imlib2/xrender benchmarks

[Adam Jackson]

On Friday 25 February 2005 18:13, Cameron wrote:
> Thank you for the info.
>
> Would it be helpful to anyone if I provided some benchmarks of render vs
> imlib2 without hardware render acceleration?

Nah, we pretty much already know that software Render is dog slow.  Profile 
data, however...

> I would also love to help fix the situation. I am solid with C, and I
> know a bit of x86 assembly. However, I'm lacking an understanding of low
> level graphics. How could I go learning about that topic?

If you want to see how the low-level Render acceleration hooks work in XAA, 
check out (and ideally help finish) some of the open RenderAccel bugs in 
bugzilla:

https://bugs.freedesktop.org/show_bug.cgi?id=1265
https://bugs.freedesktop.org/show_bug.cgi?id=1293

Those two are mostly complete and just have some correctness issues to be 
resolved before they can be applied to head.  RenderAccel in KAA is slightly 
more general than what's in XAA but the same basic ideas apply.  I can 
annotate the tdfx patch if it would help people understand.  The final stage 
of getting the hardware to accelerate the actual blend will look pretty 
similar to these no matter what AA is in use.

However a more useful thing to do would be to fix the algorithms used in the 
software blend case to not be quite so braindead.  For this, you'd be looking 
at programs/Xserver/render/ for the core, programs/Xserver/fb/fbpict.c for 
the framebuffer interface, and optionally 
programs/Xserver/hw/xfree86/xaa/xaaPict.c for the XAA-level hooks.

- ajax
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 189 bytes
Desc: not available
URL: <http://lists.x.org/archives/xorg/attachments/20050225/d0c8f4ae/attachment.pgp>