00:05reepca: Blargh. So I found out after waiting for an HDMI->VGA adapter that works for about 20 days that my 8400 GS, despite having 3 outputs, can only actually display 2 independent ones at a time. Luckily I happened to get a second one, not knowing that the problems with the first one were caused by a bad HDMI->VGA adapter. And now I am attempting to get three monitors all running at once using both cards. What do I need to know/do?
01:03imirkin: reepca: there are only 2 CRTC's on that hardware. you have to move up to kepler before you get 4 CRTC's
01:04imirkin: reepca: do you have 1 or 2 GPUs?
01:04reepca: imirkin: 2 GPUs
01:05imirkin: ah. then you want reverse-prime to use one of the outputs on the other gpu
01:05imirkin: read the bit about "Using outputs on discrete GPU"
01:05imirkin: your situation is analogous
01:09reepca: xrandr --listproviders indicates that both cards have "name" nouveau - does --setprovideroffloadsink accept a different way of identification?
01:09reepca: err wrong flag, meant --setprovideroutputsource
01:33reepca: Aha! it works! Many thanks imirkin, how would I go about causing this to take effect on startup?
02:51reepca: So, how would one go about getting --setprovideroutputsource sorts of options into an xorg configuration file?
02:52coredump: rhyskidd: I think I will give it more some months before trying again then.
02:52coredump: It's not I am needing this card just now anyway
15:08imirkin: reepca: there's no way to tell xorg to do it. many DE's will have a way of doing it when they start, i think.
15:10reepca: imirkin: alright, that'll save me some time looking in the wrong place.
15:12imirkin: you could obviously have something in your, e.g., .xsession
15:12imirkin: or whatever thing happens on startup for you
15:19imirkin: reepca: and i assume you worked out that you could just say "--setprovideroutputsource 1 0"
15:20imirkin: that works in like 99.999999745% of situations
15:20reepca: huh, actually I didn't figure that out, I was using the hex identifier.
15:21imirkin: that works too :)
16:15rhyskidd: imirkin: hrrmm, not seeing any of the known PGOB registers touched by the blob on GP107M
16:16imirkin: well, it was just a theory
16:16imirkin: sounded pretty familiar to the earlier situation
16:16imirkin: where everything works but GR just refuses to power up
16:35Tom^: karolherbst: sure i made a hack to control the pstate from cli with sudo/root or setuid :p
16:35Tom^: karolherbst: https://github.com/gulafaran/nouveauctl/blob/master/src/main.c
17:06whompy: imirkin: I'm forgetting which video player you prefer for nouveau users
17:07imirkin: whompy: i use mplayer.
17:08whompy: Sweet. That has been my perferred one as well. Time to remember how to configure hardware acceleration.
17:14whompy: Ha, I managed to forget about that page in its entirety
17:27imirkin: well, hope it helps
17:27imirkin: note that it's not perfect. there's an outstanding bug that i've been needing to investigate for ages with h264 and haven't quite had the willpower to do it
17:28imirkin: or rather, i've already spent a ton of time investigating it without results
17:28imirkin: but that was a long time ago
17:28whompy: Yeah, I recall seeing that way back when. So far it's working well. I screwed something up on my side with something entirely different, but the decoding seems fine
17:30imirkin: the issues seem to hit esp in what seem to be high-motion scenes
17:30imirkin: e.g. when the whole screen pans slowly
17:30imirkin: which in turn generates a TON of motion vectors
17:30imirkin: instead of replacing whole MB's with new data
17:45vedranm: hov interesting are bug reports "running piglit makes my dmesg scream on 9500GT"?
17:46vedranm: I presume most of those are known bugs
17:49imirkin: vedranm: esp if you're not running it single-threaded
17:50imirkin: vedranm: i'm not specifically aware of any issues, but unless you have specifics like "test X causes message Y", not worth filing a bug.
17:50imirkin: also, sometimes it's fine for a message to be generated
17:50imirkin: so check here before filing bugs
17:53vedranm: imirkin: OK
18:55mangix: karolherbst: any idea on why this commit breaks 3D for me? https://github.com/karolherbst/nouveau/commit/e9e1c027048f6f24952496c2a591dbab9515c1f8
18:55mangix: your branch works when i revert that
18:57imirkin: "breaks 3D"?
18:57imirkin: that change messes up EDID for a lot of people
18:57mangix: GNOME does not start
18:57mangix: that's what i meant
18:58imirkin: welcome to the "lot of people" :)
19:00mangix: hope it gets fixed by the time fedora gets 4.12
19:00karolherbst: mangix: no clue
19:00imirkin: 4.12 is already out and no fix.
19:01imirkin: one of a mountain of things on ben's list
19:01mangix: fedora gets it at .5 i think
19:02mangix: i happened to install the blob a while back
19:02mangix: big mistake
19:02mangix: now i need to reinstall fedora
19:03imirkin: while i can't speak to its integration with various distros, the nvidia blob is a pretty solid driver stack, esp in terms of reliability
19:03mangix: that thing is actually cancerous
19:03mangix: no way
19:03imirkin: that said i haven't actually run it myself in several years
19:04mangix: it breaks backlight control for me. also, i can't turn the screen off
19:04imirkin: but it doesn't crash :)
19:04mangix: it just stays on
19:04mangix: imirkin: if i put efforr i can probably crash it
19:05mangix: also, if i uninstall it, i get massive tearing in every application
19:05imirkin: whereas with nouveau, if you put effort in, it might not crash :)
20:24karolherbst: imirkin: may I remind you of my gallium precise series?
20:39karolherbst: mupuf: https://gist.github.com/karolherbst/b728e12dce153a775728c6199a66fa92
22:17karolherbst: mhh, those stars are wrong though, cause 0f is still selected, just the clocks are lower
22:18karolherbst: but maybe that is fine