00:09 skeggsb_: gp100/gp104 kernel support is pushed \o/
00:09 skeggsb_: i'll push the gp100 mesa patch on monday after i re-test it quickly, as it didn't apply cleanly
00:10 skeggsb_: https://github.com/skeggsb/Mesa/commits/master in the meantime
00:11 skeggsb_: gnurou: ^^^
00:11 skeggsb_: not sure if you're handling linux-firmware for gp100?
00:17 imirkin: yay, congrats!
00:17 imirkin: too bad you missed the 12.0 release. o well.
00:19 imirkin: skeggsb_: that patch looks good to me... what was the issue with compute btw? or just didn't get a chance to test it?
00:20 skeggsb_: i just didn't test it
00:20 skeggsb_: oh
00:20 skeggsb_: and it's different
00:20 skeggsb_: the launch descriptor format has changed
00:21 skeggsb_: the class seems much the same otherwise, i have some test apps using it
00:23 imirkin: of course it did
00:23 imirkin: why *not* mess with the launch descriptor format
00:23 imirkin: if they couldn't change texture argument order ... gotta mess with something!
03:46 gnurou: skeggsb_: excellent news! on my end I got permission to push the GP100 firmware that you already have to linux-firmware
03:47 gnurou: *no* GP104/GP106 for now though... >_<
03:59 imirkin: surprising
04:00 imirkin: GP100 is the one that's not available for purchase, right?
04:14 skeggsb_: i think you can, but they're in crazy expensive boards
04:15 skeggsb_: hm, maybe not, it's saying the pcie ones are q4
04:16 imirkin: right, if you buy it a datacenter at a time, you can get the nvlink ones
04:17 imirkin: afaik the pcie gp100's were only announced
04:55 uramekus: is there any place that i can find good documentation about whats done/what devs know about GK104 cards ? ive read quite a bit and i feel its not enough
04:57 imirkin: uramekus: http://envytools.readthedocs.io/en/latest/
04:57 uramekus: ive read the graphicstackoverview and 40% of introdutorycouse (didnt read the pdfs)
04:57 uramekus: oh
04:57 uramekus: envytools
04:57 imirkin: uramekus: and https://github.com/envytools/envytools/tree/master/rnndb/
04:58 imirkin: if you have particular questions ... feel free to ask
04:58 imirkin: those wiki pages are, afaik, totally useless i'm afraid
04:58 uramekus: ive tried to acess http://nouveau.readthedocs.io/en/latest/ but nothing, not even a redirect , so ightout these were blnk pages
04:58 imirkin: the things they say aren't strictly wrong, just ... not extremely useful
04:58 uramekus: they are kinda useless, but im kinda afraid to start coding
04:59 imirkin: what are you interested in working on?
04:59 imirkin: i can probably direct you to some relevant items
05:00 imirkin: that envytools hwdocs thing isn't readable either, unfortunately. it can be used as a reference, but it's a lot more useful when you already know how everything works and need a refresher than understanding it for the first time
05:02 imirkin: skeggsb_: btw, you might be interested in https://github.com/envytools/envytools/commit/f96a739b1687becb8a031787a3c968da5073fefd
05:02 uramekus: welp, i got a used 770 with broken DP, im really interested on learning kernelspace things, Video Accel would be a nice thing to start i think
05:02 imirkin: skeggsb_: i think RSpliet found it in a pascal or maxwell vbios
05:02 uramekus: maybe reclocking but seems a little bit advanced for me
05:03 imirkin: uramekus: well, step 1 is to get the hw working well on your system
05:03 imirkin: or at least as well as the state of the art will allow
05:03 imirkin: https://nouveau.freedesktop.org/wiki/VideoAcceleration/
05:03 imirkin: should show you how to get vdpau going
05:04 skeggsb_: imirkin: yeah, i spoke to him about it, i don't think we need to implement it - it's in tables that only pmu devinit executes
05:04 imirkin: if you are an expert on h264, that'd be immensely useful, coz there's something wrong in something we do somewhere but ... hasnt' been trakced down
05:04 imirkin: skeggsb_: could be. just a heads up if you hadn't already seen it. sounds like you're on top of things though :)
05:37 uramekus: well i do understand a little bit of h264 (nothing that you guys could surpass me reading the right docs in like 8hrs ) but from everything i saw i dont think i could help with it, i understand nouveau very little, despite being a user in 2yrs and being a lurker since start of the year
05:40 uramekus: imirkin: what do you reccomend for me to starting? even if all the work is already there and i would only have to reinvent the wheel? for me like, nvkm is clear but still confusing
05:44 uramekus: oh i found https://trello.com/b/ZudRDiTL/nouveau , nevermind
06:43 karolherbst: uhh nouveau pascal support
07:30 mupuf_: skeggsb_: congrats!
07:31 mupuf_: So, gp100 support is purely theoritical or they did send you one of these crazy expensive boards?
07:56 pmoreau: skeggsb_: Congratulations!
08:03 Yoshimo: skeggsb_: in case you don't mind, do you have the lspci output of a pascal card at hand?
08:04 skeggsb_: mupuf_: no, it works, including 3d
08:04 skeggsb_: Yoshimo: no, i don't
10:26 Arguggi: Me again, my gtx 970 with Nouveau is going to "No signal" when I try to boot. I put the journalctl log here: https://irc.arguggi.co.uk/nouveau.log . I see there are some errors like "nouveau 0000:01:00.0: fifo: read fault at 00ffb90000 engine 1f [] client 03 [DNISO] reason 0d [REGION_VIOLATION] on channel -1 [0000000000 unknown]"
10:27 Arguggi: I'm also seeing "nouveau 0000:01:00.0: DRM: Pointer to flat panel table invalid", any idea if I can fix this or should I just wait?
11:01 karolherbst: Arguggi: it is always helpful to provide the vbios and an mmiotrace of the nvidia driver
11:03 Arguggi: karolherbst, sure, I'll check on how to do that and upload them somewhere
11:17 Arguggi: karolherbst, (I'm running the nvidia drivers now since on nouveau I only get no signal). For the vbios I found this page (https://nouveau.freedesktop.org/wiki/DumpingVideoBios/) and the dd method seems to work: https://irc.arguggi.co.uk/vbios.rom
11:18 Arguggi: Not sure on how to get a mmiotrace since I can't even boot with nouveau
11:20 karolherbst: you should get the mmiotrace on nvidia
11:24 Arguggi: karolherbst, should i just start X and then stop the trace or should I do something in particular?
12:07 hakzsam: skeggsb_, yay!