AVL vs Red-Black tree

[Paolo Carlini]

Benjamin Kosnik wrote:

> I'd be more interested in looking at this if you could provide 
> performance measurements based on libstdc++'s own performance 
> testsuite. Ie, "make check-performance."

Good point.

> If you feel that coverage in that testsuite is lacking, please feel 
> free to suggest improvements...

In other terms, as soon as the legal issues are cleared, it would be 
*very* nice if Daniel could start with contributing performance 
testcases clearly showing the differences between an implementation 
based on Red-Black trees and one based on AVL trees. I think something 
is mostly ready, in the tarball on sourceforge.

Paolo.