMMMM
08:59hackafromuk: imirkin: in my visions lower your ego just a little bit, and you find yourself in a position where lot less bugs go through, because you do not waste energy on pointless stuff, rather then being operator which unfortunently is needed people still excpect you and i to be a programmer
08:59pmoreau: redbeardt_w2: Care to paste your xorg configuration?
09:01hackafromuk: so we know that sort of envytools work and the groundwork of kernel was done by, partly some swedish company and voluntary french hackers, no doubt imirkin could had done it as well if he would not waste time on pointless stuff
09:02hackafromuk: what is wrong with my ID, you only have to avoid conflicts to succeed, and you still every time get into this concpiracy shit who ignores who
09:02hackafromuk: and the outcome as for acheivements is almost non existant hence
09:02redbeardt_w2: pmoreau: http://pastebin.com/ApMva1hJ
09:03redbeardt_w2: pmoreau: Oh whoops. I cut off the top of it
09:03redbeardt_w2: pmoreau: Oh wait no I didn't. A tab snuck in. Hell.
09:05pmoreau: Why have two different devices for the Intel card? And you are only adding the Intel card,to screen0, but not the NVIDIA one, why?
09:05hackafromuk: i just talked about recently lxc with nova graphics , i am beginning to work on graphics
09:05hackafromuk: i have 4 + 4 core machine, i think i isolate a container to work on memhardwalled isolated cores per card..
09:06hackafromuk: control it remotely , so when i get a lockup in the container cores, so that i would not need to restart the machine
09:07redbeardt_w2: pmoreau: I actually thought the first device represented the monitor. :< When I was researching how to do these configs, the 'vesa' driver was specified there, which I don't even have.
09:08redbeardt_w2: pmoreau: Can you really add two vga devices to one screen?
09:08pmoreau: The screen section here is virtual, not physical
09:09pmoreau: But honestly, you’re better off without any configuration, it should work out-of-the-box
09:09hackafromuk: but yeah that requires two phsycal cards, i have mobo..where i infact have two pcie slots + intel integrated
09:09pmoreau: imirkin: line 483 "WARNING: value %76 not uniquely defined" and line 510: "WARNING: value %89 not uniquely defined" I guess that’s not helping :-)
09:10hackafromuk: i think overall this relevant stuff, we should advertise methods to work on #nouveau of volunteers
09:10redbeardt_w2: I recall that it didn't work out of the box, but honesly I can't remember exactly how. Meh! I'll copy out the confs and see what happens.
09:11pmoreau: Could have been because it used xf86-video-nouveau, but since you removed it, it should work better
09:15redbeardt_w: pmoreau: new and shiny sans-xorg config Xorg.0.log http://pastebin.com/eHxjfc0n
09:15hackafromuk: yeah i've reconed i have some tail a long time ago, it really should not matter, in my opinion i've forwared almost coherent on ok information
09:16redbeardt_w: pmoreau: Functionally it all seems about the same. Still only have my intel GPU showing up in xrandr --listproviders
09:16hackafromuk: so Calinou: hi, i still think for the future you want to work on FPGA's, when you want to do an accelerator performing well, consuming low power as you talked
09:18hackafromuk: you get it cheaply, you want as big resources as there are for pascal, so to do it cheaply you need to interconnect two arria 10's or kintex ultrascale +-s
09:18hackafromuk: because yeah the high end models would be bit more expensive
09:22hackafromuk: and they probably have pascal sort of level high-end polaris too, phoronix ranted about those though
09:23pmoreau: Your dmesg had a log of suspend/resume, could be that the card just goes to sleep and disappears.
09:23hackafromuk: on ultrascale+ and arria 10, they use a new technology to pack twice as much transistors onto the die with 16finfet
09:23hackafromuk: now the clb slice is the same and is 16 6 input luts
09:23pmoreau: What happens if you run `DRI_PRIME=0 glxinfo | grep OpenGL`, and same for DRI_PRIME=1?
