You are not logged in.

#1 2011-09-14 03:56:18

Gullible Jones
Member
Registered: 2004-12-29
Posts: 4,863

Benchmarking GTK

Because of my experience with GTK on the Mac G4, I did a very crude benchmark of the Murrine vs. default GTK engines. The computer used is an Asus Eee 1005HAB running Debian Squeeze (with Xfce). I tested all actions, with 500 iterations per action. Here are the results. Note the bold lines...

default engine wrote:

GtkEntry - time:  0.70
GtkComboBox - time: 11.05
GtkComboBoxEntry - time:  7.66
GtkSpinButton - time:  1.10
GtkProgressBar - time:  0.62
GtkToggleButton - time:  1.48
GtkCheckButton - time:  1.15
GtkRadioButton - time:  2.03
GtkTextView - Add text - time: 52.35
GtkTextView - Scroll - time:  4.35
GtkDrawingArea - Lines - time: 15.73
GtkDrawingArea - Circles - time: 20.59
GtkDrawingArea - Text - time:  8.94
GtkDrawingArea - Pixbufs - time:  1.07
---
Total time: 128.83

Murrine engine wrote:

GtkEntry - time:  0.70
GtkComboBox - time: 15.95
GtkComboBoxEntry - time:  8.82
GtkSpinButton - time:  2.04
GtkProgressBar - time:  1.13
GtkToggleButton - time:  2.89
GtkCheckButton - time:  2.15
GtkRadioButton - time:  3.76
GtkTextView - Add text - time: 52.41
GtkTextView - Scroll - time:  4.66
GtkDrawingArea - Lines - time: 15.67
GtkDrawingArea - Circles - time: 20.52
GtkDrawingArea - Text - time:  8.87
GtkDrawingArea - Pixbufs - time:  1.06
---
Total time: 140.65

So I learned that there is no functional difference between engine speeds on a fast machine like this EeePC. Big surprise. OTOH, I also learned that GTK takes longer to add text than it takes to do anything else - which probably means that it's slow to render fonts. Perhaps due to Pango? At any rate I wonder if there's a way to optimize the font rendering - because it's clearly a problem, and quite independent of the engine used.

Offline

#2 2011-09-14 06:15:57

moetunes
Member
From: A comfortable couch
Registered: 2010-10-09
Posts: 1,033

Re: Benchmarking GTK

I wonder if it will make a difference choosing a different font. which one was used for the tests?


You're just jealous because the voices only talk to me.

Offline

#3 2011-09-14 06:39:14

kokoko3k
Member
Registered: 2008-11-14
Posts: 2,393

Re: Benchmarking GTK

Gullible Jones wrote:

So I learned that there is no functional difference between engine speeds on a fast machine like this EeePC.

Nope, you learned that murrine engine is almost fast as the default one.
Try the same with some engine which does gradients wink


Help me to improve ssh-rdp !
Retroarch User? Try my koko-aio shader !

Offline

#4 2011-09-14 12:24:46

Gullible Jones
Member
Registered: 2004-12-29
Posts: 4,863

Re: Benchmarking GTK

Moetunes: Hmm good point. I was using Dejavu Sans 8.

kokoko3k: Judging from the appearance of Albatross, the Murrine engine can do gradients.

Offline

#5 2011-09-14 15:45:23

Gullible Jones
Member
Registered: 2004-12-29
Posts: 4,863

Re: Benchmarking GTK

Wow. This is not a GTK problem.

GtkEntry - time:  0.66
GtkComboBox - time:  9.83
GtkComboBoxEntry - time:  6.03
GtkSpinButton - time:  1.09
GtkProgressBar - time:  0.58
GtkToggleButton - time:  1.44
GtkCheckButton - time:  1.14
GtkRadioButton - time:  2.03
GtkTextView - Add text - time: 17.71
GtkTextView - Scroll - time:  4.07
GtkDrawingArea - Lines - time: 15.75
GtkDrawingArea - Circles - time: 20.45
GtkDrawingArea - Text - time:  8.83
GtkDrawingArea - Pixbufs - time:  1.08
---
Total time: 90.72

That's a more than 100% difference in font rendering speed... Just by switching from Dejavu Sans to Liberation Sans. Dejavu Sans is apparently *slow*.

(And it's also the default font for almost every distro. Go figure.)

Last edited by Gullible Jones (2011-09-14 15:45:38)

Offline

#6 2011-09-14 17:21:58

kokoko3k
Member
Registered: 2008-11-14
Posts: 2,393

Re: Benchmarking GTK

Gullible Jones wrote:

Moetunes: Hmm good point. I was using Dejavu Sans 8.

kokoko3k: Judging from the appearance of Albatross, the Murrine engine can do gradients.

Murrine can draw gradients, but depending on the style used they may be heavy or not.
I benchmarked qtcurve for qt widgets, and results varies very much when changing settings; same is for gtk.


Help me to improve ssh-rdp !
Retroarch User? Try my koko-aio shader !

Offline

#7 2011-09-15 00:42:30

moetunes
Member
From: A comfortable couch
Registered: 2010-10-09
Posts: 1,033

Re: Benchmarking GTK

Gullible Jones wrote:

Wow. This is not a GTK problem.

That's a more than 100% difference in font rendering speed... Just by switching from Dejavu Sans to Liberation Sans. Dejavu Sans is apparently *slow*.

(And it's also the default font for almost every distro. Go figure.)

It'd be nice to see how different fonts fair in this test. Maybe a half dozen or so of the most popular ones...
Cheers


You're just jealous because the voices only talk to me.

Offline

Board footer

Powered by FluxBB