09:09 AndrewR: imirkin, https://pastebin.com/5jqmuC4V - this hack on top of master restores blender ....
09:22 AndrewR: imirkin, also, this minimal change apparently fix audiosurf2's custom mod I was playing: https://pastebin.com/RUG5aKVH (not sure how much change in nouveau must be done for enabling all 64 *4 inputs,as blob apparently does .... from our previous talk on this)
09:22 AndrewR: imirkin, number of shader inputs advertized must much number of shader outputs, or I missed too much in how it all work?
09:30 AndrewR: imirkin, grr, 4x32 (for 128 fragment shader inputs)
10:01 kodomo: Hi all! Don't know whether I'm in the right channel for my problem, but I'm trying to get a GeForce M150 rev.a1 up and running with nouveau on a Lenovo T480s, so I thought I might try... :)
10:02 kodomo: In principle, all looks fine, but when I try to run glxinfo on the NVIDIA card, I get nvc0_screen_create:1111 - Error allocating PGRAPH context for M2MF: -16
10:03 kodomo: I'm using linux-5.2.2 on a gentoo, XOrg-1.20.7 with libglvnd and DRI level 2 .
10:06 kodomo: Xorg.0.log confirms both drivers (modeset & nouveau) loaded, I see 2 providers, I have --setprovideroffloadsink prior to attempting DRI_PRIME=1 glxinfo - which set 'associated providers: 1' on both providers... so everything _should_ be in order, shouldn't it? *scratch*
10:07 Moiman: perhaps try newer kernel
10:07 kodomo: The MX150 is a 1030 (Pascal series) to the best of my knowledge...
10:11 kodomo: Moiman: So you think it's a kernel related error message (tbh, I don't understand the meaning of the message itself) cando... newest gentoo one would be 5.6 for the time being...
10:12 HdkR: GP108 had issues, newer kernel may fix it
10:12 kodomo: ok - will try, then
10:30 Antoine-: Hello, I am using xubuntu 19.10, on a vaio laptop with an nvidia 640M LE (and intel hd 4000). I can't get sound through the HDMI. Could it be related to nouveau?
10:31 Antoine-: (I tried switching to nvidia, with no luck)
10:45 kodomo: Antoine-: Not an expert, myself, but if I got this right, sending audio through your NVIDIA card would only make sense, if its connectors were actually interconnected with the outside - which would imply the presence of a hardware mux, which is quite unlikely, these days. (check if your /sys/kernel/debug/vgaswitcheroo/switch lists an audio client for the nvidia - I would guess it doesn't)
10:46 kodomo: Antoine-: check https://nouveau.freedesktop.org/wiki/Optimus/ and https://01.org/linuxgraphics/gfx-docs/drm/gpu/vga-switcheroo.html
10:47 kodomo: You're probably using the nvidia to offload work from the intel one, rather than connecting it to the outside.
10:49 Antoine-: kodomo: It seems like /sys/kernel/debug/vgaswitcheroo doesn't exist
10:50 Antoine-: So you believe my issue isn't related to nvidia drivers then
10:52 kodomo: Antoine-: If it doesn't exist, you haven't compiled the multihead support into your kernel, I guess. But its presence is not crucial for your problem... The paraphrasation of my statement would be: Your problem is most likely not relating to nouveau, because the sound will likely not be output via the NVIDIA device, anyway ;)
10:53 kodomo: read the two links I posted to understand what I mean
10:56 Antoine-: Ok :)
10:57 Antoine-: Thanks for your help!
10:57 kodomo: np - good solution hunting to you!
12:31 imirkin: Antoine-: which kernel are you using?
12:31 imirkin: kodomo: did updating kernel help? it seems like accel isn't loading, you should have stuff in dmesg about that
13:35 AndrewR: imirkin, hi ...
13:38 kodomo: imirkin: got interrupted by more urgent work, doing this as a side activity, right now, will tell you the result. Thank you very much for asking :)
13:39 imirkin: AndrewR: hey. i haven't looked at your thing at all yet, sorry
13:39 imirkin: AndrewR: likely i won't have time during the day today either, maybe tonight
13:43 imirkin: Antoine-: also, is the HDMI output on the nvidia card, or the intel one?
13:47 AndrewR: imirkin, ok, i have my hacks saved on hdd
14:11 imirkin: AndrewR: if i don't come up with something, i'll add a CAP to disable this thing
14:11 imirkin: ideally i can figure out what we're not flushing
14:11 imirkin: or where we're not waiting
14:38 kodomo: imirkin: upgraded to 5.4.28 - same error: nvc0_screen_create:1111 - Error allocating PGRAPH context for M2MF: -16 (and glxinfo becomes defunc - killed it by -9 )
14:38 imirkin: pastebin dmesg
14:39 imirkin: do you have the firmware btw?
