[llvm] r176408 - recommit r172363 & r171325 (reverted in r172756)

[Shuxin Yang]

>
> Shuxin: I'm not breaking any assumption here.
If you don't think you break the assumption,  tell us your immediate 
reaction to
the class name "ObjectSizeOffsetVisitor".

It seem so odd to me that  xxx_ObjectWhatever() return a meaningful 
value however we have no idea which
object it is talking about.

 >If you notice, we already do a similar analysis for select instructions.
Show me the code.
On the other hand,  there are some code showing similar usage
dose not necessarily mean that piece of code is correct, or it is good 
practice to write the code this way.

This is contrived usage of these functions:
   p1 = &object1.field1
   p2 = &object2.field2
   p3 = phi(p1, p2);

   get-object-offset() would return offset(object.filed1).  the 
optimizer randomly take object1 as
the "underlying object", and further analysis reveals that the location 
of &object1.field1 is either used as
char or int32, a contrived type-based alias alias would mistakenly make 
a conclusion that *p3
will not alias a load of floating point type, which could be wrong 
because object2.field2 could be of fp type.

   In my option, it better to hack for same specific cases than to 
convert a common function into a
dangerous, error-prone one.

> Phi nodes were not processed because nobody (I mean, me) implemented 
> the code.
> I agree BasicAA could use more comments. I'll add a comment there.
>
> Nuno