09:58dboyan: karolherbst: I finally had a working system with nvidia 340 driver. But discovered that it only supports nvenc sdk 4, which doesn't compile. I decide to write my own demos.
10:00dboyan: I think I can some extra info from older versions without digging into uvm. I seem to find output buffer location of VENC object in the command stream you gave me.
10:00dboyan: Because I saw reading in that buffer
12:46karolherbst: dboyan: nice :)
13:47dboyan: karolherbst: well, the codec cannot initialize on 340.102. Also got the original sdk v4 to compile, and it failed with the same error.
13:48dboyan: The very same program runs fine on 375.26, mysterious
14:38ystreet00: this is for nvenc?
14:39ystreet00: 340 sounds too old for that
14:40ystreet00: the different sdk versions need at least some driver version
14:45dboyan: ystreet00: the readme nvenc sdk v4 says it can run on "R340" drivers
14:48dboyan: And the v4 demo cannot run as-is on 375.26. Now i doubt the so-called "backward compability" nvidia said.
14:49dboyan: I hacked the v4 demo, captured an mmt trace, hoping that i can find something interesting
14:51dboyan: The sad thing is that demmt doesnt't quite understand the fifo cmds. Got no annotation at all
16:40karolherbst: dboyan: yeah, well, you should figure out how those commands work ;) that's part of reeing nvenc
23:01wrl: hey i just updated my system to 4.10.1 and my quadro fx 1800 will not wake up
23:01wrl: the displays went to sleep and i cannot get them to come back on. system is still up, can SSH in
23:02skeggsb: wrl: can you post your Xorg and kernel logs to a pastebin somewhere please?
23:02wrl: skeggsb: sure. kernel logs just dmesg?
23:03wrl: there is nothing showing up in either, actually
23:03wrl: nothing relevant i mean
23:05wrl: Xorg log: https://hastebin.com/pifinatuda.txt
23:06wrl: dmesg: https://hastebin.com/conemuvumu.txt
23:06wrl: lines about the keyboard at the bottom are me unplugging and replugging the kbd thinking it was the problem
23:07skeggsb: wrl: hm, ok, can you boot with "log_buf_len=8M drm.debug=0x14 nouveau.debug=disp=trace" and try to reproduce?
23:07wrl: skeggsb: that as part of the kernel cmdline?
23:07skeggsb: it's not the issue i was expecting it to be :)
23:10wrl: skeggsb: is there a quick command to put the displays to sleep?
23:10skeggsb: xset dpms force off
23:10wrl: and then just moving the mouse or tapping a key should get them to come back on?
23:11wrl: hm. ok, that was fine, couldn't reproduce that way
23:11wrl: may just need to walk away for 10 minutes and see what happens
23:21Echelon9: skeggsb: Hi. Wondered if you had a chance to look at the mem type number for HBM2? Have been talking with karolherbst about my first nouveau patch
23:21skeggsb: haven't had a chance yet
23:22Echelon9: I'm also working on the hwmon deprecation warning in dmesg, as that's fairly straightforward to fix
23:22skeggsb: sounds good :)