06:45 fdobridge_: <o​rowith2os> any Turing+ NVIDIA workstation GPUs are free game for NVK too, right?
06:49 fdobridge_: <o​rowith2os> looking at ebay and the like and hoping I can buy something Soon™ 🥹
06:51 fdobridge_: <o​rowith2os> well, that or some Insanely Old GPUs that don't need firmware to reclock
06:52 raket: Maybe unrelated, i'm using gtx 780ti to play something that uses opengl and i'm more than happy about it :-)
06:53 fdobridge_: <o​rowith2os> it would appear workstation stuff (that's Turing+) is Still Expensive anyways
06:53 fdobridge_: <o​rowith2os> I'd be better off with some normal gaming cards
06:54 fdobridge_: <o​rowith2os> or, hear me out.... somehow I actually save my money
06:54 fdobridge_: <o​rowith2os> 😳
08:29 fdobridge_: <k​arolherbst🐧🦀> heh 😄
08:29 fdobridge_: <k​arolherbst🐧🦀> @orowith2os if you promise to work on Nouveau, I could give you a GPU 😄
08:31 fdobridge_: <k​arolherbst🐧🦀> though the beefiest I got I could hand over is the GeForce GTX Titan, comparable to a 2060 with less features
10:01 fdobridge_: <S​id> testing counts as working on it, right?
10:01 fdobridge_: <S​id> right?
10:01 fdobridge_: <S​id> :p
10:04 fdobridge_: <k​arolherbst🐧🦀> let me guess, the "testing" you speak of is mostly playing games for hours 😛
10:05 fdobridge_: <S​id> I can also do compute workloads :D
10:16 fdobridge_: <k​arolherbst🐧🦀> mhhh
13:59 fdobridge_: <z​mike.> ```
13:59 fdobridge_: <z​mike.> Slowest tests:
13:59 fdobridge_: <z​mike.> dEQP-GLES3.functional.shaders.texture_functions.texturesize.sampler3d_fixed_fragment (14104.74s)
13:59 fdobridge_: <z​mike.> ```
14:00 fdobridge_: <k​arolherbst🐧🦀> could be worse
14:03 fdobridge_: <m​ohamexiety> damn, you could fit in an entire [multithreaded] Vk CTS run in that \:o
14:40 fdobridge_: <!​DodoNVK (she) 🇱🇹> Is it slow on every driver? 🤔
15:06 fdobridge_: <o​rowith2os> I'd 100% be willing to bug test the things that are known to almost-work in daily use
15:06 fdobridge_: <o​rowith2os> And provide logs and assistance and the like
15:07 fdobridge_: <o​rowith2os> But I also just got approved for a second job, so my time there is limited
15:08 fdobridge_: <k​arolherbst🐧🦀> I see...
15:35 EisNerd: any chance to get sth useful regarding vulkan, nvk with a kepler card? And how to enable it?
15:37 DodoGTA: EisNerd: It's currently experimental for pre-Turing so you have to set an environment variable
15:39 fdobridge_: <g​fxstrand> EisNerd: Kepler does work if you set `NVK_I_WANT_A_BROKEN_VULKAN_DRIVER=1` but it's a bit sketchy.
15:40 karolherbst: I wished I had more time for stuff 🙃
15:41 EisNerd: fdobridge_: no need to recompile?
15:41 karolherbst: correct
15:41 fdobridge_: <g​fxstrand> Nope! Just set the environment variable and off you go.
15:42 EisNerd: NVK_I_WANT_A_BROKEN_VULKAN_DRIVER=1 vulkaninfo should work? on a optimus setup?
15:42 karolherbst: in theory
15:43 EisNerd: it still shows me only the intel
15:43 karolherbst: what kernel version are you running on?
15:43 karolherbst: you'll need something super recent like 6.6 or so?
15:45 EisNerd: 6.7 and 23.3.2 mesa
15:45 EisNerd: with intel and nouveau
15:46 karolherbst: mhh, you'd have to check if your distribution enables nvk at all
15:46 karolherbst: like if there is an icd file inside /usr/share/vulkan/icd.d/
15:46 EisNerd: configure log?
