[flang-commits] [clang] [flang] [llvm] [mlir] [OpenMP][flang] Add initial support for by-ref reductions on the GPU (PR #165714)

[Kareem Ergawy via flang-commits]

================
@@ -1485,6 +1490,18 @@ class OpenMPIRBuilder {
     /// reduction. If null, the implementation will use the non-atomic version
     /// along with the appropriate synchronization mechanisms.
     ReductionGenAtomicCBTy AtomicReductionGen;
+
+    /// For by-ref reductions, we need to keep track of 2 extra types that are
+    /// potentially different:
+    /// * The allocated type is the type of the storage allocated by the
+    /// reduction op's `alloc` region. For example, for allocatables and arrays,
+    /// this type would be the descriptor/box struct.
+    Type *ByRefAllocatedType;
+    /// * The by-ref element type is the type of the actual storage needed for
----------------
ergawy wrote:

Done.

https://github.com/llvm/llvm-project/pull/165714