09:24hackafromuk: that means a rouhgly the highest mid level chips are with 10billion transistors
09:25redbeardt_w: pmoreau: http://pastebin.com/0umD9LfH
09:25pmoreau: DRI_PRIME=1, not 17 :-D
09:25pmoreau: And no need for sudo
09:25redbeardt_w: .. crap
09:26redbeardt_w: pmoreau: http://pastebin.com/2V5vFrAb
09:26pmoreau: There ya go!
09:26pmoreau: Enjoy :-)
09:27pmoreau: I guess the GM108 is asleep when you run `xrandr -listproviders`, and does not wake-up, resulting in it not being listed.
09:28pmoreau: However, it does wake-up properly when you run commands using it.
09:28redbeardt_w: pmoreau: sneaky little bastage
09:29pmoreau: And with DRI3, there is no need for -setprovideroffloadsink and all that stuff: it is just plug&play (when it works of course…) :-0
09:30redbeardt_w: pmoreau: so if I kick up something 3D the nvidia should just rock up and start rendering?
09:31pmoreau: As long as you run it with `DRI_PRIME=1`, it should
09:32redbeardt_w: pmoreau: well geez
09:32redbeardt_w: pmoreau: thanks for all your help, i'll have to go to some testing now ;)
09:33pmoreau: You’re welcome! However, there is no reclocking yet fro Maxwell v1, so performance is not going to be stellar, maybe as good as the Intel one.
09:33hackafromuk: Calinou: it's real old page, but that is how old the theory is
09:34pmoreau: I think Karol’s patches that will land in 4.9 do apply for Maxwell v1 as well, and not just Kepler
09:34redbeardt_w: pmoreau: I've been having even more trouble with the Nvidia drivers honestly, so at this point I'm not even that fussed
09:35Tom^: pmoreau: but wasnt there something about the fan too, that nouveau cant control it yet?
09:35hackafromuk: Calinou: they are being smart there to do hw multi-clock domains leaving the freq around 400-1000mhz
09:35Tom^: pmoreau: so upclocking would mean overheating? :p
09:35pmoreau: Tom^: That's for v2 IIRC, since fan require signed firmware there
09:36Tom^: mhm ok
09:36pmoreau: But v1 should be fine as it does not require signed firmware.
09:38hackafromuk: the technique of such would produce lot less heat, as the signals are all over the chip, but would perform equally to 3ghz clock frequency
09:40hackafromuk: we can't go on like this that we flesh out a giant bus called pcie 3.0 or whatever and mount a giant cooler or radiator passive one to cool all the bad design
09:41hackafromuk: it's really crazy imo, there are wiser methods, for low power, low heat gen and high performance
09:42hackafromuk: Calinou: neither the nvidia or amd rom's is/are something where you at all want to learn anything
09:45hackafromuk: cause what they do there is classically just all wrong
09:46hackafromuk: i.e you can use the isa, but non-the less power management is absolute done all wrong there anyways
09:46redbeardt_w: as a layman in this realm, this is an interesting thingto hear since i assume nvidia and amd's devs must be godlike
09:47Tom^: meh they just have access to documentation and source code.
09:47Tom^: thats cheating
09:48redbeardt_w: at first it sounds it sounds like you're saying there's room for competition in that market, but maybe the barrier to entry is just too high
09:50hackafromuk: no, there is a reason why fpga designers are voted into hall of fame, though i do not know precisely what is it, but it's for very special occations, but it just happens they do things correctly
09:53pmoreau: imirkin: I am not properly deleting instructions when doing the split, cause the warning comes from the initial mul64 still holding a def on %76 even though that instructions was removed.
09:58pmoreau: imirkin: I replaced `bld.remove(i)` by a `delete_Instruction(func->getProgram(), i)` and that got rid of the definition as well.
09:58hackafromuk: i am working on that kind of source code, octavo and flexgrip source codes i have, and i posted lots of them to here, they perform very well, but they both yet do not take advantage of hw clock shifted domains
09:59hackafromuk: allthough in flexgrip, i got access to that source code, it is marked there for usage, but is commented out
09:59hackafromuk: but this is actually exactly what you want to do to clock the circuit anything that is beyong 1ghz which starts to generate excess heat
10:05hackafromuk: so yes according to that 98year pdf, so long time ago, on programmable devices those sort of things can be managed correctly, so yes grvi phalanx yank was pretty intelligent, to say he gets 100k mips throughput, yeah when he brings the clock phases into play he gets up to 8times more throughput
10:11hackafromuk: those are yeah , as i have heard they have timing allready correctly done in hw..when you'd want to do clock divider or multiplier in software you gotta the timing right
10:12karolherbst: the heck is ?:
10:12pmoreau: ternary operator?
