KDE 4.11 Haswell Desktop Effects Performance

[phoronix]

Phoronix: KDE 4.11 Haswell Desktop Effects Performance

For those using Intel graphics does toggling the "suspend desktop effects for fullscreen windows" KDE Kwin window manager option lead to better OpenGL Linux gaming performance? Here's some new benchmarks to answer that question for Intel Core i7 hardware...

http://www.phoronix.com/vr.php?view=MTQ5MjE

[mgraesslin]

what a wonderful benchmark you did there, Michael.

void Options::setUnredirectFullscreen(bool unredirectFullscreen)
{
if (GLPlatform::instance()->driver() == Driver_Intel)
unredirectFullscreen = false; // bug #252817

whoops, the option is ignored for all Intel hardware.

[darkbasic]

This problem is from 2011, why didn't Intel still fix it?

[mdias]

This problem is from 2011, why didn't Intel still fix it?

Probably because the bug hasn't been reported to the mesa bugtracker (I can't find it).

on topic: I see a noticeable increase in performance when I disable desktop effects using alt+shift+F12 when gaming in a window, so I would think fullscreen windows would also get a noticeable increase if compositing is disabled.

[josk]

I'm not sure how good that suspend compositing in fullscreen option really is? Maybe someone of you have benchmarked that option against the blocked compositing. Please share the results if you have benchmarked that?

So I have always just totally blocked the compositing when I run softwares like Steam and mplayer (that fixed all 1080p tearing problems). And I use binary nvidia drivers.

So if you don't know how to totally block the compositing...

Shortcut key combination to toggle compositing on/off

Shift+alt+f12

Or automatically for certain programs:

System settings > Window Behavior > Window rules > New

First tab:

- Description = "Name the setting"

- Window class (application) = "Binary name of the running program ex. steam"

Last tab:

- Check the BLOCK COMPOSITING, select FORCE and YES options.

[smitty3268]

This problem is from 2011, why didn't Intel still fix it?

It's not clear that it hasn't been fixed.

The first bug from a couple years ago looks like it was. There may be another problem present only in the i915g driver, which Intel doesn't support (they have a classic driver instead) but it's not clear whether that is still currently present or not.

[Michael]

what a wonderful benchmark you did there, Michael.

void Options::setUnredirectFullscreen(bool unredirectFullscreen)
{
if (GLPlatform::instance()->driver() == Driver_Intel)
unredirectFullscreen = false; // bug #252817

whoops, the option is ignored for all Intel hardware.

Shouldn't the user be informed when setting an option that is broken for their hardware rather than silently ignoring it?

[Bathroom Humor]

Shouldn't the user be informed when setting an option that is broken for their hardware rather than silently ignoring it?

Yes, probably.
But at least you did say it could just be an issue for Intel drivers. I guess next time you can test it on Nvidia cards, more testing must be done.
It is pretty odd that such a noticable performance bug affects such a large percentage of the users. :U

[Apopas]

...

lol that nick rocks!

[GreatEmerald]

on topic: I see a noticeable increase in performance when I disable desktop effects using alt+shift+F12 when gaming in a window, so I would think fullscreen windows would also get a noticeable increase if compositing is disabled.

Indeed. I'm on Ivy Bridge and I can see quite a difference. Not as huge of a difference as on Nouveau, though, but still pretty notable.