00:00AndrewR: karolherbst, kernel_image_methods/test_kernel_image_methods still supposed to fail ? ("ERROR: Returned width did not validate" and few more of those for each format), or I mis-applied your patches?
00:00karolherbst: ehhh...
00:00karolherbst: I think it should work?
00:00karolherbst: dunno
00:01karolherbst: no idea if nv50 misses things
00:01karolherbst: but probably it does
00:04AndrewR: karolherbst, ok ....
00:04karolherbst: atm I just don't have the means of checking that stuff with nv50
00:04karolherbst: but this will probably change soon
00:05karolherbst: but it might be easier for pmoreau to just look into it and finally submit patches for nv50 :p
00:39AndrewR: https://pastebin.com/nMJ4BPpF - current piglit problems on nv92 (08 oct without compute patches, 25 oct - with them + some fixes)
00:48AndrewR: https://cloud.mail.ru/public/5DH8/5eVQKvWCo - various results of piglit tests from my machine ....
10:18greg904: Hi, when I use the nouveau driver with my laptop with a GTX 1070, changing backlight brightness with /sys/class/backlight/nv_backlight/brightness doesn't work (the value changes but nothing happens). I have a mmiotrace of what the proprietary driver does (I don't know if that's useful): https://pastebin.com/8zUzuruS. Anyone has an idea?
10:19RSpliet: greg904: do you have any other means of changing the backlight?
10:20RSpliet: do the hotkeys on your laptop work?
10:20RSpliet: (assuming it's a laptop...)
10:20greg904: no, the hotkeys don't work
10:21greg904: (but also i'm not running a DE so maybe that's expected)
10:22greg904: i tried this on sway (wayland) and X with i3
10:41greg904: I forgot to mention that I think my laptop's internal display is connected with DisplayPort. Maybe nouveau doesn't support setting brightness for a DisplayPort display?
11:30RSpliet: This may be a factor. Brightness is an annoying area because there's shared responsibilities between the UEFI, nouveau and often a third party backlight module too. It's always a surprise who needs to do what
12:59greg904: The intel i915 driver uses the "drm_dp_dpcd_write" function with "DP_EDP_BACKLIGHT_BRIGHTNESS_MSB" to set the backlight brightness and when I patched the nouveau code to use it as well, it worked on my laptop. I don't know what what to do from here since I am not sure when the original code path for DP in nouveau_backlight.c must be used and when
12:59greg904: the drm_dp_dpcd_write must be used instead (I'm also not sure if this makes any sense since I don't know anything about kernel dev/nouveau/DP)
13:02karolherbst: greg904: amoled screen?
13:03karolherbst: anyway, Lyude is looking into this area atm anyway and I think Lyude has a few patches to fix it all up
13:04greg904: not an amoled screen
17:50Lyude: greg904: yeah I'm working on some patches, I'm currently dealing with troubleshooting some other issues with eDP on nouveau that are making it so that I can't test my brightness patches properly
18:14greg904: ok great, hope you will solve the issues.
18:17Lyude: greg904: hopefully :), feel free to check back in a week or two and hopefully I'll have more progress by then, my email is lyude at redhat dot com (in case I forget to poke you when I've made progress :)
19:02FulanoDaSilva: Hello, i've never used nor understand IRC but i know that this is not the right place for this, so firstly, my apologies.Please ignore if something is wrong, i'm brazilian (don't speak english). I would like to buy the graphics card that has the best possible compatibility at this moment, but as nouveau page don't state this in a clear, simple
19:02FulanoDaSilva: manner (it has the FeatureMatrix but it doesn't help for non technical users) and i couldn't find the answer in any other place so i had no choice but to post this here. At this time which one is considered to be the best option? Again, my apologies.
19:30rowbee: it's me again. update-glx is causing me problems... when the nvidia driver is loaded /usr/lib/glx is supposed to point to /usr/lib/nvidia, but when nouveau is loaded it's supposed to point to /usr/lib/mesa-diverted
19:30rowbee: anyway i can do that simply?
