00:00imirkin_: ah ok - yeah, the high end laptops do that sometimes
00:19nyef`: mangix: I've been getting the impression that my hardware disables the integrated graphics if there's an eDP panel connected or an nvidia card in the mxm slot, but possibly not if there's a radeon in the mxm slot. I've yet to experiment with this in either direction, though.
00:20nyef`: mangix: Dare I ask what sort of laptop you're using?
00:34mangix: nyef`: CLEVO P870M
00:36mangix: i think I have an early sample or something. The SoundBlaster logo is not engraved
00:37imirkin_: SB16? :)
00:38imirkin_: they don't make 'em like they used to
00:38mwk: more madness...
00:39imirkin_: [that's the original SB, not the much more advanced 16-bit model... twice as many bits!]
00:42imirkin_: mwk: are there saturn simulators that need this info?
00:45mwk: saturn never used NV1...
00:45imirkin_: what used it?
00:45mooch: just pc lmao
00:45mwk: PCI graphics cards
00:45mwk: not very popular ones.
00:45imirkin_: wasn't saturn somehow similar to it though?
00:45mooch: not really
00:45mooch: that's a misconception
00:45mwk: they had a deal with sega, but it never came to anything
00:45imirkin_: huh ok
00:46mooch: saturn basically deforms sprites
00:46mwk: sega kind of figured nvidia wasn't exactly to be taken seriously at the time
00:46nyef`: mangix: CLEVO, huh? Neat. I'm running an Alienware m17x R4, myself.
00:46mwk: NV1 does support saturn gamepad ports, but that's just nvidia adopting the connector, it never went the other way
00:47mangix: mine was second hand. friend sold it since he wanted to 1080s in SLI
00:47mangix: this chassis doesn't allow it
00:47imirkin_: mwk: ah ok.
00:48mwk: so: if I'm lucky, mooch will implement that stuff in her NVwhatever simulator (probably only after I describe the NV3 version, which doesn't differ all that much); if I'm not, it's just going to be a museum piece.
00:49mwk: well, I suppose it's going to be a museum piece either way, what else do you call ancient GPU simulators...
00:49mooch: i hope i'll be able to, but man, this shit is complex
00:49mwk: mooch: ROPs are reasonably simple, you could just steal the code straight out of hwtest
00:49mwk: rasterization, on the other hand...
00:53mooch: lmao 4chan's emugen thinks i stole my entire emulator from envytools
00:54mwk: would be nice if it was that simple to do, wouldn't it...
00:54mooch: envytools didn't even have info on the VESA parts of NV3 until i put it there
00:54mwk: for that matter, would be nice if envytools were complete enough to actually do that.
01:13mwk: ... and the final bit of madness for today.
01:13mwk: that'd be all
01:14mwk: the bitwise operations are too fucked up to document right now.
01:14mwk: so is blending
01:14mwk: and let's not even think about dithering
01:16nyef`: It's not some set of sixteen bitwise operations?
01:16mwk: 256 of them, since there are 3 inputs involved
01:17nyef`: ... Okay then.
01:17mwk: also, the hardware is very nice for you, and it also allows you to use the 3 inputs in any order
01:17mwk: including some orders that involve repetitions
01:17mwk: because why not
01:18nyef`: Sounds logical to me, especially for an initial design.
01:18nyef`: ... Or for a second system.
01:18mwk: then it converts the bitwise operation selector to the canonical bitwise operation selector with the inputs in the canonical order
01:19mwk: and, if it notices the final selector happens to equal "destination passthru", it discards the pixel entirely
01:19mwk: (which is observable, since "destination passthru" would actually modify framebuffer for some pixel formats...)
01:20mwk: ... unless that particular optimization is disabled in a certain config bit.
01:22nyef`: So, fun and games and overbuilt systems?
01:22nyef`: Well, overbuilt in some directions, and anemic in others?
01:23mwk: pretty much yes.
01:25mwk: + some hilarious hardware bugs
01:25mwk: I particularly like how they put the "plane masking" stage before the dithering stage
01:30mwk: umm wtf
01:30mwk: color key is performed after the ROP?
01:33mwk: I should decompile that NV1 driver some day and figure out just wtf their weirdo LEGO programming model is supposed to allow
01:55mwk: the nice part is that a lot of shit in the ROP is starting to make sense now
01:57mwk: it seems I can simplify the hwtest implementation a lot
01:57mwk: it's quite horrible right now
02:36imirkin: orbea: when replaying your trace against a "production" mesa 13.0.3 build, i get the same segfault. if i run it against ... something i happened to have built, it's fine.
02:37imirkin: [basically against head]
02:37imirkin: orbea: can you check if you still have the issue against mesa git?
02:37orbea: strange, I was using mesa git
02:37imirkin: orbea: ok. well my random build is also with -O0
02:37imirkin: perhaps that affects things
02:38orbea: it was with commit de0b0a3a9c
02:38orbea: that was with -O0 -g
02:38imirkin: my build is at bdeb22b, but wtvr
02:39orbea: its been happening for a long time too
02:39imirkin: er wait. wtf.
02:39imirkin: OpenGL core profile version string: 4.3 (Core Profile) Mesa 17.0.0-devel (git-bdeb22b)
02:39imirkin: $ git describe bdeb22b
02:40imirkin: ok. not sure what build i have. let me just rebuild head.
02:46orbea: interesting...the apitrace doesn't segfault with LIBGL_DRI3_DISABLE=1
02:47orbea: i wonder if it will just work with modesetting too...
02:47orbea: oh, wait, I'm already using modesetting and its not crashing...