10:13karolherbst: I meant as in "a ?: b"
10:13pmoreau: <=> a ? a : b
10:13karolherbst: yeah, I am reading through the irc logs currently
10:13pmoreau: with a being evaluated only once
10:13karolherbst: that really looks like it shouldn't be user like ever
10:13Tom^: karolherbst: https://en.wikipedia.org/wiki/%3F:#C
10:14karolherbst: ohh I see
10:14karolherbst: like strcpy(a) ?: otherString;
10:16karolherbst: pmoreau: right, maxwell1 as well
10:16hackafromuk: so the usage is like you just add those phases into clock signal list and it's very simple, vendors there clock the circuits according to gpu's but instead doing it wrong they do it correctly as in any theory of programming there
10:16pmoreau: I didn’t know you could do something like `(argc > 1 ? a : b) = 1;` O.O
10:17hackafromuk: i think what they called in verilog was continuous assginement with sensitivety list
10:17hackafromuk: that means if any of the incoming signals come available, any of the phase output is also sent
10:18karolherbst: pmoreau: C allows a lot of silly things ;)
10:18karolherbst: but I think that construct looks more usefull in C++ if you think about references
10:19pmoreau: karolherbst: And what is the fan status for maxwell1?
10:20hackafromuk: so the ram is clocked at 800mhz with the phases it's equivalent of 6000mhz, and the pll+hwmultiplier and phases are 200mhz *2 *8 around 3ghz for the shader core
10:20karolherbst: just maxwell2 has problems, cause the regs are write protected
10:24hackafromuk: redbeardt: well amd devs are working with the driver on design that is done by someone else, as devs did not design the hw, they are not probably relevant here, they work on sw bits and yeah they should have all the source codes access
10:25hackafromuk: redbeardt: non the less, asic could be done right so it would not generate as much heat, this just needs some certain degree of additional programmability in the circuit, and imo it's not what you'd call an asic anymore then
10:25hackafromuk: in other words totally pointless mission
10:30pmoreau: So, it’s working a bit better now, but also slightly worse at the same time: now it’s fully optimising as if it was not looping
10:32hackafromuk: and mostly the meaning being, before hopelessly picketing and protesting about how bad things are, before one can try to think about the correct ways to do things, cause they also there
10:32hackafromuk: i know americans regardless how cruel they have been have all sorts of correct theories too developed at least, and so do other countries probably
10:38hackafromuk: the design though it's easy in current world, there are still certain amounts of wrong design paths and correct ones, but though it is easy today the design is still hugely important how much heat and performance got of it
10:41hackafromuk: and it's easy only probably cause long time different proffessors around the world have worked polishing the theories and stuff there
10:42hackafromuk: synthesisers and code to do realize the theories also, nowdays it's easy cause of that
10:47juninhopo: i can do a script when i am banned, that i log in right away, this guy is bit idiotic
10:51Tom^: imirkin: morning. :p
10:52hollandnow: they want to ban all the world, because of having too few brains
10:53Tom^: you could set it to uh is it +r for a while
10:53imirkin: yeah, i think i will
10:53pmoreau: Nothing better than some kicking when starting a new day …
10:53imirkin: everyone - make sure you're logged into services if you want to talk
10:59aproperyank: i have like 40000 of ip's, and can make a script to register different nicks on freenode, so give up
11:00Tom^: you could also temporarly ban kiwiirc :D for a day or two and perhaps he gives up
11:00imirkin: hmmmm.... should i +m it for a while and voice everyone?
11:01Tom^: that could work aswell
11:01imirkin: meh, it's not just kiwiirc...
11:01imirkin: there's like 100 diff web irc thingies
11:01MichaelLong: why of all things has he selected this channel....
