// Create return op for device computation region in the paralle_execute op
  Operation* after_op_r = builder.create<mlir::tf_device::ReturnOp>(
      new_device_cluster.getLoc(), device_results);

  builder.setInsertionPointToEnd(&parallel_execute_device_block);

  // Create return op for the new device cluster op
  builder.create<mlir::tf_device::ReturnOp>(device_cluster.getLoc(),

    return {0, 2, 3, 1};
  } else {
    return {};
  }
}

// Shuffle elements in the `attr` according to the permutation. Optional
// `inner_size` allows to shuffle array attributes created from rank 2 tensors
// on outer dimension only.
ArrayAttr ShuffleArrayAttr(ArrayAttr attr, ArrayRef<int64_t> permutation,
                           int inner_size) {
  if (attr.empty()) return attr;

    // If the op is rewritten, FusedLoc can be created.
    for (Location child_loc : mlir::cast<FusedLoc>(loc).getLocations()) {
      std::optional<QuantizationUnitLoc::QuantizationUnit> found_unit =
          FindQuantizationUnitFromLoc(child_loc);
      if (found_unit.has_value()) return found_unit;
    }
  } else if (isa<CallSiteLoc>(loc)) {
    // If the graph is inlined, CallSiteLoc can be created.
    return FindQuantizationUnitFromLoc(

  for (Operation* subgraph_op : llvm::reverse(reverse_subgraph)) {
    // Create a deep copy of the subgraph ops' operands to the func op.
    stablehlo_func_op.getBody().begin()->push_back(subgraph_op->clone(mapper));
  }

  SmallVector<Value> result_values;
  for (const Value original_output_value : outputs) {
    // Use the mapped values in the newly created function that correspond to
    // outputs in the original function.

    "Resource ops decomposition did not converge";

// Decomposing resource ops should not take more than a few iterations (2-3) to
// converge as only a few patterns create new resource ops that can be further
// decomposed. The rest of the iterations are enough to clean up any dead ops
// created by decomposition.
constexpr int kMaxIterations = 20;

// Populates `reachable_functions` with all functions that can be reached from
// device cluster ops.

  // Guarantee all functions have one use, which enables more exact shape
  // inference.
  pm.addPass(mlir::TF::CreateGuaranteeAllFuncsOneUsePass());
  // Run shape inference so that tf_executor/tf_device ops created later will
  // likely to inherit more concrete types.
  pm.addPass(mlir::TF::CreateTFShapeInferencePass());
  pm.addNestedPass<FuncOp>(mlir::TFTPU::CreateTPUPartitionedOpConversionPass());
  pm.addNestedPass<FuncOp>(

      // Create new ConstantOp-ConvertOp-Operation sequences. At this moment,
      // old ConstantOp is guaranteed to have one F32->F16 convert op regardless
      // of its number of users.
      rewriter.setInsertionPointAfter(op);
      // create new F16 constant op in that location
      ConstantOp new_const = rewriter.create<ConstantOp>(

constexpr char kXlaOutsideCompilationAttr[] = "_xla_outside_compilation";

bool HasOutsideCompilationAttribute(Operation* op) {
  return op->getAttrOfType<StringAttr>(kXlaOutsideCompilationAttr) != nullptr;
}

// Finds op that created a given value. If the value is a BlockArgument, this
// returns the owner of the Block.
Operation* GetOpOfValue(Value value) {
  if (auto block_arg = mlir::dyn_cast<BlockArgument>(value))

  auto builder = OpBuilder::atBlockTerminator(&main_graph_op.GetBody());
  // Create an IslandOp that will wrap the IdentityOp. Add a control dependency
  // for the newly copied save function.
  auto wrapper_island_op = builder.create<IslandOp>(
      name_loc, TypeRange{main_file_prefix_arg.getType()},
      tf_executor::ControlType::get(&ctx), ValueRange(control_inputs));

  const std::string dump_file_path =
      tsl::io::JoinPath(*dump_dir, dump_file_name);
  TF_ASSIGN_OR_RETURN(std::unique_ptr<llvm::raw_ostream> file,
                      WritableFileWrapper::Create(dump_file_path));

  LOG(INFO) << "IR dump file created: " << dump_file_path;
  return file;
}

class PrinterConfig : public mlir::PassManager::IRPrinterConfig {
 public:
  explicit PrinterConfig(