00:17 fdobridge: <z​mike.> @airlied probably be a good thing if https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/28207 was merged before you flipped the zink switch
00:21 fdobridge: <a​irlied> might be, I just left it in Draft so I didn't forget about it, not like nvc0 has any testing 😛
00:21 fdobridge: <a​irlied> I think CI would need a newer kernel with hand applied patches to test this
00:22 fdobridge: <a​irlied> at least to test modifiers
00:23 fdobridge: <z​mike.> yeah, but even having a kernel without modifiers would be better than having no ci at all
00:24 fdobridge: <z​mike.> I mean, modifiers don't really affect cts anyway since we use surfaceless in ci
00:35 fdobridge: <c​langcat> Good news it's on the NVIDIA GPU XD
00:35 fdobridge: <c​langcat> https://cdn.discordapp.com/attachments/1034184951790305330/1240462537611284480/grafik.png?ex=6646a643&is=664554c3&hm=cadf0bc89695b263f8093166d7eac3eda585d82edbe425beff25bb1de8ce5c73&
00:46 fdobridge: <a​irlied> oh wow, started sddm, got a one column of garbage rendering, blamed zink/nvk, tried nvc0, still there
00:46 fdobridge: <a​irlied> stays throughout the plasma session
01:12 fdobridge: <c​langcat> On a positive not VK cube runs
01:12 fdobridge: <c​langcat> https://cdn.discordapp.com/attachments/1034184951790305330/1240471891945324616/grafik.png?ex=6646aef9&is=66455d79&hm=957c8389b1def611a756e53d5e05ff6e726f503173e436345d5f4ba61b9fafb0&
01:12 fdobridge: <c​langcat> so I think I just dunno how to use wine
01:18 fdobridge: <c​langcat> so I think I just dunno how to use wine. Either way nvk is working. (edited)
02:56 fdobridge: <a​irlied> @skeggsb9778 https://lwn.net/Articles/931943/ that was something I half remembered on how we should do things properly
03:48 fdobridge: <s​keggsb9778> @airlied did that end up going anyway? i'm all for more sane dma apis
03:49 fdobridge: <s​keggsb9778> @airlied did that end up going anywhere? i'm all for more sane dma apis (edited)
03:49 fdobridge: <a​irlied> not yet, but I haven't looked to closely, but I think it's where it should end up going to fix all these messy sg page vs not a page things
03:52 fdobridge: <s​keggsb9778> it'd be pretty nice to have the rust versions be sane from the beginning at least
06:41 fdobridge: <k​arolherbst🐧🦀> oh.. I have to run the android emulator on top of zink, because if there is one multithreading application crashing your driver quickly, it's that one
08:35 fdobridge: <r​inlovesyou> 24.1 got locked in yesterday iirc so any work now would be for 24.2
09:07 fdobridge: <v​alentineburley> Fixes get backported to releases, so they were asking if that counted as a fix
09:14 fdobridge: <p​ixelcluster> the feature branchpoint was a while ago
09:14 fdobridge: <p​ixelcluster> start of april
09:15 fdobridge: <r​inlovesyou> Ah, yeah that's definitely more than just a fix
11:05 fdobridge: <p​avlo_kozlenko> you use nouveau-vulkan-git?
11:05 fdobridge: <p​avlo_kozlenko> or stock arch mesa
11:06 fdobridge: <p​avlo_kozlenko> or stock arch mesa? (edited)
11:06 fdobridge: <p​avlo_kozlenko> 24.0.7
11:24 fdobridge: <c​langcat> I tired mesa git master. Vulkan and open GL drivers and Faith's kernels nvk branch.
11:25 fdobridge: <c​langcat> I did get it working I think it was just lutris being bad
11:27 fdobridge: <p​avlo_kozlenko> you can use native opengl
11:27 fdobridge: <p​avlo_kozlenko> for warcraft
11:27 fdobridge: <p​avlo_kozlenko> parametr line: -opengl
11:57 fdobridge: <a​huillet> https://github.com/NVIDIA/open-gpu-doc/commit/46c065c40afce4f014f9725d1d6546b7d120be66 some new methods
12:07 fdobridge: <a​huillet> I believe Nouveau exposes neither functionality ATM (?)