11:02imirkin: saw people getting things done and couldn't deal with that?
11:02imirkin: who knows - why do trolls troll
11:02imirkin:is not a psychiatrist
11:03imirkin: ah, an oldie but a goodie
11:04Tom^: +S "Only users connected via SSL may join the channel while this mode is set" "Keep in mind that this also blocks all webchat users, as they are not marked as connected via SSL."
11:06Tom^: but i guess having +m and making a tiny ircbot query the user joining telling them to ask one of the ops for +v to be able to speak is the most effective and non blocking one
11:06imirkin: or just auto-+v everyone by default except webirc?
11:07Tom^: that would work aswell
11:07Tom^: just those poor fellows who has issues and uses a webirc will be a bit silenced
11:07imirkin: seems like a lot of effort for such a dumb situation
11:11imirkin: bbl... if he comes back, ban him
11:12Tom^: will do
11:41Tom^: that did not go as planned
11:41Tom^: pmoreau: halp
11:41pmoreau: I never used modes on IRC, so…
11:42Tom^: ok panic situation over. but yea that second ban bans his nickserv user
11:42Tom^: somehow my client interpreted it as two bans :P
11:42Calinou: IRC: Global Offensive
11:42Calinou: Phoronix needs to make an article about this
11:42pmoreau: Tom^: samehere, I saw two bans
11:44Tom^: http://freenode.net/kb/answer/extbans so, with +r and just banning his nickserv users. hes gonna have a painful time remaking new ones. iirc it takes valid email accounts etc
11:57karolherbst: why is Tom^ able to ban people :O
11:57Tom^: good question.
11:57Tom^: i guess it gives you guys more time for coding and less time spent on trolls
11:59pmoreau: Possibly because he also is always there and ready to help. :-)
12:00karolherbst: well, I can't ban people :(
12:00Tom^: karolherbst: i could op you and then you can :)
12:01karolherbst: I was just wondering why
12:01Tom^: yea idk, "[12:51:08] -- | Mode #nouveau [+o Tom^] by imirkin "
12:01karolherbst: ohh I see
12:01Tom^: i dont have chanserv access or so
12:01pmoreau: but since you do not have a bouncer IIRC, you’ll lose that power when you log off.
12:02karolherbst: +0 is just temporary?
12:02Tom^: until i rejoin yes
12:02karolherbst: how was the way to add it to the chanserv thingy?
12:03karolherbst: I mean you can always get op rights back by /chanserv op #channel nick
12:03pmoreau: (that’s what I assume, I could be wrong, and maybe if you reconnect it will reapply +o)
12:03karolherbst: but no idea what ther equiernments was
12:03karolherbst: pmoreau: well usually you log in without op rights and have to get them through chanserv
12:04Tom^: i think the owner of the channel has to register the user with chanserv, and then he can simply /msg chanserv op #nouveau
12:04karolherbst: ahh /msg chanserv flags #nouveau nick +oO
12:04karolherbst: o is for the ability to op yourself
12:04karolherbst: O is auto-op on channel join
12:07karolherbst: Tom^: I am sure every op can add others as ops
12:08imirkin: it was a temp measure while joss keeps trying to spam
12:08Tom^: i thought ChanServ was sort of an bot where you have to be added to , to get privileges
12:08Tom^: im happy to help whenever required.
12:08imirkin: thanks =]
12:09karolherbst: imirkin: yeah I figured, but there was no notice so I thought Tom^ had op rights since always or so
12:09Tom^: karolherbst: that would have been mindblowing eh? :)
12:09karolherbst: ohh no I see the message
12:09karolherbst: I just skipped over it, all fine then
12:09imirkin: so you don't feel left out
12:10karolherbst: meh, I loose them after I lgout :(
12:11karolherbst: imirkin: if you don't mind you could set the +o flag through chanserv, then I can op myself whenever I want to
12:12imirkin: mmmm not sure i can
12:12karolherbst: /msg chanserv flags #nouveau karolherbst +o
12:13karolherbst: but I could imagine that nicks need the right to do that as well
12:13imirkin: right - i mean i don't think chanserv will let me
12:13karolherbst: yeah, I understood, just helped you trying that out ;)
12:14imirkin: -ChanServ- You are not authorized to execute this command.
