Currently the only performance data we get for the DisplayList mechanism is:

• If we should do something so time consuming that it would show up as increased BUILD times for framework benchmarks
• The DL complexity benchmarks which are focused more on rendering than the cost of construction or dispatch
• The display_list_rendertests unit test which is focused on measuring output accuracy and which uses the software renderer so it really doesn't reflect the cost of any operations performed by the DisplayList code itself

We should have more direct benchmarks that look at:

• The speed of constructing a DisplayList
• The speed of getting the bounds of a DisplayList (which is currently a lazy second pass over the DL)
• The speed of getting an RTree from a DisplayList (again - lazily computed as another pass over the DL)
• The speed of dispatching a DisplayList (to a null dispatcher)
• The speed of comparing 2 DisplayList objects