12:09 fdobridge: <!​DodoNVK (she) 🇱🇹> I think multiview is already exposed
12:14 fdobridge: <c​langcat> So I found a strange bug. If I run warcraft 3 through gamescope on Nouveau gamescope crashes with
12:14 fdobridge: <c​langcat> > vulkan: vk.QueueSubmit( cmdBuffer->queue(), 1, &submitInfo, VK_NULL_HANDLE ) failed! (VkResult: -4(Device Lost))
12:14 fdobridge: <c​langcat>
12:14 fdobridge: <c​langcat> But if I run gamescope on the AMDGPU and the NVIDIA(NVK) GPU to render the game it works fine. But if I run both on the AMDGPU it works fine
12:14 fdobridge: <c​langcat> So I found a strange "bug". If I run warcraft 3 through gamescope on Nouveau gamescope crashes with
12:14 fdobridge: <c​langcat> > vulkan: vk.QueueSubmit( cmdBuffer->queue(), 1, &submitInfo, VK_NULL_HANDLE ) failed! (VkResult: -4(Device Lost))
12:14 fdobridge: <c​langcat>
12:14 fdobridge: <c​langcat> But if I run gamescope on the AMDGPU and the NVIDIA(NVK) GPU to render the game it works fine. But if I run both on the AMDGPU it works fine (edited)
12:14 fdobridge: <m​arysaka> It is but not using any methods for that
12:56 fdobridge: <t​riang3l> I wish I discovered one interesting workaround for a very inconvenient bug in `radeon`'s command parser a couple of months earlier to create a shipping userspace dependency on it :blobcatnotlikethis:
13:36 fdobridge: <g​fxstrand> Thanks! I'm not all that excited for multiview but variable-rate shading is really nice. 💜
13:56 fdobridge: <r​edsheep> Looks like that should help check off another 12_2 feature for vkd3d proton, very nice
13:57 fdobridge: <r​edsheep> Which reminds me, am I missing something obvious or is there no convenient way to see what is missing to have all of the features for a vulkan profile?
13:57 fdobridge: <r​edsheep> I see the newer vkd3d proton releases have changed the profile quite a bit but seeing exactly where we stand isn't as easy as I hoped
14:00 fdobridge: <r​edsheep> Right now the tracker is based on a profile that is a year old, and since then they've added things to distinguish between optimal support and minimum support
14:28 fdobridge: <a​huillet> what tracker?
14:28 fdobridge: <r​edsheep> https://gitlab.freedesktop.org/mesa/mesa/-/issues/9479
14:30 fdobridge: <r​edsheep> Based on what I've seen looking at the latest vkd3d-proton profile shooting for 12_0 optimal is probably a better next feature goal than basic 12_2
14:32 fdobridge: <r​edsheep> Just by going off of what appears in the gui version here: https://github.com/HansKristian-Work/vkd3d-proton/blob/master/profiles/PROFILES.md
15:45 fdobridge: <g​fxstrand> I'm honestly not sure. I think there are tools for diffing profiles but IDK if there's a tool for "what am I missing?". Maybe there is?
15:46 fdobridge: <g​fxstrand> We have a decent multiview emulation. You can do a bit better if you use the HW feature but, from chats I've had with the NVIDIA Vulkan folks, it's really hard to get actual benefit. Optimizing the geometry pipeline just isn't that important.
15:47 fdobridge: <g​fxstrand> And the kinds of shenanigans you have to do to optimize what's in the Vulkan spec to what HW can do are not trivial. It was hell to get that working on Intel and it's basically the same from a HW point of view. There's an NV extension which rewrites multiview to be more useful but IDK that anyone uses it.
16:01 fdobridge: <r​edsheep> I looked for it in the vulkan profile docs for like an hour, unless it's an undocumented feature of the profile merge tool I don't think I it exists
16:02 fdobridge: <r​edsheep> Probably just needs a simple python script but it's been over 10 years since I wrote one of those
16:05 fdobridge: <r​edsheep> @asdqueerfromeu Do you have a script written up already, or did you compare the profiles manually?
16:06 fdobridge: <!​DodoNVK (she) 🇱🇹> No script is used there
16:07 fdobridge: <r​edsheep> I'll see if I can write one, it would be nice to have an easier way
17:05 bencoh: 48
17:30 fdobridge: <g​fxstrand> https://www.collabora.com/news-and-blog/news-and-events/implementing-drm-format-modifiers-in-nvk.html