01:57 rhyskidd: tu116 doc patches incoming to envytools
01:58 rhyskidd: https://github.com/envytools/envytools/pull/185
01:59 imirkin: a few too many ????'s
01:59 rhyskidd: just for the hda pci id
01:59 rhyskidd: oh, and some aspects of the CE[x] etc
02:00 rhyskidd: we probably need to update elements of that hw/gpu.rst table for a bunch of gpus after ~maxwell
02:00 imirkin: also ... released on feb 22? it's feb 20 :)
02:00 rhyskidd: yup :)
11:02 AndrewR: hello ... i tried to update mesa, and vdpau (presentation part) started to fail with mplayer: mplayer: nv50/nv50_state.c:659: nv50_bind_sampler_states: Assertion `!"unexpected shader type"' failed.
13:46 imirkin: AndrewR: ugh, probably the recent work to use compute in vdpau presentation
14:01 HdkR: Oh? Using compute for vdpau decoding? Is this common code or Nouveau specific?
14:04 imirkin: common code
14:04 imirkin: not for decoding
14:04 imirkin: but for the filtering/presentation logic
14:08 HdkR: ah
14:08 HdkR: Is that opt in for hardware that doesn't have an async compute queue?
14:09 HdkR: Or does it even matter if it is async presenting?
14:09 imirkin: no clue. just saw the commits land, didn't really pay attention
15:10 AndrewR: https://cgit.freedesktop.org/mesa/mesa/tree/src/gallium/drivers/nouveau/nv50/nv50_screen.c - sets case PIPE_CAP_COMPUTE: to 1 ... ?
15:10 AndrewR: so new code added https://cgit.freedesktop.org/mesa/mesa/commit/?id=9364d66cb7f7deb83876a44bb4e29e8105141c16 assumes nouveau can compute on nv50 :}
15:13 imirkin: yeah
15:13 imirkin: that's what i assumed... PIPE_CAP_COMPUTE is overloaded
15:13 imirkin: we should just set it to 0 on nouveau
15:13 AndrewR: https://cgit.freedesktop.org/mesa/mesa/commit/src/gallium/drivers/nouveau/nv50/nv50_screen.c?id=ff72440b40211326eda118232fabd53965410afd ? *when this compute support was added) ...
15:13 imirkin: er, nv50
15:14 imirkin: iirc the last remnants of tgsi-based opencl are gone, so it does zero good
15:14 AndrewR: imirkin, it was harmless until those vl reworks
15:15 AndrewR: imirkin, well, I can hack it locally, and wait for official fix
15:15 AndrewR: imirkin, hack it to 0
15:20 AndrewR: imirkin, https://pastebin.com/RvVV9Zdt fixes mplayer for me
15:21 imirkin: cool
15:21 imirkin: feel free to send a patch =]
15:48 AndrewR: imirkin, may be it can be fixed other way around? Is adding samplers actually hard, or I can just copy nv50_vp_sampler_states_bind (), say? and add case into nv50_bind_sampler_states ...
15:51 imirkin_: if it doesn't need images, we could actually implement the relevant compute functionality on nv50
15:51 imirkin_: i do have one plugged in (G84), so i could play with it
15:53 imirkin_: nope - it needs images.
15:53 imirkin_: iirc there's kinda-sorta image support in nv50 backend now
15:53 imirkin_: but it would need a lot of work to be brought up to today's standards
16:02 AndrewR: imirkin_, then guess i'll stay with compute disabled (tried to hack samplers in, now mplayer just segfaults)
16:03 RSpliet: imirkin_: btw, my bouncer died yesterday evening in an unfortunate bouncing accident. Which is also my mailserver... so that got served too. If you need any further piglit tests on Kepler, best send an e-mail to my uni address.
16:03 imirkin_: RSpliet: ok, will do
16:03 imirkin_: RSpliet: i'm kinda stuck though
16:03 AndrewR: imirkin_, https://pastebin.com/4xz1jcqb
16:04 RSpliet: imirkin_: well, no news is good news as far as I'm concerned...
16:04 imirkin_: AndrewR: lol yeah, that won't work for a lot of reasons :)
16:04 AndrewR: imirkin_ :}
16:54 imirkin_: AndrewR: looks like https://github.com/curro/mesa/commits/nv50-compute has some logic for bringing up compute on nv50
16:54 imirkin_: from 2011 :)
16:58 imirkin_: too bad none of that stuff landed. curro writes really nice code =/
17:04 RSpliet: imirkin_: I thought pmoreau started with reviving some of that for his OpenCL work?
17:04 imirkin_: probably
17:04 imirkin_: but i think just the basics
17:04 imirkin_: not the advanced stuff like images, etc
20:19 ioria: hi, geforce rtx 2060 is supported on linux mainstream kernels ?
20:21 imirkin_: define supported
20:21 imirkin_: afaik modesetting should work with the upcoming 5.0 kernel
20:22 imirkin_: no accel of any kind though - just a dumb framebuffer
20:23 ioria: imirkin_, thank you
20:23 ioria: imirkin_, supported = working :þ
20:24 imirkin_: works great as a space-heater
20:24 imirkin_: so ... supported?
20:24 ioria: imirkin_, guess not ?
20:24 imirkin_: different people need different things
20:24 imirkin_: for people who just need screens to light up -- supported
20:24 imirkin_: for people who want to play 3d games -- not supported
20:24 ioria: imirkin_, i see; tx
20:25 imirkin_: for people who need a space heater -- supported ;)
20:25 imirkin_: although i recommend the GTX 480 for its space-heating capabilities
20:26 ioria: thank you again; bye
20:28 HdkR: If only such a company would release documentation and work with the community to have Day 0 support of hardware
20:29 imirkin_: they make great space heaters ... what can i say
20:29 imirkin_: others have tried, but no one rises to the occasion like they do :)
20:29 HdkR: :D
20:29 imirkin_: and it can get *cold* in nyc :)
20:29 HdkR: The RX 295X2 was close
20:30 HdkR: R9*
20:30 imirkin_: yeah ... both companies have put out some good stuff over the years
20:30 imirkin_: FX 5800 or whatever was good for the day
20:30 HdkR: My threadripper system puts out some pretty good heat
20:32 HdkR: Just need to get the CPU and GPU running full out at once
20:32 HdkR: Tall order