> OK. I guess the problem is that the call to the stub doesn't have an oop > map.
Assuming you can get the oop map in stub, deoptimization could still be delayed until the stub is exited. Is that considered a problem? Actually I'm not even sure the loop synthesized from IR nodes is feasible: what would the JVM state at the safepoint be given that there's no actual java code for it? What would you deoptimize to if you were to deoptimize? Roland.
