RFR: C2 reloads oop from memory in the middle of the LRB

[Roman Kennke]

I've run the patch against the failing test, and it's good now. I've
also run hotspot_gc_shenandoah and it's good. Go!

Thank you!!
Roman


> 
> http://cr.openjdk.java.net/~roland/shenandoah/nofwdptr-matcher-fix/webrev.00/
> 
> The matcher must find Nodes that must be emitted only once such as a
> field/array Load. It does that by finding shared nodes (nodes with
> multiple uses). With compressed oops, the actual pattern for an oop load
> is (DecodeN (LoadN )) so the Load is not shared but the DecodeN
> is. There is code to deal with that:
> 
>         if( mop == Op_AddP && m->in(AddPNode::Base)->is_DecodeNarrowPtr()) {
>           // Bases used in addresses must be shared but since
>           // they are shared through a DecodeN they may appear
>           // to have a single use so force sharing here.
>           set_shared(m->in(AddPNode::Base)->in(1));
>         }
> 
> But with the load of the mark work, offset is null so there's not AddP
> and that logic doesn't work.
> 
> Roland.
>