07:37Guest98428: that moment when you type your pw wrong and use an uppercase letter instead of a lowercase one…
09:53karolherbst: imirkin, hakzsam: VP ported bioshock with eon
16:48kloofy: impulse compulsive disordered lamb called mareko has not removed a ban from me on #radeon
16:48kloofy: compulse impulsive goat maybe even
17:32Calinou: ask them about firmware!
17:40kloofy: if you talked to me, than i can say there is not any bit about firmwares information i need to know about, rather i'd just use the driver and make the firmware my own, for the t-ask question try to give bit of an answer to him too
17:40kloofy: well namely t-ask i've been wrapped my head around how to the scheduling correctly for say about 5months or more now
17:41kloofy: and there appears to be a way to do it on all asics that i care about, basically deprecating that brainfart which was done earlier that you talked about
17:42kloofy: but.., it's as said i do not care about that scheduler my own
17:47kloofy: t-ask: and on linux the major things is coopratively get rid of bugs inside the open source drivers, without wasting time for conflicts all the time
17:47kloofy: the firmware is rather small and easy hunk to be implemented, but the large hunk is the driver
17:51kloofy: and yeah we talk about basic compiler hack and the scheduler hack alltogether boosting most the chips enourmously, it's just i don't maybe even want to work on it snatching more time off of myself, and hence blocking my really small chanches to progress ahead in life
17:53kloofy: i've been sabbotaged and violated my whole life, the team work like the driver is , similar projects i can't real attend in, i.e i mostly decided to go on individually
17:59Sophira:offers hugs, if they're wanted and/or needed.
18:00Sophira:is not really involved as of yet, but will be getting more involved.
19:02karolherbst: mupuf: how long do you need reator?
19:10hakzsam: karolherbst, sorry it's me
19:10hakzsam: but I'm done
19:11karolherbst: ohh I see
19:20kloofy: i was about to order a sea island island based ultrabook, actually core i5 low power + hd8550 but i just barely missed the offer, i was out since one of the friends rang me
19:21kloofy: anyways... what is the driver that works the best, i mean all are candidates for firmware reimplementations, it should not be very hard work, but i agree the hw especially sff sucks on the net what they have in offers
19:22kloofy: so we could write a bit of software and direct the market to the right ways
19:27kloofy: Calinou: it's just my understanding about things that firmware is barely about power management, maybe couple of other things on asics, most of the things is unimportant all that matters is that isa bits are placed wisely to registers doing that non sparingly
19:27kloofy: on top of that heavy lifting one could choose to write the minor hw stuff
19:37kloofy: all that kernel and compiler llvm stuff or gpu compiler, means one of the biggest complications, the moment i got a limp like traliling behind, i knew it's potentially in the end, i wanted to finnish the research young enough, but was blocked there
19:40kloofy: but those who gambled with my life in greater details screwing it all around, are pretty much happy regardless, they succeeded with the mission
19:40kloofy: i dunno what are they though, feels like carma loan from some
19:42kloofy: it's the easiest thing to think about when bunch of really clueless people have wanted decades to see my death
19:43kloofy: they are angry because they did not get much proper things from the god
19:46kloofy: linus once said when you start with a slogan like this, there is nowhere to go, when starting early one with a need to crack up another person and feeling that need throughout the life and putting a lot of emphasis and importance on that
19:46urmet: keegi tegi liiga? tahad kalli saada?
19:46kloofy: basically there is nowhere to go from there, morans ready
19:47kloofy: urmet: minu vastu enamus ainult liiga teevadki
19:47kloofy: ei ole mingeid kalliandjaid olemaski
19:51karolherbst: somebody turned on my laptop at work...
