[flang-commits] [flang] [flang] Fix for atand(Y, X), and implment atan2d(Y, X) (PR #79002)

[Yi Wu via flang-commits]

================
@@ -333,11 +333,15 @@ static const IntrinsicInterface genericIntrinsicFunction[]{
                 common::Intent::In, {ArgFlag::canBeNull}}},
         DefaultLogical, Rank::elemental, IntrinsicClass::inquiryFunction},
     {"atan", {{"x", SameFloating}}, SameFloating},
-    {"atand", {{"x", SameFloating}}, SameFloating},
     {"atan", {{"y", OperandReal}, {"x", OperandReal}}, OperandReal},
-    {"atand", {{"y", OperandReal}, {"x", OperandReal}}, OperandReal},
+    {"atand",
+        {{"y", SameReal, Rank::scalar, Optionality::required,
+             common::Intent::In},
+            {"x", SameReal, Rank::scalar, Optionality::optional,
+                common::Intent::In}},
+        SameReal, Rank::scalar, IntrinsicClass::inquiryFunction},
----------------
yi-wu-arm wrote:

It is possible to have intrinsics with the same name but different para list here, like how it was done before. Having an `atand(X)`, and `atand(Y,X)`. But in the `intrinsicCall.cpp`, they seem only different from each other by `args.size()`. I do have another implementation by have both `atand(X)` and `atand(Y,X)`, and treat them both in `genAtand` by using `args.size()` for each case. I'm not sure if using `args.size()` is the safe/correct way to do this. I can use `atand([Y],X)` if `args.size()` could not be used.

The reason for me to use `atand(Y,[X])` here is that if X is not present, it would be treated as `atand(X)`. This definitely does not make sense in terms of naming, because I am actually using Y as X for the one input version `atand(X)`.

I will change it back from `SameReal` to `OperandReal` regard to the relaxation of same kind rule, and also update all other `atan` with the zero check.


https://github.com/llvm/llvm-project/pull/79002