02:49imirkin: ok. the stupid git_sha1.h isn't updating.
02:49imirkin: i prolly have a stale copy =/
02:49imirkin: someone moved it
02:49imirkin: OpenGL core profile version string: 4.3 (Core Profile) Mesa 17.1.0-devel (git-6a921ea)
02:50imirkin: orbea: https://hastebin.com/oyefiyeqoc.pas
02:51imirkin: anyways, no crashes at head. valgrind's happy.
02:52orbea: that is with the nouveau ddx?
02:52orbea: OpenGL core profile version string: 4.3 (Core Profile) Mesa 17.1.0-devel (git-de0b0a3a9c)
02:55imirkin: no, mesa
02:55imirkin: yes, with the nouveau ddx
02:55imirkin: oh yeah, commit 6a921ea isn't upstream
02:56orbea: well I get this http://dpaste.com/0EQTZGX
02:57orbea: however http://dpaste.com/19MQ675
02:57imirkin: i mean ... i believe you ... just ... weird.
02:58imirkin: i do get the same error with mesa 13.0.3
03:01orbea: what is commit 6a921ea? I dont see it in the mesa git.
03:02imirkin: it's a local commit
03:02imirkin: i just rebased onto HEAD
03:03orbea: i should try a newer apitrace actually
03:04imirkin: i have a very old one :)
03:04orbea: I have 7.1 right now
03:04imirkin: er, that's what i have too
03:04imirkin: i thought i had like 6.0
03:07orbea: not much diff with the apitrace master http://dpaste.com/1C5C7NT
03:10imirkin: ok, so the issue is with the depth buffer
03:10imirkin: can you pull it up in gdb?
03:11imirkin: i.e. gdb --args glretrace foo.trace
03:11orbea: let me try
03:13imirkin: this might be easier over pm
04:09imirkin: skeggsb: objections if i wholesale copy e.g. nv10CalcArbitration() from xf86-video-nv?
04:09imirkin: it looks like someone took it and gutted it at some point
04:09imirkin: all those functions are clear as mud
04:09imirkin: but the different paths definitely do something different
04:09imirkin: whether that matters, who knows
05:40jpsollie: hello everybody, can someone tell me if nouveau already supports some openCL functions? When I google it, I see messages from 2012 saying there is some nouveau openCL functionality, but other (more recent) users say there isn't
05:43jpsollie: I am confused, as I prefer using the nouveau drivers for regular work, and would prefer not to change because of this requirement, I just need to know how far you people are
06:11jpsollie: okay, I found some info in the logs, looks like I need to wait
06:13jpsollie: sorry for disturbing, keep up the good work!
09:42mwk: mode #nouveau -b *!*95ca62ba@*.com/ip.188.8.131.52
09:50PEtee: Hey guys ! Does anyone succeed to run nvidia gtx970M with nouveau driver in Debian stretch ?
10:09PEtee: Hey guys ! Does anyone succeed to run nvidia gtx970M with nouveau driver in Debian stretch ?
10:48RSpliet: PEtee: This is not distro support. Upstream software should support your GPU in part, Debian is often way behind on upstream. Talk to Debian folks if you need help running modern software
13:47Peter___: Does anyone succeed to run nouveau driver for nvidia gtx 970m in debian stretch ? thanks
14:05imirkin: orbea: can you see if https://patchwork.freedesktop.org/patch/134178/ has any effect on your issue/
14:17orbea: imirkin: i'll try it
14:29orbea: imirkin: unfortunately it does not seem to.
14:29Peter___: Does anyone succeed to run nouveau driver for nvidia gtx 970m in debian stretch ? thanks
14:29MarcinWieczorek: oh a space
16:33maasikas21: now I got real extensions working the way expected, well as i said, you do not seem to understand much, there is around 100snippets how to make yet more performance, that we specialize on slow bugfree rendering this is a crapy amatateur slogan!
18:45salamanderrake: is there a ubuntu ppa that has the lates nouveau drivers?
19:03Calinou: salamanderrake: oibaf's PPA I guess
19:03Calinou: beware, it is very cutting-edge and may break your system
19:03salamanderrake: ok, thanks.
19:05Yoshimo: Calinou: wasn't there a different one with a more recent version of llvm ?
19:05Calinou: I don't know
21:15maasikas21: jest eti problemõ?
21:15maasikas21: tõmmake nahhui raisk!
21:17maasikas21: i can be responsible for my own actions, and i fucking
21:17maasikas21: demolish you shitbags
21:19maasikas21: tell me what the hell is the problem?
21:20maasikas21: i gonna fucking show you an assault!
21:28maasikas21: tere hommikust lollakad!
21:28maasikas21: tyra te olete ikka täis saanid ju!
21:29maasikas21: kuidas te seda lollpead kommenteerite juumal kyll noh, täis napakad!
23:09ConfusedLizzard: Sorry, my Internet made a backflip
23:12ConfusedLizzard: Basicly, my laptop running a GTX 940m does not play nice with my GPU when booting Arch Linux, OpenSUSE and Kubuntu (The other *buntu Flavours work). Here is the log https://ptpb.pw/3jhV
23:15imirkin_: heh, very not nice
23:16imirkin_: obviously that log has overflowed whatever was at the start
23:16imirkin_: getting the full log, or more importantly, the first group of errors, would be interesting
23:17ConfusedLizzard: How could I provide you this info?
23:17imirkin_: ConfusedLizzard: usually these logs show up in various system log files
23:22ConfusedLizzard1: What the... sorry imirkin_ my provider thinks this is funny