GNOME 40 Mutter Moves Input Work To A Separate Thread

[phoronix]

Phoronix: GNOME 40 Mutter Moves Input Work To A Separate Thread

An exciting addition for GNOME 40 is that the Mutter compositor will be punting the input work off to a separate CPU thread...

http://www.phoronix.com/scan.php?pag...e-Input-Thread

[tildearrow]

Finally. How many years have we been waiting for this?

[cl333r]

Finally. How many years have we been waiting for this?

I don't know but boy does time fly - I just checked Linus Torvalds is already 50 and soon to be 51.

[Mthw]

Just wondering, do Plasma and Plasma Wayland have this feature already? It often felt like, at least on Wayland, it is not there.

[fenixex]

No, plasma's cursors starts lagging hard when CPU usage is high.

[Mthw]

No, plasma's cursors starts lagging hard when CPU usage is high.

Additionally I have an issue where the framerate drops randomly (and stays there) to ~25 fps even though CPU and GPU usage is low/none. It only seems to affect my laptop with AMD CPU and GPU, which is weird. Has anyone else seen something similar? It has been reported but almost no one seems to have the same issue: https://bugs.kde.org/show_bug.cgi?id=424398

[144Hz]

The review process and the CI is really good now. Adahl is super picky and thorough. CI happens on commit level, GnomeOS level and distributor level.

Just like upstreams are meant to be.

[Britoid]

The review process and the CI is really good now. Adahl is super picky and thorough. CI happens on commit level, GnomeOS level and distributor level.

Just like upstreams are meant to be.

You make me embarrassed to use GNOME.

[Mancooo]

Article is not correct, the merge-request literally says:

"To set the expectations right, I should probably also mention what this branch does not do:

• It does not perform KMS updates in the input thread. All KMS interaction still does happen in the main thread, so the pointer can visually stall, even though the input thread isn't missing events. Ideally, KMS management itself should be eventually taken out of the main thread. This branch sets the seed so that the MetaKmsCursorRenderer can possibly be managed there, so that KMS and the input threads could talk between themselves without interaction from the main thread"

[Mancooo]

Can't edit my post, sorry for the double post.

The merge-request that also needs to be merged (for the input problem to be solved) is this one: https://gitlab.gnome.org/GNOME/mutte..._requests/1241