12:14karolherbst: we really need to care about stuff like that, not that in the end everybody is gone who is able to do that
12:14karolherbst: I am sure skeggsb can?
12:14imirkin: yeah, i think so
12:21imirkin: looks like only skeggsb and marcheu can add users
14:23redbeardt: pmoreau: just incase you're interested, i fired up rocket league with DRI_PRIME=1 and the performance was actually significantly worse than when running under my intel device
14:24redbeardt: pmoreau: but hey at least I know it works!
15:04pmoreau: redbeardt: :-D You could try https://github.com/karolherbst/nouveau/tree/stable_reclocking_kepler_v6 if you are interested
15:07karolherbst: current master should also work now
15:07karolherbst: pmoreau: do you know what gpu he has?
15:15pmoreau: 940M IIRC, a GM108
15:15karolherbst: I see
15:15karolherbst: yeah, should work on master
15:15pmoreau: True, Ben merged them
16:28imirkin: redbeardt: it's expected since you're in the lowest perf level. either wait for kernel v4.9 or grab an upstream branch of the nouveau module
17:37genericum: does opengl-stereo work with nouveau ?
17:37karolherbst: genericum: you mean the stereo port on the gpu? Or just "normal" stereo implemented within opengl?
17:38genericum: I mean 'normal' stereo
17:38karolherbst: mhh, should work I guess
17:39karolherbst: I don't see why not
17:40genericum: i had a windows application that showed only one eye if it wasn't run on quadro.
17:40karolherbst: uhh, well that application might have used nvidia extensions then
17:40karolherbst: no clue though
17:40imirkin: genericum: the GL stereo stuff isn't supported
17:40genericum: wouldn't be much work to port it on linux if it would work there
17:40imirkin: if you're talking about the GLX stereo visuals
17:41genericum: imirkin: exactly
17:41imirkin: however a random application that just so happens to draw two eyes' worth of stuff won't know the difference
17:42genericum: sure you can just render 2 perspectives in one window ..
17:42genericum: I do need hdmi-3d though :(
17:42imirkin: ok, well that's probably a yet-other can of worms
17:43imirkin: the original GLX stereo stuff was meant for those 3-pin port thingies
17:43imirkin: i actually have a card with one of those, but no hw to attach to it
17:43imirkin: afaik the hdmi 3d situation in linux is presently lacking
17:43imirkin: but again, i don't have any hw to play with - in theory it shouldn't be difficult
17:44imirkin: it's just a wide image with the 2 eyes and then the TV does the magic iirc
17:44imirkin: (or tall image, there are diff ways of doing it)
17:46imirkin: patches welcome :)
17:47genericum: yes top/bottom should work with hdmi3d
17:51genericum: where do i start looking for GF100 driver code?
17:51imirkin: there are a few steps
17:51imirkin: step 1: get the screen to come up as 3d
17:51imirkin: step 2: pipe the knowledge that a 3d screen is connected through the stack
17:52imirkin: step 3: expose the new visuals
17:52imirkin: for step 1, you want to look at drivers/gpu/drm/nouveau in the linux kernel
17:52imirkin: specifically nv50_display.c and nvkm/engine/disp/*
17:53imirkin: also iirc someone started on some of this work for intel gpu's, but i have no clue whether it landed
17:53imirkin: or how far it went
17:54genericum: imirkin: iirc is the opensource intel driver ?
17:54imirkin: please rephrase
17:54imirkin: oh. iirc = if i recall correctly
17:54imirkin: the intel driver in the kernel is called 'i915'
17:55genericum: imirkin: isn't this a binary ?
17:56genericum: imirkin: no i was wrong ..
18:16genericum: is there any documentation on the hardware, aside from the kernel ?
18:24karolherbst: genericum: there is nearly none provided by nvidia, if that is what you mean. Otherwise look at envytools
18:43karolherbst: mupuf, hakzsam, pmoreau: do you have some kind of latex template for your presentation or did you just clicked some slides together with libreoffice?
