Mesa EGL X11/Wayland Code Receives Optimization For Multi-GPU/PRIME Systems

[phoronix]

Phoronix: Mesa EGL X11/Wayland Code Receives Optimization For Multi-GPU/PRIME Systems

Code merged last week to Mesa 23.1 by AMD ensures that a linear copy buffer is made on the display/scanout GPU when dealing with EGL contexts under Wayland or X11. This follows an optimization made last year to Mesa's GLX code within X.Org environments for enhancing the PRIME/multi-GPU support...

https://www.phoronix.com/news/Mesa-EGL-Wayland-Scanout-Linear

[pracedru]

Why didn't this happen before now?
Prime has been needing this for more than a decade.

[ResponseWriter]

Couldn't spot any performance numbers in those merge requests. Does anyone happen to have measured it? I have a laptop with Intel iGPU plus AMD dGPU (GCN 1.0 era, IIRC) but the latter never seemed to give any noticeably greater performance while using a ton more power, though hopefully puts it in better shape than when I last tried it.

[lumks]

Why didn't this happen before now?

Because Linux (ecosystem) is not about speed, but collaboration and doing things in a maintainable way.

[grigi]

I thought Wayland already did this, as to the reason that I get so much less stutter when rendering on dGPU and outputting to the dGPU-attached display under wayland as under X.org it was uselessly copying the render result to the compositor (that's on the iGPU) and then back to the dGPU framebuffer.

Looking forward to seeing if this makes a difference for my setup.

[MastaG]

Will this also work with Intel iGPU + nVidia dGPU?

[pracedru]

Because Linux (ecosystem) is not about speed, but collaboration and doing things in a maintainable way.

What a strange answer. So are you saying that it takes more than a decade to find the solution on collaborating with this in a maintainable way? My answer would be: obviously not.

[lumks]

What a strange answer. So are you saying that it takes more than a decade to find the solution on collaborating with this in a maintainable way? My answer would be: obviously not.

It depends on what you do. Some things can take longer. Not to find it, but to work it out (design/spec) and have someone to implement and someone who is willing to maintain it.

[pracedru]

It depends on what you do. Some things can take longer. Not to find it, but to work it out (design/spec) and have someone to implement and someone who is willing to maintain it.

I am pretty sure there is another reason.

[Ph42oN]

Can this further improve framebuffer copy performance on my laptop with ryzen 2500u+RX 560x?

Originally framebuffer copy was huge bottleneck on this laptop forcing me to use windows to play games, somewhere around kernel 5.15 it was improved making many games perform as they should, but in some cases there is still performance loss from framebuffer copy.