Xrender cursor performance / double buffering?
wcohen at redhat.com
Tue Apr 17 13:14:12 EDT 2007
Jim Gettys wrote:
> Heh. More and more cursors are images these days, rather than two
> plane. The question is if we hit the hardware alpha blending unit for
> cursors or not. I'd suspect not, but profiling would provide the answer
> if you don't know instantly...
> - Jim
What would be the experiment to test cursor overhead? Start up an application ,
wait for the application to warm up, start up profiling, wiggle the cursor
around a lot, stop profiling, and then do the analysis?
Is there someway to drive the cursor around without having someone physically
using the touchpad? In the past I asked about dogtail running on the olpc.
However, it appears that doesn't run because the GTK accessibility code isn't
available on olpc machine. Are there other testing frameworks that work with
> On Tue, 2007-04-17 at 08:47 -0600, Jordan Crouse wrote:
>> On 17/04/07 16:25 +0200, Bert Freudenberg wrote:
>>> XO is not using the HW cursor anyway because, as you point out, that
>>> only supports 2 bit cursors (1 bit black/white, 1 bit alpha).
>> So I'm confused. You know that the hardware can only do two bit cursors.
>> Yet you are ignoring that, and purposely taking a performance hit so
>> you can have a more colorful cursor. Why?
More information about the Devel