15:59fdobridge: <![NVK Whacker] Echo (she) 🇱🇹> `fixes\player\drive_c\Program Files (x86)\Microsoft\EdgeWebView\Application\109.0.1518.115\msedgewebview2.exe: ../mesa/src/nouveau/vulkan/nvk_cmd_draw.c:603: nvk_CmdBeginRendering: Assertion 'level->tiling.is_tiled' failed.` 🌐
16:05fdobridge: <rhed0x> is it trying to render to a linear image?
16:13TimurTabi: gitlab.freedesktop.org seems to be down
16:54fdobridge: <pixelcluster> gitlab upgrade is currently underway
18:39fdobridge: <gfxstrand> Yeah... Render to linear isn't a thing on NVIDIA. I mean it is but the restrictions are insane. We need to implement it for modifier support but the implementation is going to involve blits back and forth to a tilted shadow copy. 🙃
18:39fdobridge: <![NVK Whacker] Echo (she) 🇱🇹> I wonder if this is triggered because of my PRIME setup
18:40fdobridge: <gfxstrand> Shouldn't. We handle that properly in the WSI layers.
18:41fdobridge: <gfxstrand> It's possible, though, that there's some weird interaction between prime and whatever wine is doing. 🤷🏻♀️
18:41fdobridge: <![NVK Whacker] Echo (she) 🇱🇹> The assert also only gets triggered on winewayland for some reason (winex11 is fine)
18:41fdobridge: <gfxstrand> Like, I wouldn't be at all surprised if Edge+DXVK is hitting fun corners.
18:42fdobridge: <gfxstrand> Oh...
18:42fdobridge: <gfxstrand> Yeah, it's prime + Wayland. I thought I'd fixed that. I guess I should probably go fix it harder. 🙄
18:43fdobridge: <![NVK Whacker] Echo (she) 🇱🇹> I use XWayland for winex11 though (not native X11)
18:44fdobridge: <rhed0x> dxvk shouldn't use linear images
18:44fdobridge: <rhed0x> should be a bug if it does
18:45fdobridge: <gfxstrand> It's not DXVK. It's busted Wayland PRIME code
18:50fdobridge: <gfxstrand> I had a merge request for this. I don't remember what happened to it. And gitlab is down this week-end so I can't look it up.
18:51fdobridge: <gfxstrand> Also, the WSI code has way too many little flags and half of them aren't properly respected by all the back-ends. It's such a mess...
18:51fdobridge: <gfxstrand> I hate WSI
19:01fdobridge: <![NVK Whacker] Echo (she) 🇱🇹> Is it worse than Direct3D?
19:03fdobridge: <gfxstrand> Window systems are just horrible. The problems they have to solve are also horrible.
19:04fdobridge: <rhed0x> WSI in d3d is actually very easy
19:04fdobridge: <rhed0x> the Vulkan WSI stuff is A LOT more complicated
19:06fdobridge: <![NVK Whacker] Echo (she) 🇱🇹> I mean I remember the "I hate this API" messages when you were talking about Direct3D 9 and 12 at least
19:06fdobridge: <rhed0x> yeah but not WSI
19:07fdobridge: <rhed0x> although d3d9 WSI has its own oddities
19:07fdobridge: <rhed0x> it's everything else in d3d9 that sucks 🙃
19:13fdobridge: <karolherbst🐧🦀> though granted, MS can just dictate how their windowing system works, it's not like they have to deal with 5 different ones
19:14fdobridge: <rhed0x> yup
19:14fdobridge: <karolherbst🐧🦀> not sure we have 5, might be 6 or 7 or even just 4
19:14fdobridge: <rhed0x> x11, Wayland, surfaceflinger, Windows, what else?
19:14fdobridge: <rhed0x> embedded stuff I guess
19:15fdobridge: <karolherbst🐧🦀> drm, and x11 has kinda two
19:15fdobridge: <rhed0x> ah right, raw DRM too
19:15fdobridge: <karolherbst🐧🦀> not sure if the differences between dri2 and dri3 are super relevant
19:15fdobridge: <karolherbst🐧🦀> but...
19:15fdobridge: <karolherbst🐧🦀> anyway, we have a bunch
19:17fdobridge: <karolherbst🐧🦀> is anybody doing vulkan on raw drm these days?
19:17fdobridge: <karolherbst🐧🦀> wondering how easy that is with SDL2..
19:18fdobridge: <airlied> pretty sure that's how VR works
19:18fdobridge: <karolherbst🐧🦀> right, makes sense
19:18fdobridge: <rhed0x> maybe gamescope?
19:25fdobridge: <gfxstrand> Yeah, and Microsoft has made some non-trivial compromises to make it that easy.
19:27fdobridge: <gfxstrand> Which isn't to say that Linux window systems couldn't or shouldn't make or have made similar compromises. Just saying that DXGI swapchains aren't an existence proof that it's not as hard as people say. They're just proof that it's possible to layer enough hacks and compromises to make things pretty easy for full-screen games.
19:29fdobridge: <gfxstrand> Which isn't to say that Linux window systems couldn't or shouldn't make or have made similar compromises. Just saying that DXGI swapchains aren't an existence proof that WSI is or can be easy. They're just proof that it's possible to layer enough hacks and compromises to make things pretty easy for full-screen games. (edited)
19:33fdobridge: <airlied> I feel wsi was one area vulkan could have gone a bit more implicit 😛
19:48fdobridge: <rhed0x> fair opinion considering everyone hates it :)
20:51fdobridge: <gfxstrand> It's easy to say that. Much harder to figure out how to thread that particular needle.
20:51fdobridge: <rhed0x> definitely
21:36fdobridge: <airlied> make it like SDL 😛
23:52fdobridge: <prop_energy_ball> Yes, Gamescope
23:53fdobridge: <prop_energy_ball> SteamVR just uses KHR_display
23:55fdobridge: <prop_energy_ball> I don't think any of those compromises are exclusive to full-screen games or games at all.
23:56fdobridge: <prop_energy_ball> Also while I am here, allocating my own images for swapchain when?
23:56fdobridge: <prop_energy_ball> 🐸
23:56fdobridge: <prop_energy_ball> I've been half tempted to write an ext for that recently...
23:57fdobridge: <prop_energy_ball> I'd like to remove the swapchain copies (virtual swapchain thingy) in DXVK and vkd3d-proton at some point
23:58fdobridge: <prop_energy_ball> It'd be possible if we had some additional guarantees from the swapchain and that I think. Both of those guarantees can be sated by Gamescope at least and probably other Wayland compositors.