Complexity of 'SI + interfaces' vs. 'MI'

How can a compiler take advance of the fact that an OO language has
only single inheritance + interfaces instead of multiple inheritance
(a-la Java)? Surely it is freed from the "diamond problem": it will
always know from which class a method is inherited.


But can the compiler exploit this situation to produce better
(simpler, smaller and faster) code? Can the adjustment-table in vtable
be eliminated? Can the thunk functions be simplified?


What optimizations (I'm mostly interested in runtime ones) can be
implemented once some assumption are relaxed or strengthened?


--
Gioele
[Seems to me it's mostly leaving out code for complications that
can't happen. The adjustment table certainly goes away. -John]

Return to the comp.compilers page.
Search the comp.compilers archives again.