[llvm-branch-commits] [flang] [Flang][OpenMP] Derived type explicit allocatable member mapping (PR #96266)

[via llvm-branch-commits]

================
@@ -141,6 +143,110 @@ createMapInfoOp(fir::FirOpBuilder &builder, mlir::Location loc,
   return op;
 }
 
+omp::ObjectList gatherObjects(omp::Object obj,
+                              semantics::SemanticsContext &semaCtx) {
+  omp::ObjectList objList;
+  std::optional<omp::Object> baseObj = getBaseObject(obj, semaCtx);
+  while (baseObj.has_value()) {
+    objList.push_back(baseObj.value());
+    baseObj = getBaseObject(baseObj.value(), semaCtx);
+  }
+  return omp::ObjectList{llvm::reverse(objList)};
+}
+
+bool duplicateMemberMapInfo(OmpMapMemberIndicesData &parentMembers,
+                            llvm::SmallVectorImpl<int> &memberIndices) {
+  // A variation of std:equal that supports non-equal length index lists for our
+  // specific use-case, if one is larger than the other, we use -1, the default
+  // filler element in place of the smaller vector, this prevents UB from over
+  // indexing and removes the need for us to do any filling of intermediate
+  // index lists we'll discard.
+  auto isEqual = [](auto first1, auto last1, auto first2, auto last2) {
+    int v1, v2;
+    for (; first1 != last1; ++first1, ++first2) {
+      v1 = (first1 == last1) ? -1 : *first1;
+      v2 = (first2 == last2) ? -1 : *first2;
+
+      if (!(v1 == v2))
+        return false;
+    }
+    return true;
+  };
+
+  for (auto memberData : parentMembers.memberPlacementIndices)
+    if (isEqual(memberData.begin(), memberData.end(), memberIndices.begin(),
+                memberIndices.end()))
+      return true;
+  return false;
+}
----------------
agozillon wrote:

Happy to change the name, but unfortunately std::equal will not work and will result in the previously stated UB, I had this previously :-) The implementation is effectively the exact same (provided I am looking at the correct overload at least) as the above implementation, minus line 168 which supplements -1's when we over index, without this in the original implementation of std::equal we over index the 2nd range in certain cases and end up accessing random values which yields incorrect results. So I'll try to do what I stated in the previous comment and see how that goes!

https://github.com/llvm/llvm-project/pull/96266