That's interesting. Do I understand correctly though that the performance of CTL is mostly on a par with that of the STL?

Yes, I'd say at worst its 20% slower. The implementation is naive and readable (and therefor hackable). I did not have the time to implement all the heuristics that make STL as fast as it. The goal is hackability and ease of use (think text book function examples), while giving embedded engineers (or even desktop engineers with an aversion to C++) a chance to focus on getting their work done.

That, and lighting fast compile times, where programming feels like (oh, the heresy) a _drastically reduced C++_.