View Single Post
Old 04-04-06, 11:16 PM   #68
d13f00l
Registered User
 
Join Date: Apr 2005
Posts: 197
Default Re: Slow AA text rendering in KDE

Quote:
Originally Posted by Linuxhippy
1.) If KDE application 1 performs bad with one NVidia card and very good with another one its in my eyes not KDEs nore the applications fault. Its simply a speed regression either in the chip or in the driver!

2.) Why X? I would recommend to get some ground knowledge before generating tons of traffic blaming other guys which are not involved into this problem at all. It has never been and will never be X's responsibility to accalerate anything - thats why drivers exist.

lg Clemens
In regards to font acceleration, I was referring to the nvidia X driver. Right now, the driver does non grayscale font aa in software. Cool link. I can't wait till Mustang is stable, it's supposed to have a lot of work done in the GTK L&F for Swing as well, yum. Dolphin should be pretty interesting too.

1) Well, it's really not that case here, on my PC. What a strange issue.
My GF4MX440 works just as bad with the latest driver as my 6200 with font AA. And, like someone else said, KWrite seems to render each letter indivudually, while kedit buffers each line. I joined #kde-devel and brought it up there as well, the developer who was looking into it had a dual Xeon system, and the problem didn't affect him as his system is fast enough to compensate. Others had the issue, and it happens with other drivers as well. I have not tried that QT application on my GF4MX, I'm just judging by KWrite performance between my two cards. Renderaccell is enabled with both.

2) Why do you think Microsoft is coming out with Aero, and why does XGl exist? It's a pain in the arse for driver developers to write specific code for each OS to accelerate various 2d video functions, and it brings around a lot of room for error when it's not done right. We're going to see a shift in responsibility, and it's for the better. Rather than Nvidia focusing on accelerating functions X provides, they can concentrate more on optimizing their OGL/DX API implementations, which in effect would speed up everything. Not to mention, a single codebase in X for accelerating various functions(font accell, transparent windows) would work and be accelerated on basically any card that supports OGL. It will save everyone time and the hassle once complete.

Last edited by d13f00l; 04-04-06 at 11:33 PM.
d13f00l is offline   Reply With Quote