Comments on: Improving input latency

By: vignatti

vignatti — Fri, 13 Feb 2009 04:02:37 +0000

a multithreaded X server won’t solve all issues that we have today. See my old posts.

By: Mark

Mark — Sat, 17 Jan 2009 11:51:33 +0000

Low latency mouse would be nice. But it should be doable in user-space by making X multithreaded and having proper controls on how long can one thread/process hog the rendering pipeline. This is all needed anyway for console switches.

Don’t forget about multiple screens, etc…

By: vignatti

vignatti — Fri, 16 Jan 2009 19:32:52 +0000

@ LGB:

I’ll not enter in monolithic VS microkernel discussion and they merits ;)

By: LGB

LGB — Fri, 16 Jan 2009 15:17:21 +0000

Okey, but pushing things into kernel done “shurtcuts” there to have better performance etc would have no end then: we could bring firefox into kernel too, and it would be faster because of the the lack of context switches between user/kernel space, and so on :) I think everything should be moved OUT from kernel (and this is done by kernel developers too) instead what is possible, but push in things which works better there (eg: kernel based mode switching ….).

By: Cursor’s update inside kernel only « Tiago Vignatti

Cursor’s update inside kernel only « Tiago Vignatti — Wed, 14 Jan 2009 22:48:49 +0000

[...] DRM based cursors obtained by short-circuit the kernel input layer and DRM module. This would solve all cursor’s latency issues that we have in our current [...]

By: keep it going… « Tiago Vignatti

keep it going… « Tiago Vignatti — Thu, 07 Aug 2008 23:22:33 +0000

[...] it going… Given that GSoC ‘08 is getting close to the end, strategy number 2 showed more feasible to proceed my work. Strategy #3 would be a lot of fun but would imply a hell [...]

By: vignatti

vignatti — Wed, 30 Jul 2008 15:31:01 +0000

to Robert:

“Number three would make anyone wanting to play with new input methods play kernel programmer for the day. Which brings its own set of restrictions and pain.”, oh god. For sure this is not a good argument of design. And thanks for remember me that X is different from linux…

By: vignatti

vignatti — Wed, 30 Jul 2008 15:27:05 +0000

to Rudd-O:

Xinput extension would exists in strategy #3 at all inside the X server. You also mentioned that “cursor latency is really not a problem in modern desktops with enough RAM”. And if I want to run Xorg in my mobile?

By: Robert

Robert — Wed, 30 Jul 2008 13:43:13 +0000

Number three would make anyone wanting to play with new input methods play kernel programmer for the day.

Which brings its own set of restrictions and pain.

And remember X != Linux.

By: Rudd-O

Rudd-O — Wed, 30 Jul 2008 09:37:05 +0000

Strategy 3 is wrong. You’d lose all the new XInput goodness for multitouch. And cursor latency is really not a problem in modern desktops with enough RAM, but multitouch IS a concern. Wanna avoid cursor jumpiness? Then work harder on making shared libs consume less RAM.

Strategies 1+2 seem much more reasonable.