[flang-commits] [flang] [flang][OpenMP] Implement flexible OpenMP clause representation (PR #81621)

[Sergio Afonso via flang-commits]

================
@@ -142,6 +142,1122 @@ static void genNestedEvaluations(Fortran::lower::AbstractConverter &converter,
     converter.genEval(e);
 }
 
+//===----------------------------------------------------------------------===//
+// Clauses
+//===----------------------------------------------------------------------===//
+
+namespace detail {
+template <typename C> //
+llvm::omp::Clause getClauseIdForClass(C &&) {
+  using namespace Fortran;
+  using A = llvm::remove_cvref_t<C>; // A is referenced in OMP.inc
+  // The code included below contains a sequence of checks like the following
+  // for each OpenMP clause
+  //   if constexpr (std::is_same_v<A, parser::OmpClause::AcqRel>)
+  //     return llvm::omp::Clause::OMPC_acq_rel;
+  //   [...]
+#define GEN_FLANG_CLAUSE_PARSER_KIND_MAP
+#include "llvm/Frontend/OpenMP/OMP.inc"
+}
+} // namespace detail
+
+static llvm::omp::Clause getClauseId(const Fortran::parser::OmpClause &clause) {
+  return std::visit([](auto &&s) { return detail::getClauseIdForClass(s); },
+                    clause.u);
+}
+
+namespace omp {
+using namespace Fortran;
+using SomeType = evaluate::SomeType;
+using SomeExpr = semantics::SomeExpr;
+using MaybeExpr = semantics::MaybeExpr;
+
+template <typename T> //
+using List = std::vector<T>;
----------------
skatrak wrote:

Ah yes, I'm aware of this limitation. Do you think it would be possible to store const references to those larger objects rather than full copies? I'm guessing these might be Symbol or Expression-related types, which I've seen trigger this sort of thing before. If we can't get around copying these big structures, then either keeping `std::vector` or `llvm::SmallVector<T, 0>` would be necessary.

https://github.com/llvm/llvm-project/pull/81621