Around 19 o'clock on Jun 24, Billy Biggs wrote:

> I thought the interest was for better looking text and graphics.  I think
> it's a reasonable goal to move intelligence out of the server: let clients
> do high-quality compositing with better precision if desired, and only do
> in the server what is accellerated in the hardware.

Certainly the overall goal is to have better looking text and graphics.
I think we can pretty easily agree on what that means in this particular 
context, the discussion here centers on where that work should be done.

At the lowest levels, the Render extension is partially about providing
access to hardware acceleration, but it is also about providing reasonable
performance for most applications over a network.

In the context of gamma correction, I believe the correct question is to 
discuss how the compositing operator should be extended so that clients 
can do the right thing when Render is combined with suitable client-side 
code.  Essentially, I'd like to ensure that applications need never do
GetImage/<render-render-render>/PutImage for basic operations.  It's fine 
by me if they need to do extra work during PutImage to prepare the images 
as needed.

> I guess the alternative is to let Xft do its own compositing in
> certain cases?

I'd like to avoid that in this case; Xft provides a relatively low level 
of functionality which should be supported entire in the server.  That is 
so text applications will continue to operate efficiently over a network.

Text, in many ways, is an important special case for Render -- the balance 
between server/client should be shifted towards the server in this case, 
even though many of the operations may not commonly be accelerated in 
hardware.

Keith Packard        XFree86 Core Team        HP Cambridge Research Lab


_______________________________________________
Render mailing list
[EMAIL PROTECTED]
http://XFree86.Org/mailman/listinfo/render

Reply via email to