1 |
On Friday 15 August 2008 15:41:04 Volker Armin Hemmann |
2 |
wrote: |
3 |
> On Freitag, 15. August 2008, Alan McKinnon wrote: |
4 |
> > Oddly enough, 3D is very very snappy. It's the 2D |
5 |
stuff that cripples the |
6 |
> > performance, and on a desktop, 2D is exactly what |
7 |
you want. |
8 |
> |
9 |
> AFAIK the problem is that the 8XXX series doesn't |
10 |
really have 2d hardware |
11 |
> anymore - everything is done in the 3d part - and the |
12 |
drivers to support |
13 |
> that are harder to write than anticipated. |
14 |
|
15 |
Sorry for the delayed reply. nvnews was unreachable for |
16 |
me all yesterday afternoon. This post is straight from |
17 |
the horses mouth, from Aaron Platter. He's the nvidia |
18 |
driver author: |
19 |
|
20 |
http://www.nvnews.net/vbulletin/showpost.php?p=1704502&postcount=17 |
21 |
|
22 |
"Compiz window resizing is slow for a number of reasons |
23 |
mostly related to the inefficient way that the X server |
24 |
resizes redirected windows. Performance should be |
25 |
significantly better with InitialPixmapPlacement=2, |
26 |
except that most window decorators use currently- |
27 |
unaccelerated convolution filters for their shadows, |
28 |
bogging it down. Please try solid resizing with IPP=2 |
29 |
and the window decorator disabled, to see how much of an |
30 |
effect it has. Improving the performance of convolution |
31 |
filters and redirected window resizing in general is |
32 |
something we're working on for future driver releases. |
33 |
|
34 |
"Also, you should see a significant improvement in |
35 |
redirected window resizing performance with IPP=1 if you |
36 |
use one of the xserver 1.5 prereleases." |
37 |
|
38 |
-- |
39 |
alan dot mckinnon at gmail dot com |