[clang] [Clang] Add `__builtin_[ordered_]reduce_fadd` for ordered/unordered fp reductions (PR #176160)

[Benjamin Maxwell via cfe-commits]

================
@@ -3628,6 +3630,44 @@ Sema::CheckBuiltinFunctionCall(FunctionDecl *FDecl, unsigned BuiltinID,
     break;
   }
 
+  case Builtin::BI__builtin_reduce_addf:
+  case Builtin::BI__builtin_ordered_reduce_addf: {
+    if (checkArgCountRange(TheCall, 1, 2))
+      return ExprError();
+
+    ExprResult Vec = UsualUnaryConversions(TheCall->getArg(0));
+    if (Vec.isInvalid())
+      return ExprError();
+
+    TheCall->setArg(0, Vec.get());
+
+    QualType ElTy = getVectorElementType(Context, Vec.get()->getType());
+    if (ElTy.isNull() || !ElTy->isRealFloatingType()) {
+      Diag(Vec.get()->getBeginLoc(), diag::err_builtin_invalid_arg_type)
+          << 1 << /* vector of */ 4 << /* no int */ 0 << /* fp */ 1
+          << Vec.get()->getType();
+      return ExprError();
+    }
+
+    if (TheCall->getNumArgs() == 2) {
+      ExprResult StartValue = UsualUnaryConversions(TheCall->getArg(1));
+      if (StartValue.isInvalid())
+        return ExprError();
+
+      if (!StartValue.get()->getType()->isRealFloatingType()) {
----------------
MacDue wrote:

I defined the innit value as being cast to the element type to allow things to work with Clang's floating-point promotions, which means for example, promotes Float16 to float32 if it's not a natively supported scalar. But maybe, I can avoid emitting the promotions in the first place. 

https://github.com/llvm/llvm-project/pull/176160