> Here, I don't really ask for tweaking something but more to remove an assert
> which do something which is unrelated to the current algorithm.
> In my opinion, it's a debug assert used during the development
> that slip into the production code. The fact that the range [-128, 127] should be
> covered by the cache is mandated by the JLS, but if you really want an assert
> you should move it when high and low are initialized (they are final)
> and not where they are used.
> I also agree that the inlining heuristic should be changed and
> I'm sure that everybody that have looked to the inlining heuristic code of Hotspot will agree with
> but this is somehow unrelated to the problem.
> This is how I've found this assert,
> it doesn't change the fact that this assert should not be there.
OT: It would be an interesting discussion:
"What is a debug assert, and what not".