19:30rowbee: if it's set up the wrong way the following happen:
19:31rowbee: nouveau loaded with /usr/lib/nvidia link: opengl stuff doesn't work, can fix with sudo update-glx --set glx /usr/lib/mesa-diverted
19:31linkmauve: FulanoDaSilva, you’ll probably have better chance with AMD or Intel GPUs.
19:31rowbee: nvidia loaded with /usr/lib/mesa-diverted link: xorg completely shits the bed and i have to do the update-glx AND reboot
19:31linkmauve: Nvidia is really hostile to people using their hardware with free drivers.
19:32rowbee: i wonder how much it'd bork my system to use a direct "ln" command instead of going through update-glx...
19:34ericonr: FulanoDaSilva: hello fellow brazilian :p message me if you need any language help. anyway, you probably want the GPUs with reclocking support
19:34ericonr: if you really want to stick with nvidia
19:38ericonr: reclocking support means they can actually run at close to their full speed. The newer ones run quite slow :/
19:38rowbee: just switched to /usr/lib/mesa-diverted with xorg running on nvidia, glxgears still working
19:38rowbee: weird
19:38rowbee: it only breaks if mesa-diverted is linked during boot
19:42rowbee: FulanoDaSilva: amdgpu driver has been pretty good imo, since it's the open source driver that AMD is directly collaborating with
19:52Lyude: it isn't just collaborating, it's the official driver :P
19:52Lyude: (amdgpu-pro also exists. but no one cares about amdgpu-pro)
20:00FulanoDaSilva: Thank you all for the answers. I would not like to use "AMD + it's open source drivers" because as far as i know it will not be 100% free as with "nvidia + nouveau", the last option would not depend on anything proprietary at all
20:00Lyude: no, that's not true
20:00Lyude: unless you are using an older GPU before maxwell 2 you will need closed firmware for things to run on nouveau (which is shipped in linux-firmware by default)
20:01Lyude: just about every modern GPU these days needs firmware
20:07FulanoDaSilva: So can i really use AMD and be AS FREE AS IF I WAS USING NOUVEAU? are you sure?
20:20Lyude: FulanoDaSilva: yes
20:21Lyude: FulanoDaSilva: some of the firmware on the older cards is actually made by us, but pretty much everything maxwell 2 and forward has mandatory firmware signing so we can't upload our own fw unfortuantely
20:31FulanoDaSilva: Which card do you recommend for AMD, and which one would you recommend me if i still choose to go with nvidia? (considering freedom as the main criterion)
20:33FulanoDaSilva: Remembering that the most important factor is the freedom!
20:46xexaxo1: rowbee: consider updating your Debian(based) distro. GLVND has been a thing for a while now, so you don't need the diverted machinery
20:47rowbee: i do have glvnd
20:47rowbee: and i'm on bullseye (testing)
20:47xexaxo1: with glvnd the different drivers can coexist at the same time.
20:47rowbee: the problem is that if i don't update-glx, then i get "extension GLX not supported" when i try to run glxgears with nouveau
23:50Lyude: skeggsb: for adding new NVIF methods (I'm pretty certain I'm going to need one for paneling the edp panel on/off outside of nvkm), does it matter if I allow said methods to be called before an outp has been assigned to a given ior?
23:52skeggsb: Lyude: just hack it in however you like, i need to fix all the disp stuff at some point
23:54Lyude: skeggsb: what kind of changes are you thinking of for that btw?
23:56skeggsb: objects for head/outp/conn, killing the clunky disp methods. and, ripping link training out of supervisor handling and making it explicit
23:56Lyude: oh cool :)
23:56Lyude: was hoping for the last one at some point but haven't had time to explore it further
23:56skeggsb: it's very needed, that was a poor choice made with the best intentions :P
23:57skeggsb: (avoiding a two-stage modeset)
23:58skeggsb: probably some better way to deal with SOR routing too, though i'm not sure exacly what yet
23:59Lyude: btw - any opinions on the twod stuff I asked about the other day, regarding using it for the time being to clear surfaces to black from the kernel module? (also-core surfaces do work :)