[clang] [CIR] Upstream ArraySubscriptExpr for fixed size array (PR #134536)

[Andy Kaylor via cfe-commits]

================
@@ -232,6 +233,161 @@ LValue CIRGenFunction::emitUnaryOpLValue(const UnaryOperator *e) {
   llvm_unreachable("Unknown unary operator kind!");
 }
 
+/// If the specified expr is a simple decay from an array to pointer,
+/// return the array subexpression.
+/// FIXME: this could be abstracted into a common AST helper.
+static const Expr *isSimpleArrayDecayOperand(const Expr *e) {
+  // If this isn't just an array->pointer decay, bail out.
+  const auto *castExpr = dyn_cast<CastExpr>(e);
+  if (!castExpr || castExpr->getCastKind() != CK_ArrayToPointerDecay)
+    return nullptr;
+
+  // If this is a decay from variable width array, bail out.
+  const Expr *subExpr = castExpr->getSubExpr();
+  if (subExpr->getType()->isVariableArrayType())
+    return nullptr;
+
+  return subExpr;
+}
+
+static mlir::IntegerAttr getConstantIndexOrNull(mlir::Value idx) {
+  // TODO(cir): should we consider using MLIRs IndexType instead of IntegerAttr?
+  if (auto constantOp = dyn_cast<cir::ConstantOp>(idx.getDefiningOp()))
+    return mlir::dyn_cast<mlir::IntegerAttr>(constantOp.getValue());
+  return {};
+}
+
+static CharUnits getArrayElementAlign(CharUnits arrayAlign, mlir::Value idx,
+                                      CharUnits eltSize) {
+  // If we have a constant index, we can use the exact offset of the
+  // element we're accessing.
+  const mlir::IntegerAttr constantIdx = getConstantIndexOrNull(idx);
+  if (constantIdx) {
+    const CharUnits offset = constantIdx.getValue().getZExtValue() * eltSize;
+    return arrayAlign.alignmentAtOffset(offset);
----------------
andykaylor wrote:

In order to find a test case for this, I modified the equivalent classic clang codegen implementation so that it would always go to the else case here. That caused 15 tests to fail, but we don't have enough support implemented to run any of the failing tests with CIR. There were 12 OpenMP failures, two PowerPC intrinsics, and one HLSL.

https://github.com/llvm/llvm-project/pull/134536