A Look At The Many Features To X.Org Server 1.20

[Happy Heyoka]

The only insane ones are the fanboys/girls who act like GBM is perfect for all users/uses. [...snip...]Unfortunately, things aren't that simple.

Care to elaborate?

[DanL]

Care to elaborate?

Start here: https://www.phoronix.com/scan.php?pa...ice-Memory-API
You can also search Phoronix for several more articles about the proposed "memory device allocator".

[Happy Heyoka]

Start here: https://www.phoronix.com/scan.php?pa...ice-Memory-API
You can also search Phoronix for several more articles about the proposed "memory device allocator".

Yes, familiar with all of that - was already of the understanding that GBM came first, that Nvidia was aware and that then the presented their buffer scheme as is and expected everyone else to implement it; and by golly, a lot of people who have far fewer resources than Nvidia decided not to (that's about as succinctly as I can put it, and it's fairly simple and without accusing Nvidia of anything other than being uninterested in having to work for Wayland support like everyone else).

The arguments from various developers (desktop, wm etc) about not wanting multiple parallel code paths to work on with different buffer infrastructure are totally with merit from my point of view.

I could easily accuse Nvidia of "embrace and extend" which is a tactic a few of the older readers may remember (which is along the lines of what they did, imho with Cuda vs OpenCL)

[DanL]

The arguments from various developers (desktop, wm etc) about not wanting multiple parallel code paths to work on with different buffer infrastructure are totally with merit from my point of view.

It's not ideal, but that's the reality at the moment. The EGLstream patch is a band-aid/stopgap until a better solution comes along.

[Ansla]

You think so? Linux desktop users are minority. Nvidia earns money on professional users. They don't need Wayland or GBM - desktop users need it. So I don't think Nvidia would listen desktop users.

Are you assuming that all professionals use the cards just for number crunching on headless servers? I think a significant portion of them do rendering and would benefit from having the result of that rendering be presented using Wayland rather than X.

To be fair - GBM is Mesa solution and Nvidia doesn't use Mesa. It's free software, so You can't force everybody to use Mesa. EGL in Xwayland isn't for Nvidia. It's for users. Thanks to this option more users can use Xwayland. Is this bad?

While GBM was designed with mesa in mind it has nothing no dependency on it. AMDGPU-PRO uses it just fine and it only uses mesa for video acceleration, not OpenGL or Vulkan. What GBM needs is a kernel driver that integrates with KMS. And that is what nVidia is unwilling to provide. It would require them to integrate much deeper with the kernel, much closer to actually having a native Linux driver, as opposed to the current thin wrapper around their Windows driver.