Jeroen Frijters writes: > Andrew Haley wrote: > > David Holmes writes: > > > That was my initial thought but the generalization is > > trivial and useful. > > > For example, given the XXX/VMxxx split having only > > > getCallerClass wouldn't allow XXX to defer to VMxxx and have > > > VMxxx find the real caller. > > > > Perhaps we don't need to defer to VMxxx if we have a portable way to > > do getCallerClass. > > I don't see how.
Well, if the only reason for the VMxxx method in some cases it to find the caller's class, and we have a well-defined getCallerClass(), then we don't in that particular case need a VMxxx method. > My desire to have a getCallerClass is because I want to be able to > support method inlining properly. If the JIT knows about > getCallerClass, it can do the right thing. The generalization is > not inline proof. I don't think the generalization can be used reliably in the presence of optimization. It'll only work with the simplest VM designs unless we keep a lot of extra information around. Andrew. _______________________________________________ Classpath mailing list [EMAIL PROTECTED] http://mail.gnu.org/mailman/listinfo/classpath