19:52urmet: NSA probably
19:53kloofy: i head off to sleep, for some yeah only suffering is actual in life, i named some others who have managed to get such an envy, this is hypothesis that some think for their misfortune they must just payback for them who were more fortunate with their birth, it equally balenced stuff in their worlds
19:53karolherbst: that didn't work out so well, did it
19:56hakzsam: karolherbst, do you have a F1 2015 trace somewhere btw? I lost mine
19:56karolherbst: mhh I don't have that game, so most likely not
19:56karolherbst: nope, doesn't seem that wasy
19:59karolherbst: I give up
20:11kloofy: karolherbst: well somebody woked against me only for 20years in a row too, good luck to you , seems the nation here is world wide known for the quassimodos and retards killing off their biggest hopes, so we lost valgre and mäng so we lust me "unfortunently"
20:14mupuf: kloofy: hey, could you keep it technical here? please
20:15urmet: karolherbst: call your network guy and let him disconnect your laptop :P
20:15karolherbst: urmet: we don't have one :D
20:16karolherbst: well, I will check tomorow what that is all about, maybe a pidgin got into the office or something.. or somebody really worked until late
20:17kloofy: mupuf: no i am about to leave it does not matter at all if that is technical or not, what i've learned the same suffered dinner of a jesus that i face
20:17urmet: pigeons usually don't know how to use irc clients..
20:17karolherbst: urmet: I put my laptop into suspend, so irc reconnects automatically ;)
20:18kloofy: mupuf: though what is the technical stuff you miss, as i understand it's ok with you and hakzsam and most of the french overall, most the germans too, i seem to get along with you
20:18urmet: kobi magama ja ära spämmi
20:19kloofy: urmet: ära sega vahele
20:19urmet: ainult sina segad. kõiki
20:20mupuf: kloofy: I want to see you work towards something for nouveau and talk about/around it.
20:20kloofy: urmet: sa oled saja protsendiliselt segane inimene nagu siin eestis on neid enamasti kohatud, ära hakka siin midagi rääkima, te olete margi eestil täis teinud
20:20kloofy: mupuf: well roughly...
20:20urmet: well, i vote for kickban
20:20kloofy: mupuf: with nouveau everything looks fine, except few of the bugs inside the driver about multithreading which i covered
20:21kloofy: and the scheduling for the instructions which is also easy
20:21mupuf: ah ah, estonian. Feels like I could get it ... but not at all :D
20:21mupuf: well, how about the sched instructions for maxwell
20:21kloofy: this guy is moran
20:21mupuf: it is not easy and if it is for you, prove it ;)
20:22kloofy: mupuf: well they are just masks, but i could it better, you could back around 7 masks every 7 instructions too
20:22kloofy: mupuf: well i showd the patent how the memory controller actually works
20:22mupuf: how is this related to sched?
20:22kloofy: it keeps the memory back information like tiling and such on the regs
20:23kloofy: mupuf: well you can use those regs to put the masks into
20:23imirkin_: guys - /ignore works quite nicely.
20:25urmet: so much cleaner
20:25kloofy: mupuf: you do not waste cache that way at all
20:26kloofy: mupuf: but mwk actually processed this information, and i like his style
20:28kloofy: 65536 instruction slots means you can put can have the tiling regs encode the masks , it's all there is nothing to prove thow
20:29kloofy: 47 bit is a tag called program counter other remaining bits are so called tiling information, which show which are the bnk witdhs which is the physical location
20:32hakzsam: karolherbst, where is located your traces repo?
20:33karolherbst: hakzsam: google-drive
20:33hakzsam: link please? :)
20:33kloofy: mupuf: you mean how many of those max 65535 slots have corresponding tiling 64-47 reg bits held on the cirduit, well all of them!
20:34mupuf: karolherbst: what happened to the idea of sharing them on mupuf.org?
20:34hakzsam: mupuf, oh, you also have a repo?
