EduardoS (no.delete@this.spam.com) on November 23, 2012 10:13 am wrote:
> Wilco (Wilco.Dijkstra.delete@this.ntlworld.com) on November 21, 2012 7:04 pm wrote:
> > If you have the ability to dynamically switch between logging and no logging at runtime then you always
> > have some overhead.
>
> Unless you can recompile the code*... Oh, you can!

Recompiling the code is overhead by definition, and quite large at that compared to the cost of a virtual call.

> I think you missed the point, "the VM can recompile if something change", the
> trigger may be in the cold path that causes the change, not in the hot path.

Sure, but you have to consider the general case. Assume we have several classes which each have several objects for different purposes (logging or whatever), and in each class you repeatedly switch between different subclasses for those objects (in a cold path, say only 10 times a second). Every switch causes recompilation of all functions in that class and all other functions which they were inlined into. Do you understand the problem with your "optimization" now?

Wilco