[clang] [clang] Factor out OpenACC part of `Sema` (PR #84184)

[Erich Keane via cfe-commits]

================
@@ -0,0 +1,67 @@
+//===----- SemaOpenACC.h - Semantic Analysis for OpenACC constructs -------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+/// \file
+/// This file declares semantic analysis for OpenACC constructs and
+/// clauses.
+///
+//===----------------------------------------------------------------------===//
+
+#ifndef LLVM_CLANG_SEMA_SEMAOPENACC_H
+#define LLVM_CLANG_SEMA_SEMAOPENACC_H
+
+#include "clang/AST/DeclGroup.h"
+#include "clang/Basic/OpenACCKinds.h"
+#include "clang/Basic/SourceLocation.h"
+#include "clang/Sema/Ownership.h"
+
+namespace clang {
+
+class Sema;
----------------
erichkeane wrote:

I think it is less of an OpenACC discussion and more of a general use case discussion FWIW.  BUT I think we DO expect a lot of the 'child' `Sema`s to call into Sema quite a lot.

Very much of the OpenMP implementation (and soon to be OpenACC implementation) calls into Constant evaluation functions, etc.  BUT I think the hope here is that the calls will be '1 way', in that Sema won't call into the OpenACC implementation ever, but the OpenACC will call into Sema reasonably often (as it does for OMP).

https://github.com/llvm/llvm-project/pull/84184