EXA performance and bugs

[Pierre Ossman]

Eric Anholt wrote:
> On Wed, 2005-09-28 at 17:51 +0200, Pierre Ossman wrote:
> 
>> * Default mouse pointer is a colored, semi-transparent square.
> 
> Please submit this as a bugzilla entry with what card you have, exact
> date of CVS, and the xorg.conf file attached.
> 

Done:
https://bugs.freedesktop.org/show_bug.cgi?id=4633

> 
>> * Small rendering artifacts. Previous thread said these were caused by
>>cache effects.
> 
> Are you running current CVS?  I haven't had any reports of rendering
> artifacts with current CVS.
> 

Yes. Entered a bug report with attached screen shot at:
https://bugs.freedesktop.org/show_bug.cgi?id=4634

> 
>> * Glyph rendering fails when radeon module isn't loaded. Red Hat's rhgb
>>starts the X server in some funky way without loading the radeon module.
>>I guess the achieve this by not enabling dri (haven't look to closely at
>>it).
> 
> 
> Fixed this one (I'm pretty sure -- you didn't describe it much) a week
> ago.
> 

Quite right. The CVS from 25th failed, but from the 28th works.

> 
>>I also have varying performance. I don't know if this is expected or not
>>but what I'm seeing is that rendering slows to a near halt if I drag a
>>window over another window and the desktop. Dragging over just one other
>>window gives full performance. Dragging over several other windows seems
>>to degrade performance gradually for each window added to the mix.
> 
> 
> Things get slow when memory gets tight.  We need more work on this.
> 

Is this when card memory or system memory gets tight? Both?

> 
>>I also see the problem with slow rendering in gnome-terminal. It is
>>enough to have EXA turned on. Composite makes no difference.
> 
> 
> If you're using subpixel antialiasing, yes, that's slow.
> 

I do. But there were no such problems with XAA. Are some operations
unaccelerated with EXA that were accelerated with XAA?

Rgds
Pierre