[aarch64-port-dev ] RFR (trivial) : fix aarch64-8u type profile bug

[Yangfei (Felix)]

> > This also reminds me of another two aarch64-specific profiling issues:
> > https://bugs.openjdk.java.net/browse/JDK-8188221
> > https://bugs.openjdk.java.net/browse/JDK-8189439
> >
> > I think they also should be incorporated in aarch64 8u.  What do you think?
> 
> I've always been reluctant to backport performance-only patches to 8u, but I
> admit that version will be around for a long time, so OK.
> 

Looks like the upstream patches can be simplified: 'mdp' is loaded by test_method_data_pointer which is called by profile_return_type & profile_parameters_type.  

Patch:
diff -r 03fce7b04b42 src/hotspot/cpu/aarch64/templateInterpreterGenerator_aarch64.cpp
--- a/src/hotspot/cpu/aarch64/templateInterpreterGenerator_aarch64.cpp  Tue Sep 24 03:28:42 2019 -0400
+++ b/src/hotspot/cpu/aarch64/templateInterpreterGenerator_aarch64.cpp  Wed Sep 25 16:15:57 2019 +0800
@@ -444,7 +444,6 @@
     Register obj = r0;
     Register mdp = r1;
     Register tmp = r2;
-    __ ldr(mdp, Address(rmethod, Method::method_data_offset()));
     __ profile_return_type(mdp, obj, tmp);
   }

@@ -1634,13 +1633,8 @@
   __ mov(rscratch2, true);
   __ strb(rscratch2, do_not_unlock_if_synchronized);

-  Label no_mdp;
   Register mdp = r3;
-  __ ldr(mdp, Address(rmethod, Method::method_data_offset()));
-  __ cbz(mdp, no_mdp);
-  __ add(mdp, mdp, in_bytes(MethodData::data_offset()));
   __ profile_parameters_type(mdp, r1, r2);
-  __ bind(no_mdp);

   // increment invocation count & check for overflow
   Label invocation_counter_overflow;