What about cursor support?

[Keith Whitwell]

Jon Smirl wrote:
> On 5/20/05, Keith Whitwell <keith at tungstengraphics.com> wrote:
> 
>>Isn't the cursor just the topmost composited element?
> 
> 
> Don't you want hardware cursor support so the mouse won't lag when the
> CPU gets loaded?
> 

Depends if you want all the GL goodies in your cursor, or if you're 
happy with whatever the hardware happens to implement...  Which might 
not be very nice...

Lots of cards don't support rgb or argb cursors so even under current 
X.org we don't have hardware cursor a lot of the time and nobody notices 
except when it gets screwed up by direct rendering.

If you think about what's involved, software cursor shouldn't be slow. 
A clever compositor would only have to redo the tiny square under the 
cursor, assisted by saving the old contents.  Hopefully hardware can 
manage that at a reasonable rate...

Cursor moves are triggered by the kernel already, so scheduling cursor 
updates under high CPU loads is something we're already relying on the 
kernel for help with.

Anyway, all the above aside, it should be pretty easy to come up with an 
extension that encapsulates the traditional hardware sprite cursor 
model.  I think an extension is the right approach as it doesn't tie the 
base proposal into one of the creakier legacy hardware concepts still 
out there.

Keith