14:40 kodomo: hrm... I see an init failed with nouveau... this will take a moment to paste the entire dmesg ... gimme a sec...
14:41 imirkin: just copy/paste, no?
14:45 kodomo: page by page.... there's tons of errors, I'm afraid... and I already did a copy-paste error :P - maybe I should limit myself to the last page or so...
14:45 imirkin: huh?
14:46 imirkin: page by page?
14:46 imirkin: why not just do the whole thing at once? seems easier
14:46 kodomo: imirkin: because it's too long to fit the window size of my xterm?
14:46 imirkin: and your terminal doesn't support scrolling?
14:47 imirkin: i know it's a fairly novel feature, but i expect any terminal written since ~1980 to have it
14:47 kodomo: xterm is... *cough* ... old... *g*
14:47 imirkin: yeah
14:47 imirkin: and it supports scrolling
14:47 imirkin: you're doing something wrong.
14:47 imirkin: just run "dmesg"
14:47 kodomo: not by clicking, it doesn't...
14:47 imirkin: then scroll up, double click on the first line
14:48 imirkin: er, triple-click
14:48 imirkin: then scroll all the way to the bottom
14:48 imirkin: and left-click on the last line
14:48 imirkin: scroll wheel should work
14:48 imirkin: if you don't have one, shift page-up/down do it
14:49 imirkin: you can also double/triple-click + drag
14:49 kodomo: wow! I live in modern times 8D
14:49 kodomo: *lol* I've been using it so long and didn't know it could do that ^^
14:50 imirkin: there's a maximum scrollback buffer size which you can increase too
14:50 imirkin: via .Xresources - you can look at the man page for the various settings
14:50 kodomo: that I know - and I might need to... I'll first paste what I have
14:50 imirkin: i usually set it to 10000 -- the default tends to be too small for my taste
14:50 kodomo: | less tells me that dmesg got overflowed, anyway...
14:50 kodomo: ...boot output's already gone
14:51 kodomo: https://pastebin.com/70bQYe2j
14:51 imirkin: laptop?
14:52 kodomo: yes - as I wrote: Lenovo T480s
14:52 imirkin: you assume i read carefully
14:52 imirkin: you probably need some patches which landed recently around PM
14:52 kodomo: w/ MX150 rev.a1 (1030 - Pascal, afaik)
14:52 imirkin: these are available in v5.7-rc1, but should also be making their way to various stable trees
14:53 imirkin: you're on the 5.4 series ... let me check what's going on there...
14:53 imirkin: v5.4.35 (released 6h ago) should have it.
14:54 kodomo: imirkin: thx... the 5.7 doesn't even seem to have a gentoo .ebuild, yet... and when I tried to make clean in 5.6, it failed, telling me there was a rule missing (not exactly confidence-inspiring ;) )
14:54 imirkin: my guess is that will resolve all your issues
14:54 imirkin: you should be able to trivially make a v5.4.35-based gentoo ebuild, i think
14:54 imirkin: just copy the old one, bump the versions, and that's it
14:54 imirkin: (assuming you already have a local overlay set up, if not, i'd strongly encourage it)
14:54 kodomo: 5.4.35... gonna check whether I can lay my hands on that one...
14:55 imirkin: https://packages.gentoo.org/packages/sys-kernel/vanilla-kernel
14:55 imirkin: not there yet
14:55 kodomo: yup - willdo (as I said: interleaving work items, right now - might take some time - but will do)
14:55 imirkin: but like i said, you can make an ebuild yourself for it pretty easily
14:55 kodomo: Thank you :) will report back once I've tried it.
14:56 imirkin: also v5.6.7 should have the patches in question as well
18:41 imirkin: skeggsb: does the gpu snoop cpu reads/writes for "gart" bo's on nv50+?
21:16 skeggsb: imirkin: yep
21:17 imirkin: =/
21:17 imirkin: it would have been such a nice explanation if it was using that NOSNOOP thing
21:17 imirkin: basically i'm having trouble with coherent vbo's
21:18 imirkin: i flush the cache, but it appears to not be enough
21:18 imirkin: going to try a WFI
21:18 imirkin: noooope
23:45 aplund: The feature matrix lists opencl as a WIP. Is there any available code?