NVIDIA Working On An EGLStreams Back-End For KDE On Wayland

[Britoid]

Good thing KDE stood its ground!
If NV decides to not adhere to standards, it's their burden to support their isolated solution.

There was no "standard", which is why there's now a mess. Each EGL implementation made their own, Mesa created GBM and nVidia EGLStreams.

For x reasons nVidia apparently can't use GBM, which is why they suggested making a new universal method that supports all use cases.

[tildearrow]

Since it's EGL based does it mean the compositor will only be able to use OpenGL and never Vulkan?

I was about to ask this... But here I have another question.

Would EGLStreams mean we are forced to use OpenGL for a Vulkan compositor, and as such we wouldn't ever have a pure Vulkan compositor on NVIDIA?

[LinAGKar]

I am talking about the traditional X model here, including the GLX protocol, and NOT DRI/DRM crap that infested things later. The original X model of having the X server solely in charge of handling the device drivers is the right way. The wrong way is putting window managers in charge of hardware and even more insanely putting every single application in charge of the hardware. With Wayland you have the window manager, and all applications directly vying for control over the same GPU hardware, this is a recipe for disaster.

Some have said that the X11 model anyway had applications talking to video drivers directly. This is incorrect. Under the traditional X client/server model, applications would not be anywhere near driver code, they sent all of their graphics commands over X protocol, including GLX, to the X server. The X server has the video drivers and can use its own accelerated drivers to send OpenGL commands onto the GPU. This is called AIGLX, Accelerated Indirect GLX and it is clearly by far the best model. ONLY in the case where DRI/DRM is used would an X application have video card drivers, but AIGLX can be used instead of DRI/DRM.

I fail to see the benefit of high latency, severe CPU overhead, and poor OpenGL/Vulkan support.

[Britoid]

I was about to ask this... But here I have another question.

Would EGLStreams mean we are forced to use OpenGL for a Vulkan compositor, and as such we wouldn't ever have a pure Vulkan compositor on NVIDIA?

Vulkan has VK_KHR_display which afaik does what GBM/EGLStreams does in a more limited fashion.

[shmerl]

I was about to ask this... But here I have another question.

Would EGLStreams mean we are forced to use OpenGL for a Vulkan compositor, and as such we wouldn't ever have a pure Vulkan compositor on NVIDIA?

Vulkan compositor won't need EGLstreams or GBM. But someone needs to implement it first: https://cgit.kde.org/kwin.git/log/?h=fredrik/vulkan

[lunarcloud]

How old is that laptop? Because anything remotely modern has the Intel iGPU connected to all display outputs, and the NVIDIA (or AMD) dedicated GPU run headless as a 3D accelerator. Intel's iGPU does support second monitors fine.

https://www.amazon.com/Acer-Predator-i7-7700HQ-1060-6GB-G3-571-77QK/dp/B06Y4GZS9C

As far as I can tell, I haven't been able to use the HDMI port when the nvidia drivers aren't working and exclusively using the intel drivers.

I am under the impression that Nouveau [...] delivers a low-performing but workable solution at this point of time.

That's fine, I guess, but I'd rather use the nvidia drivers than under-use the hardware I have.
It's not worth using Wayland if my 3D tools or games work worse because the driver I'd have to use.

I'd obviously prefer Nvidia actually implement the real Wayland API stuff, but if they're gonna maintain their own code paths, neat.

[totoz]

How old is that laptop? Because anything remotely modern has the Intel iGPU connected to all display outputs, and the NVIDIA (or AMD) dedicated GPU run headless as a 3D accelerator. Intel's iGPU does support second monitors fine.

Some laptops still connect the display panel to the dGPU in order to get G-SYNC working (for NVIDIA laptops featuring G-SYNC).

[Azrael5]

One of the best news ever.

[bearoso]

EGLStreams doesn’t maintain buffer contents after composition, which is kind of weird, and I highly suspect it’s a workaround for a serious deficiency with nvidia hardware. Because of this, I don’t think that nvidia can implement GBM or any proposed universal buffer format without it taking a more constricted approach than GBM.

[Charlie68]

It 's incredible to read the nostalgic post again on Xorg, resign yourself, certainly we are at the mercy of people who do not understand anything, which makes it clear that there is only one person or two who understand everyone else are idiots, which qualifies them .
However good news for KDE's Nvidia users.