GNOME Shell + Mutter Had A Busy November With Some Big Performance Optimizations

[Niarbeht]

steam hardware survey isn't accurate imo. It just randomly picks people, asks if they want to opt in to the survey.

That's how surveys work.

[sarmad]

Gnome Shell was already smooth from day one on an 2005 LG LT20 convertible notebook running on 1.4Ghz Intel Pentium M and 2GB RAM (upgradable to 4GB RAM max). Same with 2007 Sony VIAO equipped with Intel Core Duo, Intel 855 graphic card and 4GB RAM.

You must've been launching no more than 3 windows then.

Jokes aside, if I remember correctly Gnome was relatively smooth at first, then it slowed down and kept slowing down with every new release until they realized it's no longer funny and started fixing the performance. Hopefully in a couple of releases it will be in great shape for all use cases (as opposed to currently being smooth in some cases but not others).

[amehaye]

Anyone knows what's up with Mr. Daniel Van Vugt?

He contributed a great many performance MRs to Mutter and GNOME Shell, but hasn't been active in the last month or so. I hope that everything is well with him.

[Michael]

Anyone knows what's up with Mr. Daniel Van Vugt?

He contributed a great many performance MRs to Mutter and GNOME Shell, but hasn't been active in the last month or so. I hope that everything is well with him.

HE was on vacation the past month

[pal666]

And he's right that it was still faster than gnome 3 is today.

it's easy to be faster when doing less. text console is even faster than gnome 2

[pal666]

In an ideal world, everyone writes bug free highly efficient code. We don't live in that world

people can write arbitrary things, maintainer's job is to say no to saboteurs disguised as volunteers. speed and bugginess can be measured objectively

[RahulSundaram]

people can write arbitrary things, maintainer's job is to say no to saboteurs disguised as volunteers. speed and bugginess can be measured objectively

That's taking it to extremes. Just because someone contributes code that is less than ideal doesn't make them saboteurs. Maintainers are not infallible either. Performance is measurable but is very much relative. Maintainers might very well choose more maintainable code over performance. Bugs on the other are much harder to measure objectively. Realistically one can only test for known issues and avoid regressions.