15:47 EisNerd: only intel icd files
15:47 karolherbst: yeah.. then you probably don't have nvk installed at all
15:47 karolherbst: the files should start with nouveau_icd.
15:47 EisNerd: ok how to change this?
15:48 EisNerd: some configure I guess
15:48 karolherbst: yeah
15:48 karolherbst: "nouveau-experimental" needs to be added to `vulkan-drivers`
15:49 EisNerd: ??
15:49 EisNerd: guess I need to check configure phase
15:49 karolherbst: yeah
15:49 karolherbst: check the packaging and see if you can do something there. Depending on the distribution it's trivial to do that or not
15:49 karolherbst: but there might be third-party repositories enabling it
15:50 karolherbst: but that all depends on the distribution you are on
15:50 EisNerd: it is trivial I would guess => Gnetoo FTW
15:51 karolherbst: ahh yeah
15:52 EisNerd: anything needed in the kernel config?
15:52 Sid127: nope
15:53 Sid127: also hello funny people in my terminal
15:53 EisNerd: -Dvulkan-drivers=intel,intel_hasvk should also list nouveau?
15:53 karolherbst: EisNerd: "vulkan_enable video_cards_nouveau nouveau-experimental" would need to be added where all those vulkan_enable things are I think
15:53 karolherbst: Sid127: hiii
15:54 karolherbst: (in the ebuild)
15:54 Sid127: EisNerd: -Dvulkan-drivers=intel,intel_hasvk,nouveau-experimental should do it
15:54 EisNerd: ok I'll try
15:55 Sid127: I wanna test something out for nouveau since I have some free time rn but I don't wanna test games again because I already know what happens :\
15:55 Sid127: any ideas?
15:55 karolherbst: somebody might want to add it to gentoo properly, but not sure if nvk is ready enough for general usage yet :D
15:55 Sid127: any weird kernel module thing that needs testing? patches? idk
15:55 karolherbst: Sid127: always hard to answer such questions, but worst case you can plug in all the displays you have and see if that works :D
15:56 Sid127: I have exactly 1 display :D
15:56 karolherbst: mhhh
15:56 karolherbst: TVs or other funky hardware?
15:56 Sid127: do have a dummy edid plug though, supports upto 4k@60
15:56 karolherbst: same
15:56 Sid127: nothing in the dorms sadly
15:56 Sid127: unless I can use my phone as a display, but that's unlikely
15:57 karolherbst: yeah...
15:57 EisNerd: I'd guess it should be a onelliner
15:57 EisNerd: vulkan-drivers : intel,intel_hasvk,nouveau-experimental
15:57 karolherbst: Sid127: could always see how well zink works driving a desktop
15:57 EisNerd: yepp
15:57 Sid127: could do, yeah
15:57 Sid127: could also test how well nvk runs wlroots based compositors
15:57 Sid127: I know sway has a vulkan renderer
15:58 karolherbst: ahh yeah, that would be helpful
15:58 Sid127: so, plasma wayland + zink or sway vulkan
15:58 karolherbst: sounds like a good idea
15:59 Sid127: both it is then
15:59 karolherbst: yeah.. so if we really want to go with zink as the main GL impl, we shouldn't regress running a desktop and stuff :D
15:59 Sid127: now to see how the heck I can get plasma wayland to render on my dGPU while also using zink
15:59 Sid127: e-e
15:59 karolherbst: uhhh
15:59 karolherbst: put your laptop into discrete only mode in the firmware :D
16:00 Sid127: that doesn't output anything on the laptop display
16:00 karolherbst: mhhhh
16:00 Sid127: I did try it a while ago to see if it'd improve framerates
16:00 karolherbst: it should tho
16:00 Sid127: sadly doesn't :\
16:00 Sid127: damn you Acer from 2019
16:00 karolherbst: sounds like a nouveau bug to me then
16:00 karolherbst: or just broken on the firmware level?
16:00 Sid127: oh, it doesn't even on proprietary
16:01 karolherbst: pain
16:01 Sid127: it renders to hdmi, not laptop display
16:01 karolherbst: oh no...
16:01 karolherbst: seems like the internal display doesn't mux then...
