[flang-commits] [flang] [flang] Allow lowering of sub-expressions to be overridden (PR #69944)

[via flang-commits]

================
@@ -269,11 +220,37 @@ static inline void genOmpAccAtomicUpdateStatement(
   //    }
   //    return
   //  }
-  mlir::Value updateVar = converter.getSymbolAddress(*name->symbol);
-  if (auto decl = updateVar.getDefiningOp<hlfir::DeclareOp>())
-    updateVar = decl.getBase();
 
-  firOpBuilder.setInsertionPointAfter(tempOp);
+  Fortran::lower::ExprToValueMap exprValueOverrides;
+  // Lower any non atomic sub-expression before the atomic operation, and
+  // map its lowered value to the semantic representation.
+  const Fortran::lower::SomeExpr *nonAtomicSubExpr{nullptr};
+  std::visit(
+      [&](const auto &op) -> void {
+        using T = std::decay_t<decltype(op)>;
+        if constexpr (std::is_base_of<Fortran::parser::Expr::IntrinsicBinary,
+                                      T>::value) {
+          const auto &exprLeft{std::get<0>(op.t)};
+          const auto &exprRight{std::get<1>(op.t)};
+          if (exprLeft.value().source == assignmentStmtVariable.GetSource())
+            nonAtomicSubExpr = Fortran::semantics::GetExpr(exprRight);
+          else
+            nonAtomicSubExpr = Fortran::semantics::GetExpr(exprLeft);
+        }
----------------
jeanPerier wrote:

I suppose you can indeed go through the intrinsic function arguments here too yes, @NimishMishra, I do not plan to work on this, so I am happy if you do in a later patch!

https://github.com/llvm/llvm-project/pull/69944