Acceleration of 2D operations for kdrive

[Keith Packard]

On Fri, 2007-02-02 at 07:23 -0600, David Hagood wrote:

> Each trace on such a 
> display consists of 500 to 1000 line segments, and there can be multiple 
> traces per object - so we are running about 150 THOUSAND line segments a 
> second.

For short lines, software is almost always faster; the cost of setting
up the line and synchronizing with the hardware swamps the benefit of
using it. For long lines, you fill the screen fast enough that using the
hardware doesn't make a huge difference. That's why I didn't bother
adding hardware line hooks.

Of course, changing kdrive to add line acceleration should be easy; as
you can see, the structure within the driver is not complicated. Patches
are always welcome. It's not like kdrive is fixed in any way; there is
no ABI or API guarantee necessary as the drivers are statically linked
to the X server.

> You can bet whatever body part you wish WE NEED ACCELERATED LINE DRAWS.

I would recommend measuring the difference between hardware and software
before you place your wager; most 2D line engines are not all that
impressive. Your's may be different, but until you measure it, you won't
know.

-- 
keith.packard at intel.com
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 189 bytes
Desc: This is a digitally signed message part
URL: <http://lists.x.org/archives/xorg/attachments/20070202/c3a36355/attachment.pgp>