16:01 karolherbst: however
16:01 karolherbst: plasma would still run on your dGPU this way :P
16:01 Sid127: something like that
16:01 Sid127: well, it sure would, but I'd have no way to see what's wrong
16:02 karolherbst: mhh?
16:02 karolherbst: I thought it would then just display on your HDMI display?
16:02 Sid127: I'd have no way to see if it renders correctly
16:02 Sid127: it would if I had an HDMI display e-e
16:02 karolherbst: ohhhh
16:02 karolherbst: I see
16:03 Sid127: yeah, I've only got the laptop panel with me
16:03 karolherbst: mhhh
16:15 EisNerd: loader interface version should be 5
16:15 EisNerd: but is 4
16:16 fdobridge_: <t​om3026> well dont forget DRI_PRIME=1
16:17 Sid127: vulkaninfo should show it without that too
16:18 EisNerd: it does
16:18 Sid127: loader interface warning can be ignored for now iirc
16:18 EisNerd: but only vulakn 1.0 version
16:23 EisNerd: WARNING: NVK is not a conformant Vulkan implementation, testing use only. \n Selected GPU 0: Quadro K2100M, type: DiscreteGpu
16:23 EisNerd: vkcube, works
16:25 EisNerd: now I only need to get the missign text in heroic-launcher fixed, then I can try sth real
16:41 fdobridge_: <z​mike.> @gfxstrand how far off do you think you are from getting your common shobj thing landed? I think we're about to duplicate it all in radv
17:05 i509vcb: I need to find some time to try to make T234 run on nouveau. It's GSP based and the firmware does exist. The only issue is that the display hardware is different from past tegra chips...
17:09 fdobridge_: <g​fxstrand> Hoping to have an MR posted today
17:10 fdobridge_: <g​fxstrand> I'm doing a bit of cleanup right now. Will do another CTS run (last one looked mostly okay) and add some comments places and post it.
17:16 fdobridge_: <z​mike.> Awesome
17:55 fdobridge_: <a​irlied> T234 is not gsp based at all
18:25 fdobridge_: <m​akinbacon21> is it not
18:26 fdobridge_: <m​akinbacon21> it looked from driver that it did some gsp-ish stuff
18:26 fdobridge_: <m​akinbacon21> but loaded that weird acr gsp thing
18:28 fdobridge_: <m​akinbacon21> been looking into it
18:29 fdobridge_: <m​akinbacon21> the display side of things is essentially the same as desktop
18:29 fdobridge_: <m​akinbacon21> the render side is still nvgpu
18:29 fdobridge_: <m​akinbacon21> i figure the render stuff would prolly mostly be gp10b kang with changes for ampere
18:30 fdobridge_: <m​akinbacon21> also fwiw i think tegradrm has been getting changes in nvidia's sources, so possible the display side of things will resolve itself to some extent
18:31 fdobridge_: <m​akinbacon21> https://nv-tegra.nvidia.com/r/gitweb?p=linux-nv-oot.git;a=history;f=drivers/gpu/drm/tegra;h=cf22c7be2188cab302096166a5d6d7a20e84cef8;hb=f255a170a7d1603400812a301bb88f2ec3043b36
18:31 fdobridge_: <m​akinbacon21> but i guess thats more vic and nvenc and not the display stuff itself
18:43 i509vcb: acr gsp stuff?
18:44 i509vcb: I guess the 234 is not GSP based then, but at least nvgpu exists for reference
19:03 fdobridge_: <k​arolherbst🐧🦀> @airlied why shouldn't T234 be tegra based though?
19:03 fdobridge_: <k​arolherbst🐧🦀> @airlied why shouldn't T234 be gsp based though? (edited)
19:03 fdobridge_: <k​arolherbst🐧🦀> it's an ampere one
19:03 fdobridge_: <k​arolherbst🐧🦀> or are the tegra ones just special?
