Josh Saddler - Latest comments on Benchmarks: gtk+ engines

In response to: Benchmarks: gtk+ engines

Josh Saddler [Member] — Thu, 08 Jan 2009 21:25:51 +0000

@Alejandro:

See comment number 3.

In response to: Benchmarks: gtk+ engines

Alejandro M. [Visitor] — Thu, 08 Jan 2009 20:32:31 +0000

may you benchmark murrine too?

In response to: Benchmarks: gtk+ engines

665 [Visitor] — Fri, 02 Jan 2009 17:59:15 +0000

Sorry guys, i prefer pixmap themes, its the only engine where you can see real creativity & difference

other themes should be classified as *-LOOK.org-spam

In response to: Benchmarks: gtk+ engines

Donalds [Visitor] — Sun, 14 Dec 2008 23:35:00 +0000

I'm getting different results which probably means this test have little meaning, using "Simple" GTK theme I get average 4.25 (4.50 something with Mist) and using my favourite Aurora (with Aurora Elegant scheme) I get average of 6.34.

So if I get the extra eyecandy for 2 seconds extra I'm fairly OK with that.

Everything on my system is latest of whatever is in ~ on gentoo, nothing else, and my system is far from super (amd x2 3800+ on a geforce 7600gt).

Thoughts?

In response to: Benchmarks: gtk+ engines

Mart Raudsepp [Visitor] — Sun, 14 Dec 2008 17:42:54 +0000

FYI, the USE=glitz on your cairo has no relevance to the themes. Nothing in gtk+ creates cairo glitz surfaces, or at least any of the theme engines I've ever seen. They quite probably all create xlib surfaces and in-memory (not drawn to screen but as a cache thing) surfaces only, so that glitz doesn't do any good for you as far as theme performance goes.

Also, your runtimes appear to be very short. I'd highly suggest going with more runs, like gtkperf -a -c 500 or the like, to get at least 30 seconds - but more is always even better if you can wait out.

About gtkperf, I really should fix it to do something useful in the end for the GtkDrawingArea tests. They seem to currently all use gdk_draw_* functions, which nothing modern uses anymore. They should get a patch to use cairo like real apps do. But that's something for upstream then.

Glad you found gtkperf useful. I packaged it to have an additional measuring point for some xorg-server patches performance affect measurement -- the EXA glyph cache patches that 1.5.3 will or did get added on, make a huuuuge difference for many drawing op - text.

In response to: Benchmarks: gtk+ engines

F [Visitor] — Sun, 14 Dec 2008 14:07:33 +0000

I personally find Rezlooks dull and nauseating, but subjective impressions aside I did some testing on Intel P-M 2.26GHz/i915GM and to my surprise the differences were not very substantial: Aurora - 16.25 Candido - 16.07 Murrine - 15.29 Rezlooks - 13.07 Xfce - 13.05 I think the current state of intel graphics drivers is to blame, hopefully with GEM and UXA the situation will improve.

In response to: Benchmarks: gtk+ engines

Josh Saddler [Member] — Sun, 14 Dec 2008 10:24:14 +0000

@f:

I loathe Murrine with a fierce, burning passion. I hate it with every fiber of my being. It's just about the ugliest, most inconsistent engine ever to have been spawned from the deepest, darkest, slime-smeared pit.

Of course, I do use it on my Ubuntu Studio laptop.*

* Only because the Ubuntu folks came up with the only halfway decent, usable dark gtk+ theme for any engine. As a default theme, it's among the better ones out there. Every other Murrine theme I've seen is pure horsepiss, however.

@Milan:

Thanks for that; I forgot to stick my nvidia-drivers version in there. Fixed.

In response to: Benchmarks: gtk+ engines

Milan [Visitor] — Sun, 14 Dec 2008 10:21:29 +0000

I just tried gtkperf with nimbus and aurora, nimbus score is 4.50 while aurora is 7.15. My configuration is actually slower then yours, 4000+ with nvidia 6100 and 2gb ram. I guess you can blame nvidia drivers for that, on newer cards 2d is awfull.