20:34mupuf: I mean, fine by me, but I was looking for these traces but couldn't fine them :D
20:34mupuf: hakzsam: sftp
20:34karolherbst: yeah, I kind of forgot the access stuff
20:34karolherbst: I couldn'T access at all
20:34karolherbst: something along those lines
20:34mupuf: ah ah
20:35mupuf: yeah, right, I may also have forgotten
20:35karolherbst: I linked to them from bugzilla bugs
20:35mupuf: free storage on gdrive is also fine
20:35karolherbst: well 15GB
20:35karolherbst: ain't that much for traces
20:36hakzsam: it's not yeah
20:36mupuf: true, I have 1TB of storage on the server
20:37karolherbst: mupuf: maybe you could set it up for git lfs?
20:38mupuf: git lfs?
20:38mupuf: what's wrong with sftp?
20:38mupuf: it is more efficient AFAICT
20:38karolherbst: I like git
20:38karolherbst: and with git lfs you don't download the files except you want to
20:38urmet: isn't git lfs some kind of http(s) shenanigans to get it working?
20:40karolherbst: mupuf: actually we could move the vbios repository to git lfs as an example
20:41mupuf:has no idea what git lfs is
20:41karolherbst: large file storage
20:42karolherbst: it replases big files withing git by text files with a pointer to something
20:42karolherbst: and the big files are uploaded on a git lfs server
20:43hakzsam: seems an interesting stuff
20:43kloofy: mupuf: so in what sense are you being technical quasimodos like urmet make you'r life miserable when you try to talk technical stuff to me, plus they specialise on stalking me and driving my female fans crazy
20:43karolherbst: I think on github you get like 100MB free storage .D
20:43karolherbst: but there are many omplementations you can host anywhere
20:44urmet: 100MB? :O the possibilities
20:44karolherbst: 5$ per month for 50gb
20:44karolherbst: ohh actually, 1GB are for free
20:44Calinou: *bandwidth* is limited to 1 GB too
20:44Calinou: TL;DR it's useless
20:45karolherbst: Calinou: ... seriously?
20:45karolherbst: Calinou: you are aware that you can simply use your own git lfs server?
20:45Calinou: GitHub's free LFS tier is plain useless though
20:47karolherbst: hakzsam: you can specify pattern of files which get autoamtically uploaded to git lfs, like "git lfs track "*.xz"
20:47karolherbst: and then it integrates into the normal git add/commit workflow
20:47hakzsam: yeah, I should try
20:48hakzsam: karolherbst, btw, the divinity traces seem corrupted, I can't replay them
20:48karolherbst: uhh, why do I have an unlimited bitbucket account? :D
20:48karolherbst: hakzsam: they shouldn't be
20:49karolherbst: hakzsam: it may crash though, bcause of sillyness
20:49hakzsam: but they do here :)
20:49karolherbst: even the catalyst one?
20:49kloofy: mupuf: i know you're ok, but i am commercializing my research and i do not want to deal with scammers and stalkers daily basis anymore, they along with imirkin do not convince me much
20:49karolherbst: ohh wait
20:49karolherbst: hakzsam: maxwell or kepler?
20:49kloofy: so have good one mupuf, cheers.
20:50hakzsam: but actually, apitrace crashes
20:50karolherbst: you need gl 4.2 or so
20:50hakzsam: I have locally :p
20:50karolherbst: let me check
20:51karolherbst: the game engine produces pretty bad stuff somehow, so it might be normal
20:51karolherbst: but the catalyst one should work
20:52karolherbst: .. how nice, somebody turned by laptop off
20:52karolherbst: ohh wait, maybe that's mac doing updates or so
20:52urmet: mac? D:
20:52karolherbst: well, I can't choose what I get at work
20:52karolherbst: hakzsam: right, midshader thing
20:53karolherbst: you need that
20:53hakzsam: oh okay
20:53karolherbst: ... the heck
20:54karolherbst: mhh, funny that both traces run here
20:54karolherbst: the one shouldn't
20:54karolherbst: maybe the other trace was on nouveau
20:55karolherbst: yep, seems that way
20:55karolherbst: both traces run fine here
20:55hakzsam: with allow_glsl_extension_directive_midshader ?