19:05 i509vcb: Yeah this discussion is confusing since I definitely saw gsp stuff in nvgpu, but I'm certain still lost lol
19:10 i509vcb: Anyways I do have T234 hardware so I guess if someone else has something I'm around
19:44 fdobridge_: <z​mike.> more simple unit test fails https://gitlab.freedesktop.org/mesa/mesa/-/issues/10426
19:58 fdobridge_: <a​irlied> @karolherbst is t234 is orin, the software nvidia releases for it is not GSP based, and it's not supported by their current GSP effort, it might have a GSP, but nvidia aren't currently using it like they do on discrete
19:59 fdobridge_: <k​arolherbst🐧🦀> yeah, fair
19:59 fdobridge_: <a​irlied> @karolherbst if t234 is orin, the software nvidia releases for it is not GSP based, and it's not supported by their current GSP effort, it might have a GSP, but nvidia aren't currently using it like they do on discrete (edited)
20:02 i509vcb: T234 is Orin, yes
20:03 i509vcb: I guess if nvidia isn't using GSP in it's releases for T234 then I guess there isn't some weird signed reclocking madness then
20:05 fdobridge_: <a​irlied> nvgpu I assume supports it
20:55 fdobridge_: <m​akinbacon21> yes, but only r34 and up of
20:55 fdobridge_: <m​akinbacon21> yes, but only r34 and up ofc (edited)
20:56 fdobridge_: <m​akinbacon21> there are fw bins for t234 that comprise a blob named acr-gsp
20:57 fdobridge_: <m​akinbacon21> @airlied there is a ton of gsp stuff in nvgpu
20:57 fdobridge_: <m​akinbacon21> looks like support was added for gv100
20:57 fdobridge_: <m​akinbacon21> https://gitlab.incom.co/search?group_id=5&project_id=518&repository_ref=lineage-21&scope=commits&search=gsp
20:58 fdobridge_: <m​akinbacon21> and they very clearly support ga10b
20:58 fdobridge_: <m​akinbacon21> and they very clearly support ga10b with gsp stuff (edited)
20:59 fdobridge_: <m​akinbacon21> example
20:59 fdobridge_: <m​akinbacon21> https://gitlab.incom.co/CM-Shield/android_kernel_nvidia_nvgpu/-/commit/2a98d20263ef4355e2d2c38b29350afc39a75f7c
20:59 fdobridge_: <m​akinbacon21> it looks like the split acr gsp fw thing was originally implemented with this: https://gitlab.incom.co/CM-Shield/android_kernel_nvidia_nvgpu/-/commit/4efdc362175c67f93d3546727c8825686619c1cb
20:59 fdobridge_: <a​irlied> yeah I think they have fw for the gsp processor just not the same as elsewhere
20:59 fdobridge_: <m​akinbacon21> it looks like the split acr gsp fw thing was originally implemented with this: https://gitlab.incom.co/CM-Shield/android_kernel_nvidia_nvgpu/-/commit/4efdc362175c67f93d3546727c8825686619c1cb
20:59 fdobridge_: <m​akinbacon21> then iterated upon (edited)
21:00 fdobridge_: <m​akinbacon21> it looks like the split acr gsp fw thing was originally implemented with this: https://gitlab.incom.co/CM-Shield/android_kernel_nvidia_nvgpu/-/commit/4efdc362175c67f93d3546727c8825686619c1cb
21:00 fdobridge_: <m​akinbacon21> then iterated upon (edited)
21:00 fdobridge_: <a​irlied> and they are using it for some features
21:00 fdobridge_: <m​akinbacon21> ya
21:02 fdobridge_: <m​akinbacon21> but yea not sure how much would be needed to add support
22:17 fdobridge_: <a​irlied> #define LSF_FALCON_ID_GSPLITE 🙂
22:17 fdobridge_: <a​irlied> kinda implies a smaller gsp
22:21 Lyude: BTW - is there any reason we're on a slightly older version of the kernel? or has stuff just not been rebased quite yet? (noticed that it's asking for slightly older versions of the rust tools)
22:44 Lyude: eyyy rust analyzer works :)
23:40 Lyude: ah, I've gotta implement a type for a platform driver I see
23:54 Lyude: airlied, dakr: either of you know how to fix this? ERROR: modpost: "rust_build_error" [drivers/gpu/drm/nova/nova.ko] undefined!