18:43karolherbst: because I would just use the same layout you used
18:46hakzsam: karolherbst, https://docs.google.com/presentation/d/1q_rUoK7cjne-CzpMGFwHxTL7Tb_XDR2c80UtEAeTdSM/edit?ts=57b2f4c2#slide=id.p
18:46karolherbst: ohh, google docs
18:48karolherbst: hakzsam: that's mine so far :D https://drive.google.com/file/d/0B78S7GSrzebIV29wQXh3YnNtdm8/view?usp=sharing
18:48karolherbst: I thought I use something nice to look at :/ but if you don't have anything, I won't too!
18:48hakzsam: yeah, not started yet
18:49hakzsam: my plan is to write "my part" tomorrow :)
18:50karolherbst: well if you plan to do that in latex, I could give you my "framework" for doing it painless
18:50hakzsam: no, I will use gdocs
18:51karolherbst: that "..." after the GP100 accel comment :D
18:52karolherbst: power consumption monitoring was merged with 4.6 by the way
18:52karolherbst: but it was broken for 4.6 and 4.7 and finally reliable to use with 4.8
18:53karolherbst: I think
18:53karolherbst: maybe the last fixed just went in for 4.9
18:53karolherbst: mhh seems that way
18:54karolherbst: hakzsam: also, sure you added gm107 and gk110 isa support within 2014? ;)
18:55karolherbst: ohh I can actually change stuff :O
18:55karolherbst: better revert what I changed
18:55karolherbst: mhh, maybe 2014 is right though, what do I know
18:55hakzsam: it's the talk from the past year
18:56karolherbst: I see
18:56karolherbst: so parts are old and parts are new
18:56hakzsam: only the date in the first slide is correct ;)
18:57karolherbst: k, understood
18:57hakzsam: maybe mupuf has worked on some parts
18:57karolherbst: I think the kernel libdrm part is updated
18:58hakzsam: and the second slide :)
18:58hakzsam: err, the introduction one
18:58karolherbst: right, the year changed
18:59karolherbst: I need a high res version of that logo though :O
19:03karolherbst: you really should use latex though :p (simply because it is much more awesome!)
19:03hakzsam: I know, but I will use the existing template on gdocs this time
19:17pmoreau: karolherbst, hakzsam: We worked with mupuf a bit on the new presentation
19:18pmoreau: But I don’t have a link to the new Google Doc to share…
19:19karolherbst: ohh crap, silly latex can't handle _ ...
19:21hakzsam: pmoreau, new google doc? did you create a new one?
19:35pmoreau: mupuf did
19:35mupuf: hakzsam: you have it
19:35pmoreau: A copy of the previous one, and then updated it
19:35hakzsam: ah okay
20:14imirkin: GK110 became really usable in mesa 10.2. iirc GM107 isa was added in 10.3
20:45pmoreau: So the computed liveness of %68 is "livei(%68): [18 21) [22 24) [42 45)", whereas it should be "[18 45)" AFAICT, from the define, to the end of the loop, or should it also go back to the beginning of the loop?
21:20mwk: seems I missed a good party, everyone got an @
21:20karolherbst: you sure did
21:20imirkin: i figured it'd be good to have a few people with ops around
21:20imirkin: i'll -r and deop everyone in a few days i think
21:20mwk: no, I mereely got drunk, I don't have an @
21:20karolherbst: can't suspend my laptop then I figure
21:21mwk: ok, so it wasn't one of you?
21:21imirkin: it was me
21:21mwk: I figured I'd thank whoever threw me into a train
21:21imirkin: i gave people ops to deal with the joss situation
21:22mwk: oh, that thing
21:22imirkin: since i wasn't going to be here 24/7
21:25mwk: imirkin: good thinking, anyhow
21:29mwk: imirkin: now you made me look wtf +r is in the freenode docs :(
21:55karolherbst: the UNK40 P table looks important
21:56karolherbst: one value range: 6455-15250
21:57karolherbst: the other: 6505-42643
22:12karolherbst: uhh the vpstate table indeed contains domain clocks, but I never saw them :O