RFR: Simpler obj equals barrier for C2

[Roland Westrelin]

> Would it also require the memphi if we pulled up the
> resolve_for_obj_equals() to right after we got the args? E.g. something
> like:
>
>   args[0] = argument(0);
>   args[1] = argument(1);
>   access_resolve_for_obj_equals(args[0], args[1]);

I took a closer look and I was wrong. I don't see a need for the mem
phi. There would be a need for it if the barriers were right before the
CmpP and that's what the mem phi code that you removed seems to
assume. So your webrev is good.

Roland.