20:55karolherbst: I have it set, yes
20:56hakzsam: okay, I will try again
20:56karolherbst: I think they need it because they include the shader themself
20:56karolherbst: and don't pace the #extension decleration at the top
20:57karolherbst: the nouveau path doesn't even run on radeon/intel without issues
21:01hakzsam: karolherbst, works fine now
21:02karolherbst: if you want to run the game, you also need the patches/hacks mentioned in the bug tracker
21:02hakzsam: all your traces replay fine on gm107 (except the TressFX one as expected)
21:02karolherbst: even the metro one?
21:03karolherbst: ohh wait, I trimmed it
21:03karolherbst: hakzsam: can you replay the metro one and keep an eye on your sys ram usage?
21:03karolherbst: and see if you notice anything odd
21:03hakzsam: nope, the metro just crashes
21:03hakzsam: process killed ;)
21:03karolherbst: let me check something
21:04karolherbst: I think there is a resource problem with metro
21:04karolherbst: if you have time and are motivated
21:04karolherbst: it claims like 6GB or memory on my system
21:04hakzsam: motivated yes, time no :)
21:05karolherbst: wow, it froze my entire screen
21:05karolherbst: allthough I run intel
21:06hakzsam: imirkin, wtf, flickering in talos still happens on gm107
21:07hakzsam: karolherbst, well, I have almost fixed all UE4 demos, and I would like to fix F1 2015 before metro
21:08imirkin_: iirc metro ll has never worked on nouveau
21:08imirkin_: we never unmap objects
21:08imirkin_: but we should
21:08imirkin_: i think it's getting killed by that
21:08karolherbst: hakzsam: no worries :)
21:08hakzsam: imirkin_, probably
21:08karolherbst: imirkin_: I cat meminfo, let me check what increases
21:08imirkin_: might not even be in meminfo
21:08hakzsam: imirkin_, but I wonder why talos still has this flickering issue
21:08imirkin_: it's vram mappings or something dumb
21:08imirkin_: hakzsam: differnet issue.
21:09hakzsam: well, we fixed that recently, right?
21:09hakzsam: at least on fermi it does
21:09karolherbst: imirkin_: yep, seems you are right
21:09karolherbst: SUnreclaim increases significantly though
21:09karolherbst: well, 30MB, which is till +60%
21:09imirkin_: hakzsam: on gm107?
21:10hakzsam: karolherbst, hehe, the tomb raider fire issue is fun :)
21:11karolherbst: one of the best
21:11imirkin_: hakzsam: there's a million reasons why it could be messing up on gm107
21:12hakzsam: but it's exactly the same issue we had
21:12imirkin_: well, note that GM107 takes some slightly different paths through the vbo logic
21:13imirkin_: otoh talos only uses vbo's, so in practice, it should be identical
21:33karolherbst: mupuf: we can fake the vbios of maxwell2 gpus, right? I am currently not that much sure
21:39karolherbst: 100% sure?
21:40karolherbst: silly maxwell2 gpus
21:43karolherbst: this really destroy all the fun
21:45mupuf: we just need to fix it
21:45mupuf: it is probably still possible
21:45karolherbst: I see
21:45mupuf: but nvidia refuses to accept my vbios
21:45imirkin_: there might be a signature thing going on there too
21:45karolherbst: let me try it
21:45karolherbst: I am sure you did it wrong :P
21:47karolherbst: mupuf: how much vram does your gm206 have?
21:47karolherbst: 4 or 2GB?
