John Martin wrote:
> Bart Smaalders wrote:
>>
>> The X community could really use a better way of evaluating delivered
>> performance than XMarks - but someone needs to step up to the plate
>> and examine the actual protocol traffic used by our desktops today,
>> and develop a benchmark that will act as a better predictor of actual
>> delivered desktop graphics performance...
>>
> The protocol from clients is probably becoming noise.  What happens
> to a simple XDrawLine request after it gets to the server depends greatly
> on the desktop being used.  The conventional model was the request
> was sent directly to the graphics board and the rendering happened to
> the onscreen memory.  We are on the cusp of switching to compositing
> desktops, so the rendering may now happen to offscreen graphics memory
> or even system memory.  I believe you will soon find matching hardware
> to expected desktop graphics performance depends more heavily on the 
> desktop
> being used than the client rendering commands.
> 

<I'm using compiz myself, when I don't need to use bugster (grr).>

If what you mean is that I need to select first if I want to use
a compositing window manager+xserver, that makes perfect sense.

Don't I still care about delivered client performance, though?
Or does it all go so fast that it doesn't matter?

- Bart




-- 
Bart Smaalders                  Solaris Kernel Performance
barts at cyber.eng.sun.com              http://blogs.sun.com/barts

Reply via email to