[flang-commits] [flang] [flang] AliasAnalysis: distinguish addr of arg vs. addr in arg (PR #87723)

[Joel E. Denny via flang-commits]

================
@@ -399,19 +404,30 @@ AliasAnalysis::Source AliasAnalysis::getSource(mlir::Value v) {
   if (!defOp && type == SourceKind::Unknown)
     // Check if the memory source is coming through a dummy argument.
     if (isDummyArgument(v)) {
-      type = SourceKind::Argument;
       ty = v.getType();
       if (fir::valueHasFirAttribute(v, fir::getTargetAttrName()))
         attributes.set(Attribute::Target);
-
       if (Source::isPointerReference(ty))
         attributes.set(Attribute::Pointer);
+      if (followBoxAddr && fir::isa_ref_type(ty))
----------------
jdenny-ornl wrote:

I'm not sure I understand your meaning.  Given the following:

```
subroutine test(p)
  real, pointer, intent(in) :: p
end subroutine test
```

flang-new generates:

```
func.func @_QPtest(%arg0: !fir.ref<!fir.box<!fir.ptr<f32>>> {fir.bindc_name = "p"}) {
  %0:2 = hlfir.declare %arg0 {fortran_attrs = #fir.var_attrs<intent_in, pointer>, uniq_name = "_QFtestEp"} : (!fir.ref<!fir.box<!fir.ptr<f32>>>) -> (!fir.ref<!fir.box<!fir.ptr<f32>>>, !fir.ref<!fir.box<!fir.ptr<f32>>>)
  return
}
```

If I drop the `intent(in)`, the only difference is that the `intent_in` attribute is lost.

Would you please provide an example of the case you have in mind?  Thanks.

https://github.com/llvm/llvm-project/pull/87723