|
|
#1 | |
|
Registered User
Join Date: May 2006
Posts: 477
|
The previous reported X 1.9 performance regressions were with KDE users, but Gnome is also affected.
Compiz is enabled on all tests below gtkperf - 1000 runs - gtk-murrine-engine 0.90.3 (Dust Theme) - Fedora 13 x86_64 (X.org 1.8) Code:
GtkPerf 0.40 - Starting testing: Mon Oct 18 22:42:49 2010 GtkEntry - time: 0.16 GtkComboBox - time: 5.17 GtkComboBoxEntry - time: 3.32 GtkSpinButton - time: 0.57 GtkProgressBar - time: 0.98 GtkToggleButton - time: 1.10 GtkCheckButton - time: 0.53 GtkRadioButton - time: 0.90 GtkTextView - Add text - time: 15.45 GtkTextView - Scroll - time: 2.18 GtkDrawingArea - Lines - time: 4.90 GtkDrawingArea - Circles - time: 7.20 GtkDrawingArea - Text - time: 2.06 GtkDrawingArea - Pixbufs - time: 0.46 --- Total time: 44.99 Code:
GtkPerf 0.40 - Starting testing: Sun Nov 7 16:50:32 2010 GtkEntry - time: 0.18 GtkComboBox - time: 42.00 GtkComboBoxEntry - time: 21.41 GtkSpinButton - time: 15.53 GtkProgressBar - time: 28.80 GtkToggleButton - time: 20.91 GtkCheckButton - time: 1.67 GtkRadioButton - time: 1.98 GtkTextView - Add text - time: 14.65 GtkTextView - Scroll - time: 6.47 GtkDrawingArea - Lines - time: 8.25 GtkDrawingArea - Circles - time: 11.37 GtkDrawingArea - Text - time: 2.74 GtkDrawingArea - Pixbufs - time: 0.61 --- Total time: 176.57 Code:
GtkPerf 0.40 - Starting testing: Sun Nov 7 16:40:49 2010 GtkEntry - time: 0.17 GtkComboBox - time: 16.45 GtkComboBoxEntry - time: 15.67 GtkSpinButton - time: 14.75 GtkProgressBar - time: 1.59 GtkToggleButton - time: 1.44 GtkCheckButton - time: 0.60 GtkRadioButton - time: 1.00 GtkTextView - Add text - time: 14.59 GtkTextView - Scroll - time: 2.64 GtkDrawingArea - Lines - time: 6.27 GtkDrawingArea - Circles - time: 8.52 GtkDrawingArea - Text - time: 2.71 GtkDrawingArea - Pixbufs - time: 0.47 --- Total time: 86.88 Last edited by mooninite; 11-10-10 at 05:44 PM. |
|
|
|
|
|
|
#2 | |
|
Registered User
Join Date: Jul 2005
Posts: 61
|
Wouldn't it be more useful to compare the same versions of gtk-engines?
The regression could be from gtk-murrine-engine 0.90.3 to gtk-murrine-engine 0.98.1.1. |
|
|
|
|
|
|
#3 | |
|
Registered User
Join Date: May 2006
Posts: 477
|
Quote:
I provided the generic gtk-engine result as well, to discount such an idea. GTK performance still suffers from X.org 1.9. |
|
|
|
|
|
|
#4 | |
|
Registered User
Join Date: Feb 2006
Posts: 9
|
I am experiencing the same issue. What I did to alleviate the issue somewhat was to go from a high glossy gradient look to plain (clearlooks -> clearlooks classic.) Now its still half the speed of what it usually is but better than the 4 times slower that was with the heavy glossy look.
Hopefully this will give some ideas where things got out of hand in the code. |
|
|
|
|
|
|
#5 |
|
Registered User
Join Date: May 2006
Posts: 477
|
After being pointed to this thread by a friend, I thought I'd downgrade cairo as well just to try it out.
After rebooting - 2D is back to normal. gtkperf - 1000 runs - gtk-murrine-engine 0.98.1.1 (Dust Theme) - Fedora 14 x86_64 (X.org 1.9) - cairo 1.8 Code:
GtkPerf 0.40 - Starting testing: Wed Nov 10 17:40:56 2010 GtkEntry - time: 0.19 GtkComboBox - time: 6.23 GtkComboBoxEntry - time: 3.89 GtkSpinButton - time: 1.19 GtkProgressBar - time: 0.96 GtkToggleButton - time: 1.06 GtkCheckButton - time: 0.72 GtkRadioButton - time: 1.49 GtkTextView - Add text - time: 15.52 GtkTextView - Scroll - time: 2.84 GtkDrawingArea - Lines - time: 8.93 GtkDrawingArea - Circles - time: 11.62 GtkDrawingArea - Text - time: 4.07 GtkDrawingArea - Pixbufs - time: 0.64 --- Total time: 59.34 |
|
|
|
|
|
#6 | |
|
Registered User
Join Date: Jul 2007
Posts: 38
|
Fedora's cairo doesn't include cairo-1.10.0-buggy_gradients.patch, so file a bug with Fedora's bugzilla.
|
|
|
|
|
|
|
#7 |
|
Registered User
Join Date: May 2006
Posts: 477
|
One had been created. I'll go ahead and link it here.
https://bugzilla.redhat.com/show_bug.cgi?id=652124 |
|
|
|
|
|
#8 |
|
Registered User
Join Date: Jun 2006
Posts: 680
|
cairo-1.10.0-3.fc14.src.rpm with cairo-1.10.0-buggy_gradients.patch for Fedora 14.
Download and rebuild using Code:
rpmbuild --rebuild cairo-1.10.0-3.fc14.src.rpm Run `rpm -Uvh packagename1 packagename2 packagename3` to update packages. This patch indeed fixes the problems with gradients. |
|
|
|
|
|
#9 |
|
NVIDIA Corporation
Join Date: Sep 2007
Posts: 254
|
Please note that this should be fixed with 270.41.03. I suggest un-applying the buggy-gradients patch from cairo when using this driver as server-side gradients can be very beneficial to performance now that this bug is out of the way.
|
|
|
|
|
|
#10 | |
|
Registered User
Join Date: May 2006
Posts: 477
|
Quote:
I can only see about a 5% performance difference though (server-side 5% faster) using gtkperf (1000 runs). |
|
|
|
|
|
|
#11 |
|
NVIDIA Corporation
Join Date: Sep 2007
Posts: 254
|
Yeah, for general UI rendering the trapezoid strips/gradients cover a pretty small area, so the benefit of hardware acceleration isn't dramatic. I suggest doing the same experiment with cairo-perf-trace or any benchmark that uses Cairo to render vector graphics on large areas.
|
|
|
|
|
|
#12 | |
|
Registered User
Join Date: Jan 2007
Posts: 119
|
Anybody noticed a regression about this?
here is what i emailed to nvidia support: Quote:
|
|
|
|
|
![]() |
| Thread Tools | |
|
|