21:47mupuf: imirkin_: yes, there is something like this
21:48karolherbst: seems like 4GB
21:52karolherbst: mupuf: iit works for me
21:53karolherbst: well as in I whiped the vmap tbale and nvidia refuses to load
21:53karolherbst: "NVRM: RmInitAdapter failed! (0x30:0xffff:669)"
21:53mupuf: karolherbst: try without even changing one bit
21:53mupuf: it will refuse to load ;)
21:57karolherbst: you are right
21:57karolherbst: silly nvidia :/
23:03daidoji: hello, I've been having some type of race condition that when running Chrome (or sometimes FF, maybe flash related)
23:03daidoji: Xorg will hang entirely
23:04daidoji: here are some logs https://gist.github.com/daidoji/e7badeb8f2914e0f6574d6070cd2086d
23:04daidoji: with these errors
23:04daidoji: I've attempted to peruse Google but it seems like this type of issue should have been fixed several kernel versions back or something and nothing that seemed to directly apply to me
23:04daidoji: any help would be appreciated
23:05imirkin_: daidoji: that log is just saying "kernel driver is hung"
23:06imirkin_: daidoji: a kernel log could have interesting info
23:06daidoji: imirkin_: dmesg and syslog don't show anything on reboot
23:06daidoji: although I will wait until it happens again to past those logs if you think they'd be helpful
23:07daidoji: I'm not entirely able to reproduce deterministically though
23:07imirkin_: give it at least 2 minutes before rebooting
23:07daidoji: imirkin_: oh okay, what will that do?
23:07imirkin_: also it seems like more applications are deciding that parallel GL contexts are a good idea
23:07imirkin_: which is something nouveau totally sucks at
23:07daidoji: imirkin_: oh really?
23:07imirkin_: i have a branch if you feel like testing random patches: https://github.com/imirkin/mesa/commits/locking
23:08daidoji: imirkin_: do you think that would work?
23:08imirkin_: this isn't super-tested on nv50, i mostly did the initial devel on nvc0
23:08imirkin_: mmmm... not necessarily
23:08daidoji: I might try that
23:08imirkin_: the basic deal is that there's a bit of an expectation mismatch
23:09imirkin_: it used to be, back in the bad old days, that you played games with GL, and then your X DDX handled the rest via various operations
23:09imirkin_: and so the solution to game $x crashing your box was "save your work before starting game $x"
23:09imirkin_: or "stop playing game $x"
23:09imirkin_: meanwhile the X DDX was super-duper stable
23:10imirkin_: now everyone and their mother thinks sticking opengl calls into random applications is a great idea
23:10daidoji: oh its not?
23:10imirkin_: the opengl driver isn't *nearly* as hardened as the X DDX
23:10imirkin_: and it's 100 times more complex
23:11imirkin_: and the kernel driver isn't hardened against a ton of applications trying to create contexts
23:11imirkin_: you could argue that these are all nouveau bugs - and you'd be right
23:11imirkin_: but that doesn't change the fact that they exist :)
23:11daidoji: oh I see
23:12imirkin_: basically application usage evolved faster than the driver hardened
23:12daidoji: so opengl calls in a bunch of different applications cause issues in nouveau -> mesa -> X DDX -> kernel?
23:12daidoji: is that the right heirarchy?
23:12imirkin_: nouveau isn't a thing as such
23:13imirkin_: it's a little bit in mesa, it's a little bit in X, and it's a little bit in the kernel
23:13imirkin_: (or big, depending on how you look at it)
23:13imirkin_: also mesa interacts with the kernel directly
23:14imirkin_: and it probably doesn't help matters that i'm a luddite and run a desktop and applications from 1965, so i don't always run into these issues as much as "regular" users
23:14daidoji: thats why it can hang my whole system?
23:14daidoji: imirkin_: oh perhaps, but thanks for doing all the free work
23:15imirkin_: anyways... seeing a dmesg after the hang would be interesting
23:15imirkin_: there's another issue on tesla gpu's that's never been properly diagnosed
23:15imirkin_: where you get an error with something about 0x00406040
23:15imirkin_: and then poof
23:15imirkin_: anyways, i gtg
23:15imirkin_: much like the gpu does in those situations
23:18daidoji: imirkin_: roger, thanks. I'll try and get that dmesg the next time it happens