In the current DGA access implementation is not optimized and therefore
full performance enhancement with DGA access can not be be observed. The
following items should be considered:

1. Check for the DGA access capability of a screen and generate an indication
   flag in the screen structure to avoid querying DGA lock every time
   for remote X rendering.
2. Check for clipping changes so that lock can be faster, without having
   to query the clip every time.
3. Optimize the logic for cached info per drawable to get to the result
   quicker.
4. Check why the device is opened and closed with every getLock() call.
   See 4246160.